projects/linux·blob master · refs

643 lines · 15.7 KiB

1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * IPv6 input
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <[email protected]>
8 * Ian P. Morris <[email protected]>
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#include <net/inet6_hashtables.h>
48
49static void tcp_v6_early_demux(struct sk_buff *skb)
50{
51 struct net *net = dev_net_rcu(skb->dev);
52 const struct ipv6hdr *hdr;
53 const struct tcphdr *th;
54 struct sock *sk;
55
56 if (skb->pkt_type != PACKET_HOST)
57 return;
58
59 if (!pskb_may_pull(skb, skb_transport_offset(skb) +
60 sizeof(struct tcphdr)))
61 return;
62
63 hdr = ipv6_hdr(skb);
64 th = tcp_hdr(skb);
65
66 if (th->doff < sizeof(struct tcphdr) / 4)
67 return;
68
69 /* Note : We use inet6_iif() here, not tcp_v6_iif() */
70 sk = __inet6_lookup_established(net, &hdr->saddr, th->source,
71 &hdr->daddr, ntohs(th->dest),
72 inet6_iif(skb), inet6_sdif(skb));
73 if (sk) {
74 skb->sk = sk;
75 skb->destructor = sock_edemux;
76 if (sk_fullsock(sk)) {
77 struct dst_entry *dst = rcu_dereference(sk->sk_rx_dst);
78
79 if (dst)
80 dst = dst_check(dst, sk->sk_rx_dst_cookie);
81 if (dst &&
82 sk->sk_rx_dst_ifindex == skb->skb_iif)
83 skb_dst_set_noref(skb, dst);
84 }
85 }
86}
87
88static void ip6_rcv_finish_core(struct net *net, struct sock *sk,
89 struct sk_buff *skb)
90{
91 if (READ_ONCE(net->ipv4.sysctl_ip_early_demux) &&
92 !skb_dst(skb) && !skb->sk) {
93 switch (ipv6_hdr(skb)->nexthdr) {
94 case IPPROTO_TCP:
95 if (READ_ONCE(net->ipv4.sysctl_tcp_early_demux))
96 tcp_v6_early_demux(skb);
97 break;
98 case IPPROTO_UDP:
99 if (READ_ONCE(net->ipv4.sysctl_udp_early_demux))
100 udp_v6_early_demux(skb);
101 break;
102 }
103 }
104
105 if (!skb_valid_dst(skb))
106 ip6_route_input(skb);
107}
108
109int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
110{
111 /* if ingress device is enslaved to an L3 master device pass the
112 * skb to its handler for processing
113 */
114 skb = l3mdev_ip6_rcv(skb);
115 if (!skb)
116 return NET_RX_SUCCESS;
117 ip6_rcv_finish_core(net, sk, skb);
118
119 return dst_input(skb);
120}
121
122static void ip6_sublist_rcv_finish(struct list_head *head)
123{
124 struct sk_buff *skb, *next;
125
126 list_for_each_entry_safe(skb, next, head, list) {
127 skb_list_del_init(skb);
128 dst_input(skb);
129 }
130}
131
132static bool ip6_can_use_hint(const struct sk_buff *skb,
133 const struct sk_buff *hint)
134{
135 return hint && !skb_dst(skb) &&
136 ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr);
137}
138
139static struct sk_buff *ip6_extract_route_hint(const struct net *net,
140 struct sk_buff *skb)
141{
142 if (fib6_routes_require_src(net) || fib6_has_custom_rules(net) ||
143 IP6CB(skb)->flags & IP6SKB_MULTIPATH)
144 return NULL;
145
146 return skb;
147}
148
149static void ip6_list_rcv_finish(struct net *net, struct sock *sk,
150 struct list_head *head)
151{
152 struct sk_buff *skb, *next, *hint = NULL;
153 struct dst_entry *curr_dst = NULL;
154 LIST_HEAD(sublist);
155
156 list_for_each_entry_safe(skb, next, head, list) {
157 struct dst_entry *dst;
158
159 skb_list_del_init(skb);
160 /* if ingress device is enslaved to an L3 master device pass the
161 * skb to its handler for processing
162 */
163 skb = l3mdev_ip6_rcv(skb);
164 if (!skb)
165 continue;
166
167 if (ip6_can_use_hint(skb, hint))
168 skb_dst_copy(skb, hint);
169 else
170 ip6_rcv_finish_core(net, sk, skb);
171 dst = skb_dst(skb);
172 if (curr_dst != dst) {
173 hint = ip6_extract_route_hint(net, skb);
174
175 /* dispatch old sublist */
176 if (!list_empty(&sublist))
177 ip6_sublist_rcv_finish(&sublist);
178 /* start new sublist */
179 INIT_LIST_HEAD(&sublist);
180 curr_dst = dst;
181 }
182 list_add_tail(&skb->list, &sublist);
183 }
184 /* dispatch final sublist */
185 ip6_sublist_rcv_finish(&sublist);
186}
187
188static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
189 struct net *net)
190{
191 enum skb_drop_reason reason;
192 const struct ipv6hdr *hdr;
193 u32 pkt_len;
194 struct inet6_dev *idev;
195
196 if (skb->pkt_type == PACKET_OTHERHOST) {
197 dev_core_stats_rx_otherhost_dropped_inc(skb->dev);
198 kfree_skb_reason(skb, SKB_DROP_REASON_OTHERHOST);
199 return NULL;
200 }
201
202 rcu_read_lock();
203
204 idev = __in6_dev_get(skb->dev);
205
206 __IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
207
208 SKB_DR_SET(reason, NOT_SPECIFIED);
209 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
210 !idev || unlikely(READ_ONCE(idev->cnf.disable_ipv6))) {
211 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
212 if (idev && unlikely(READ_ONCE(idev->cnf.disable_ipv6)))
213 SKB_DR_SET(reason, IPV6DISABLED);
214 goto drop;
215 }
216
217 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
218
219 /*
220 * Store incoming device index. When the packet will
221 * be queued, we cannot refer to skb->dev anymore.
