ipv6_sockglue.c revision ba89966c1984513f4f2cc0a6c182266be44ddd03 
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
58DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
59
60static struct packet_type ipv6_packet_type = {
61	.type = __constant_htons(ETH_P_IPV6),
62	.func = ipv6_rcv,
63};
64
65struct ip6_ra_chain *ip6_ra_chain;
66DEFINE_RWLOCK(ip6_ra_lock);
67
68int 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
112int ipv6_setsockopt(struct sock *sk, int level, int optname,
113		    char __user *optval, int optlen)
114{
115	struct ipv6_pinfo *np = inet6_sk(sk);
116	int val, valbool;
117	int retv = -ENOPROTOOPT;
118
119	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
120		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
121
122	if(level!=SOL_IPV6)
123		goto out;
124
125	if (optval == NULL)
126		val=0;
127	else if (get_user(val, (int __user *) optval))
128		return -EFAULT;
129
130	valbool = (val!=0);
131
132	lock_sock(sk);
133
134	switch (optname) {
135
136	case IPV6_ADDRFORM:
137		if (val == PF_INET) {
138			struct ipv6_txoptions *opt;
139			struct sk_buff *pktopt;
140
141			if (sk->sk_protocol != IPPROTO_UDP &&
142			    sk->sk_protocol != IPPROTO_TCP)
143				break;
144
145			if (sk->sk_state != TCP_ESTABLISHED) {
146				retv = -ENOTCONN;
147				break;
148			}
149
150			if (ipv6_only_sock(sk) ||
151			    !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
152				retv = -EADDRNOTAVAIL;
153				break;
154			}
155
156			fl6_free_socklist(sk);
157			ipv6_sock_mc_close(sk);
158
159			/*
160			 * Sock is moving from IPv6 to IPv4 (sk_prot), so
161			 * remove it from the refcnt debug socks count in the
162			 * original family...
163			 */
164			sk_refcnt_debug_dec(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			/*
196			 * ... and add it to the refcnt debug socks count
197			 * in the new family. -acme
198			 */
199			sk_refcnt_debug_inc(sk);
200			module_put(THIS_MODULE);
201			retv = 0;
202			break;
203		}
204		goto e_inval;
205
206	case IPV6_V6ONLY:
207		if (inet_sk(sk)->num)
208			goto e_inval;
209		np->ipv6only = valbool;
210		retv = 0;
211		break;
212
213	case IPV6_PKTINFO:
214		np->rxopt.bits.rxinfo = valbool;
215		retv = 0;
216		break;
217
218	case IPV6_HOPLIMIT:
219		np->rxopt.bits.rxhlim = valbool;
220		retv = 0;
221		break;
222
223	case IPV6_RTHDR:
224		if (val < 0 || val > 2)
225			goto e_inval;
226		np->rxopt.bits.srcrt = val;
227		retv = 0;
228		break;
229
230	case IPV6_HOPOPTS:
231		np->rxopt.bits.hopopts = valbool;
232		retv = 0;
233		break;
234
235	case IPV6_DSTOPTS:
236		np->rxopt.bits.dstopts = valbool;
237		retv = 0;
238		break;
239
240	case IPV6_FLOWINFO:
241		np->rxopt.bits.rxflow = valbool;
242		retv = 0;
243		break;
244
245	case IPV6_PKTOPTIONS:
246	{
247		struct ipv6_txoptions *opt = NULL;
248		struct msghdr msg;
249		struct flowi fl;
250		int junk;
251
252		fl.fl6_flowlabel = 0;
253		fl.oif = sk->sk_bound_dev_if;
254
255		if (optlen == 0)
256			goto update;
257
258		/* 1K is probably excessive
259		 * 1K is surely not enough, 2K per standard header is 16K.
