Use sk_mark for IPv6 routing lookups
[linux-2.6.git] / net / ipv6 / ipv6_sockglue.c
1 /*
2  *      IPv6 BSD socket options interface
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      Based on linux/net/ipv4/ip_sockglue.c
9  *
10  *      This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  *
15  *      FIXME: Make the setsockopt code POSIX compliant: That is
16  *
17  *      o       Truncate getsockopt returns
18  *      o       Return an optlen of the truncated length if need be
19  *
20  *      Changes:
21  *      David L Stevens <dlstevens@us.ibm.com>:
22  *              - added multicast source filtering API for MLDv2
23  */
24
25 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/mroute6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/init.h>
37 #include <linux/sysctl.h>
38 #include <linux/netfilter.h>
39
40 #include <net/sock.h>
41 #include <net/snmp.h>
42 #include <net/ipv6.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/transp_v6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/inet_common.h>
49 #include <net/tcp.h>
50 #include <net/udp.h>
51 #include <net/udplite.h>
52 #include <net/xfrm.h>
53 #include <net/compat.h>
54
55 #include <asm/uaccess.h>
56
57 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
58
59 struct ip6_ra_chain *ip6_ra_chain;
60 DEFINE_RWLOCK(ip6_ra_lock);
61
62 int ip6_ra_control(struct sock *sk, int sel)
63 {
64         struct ip6_ra_chain *ra, *new_ra, **rap;
65
66         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
67         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
68                 return -ENOPROTOOPT;
69
70         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
71
72         write_lock_bh(&ip6_ra_lock);
73         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
74                 if (ra->sk == sk) {
75                         if (sel>=0) {
76                                 write_unlock_bh(&ip6_ra_lock);
77                                 kfree(new_ra);
78                                 return -EADDRINUSE;
79                         }
80
81                         *rap = ra->next;
82                         write_unlock_bh(&ip6_ra_lock);
83
84                         sock_put(sk);
85                         kfree(ra);
86                         return 0;
87                 }
88         }
89         if (new_ra == NULL) {
90                 write_unlock_bh(&ip6_ra_lock);
91                 return -ENOBUFS;
92         }
93         new_ra->sk = sk;
94         new_ra->sel = sel;
95         new_ra->next = ra;
96         *rap = new_ra;
97         sock_hold(sk);
98         write_unlock_bh(&ip6_ra_lock);
99         return 0;
100 }
101
102 static
103 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
104                                            struct ipv6_txoptions *opt)
105 {
106         if (inet_sk(sk)->is_icsk) {
107                 if (opt &&
108                     !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
109                     inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
110                         struct inet_connection_sock *icsk = inet_csk(sk);
111                         icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
112                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
113                 }
114                 opt = xchg(&inet6_sk(sk)->opt, opt);
115         } else {
116                 write_lock(&sk->sk_dst_lock);
117                 opt = xchg(&inet6_sk(sk)->opt, opt);
118                 write_unlock(&sk->sk_dst_lock);
119         }
120         sk_dst_reset(sk);
121
122         return opt;
123 }
124
125 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
126                     char __user *optval, unsigned int optlen)
127 {
128         struct ipv6_pinfo *np = inet6_sk(sk);
129         struct net *net = sock_net(sk);
130         int val, valbool;
131         int retv = -ENOPROTOOPT;
132
133         if (optval == NULL)
134                 val=0;
135         else {
136                 if (optlen >= sizeof(int)) {
137                         if (get_user(val, (int __user *) optval))
138                                 return -EFAULT;
139                 } else
140                         val = 0;
141         }
142
143         valbool = (val!=0);
144
145         if (ip6_mroute_opt(optname))
146                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
147
148         lock_sock(sk);
149
150         switch (optname) {
151
152         case IPV6_ADDRFORM:
153                 if (optlen < sizeof(int))
154                         goto e_inval;
155                 if (val == PF_INET) {
156                         struct ipv6_txoptions *opt;
157                         struct sk_buff *pktopt;
158
159                         if (sk->sk_type == SOCK_RAW)
160                                 break;
161
162                         if (sk->sk_protocol == IPPROTO_UDP ||
163                             sk->sk_protocol == IPPROTO_UDPLITE) {
164                                 struct udp_sock *up = udp_sk(sk);
165                                 if (up->pending == AF_INET6) {
166                                         retv = -EBUSY;
167                                         break;
168                                 }
169                         } else if (sk->sk_protocol != IPPROTO_TCP)
170                                 break;
171
172                         if (sk->sk_state != TCP_ESTABLISHED) {
173                                 retv = -ENOTCONN;
174                                 break;
175                         }
176
177                         if (ipv6_only_sock(sk) ||
178                             !ipv6_addr_v4mapped(&np->daddr)) {
179                                 retv = -EADDRNOTAVAIL;
180                                 break;
181                         }
182
183                         fl6_free_socklist(sk);
184                         ipv6_sock_mc_close(sk);
185
186                         /*
187                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
188                          * remove it from the refcnt debug socks count in the
189                          * original family...