222 *
223 * BTW, when we send a packet for our own local address on a
224 * non-loopback interface (e.g. ethX), it is being delivered
225 * via the loopback interface (lo) here; skb->dev = loopback_dev.
226 * It, however, should be considered as if it is being
227 * arrived via the sending interface (ethX), because of the
228 * nature of scoping architecture. --yoshfuji
229 */
230 IP6CB(skb)->iif = skb_valid_dst(skb) ?
231 ip6_dst_idev(skb_dst(skb))->dev->ifindex :
232 dev->ifindex;
233
234 if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
235 goto err;
236
237 hdr = ipv6_hdr(skb);
238
239 if (hdr->version != 6) {
240 SKB_DR_SET(reason, UNHANDLED_PROTO);
241 goto err;
242 }
243
244 __IP6_ADD_STATS(net, idev,
245 IPSTATS_MIB_NOECTPKTS +
246 (ipv6_get_dsfield(hdr) & INET_ECN_MASK),
247 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
248 /*
249 * RFC4291 2.5.3
250 * The loopback address must not be used as the source address in IPv6
251 * packets that are sent outside of a single node. [..]
252 * A packet received on an interface with a destination address
253 * of loopback must be dropped.
254 */
255 if ((ipv6_addr_loopback(&hdr->saddr) ||
256 ipv6_addr_loopback(&hdr->daddr)) &&
257 !(dev->flags & IFF_LOOPBACK) &&
258 !netif_is_l3_master(dev))
259 goto err;
260
261 /* RFC4291 Errata ID: 3480
262 * Interface-Local scope spans only a single interface on a
263 * node and is useful only for loopback transmission of
264 * multicast. Packets with interface-local scope received
265 * from another node must be discarded.
266 */
267 if (!(skb->pkt_type == PACKET_LOOPBACK ||
268 dev->flags & IFF_LOOPBACK) &&
269 ipv6_addr_is_multicast(&hdr->daddr) &&
270 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
271 goto err;
272
273 /* If enabled, drop unicast packets that were encapsulated in link-layer
274 * multicast or broadcast to protected against the so-called "hole-196"
275 * attack in 802.11 wireless.
276 */
277 if (!ipv6_addr_is_multicast(&hdr->daddr) &&
278 (skb->pkt_type == PACKET_BROADCAST ||
279 skb->pkt_type == PACKET_MULTICAST) &&
280 READ_ONCE(idev->cnf.drop_unicast_in_l2_multicast)) {
281 SKB_DR_SET(reason, UNICAST_IN_L2_MULTICAST);
282 goto err;
283 }
284
285 /* RFC4291 2.7
286 * Nodes must not originate a packet to a multicast address whose scope
287 * field contains the reserved value 0; if such a packet is received, it
288 * must be silently dropped.
289 */
290 if (ipv6_addr_is_multicast(&hdr->daddr) &&
291 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
292 goto err;
293
294 /*
295 * RFC4291 2.7
296 * Multicast addresses must not be used as source addresses in IPv6
297 * packets or appear in any Routing header.
