These changes are the raw update to linux-4.4.6-rt14. Kernel sources
[kvmfornfv.git] / kernel / net / ipv6 / ip6_input.c
1 /*
2  *      IPv6 input
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Ian P. Morris           <I.P.Morris@soton.ac.uk>
8  *
9  *      Based in linux/net/ipv4/ip_input.c
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16 /* Changes
17  *
18  *      Mitsuru KANDA @USAGI and
19  *      YOSHIFUJI Hideaki @USAGI: Remove ipv6_parse_exthdrs().
20  */
21
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/netdevice.h>
28 #include <linux/in6.h>
29 #include <linux/icmpv6.h>
30 #include <linux/mroute6.h>
31 #include <linux/slab.h>
32
33 #include <linux/netfilter.h>
34 #include <linux/netfilter_ipv6.h>
35
36 #include <net/sock.h>
37 #include <net/snmp.h>
38
39 #include <net/ipv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/rawv6.h>
43 #include <net/ndisc.h>
44 #include <net/ip6_route.h>
45 #include <net/addrconf.h>
46 #include <net/xfrm.h>
47 #include <net/inet_ecn.h>
48 #include <net/dst_metadata.h>
49
50 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
51 {
52         if (sysctl_ip_early_demux && !skb_dst(skb) && skb->sk == NULL) {
53                 const struct inet6_protocol *ipprot;
54
55                 ipprot = rcu_dereference(inet6_protos[ipv6_hdr(skb)->nexthdr]);
56                 if (ipprot && ipprot->early_demux)
57                         ipprot->early_demux(skb);
58         }
59         if (!skb_valid_dst(skb))
60                 ip6_route_input(skb);
61
62         return dst_input(skb);
63 }
64
65 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
66 {
67         const struct ipv6hdr *hdr;
68         u32 pkt_len;
69         struct inet6_dev *idev;
70         struct net *net = dev_net(skb->dev);
71
72         if (skb->pkt_type == PACKET_OTHERHOST) {
73                 kfree_skb(skb);
74                 return NET_RX_DROP;
75         }
76
77         rcu_read_lock();
78
79         idev = __in6_dev_get(skb->dev);
80
81         IP6_UPD_PO_STATS_BH(net, idev, IPSTATS_MIB_IN, skb->len);
82
83         if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
84             !idev || unlikely(idev->cnf.disable_ipv6)) {
85                 IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDISCARDS);
86                 goto drop;
87         }
88
89         memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
90
91         /*
92          * Store incoming device index. When the packet will
93          * be queued, we cannot refer to skb->dev anymore.
94          *
95          * BTW, when we send a packet for our own local address on a
96          * non-loopback interface (e.g. ethX), it is being delivered
97          * via the loopback interface (lo) here; skb->dev = loopback_dev.
98          * It, however, should be considered as if it is being
99          * arrived via the sending interface (ethX), because of the
100          * nature of scoping architecture. --yoshfuji
101          */
102         IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
103
104         if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
105                 goto err;
106
107         hdr = ipv6_hdr(skb);
108
109         if (hdr->version != 6)
110                 goto err;
111
112         IP6_ADD_STATS_BH(net, idev,
113                          IPSTATS_MIB_NOECTPKTS +
114                                 (ipv6_get_dsfield(hdr) & INET_ECN_MASK),
115                          max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
116         /*
117          * RFC4291 2.5.3
118          * A packet received on an interface with a destination address
119          * of loopback must be dropped.
120          */
121         if (!(dev->flags & IFF_LOOPBACK) &&
122             ipv6_addr_loopback(&hdr->daddr))
123                 goto err;
124
125         /* RFC4291 Errata ID: 3480
126          * Interface-Local scope spans only a single interface on a
127          * node and is useful only for loopback transmission of
128          * multicast.  Packets with interface-local scope received
129          * from another node must be discarded.
130          */
131         if (!(skb->pkt_type == PACKET_LOOPBACK ||
132               dev->flags & IFF_LOOPBACK) &&
133             ipv6_addr_is_multicast(&hdr->daddr) &&
134             IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
135                 goto err;
136
137         /* RFC4291 2.7
138          * Nodes must not originate a packet to a multicast address whose scope
139          * field contains the reserved value 0; if such a packet is received, it
140          * must be silently dropped.
141          */
142         if (ipv6_addr_is_multicast(&hdr->daddr) &&
143             IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
144                 goto err;
145
146         /*
147          * RFC4291 2.7
148          * Multicast addresses must not be used as source addresses in IPv6
149          * packets or appear in any Routing header.
