Merge branch 'master' of /repos/git/net-next-2.6
[pandora-kernel.git] / net / ipv6 / raw.c
1 /*
2  *      RAW sockets for IPv6
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      Adapted from linux/net/ipv4/raw.c
9  *
10  *      Fixes:
11  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
12  *      YOSHIFUJI,H.@USAGI      :       raw checksum (RFC2292(bis) compliance)
13  *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
14  *
15  *      This program is free software; you can redistribute it and/or
16  *      modify it under the terms of the GNU General Public License
17  *      as published by the Free Software Foundation; either version
18  *      2 of the License, or (at your option) any later version.
19  */
20
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <asm/uaccess.h>
35 #include <asm/ioctls.h>
36
37 #include <net/net_namespace.h>
38 #include <net/ip.h>
39 #include <net/sock.h>
40 #include <net/snmp.h>
41
42 #include <net/ipv6.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/ip6_route.h>
46 #include <net/ip6_checksum.h>
47 #include <net/addrconf.h>
48 #include <net/transp_v6.h>
49 #include <net/udp.h>
50 #include <net/inet_common.h>
51 #include <net/tcp_states.h>
52 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
53 #include <net/mip6.h>
54 #endif
55 #include <linux/mroute6.h>
56
57 #include <net/raw.h>
58 #include <net/rawv6.h>
59 #include <net/xfrm.h>
60
61 #include <linux/proc_fs.h>
62 #include <linux/seq_file.h>
63
64 static struct raw_hashinfo raw_v6_hashinfo = {
65         .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
66 };
67
68 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
69                 unsigned short num, struct in6_addr *loc_addr,
70                 struct in6_addr *rmt_addr, int dif)
71 {
72         struct hlist_node *node;
73         int is_multicast = ipv6_addr_is_multicast(loc_addr);
74
75         sk_for_each_from(sk, node)
76                 if (inet_sk(sk)->inet_num == num) {
77                         struct ipv6_pinfo *np = inet6_sk(sk);
78
79                         if (!net_eq(sock_net(sk), net))
80                                 continue;
81
82                         if (!ipv6_addr_any(&np->daddr) &&
83                             !ipv6_addr_equal(&np->daddr, rmt_addr))
84                                 continue;
85
86                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
87                                 continue;
88
89                         if (!ipv6_addr_any(&np->rcv_saddr)) {
90                                 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
91                                         goto found;
92                                 if (is_multicast &&
93                                     inet6_mc_check(sk, loc_addr, rmt_addr))
94                                         goto found;
95                                 continue;
96                         }
97                         goto found;
98                 }
99         sk = NULL;
100 found:
101         return sk;
102 }
103
104 /*
105  *      0 - deliver
106  *      1 - block
107  */
108 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
109 {
110         struct icmp6hdr *icmph;
111         struct raw6_sock *rp = raw6_sk(sk);
112
113         if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
114                 __u32 *data = &rp->filter.data[0];
115                 int bit_nr;
116
117                 icmph = (struct icmp6hdr *) skb->data;
118                 bit_nr = icmph->icmp6_type;
119
120                 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
121         }
122         return 0;
123 }
124
125 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
126 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
127
128 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
129                                            struct sk_buff *skb))
130 {
131         rcu_assign_pointer(mh_filter, filter);
132         return 0;
133 }
134 EXPORT_SYMBOL(rawv6_mh_filter_register);
135
136 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
137                                              struct sk_buff *skb))
138 {
139         rcu_assign_pointer(mh_filter, NULL);
140         synchronize_rcu();
141         return 0;
142 }
143 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
144
145 #endif
146
147 /*
148  *      demultiplex raw sockets.
149  *      (should consider queueing the skb in the sock receive_queue
150  *      without calling rawv6.c)
151  *
152  *      Caller owns SKB so we must make clones.
153  */
154 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
155 {
156         struct in6_addr *saddr;
157         struct in6_addr *daddr;
158         struct sock *sk;
159         int delivered = 0;
160         __u8 hash;
161         struct net *net;
162
163         saddr = &ipv6_hdr(skb)->saddr;
164         daddr = saddr + 1;
165
166         hash = nexthdr & (MAX_INET_PROTOS - 1);
167
168         read_lock(&raw_v6_hashinfo.lock);
169         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
170
171         if (sk == NULL)
172                 goto out;
173
174         net = dev_net(skb->dev);
175         sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
176
177         while (sk) {
178                 int filtered;
179
180                 delivered = 1;
181                 switch (nexthdr) {
182                 case IPPROTO_ICMPV6:
183                         filtered = icmpv6_filter(sk, skb);
184                         break;
185
186 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
187                 case IPPROTO_MH:
188                 {
189                         /* XXX: To validate MH only once for each packet,
190                          * this is placed here. It should be after checking
191                          * xfrm policy, however it doesn't. The checking xfrm
192                          * policy is placed in rawv6_rcv() because it is
193                          * required for each socket.
