[IPSEC]: Restrict socket policy loading to CAP_NET_ADMIN.
[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  *      $Id: ipv6_sockglue.c,v 1.41 2002/02/01 22:01:04 davem Exp $
11  *
12  *      This program is free software; you can redistribute it and/or
13  *      modify it under the terms of the GNU General Public License
14  *      as published by the Free Software Foundation; either version
15  *      2 of the License, or (at your option) any later version.
16  *
17  *      FIXME: Make the setsockopt code POSIX compliant: That is
18  *
19  *      o       Return -EINVAL for setsockopt of short lengths
20  *      o       Truncate getsockopt returns
21  *      o       Return an optlen of the truncated length if need be
22  *
23  *      Changes:
24  *      David L Stevens <dlstevens@us.ibm.com>:
25  *              - added multicast source filtering API for MLDv2
26  */
27
28 #include <linux/module.h>
29 #include <linux/config.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/sched.h>
35 #include <linux/net.h>
36 #include <linux/in6.h>
37 #include <linux/netdevice.h>
38 #include <linux/if_arp.h>
39 #include <linux/init.h>
40 #include <linux/sysctl.h>
41 #include <linux/netfilter.h>
42
43 #include <net/sock.h>
44 #include <net/snmp.h>
45 #include <net/ipv6.h>
46 #include <net/ndisc.h>
47 #include <net/protocol.h>
48 #include <net/transp_v6.h>
49 #include <net/ip6_route.h>
50 #include <net/addrconf.h>
51 #include <net/inet_common.h>
52 #include <net/tcp.h>
53 #include <net/udp.h>
54 #include <net/xfrm.h>
55
56 #include <asm/uaccess.h>
57
58 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics);
59
60 static struct packet_type ipv6_packet_type = {
61         .type = __constant_htons(ETH_P_IPV6), 
62         .func = ipv6_rcv,
63 };
64
65 struct ip6_ra_chain *ip6_ra_chain;
66 DEFINE_RWLOCK(ip6_ra_lock);
67
68 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
69 {
70         struct ip6_ra_chain *ra, *new_ra, **rap;
71
72         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
73         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
74                 return -EINVAL;
75
76         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
77
78         write_lock_bh(&ip6_ra_lock);
79         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
80                 if (ra->sk == sk) {
81                         if (sel>=0) {
82                                 write_unlock_bh(&ip6_ra_lock);
83                                 if (new_ra)
84                                         kfree(new_ra);
85                                 return -EADDRINUSE;
86                         }
87
88                         *rap = ra->next;
89                         write_unlock_bh(&ip6_ra_lock);
90
91                         if (ra->destructor)
92                                 ra->destructor(sk);
93                         sock_put(sk);
94                         kfree(ra);
95                         return 0;
96                 }
97         }
98         if (new_ra == NULL) {
99                 write_unlock_bh(&ip6_ra_lock);
100                 return -ENOBUFS;
101         }
102         new_ra->sk = sk;
103         new_ra->sel = sel;
104         new_ra->destructor = destructor;
105         new_ra->next = ra;
106         *rap = new_ra;
107         sock_hold(sk);
108         write_unlock_bh(&ip6_ra_lock);
109         return 0;
110 }
111
112 extern int ip6_mc_source(int add, int omode, struct sock *sk,
113         struct group_source_req *pgsr);
114 extern int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf);
115 extern int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
116         struct group_filter __user *optval, int __user *optlen);
117
118
119 int ipv6_setsockopt(struct sock *sk, int level, int optname,
120                     char __user *optval, int optlen)
121 {
122         struct ipv6_pinfo *np = inet6_sk(sk);
123         int val, valbool;
124         int retv = -ENOPROTOOPT;
125
126         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
127                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
128
129         if(level!=SOL_IPV6)
130                 goto out;
131
132         if (optval == NULL)
133                 val=0;
134         else if (get_user(val, (int __user *) optval))
135                 return -EFAULT;
136
137         valbool = (val!=0);
138
139         lock_sock(sk);
140
141         switch (optname) {
142
143         case IPV6_ADDRFORM:
144                 if (val == PF_INET) {
145                         struct ipv6_txoptions *opt;
146                         struct sk_buff *pktopt;
147
148                         if (sk->sk_protocol != IPPROTO_UDP &&
149                             sk->sk_protocol != IPPROTO_TCP)
150                                 break;
151
152                         if (sk->sk_state != TCP_ESTABLISHED) {
153                                 retv = -ENOTCONN;
154                                 break;
155                         }
156
157                         if (ipv6_only_sock(sk) ||
158                             !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
159                                 retv = -EADDRNOTAVAIL;
160                                 break;
161                         }
162
163                         fl6_free_socklist(sk);
164                         ipv6_sock_mc_close(sk);
165
166                         if (sk->sk_protocol == IPPROTO_TCP) {
167                                 struct tcp_sock *tp = tcp_sk(sk);
168
169                                 local_bh_disable();
170                                 sock_prot_dec_use(sk->sk_prot);
171                                 sock_prot_inc_use(&tcp_prot);
172                                 local_bh_enable();
173                                 sk->sk_prot = &tcp_prot;
174                                 tp->af_specific = &ipv4_specific;
175                                 sk->sk_socket->ops = &inet_stream_ops;
176                                 sk->sk_family = PF_INET;
177                                 tcp_sync_mss(sk, tp->pmtu_cookie);
178                         } else {
179                                 local_bh_disable();
180                                 sock_prot_dec_use(sk->sk_prot);
181                                 sock_prot_inc_use(&udp_prot);
182                                 local_bh_enable();
183                                 sk->sk_prot = &udp_prot;
184                                 sk->sk_socket->ops = &inet_dgram_ops;
185                                 sk->sk_family = PF_INET;
186                         }
187                         opt = xchg(&np->opt, NULL);
188                         if (opt)
189                                 sock_kfree_s(sk, opt, opt->tot_len);
190                         pktopt = xchg(&np->pktoptions, NULL);
191                         if (pktopt)
192                                 kfree_skb(pktopt);
193
194                         sk->sk_destruct = inet_sock_destruct;
195 #ifdef INET_REFCNT_DEBUG
196                         atomic_dec(&inet6_sock_nr);
197 #endif
198                         module_put(THIS_MODULE);
199                         retv = 0;
200                         break;
201                 }
202                 goto e_inval;
203
204         case IPV6_V6ONLY:
205                 if (inet_sk(sk)->num)
206                         goto e_inval;
207                 np->ipv6only = valbool;
208                 retv = 0;
209                 break;
210
211         case IPV6_PKTINFO:
212                 np->rxopt.bits.rxinfo = valbool;
213                 retv = 0;
214                 break;
215
216         case IPV6_HOPLIMIT:
217                 np->rxopt.bits.rxhlim = valbool;
218                 retv = 0;
219                 break;
220
221         case IPV6_RTHDR:
222                 if (val < 0 || val > 2)
223                         goto e_inval;
224                 np->rxopt.bits.srcrt = val;
225                 retv = 0;
226                 break;
227
228         case IPV6_HOPOPTS:
229                 np->rxopt.bits.hopopts = valbool;
230                 retv = 0;
231                 break;
232
233         case IPV6_DSTOPTS:
234                 np->rxopt.bits.dstopts = valbool;
235                 retv = 0;
236                 break;
237
238         case IPV6_FLOWINFO:
239                 np->rxopt.bits.rxflow = valbool;
240                 retv = 0;
241                 break;
242
243         case IPV6_PKTOPTIONS:
244         {
245                 struct ipv6_txoptions *opt = NULL;
246                 struct msghdr msg;
247                 struct flowi fl;
248                 int junk;
249
250                 fl.fl6_flowlabel = 0;
251                 fl.oif = sk->sk_bound_dev_if;
252
253                 if (optlen == 0)
254                         goto update;
255
256                 /* 1K is probably excessive
257                  * 1K is surely not enough, 2K per standard header is 16K.
