[IPV6] MROUTE: Support multicast forwarding.
[linux-2.6.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  *      $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
11  *
12  *      Fixes:
13  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
14  *      YOSHIFUJI,H.@USAGI      :       raw checksum (RFC2292(bis) compliance)
15  *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
16  *
17  *      This program is free software; you can redistribute it and/or
18  *      modify it under the terms of the GNU General Public License
19  *      as published by the Free Software Foundation; either version
20  *      2 of the License, or (at your option) any later version.
21  */
22
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv6.h>
34 #include <linux/skbuff.h>
35 #include <asm/uaccess.h>
36 #include <asm/ioctls.h>
37
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
54 #include <net/mip6.h>
55 #endif
56 #include <linux/mroute6.h>
57
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
61
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64
65 static struct raw_hashinfo raw_v6_hashinfo = {
66         .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
67 };
68
69 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
70                 unsigned short num, struct in6_addr *loc_addr,
71                 struct in6_addr *rmt_addr, int dif)
72 {
73         struct hlist_node *node;
74         int is_multicast = ipv6_addr_is_multicast(loc_addr);
75
76         sk_for_each_from(sk, node)
77                 if (inet_sk(sk)->num == num) {
78                         struct ipv6_pinfo *np = inet6_sk(sk);
79
80                         if (!net_eq(sock_net(sk), net))
81                                 continue;
82
83                         if (!ipv6_addr_any(&np->daddr) &&
84                             !ipv6_addr_equal(&np->daddr, rmt_addr))
85                                 continue;
86
87                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
88                                 continue;
89
90                         if (!ipv6_addr_any(&np->rcv_saddr)) {
91                                 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
92                                         goto found;
93                                 if (is_multicast &&
94                                     inet6_mc_check(sk, loc_addr, rmt_addr))
95                                         goto found;
96                                 continue;
97                         }
98                         goto found;
99                 }
100         sk = NULL;
101 found:
102         return sk;
103 }
104
105 /*
106  *      0 - deliver
107  *      1 - block
108  */
109 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
110 {
111         struct icmp6hdr *icmph;
112         struct raw6_sock *rp = raw6_sk(sk);
113
114         if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
115                 __u32 *data = &rp->filter.data[0];
116                 int bit_nr;
117
118                 icmph = (struct icmp6hdr *) skb->data;
119                 bit_nr = icmph->icmp6_type;
120
121                 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
122         }
123         return 0;
124 }
125
126 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
127 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
128
129 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
130                                            struct sk_buff *skb))
131 {
132         rcu_assign_pointer(mh_filter, filter);
133         return 0;
134 }
135 EXPORT_SYMBOL(rawv6_mh_filter_register);
136
137 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
138                                              struct sk_buff *skb))
139 {
140         rcu_assign_pointer(mh_filter, NULL);
141         synchronize_rcu();
142         return 0;
143 }
144 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
145
146 #endif
147
148 /*
149  *      demultiplex raw sockets.
150  *      (should consider queueing the skb in the sock receive_queue
151  *      without calling rawv6.c)
152  *
153  *      Caller owns SKB so we must make clones.
154  */
155 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
156 {
157         struct in6_addr *saddr;
158         struct in6_addr *daddr;
159         struct sock *sk;
160         int delivered = 0;
161         __u8 hash;
162         struct net *net;
163
164         saddr = &ipv6_hdr(skb)->saddr;
165         daddr = saddr + 1;
166
167         hash = nexthdr & (MAX_INET_PROTOS - 1);
168
169         read_lock(&raw_v6_hashinfo.lock);
170         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
171
172         if (sk == NULL)
173                 goto out;
174
175         net = dev_net(skb->dev);
176         sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
177
178         while (sk) {
179                 int filtered;
180
181                 delivered = 1;
182                 switch (nexthdr) {
183                 case IPPROTO_ICMPV6:
184                         filtered = icmpv6_filter(sk, skb);
185                         break;
186
187 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
188                 case IPPROTO_MH:
189                 {
190                         /* XXX: To validate MH only once for each packet,
191                          * this is placed here. It should be after checking
192                          * xfrm policy, however it doesn't. The checking xfrm
193                          * policy is placed in rawv6_rcv() because it is
194                          * required for each socket.