190                          */
191                         sk_refcnt_debug_dec(sk);
192
193                         if (sk->sk_protocol == IPPROTO_TCP) {
194                                 struct inet_connection_sock *icsk = inet_csk(sk);
195                                 local_bh_disable();
196                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
197                                 sock_prot_inuse_add(net, &tcp_prot, 1);
198                                 local_bh_enable();
199                                 sk->sk_prot = &tcp_prot;
200                                 icsk->icsk_af_ops = &ipv4_specific;
201                                 sk->sk_socket->ops = &inet_stream_ops;
202                                 sk->sk_family = PF_INET;
203                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
204                         } else {
205                                 struct proto *prot = &udp_prot;
206
207                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
208                                         prot = &udplite_prot;
209                                 local_bh_disable();
210                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
211                                 sock_prot_inuse_add(net, prot, 1);
212                                 local_bh_enable();
213                                 sk->sk_prot = prot;
214                                 sk->sk_socket->ops = &inet_dgram_ops;
215                                 sk->sk_family = PF_INET;
216                         }
217                         opt = xchg(&np->opt, NULL);
218                         if (opt)
219                                 sock_kfree_s(sk, opt, opt->tot_len);
220                         pktopt = xchg(&np->pktoptions, NULL);
221                         kfree_skb(pktopt);
222
223                         sk->sk_destruct = inet_sock_destruct;
224                         /*
225                          * ... and add it to the refcnt debug socks count
226                          * in the new family. -acme
227                          */
228                         sk_refcnt_debug_inc(sk);
229                         module_put(THIS_MODULE);
230                         retv = 0;
231                         break;
232                 }
233                 goto e_inval;
234
235         case IPV6_V6ONLY:
236                 if (optlen < sizeof(int) ||
237                     inet_sk(sk)->num)
238                         goto e_inval;
239                 np->ipv6only = valbool;
240                 retv = 0;
241                 break;
242
243         case IPV6_RECVPKTINFO:
244                 if (optlen < sizeof(int))
245                         goto e_inval;
246                 np->rxopt.bits.rxinfo = valbool;
247                 retv = 0;
248                 break;
249
250         case IPV6_2292PKTINFO:
251                 if (optlen < sizeof(int))
252                         goto e_inval;
253                 np->rxopt.bits.rxoinfo = valbool;
254                 retv = 0;
255                 break;
256
257         case IPV6_RECVHOPLIMIT:
258                 if (optlen < sizeof(int))
259                         goto e_inval;
260                 np->rxopt.bits.rxhlim = valbool;
261                 retv = 0;
262                 break;
263
264         case IPV6_2292HOPLIMIT:
265                 if (optlen < sizeof(int))
266                         goto e_inval;
267                 np->rxopt.bits.rxohlim = valbool;
268                 retv = 0;
269                 break;
270
271         case IPV6_RECVRTHDR:
272                 if (optlen < sizeof(int))
273                         goto e_inval;
274                 np->rxopt.bits.srcrt = valbool;
275                 retv = 0;
276                 break;
277
278         case IPV6_2292RTHDR:
279                 if (optlen < sizeof(int))
280                         goto e_inval;
281                 np->rxopt.bits.osrcrt = valbool;
282                 retv = 0;
283                 break;
284
285         case IPV6_RECVHOPOPTS:
286                 if (optlen < sizeof(int))
287                         goto e_inval;
288                 np->rxopt.bits.hopopts = valbool;
289                 retv = 0;
290                 break;
291
292         case IPV6_2292HOPOPTS:
293                 if (optlen < sizeof(int))
294                         goto e_inval;
295                 np->rxopt.bits.ohopopts = valbool;
296                 retv = 0;
297                 break;
298
299         case IPV6_RECVDSTOPTS:
300                 if (optlen < sizeof(int))
301                         goto e_inval;
302                 np->rxopt.bits.dstopts = valbool;
303                 retv = 0;
304                 break;
305
306         case IPV6_2292DSTOPTS:
