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 *));
ip6_rcv_finish_core(struct net * net,struct sock * sk,struct sk_buff * skb)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
ip6_rcv_finish(struct net * net,struct sock * sk,struct sk_buff * skb)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
ip6_sublist_rcv_finish(struct list_head * head)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
ip6_list_rcv_finish(struct net * net,struct sock * sk,struct list_head * head)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
ip6_rcv_core(struct sk_buff * skb,struct net_device * dev,struct net * net)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 /* While RFC4291 is not explicit about v4mapped addresses
227 * in IPv6 headers, it seems clear linux dual-stack
228 * model can not deal properly with these.
229 * Security models could be fooled by ::ffff:127.0.0.1 for example.
230 *
231 * https://tools.ietf.org/html/draft-itojun-v6ops-v4mapped-harmful-02
232 */
233 if (ipv6_addr_v4mapped(&hdr->saddr))
234 goto err;
235
236 skb->transport_header = skb->network_header + sizeof(*hdr);
237 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
238
239 pkt_len = ntohs(hdr->payload_len);
240
241 /* pkt_len may be zero if Jumbo payload option is present */
242 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
243 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
244 __IP6_INC_STATS(net,
245 idev, IPSTATS_MIB_INTRUNCATEDPKTS);
246 goto drop;
247 }
248 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
249 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
250 goto drop;
251 }
252 hdr = ipv6_hdr(skb);
253 }
254
255 if (hdr->nexthdr == NEXTHDR_HOP) {
256 if (ipv6_parse_hopopts(skb) < 0) {
257 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
258 rcu_read_unlock();
259 return NULL;
260 }
261 }
262
263 rcu_read_unlock();
264
265 /* Must drop socket now because of tproxy. */
266 skb_orphan(skb);
267
268 return skb;
269 err:
270 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
271 drop:
272 rcu_read_unlock();
273 kfree_skb(skb);
274 return NULL;
275 }
276
ipv6_rcv(struct sk_buff * skb,struct net_device * dev,struct packet_type * pt,struct net_device * orig_dev)277 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
278 {
279 struct net *net = dev_net(skb->dev);
280
281 skb = ip6_rcv_core(skb, dev, net);
282 if (skb == NULL)
283 return NET_RX_DROP;
284 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
285 net, NULL, skb, dev, NULL,
286 ip6_rcv_finish);
287 }
288
ip6_sublist_rcv(struct list_head * head,struct net_device * dev,struct net * net)289 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
290 struct net *net)
291 {
292 NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL,
293 head, dev, NULL, ip6_rcv_finish);
294 ip6_list_rcv_finish(net, NULL, head);
295 }
296
297 /* Receive a list of IPv6 packets */
ipv6_list_rcv(struct list_head * head,struct packet_type * pt,struct net_device * orig_dev)298 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
299 struct net_device *orig_dev)
300 {
301 struct net_device *curr_dev = NULL;
302 struct net *curr_net = NULL;
303 struct sk_buff *skb, *next;
304 struct list_head sublist;
305
306 INIT_LIST_HEAD(&sublist);
307 list_for_each_entry_safe(skb, next, head, list) {
308 struct net_device *dev = skb->dev;
309 struct net *net = dev_net(dev);
310
311 skb_list_del_init(skb);
312 skb = ip6_rcv_core(skb, dev, net);
313 if (skb == NULL)
314 continue;
315
316 if (curr_dev != dev || curr_net != net) {
317 /* dispatch old sublist */
318 if (!list_empty(&sublist))
319 ip6_sublist_rcv(&sublist, curr_dev, curr_net);
320 /* start new sublist */
321 INIT_LIST_HEAD(&sublist);
322 curr_dev = dev;
323 curr_net = net;
324 }
325 list_add_tail(&skb->list, &sublist);
326 }
327 /* dispatch final sublist */
328 ip6_sublist_rcv(&sublist, curr_dev, curr_net);
329 }
330
331 INDIRECT_CALLABLE_DECLARE(int udpv6_rcv(struct sk_buff *));
332 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
333
334 /*
335 * Deliver the packet to the host
336 */
ip6_protocol_deliver_rcu(struct net * net,struct sk_buff * skb,int nexthdr,bool have_final)337 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
338 bool have_final)
339 {
340 const struct inet6_protocol *ipprot;
341 struct inet6_dev *idev;
342 unsigned int nhoff;
343 bool raw;
344
345 /*
346 * Parse extension headers
347 */
348
349 resubmit:
350 idev = ip6_dst_idev(skb_dst(skb));
351 nhoff = IP6CB(skb)->nhoff;
352 if (!have_final) {
353 if (!pskb_pull(skb, skb_transport_offset(skb)))
354 goto discard;
355 nexthdr = skb_network_header(skb)[nhoff];
356 }
357
358 resubmit_final:
359 raw = raw6_local_deliver(skb, nexthdr);
360 ipprot = rcu_dereference(inet6_protos[nexthdr]);
361 if (ipprot) {
362 int ret;
363
364 if (have_final) {
365 if (!(ipprot->flags & INET6_PROTO_FINAL)) {
366 /* Once we've seen a final protocol don't
367 * allow encapsulation on any non-final
368 * ones. This allows foo in UDP encapsulation
369 * to work.