195                          */
196                         int (*filter)(struct sock *sock, struct sk_buff *skb);
197
198                         filter = rcu_dereference(mh_filter);
199                         filtered = filter ? filter(sk, skb) : 0;
200                         break;
201                 }
202 #endif
203                 default:
204                         filtered = 0;
205                         break;
206                 }
207
208                 if (filtered < 0)
209                         break;
210                 if (filtered == 0) {
211                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
212
213                         /* Not releasing hash table! */
214                         if (clone) {
215                                 nf_reset(clone);
216                                 rawv6_rcv(sk, clone);
217                         }
218                 }
219                 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
220                                      IP6CB(skb)->iif);
221         }
222 out:
223         read_unlock(&raw_v6_hashinfo.lock);
224         return delivered;
225 }
226
227 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
228 {
229         struct sock *raw_sk;
230
231         raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
232         if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
233                 raw_sk = NULL;
234
235         return raw_sk != NULL;
236 }
237
238 /* This cleans up af_inet6 a bit. -DaveM */
239 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
240 {
241         struct inet_sock *inet = inet_sk(sk);
242         struct ipv6_pinfo *np = inet6_sk(sk);
243         struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
244         __be32 v4addr = 0;
245         int addr_type;
246         int err;
247
248         if (addr_len < SIN6_LEN_RFC2133)
249                 return -EINVAL;
250         addr_type = ipv6_addr_type(&addr->sin6_addr);
251
252         /* Raw sockets are IPv6 only */
253         if (addr_type == IPV6_ADDR_MAPPED)
254                 return(-EADDRNOTAVAIL);
255
256         lock_sock(sk);
257
258         err = -EINVAL;
259         if (sk->sk_state != TCP_CLOSE)
260                 goto out;
261
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;
278
279                         dev = dev_get_by_index(sock_net(sk), sk->sk_bound_dev_if);
280                         if (!dev) {
281                                 err = -ENODEV;
282                                 goto out;
283                         }
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                                 if (dev)
295                                         dev_put(dev);
296                                 goto out;
297                         }
298                 }
299                 if (dev)
300                         dev_put(dev);
301         }
302
303         inet->rcv_saddr = inet->saddr = v4addr;
304         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
305         if (!(addr_type & IPV6_ADDR_MULTICAST))
306                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
307         err = 0;
308 out:
309         release_sock(sk);
310         return err;
311 }
312
313 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
314                struct inet6_skb_parm *opt,
315                int type, int code, int offset, __be32 info)
316 {
317         struct inet_sock *inet = inet_sk(sk);
318         struct ipv6_pinfo *np = inet6_sk(sk);
319         int err;
320         int harderr;
321
322         /* Report error on raw socket, if:
323            1. User requested recverr.
324            2. Socket is connected (otherwise the error indication
325               is useless without recverr and error is hard.
326          */
327         if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
328                 return;
329
330         harderr = icmpv6_err_convert(type, code, &err);
331         if (type == ICMPV6_PKT_TOOBIG)
332                 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
333
334         if (np->recverr) {
335                 u8 *payload = skb->data;
336                 if (!inet->hdrincl)
337                         payload += offset;
338                 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
339         }
340
341         if (np->recverr || harderr) {
342                 sk->sk_err = err;
343                 sk->sk_error_report(sk);
344         }
345 }
346
347 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
348                 int type, int code, int inner_offset, __be32 info)
349 {
350         struct sock *sk;
351         int hash;
352         struct in6_addr *saddr, *daddr;
353         struct net *net;
354
355         hash = nexthdr & (RAW_HTABLE_SIZE - 1);
356
357         read_lock(&raw_v6_hashinfo.lock);
358         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
359         if (sk != NULL) {
360                 saddr = &ipv6_hdr(skb)->saddr;
361                 daddr = &ipv6_hdr(skb)->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 0;
381         }
382
383         /* Charge it to the socket. */
384         if (sock_queue_rcv_skb(sk,skb)<0) {
385                 atomic_inc(&sk->sk_drops);
386                 kfree_skb(skb);
387                 return 0;
388         }
389
390         return 0;
391 }
392
393 /*
394  *      This is next to useless...
395  *      if we demultiplex in network layer we don't need the extra call
396  *      just to queue the skb...
397  *      maybe we could have the network decide upon a hint if it
398  *      should call raw_rcv for demultiplexing
399  */
400 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
401 {
402         struct inet_sock *inet = inet_sk(sk);
403         struct raw6_sock *rp = raw6_sk(sk);
404
405         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
406                 atomic_inc(&sk->sk_drops);
407                 kfree_skb(skb);
408                 return NET_RX_DROP;
409         }
410
411         if (!rp->checksum)
412                 skb->ip_summed = CHECKSUM_UNNECESSARY;
413
414         if (skb->ip_summed == CHECKSUM_COMPLETE) {
415                 skb_postpull_rcsum(skb, skb_network_header(skb),
416                                    skb_network_header_len(skb));
417                 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
418                                      &ipv6_hdr(skb)->daddr,
419                                      skb->len, inet->num, skb->csum))
420                         skb->ip_summed = CHECKSUM_UNNECESSARY;
421         }
422         if (!skb_csum_unnecessary(skb))
423                 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
424                                                          &ipv6_hdr(skb)->daddr,
425                                                          skb->len,
426                                                          inet->num, 0));
427
428         if (inet->hdrincl) {
429                 if (skb_checksum_complete(skb)) {
430                         atomic_inc(&sk->sk_drops);
431                         kfree_skb(skb);
432                         return 0;
433                 }
434         }
435
436         rawv6_rcv_skb(sk, skb);
437         return 0;
438 }
439
440
441 /*
442  *      This should be easy, if there is something there
443  *      we return it, otherwise we block.