307                 if (optlen < sizeof(int))
308                         goto e_inval;
309                 np->rxopt.bits.odstopts = valbool;
310                 retv = 0;
311                 break;
312
313         case IPV6_TCLASS:
314                 if (optlen < sizeof(int))
315                         goto e_inval;
316                 if (val < -1 || val > 0xff)
317                         goto e_inval;
318                 /* RFC 3542, 6.5: default traffic class of 0x0 */
319                 if (val == -1)
320                         val = 0;
321                 np->tclass = val;
322                 retv = 0;
323                 break;
324
325         case IPV6_RECVTCLASS:
326                 if (optlen < sizeof(int))
327                         goto e_inval;
328                 np->rxopt.bits.rxtclass = valbool;
329                 retv = 0;
330                 break;
331
332         case IPV6_FLOWINFO:
333                 if (optlen < sizeof(int))
334                         goto e_inval;
335                 np->rxopt.bits.rxflow = valbool;
336                 retv = 0;
337                 break;
338
339         case IPV6_HOPOPTS:
340         case IPV6_RTHDRDSTOPTS:
341         case IPV6_RTHDR:
342         case IPV6_DSTOPTS:
343         {
344                 struct ipv6_txoptions *opt;
345
346                 /* remove any sticky options header with a zero option
347                  * length, per RFC3542.
348                  */
349                 if (optlen == 0)
350                         optval = NULL;
351                 else if (optval == NULL)
352                         goto e_inval;
353                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
354                          optlen & 0x7 || optlen > 8 * 255)
355                         goto e_inval;
356
357                 /* hop-by-hop / destination options are privileged option */
358                 retv = -EPERM;
359                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
360                         break;
361
362                 opt = ipv6_renew_options(sk, np->opt, optname,
363                                          (struct ipv6_opt_hdr __user *)optval,
364                                          optlen);
365                 if (IS_ERR(opt)) {
366                         retv = PTR_ERR(opt);
367                         break;
368                 }
369
370                 /* routing header option needs extra check */
371                 retv = -EINVAL;
372                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
373                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
374                         switch (rthdr->type) {
375 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
376                         case IPV6_SRCRT_TYPE_2:
377                                 if (rthdr->hdrlen != 2 ||
378                                     rthdr->segments_left != 1)
379                                         goto sticky_done;
380
381                                 break;
382 #endif
383                         default:
384                                 goto sticky_done;
385                         }
386                 }
387
388                 retv = 0;
389                 opt = ipv6_update_options(sk, opt);
390 sticky_done:
391                 if (opt)
392                         sock_kfree_s(sk, opt, opt->tot_len);
393                 break;
394         }
395
396         case IPV6_PKTINFO:
397         {
398                 struct in6_pktinfo pkt;
399
400                 if (optlen == 0)
401                         goto e_inval;
402                 else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
403                         goto e_inval;
404
405                 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
406                                 retv = -EFAULT;
407                                 break;
408                 }
409                 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
410                         goto e_inval;
411
412                 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
413                 ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
414                 retv = 0;
415                 break;
416         }
417
418         case IPV6_2292PKTOPTIONS:
419         {
420                 struct ipv6_txoptions *opt = NULL;
421                 struct msghdr msg;
422                 struct flowi fl;
423                 int junk;
424
425                 fl.fl6_flowlabel = 0;
426                 fl.oif = sk->sk_bound_dev_if;
427                 fl.mark = sk->sk_mark;
428
429                 if (optlen == 0)
430                         goto update;
431
432                 /* 1K is probably excessive
433                  * 1K is surely not enough, 2K per standard header is 16K.