260		 */
261		retv = -EINVAL;
262		if (optlen > 64*1024)
263			break;
264
265		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
266		retv = -ENOBUFS;
267		if (opt == NULL)
268			break;
269
270		memset(opt, 0, sizeof(*opt));
271		opt->tot_len = sizeof(*opt) + optlen;
272		retv = -EFAULT;
273		if (copy_from_user(opt+1, optval, optlen))
274			goto done;
275
276		msg.msg_controllen = optlen;
277		msg.msg_control = (void*)(opt+1);
278
279		retv = datagram_send_ctl(&msg, &fl, opt, &junk);
280		if (retv)
281			goto done;
282update:
283		retv = 0;
284		if (sk->sk_type == SOCK_STREAM) {
285			if (opt) {
286				struct tcp_sock *tp = tcp_sk(sk);
287				if (!((1 << sk->sk_state) &
288				      (TCPF_LISTEN | TCPF_CLOSE))
289				    && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
290					tp->ext_header_len = opt->opt_flen + opt->opt_nflen;
291					tcp_sync_mss(sk, tp->pmtu_cookie);
292				}
293			}
294			opt = xchg(&np->opt, opt);
295			sk_dst_reset(sk);
296		} else {
297			write_lock(&sk->sk_dst_lock);
298			opt = xchg(&np->opt, opt);
299			write_unlock(&sk->sk_dst_lock);
300			sk_dst_reset(sk);
301		}
302
303done:
304		if (opt)
305			sock_kfree_s(sk, opt, opt->tot_len);
306		break;
307	}
308	case IPV6_UNICAST_HOPS:
309		if (val > 255 || val < -1)
310			goto e_inval;
311		np->hop_limit = val;
312		retv = 0;
313		break;
314
315	case IPV6_MULTICAST_HOPS:
316		if (sk->sk_type == SOCK_STREAM)
317			goto e_inval;
318		if (val > 255 || val < -1)
319			goto e_inval;
320		np->mcast_hops = val;
321		retv = 0;
322		break;
323
324	case IPV6_MULTICAST_LOOP:
325		np->mc_loop = valbool;
326		retv = 0;
327		break;
328
329	case IPV6_MULTICAST_IF:
330		if (sk->sk_type == SOCK_STREAM)
331			goto e_inval;
332		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
333			goto e_inval;
334
335		if (__dev_get_by_index(val) == NULL) {
336			retv = -ENODEV;
337			break;
338		}
339		np->mcast_oif = val;
340		retv = 0;
341		break;
342	case IPV6_ADD_MEMBERSHIP:
343	case IPV6_DROP_MEMBERSHIP:
344	{
345		struct ipv6_mreq mreq;
346
347		retv = -EFAULT;
348		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
349			break;
350
351		if (optname == IPV6_ADD_MEMBERSHIP)
352			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
353		else
354			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
355		break;
356	}
357	case IPV6_JOIN_ANYCAST:
358	case IPV6_LEAVE_ANYCAST:
359	{
360		struct ipv6_mreq mreq;
361
362		if (optlen != sizeof(struct ipv6_mreq))
363			goto e_inval;
364
365		retv = -EFAULT;
366		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
367			break;
368
369		if (optname == IPV6_JOIN_ANYCAST)
370			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
371		else
372			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
373		break;
374	}
375	case MCAST_JOIN_GROUP:
376	case MCAST_LEAVE_GROUP:
377	{
378		struct group_req greq;
379		struct sockaddr_in6 *psin6;
380
381		retv = -EFAULT;
382		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
383			break;
384		if (greq.gr_group.ss_family != AF_INET6) {
385			retv = -EADDRNOTAVAIL;
386			break;
387		}
388		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
389		if (optname == MCAST_JOIN_GROUP)
390			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
391				&psin6->sin6_addr);
392		else
393			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
394				&psin6->sin6_addr);
395		break;
396	}
397	case MCAST_JOIN_SOURCE_GROUP:
398	case MCAST_LEAVE_SOURCE_GROUP:
399	case MCAST_BLOCK_SOURCE:
400	case MCAST_UNBLOCK_SOURCE:
401	{
402		struct group_source_req greqs;
403		int omode, add;
404
405		if (optlen != sizeof(struct group_source_req))
406			goto e_inval;
407		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
408			retv = -EFAULT;
409			break;
410		}
411		if (greqs.gsr_group.ss_family != AF_INET6 ||
412		    greqs.gsr_source.ss_family != AF_INET6) {
413			retv = -EADDRNOTAVAIL;
414			break;
415		}
416		if (optname == MCAST_BLOCK_SOURCE) {
417			omode = MCAST_EXCLUDE;
418			add = 1;
419		} else if (optname == MCAST_UNBLOCK_SOURCE) {
420			omode = MCAST_EXCLUDE;
421			add = 0;
422		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
423			struct sockaddr_in6 *psin6;
424
425			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
426			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
427				&psin6->sin6_addr);
428			/* prior join w/ different source is ok */
429			if (retv && retv != -EADDRINUSE)
430				break;
431			omode = MCAST_INCLUDE;
432			add = 1;
433		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
434			omode = MCAST_INCLUDE;
435			add = 0;
436		}
437		retv = ip6_mc_source(add, omode, sk, &greqs);
438		break;
439	}
440	case MCAST_MSFILTER:
441	{
442		extern int sysctl_mld_max_msf;
443		struct group_filter *gsf;
444
445		if (optlen < GROUP_FILTER_SIZE(0))
446			goto e_inval;
447		if (optlen > sysctl_optmem_max) {
448			retv = -ENOBUFS;
449			break;
450		}
451		gsf = (struct group_filter *)kmalloc(optlen,GFP_KERNEL);
452		if (gsf == 0) {
453			retv = -ENOBUFS;
454			break;
455		}
456		retv = -EFAULT;
457		if (copy_from_user(gsf, optval, optlen)) {
458			kfree(gsf);
459			break;
460		}
461		/* numsrc >= (4G-140)/128 overflow in 32 bits */
462		if (gsf->gf_numsrc >= 0x1ffffffU ||
463		    gsf->gf_numsrc > sysctl_mld_max_msf) {
464			kfree(gsf);
465			retv = -ENOBUFS;
466			break;
467		}
468		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
469			kfree(gsf);
470			retv = -EINVAL;
471			break;
472		}
473		retv = ip6_mc_msfilter(sk, gsf);
474		kfree(gsf);
475
476		break;
477	}
478	case IPV6_ROUTER_ALERT:
479		retv = ip6_ra_control(sk, val, NULL);
480		break;
481	case IPV6_MTU_DISCOVER:
482		if (val<0 || val>2)
483			goto e_inval;
484		np->pmtudisc = val;
485		retv = 0;
486		break;
487	case IPV6_MTU:
488		if (val && val < IPV6_MIN_MTU)
489			goto e_inval;
490		np->frag_size = val;
491		retv = 0;
492		break;
493	case IPV6_RECVERR:
494		np->recverr = valbool;
495		if (!val)
496			skb_queue_purge(&sk->sk_error_queue);
497		retv = 0;
498		break;
499	case IPV6_FLOWINFO_SEND:
500		np->sndflow = valbool;
501		retv = 0;
502		break;
503	case IPV6_FLOWLABEL_MGR:
504		retv = ipv6_flowlabel_opt(sk, optval, optlen);
505		break;
506	case IPV6_IPSEC_POLICY:
507	case IPV6_XFRM_POLICY:
508		retv = -EPERM;
509		if (!capable(CAP_NET_ADMIN))
510			break;
511		retv = xfrm_user_policy(sk, optname, optval, optlen);
512		break;
513
514#ifdef CONFIG_NETFILTER
515	default:
516		retv = nf_setsockopt(sk, PF_INET6, optname, optval,
517					    optlen);
518		break;
519#endif
520
521	}
522	release_sock(sk);
523
524out:
525	return retv;
526
527e_inval:
528	release_sock(sk);
529	return -EINVAL;
530}
531
532int ipv6_getsockopt(struct sock *sk, int level, int optname,
533		    char __user *optval, int __user *optlen)
534{
535	struct ipv6_pinfo *np = inet6_sk(sk);
536	int len;
537	int val;
538
539	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
540		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
541	if(level!=SOL_IPV6)
542		return -ENOPROTOOPT;
543	if (get_user(len, optlen))
