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 INDIRECT_CALLABLE_DECLARE(void tcp_v6_early_demux(struct sk_buff *));
49 static void ip6_rcv_finish_core(struct net *net, struct sock *sk,
52 void (*edemux)(struct sk_buff *skb);
54 if (net->ipv4.sysctl_ip_early_demux && !skb_dst(skb) && skb->sk == NULL) {
55 const struct inet6_protocol *ipprot;
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);
62 if (!skb_valid_dst(skb))
66 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
68 /* if ingress device is enslaved to an L3 master device pass the
69 * skb to its handler for processing
71 skb = l3mdev_ip6_rcv(skb);
73 return NET_RX_SUCCESS;
74 ip6_rcv_finish_core(net, sk, skb);
76 return dst_input(skb);
79 static void ip6_sublist_rcv_finish(struct list_head *head)
81 struct sk_buff *skb, *next;
83 list_for_each_entry_safe(skb, next, head, list) {
84 skb_list_del_init(skb);
89 static bool ip6_can_use_hint(const struct sk_buff *skb,
90 const struct sk_buff *hint)
92 return hint && !skb_dst(skb) &&
93 ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr);
96 static struct sk_buff *ip6_extract_route_hint(const struct net *net,
99 if (fib6_routes_require_src(net) || fib6_has_custom_rules(net))
105 static void ip6_list_rcv_finish(struct net *net, struct sock *sk,
106 struct list_head *head)
108 struct sk_buff *skb, *next, *hint = NULL;
109 struct dst_entry *curr_dst = NULL;
110 struct list_head sublist;
112 INIT_LIST_HEAD(&sublist);
113 list_for_each_entry_safe(skb, next, head, list) {
114 struct dst_entry *dst;
116 skb_list_del_init(skb);
117 /* if ingress device is enslaved to an L3 master device pass the
118 * skb to its handler for processing
120 skb = l3mdev_ip6_rcv(skb);
124 if (ip6_can_use_hint(skb, hint))
125 skb_dst_copy(skb, hint);
127 ip6_rcv_finish_core(net, sk, skb);
129 if (curr_dst != dst) {
130 hint = ip6_extract_route_hint(net, skb);
132 /* dispatch old sublist */
133 if (!list_empty(&sublist))
134 ip6_sublist_rcv_finish(&sublist);
135 /* start new sublist */
136 INIT_LIST_HEAD(&sublist);
139 list_add_tail(&skb->list, &sublist);
141 /* dispatch final sublist */
142 ip6_sublist_rcv_finish(&sublist);
145 static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev,
148 enum skb_drop_reason reason;
149 const struct ipv6hdr *hdr;
151 struct inet6_dev *idev;
153 if (skb->pkt_type == PACKET_OTHERHOST) {
154 dev_core_stats_rx_otherhost_dropped_inc(skb->dev);
155 kfree_skb_reason(skb, SKB_DROP_REASON_OTHERHOST);
161 idev = __in6_dev_get(skb->dev);
163 __IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
165 SKB_DR_SET(reason, NOT_SPECIFIED);
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 if (idev && unlikely(idev->cnf.disable_ipv6))
170 SKB_DR_SET(reason, IPV6DISABLED);
174 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
177 * Store incoming device index. When the packet will
178 * be queued, we cannot refer to skb->dev anymore.
180 * BTW, when we send a packet for our own local address on a
181 * non-loopback interface (e.g. ethX), it is being delivered
182 * via the loopback interface (lo) here; skb->dev = loopback_dev.
183 * It, however, should be considered as if it is being
184 * arrived via the sending interface (ethX), because of the
185 * nature of scoping architecture. --yoshfuji
187 IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
189 if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
194 if (hdr->version != 6) {
195 SKB_DR_SET(reason, UNHANDLED_PROTO);
199 __IP6_ADD_STATS(net, idev,
200 IPSTATS_MIB_NOECTPKTS +
201 (ipv6_get_dsfield(hdr) & INET_ECN_MASK),
202 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
205 * The loopback address must not be used as the source address in IPv6
206 * packets that are sent outside of a single node. [..]