434                  */
435                 retv = -EINVAL;
436                 if (optlen > 64*1024)
437                         break;
438
439                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
440                 retv = -ENOBUFS;
441                 if (opt == NULL)
442                         break;
443
444                 memset(opt, 0, sizeof(*opt));
445                 opt->tot_len = sizeof(*opt) + optlen;
446                 retv = -EFAULT;
447                 if (copy_from_user(opt+1, optval, optlen))
448                         goto done;
449
450                 msg.msg_controllen = optlen;
451                 msg.msg_control = (void*)(opt+1);
452
453                 retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk);
454                 if (retv)
455                         goto done;
456 update:
457                 retv = 0;
458                 opt = ipv6_update_options(sk, opt);
459 done:
460                 if (opt)
461                         sock_kfree_s(sk, opt, opt->tot_len);
462                 break;
463         }
464         case IPV6_UNICAST_HOPS:
465                 if (optlen < sizeof(int))
466                         goto e_inval;
467                 if (val > 255 || val < -1)
468                         goto e_inval;
469                 np->hop_limit = val;
470                 retv = 0;
471                 break;
472
473         case IPV6_MULTICAST_HOPS:
474                 if (sk->sk_type == SOCK_STREAM)
475                         break;
476                 if (optlen < sizeof(int))
477                         goto e_inval;
478                 if (val > 255 || val < -1)
479                         goto e_inval;
480                 np->mcast_hops = val;
481                 retv = 0;
482                 break;
483
484         case IPV6_MULTICAST_LOOP:
485                 if (optlen < sizeof(int))
486                         goto e_inval;
487                 if (val != valbool)
488                         goto e_inval;
489                 np->mc_loop = valbool;
490                 retv = 0;
491                 break;
492
493         case IPV6_MULTICAST_IF:
494                 if (sk->sk_type == SOCK_STREAM)
495                         break;
496                 if (optlen < sizeof(int))
497                         goto e_inval;
498
499                 if (val) {
500                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
501                                 goto e_inval;
502
503                         if (__dev_get_by_index(net, val) == NULL) {
504                                 retv = -ENODEV;
505                                 break;
506                         }
507                 }
508                 np->mcast_oif = val;
509                 retv = 0;
510                 break;
511         case IPV6_ADD_MEMBERSHIP:
512         case IPV6_DROP_MEMBERSHIP:
513         {
514                 struct ipv6_mreq mreq;
515
516                 if (optlen < sizeof(struct ipv6_mreq))
517                         goto e_inval;
518
519                 retv = -EPROTO;
520                 if (inet_sk(sk)->is_icsk)
521                         break;
522
523                 retv = -EFAULT;
524                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
525                         break;
526
527                 if (optname == IPV6_ADD_MEMBERSHIP)
528                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
529                 else
530                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
531                 break;
532         }
533         case IPV6_JOIN_ANYCAST:
534         case IPV6_LEAVE_ANYCAST:
535         {
536                 struct ipv6_mreq mreq;
537
538                 if (optlen < sizeof(struct ipv6_mreq))
539                         goto e_inval;
540
541                 retv = -EFAULT;
542                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
543                         break;
544
545                 if (optname == IPV6_JOIN_ANYCAST)
546                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
547                 else
548                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
549                 break;
550         }
551         case MCAST_JOIN_GROUP:
552         case MCAST_LEAVE_GROUP:
553         {
554                 struct group_req greq;
555                 struct sockaddr_in6 *psin6;
556
557                 if (optlen < sizeof(struct group_req))
558                         goto e_inval;
559
560                 retv = -EFAULT;
561                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
562                         break;
563                 if (greq.gr_group.ss_family != AF_INET6) {
564                         retv = -EADDRNOTAVAIL;
565                         break;
566                 }
567                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
568                 if (optname == MCAST_JOIN_GROUP)
569                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
570                                 &psin6->sin6_addr);
571                 else
572                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
573                                 &psin6->sin6_addr);
574                 break;
575         }
576         case MCAST_JOIN_SOURCE_GROUP:
577         case MCAST_LEAVE_SOURCE_GROUP:
578         case MCAST_BLOCK_SOURCE:
579         case MCAST_UNBLOCK_SOURCE:
580         {
581                 struct group_source_req greqs;
582                 int omode, add;
583
584                 if (optlen < sizeof(struct group_source_req))
585                         goto e_inval;