194                          */
195                         int (*filter)(struct sock *sock, struct sk_buff *skb);
196
197                         filter = rcu_dereference(mh_filter);
198                         filtered = filter ? filter(sk, skb) : 0;
199                         break;
200                 }
201 #endif
202                 default:
203                         filtered = 0;
204                         break;
205                 }
206
207                 if (filtered < 0)
208                         break;
209                 if (filtered == 0) {
210                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
211
212                         /* Not releasing hash table! */
213                         if (clone) {
214                                 nf_reset(clone);
215                                 rawv6_rcv(sk, clone);
216                         }
217                 }
218                 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
219                                      IP6CB(skb)->iif);
220         }
221 out:
222         read_unlock(&raw_v6_hashinfo.lock);
223         return delivered;
224 }
225
226 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
227 {
228         struct sock *raw_sk;
229
230         raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
231         if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
232                 raw_sk = NULL;
233
234         return raw_sk != NULL;
235 }
236
237 /* This cleans up af_inet6 a bit. -DaveM */
238 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
239 {
240         struct inet_sock *inet = inet_sk(sk);
241         struct ipv6_pinfo *np = inet6_sk(sk);
242         struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
243         __be32 v4addr = 0;
244         int addr_type;
245         int err;
246
247         if (addr_len < SIN6_LEN_RFC2133)
248                 return -EINVAL;
249         addr_type = ipv6_addr_type(&addr->sin6_addr);
250
251         /* Raw sockets are IPv6 only */
252         if (addr_type == IPV6_ADDR_MAPPED)
253                 return -EADDRNOTAVAIL;
254
255         lock_sock(sk);
256
257         err = -EINVAL;
258         if (sk->sk_state != TCP_CLOSE)
259                 goto out;
260
261         rcu_read_lock();
262         /* Check if the address belongs to the host. */
263         if (addr_type != IPV6_ADDR_ANY) {
264                 struct net_device *dev = NULL;
265
266                 if (addr_type & IPV6_ADDR_LINKLOCAL) {
267                         if (addr_len >= sizeof(struct sockaddr_in6) &&
268                             addr->sin6_scope_id) {
269                                 /* Override any existing binding, if another
270                                  * one is supplied by user.
271                                  */
272                                 sk->sk_bound_dev_if = addr->sin6_scope_id;
273                         }
274
275                         /* Binding to link-local address requires an interface */
276                         if (!sk->sk_bound_dev_if)
277                                 goto out_unlock;
278
279                         err = -ENODEV;
280                         dev = dev_get_by_index_rcu(sock_net(sk),
281                                                    sk->sk_bound_dev_if);
282                         if (!dev)
283                                 goto out_unlock;
284                 }
285
286                 /* ipv4 addr of the socket is invalid.  Only the
287                  * unspecified and mapped address have a v4 equivalent.
288                  */
289                 v4addr = LOOPBACK4_IPV6;
290                 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
291                         err = -EADDRNOTAVAIL;
292                         if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
293                                            dev, 0)) {
294                                 goto out_unlock;
295                         }
296                 }
297         }
298
299         inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
300         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
301         if (!(addr_type & IPV6_ADDR_MULTICAST))
302                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
303         err = 0;
304 out_unlock:
305         rcu_read_unlock();
306 out:
307         release_sock(sk);
308         return err;
309 }
310
311 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
312                struct inet6_skb_parm *opt,
313                u8 type, u8 code, int offset, __be32 info)
314 {
315         struct inet_sock *inet = inet_sk(sk);
316         struct ipv6_pinfo *np = inet6_sk(sk);
317         int err;
318         int harderr;
319
320         /* Report error on raw socket, if:
321            1. User requested recverr.
322            2. Socket is connected (otherwise the error indication
323               is useless without recverr and error is hard.
324          */
325         if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
326                 return;
327
328         harderr = icmpv6_err_convert(type, code, &err);
329         if (type == ICMPV6_PKT_TOOBIG)
330                 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
331
332         if (np->recverr) {
333                 u8 *payload = skb->data;
334                 if (!inet->hdrincl)
335                         payload += offset;
336                 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
337         }
338
339         if (np->recverr || harderr) {
340                 sk->sk_err = err;
341                 sk->sk_error_report(sk);
342         }
343 }
344
345 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
346                 u8 type, u8 code, int inner_offset, __be32 info)
347 {
348         struct sock *sk;
349         int hash;
350         struct in6_addr *saddr, *daddr;
351         struct net *net;
352
353         hash = nexthdr & (RAW_HTABLE_SIZE - 1);
354
355         read_lock(&raw_v6_hashinfo.lock);
356         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
357         if (sk != NULL) {
358                 /* Note: ipv6_hdr(skb) != skb->data */
359                 struct ipv6hdr *ip6h = (struct ipv6hdr *)skb->data;
360                 saddr = &ip6h->saddr;
361                 daddr = &ip6h->daddr;
362                 net = dev_net(skb->dev);
363
364                 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
365                                                 IP6CB(skb)->iif))) {
366                         rawv6_err(sk, skb, NULL, type, code,
367                                         inner_offset, info);
368                         sk = sk_next(sk);
369                 }
370         }
371         read_unlock(&raw_v6_hashinfo.lock);
372 }
373
374 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
375 {
376         if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
377             skb_checksum_complete(skb)) {
378                 atomic_inc(&sk->sk_drops);
379                 kfree_skb(skb);
380                 return NET_RX_DROP;
381         }
382
383         /* Charge it to the socket. */
384         if (sock_queue_rcv_skb(sk, skb) < 0) {
385                 kfree_skb(skb);
386                 return NET_RX_DROP;
387         }
388
389         return 0;
390 }
391
392 /*
393  *      This is next to useless...