298 */
299 if (ipv6_addr_is_multicast(&hdr->saddr))
300 goto err;
301
302 skb->transport_header = skb->network_header + sizeof(*hdr);
303 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
304
305 pkt_len = ipv6_payload_len(skb, hdr);
306
307 /* pkt_len may be zero if Jumbo payload option is present */
308 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
309 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
310 __IP6_INC_STATS(net,
311 idev, IPSTATS_MIB_INTRUNCATEDPKTS);
312 SKB_DR_SET(reason, PKT_TOO_SMALL);
313 goto drop;
314 }
315 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
316 goto err;
317 hdr = ipv6_hdr(skb);
318 }
319
320 if (hdr->nexthdr == NEXTHDR_HOP) {
321 if (ipv6_parse_hopopts(skb) < 0) {
322 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
323 rcu_read_unlock();
324 return NULL;
325 }
326 }
327
328 rcu_read_unlock();
329
330 /* Must drop socket now because of tproxy. */
331 if (!skb_sk_is_prefetched(skb))
332 skb_orphan(skb);
333
334 return skb;
335err:
336 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
337 SKB_DR_OR(reason, IP_INHDR);
338drop:
339 rcu_read_unlock();
340 kfree_skb_reason(skb, reason);
341 return NULL;
342}
343
344int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
345{
346 struct net *net = dev_net(skb->dev);
347
348 skb = ip6_rcv_core(skb, dev, net);
349 if (skb == NULL)
350 return NET_RX_DROP;
351 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
352 net, NULL, skb, dev, NULL,
353 ip6_rcv_finish);
354}
355
356static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
357 struct net *net)
358{
359 NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL,
360 head, dev, NULL, ip6_rcv_finish);
361 ip6_list_rcv_finish(net, NULL, head);
362}
363
364/* Receive a list of IPv6 packets */
365void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
366 struct net_device *orig_dev)
367{
368 struct net_device *curr_dev = NULL;
369 struct net *curr_net = NULL;
370 struct sk_buff *skb, *next;
371 LIST_HEAD(sublist);
372
373 list_for_each_entry_safe(skb, next, head, list) {
374 struct net_device *dev = skb->dev;
375 struct net *net = dev_net(dev);
376
377 skb_list_del_init(skb);
378 skb = ip6_rcv_core(skb, dev, net);
379 if (skb == NULL)
380 continue;
381
382 if (curr_dev != dev || curr_net != net) {
383 /* dispatch old sublist */
384 if (!list_empty(&sublist))
385 ip6_sublist_rcv(&sublist, curr_dev, curr_net);
386 /* start new sublist */
387 INIT_LIST_HEAD(&sublist);
388 curr_dev = dev;
389 curr_net = net;
390 }
391 list_add_tail(&skb->list, &sublist);
392 }
393 /* dispatch final sublist */
394 if (!list_empty(&sublist))
395 ip6_sublist_rcv(&sublist, curr_dev, curr_net);
396}
397
398INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
399
400/*
401 * Deliver the packet to the host
402 */
403void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
404 bool have_final)
405{
406 int exthdr_cnt = IP6CB(skb)->flags & IP6SKB_HOPBYHOP ? 1 : 0;
407 const struct inet6_protocol *ipprot;
408 struct inet6_dev *idev;
409 unsigned int nhoff;
410 SKB_DR(reason);
411 bool raw;
412
413 /*
414 * Parse extension headers
415 */
416
417resubmit:
418 idev = ip6_dst_idev(skb_dst(skb));
419 nhoff = IP6CB(skb)->nhoff;
420 if (!have_final) {
421 if (!pskb_pull(skb, skb_transport_offset(skb)))
422 goto discard;
423 nexthdr = skb_network_header(skb)[nhoff];
424 }
425
426resubmit_final:
427 raw = raw6_local_deliver(skb, nexthdr);
428 ipprot = rcu_dereference(inet6_protos[nexthdr]);
429 if (ipprot) {
430 int ret;
431
432 if (have_final) {
433 if (!(ipprot->flags & INET6_PROTO_FINAL)) {
434 /* Once we've seen a final protocol don't
435 * allow encapsulation on any non-final
436 * ones. This allows foo in UDP encapsulation
437 * to work.
438 */
439 goto discard;
440 }
441 } else if (ipprot->flags & INET6_PROTO_FINAL) {
442 const struct ipv6hdr *hdr;
443 int sdif = inet6_sdif(skb);
444 struct net_device *dev;
445
446 /* Only do this once for first final protocol */
447 have_final = true;
448
449
450 skb_postpull_rcsum(skb, skb_network_header(skb),
451 skb_network_header_len(skb));
452 hdr = ipv6_hdr(skb);
453
454 /* skb->dev passed may be master dev for vrfs. */
455 if (sdif) {
456 dev = dev_get_by_index_rcu(net, sdif);
457 if (!dev)
458 goto discard;
459 } else {
460 dev = skb->dev;
461 }
462
463 if (ipv6_addr_is_multicast(&hdr->daddr) &&
464 !ipv6_chk_mcast_addr(dev, &hdr->daddr,
465 &hdr->saddr) &&
466 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb))) {
467 SKB_DR_SET(reason, IP_INADDRERRORS);
468 goto discard;
469 }
470 }
471 if (!(ipprot->flags & INET6_PROTO_NOPOLICY)) {
472 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
473 SKB_DR_SET(reason, XFRM_POLICY);
474 goto discard;
475 }
476 nf_reset_ct(skb);
477 }
478
479 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
480 skb);
481 if (ret > 0) {
482 if (ipprot->flags & INET6_PROTO_FINAL) {
483 /* Not an extension header, most likely UDP
484 * encapsulation. Use return value as nexthdr
485 * protocol not nhoff (which presumably is
486 * not set by handler).