586                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
587                         retv = -EFAULT;
588                         break;
589                 }
590                 if (greqs.gsr_group.ss_family != AF_INET6 ||
591                     greqs.gsr_source.ss_family != AF_INET6) {
592                         retv = -EADDRNOTAVAIL;
593                         break;
594                 }
595                 if (optname == MCAST_BLOCK_SOURCE) {
596                         omode = MCAST_EXCLUDE;
597                         add = 1;
598                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
599                         omode = MCAST_EXCLUDE;
600                         add = 0;
601                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
602                         struct sockaddr_in6 *psin6;
603
604                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
605                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
606                                 &psin6->sin6_addr);
607                         /* prior join w/ different source is ok */
608                         if (retv && retv != -EADDRINUSE)
609                                 break;
610                         omode = MCAST_INCLUDE;
611                         add = 1;
612                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
613                         omode = MCAST_INCLUDE;
614                         add = 0;
615                 }
616                 retv = ip6_mc_source(add, omode, sk, &greqs);
617                 break;
618         }
619         case MCAST_MSFILTER:
620         {
621                 extern int sysctl_mld_max_msf;
622                 struct group_filter *gsf;
623
624                 if (optlen < GROUP_FILTER_SIZE(0))
625                         goto e_inval;
626                 if (optlen > sysctl_optmem_max) {
627                         retv = -ENOBUFS;
628                         break;
629                 }
630                 gsf = kmalloc(optlen,GFP_KERNEL);
631                 if (!gsf) {
632                         retv = -ENOBUFS;
633                         break;
634                 }
635                 retv = -EFAULT;
636                 if (copy_from_user(gsf, optval, optlen)) {
637                         kfree(gsf);
638                         break;
639                 }
640                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
641                 if (gsf->gf_numsrc >= 0x1ffffffU ||
642                     gsf->gf_numsrc > sysctl_mld_max_msf) {
643                         kfree(gsf);
644                         retv = -ENOBUFS;
645                         break;
646                 }
647                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
648                         kfree(gsf);
649                         retv = -EINVAL;
650                         break;
651                 }
652                 retv = ip6_mc_msfilter(sk, gsf);
653                 kfree(gsf);
654
655                 break;
656         }
657         case IPV6_ROUTER_ALERT:
658                 if (optlen < sizeof(int))
659                         goto e_inval;
660                 retv = ip6_ra_control(sk, val);
661                 break;
662         case IPV6_MTU_DISCOVER:
663                 if (optlen < sizeof(int))
664                         goto e_inval;
665                 if (val<0 || val>3)
666                         goto e_inval;
667                 np->pmtudisc = val;
668                 retv = 0;
669                 break;
670         case IPV6_MTU:
671                 if (optlen < sizeof(int))
672                         goto e_inval;
673                 if (val && val < IPV6_MIN_MTU)
674                         goto e_inval;
675                 np->frag_size = val;
676                 retv = 0;
677                 break;
678         case IPV6_RECVERR:
679                 if (optlen < sizeof(int))
680                         goto e_inval;
681                 np->recverr = valbool;
682                 if (!val)
683                         skb_queue_purge(&sk->sk_error_queue);
684                 retv = 0;
685                 break;
686         case IPV6_FLOWINFO_SEND:
687                 if (optlen < sizeof(int))
688                         goto e_inval;
689                 np->sndflow = valbool;
690                 retv = 0;
691                 break;
692         case IPV6_FLOWLABEL_MGR:
693                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
694                 break;
695         case IPV6_IPSEC_POLICY:
696         case IPV6_XFRM_POLICY:
697                 retv = -EPERM;
698                 if (!capable(CAP_NET_ADMIN))
699                         break;
700                 retv = xfrm_user_policy(sk, optname, optval, optlen);
701                 break;
702
703         case IPV6_ADDR_PREFERENCES:
704             {
705                 unsigned int pref = 0;
706                 unsigned int prefmask = ~0;
707
708                 if (optlen < sizeof(int))
709                         goto e_inval;
710
711                 retv = -EINVAL;
712
713                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
714                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
715                                IPV6_PREFER_SRC_TMP|
716                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
717                 case IPV6_PREFER_SRC_PUBLIC:
718                         pref |= IPV6_PREFER_SRC_PUBLIC;
719                         break;
720                 case IPV6_PREFER_SRC_TMP:
721                         pref |= IPV6_PREFER_SRC_TMP;
722                         break;