394  *      if we demultiplex in network layer we don't need the extra call
395  *      just to queue the skb...
396  *      maybe we could have the network decide upon a hint if it
397  *      should call raw_rcv for demultiplexing
398  */
399 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
400 {
401         struct inet_sock *inet = inet_sk(sk);
402         struct raw6_sock *rp = raw6_sk(sk);
403
404         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
405                 atomic_inc(&sk->sk_drops);
406                 kfree_skb(skb);
407                 return NET_RX_DROP;
408         }
409
410         if (!rp->checksum)
411                 skb->ip_summed = CHECKSUM_UNNECESSARY;
412
413         if (skb->ip_summed == CHECKSUM_COMPLETE) {
414                 skb_postpull_rcsum(skb, skb_network_header(skb),
415                                    skb_network_header_len(skb));
416                 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
417                                      &ipv6_hdr(skb)->daddr,
418                                      skb->len, inet->inet_num, skb->csum))
419                         skb->ip_summed = CHECKSUM_UNNECESSARY;
420         }
421         if (!skb_csum_unnecessary(skb))
422                 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
423                                                          &ipv6_hdr(skb)->daddr,
424                                                          skb->len,
425                                                          inet->inet_num, 0));
426
427         if (inet->hdrincl) {
428                 if (skb_checksum_complete(skb)) {
429                         atomic_inc(&sk->sk_drops);
430                         kfree_skb(skb);
431                         return NET_RX_DROP;
432                 }
433         }
434
435         rawv6_rcv_skb(sk, skb);
436         return 0;
437 }
438
439
440 /*
441  *      This should be easy, if there is something there
442  *      we return it, otherwise we block.
443  */
444
445 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
446                   struct msghdr *msg, size_t len,
447                   int noblock, int flags, int *addr_len)
448 {
449         struct ipv6_pinfo *np = inet6_sk(sk);
450         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
451         struct sk_buff *skb;
452         size_t copied;
453         int err;
454
455         if (flags & MSG_OOB)
456                 return -EOPNOTSUPP;
457
458         if (addr_len)
459                 *addr_len=sizeof(*sin6);
460
461         if (flags & MSG_ERRQUEUE)
462                 return ipv6_recv_error(sk, msg, len);
463
464         skb = skb_recv_datagram(sk, flags, noblock, &err);
465         if (!skb)
466                 goto out;
467
468         copied = skb->len;
469         if (copied > len) {
470                 copied = len;
471                 msg->msg_flags |= MSG_TRUNC;
472         }
473
474         if (skb_csum_unnecessary(skb)) {
475                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
476         } else if (msg->msg_flags&MSG_TRUNC) {
477                 if (__skb_checksum_complete(skb))
478                         goto csum_copy_err;
479                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
480         } else {
481                 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
482                 if (err == -EINVAL)
483                         goto csum_copy_err;
484         }
485         if (err)
486                 goto out_free;
487
488         /* Copy the address. */
489         if (sin6) {
490                 sin6->sin6_family = AF_INET6;
491                 sin6->sin6_port = 0;
492                 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
493                 sin6->sin6_flowinfo = 0;
494                 sin6->sin6_scope_id = 0;
495                 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
496                         sin6->sin6_scope_id = IP6CB(skb)->iif;
497         }
498
499         sock_recv_ts_and_drops(msg, sk, skb);
500
501         if (np->rxopt.all)
502                 datagram_recv_ctl(sk, msg, skb);
503
504         err = copied;
505         if (flags & MSG_TRUNC)
506                 err = skb->len;
507
508 out_free:
509         skb_free_datagram(sk, skb);
510 out:
511         return err;
512
513 csum_copy_err:
514         skb_kill_datagram(sk, skb, flags);
515
516         /* Error for blocking case is chosen to masquerade
517            as some normal condition.
518          */
519         err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
520         goto out;
521 }
522
523 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
524                                      struct raw6_sock *rp)
525 {
526         struct sk_buff *skb;
527         int err = 0;
528         int offset;
529         int len;
530         int total_len;
531         __wsum tmp_csum;
532         __sum16 csum;
533
534         if (!rp->checksum)
535                 goto send;
536
537         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
538                 goto out;
539
540         offset = rp->offset;
541         total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
542                                                 skb->data);
543         if (offset >= total_len - 1) {
544                 err = -EINVAL;
545                 ip6_flush_pending_frames(sk);
546                 goto out;
547         }
548
549         /* should be check HW csum miyazawa */
550         if (skb_queue_len(&sk->sk_write_queue) == 1) {
551                 /*
552                  * Only one fragment on the socket.