444  */
445
446 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
447                   struct msghdr *msg, size_t len,
448                   int noblock, int flags, int *addr_len)
449 {
450         struct ipv6_pinfo *np = inet6_sk(sk);
451         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
452         struct sk_buff *skb;
453         size_t copied;
454         int err;
455
456         if (flags & MSG_OOB)
457                 return -EOPNOTSUPP;
458
459         if (addr_len)
460                 *addr_len=sizeof(*sin6);
461
462         if (flags & MSG_ERRQUEUE)
463                 return ipv6_recv_error(sk, msg, len);
464
465         skb = skb_recv_datagram(sk, flags, noblock, &err);
466         if (!skb)
467                 goto out;
468
469         copied = skb->len;
470         if (copied > len) {
471                 copied = len;
472                 msg->msg_flags |= MSG_TRUNC;
473         }
474
475         if (skb_csum_unnecessary(skb)) {
476                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
477         } else if (msg->msg_flags&MSG_TRUNC) {
478                 if (__skb_checksum_complete(skb))
479                         goto csum_copy_err;
480                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
481         } else {
482                 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
483                 if (err == -EINVAL)
484                         goto csum_copy_err;
485         }
486         if (err)
487                 goto out_free;
488
489         /* Copy the address. */
490         if (sin6) {
491                 sin6->sin6_family = AF_INET6;
492                 sin6->sin6_port = 0;
493                 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
494                 sin6->sin6_flowinfo = 0;
495                 sin6->sin6_scope_id = 0;
496                 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
497                         sin6->sin6_scope_id = IP6CB(skb)->iif;
498         }
499
500         sock_recv_timestamp(msg, sk, skb);
501
502         if (np->rxopt.all)
503                 datagram_recv_ctl(sk, msg, skb);
504
505         err = copied;
506         if (flags & MSG_TRUNC)
507                 err = skb->len;
508
509 out_free:
510         skb_free_datagram(sk, skb);
511 out:
512         return err;
513
514 csum_copy_err:
515         skb_kill_datagram(sk, skb, flags);
516
517         /* Error for blocking case is chosen to masquerade
518            as some normal condition.
519          */
520         err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
521         atomic_inc(&sk->sk_drops);
522         goto out;
523 }
524
525 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
526                                      struct raw6_sock *rp)
527 {
528         struct sk_buff *skb;
529         int err = 0;
530         int offset;
531         int len;
532         int total_len;
533         __wsum tmp_csum;
534         __sum16 csum;
535
536         if (!rp->checksum)
537                 goto send;
538
539         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
540                 goto out;
541
542         offset = rp->offset;
543         total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
544                                                 skb->data);
545         if (offset >= total_len - 1) {
546                 err = -EINVAL;
547                 ip6_flush_pending_frames(sk);
548                 goto out;
549         }
550
551         /* should be check HW csum miyazawa */
552         if (skb_queue_len(&sk->sk_write_queue) == 1) {
553                 /*
554                  * Only one fragment on the socket.