207 * A packet received on an interface with a destination address
208 * of loopback must be dropped.
210 if ((ipv6_addr_loopback(&hdr->saddr) ||
211 ipv6_addr_loopback(&hdr->daddr)) &&
212 !(dev->flags & IFF_LOOPBACK) &&
213 !netif_is_l3_master(dev))
216 /* RFC4291 Errata ID: 3480
217 * Interface-Local scope spans only a single interface on a
218 * node and is useful only for loopback transmission of
219 * multicast. Packets with interface-local scope received
220 * from another node must be discarded.
222 if (!(skb->pkt_type == PACKET_LOOPBACK ||
223 dev->flags & IFF_LOOPBACK) &&
224 ipv6_addr_is_multicast(&hdr->daddr) &&
225 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
228 /* If enabled, drop unicast packets that were encapsulated in link-layer
229 * multicast or broadcast to protected against the so-called "hole-196"
230 * attack in 802.11 wireless.
232 if (!ipv6_addr_is_multicast(&hdr->daddr) &&
233 (skb->pkt_type == PACKET_BROADCAST ||
234 skb->pkt_type == PACKET_MULTICAST) &&
235 idev->cnf.drop_unicast_in_l2_multicast) {
236 SKB_DR_SET(reason, UNICAST_IN_L2_MULTICAST);
241 * Nodes must not originate a packet to a multicast address whose scope
242 * field contains the reserved value 0; if such a packet is received, it
243 * must be silently dropped.
245 if (ipv6_addr_is_multicast(&hdr->daddr) &&
246 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
251 * Multicast addresses must not be used as source addresses in IPv6
252 * packets or appear in any Routing header.
254 if (ipv6_addr_is_multicast(&hdr->saddr))
257 skb->transport_header = skb->network_header + sizeof(*hdr);
258 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
260 pkt_len = ntohs(hdr->payload_len);
262 /* pkt_len may be zero if Jumbo payload option is present */
263 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
264 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
266 idev, IPSTATS_MIB_INTRUNCATEDPKTS);
267 SKB_DR_SET(reason, PKT_TOO_SMALL);
270 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
275 if (hdr->nexthdr == NEXTHDR_HOP) {
276 if (ipv6_parse_hopopts(skb) < 0) {
277 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
285 /* Must drop socket now because of tproxy. */
286 if (!skb_sk_is_prefetched(skb))
291 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
292 SKB_DR_OR(reason, IP_INHDR);
295 kfree_skb_reason(skb, reason);
299 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
301 struct net *net = dev_net(skb->dev);
303 skb = ip6_rcv_core(skb, dev, net);
306 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
307 net, NULL, skb, dev, NULL,
311 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev,
314 NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL,
315 head, dev, NULL, ip6_rcv_finish);
316 ip6_list_rcv_finish(net, NULL, head);
319 /* Receive a list of IPv6 packets */
320 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
321 struct net_device *orig_dev)
323 struct net_device *curr_dev = NULL;
324 struct net *curr_net = NULL;
325 struct sk_buff *skb, *next;
326 struct list_head sublist;
328 INIT_LIST_HEAD(&sublist);
329 list_for_each_entry_safe(skb, next, head, list) {
330 struct net_device *dev = skb->dev;
331 struct net *net = dev_net(dev);
333 skb_list_del_init(skb);
334 skb = ip6_rcv_core(skb, dev, net);
338 if (curr_dev != dev || curr_net != net) {
339 /* dispatch old sublist */
340 if (!list_empty(&sublist))
341 ip6_sublist_rcv(&sublist, curr_dev, curr_net);
342 /* start new sublist */
343 INIT_LIST_HEAD(&sublist);
347 list_add_tail(&skb->list, &sublist);
349 /* dispatch final sublist */
350 if (!list_empty(&sublist))
351 ip6_sublist_rcv(&sublist, curr_dev, curr_net);
354 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *));
357 * Deliver the packet to the host
359 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
362 const struct inet6_protocol *ipprot;