553                  */
554                 tmp_csum = skb->csum;
555         } else {
556                 struct sk_buff *csum_skb = NULL;
557                 tmp_csum = 0;
558
559                 skb_queue_walk(&sk->sk_write_queue, skb) {
560                         tmp_csum = csum_add(tmp_csum, skb->csum);
561
562                         if (csum_skb)
563                                 continue;
564
565                         len = skb->len - skb_transport_offset(skb);
566                         if (offset >= len) {
567                                 offset -= len;
568                                 continue;
569                         }
570
571                         csum_skb = skb;
572                 }
573
574                 skb = csum_skb;
575         }
576
577         offset += skb_transport_offset(skb);
578         if (skb_copy_bits(skb, offset, &csum, 2))
579                 BUG();
580
581         /* in case cksum was not initialized */
582         if (unlikely(csum))
583                 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
584
585         csum = csum_ipv6_magic(&fl->fl6_src,
586                                    &fl->fl6_dst,
587                                    total_len, fl->proto, tmp_csum);
588
589         if (csum == 0 && fl->proto == IPPROTO_UDP)
590                 csum = CSUM_MANGLED_0;
591
592         if (skb_store_bits(skb, offset, &csum, 2))
593                 BUG();
594
595 send:
596         err = ip6_push_pending_frames(sk);
597 out:
598         return err;
599 }
600
601 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
602                         struct flowi *fl, struct rt6_info *rt,
603                         unsigned int flags)
604 {
605         struct ipv6_pinfo *np = inet6_sk(sk);
606         struct ipv6hdr *iph;
607         struct sk_buff *skb;
608         int err;
609
610         if (length > rt->u.dst.dev->mtu) {
611                 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
612                 return -EMSGSIZE;
613         }
614         if (flags&MSG_PROBE)
615                 goto out;
616
617         skb = sock_alloc_send_skb(sk,
618                                   length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
619                                   flags & MSG_DONTWAIT, &err);
620         if (skb == NULL)
621                 goto error;
622         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
623
624         skb->priority = sk->sk_priority;
625         skb->mark = sk->sk_mark;
626         skb_dst_set(skb, dst_clone(&rt->u.dst));
627
628         skb_put(skb, length);
629         skb_reset_network_header(skb);
630         iph = ipv6_hdr(skb);
631
632         skb->ip_summed = CHECKSUM_NONE;
633
634         skb->transport_header = skb->network_header;
635         err = memcpy_fromiovecend((void *)iph, from, 0, length);
636         if (err)
637                 goto error_fault;
638
639         IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
640         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
641                       rt->u.dst.dev, dst_output);
642         if (err > 0)
643                 err = net_xmit_errno(err);
644         if (err)
645                 goto error;
646 out:
647         return 0;
648
649 error_fault:
650         err = -EFAULT;
651         kfree_skb(skb);
652 error:
653         IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
654         if (err == -ENOBUFS && !np->recverr)
655                 err = 0;
656         return err;
657 }
658
659 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
660 {
661         struct iovec *iov;
662         u8 __user *type = NULL;
663         u8 __user *code = NULL;
664         u8 len = 0;
665         int probed = 0;
666         int i;
667
668         if (!msg->msg_iov)
669                 return 0;
670
671         for (i = 0; i < msg->msg_iovlen; i++) {
672                 iov = &msg->msg_iov[i];
673                 if (!iov)
674                         continue;
675
676                 switch (fl->proto) {
677                 case IPPROTO_ICMPV6:
678                         /* check if one-byte field is readable or not. */
679                         if (iov->iov_base && iov->iov_len < 1)
680                                 break;
681
682                         if (!type) {
683                                 type = iov->iov_base;
684                                 /* check if code field is readable or not. */
685                                 if (iov->iov_len > 1)
686                                         code = type + 1;
687                         } else if (!code)
688                                 code = iov->iov_base;
689
690                         if (type && code) {
691                                 if (get_user(fl->fl_icmp_type, type) ||
692                                     get_user(fl->fl_icmp_code, code))
693                                         return -EFAULT;
694                                 probed = 1;
695                         }
696                         break;
697                 case IPPROTO_MH:
698                         if (iov->iov_base && iov->iov_len < 1)
699                                 break;
700                         /* check if type field is readable or not. */
701                         if (iov->iov_len > 2 - len) {
702                                 u8 __user *p = iov->iov_base;
703                                 if (get_user(fl->fl_mh_type, &p[2 - len]))
704                                         return -EFAULT;
705                                 probed = 1;
706                         } else
707                                 len += iov->iov_len;
708
709                         break;
710                 default:
711                         probed = 1;
712                         break;
713                 }
714                 if (probed)
715                         break;
716         }
717         return 0;
718 }
719
720 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
721                    struct msghdr *msg, size_t len)
722 {
723         struct ipv6_txoptions opt_space;
724         struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
725         struct in6_addr *daddr, *final_p = NULL, final;
726         struct inet_sock *inet = inet_sk(sk);
727         struct ipv6_pinfo *np = inet6_sk(sk);
728         struct raw6_sock *rp = raw6_sk(sk);
729         struct ipv6_txoptions *opt = NULL;
730         struct ip6_flowlabel *flowlabel = NULL;
731         struct dst_entry *dst = NULL;
732         struct flowi fl;
733         int addr_len = msg->msg_namelen;
734         int hlimit = -1;
735         int tclass = -1;
736         u16 proto;
737         int err;
738
739         /* Rough check on arithmetic overflow,
740            better check is made in ip6_append_data().