723                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
724                         break;
725                 case 0:
726                         goto pref_skip_pubtmp;
727                 default:
728                         goto e_inval;
729                 }
730
731                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
732                               IPV6_PREFER_SRC_TMP);
733 pref_skip_pubtmp:
734
735                 /* check HOME/COA conflicts */
736                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
737                 case IPV6_PREFER_SRC_HOME:
738                         break;
739                 case IPV6_PREFER_SRC_COA:
740                         pref |= IPV6_PREFER_SRC_COA;
741                 case 0:
742                         goto pref_skip_coa;
743                 default:
744                         goto e_inval;
745                 }
746
747                 prefmask &= ~IPV6_PREFER_SRC_COA;
748 pref_skip_coa:
749
750                 /* check CGA/NONCGA conflicts */
751                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
752                 case IPV6_PREFER_SRC_CGA:
753                 case IPV6_PREFER_SRC_NONCGA:
754                 case 0:
755                         break;
756                 default:
757                         goto e_inval;
758                 }
759
760                 np->srcprefs = (np->srcprefs & prefmask) | pref;
761                 retv = 0;
762
763                 break;
764             }
765         }
766
767         release_sock(sk);
768
769         return retv;
770
771 e_inval:
772         release_sock(sk);
773         return -EINVAL;
774 }
775
776 int ipv6_setsockopt(struct sock *sk, int level, int optname,
777                     char __user *optval, unsigned int optlen)
778 {
779         int err;
780
781         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
782                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
783
784         if (level != SOL_IPV6)
785                 return -ENOPROTOOPT;
786
787         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
788 #ifdef CONFIG_NETFILTER
789         /* we need to exclude all possible ENOPROTOOPTs except default case */
790         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
791                         optname != IPV6_XFRM_POLICY) {
792                 lock_sock(sk);
793                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
794                                 optlen);
795                 release_sock(sk);
796         }
797 #endif
798         return err;
799 }
800
801 EXPORT_SYMBOL(ipv6_setsockopt);
802
803 #ifdef CONFIG_COMPAT
804 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
805                            char __user *optval, unsigned int optlen)
806 {
807         int err;
808
809         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
810                 if (udp_prot.compat_setsockopt != NULL)
811                         return udp_prot.compat_setsockopt(sk, level, optname,
812                                                           optval, optlen);
813                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
814         }
815
816         if (level != SOL_IPV6)
817                 return -ENOPROTOOPT;
818
819         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
820                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
821                         ipv6_setsockopt);
822
823         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
824 #ifdef CONFIG_NETFILTER
825         /* we need to exclude all possible ENOPROTOOPTs except default case */
826         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
827             optname != IPV6_XFRM_POLICY) {
828                 lock_sock(sk);
829                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
830                                            optval, optlen);
831                 release_sock(sk);
832         }
833 #endif
834         return err;
835 }
836
837 EXPORT_SYMBOL(compat_ipv6_setsockopt);
838 #endif
839
840 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
841                                   int optname, char __user *optval, int len)
842 {
843         struct ipv6_opt_hdr *hdr;
844
845         if (!opt)
846                 return 0;
847
848         switch(optname) {
849         case IPV6_HOPOPTS:
850                 hdr = opt->hopopt;
851                 break;
852         case IPV6_RTHDRDSTOPTS:
853                 hdr = opt->dst0opt;
854                 break;
855         case IPV6_RTHDR:
856                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
857                 break;
858         case IPV6_DSTOPTS:
859                 hdr = opt->dst1opt;
860                 break;
861         default:
862                 return -EINVAL; /* should not happen */
863         }
864
865         if (!hdr)
866                 return 0;
867
868         len = min_t(unsigned int, len, ipv6_optlen(hdr));
869         if (copy_to_user(optval, hdr, len))
870                 return -EFAULT;
871         return len;
872 }
873
874 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
875                     char __user *optval, int __user *optlen)
876 {
877         struct ipv6_pinfo *np = inet6_sk(sk);
878         int len;
879         int val;
880
881         if (ip6_mroute_opt(optname))
882                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
883
884         if (get_user(len, optlen))
885                 return -EFAULT;
886         switch (optname) {
887         case IPV6_ADDRFORM:
888                 if (sk->sk_protocol != IPPROTO_UDP &&