555                  */
556                 tmp_csum = skb->csum;
557         } else {
558                 struct sk_buff *csum_skb = NULL;
559                 tmp_csum = 0;
560
561                 skb_queue_walk(&sk->sk_write_queue, skb) {
562                         tmp_csum = csum_add(tmp_csum, skb->csum);
563
564                         if (csum_skb)
565                                 continue;
566
567                         len = skb->len - skb_transport_offset(skb);
568                         if (offset >= len) {
569                                 offset -= len;
570                                 continue;
571                         }
572
573                         csum_skb = skb;
574                 }
575
576                 skb = csum_skb;
577         }
578
579         offset += skb_transport_offset(skb);
580         if (skb_copy_bits(skb, offset, &csum, 2))
581                 BUG();
582
583         /* in case cksum was not initialized */
584         if (unlikely(csum))
585                 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
586
587         csum = csum_ipv6_magic(&fl->fl6_src,
588                                    &fl->fl6_dst,
589                                    total_len, fl->proto, tmp_csum);
590
591         if (csum == 0 && fl->proto == IPPROTO_UDP)
592                 csum = CSUM_MANGLED_0;
593
594         if (skb_store_bits(skb, offset, &csum, 2))
595                 BUG();
596
597 send:
598         err = ip6_push_pending_frames(sk);
599 out:
600         return err;
601 }
602
603 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
604                         struct flowi *fl, struct rt6_info *rt,
605                         unsigned int flags)
606 {
607         struct ipv6_pinfo *np = inet6_sk(sk);
608         struct ipv6hdr *iph;
609         struct sk_buff *skb;
610         unsigned int hh_len;
611         int err;
612
613         if (length > rt->u.dst.dev->mtu) {
614                 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
615                 return -EMSGSIZE;
616         }
617         if (flags&MSG_PROBE)
618                 goto out;
619
620         hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
621
622         skb = sock_alloc_send_skb(sk, length+hh_len+15,
623                                   flags&MSG_DONTWAIT, &err);
624         if (skb == NULL)
625                 goto error;
626         skb_reserve(skb, hh_len);
627
628         skb->priority = sk->sk_priority;
629         skb->mark = sk->sk_mark;
630         skb->dst = dst_clone(&rt->u.dst);
631
632         skb_put(skb, length);
633         skb_reset_network_header(skb);
634         iph = ipv6_hdr(skb);
635
636         skb->ip_summed = CHECKSUM_NONE;
637
638         skb->transport_header = skb->network_header;
639         err = memcpy_fromiovecend((void *)iph, from, 0, length);
640         if (err)
641                 goto error_fault;
642
643         IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
644         err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
645                       dst_output);
646         if (err > 0)
647                 err = np->recverr ? net_xmit_errno(err) : 0;
648         if (err)
649                 goto error;
650 out:
651         return 0;
652
653 error_fault:
654         err = -EFAULT;
655         kfree_skb(skb);
656 error:
657         IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
658         return err;
659 }
660
661 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
662 {
663         struct iovec *iov;
664         u8 __user *type = NULL;
665         u8 __user *code = NULL;
666         u8 len = 0;
667         int probed = 0;
668         int i;
669
670         if (!msg->msg_iov)
671                 return 0;
672
673         for (i = 0; i < msg->msg_iovlen; i++) {
674                 iov = &msg->msg_iov[i];
675                 if (!iov)
676                         continue;
677
678                 switch (fl->proto) {
679                 case IPPROTO_ICMPV6:
680                         /* check if one-byte field is readable or not. */
681                         if (iov->iov_base && iov->iov_len < 1)
682                                 break;
683
684                         if (!type) {
685                                 type = iov->iov_base;
686                                 /* check if code field is readable or not. */
687                                 if (iov->iov_len > 1)
688                                         code = type + 1;
689                         } else if (!code)
690                                 code = iov->iov_base;
691
692                         if (type && code) {
693                                 if (get_user(fl->fl_icmp_type, type) ||
694                                     get_user(fl->fl_icmp_code, code))
695                                         return -EFAULT;
696                                 probed = 1;
697                         }
698                         break;
699                 case IPPROTO_MH:
700                         if (iov->iov_base && iov->iov_len < 1)
701                                 break;
702                         /* check if type field is readable or not. */
703                         if (iov->iov_len > 2 - len) {
704                                 u8 __user *p = iov->iov_base;
705                                 if (get_user(fl->fl_mh_type, &p[2 - len]))
706                                         return -EFAULT;
707                                 probed = 1;
708                         } else
709                                 len += iov->iov_len;
710
711                         break;
712                 default:
713                         probed = 1;
714                         break;
715                 }
716                 if (probed)
717                         break;
718         }
719         return 0;
720 }
721
722 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
723                    struct msghdr *msg, size_t len)
724 {
725         struct ipv6_txoptions opt_space;
726         struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
727         struct in6_addr *daddr, *final_p = NULL, final;
728         struct inet_sock *inet = inet_sk(sk);
729         struct ipv6_pinfo *np = inet6_sk(sk);
730         struct raw6_sock *rp = raw6_sk(sk);
731         struct ipv6_txoptions *opt = NULL;
732         struct ip6_flowlabel *flowlabel = NULL;
733         struct dst_entry *dst = NULL;
734         struct flowi fl;
735         int addr_len = msg->msg_namelen;
736         int hlimit = -1;
737         int tclass = -1;
738         u16 proto;
739         int err;
740
741         /* Rough check on arithmetic overflow,
742            better check is made in ip6_append_data().