741          */
742         if (len > INT_MAX)
743                 return -EMSGSIZE;
744
745         /* Mirror BSD error message compatibility */
746         if (msg->msg_flags & MSG_OOB)
747                 return -EOPNOTSUPP;
748
749         /*
750          *      Get and verify the address.
751          */
752         memset(&fl, 0, sizeof(fl));
753
754         fl.mark = sk->sk_mark;
755
756         if (sin6) {
757                 if (addr_len < SIN6_LEN_RFC2133)
758                         return -EINVAL;
759
760                 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
761                         return(-EAFNOSUPPORT);
762
763                 /* port is the proto value [0..255] carried in nexthdr */
764                 proto = ntohs(sin6->sin6_port);
765
766                 if (!proto)
767                         proto = inet->inet_num;
768                 else if (proto != inet->inet_num)
769                         return(-EINVAL);
770
771                 if (proto > 255)
772                         return(-EINVAL);
773
774                 daddr = &sin6->sin6_addr;
775                 if (np->sndflow) {
776                         fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
777                         if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
778                                 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
779                                 if (flowlabel == NULL)
780                                         return -EINVAL;
781                                 daddr = &flowlabel->dst;
782                         }
783                 }
784
785                 /*
786                  * Otherwise it will be difficult to maintain
787                  * sk->sk_dst_cache.
788                  */
789                 if (sk->sk_state == TCP_ESTABLISHED &&
790                     ipv6_addr_equal(daddr, &np->daddr))
791                         daddr = &np->daddr;
792
793                 if (addr_len >= sizeof(struct sockaddr_in6) &&
794                     sin6->sin6_scope_id &&
795                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
796                         fl.oif = sin6->sin6_scope_id;
797         } else {
798                 if (sk->sk_state != TCP_ESTABLISHED)
799                         return -EDESTADDRREQ;
800
801                 proto = inet->inet_num;
802                 daddr = &np->daddr;
803                 fl.fl6_flowlabel = np->flow_label;
804         }
805
806         if (fl.oif == 0)
807                 fl.oif = sk->sk_bound_dev_if;
808
809         if (msg->msg_controllen) {
810                 opt = &opt_space;
811                 memset(opt, 0, sizeof(struct ipv6_txoptions));
812                 opt->tot_len = sizeof(struct ipv6_txoptions);
813
814                 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, &tclass);
815                 if (err < 0) {
816                         fl6_sock_release(flowlabel);
817                         return err;
818                 }
819                 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
820                         flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
821                         if (flowlabel == NULL)
822                                 return -EINVAL;
823                 }
824                 if (!(opt->opt_nflen|opt->opt_flen))
825                         opt = NULL;
826         }
827         if (opt == NULL)
828                 opt = np->opt;
829         if (flowlabel)
830                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
831         opt = ipv6_fixup_options(&opt_space, opt);
832
833         fl.proto = proto;
834         err = rawv6_probe_proto_opt(&fl, msg);
835         if (err)
836                 goto out;
837
838         if (!ipv6_addr_any(daddr))
839                 ipv6_addr_copy(&fl.fl6_dst, daddr);
840         else
841                 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
842         if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
843                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
844
845         /* merge ip6_build_xmit from ip6_output */
846         if (opt && opt->srcrt) {
847                 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
848                 ipv6_addr_copy(&final, &fl.fl6_dst);
849                 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
850                 final_p = &final;
851         }
852
853         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
854                 fl.oif = np->mcast_oif;
855         security_sk_classify_flow(sk, &fl);
856
857         err = ip6_dst_lookup(sk, &dst, &fl);
858         if (err)
859                 goto out;
860         if (final_p)
861                 ipv6_addr_copy(&fl.fl6_dst, final_p);
862
863         err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
864         if (err < 0) {
865                 if (err == -EREMOTE)
866                         err = ip6_dst_blackhole(sk, &dst, &fl);
867                 if (err < 0)
868                         goto out;
869         }
870
871         if (hlimit < 0) {
872                 if (ipv6_addr_is_multicast(&fl.fl6_dst))
873                         hlimit = np->mcast_hops;
874                 else
875                         hlimit = np->hop_limit;
876                 if (hlimit < 0)
877                         hlimit = ip6_dst_hoplimit(dst);
878         }
879
880         if (tclass < 0)
881                 tclass = np->tclass;
882
883         if (msg->msg_flags&MSG_CONFIRM)
884                 goto do_confirm;
885
886 back_from_confirm:
887         if (inet->hdrincl) {
888                 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
889         } else {
890                 lock_sock(sk);
891                 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
892                         len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
893                         msg->msg_flags);
894
895                 if (err)
896                         ip6_flush_pending_frames(sk);
897                 else if (!(msg->msg_flags & MSG_MORE))