150          */
151         if (ipv6_addr_is_multicast(&hdr->saddr))
152                 goto err;
153
154         skb->transport_header = skb->network_header + sizeof(*hdr);
155         IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
156
157         pkt_len = ntohs(hdr->payload_len);
158
159         /* pkt_len may be zero if Jumbo payload option is present */
160         if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
161                 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
162                         IP6_INC_STATS_BH(net,
163                                          idev, IPSTATS_MIB_INTRUNCATEDPKTS);
164                         goto drop;
165                 }
166                 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
167                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
168                         goto drop;
169                 }
170                 hdr = ipv6_hdr(skb);
171         }
172
173         if (hdr->nexthdr == NEXTHDR_HOP) {
174                 if (ipv6_parse_hopopts(skb) < 0) {
175                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
176                         rcu_read_unlock();
177                         return NET_RX_DROP;
178                 }
179         }
180
181         rcu_read_unlock();
182
183         /* Must drop socket now because of tproxy. */
184         skb_orphan(skb);
185
186         return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
187                        net, NULL, skb, dev, NULL,
188                        ip6_rcv_finish);
189 err:
190         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
191 drop:
192         rcu_read_unlock();
193         kfree_skb(skb);
194         return NET_RX_DROP;
195 }
196
197 /*
198  *      Deliver the packet to the host
199  */
200
201
202 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
203 {
204         const struct inet6_protocol *ipprot;
205         struct inet6_dev *idev;
206         unsigned int nhoff;
207         int nexthdr;
208         bool raw;
209
210         /*
211          *      Parse extension headers
212          */
213
214         rcu_read_lock();
215 resubmit:
216         idev = ip6_dst_idev(skb_dst(skb));
217         if (!pskb_pull(skb, skb_transport_offset(skb)))
218                 goto discard;
219         nhoff = IP6CB(skb)->nhoff;
220         nexthdr = skb_network_header(skb)[nhoff];
221
222         raw = raw6_local_deliver(skb, nexthdr);
223         ipprot = rcu_dereference(inet6_protos[nexthdr]);
224         if (ipprot) {
225                 int ret;
226
227                 if (ipprot->flags & INET6_PROTO_FINAL) {
228                         const struct ipv6hdr *hdr;
229
230                         /* Free reference early: we don't need it any more,
231                            and it may hold ip_conntrack module loaded
232                            indefinitely. */
233                         nf_reset(skb);
234
235                         skb_postpull_rcsum(skb, skb_network_header(skb),
236                                            skb_network_header_len(skb));
237                         hdr = ipv6_hdr(skb);
238                         if (ipv6_addr_is_multicast(&hdr->daddr) &&
239                             !ipv6_chk_mcast_addr(skb->dev, &hdr->daddr,
240                             &hdr->saddr) &&
241                             !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
242                                 goto discard;
243                 }
244                 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
245                     !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
246                         goto discard;
247
248                 ret = ipprot->handler(skb);
249                 if (ret > 0)
250                         goto resubmit;
251                 else if (ret == 0)
252                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDELIVERS);
253         } else {
254                 if (!raw) {
255                         if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
256                                 IP6_INC_STATS_BH(net, idev,
257                                                  IPSTATS_MIB_INUNKNOWNPROTOS);
258                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
259                                             ICMPV6_UNK_NEXTHDR, nhoff);
260                         }
261                         kfree_skb(skb);
262                 } else {
263                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDELIVERS);
264                         consume_skb(skb);
265                 }
266         }
267         rcu_read_unlock();
268         return 0;
269
270 discard:
271         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDISCARDS);
272         rcu_read_unlock();
273         kfree_skb(skb);
274         return 0;
275 }
276
277
278 int ip6_input(struct sk_buff *skb)
279 {
280         return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
281                        dev_net(skb->dev), NULL, skb, skb->dev, NULL,
282                        ip6_input_finish);
283 }
284
285 int ip6_mc_input(struct sk_buff *skb)
286 {
287         const struct ipv6hdr *hdr;
288         bool deliver;
289
290         IP6_UPD_PO_STATS_BH(dev_net(skb_dst(skb)->dev),
291                          ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INMCAST,
292                          skb->len);
293
294         hdr = ipv6_hdr(skb);
295         deliver = ipv6_chk_mcast_addr(skb->dev, &hdr->daddr, NULL);
296
297 #ifdef CONFIG_IPV6_MROUTE
298         /*
299          *      IPv6 multicast router mode is now supported ;)
300          */
301         if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
302             !(ipv6_addr_type(&hdr->daddr) &
303               (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
304             likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
305                 /*
306                  * Okay, we try to forward - split and duplicate
307                  * packets.
308                  */
309                 struct sk_buff *skb2;
310                 struct inet6_skb_parm *opt = IP6CB(skb);
311
312                 /* Check for MLD */
313                 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
314                         /* Check if this is a mld message */
315                         u8 nexthdr = hdr->nexthdr;
316                         __be16 frag_off;
317                         int offset;
318
319                         /* Check if the value of Router Alert
320                          * is for MLD (0x0000).
321                          */
322                         if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
323                                 deliver = false;
324
325                                 if (!ipv6_ext_hdr(nexthdr)) {
326                                         /* BUG */
327                                         goto out;
328                                 }
329                                 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
330                                                           &nexthdr, &frag_off);
331                                 if (offset < 0)
332                                         goto out;
333
334                                 if (ipv6_is_mld(skb, nexthdr, offset))
335                                         deliver = true;
336
337                                 goto out;
338                         }
339                         /* unknown RA - process it normally */
340                 }
341
342                 if (deliver)
343                         skb2 = skb_clone(skb, GFP_ATOMIC);
344                 else {
345                         skb2 = skb;
346                         skb = NULL;
347                 }
348
349                 if (skb2) {
350                         ip6_mr_input(skb2);
351                 }
352         }
353 out:
354 #endif
355         if (likely(deliver))
356                 ip6_input(skb);
357         else {
358                 /* discard */
359                 kfree_skb(skb);
360         }
361
362         return 0;
363 }