1 // SPDX-License-Identifier: GPL-2.0-or-later
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Ian P. Morris <I.P.Morris@soton.ac.uk>
10 * Based in linux/net/ipv4/ip_input.c
14 * Mitsuru KANDA @USAGI and
15 * YOSHIFUJI Hideaki @USAGI: Remove ipv6_parse_exthdrs().
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>
30 #include <linux/netfilter.h>
31 #include <linux/netfilter_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>
45 #include <net/inet_ecn.h>
46 #include <net/dst_metadata.h>
48 static void ip6_rcv_finish_core(struct net *net, struct sock *sk,
51 if (READ_ONCE(net->ipv4.sysctl_ip_early_demux) &&
52 !skb_dst(skb) && !skb->sk) {
53 switch (ipv6_hdr(skb)->nexthdr) {
55 if (READ_ONCE(net->ipv4.sysctl_tcp_early_demux))
56 tcp_v6_early_demux(skb);
59 if (READ_ONCE(net->ipv4.sysctl_udp_early_demux))
60 udp_v6_early_demux(skb);
65 if (!skb_valid_dst(skb))
69 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
71 /* if ingress device is enslaved to an L3 master device pass the
72 * skb to its handler for processing
74 skb = l3mdev_ip6_rcv(skb);
76 return NET_RX_SUCCESS;
77 ip6_rcv_finish_core(net, sk, skb);
79 return dst_input(skb);
82 static void ip6_sublist_rcv_finish(struct list_head *head)
84 struct sk_buff *skb, *next;
86 list_for_each_entry_safe(skb, next, head, list) {
87 skb_list_del_init(skb);
92 static bool ip6_can_use_hint(const struct sk_buff *skb,
93 const struct sk_buff *hint)
95 return hint && !skb_dst(skb) &&
96 ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr);
99 static struct sk_buff *ip6_extract_route_hint(const struct net *net,
102 if (fib6_routes_require_src(net) || fib6_has_custom_rules(net))
108 static void ip6_list_rcv_finish(struct net *net, struct sock *sk,
109 struct list_head *head)
111 struct sk_buff *skb, *next, *hint = NULL;
112 struct dst_entry *curr_dst = NULL;
113 struct list_head sublist;
115 INIT_LIST_HEAD(&sublist);
116 list_for_each_entry_safe(skb, next, head, list) {
117 struct dst_entry *dst;
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
123 skb = l3mdev_ip6_rcv(skb);
127 if (ip6_can_use_hint(skb, hint))
128 skb_dst_copy(skb, hint);
130 ip6_rcv_finish_core(net, sk, skb);
132 if (curr_dst != dst) {
133 hint = ip6_extract_route_hint(net, skb);
135 /* dispatch old sublist */
136 if (!list_empty(&sublist))
137 ip6_sublist_rcv_finish(&sublist);
138 /* start new sublist */
139 INIT_LIST_HEAD(&sublist);
142 list_add_tail(&skb->list, &sublist);
144 /* dispatch final sublist */
145 ip6_sublist_rcv_finish(&sublist);
148 static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
151 const struct ipv6hdr *hdr;
153 struct inet6_dev *idev;
155 if (skb->pkt_type == PACKET_OTHERHOST) {
162 idev = __in6_dev_get(skb->dev);
164 __IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
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);
172 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
175 * Store incoming device index. When the packet will
176 * be queued, we cannot refer to skb->dev anymore.
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
185 IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
187 if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
192 if (hdr->version != 6)
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));
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.
206 if ((ipv6_addr_loopback(&hdr->saddr) ||
207 ipv6_addr_loopback(&hdr->daddr)) &&
208 !(dev->flags & IFF_LOOPBACK) &&
209 !netif_is_l3_master(dev))
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.
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)
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.
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)
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.
239 if (ipv6_addr_is_multicast(&hdr->daddr) &&
240 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
245 * Multicast addresses must not be used as source addresses in IPv6
246 * packets or appear in any Routing header.