898                         err = rawv6_push_pending_frames(sk, &fl, rp);
899                 release_sock(sk);
900         }
901 done:
902         dst_release(dst);
903 out:
904         fl6_sock_release(flowlabel);
905         return err<0?err:len;
906 do_confirm:
907         dst_confirm(dst);
908         if (!(msg->msg_flags & MSG_PROBE) || len)
909                 goto back_from_confirm;
910         err = 0;
911         goto done;
912 }
913
914 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
915                                char __user *optval, int optlen)
916 {
917         switch (optname) {
918         case ICMPV6_FILTER:
919                 if (optlen > sizeof(struct icmp6_filter))
920                         optlen = sizeof(struct icmp6_filter);
921                 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
922                         return -EFAULT;
923                 return 0;
924         default:
925                 return -ENOPROTOOPT;
926         }
927
928         return 0;
929 }
930
931 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
932                                char __user *optval, int __user *optlen)
933 {
934         int len;
935
936         switch (optname) {
937         case ICMPV6_FILTER:
938                 if (get_user(len, optlen))
939                         return -EFAULT;
940                 if (len < 0)
941                         return -EINVAL;
942                 if (len > sizeof(struct icmp6_filter))
943                         len = sizeof(struct icmp6_filter);
944                 if (put_user(len, optlen))
945                         return -EFAULT;
946                 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
947                         return -EFAULT;
948                 return 0;
949         default:
950                 return -ENOPROTOOPT;
951         }
952
953         return 0;
954 }
955
956
957 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
958                             char __user *optval, unsigned int optlen)
959 {
960         struct raw6_sock *rp = raw6_sk(sk);
961         int val;
962
963         if (get_user(val, (int __user *)optval))
964                 return -EFAULT;
965
966         switch (optname) {
967                 case IPV6_CHECKSUM:
968                         if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
969                             level == IPPROTO_IPV6) {
970                                 /*
971                                  * RFC3542 tells that IPV6_CHECKSUM socket
972                                  * option in the IPPROTO_IPV6 level is not
973                                  * allowed on ICMPv6 sockets.
974                                  * If you want to set it, use IPPROTO_RAW
975                                  * level IPV6_CHECKSUM socket option
976                                  * (Linux extension).
977                                  */
978                                 return -EINVAL;
979                         }
980
981                         /* You may get strange result with a positive odd offset;
982                            RFC2292bis agrees with me. */
983                         if (val > 0 && (val&1))
984                                 return(-EINVAL);
985                         if (val < 0) {
986                                 rp->checksum = 0;
987                         } else {
988                                 rp->checksum = 1;
989                                 rp->offset = val;
990                         }
991
992                         return 0;
993                         break;
994
995                 default:
996                         return(-ENOPROTOOPT);
997         }
998 }
999
1000 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1001                           char __user *optval, unsigned int optlen)
1002 {
1003         switch(level) {
1004                 case SOL_RAW:
1005                         break;
1006
1007                 case SOL_ICMPV6:
1008                         if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1009                                 return -EOPNOTSUPP;
1010                         return rawv6_seticmpfilter(sk, level, optname, optval,
1011                                                    optlen);
1012                 case SOL_IPV6:
1013                         if (optname == IPV6_CHECKSUM)
1014                                 break;
1015                 default:
1016                         return ipv6_setsockopt(sk, level, optname, optval,
1017                                                optlen);
1018         }
1019
1020         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1021 }
1022
1023 #ifdef CONFIG_COMPAT
1024 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1025                                    char __user *optval, unsigned int optlen)
1026 {
1027         switch (level) {
1028         case SOL_RAW:
1029                 break;
1030         case SOL_ICMPV6:
1031                 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1032                         return -EOPNOTSUPP;
1033                 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1034         case SOL_IPV6:
1035                 if (optname == IPV6_CHECKSUM)
1036                         break;
1037         default:
1038                 return compat_ipv6_setsockopt(sk, level, optname,
1039                                               optval, optlen);
1040         }
1041         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1042 }
1043 #endif
1044
1045 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1046                             char __user *optval, int __user *optlen)
1047 {
1048         struct raw6_sock *rp = raw6_sk(sk);
1049         int val, len;
1050
1051         if (get_user(len,optlen))
1052                 return -EFAULT;
1053
1054         switch (optname) {
1055         case IPV6_CHECKSUM:
1056                 /*
1057                  * We allow getsockopt() for IPPROTO_IPV6-level
1058                  * IPV6_CHECKSUM socket option on ICMPv6 sockets
1059                  * since RFC3542 is silent about it.