258                  */
259                 retv = -EINVAL;
260                 if (optlen > 64*1024)
261                         break;
262
263                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
264                 retv = -ENOBUFS;
265                 if (opt == NULL)
266                         break;
267
268                 memset(opt, 0, sizeof(*opt));
269                 opt->tot_len = sizeof(*opt) + optlen;
270                 retv = -EFAULT;
271                 if (copy_from_user(opt+1, optval, optlen))
272                         goto done;
273
274                 msg.msg_controllen = optlen;
275                 msg.msg_control = (void*)(opt+1);
276
277                 retv = datagram_send_ctl(&msg, &fl, opt, &junk);
278                 if (retv)
279                         goto done;
280 update:
281                 retv = 0;
282                 if (sk->sk_type == SOCK_STREAM) {
283                         if (opt) {
284                                 struct tcp_sock *tp = tcp_sk(sk);
285                                 if (!((1 << sk->sk_state) &
286                                       (TCPF_LISTEN | TCPF_CLOSE))
287                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
288                                         tp->ext_header_len = opt->opt_flen + opt->opt_nflen;
289                                         tcp_sync_mss(sk, tp->pmtu_cookie);
290                                 }
291                         }
292                         opt = xchg(&np->opt, opt);
293                         sk_dst_reset(sk);
294                 } else {
295                         write_lock(&sk->sk_dst_lock);
296                         opt = xchg(&np->opt, opt);
297                         write_unlock(&sk->sk_dst_lock);
298                         sk_dst_reset(sk);
299                 }
300
301 done:
302                 if (opt)
303                         sock_kfree_s(sk, opt, opt->tot_len);
304                 break;
305         }
306         case IPV6_UNICAST_HOPS:
307                 if (val > 255 || val < -1)
308                         goto e_inval;
309                 np->hop_limit = val;
310                 retv = 0;
311                 break;
312
313         case IPV6_MULTICAST_HOPS:
314                 if (sk->sk_type == SOCK_STREAM)
315                         goto e_inval;
316                 if (val > 255 || val < -1)
317                         goto e_inval;
318                 np->mcast_hops = val;
319                 retv = 0;
320                 break;
321
322         case IPV6_MULTICAST_LOOP:
323                 np->mc_loop = valbool;
324                 retv = 0;
325                 break;
326
327         case IPV6_MULTICAST_IF:
328                 if (sk->sk_type == SOCK_STREAM)
329                         goto e_inval;
330                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
331                         goto e_inval;
332
333                 if (__dev_get_by_index(val) == NULL) {
334                         retv = -ENODEV;
335                         break;
336                 }
337                 np->mcast_oif = val;
338                 retv = 0;
339                 break;
340         case IPV6_ADD_MEMBERSHIP:
341         case IPV6_DROP_MEMBERSHIP:
342         {
343                 struct ipv6_mreq mreq;
344
345                 retv = -EFAULT;
346                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
347                         break;
348
349                 if (optname == IPV6_ADD_MEMBERSHIP)
350                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
351                 else
352                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
353                 break;
354         }
355         case IPV6_JOIN_ANYCAST:
356         case IPV6_LEAVE_ANYCAST:
357         {
358                 struct ipv6_mreq mreq;
359
360                 if (optlen != sizeof(struct ipv6_mreq))
361                         goto e_inval;
362
363                 retv = -EFAULT;
364                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
365                         break;
366
367                 if (optname == IPV6_JOIN_ANYCAST)
368                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
369                 else
370                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
371                 break;
372         }
373         case MCAST_JOIN_GROUP:
374         case MCAST_LEAVE_GROUP:
375         {
376                 struct group_req greq;
377                 struct sockaddr_in6 *psin6;
378
379                 retv = -EFAULT;
380                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
381                         break;
382                 if (greq.gr_group.ss_family != AF_INET6) {
383                         retv = -EADDRNOTAVAIL;
384                         break;
385                 }
386                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
387                 if (optname == MCAST_JOIN_GROUP)
388                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
389                                 &psin6->sin6_addr);
390                 else
391                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
392                                 &psin6->sin6_addr);
393                 break;
394         }
395         case MCAST_JOIN_SOURCE_GROUP:
396         case MCAST_LEAVE_SOURCE_GROUP:
397         case MCAST_BLOCK_SOURCE:
398         case MCAST_UNBLOCK_SOURCE:
399         {