743          */
744         if (len > INT_MAX)
745                 return -EMSGSIZE;
746
747         /* Mirror BSD error message compatibility */
748         if (msg->msg_flags & MSG_OOB)
749                 return -EOPNOTSUPP;
750
751         /*
752          *      Get and verify the address.
753          */
754         memset(&fl, 0, sizeof(fl));
755
756         fl.mark = sk->sk_mark;
757
758         if (sin6) {
759                 if (addr_len < SIN6_LEN_RFC2133)
760                         return -EINVAL;
761
762                 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
763                         return(-EAFNOSUPPORT);
764
765                 /* port is the proto value [0..255] carried in nexthdr */
766                 proto = ntohs(sin6->sin6_port);
767
768                 if (!proto)
769                         proto = inet->num;
770                 else if (proto != inet->num)
771                         return(-EINVAL);
772
773                 if (proto > 255)
774                         return(-EINVAL);
775
776                 daddr = &sin6->sin6_addr;
777                 if (np->sndflow) {
778                         fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
779                         if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
780                                 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
781                                 if (flowlabel == NULL)
782                                         return -EINVAL;
783                                 daddr = &flowlabel->dst;
784                         }
785                 }
786
787                 /*
788                  * Otherwise it will be difficult to maintain
789                  * sk->sk_dst_cache.
790                  */
791                 if (sk->sk_state == TCP_ESTABLISHED &&
792                     ipv6_addr_equal(daddr, &np->daddr))
793                         daddr = &np->daddr;
794
795                 if (addr_len >= sizeof(struct sockaddr_in6) &&
796                     sin6->sin6_scope_id &&
797                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
798                         fl.oif = sin6->sin6_scope_id;
799         } else {
800                 if (sk->sk_state != TCP_ESTABLISHED)
801                         return -EDESTADDRREQ;
802
803                 proto = inet->num;
804                 daddr = &np->daddr;
805                 fl.fl6_flowlabel = np->flow_label;
806         }
807
808         if (ipv6_addr_any(daddr)) {
809                 /*
810                  * unspecified destination address
811                  * treated as error... is this correct ?
812                  */
813                 fl6_sock_release(flowlabel);
814                 return(-EINVAL);
815         }
816
817         if (fl.oif == 0)
818                 fl.oif = sk->sk_bound_dev_if;
819
820         if (msg->msg_controllen) {
821                 opt = &opt_space;
822                 memset(opt, 0, sizeof(struct ipv6_txoptions));
823                 opt->tot_len = sizeof(struct ipv6_txoptions);
824
825                 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
826                 if (err < 0) {
827                         fl6_sock_release(flowlabel);
828                         return err;
829                 }
830                 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
831                         flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
832                         if (flowlabel == NULL)
833                                 return -EINVAL;
834                 }
835                 if (!(opt->opt_nflen|opt->opt_flen))
836                         opt = NULL;
837         }
838         if (opt == NULL)
839                 opt = np->opt;
840         if (flowlabel)
841                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
842         opt = ipv6_fixup_options(&opt_space, opt);
843
844         fl.proto = proto;
845         err = rawv6_probe_proto_opt(&fl, msg);
846         if (err)
847                 goto out;
848
849         ipv6_addr_copy(&fl.fl6_dst, daddr);
850         if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
851                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
852
853         /* merge ip6_build_xmit from ip6_output */
854         if (opt && opt->srcrt) {
855                 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
856                 ipv6_addr_copy(&final, &fl.fl6_dst);
857                 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
858                 final_p = &final;
859         }
860
861         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
862                 fl.oif = np->mcast_oif;
863         security_sk_classify_flow(sk, &fl);
864
865         err = ip6_dst_lookup(sk, &dst, &fl);
866         if (err)
867                 goto out;
868         if (final_p)
869                 ipv6_addr_copy(&fl.fl6_dst, final_p);
870
871         if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
872                 if (err == -EREMOTE)
873                         err = ip6_dst_blackhole(sk, &dst, &fl);
874                 if (err < 0)
875                         goto out;
876         }
877
878         if (hlimit < 0) {
879                 if (ipv6_addr_is_multicast(&fl.fl6_dst))
880                         hlimit = np->mcast_hops;
881                 else
882                         hlimit = np->hop_limit;
883                 if (hlimit < 0)
884                         hlimit = ip6_dst_hoplimit(dst);
885         }
886
887         if (tclass < 0) {