889                     sk->sk_protocol != IPPROTO_UDPLITE &&
890                     sk->sk_protocol != IPPROTO_TCP)
891                         return -ENOPROTOOPT;
892                 if (sk->sk_state != TCP_ESTABLISHED)
893                         return -ENOTCONN;
894                 val = sk->sk_family;
895                 break;
896         case MCAST_MSFILTER:
897         {
898                 struct group_filter gsf;
899                 int err;
900
901                 if (len < GROUP_FILTER_SIZE(0))
902                         return -EINVAL;
903                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
904                         return -EFAULT;
905                 if (gsf.gf_group.ss_family != AF_INET6)
906                         return -EADDRNOTAVAIL;
907                 lock_sock(sk);
908                 err = ip6_mc_msfget(sk, &gsf,
909                         (struct group_filter __user *)optval, optlen);
910                 release_sock(sk);
911                 return err;
912         }
913
914         case IPV6_2292PKTOPTIONS:
915         {
916                 struct msghdr msg;
917                 struct sk_buff *skb;
918
919                 if (sk->sk_type != SOCK_STREAM)
920                         return -ENOPROTOOPT;
921
922                 msg.msg_control = optval;
923                 msg.msg_controllen = len;
924                 msg.msg_flags = 0;
925
926                 lock_sock(sk);
927                 skb = np->pktoptions;
928                 if (skb)
929                         atomic_inc(&skb->users);
930                 release_sock(sk);
931
932                 if (skb) {
933                         int err = datagram_recv_ctl(sk, &msg, skb);
934                         kfree_skb(skb);
935                         if (err)
936                                 return err;
937                 } else {
938                         if (np->rxopt.bits.rxinfo) {
939                                 struct in6_pktinfo src_info;
940                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
941                                         np->sticky_pktinfo.ipi6_ifindex;
942                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
943                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
944                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
945                         }
946                         if (np->rxopt.bits.rxhlim) {
947                                 int hlim = np->mcast_hops;
948                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
949                         }
950                         if (np->rxopt.bits.rxoinfo) {
951                                 struct in6_pktinfo src_info;
952                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
953                                         np->sticky_pktinfo.ipi6_ifindex;
954                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
955                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
956                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
957                         }
958                         if (np->rxopt.bits.rxohlim) {
959                                 int hlim = np->mcast_hops;
960                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
961                         }
962                 }
963                 len -= msg.msg_controllen;
964                 return put_user(len, optlen);
965         }
966         case IPV6_MTU:
967         {
968                 struct dst_entry *dst;
969                 val = 0;
970                 lock_sock(sk);
971                 dst = sk_dst_get(sk);
972                 if (dst) {
973                         val = dst_mtu(dst);
974                         dst_release(dst);
975                 }
976                 release_sock(sk);
977                 if (!val)
978                         return -ENOTCONN;
979                 break;
980         }
981
982         case IPV6_V6ONLY:
983                 val = np->ipv6only;
984                 break;
985
986         case IPV6_RECVPKTINFO:
987                 val = np->rxopt.bits.rxinfo;
988                 break;
989
990         case IPV6_2292PKTINFO:
991                 val = np->rxopt.bits.rxoinfo;
992                 break;
993
994         case IPV6_RECVHOPLIMIT:
995                 val = np->rxopt.bits.rxhlim;
996                 break;
997
998         case IPV6_2292HOPLIMIT:
999                 val = np->rxopt.bits.rxohlim;
1000                 break;
1001
1002         case IPV6_RECVRTHDR:
1003                 val = np->rxopt.bits.srcrt;
1004                 break;
1005
1006         case IPV6_2292RTHDR:
1007                 val = np->rxopt.bits.osrcrt;
1008                 break;
1009
1010         case IPV6_HOPOPTS:
1011         case IPV6_RTHDRDSTOPTS:
1012         case IPV6_RTHDR:
1013         case IPV6_DSTOPTS:
1014         {
1015
1016                 lock_sock(sk);
1017                 len = ipv6_getsockopt_sticky(sk, np->opt,
1018                                              optname, optval, len);
1019                 release_sock(sk);
1020                 /* check if ipv6_getsockopt_sticky() returns err code */
1021                 if (len < 0)
1022                         return len;
1023                 return put_user(len, optlen);
1024         }
1025
1026         case IPV6_RECVHOPOPTS:
1027                 val = np->rxopt.bits.hopopts;
1028                 break;
1029
1030         case IPV6_2292HOPOPTS:
1031                 val = np->rxopt.bits.ohopopts;
1032                 break;
1033
1034         case IPV6_RECVDSTOPTS:
1035                 val = np->rxopt.bits.dstopts;