487 */
488 nexthdr = ret;
489 goto resubmit_final;
490 } else {
491 if (unlikely(exthdr_cnt++ >= IP6_MAX_EXT_HDRS_CNT)) {
492 SKB_DR_SET(reason, IPV6_TOO_MANY_EXTHDRS);
493 goto discard;
494 }
495 goto resubmit;
496 }
497 } else if (ret == 0) {
498 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
499 }
500 } else {
501 if (!raw) {
502 if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
503 __IP6_INC_STATS(net, idev,
504 IPSTATS_MIB_INUNKNOWNPROTOS);
505 icmpv6_send(skb, ICMPV6_PARAMPROB,
506 ICMPV6_UNK_NEXTHDR, nhoff);
507 SKB_DR_SET(reason, IP_NOPROTO);
508 } else {
509 SKB_DR_SET(reason, XFRM_POLICY);
510 }
511 kfree_skb_reason(skb, reason);
512 } else {
513 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
514 consume_skb(skb);
515 }
516 }
517 return;
518
519discard:
520 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
521 kfree_skb_reason(skb, reason);
522}
523
524static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
525{
526 if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) {
527 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
528 IPSTATS_MIB_INDISCARDS);
529 kfree_skb_reason(skb, SKB_DROP_REASON_NOMEM);
530 return 0;
531 }
532
533 skb_clear_delivery_time(skb);
534 ip6_protocol_deliver_rcu(net, skb, 0, false);
535
536 return 0;
537}
538
539
540int ip6_input(struct sk_buff *skb)
541{
542 int res;
543
544 rcu_read_lock();
545 res = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
546 dev_net_rcu(skb->dev), NULL, skb, skb->dev, NULL,
547 ip6_input_finish);
548 rcu_read_unlock();
549
550 return res;
551}
552EXPORT_SYMBOL_GPL(ip6_input);
553
554int ip6_mc_input(struct sk_buff *skb)
555{
556 struct net_device *dev = skb->dev;
557 int sdif = inet6_sdif(skb);
558 const struct ipv6hdr *hdr;
559 bool deliver;
560
561 __IP6_UPD_PO_STATS(skb_dst_dev_net_rcu(skb),
562 __in6_dev_get_safely(dev), IPSTATS_MIB_INMCAST,
563 skb->len);
564
565 /* skb->dev passed may be master dev for vrfs. */
566 if (sdif) {
567 dev = dev_get_by_index_rcu(dev_net_rcu(dev), sdif);
568 if (!dev) {
569 kfree_skb(skb);
570 return -ENODEV;
571 }
572 }
573
574 hdr = ipv6_hdr(skb);
575 deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
576
577#ifdef CONFIG_IPV6_MROUTE
578 /*
579 * IPv6 multicast router mode is now supported ;)
580 */
581 if (atomic_read(&dev_net_rcu(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
582 !(ipv6_addr_type(&hdr->daddr) &
583 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
584 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
585 /*
586 * Okay, we try to forward - split and duplicate
587 * packets.
588 */
589 struct sk_buff *skb2;
590 struct inet6_skb_parm *opt = IP6CB(skb);
591
592 /* Check for MLD */
593 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
594 /* Check if this is a mld message */
595 u8 nexthdr = hdr->nexthdr;
596 __be16 frag_off;
597 int offset;
598
599 /* Check if the value of Router Alert
600 * is for MLD (0x0000).
601 */
602 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
603 deliver = false;
604
605 if (!ipv6_ext_hdr(nexthdr)) {
606 /* BUG */
607 goto out;
608 }
609 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
610 &nexthdr, &frag_off);
611 if (offset < 0)
612 goto out;
613
614 if (ipv6_is_mld(skb, nexthdr, offset))
615 deliver = true;
616
617 goto out;
618 }
619 /* unknown RA - process it normally */
620 }
621
622 if (deliver) {
623 skb2 = skb_clone(skb, GFP_ATOMIC);
624 } else {
625 skb_orphan(skb);
626 skb2 = skb;
627 skb = NULL;
628 }
629
630 if (skb2)
631 ip6_mr_input(skb2);
632 }
633out:
634#endif
635 if (likely(deliver)) {
636 ip6_input(skb);
637 } else {
638 /* discard */
639 kfree_skb(skb);
640 }
641
642 return 0;
643}