888                 tclass = np->tclass;
889                 if (tclass < 0)
890                         tclass = 0;
891         }
892
893         if (msg->msg_flags&MSG_CONFIRM)
894                 goto do_confirm;
895
896 back_from_confirm:
897         if (inet->hdrincl) {
898                 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
899         } else {
900                 lock_sock(sk);
901                 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
902                         len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
903                         msg->msg_flags);
904
905                 if (err)
906                         ip6_flush_pending_frames(sk);
907                 else if (!(msg->msg_flags & MSG_MORE))
908                         err = rawv6_push_pending_frames(sk, &fl, rp);
909                 release_sock(sk);
910         }
911 done:
912         dst_release(dst);
913 out:
914         fl6_sock_release(flowlabel);
915         return err<0?err:len;
916 do_confirm:
917         dst_confirm(dst);
918         if (!(msg->msg_flags & MSG_PROBE) || len)
919                 goto back_from_confirm;
920         err = 0;
921         goto done;
922 }
923
924 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
925                                char __user *optval, int optlen)
926 {
927         switch (optname) {
928         case ICMPV6_FILTER:
929                 if (optlen > sizeof(struct icmp6_filter))
930                         optlen = sizeof(struct icmp6_filter);
931                 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
932                         return -EFAULT;
933                 return 0;
934         default:
935                 return -ENOPROTOOPT;
936         }
937
938         return 0;
939 }
940
941 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
942                                char __user *optval, int __user *optlen)
943 {
944         int len;
945
946         switch (optname) {
947         case ICMPV6_FILTER:
948                 if (get_user(len, optlen))
949                         return -EFAULT;
950                 if (len < 0)
951                         return -EINVAL;
952                 if (len > sizeof(struct icmp6_filter))
953                         len = sizeof(struct icmp6_filter);
954                 if (put_user(len, optlen))
955                         return -EFAULT;
956                 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
957                         return -EFAULT;
958                 return 0;
959         default:
960                 return -ENOPROTOOPT;
961         }
962
963         return 0;
964 }
965
966
967 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
968                             char __user *optval, int optlen)
969 {
970         struct raw6_sock *rp = raw6_sk(sk);
971         int val;
972
973         if (get_user(val, (int __user *)optval))
974                 return -EFAULT;
975
976         switch (optname) {
977                 case IPV6_CHECKSUM:
978                         /* You may get strange result with a positive odd offset;
979                            RFC2292bis agrees with me. */
980                         if (val > 0 && (val&1))
981                                 return(-EINVAL);
982                         if (val < 0) {
983                                 rp->checksum = 0;
984                         } else {
985                                 rp->checksum = 1;
986                                 rp->offset = val;
987                         }
988
989                         return 0;
990                         break;
991
992                 default:
993                         return(-ENOPROTOOPT);
994         }
995 }
996
997 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
998                           char __user *optval, int optlen)
999 {
1000         switch(level) {
1001                 case SOL_RAW:
1002                         break;
1003
1004                 case SOL_ICMPV6:
1005                         if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1006                                 return -EOPNOTSUPP;
1007                         return rawv6_seticmpfilter(sk, level, optname, optval,
1008                                                    optlen);
1009                 case SOL_IPV6:
1010                         if (optname == IPV6_CHECKSUM)
1011                                 break;
1012                 default:
1013                         return ipv6_setsockopt(sk, level, optname, optval,
1014                                                optlen);
1015         }
1016
1017         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1018 }
1019
1020 #ifdef CONFIG_COMPAT
1021 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1022                                    char __user *optval, int optlen)
1023 {
1024         switch (level) {
1025         case SOL_RAW:
1026                 break;
1027         case SOL_ICMPV6:
1028                 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1029                         return -EOPNOTSUPP;
1030                 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1031         case SOL_IPV6:
1032                 if (optname == IPV6_CHECKSUM)
1033                         break;
1034         default:
1035                 return compat_ipv6_setsockopt(sk, level, optname,
1036                                               optval, optlen);
1037         }
1038         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1039 }
1040 #endif
1041