544		return -EFAULT;
545	switch (optname) {
546	case IPV6_ADDRFORM:
547		if (sk->sk_protocol != IPPROTO_UDP &&
548		    sk->sk_protocol != IPPROTO_TCP)
549			return -EINVAL;
550		if (sk->sk_state != TCP_ESTABLISHED)
551			return -ENOTCONN;
552		val = sk->sk_family;
553		break;
554	case MCAST_MSFILTER:
555	{
556		struct group_filter gsf;
557		int err;
558
559		if (len < GROUP_FILTER_SIZE(0))
560			return -EINVAL;
561		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
562			return -EFAULT;
563		lock_sock(sk);
564		err = ip6_mc_msfget(sk, &gsf,
565			(struct group_filter __user *)optval, optlen);
566		release_sock(sk);
567		return err;
568	}
569
570	case IPV6_PKTOPTIONS:
571	{
572		struct msghdr msg;
573		struct sk_buff *skb;
574
575		if (sk->sk_type != SOCK_STREAM)
576			return -ENOPROTOOPT;
577
578		msg.msg_control = optval;
579		msg.msg_controllen = len;
580		msg.msg_flags = 0;
581
582		lock_sock(sk);
583		skb = np->pktoptions;
584		if (skb)
585			atomic_inc(&skb->users);
586		release_sock(sk);
587
588		if (skb) {
589			int err = datagram_recv_ctl(sk, &msg, skb);
590			kfree_skb(skb);
591			if (err)
592				return err;
593		} else {
594			if (np->rxopt.bits.rxinfo) {
595				struct in6_pktinfo src_info;
596				src_info.ipi6_ifindex = np->mcast_oif;
597				ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
598				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
599			}
600			if (np->rxopt.bits.rxhlim) {
601				int hlim = np->mcast_hops;
602				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
603			}
604		}
605		len -= msg.msg_controllen;
606		return put_user(len, optlen);
607	}
608	case IPV6_MTU:
609	{
610		struct dst_entry *dst;
611		val = 0;
612		lock_sock(sk);
613		dst = sk_dst_get(sk);
614		if (dst) {
615			val = dst_mtu(dst);
616			dst_release(dst);
617		}
618		release_sock(sk);
619		if (!val)
620			return -ENOTCONN;
621		break;
622	}
623
624	case IPV6_V6ONLY:
625		val = np->ipv6only;
626		break;
627
628	case IPV6_PKTINFO:
629		val = np->rxopt.bits.rxinfo;
630		break;
631
632	case IPV6_HOPLIMIT:
633		val = np->rxopt.bits.rxhlim;
634		break;
635
636	case IPV6_RTHDR:
637		val = np->rxopt.bits.srcrt;
638		break;
639
640	case IPV6_HOPOPTS:
641		val = np->rxopt.bits.hopopts;
642		break;
643
644	case IPV6_DSTOPTS:
645		val = np->rxopt.bits.dstopts;
646		break;
647
648	case IPV6_FLOWINFO:
649		val = np->rxopt.bits.rxflow;
650		break;
651
652	case IPV6_UNICAST_HOPS:
653		val = np->hop_limit;
654		break;
655
656	case IPV6_MULTICAST_HOPS:
657		val = np->mcast_hops;
658		break;
659
660	case IPV6_MULTICAST_LOOP:
661		val = np->mc_loop;
662		break;
663
664	case IPV6_MULTICAST_IF:
665		val = np->mcast_oif;
666		break;
667
668	case IPV6_MTU_DISCOVER:
669		val = np->pmtudisc;
670		break;
671
672	case IPV6_RECVERR:
673		val = np->recverr;
674		break;
675
676	case IPV6_FLOWINFO_SEND:
677		val = np->sndflow;
678		break;
679
680	default:
681#ifdef CONFIG_NETFILTER
682		lock_sock(sk);
683		val = nf_getsockopt(sk, PF_INET6, optname, optval,
684				    &len);
685		release_sock(sk);
686		if (val >= 0)
687			val = put_user(len, optlen);
688		return val;
689#else
690		return -EINVAL;
691#endif
692	}
693	len = min_t(unsigned int, sizeof(int), len);
694	if(put_user(len, optlen))
695		return -EFAULT;
696	if(copy_to_user(optval,&val,len))
697		return -EFAULT;
698	return 0;
699}
700
701void __init ipv6_packet_init(void)
702{
703	dev_add_pack(&ipv6_packet_type);
704}
705
706void ipv6_packet_cleanup(void)
707{
708	dev_remove_pack(&ipv6_packet_type);
709}
710