248 if (ipv6_addr_is_multicast(&hdr->saddr))
251 skb->transport_header = skb->network_header + sizeof(*hdr);
252 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
254 pkt_len = ntohs(hdr->payload_len);
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) {
260 idev, IPSTATS_MIB_INTRUNCATEDPKTS);
263 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
264 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
270 if (hdr->nexthdr == NEXTHDR_HOP) {
271 if (ipv6_parse_hopopts(skb) < 0) {
272 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
280 /* Must drop socket now because of tproxy. */
281 if (!skb_sk_is_prefetched(skb))
286 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
293 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
295 struct net *net = dev_net(skb->dev);
297 skb = ip6_rcv_core(skb, dev, net);
300 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
301 net, NULL, skb, dev, NULL,
305 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
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);
313 /* Receive a list of IPv6 packets */
314 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
315 struct net_device *orig_dev)
317 struct net_device *curr_dev = NULL;
318 struct net *curr_net = NULL;
319 struct sk_buff *skb, *next;
320 struct list_head sublist;
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);
327 skb_list_del_init(skb);
328 skb = ip6_rcv_core(skb, dev, net);
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);
341 list_add_tail(&skb->list, &sublist);
343 /* dispatch final sublist */
344 if (!list_empty(&sublist))
345 ip6_sublist_rcv(&sublist, curr_dev, curr_net);
348 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
351 * Deliver the packet to the host
353 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
356 const struct inet6_protocol *ipprot;
357 struct inet6_dev *idev;
362 * Parse extension headers
366 idev = ip6_dst_idev(skb_dst(skb));
367 nhoff = IP6CB(skb)->nhoff;
369 if (!pskb_pull(skb, skb_transport_offset(skb)))
371 nexthdr = skb_network_header(skb)[nhoff];
375 raw = raw6_local_deliver(skb, nexthdr);
376 ipprot = rcu_dereference(inet6_protos[nexthdr]);
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
389 } else if (ipprot->flags & INET6_PROTO_FINAL) {
390 const struct ipv6hdr *hdr;
391 int sdif = inet6_sdif(skb);
392 struct net_device *dev;
394 /* Only do this once for first final protocol */
397 /* Free reference early: we don't need it any more,
398 and it may hold ip_conntrack module loaded
402 skb_postpull_rcsum(skb, skb_network_header(skb),
403 skb_network_header_len(skb));
406 /* skb->dev passed may be master dev for vrfs. */
408 dev = dev_get_by_index_rcu(net, sdif);
415 if (ipv6_addr_is_multicast(&hdr->daddr) &&
416 !ipv6_chk_mcast_addr(dev, &hdr->daddr,
418 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
421 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
422 !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
425 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
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).
439 } else if (ret == 0) {
440 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
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);
452 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
459 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
463 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
466 ip6_protocol_deliver_rcu(net, skb, 0, false);
473 int ip6_input(struct sk_buff *skb)
475 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
476 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
479 EXPORT_SYMBOL_GPL(ip6_input);
481 int ip6_mc_input(struct sk_buff *skb)
483 int sdif = inet6_sdif(skb);
484 const struct ipv6hdr *hdr;
485 struct net_device *dev;
488 __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
489 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
492 /* skb->dev passed may be master dev for vrfs. */
495 dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
506 deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
510 #ifdef CONFIG_IPV6_MROUTE
512 * IPv6 multicast router mode is now supported ;)
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))) {
519 * Okay, we try to forward - split and duplicate
522 struct sk_buff *skb2;
523 struct inet6_skb_parm *opt = IP6CB(skb);
526 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
527 /* Check if this is a mld message */
528 u8 nexthdr = hdr->nexthdr;
532 /* Check if the value of Router Alert
533 * is for MLD (0x0000).
535 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
538 if (!ipv6_ext_hdr(nexthdr)) {
542 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
543 &nexthdr, &frag_off);
547 if (ipv6_is_mld(skb, nexthdr, offset))
552 /* unknown RA - process it normally */
556 skb2 = skb_clone(skb, GFP_ATOMIC);