1042 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1043                             char __user *optval, int __user *optlen)
1044 {
1045         struct raw6_sock *rp = raw6_sk(sk);
1046         int val, len;
1047
1048         if (get_user(len,optlen))
1049                 return -EFAULT;
1050
1051         switch (optname) {
1052         case IPV6_CHECKSUM:
1053                 if (rp->checksum == 0)
1054                         val = -1;
1055                 else
1056                         val = rp->offset;
1057                 break;
1058
1059         default:
1060                 return -ENOPROTOOPT;
1061         }
1062
1063         len = min_t(unsigned int, sizeof(int), len);
1064
1065         if (put_user(len, optlen))
1066                 return -EFAULT;
1067         if (copy_to_user(optval,&val,len))
1068                 return -EFAULT;
1069         return 0;
1070 }
1071
1072 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1073                           char __user *optval, int __user *optlen)
1074 {
1075         switch(level) {
1076                 case SOL_RAW:
1077                         break;
1078
1079                 case SOL_ICMPV6:
1080                         if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1081                                 return -EOPNOTSUPP;
1082                         return rawv6_geticmpfilter(sk, level, optname, optval,
1083                                                    optlen);
1084                 case SOL_IPV6:
1085                         if (optname == IPV6_CHECKSUM)
1086                                 break;
1087                 default:
1088                         return ipv6_getsockopt(sk, level, optname, optval,
1089                                                optlen);
1090         }
1091
1092         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1093 }
1094
1095 #ifdef CONFIG_COMPAT
1096 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1097                                    char __user *optval, int __user *optlen)
1098 {
1099         switch (level) {
1100         case SOL_RAW:
1101                 break;
1102         case SOL_ICMPV6:
1103                 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1104                         return -EOPNOTSUPP;
1105                 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1106         case SOL_IPV6:
1107                 if (optname == IPV6_CHECKSUM)
1108                         break;
1109         default:
1110                 return compat_ipv6_getsockopt(sk, level, optname,
1111                                               optval, optlen);
1112         }
1113         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1114 }
1115 #endif
1116
1117 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1118 {
1119         switch(cmd) {
1120                 case SIOCOUTQ:
1121                 {
1122                         int amount = atomic_read(&sk->sk_wmem_alloc);
1123                         return put_user(amount, (int __user *)arg);
1124                 }
1125                 case SIOCINQ:
1126                 {
1127                         struct sk_buff *skb;
1128                         int amount = 0;
1129
1130                         spin_lock_bh(&sk->sk_receive_queue.lock);
1131                         skb = skb_peek(&sk->sk_receive_queue);
1132                         if (skb != NULL)
1133                                 amount = skb->tail - skb->transport_header;
1134                         spin_unlock_bh(&sk->sk_receive_queue.lock);
1135                         return put_user(amount, (int __user *)arg);
1136                 }
1137
1138                 default:
1139 #ifdef CONFIG_IPV6_MROUTE
1140                         return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1141 #else
1142                         return -ENOIOCTLCMD;
1143 #endif
1144         }
1145 }
1146
1147 static void rawv6_close(struct sock *sk, long timeout)
1148 {
1149         if (inet_sk(sk)->num == IPPROTO_RAW)
1150                 ip6_ra_control(sk, -1, NULL);
1151         ip6mr_sk_done(sk);
1152         sk_common_release(sk);
1153 }
1154
1155 static int rawv6_init_sk(struct sock *sk)
1156 {
1157         struct raw6_sock *rp = raw6_sk(sk);
1158
1159         switch (inet_sk(sk)->num) {
1160         case IPPROTO_ICMPV6:
1161                 rp->checksum = 1;
1162                 rp->offset   = 2;
1163                 break;
1164         case IPPROTO_MH:
1165                 rp->checksum = 1;
1166                 rp->offset   = 4;
1167                 break;
1168         default:
1169                 break;
1170         }
1171         return(0);
1172 }
1173
1174 struct proto rawv6_prot = {
1175         .name              = "RAWv6",
1176         .owner             = THIS_MODULE,
1177         .close             = rawv6_close,
1178         .connect           = ip6_datagram_connect,
1179         .disconnect        = udp_disconnect,
1180         .ioctl             = rawv6_ioctl,
1181         .init              = rawv6_init_sk,
1182         .destroy           = inet6_destroy_sock,
1183         .setsockopt        = rawv6_setsockopt,
1184         .getsockopt        = rawv6_getsockopt,
1185         .sendmsg           = rawv6_sendmsg,
1186         .recvmsg           = rawv6_recvmsg,
1187         .bind              = rawv6_bind,
1188         .backlog_rcv       = rawv6_rcv_skb,
1189         .hash              = raw_hash_sk,
1190         .unhash            = raw_unhash_sk,
1191         .obj_size          = sizeof(struct raw6_sock),
1192         .h.raw_hash        = &raw_v6_hashinfo,