1060                  */
1061                 if (rp->checksum == 0)
1062                         val = -1;
1063                 else
1064                         val = rp->offset;
1065                 break;
1066
1067         default:
1068                 return -ENOPROTOOPT;
1069         }
1070
1071         len = min_t(unsigned int, sizeof(int), len);
1072
1073         if (put_user(len, optlen))
1074                 return -EFAULT;
1075         if (copy_to_user(optval,&val,len))
1076                 return -EFAULT;
1077         return 0;
1078 }
1079
1080 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1081                           char __user *optval, int __user *optlen)
1082 {
1083         switch(level) {
1084                 case SOL_RAW:
1085                         break;
1086
1087                 case SOL_ICMPV6:
1088                         if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1089                                 return -EOPNOTSUPP;
1090                         return rawv6_geticmpfilter(sk, level, optname, optval,
1091                                                    optlen);
1092                 case SOL_IPV6:
1093                         if (optname == IPV6_CHECKSUM)
1094                                 break;
1095                 default:
1096                         return ipv6_getsockopt(sk, level, optname, optval,
1097                                                optlen);
1098         }
1099
1100         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1101 }
1102
1103 #ifdef CONFIG_COMPAT
1104 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1105                                    char __user *optval, int __user *optlen)
1106 {
1107         switch (level) {
1108         case SOL_RAW:
1109                 break;
1110         case SOL_ICMPV6:
1111                 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1112                         return -EOPNOTSUPP;
1113                 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1114         case SOL_IPV6:
1115                 if (optname == IPV6_CHECKSUM)
1116                         break;
1117         default:
1118                 return compat_ipv6_getsockopt(sk, level, optname,
1119                                               optval, optlen);
1120         }
1121         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1122 }
1123 #endif
1124
1125 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1126 {
1127         switch(cmd) {
1128                 case SIOCOUTQ:
1129                 {
1130                         int amount = sk_wmem_alloc_get(sk);
1131
1132                         return put_user(amount, (int __user *)arg);
1133                 }
1134                 case SIOCINQ:
1135                 {
1136                         struct sk_buff *skb;
1137                         int amount = 0;
1138
1139                         spin_lock_bh(&sk->sk_receive_queue.lock);
1140                         skb = skb_peek(&sk->sk_receive_queue);
1141                         if (skb != NULL)
1142                                 amount = skb->tail - skb->transport_header;
1143                         spin_unlock_bh(&sk->sk_receive_queue.lock);
1144                         return put_user(amount, (int __user *)arg);
1145                 }
1146
1147                 default:
1148 #ifdef CONFIG_IPV6_MROUTE
1149                         return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1150 #else
1151                         return -ENOIOCTLCMD;
1152 #endif
1153         }
1154 }
1155
1156 static void rawv6_close(struct sock *sk, long timeout)
1157 {
1158         if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1159                 ip6_ra_control(sk, -1);
1160         ip6mr_sk_done(sk);
1161         sk_common_release(sk);
1162 }
1163
1164 static void raw6_destroy(struct sock *sk)
1165 {
1166         lock_sock(sk);
1167         ip6_flush_pending_frames(sk);
1168         release_sock(sk);
1169
1170         inet6_destroy_sock(sk);
1171 }
1172
1173 static int rawv6_init_sk(struct sock *sk)
1174 {
1175         struct raw6_sock *rp = raw6_sk(sk);
1176
1177         switch (inet_sk(sk)->inet_num) {
1178         case IPPROTO_ICMPV6:
1179                 rp->checksum = 1;
1180                 rp->offset   = 2;
1181                 break;
1182         case IPPROTO_MH:
1183                 rp->checksum = 1;
1184                 rp->offset   = 4;
1185                 break;
1186         default:
1187                 break;
1188         }
1189         return(0);
1190 }
1191
1192 struct proto rawv6_prot = {
1193         .name              = "RAWv6",
1194         .owner             = THIS_MODULE,
1195         .close             = rawv6_close,
1196         .destroy           = raw6_destroy,
1197         .connect           = ip6_datagram_connect,
1198         .disconnect        = udp_disconnect,
1199         .ioctl             = rawv6_ioctl,
1200         .init              = rawv6_init_sk,
1201         .setsockopt        = rawv6_setsockopt,
1202         .getsockopt        = rawv6_getsockopt,
1203         .sendmsg           = rawv6_sendmsg,
1204         .recvmsg           = rawv6_recvmsg,
1205         .bind              = rawv6_bind,
1206         .backlog_rcv       = rawv6_rcv_skb,
1207         .hash              = raw_hash_sk,
1208         .unhash            = raw_unhash_sk,
1209         .obj_size          = sizeof(struct raw6_sock),
1210         .h.raw_hash        = &raw_v6_hashinfo,
1211 #ifdef CONFIG_COMPAT