400                 struct group_source_req greqs;
401                 int omode, add;
402
403                 if (optlen != sizeof(struct group_source_req))
404                         goto e_inval;
405                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
406                         retv = -EFAULT;
407                         break;
408                 }
409                 if (greqs.gsr_group.ss_family != AF_INET6 ||
410                     greqs.gsr_source.ss_family != AF_INET6) {
411                         retv = -EADDRNOTAVAIL;
412                         break;
413                 }
414                 if (optname == MCAST_BLOCK_SOURCE) {
415                         omode = MCAST_EXCLUDE;
416                         add = 1;
417                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
418                         omode = MCAST_EXCLUDE;
419                         add = 0;
420                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
421                         struct sockaddr_in6 *psin6;
422
423                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
424                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
425                                 &psin6->sin6_addr);
426                         /* prior join w/ different source is ok */
427                         if (retv && retv != -EADDRINUSE)
428                                 break;
429                         omode = MCAST_INCLUDE;
430                         add = 1;
431                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
432                         omode = MCAST_INCLUDE;
433                         add = 0;
434                 }
435                 retv = ip6_mc_source(add, omode, sk, &greqs);
436                 break;
437         }
438         case MCAST_MSFILTER:
439         {
440                 extern int sysctl_optmem_max;
441                 extern int sysctl_mld_max_msf;
442                 struct group_filter *gsf;
443
444                 if (optlen < GROUP_FILTER_SIZE(0))
445                         goto e_inval;
446                 if (optlen > sysctl_optmem_max) {
447                         retv = -ENOBUFS;
448                         break;
449                 }
450                 gsf = (struct group_filter *)kmalloc(optlen,GFP_KERNEL);
451                 if (gsf == 0) {
452                         retv = -ENOBUFS;
453                         break;
454                 }
455                 retv = -EFAULT;
456                 if (copy_from_user(gsf, optval, optlen)) {
457                         kfree(gsf);
458                         break;
459                 }
460                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
461                 if (gsf->gf_numsrc >= 0x1ffffffU ||
462                     gsf->gf_numsrc > sysctl_mld_max_msf) {
463                         kfree(gsf);
464                         retv = -ENOBUFS;
465                         break;
466                 }
467                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
468                         kfree(gsf);
469                         retv = -EINVAL;
470                         break;
471                 }
472                 retv = ip6_mc_msfilter(sk, gsf);
473                 kfree(gsf);
474
475                 break;
476         }
477         case IPV6_ROUTER_ALERT:
478                 retv = ip6_ra_control(sk, val, NULL);
479                 break;
480         case IPV6_MTU_DISCOVER:
481                 if (val<0 || val>2)
482                         goto e_inval;
483                 np->pmtudisc = val;
484                 retv = 0;
485                 break;
486         case IPV6_MTU:
487                 if (val && val < IPV6_MIN_MTU)
488                         goto e_inval;
489                 np->frag_size = val;
490                 retv = 0;
491                 break;
492         case IPV6_RECVERR:
493                 np->recverr = valbool;
494                 if (!val)
495                         skb_queue_purge(&sk->sk_error_queue);
496                 retv = 0;
497                 break;
498         case IPV6_FLOWINFO_SEND:
499                 np->sndflow = valbool;
500                 retv = 0;
501                 break;
502         case IPV6_FLOWLABEL_MGR:
503                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
504                 break;
505         case IPV6_IPSEC_POLICY:
506         case IPV6_XFRM_POLICY:
507                 retv = -EPERM;
508                 if (!capable(CAP_NET_ADMIN))
509                         break;
510                 retv = xfrm_user_policy(sk, optname, optval, optlen);
511                 break;
512
513 #ifdef CONFIG_NETFILTER
514         default:
515                 retv = nf_setsockopt(sk, PF_INET6, optname, optval, 
516                                             optlen);
517                 break;
518 #endif
519
520         }
521         release_sock(sk);
522
523 out:
524         return retv;
525
526 e_inval:
527         release_sock(sk);
528         return -EINVAL;
529 }
530
531 int ipv6_getsockopt(struct sock *sk, int level, int optname,
532                     char __user *optval, int __user *optlen)
533 {
534         struct ipv6_pinfo *np = inet6_sk(sk);
535         int len;
536         int val;
537
538         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
539                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
540         if(level!=SOL_IPV6)
541                 return -ENOPROTOOPT;