1193 #ifdef CONFIG_COMPAT
1194         .compat_setsockopt = compat_rawv6_setsockopt,
1195         .compat_getsockopt = compat_rawv6_getsockopt,
1196 #endif
1197 };
1198
1199 #ifdef CONFIG_PROC_FS
1200 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1201 {
1202         struct ipv6_pinfo *np = inet6_sk(sp);
1203         struct in6_addr *dest, *src;
1204         __u16 destp, srcp;
1205
1206         dest  = &np->daddr;
1207         src   = &np->rcv_saddr;
1208         destp = 0;
1209         srcp  = inet_sk(sp)->num;
1210         seq_printf(seq,
1211                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1212                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1213                    i,
1214                    src->s6_addr32[0], src->s6_addr32[1],
1215                    src->s6_addr32[2], src->s6_addr32[3], srcp,
1216                    dest->s6_addr32[0], dest->s6_addr32[1],
1217                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
1218                    sp->sk_state,
1219                    atomic_read(&sp->sk_wmem_alloc),
1220                    atomic_read(&sp->sk_rmem_alloc),
1221                    0, 0L, 0,
1222                    sock_i_uid(sp), 0,
1223                    sock_i_ino(sp),
1224                    atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1225 }
1226
1227 static int raw6_seq_show(struct seq_file *seq, void *v)
1228 {
1229         if (v == SEQ_START_TOKEN)
1230                 seq_printf(seq,
1231                            "  sl  "
1232                            "local_address                         "
1233                            "remote_address                        "
1234                            "st tx_queue rx_queue tr tm->when retrnsmt"
1235                            "   uid  timeout inode  drops\n");
1236         else
1237                 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1238         return 0;
1239 }
1240
1241 static const struct seq_operations raw6_seq_ops = {
1242         .start =        raw_seq_start,
1243         .next =         raw_seq_next,
1244         .stop =         raw_seq_stop,
1245         .show =         raw6_seq_show,
1246 };
1247
1248 static int raw6_seq_open(struct inode *inode, struct file *file)
1249 {
1250         return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1251 }
1252
1253 static const struct file_operations raw6_seq_fops = {
1254         .owner =        THIS_MODULE,
1255         .open =         raw6_seq_open,
1256         .read =         seq_read,
1257         .llseek =       seq_lseek,
1258         .release =      seq_release_net,
1259 };
1260
1261 static int raw6_init_net(struct net *net)
1262 {
1263         if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1264                 return -ENOMEM;
1265
1266         return 0;
1267 }
1268
1269 static void raw6_exit_net(struct net *net)
1270 {
1271         proc_net_remove(net, "raw6");
1272 }
1273
1274 static struct pernet_operations raw6_net_ops = {
1275         .init = raw6_init_net,
1276         .exit = raw6_exit_net,
1277 };
1278
1279 int __init raw6_proc_init(void)
1280 {
1281         return register_pernet_subsys(&raw6_net_ops);
1282 }
1283
1284 void raw6_proc_exit(void)
1285 {
1286         unregister_pernet_subsys(&raw6_net_ops);
1287 }
1288 #endif  /* CONFIG_PROC_FS */
1289
1290 /* Same as inet6_dgram_ops, sans udp_poll.  */
1291 static const struct proto_ops inet6_sockraw_ops = {
1292         .family            = PF_INET6,
1293         .owner             = THIS_MODULE,
1294         .release           = inet6_release,
1295         .bind              = inet6_bind,
1296         .connect           = inet_dgram_connect,        /* ok           */
1297         .socketpair        = sock_no_socketpair,        /* a do nothing */
1298         .accept            = sock_no_accept,            /* a do nothing */
1299         .getname           = inet6_getname,
1300         .poll              = datagram_poll,             /* ok           */
1301         .ioctl             = inet6_ioctl,               /* must change  */
1302         .listen            = sock_no_listen,            /* ok           */
1303         .shutdown          = inet_shutdown,             /* ok           */
1304         .setsockopt        = sock_common_setsockopt,    /* ok           */
1305         .getsockopt        = sock_common_getsockopt,    /* ok           */
1306         .sendmsg           = inet_sendmsg,              /* ok           */
1307         .recvmsg           = sock_common_recvmsg,       /* ok           */
1308         .mmap              = sock_no_mmap,
1309         .sendpage          = sock_no_sendpage,
1310 #ifdef CONFIG_COMPAT
1311         .compat_setsockopt = compat_sock_common_setsockopt,
1312         .compat_getsockopt = compat_sock_common_getsockopt,
1313 #endif
1314 };
1315
1316 static struct inet_protosw rawv6_protosw = {
1317         .type           = SOCK_RAW,
1318         .protocol       = IPPROTO_IP,   /* wild card */
1319         .prot           = &rawv6_prot,
1320         .ops            = &inet6_sockraw_ops,
1321         .capability     = CAP_NET_RAW,
1322         .no_check       = UDP_CSUM_DEFAULT,
1323         .flags          = INET_PROTOSW_REUSE,
1324 };
1325
1326 int __init rawv6_init(void)
1327 {
1328         int ret;
1329
1330         ret = inet6_register_protosw(&rawv6_protosw);
1331         if (ret)
1332                 goto out;
1333 out:
1334         return ret;
1335 }
1336
1337 void rawv6_exit(void)
1338 {
1339         inet6_unregister_protosw(&rawv6_protosw);
1340 }