1036                 break;
1037
1038         case IPV6_2292DSTOPTS:
1039                 val = np->rxopt.bits.odstopts;
1040                 break;
1041
1042         case IPV6_TCLASS:
1043                 val = np->tclass;
1044                 break;
1045
1046         case IPV6_RECVTCLASS:
1047                 val = np->rxopt.bits.rxtclass;
1048                 break;
1049
1050         case IPV6_FLOWINFO:
1051                 val = np->rxopt.bits.rxflow;
1052                 break;
1053
1054         case IPV6_UNICAST_HOPS:
1055         case IPV6_MULTICAST_HOPS:
1056         {
1057                 struct dst_entry *dst;
1058
1059                 if (optname == IPV6_UNICAST_HOPS)
1060                         val = np->hop_limit;
1061                 else
1062                         val = np->mcast_hops;
1063
1064                 dst = sk_dst_get(sk);
1065                 if (dst) {
1066                         if (val < 0)
1067                                 val = ip6_dst_hoplimit(dst);
1068                         dst_release(dst);
1069                 }
1070                 if (val < 0)
1071                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1072                 break;
1073         }
1074
1075         case IPV6_MULTICAST_LOOP:
1076                 val = np->mc_loop;
1077                 break;
1078
1079         case IPV6_MULTICAST_IF:
1080                 val = np->mcast_oif;
1081                 break;
1082
1083         case IPV6_MTU_DISCOVER:
1084                 val = np->pmtudisc;
1085                 break;
1086
1087         case IPV6_RECVERR:
1088                 val = np->recverr;
1089                 break;
1090
1091         case IPV6_FLOWINFO_SEND:
1092                 val = np->sndflow;
1093                 break;
1094
1095         case IPV6_ADDR_PREFERENCES:
1096                 val = 0;
1097
1098                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1099                         val |= IPV6_PREFER_SRC_TMP;
1100                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1101                         val |= IPV6_PREFER_SRC_PUBLIC;
1102                 else {
1103                         /* XXX: should we return system default? */
1104                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1105                 }
1106
1107                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1108                         val |= IPV6_PREFER_SRC_COA;
1109                 else
1110                         val |= IPV6_PREFER_SRC_HOME;
1111                 break;
1112
1113         default:
1114                 return -ENOPROTOOPT;
1115         }
1116         len = min_t(unsigned int, sizeof(int), len);
1117         if(put_user(len, optlen))
1118                 return -EFAULT;
1119         if(copy_to_user(optval,&val,len))
1120                 return -EFAULT;
1121         return 0;
1122 }
1123
1124 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1125                     char __user *optval, int __user *optlen)
1126 {
1127         int err;
1128
1129         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1130                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1131
1132         if(level != SOL_IPV6)
1133                 return -ENOPROTOOPT;
1134
1135         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1136 #ifdef CONFIG_NETFILTER
1137         /* we need to exclude all possible ENOPROTOOPTs except default case */
1138         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1139                 int len;
1140
1141                 if (get_user(len, optlen))
1142                         return -EFAULT;
1143
1144                 lock_sock(sk);
1145                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1146                                 &len);
1147                 release_sock(sk);
1148                 if (err >= 0)
1149                         err = put_user(len, optlen);
1150         }
1151 #endif
1152         return err;
1153 }
1154
1155 EXPORT_SYMBOL(ipv6_getsockopt);
1156
1157 #ifdef CONFIG_COMPAT
1158 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1159                            char __user *optval, int __user *optlen)
1160 {
1161         int err;
1162
1163         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1164                 if (udp_prot.compat_getsockopt != NULL)
1165                         return udp_prot.compat_getsockopt(sk, level, optname,
1166                                                           optval, optlen);
1167                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1168         }
1169
1170         if (level != SOL_IPV6)
1171                 return -ENOPROTOOPT;
1172
1173         if (optname == MCAST_MSFILTER)
1174                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1175                         ipv6_getsockopt);
1176
1177         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1178 #ifdef CONFIG_NETFILTER
1179         /* we need to exclude all possible ENOPROTOOPTs except default case */
1180         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1181                 int len;
1182
1183                 if (get_user(len, optlen))
1184                         return -EFAULT;
1185
1186                 lock_sock(sk);
1187                 err = compat_nf_getsockopt(sk, PF_INET6,
1188                                            optname, optval, &len);
1189                 release_sock(sk);
1190                 if (err >= 0)
1191                         err = put_user(len, optlen);
1192         }
1193 #endif
1194         return err;
1195 }
1196
1197 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1198 #endif
1199