542         if (get_user(len, optlen))
543                 return -EFAULT;
544         switch (optname) {
545         case IPV6_ADDRFORM:
546                 if (sk->sk_protocol != IPPROTO_UDP &&
547                     sk->sk_protocol != IPPROTO_TCP)
548                         return -EINVAL;
549                 if (sk->sk_state != TCP_ESTABLISHED)
550                         return -ENOTCONN;
551                 val = sk->sk_family;
552                 break;
553         case MCAST_MSFILTER:
554         {
555                 struct group_filter gsf;
556                 int err;
557
558                 if (len < GROUP_FILTER_SIZE(0))
559                         return -EINVAL;
560                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
561                         return -EFAULT;
562                 lock_sock(sk);
563                 err = ip6_mc_msfget(sk, &gsf,
564                         (struct group_filter __user *)optval, optlen);
565                 release_sock(sk);
566                 return err;
567         }
568
569         case IPV6_PKTOPTIONS:
570         {
571                 struct msghdr msg;
572                 struct sk_buff *skb;
573
574                 if (sk->sk_type != SOCK_STREAM)
575                         return -ENOPROTOOPT;
576
577                 msg.msg_control = optval;
578                 msg.msg_controllen = len;
579                 msg.msg_flags = 0;
580
581                 lock_sock(sk);
582                 skb = np->pktoptions;
583                 if (skb)
584                         atomic_inc(&skb->users);
585                 release_sock(sk);
586
587                 if (skb) {
588                         int err = datagram_recv_ctl(sk, &msg, skb);
589                         kfree_skb(skb);
590                         if (err)
591                                 return err;
592                 } else {
593                         if (np->rxopt.bits.rxinfo) {
594                                 struct in6_pktinfo src_info;
595                                 src_info.ipi6_ifindex = np->mcast_oif;
596                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
597                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
598                         }
599                         if (np->rxopt.bits.rxhlim) {
600                                 int hlim = np->mcast_hops;
601                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
602                         }
603                 }
604                 len -= msg.msg_controllen;
605                 return put_user(len, optlen);
606         }
607         case IPV6_MTU:
608         {
609                 struct dst_entry *dst;
610                 val = 0;        
611                 lock_sock(sk);
612                 dst = sk_dst_get(sk);
613                 if (dst) {
614                         val = dst_mtu(dst);
615                         dst_release(dst);
616                 }
617                 release_sock(sk);
618                 if (!val)
619                         return -ENOTCONN;
620                 break;
621         }
622
623         case IPV6_V6ONLY:
624                 val = np->ipv6only;
625                 break;
626
627         case IPV6_PKTINFO:
628                 val = np->rxopt.bits.rxinfo;
629                 break;
630
631         case IPV6_HOPLIMIT:
632                 val = np->rxopt.bits.rxhlim;
633                 break;
634
635         case IPV6_RTHDR:
636                 val = np->rxopt.bits.srcrt;
637                 break;
638
639         case IPV6_HOPOPTS:
640                 val = np->rxopt.bits.hopopts;
641                 break;
642
643         case IPV6_DSTOPTS:
644                 val = np->rxopt.bits.dstopts;
645                 break;
646
647         case IPV6_FLOWINFO:
648                 val = np->rxopt.bits.rxflow;
649                 break;
650
651         case IPV6_UNICAST_HOPS:
652                 val = np->hop_limit;
653                 break;
654
655         case IPV6_MULTICAST_HOPS:
656                 val = np->mcast_hops;
657                 break;
658
659         case IPV6_MULTICAST_LOOP:
660                 val = np->mc_loop;
661                 break;
662
663         case IPV6_MULTICAST_IF:
664                 val = np->mcast_oif;
665                 break;
666
667         case IPV6_MTU_DISCOVER:
668                 val = np->pmtudisc;
669                 break;
670
671         case IPV6_RECVERR:
672                 val = np->recverr;
673                 break;
674
675         case IPV6_FLOWINFO_SEND:
676                 val = np->sndflow;
677                 break;
678
679         default:
680 #ifdef CONFIG_NETFILTER
681                 lock_sock(sk);
682                 val = nf_getsockopt(sk, PF_INET6, optname, optval, 
683                                     &len);
684                 release_sock(sk);
685                 if (val >= 0)
686                         val = put_user(len, optlen);
687                 return val;
688 #else
689                 return -EINVAL;
690 #endif
691         }
692         len = min_t(unsigned int, sizeof(int), len);
693         if(put_user(len, optlen))
694                 return -EFAULT;
695         if(copy_to_user(optval,&val,len))
696                 return -EFAULT;
697         return 0;
698 }
699
700 void __init ipv6_packet_init(void)
701 {
702         dev_add_pack(&ipv6_packet_type);
703 }
704
705 void ipv6_packet_cleanup(void)
706 {
707         dev_remove_pack(&ipv6_packet_type);
708 }