1212         .compat_setsockopt = compat_rawv6_setsockopt,
1213         .compat_getsockopt = compat_rawv6_getsockopt,
1214 #endif
1215 };
1216
1217 #ifdef CONFIG_PROC_FS
1218 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1219 {
1220         struct ipv6_pinfo *np = inet6_sk(sp);
1221         struct in6_addr *dest, *src;
1222         __u16 destp, srcp;
1223
1224         dest  = &np->daddr;
1225         src   = &np->rcv_saddr;
1226         destp = 0;
1227         srcp  = inet_sk(sp)->inet_num;
1228         seq_printf(seq,
1229                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1230                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1231                    i,
1232                    src->s6_addr32[0], src->s6_addr32[1],
1233                    src->s6_addr32[2], src->s6_addr32[3], srcp,
1234                    dest->s6_addr32[0], dest->s6_addr32[1],
1235                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
1236                    sp->sk_state,
1237                    sk_wmem_alloc_get(sp),
1238                    sk_rmem_alloc_get(sp),
1239                    0, 0L, 0,
1240                    sock_i_uid(sp), 0,
1241                    sock_i_ino(sp),
1242                    atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1243 }
1244
1245 static int raw6_seq_show(struct seq_file *seq, void *v)
1246 {
1247         if (v == SEQ_START_TOKEN)
1248                 seq_printf(seq,
1249                            "  sl  "
1250                            "local_address                         "
1251                            "remote_address                        "
1252                            "st tx_queue rx_queue tr tm->when retrnsmt"
1253                            "   uid  timeout inode ref pointer drops\n");
1254         else
1255                 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1256         return 0;
1257 }
1258
1259 static const struct seq_operations raw6_seq_ops = {
1260         .start =        raw_seq_start,
1261         .next =         raw_seq_next,
1262         .stop =         raw_seq_stop,
1263         .show =         raw6_seq_show,
1264 };
1265
1266 static int raw6_seq_open(struct inode *inode, struct file *file)
1267 {
1268         return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1269 }
1270
1271 static const struct file_operations raw6_seq_fops = {
1272         .owner =        THIS_MODULE,
1273         .open =         raw6_seq_open,
1274         .read =         seq_read,
1275         .llseek =       seq_lseek,
1276         .release =      seq_release_net,
1277 };
1278
1279 static int __net_init raw6_init_net(struct net *net)
1280 {
1281         if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1282                 return -ENOMEM;
1283
1284         return 0;
1285 }
1286
1287 static void __net_exit raw6_exit_net(struct net *net)
1288 {
1289         proc_net_remove(net, "raw6");
1290 }
1291
1292 static struct pernet_operations raw6_net_ops = {
1293         .init = raw6_init_net,
1294         .exit = raw6_exit_net,
1295 };
1296
1297 int __init raw6_proc_init(void)
1298 {
1299         return register_pernet_subsys(&raw6_net_ops);
1300 }
1301
1302 void raw6_proc_exit(void)
1303 {
1304         unregister_pernet_subsys(&raw6_net_ops);
1305 }
1306 #endif  /* CONFIG_PROC_FS */
1307
1308 /* Same as inet6_dgram_ops, sans udp_poll.  */
1309 static const struct proto_ops inet6_sockraw_ops = {
1310         .family            = PF_INET6,
1311         .owner             = THIS_MODULE,
1312         .release           = inet6_release,
1313         .bind              = inet6_bind,
1314         .connect           = inet_dgram_connect,        /* ok           */
1315         .socketpair        = sock_no_socketpair,        /* a do nothing */
1316         .accept            = sock_no_accept,            /* a do nothing */
1317         .getname           = inet6_getname,
1318         .poll              = datagram_poll,             /* ok           */
1319         .ioctl             = inet6_ioctl,               /* must change  */
1320         .listen            = sock_no_listen,            /* ok           */
1321         .shutdown          = inet_shutdown,             /* ok           */
1322         .setsockopt        = sock_common_setsockopt,    /* ok           */
1323         .getsockopt        = sock_common_getsockopt,    /* ok           */
1324         .sendmsg           = inet_sendmsg,              /* ok           */
1325         .recvmsg           = sock_common_recvmsg,       /* ok           */
1326         .mmap              = sock_no_mmap,
1327         .sendpage          = sock_no_sendpage,
1328 #ifdef CONFIG_COMPAT
1329         .compat_setsockopt = compat_sock_common_setsockopt,
1330         .compat_getsockopt = compat_sock_common_getsockopt,
1331 #endif
1332 };
1333
1334 static struct inet_protosw rawv6_protosw = {
1335         .type           = SOCK_RAW,
1336         .protocol       = IPPROTO_IP,   /* wild card */
1337         .prot           = &rawv6_prot,
1338         .ops            = &inet6_sockraw_ops,
1339         .no_check       = UDP_CSUM_DEFAULT,
1340         .flags          = INET_PROTOSW_REUSE,
1341 };
1342
1343 int __init rawv6_init(void)
1344 {
1345         int ret;
1346
1347         ret = inet6_register_protosw(&rawv6_protosw);
1348         if (ret)
1349                 goto out;
1350 out:
1351         return ret;
1352 }
1353
1354 void rawv6_exit(void)
1355 {
1356         inet6_unregister_protosw(&rawv6_protosw);
1357 }