370 */
371 goto discard;
372 }
373 } else if (ipprot->flags & INET6_PROTO_FINAL) {
374 const struct ipv6hdr *hdr;
375 int sdif = inet6_sdif(skb);
376 struct net_device *dev;
377
378 /* Only do this once for first final protocol */
379 have_final = true;
380
381 /* Free reference early: we don't need it any more,
382 and it may hold ip_conntrack module loaded
383 indefinitely. */
384 nf_reset_ct(skb);
385
386 skb_postpull_rcsum(skb, skb_network_header(skb),
387 skb_network_header_len(skb));
388 hdr = ipv6_hdr(skb);
389
390 /* skb->dev passed may be master dev for vrfs. */
391 if (sdif) {
392 dev = dev_get_by_index_rcu(net, sdif);
393 if (!dev)
394 goto discard;
395 } else {
396 dev = skb->dev;
397 }
398
399 if (ipv6_addr_is_multicast(&hdr->daddr) &&
400 !ipv6_chk_mcast_addr(dev, &hdr->daddr,
401 &hdr->saddr) &&
402 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
403 goto discard;
404 }
405 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
406 !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
407 goto discard;
408
409 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
410 skb);
411 if (ret > 0) {
412 if (ipprot->flags & INET6_PROTO_FINAL) {
413 /* Not an extension header, most likely UDP
414 * encapsulation. Use return value as nexthdr
415 * protocol not nhoff (which presumably is
416 * not set by handler).
417 */
418 nexthdr = ret;
419 goto resubmit_final;
420 } else {
421 goto resubmit;
422 }
423 } else if (ret == 0) {
424 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
425 }
426 } else {
427 if (!raw) {
428 if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
429 __IP6_INC_STATS(net, idev,
430 IPSTATS_MIB_INUNKNOWNPROTOS);
431 icmpv6_send(skb, ICMPV6_PARAMPROB,
432 ICMPV6_UNK_NEXTHDR, nhoff);
433 }
434 kfree_skb(skb);
435 } else {
436 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
437 consume_skb(skb);
438 }
439 }
440 return;
441
442 discard:
443 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
444 kfree_skb(skb);
445 }
446
ip6_input_finish(struct net * net,struct sock * sk,struct sk_buff * skb)447 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
448 {
449 rcu_read_lock();
450 ip6_protocol_deliver_rcu(net, skb, 0, false);
451 rcu_read_unlock();
452
453 return 0;
454 }
455
456
ip6_input(struct sk_buff * skb)457 int ip6_input(struct sk_buff *skb)
458 {
459 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
460 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
461 ip6_input_finish);
462 }
463 EXPORT_SYMBOL_GPL(ip6_input);
464
ip6_mc_input(struct sk_buff * skb)465 int ip6_mc_input(struct sk_buff *skb)
466 {
467 int sdif = inet6_sdif(skb);
468 const struct ipv6hdr *hdr;
469 struct net_device *dev;
470 bool deliver;
471
472 __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
473 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
474 skb->len);
475
476 /* skb->dev passed may be master dev for vrfs. */
477 if (sdif) {
478 rcu_read_lock();
479 dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
480 if (!dev) {
481 rcu_read_unlock();
482 kfree_skb(skb);
483 return -ENODEV;
484 }
485 } else {
486 dev = skb->dev;
487 }
488
489 hdr = ipv6_hdr(skb);
490 deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
491 if (sdif)
492 rcu_read_unlock();
493
494 #ifdef CONFIG_IPV6_MROUTE
495 /*
496 * IPv6 multicast router mode is now supported ;)
497 */
498 if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
499 !(ipv6_addr_type(&hdr->daddr) &
500 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
501 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
502 /*
503 * Okay, we try to forward - split and duplicate
504 * packets.
505 */
506 struct sk_buff *skb2;
507 struct inet6_skb_parm *opt = IP6CB(skb);
508
509 /* Check for MLD */
510 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
511 /* Check if this is a mld message */
512 u8 nexthdr = hdr->nexthdr;
513 __be16 frag_off;
514 int offset;
515
516 /* Check if the value of Router Alert
517 * is for MLD (0x0000).
518 */
519 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
520 deliver = false;
521
522 if (!ipv6_ext_hdr(nexthdr)) {
523 /* BUG */
524 goto out;
525 }
526 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
527 &nexthdr, &frag_off);
528 if (offset < 0)
529 goto out;
530
531 if (ipv6_is_mld(skb, nexthdr, offset))
532 deliver = true;
533
534 goto out;
535 }
536 /* unknown RA - process it normally */
537 }
538
539 if (deliver)
540 skb2 = skb_clone(skb, GFP_ATOMIC);
541 else {
542 skb2 = skb;
543 skb = NULL;
544 }
545
546 if (skb2) {
547 ip6_mr_input(skb2);
548 }
549 }
550 out:
551 #endif
552 if (likely(deliver))
553 ip6_input(skb);
554 else {
555 /* discard */
556 kfree_skb(skb);
557 }
558
559 return 0;
560 }
561