363 struct inet6_dev *idev;
369 * Parse extension headers
373 idev = ip6_dst_idev(skb_dst(skb));
374 nhoff = IP6CB(skb)->nhoff;
376 if (!pskb_pull(skb, skb_transport_offset(skb)))
378 nexthdr = skb_network_header(skb)[nhoff];
382 raw = raw6_local_deliver(skb, nexthdr);
383 ipprot = rcu_dereference(inet6_protos[nexthdr]);
388 if (!(ipprot->flags & INET6_PROTO_FINAL)) {
389 /* Once we've seen a final protocol don't
390 * allow encapsulation on any non-final
391 * ones. This allows foo in UDP encapsulation
396 } else if (ipprot->flags & INET6_PROTO_FINAL) {
397 const struct ipv6hdr *hdr;
398 int sdif = inet6_sdif(skb);
399 struct net_device *dev;
401 /* Only do this once for first final protocol */
404 /* Free reference early: we don't need it any more,
405 and it may hold ip_conntrack module loaded
409 skb_postpull_rcsum(skb, skb_network_header(skb),
410 skb_network_header_len(skb));
413 /* skb->dev passed may be master dev for vrfs. */
415 dev = dev_get_by_index_rcu(net, sdif);
422 if (ipv6_addr_is_multicast(&hdr->daddr) &&
423 !ipv6_chk_mcast_addr(dev, &hdr->daddr,
425 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb))) {
426 SKB_DR_SET(reason, IP_INADDRERRORS);
430 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
431 !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
432 SKB_DR_SET(reason, XFRM_POLICY);
436 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv,
439 if (ipprot->flags & INET6_PROTO_FINAL) {
440 /* Not an extension header, most likely UDP
441 * encapsulation. Use return value as nexthdr
442 * protocol not nhoff (which presumably is
443 * not set by handler).
450 } else if (ret == 0) {
451 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
455 if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
456 __IP6_INC_STATS(net, idev,
457 IPSTATS_MIB_INUNKNOWNPROTOS);
458 icmpv6_send(skb, ICMPV6_PARAMPROB,
459 ICMPV6_UNK_NEXTHDR, nhoff);
460 SKB_DR_SET(reason, IP_NOPROTO);
462 SKB_DR_SET(reason, XFRM_POLICY);
464 kfree_skb_reason(skb, reason);
466 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
473 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
474 kfree_skb_reason(skb, reason);
477 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
479 skb_clear_delivery_time(skb);
481 ip6_protocol_deliver_rcu(net, skb, 0, false);
488 int ip6_input(struct sk_buff *skb)
490 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
491 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
494 EXPORT_SYMBOL_GPL(ip6_input);
496 int ip6_mc_input(struct sk_buff *skb)
498 int sdif = inet6_sdif(skb);
499 const struct ipv6hdr *hdr;
500 struct net_device *dev;
503 __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
504 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
507 /* skb->dev passed may be master dev for vrfs. */
510 dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
521 deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
525 #ifdef CONFIG_IPV6_MROUTE
527 * IPv6 multicast router mode is now supported ;)
529 if (atomic_read(&dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
530 !(ipv6_addr_type(&hdr->daddr) &
531 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
532 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
534 * Okay, we try to forward - split and duplicate
537 struct sk_buff *skb2;
538 struct inet6_skb_parm *opt = IP6CB(skb);
541 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
542 /* Check if this is a mld message */
543 u8 nexthdr = hdr->nexthdr;
547 /* Check if the value of Router Alert
548 * is for MLD (0x0000).
550 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
553 if (!ipv6_ext_hdr(nexthdr)) {
557 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
558 &nexthdr, &frag_off);
562 if (ipv6_is_mld(skb, nexthdr, offset))
567 /* unknown RA - process it normally */
571 skb2 = skb_clone(skb, GFP_ATOMIC);