xref: /linux/net/ipv4/ip_sockglue.c (revision 13abf8130139c2ccd4962a7e5a8902be5e6cb5a7)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		The IP to API glue.
7  *
8  * Version:	$Id: ip_sockglue.c,v 1.62 2002/02/01 22:01:04 davem Exp $
9  *
10  * Authors:	see ip.c
11  *
12  * Fixes:
13  *		Many		:	Split from ip.c , see ip.c for history.
14  *		Martin Mares	:	TOS setting fixed.
15  *		Alan Cox	:	Fixed a couple of oopses in Martin's
16  *					TOS tweaks.
17  *		Mike McLagan	:	Routing by source
18  */
19 
20 #include <linux/config.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/mm.h>
24 #include <linux/sched.h>
25 #include <linux/skbuff.h>
26 #include <linux/ip.h>
27 #include <linux/icmp.h>
28 #include <linux/netdevice.h>
29 #include <net/sock.h>
30 #include <net/ip.h>
31 #include <net/icmp.h>
32 #include <net/tcp.h>
33 #include <linux/tcp.h>
34 #include <linux/udp.h>
35 #include <linux/igmp.h>
36 #include <linux/netfilter.h>
37 #include <linux/route.h>
38 #include <linux/mroute.h>
39 #include <net/route.h>
40 #include <net/xfrm.h>
41 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
42 #include <net/transp_v6.h>
43 #endif
44 
45 #include <linux/errqueue.h>
46 #include <asm/uaccess.h>
47 
48 #define IP_CMSG_PKTINFO		1
49 #define IP_CMSG_TTL		2
50 #define IP_CMSG_TOS		4
51 #define IP_CMSG_RECVOPTS	8
52 #define IP_CMSG_RETOPTS		16
53 
54 /*
55  *	SOL_IP control messages.
56  */
57 
58 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
59 {
60 	struct in_pktinfo info;
61 	struct rtable *rt = (struct rtable *)skb->dst;
62 
63 	info.ipi_addr.s_addr = skb->nh.iph->daddr;
64 	if (rt) {
65 		info.ipi_ifindex = rt->rt_iif;
66 		info.ipi_spec_dst.s_addr = rt->rt_spec_dst;
67 	} else {
68 		info.ipi_ifindex = 0;
69 		info.ipi_spec_dst.s_addr = 0;
70 	}
71 
72 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
73 }
74 
75 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
76 {
77 	int ttl = skb->nh.iph->ttl;
78 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
79 }
80 
81 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
82 {
83 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &skb->nh.iph->tos);
84 }
85 
86 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
87 {
88 	if (IPCB(skb)->opt.optlen == 0)
89 		return;
90 
91 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen, skb->nh.iph+1);
92 }
93 
94 
95 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
96 {
97 	unsigned char optbuf[sizeof(struct ip_options) + 40];
98 	struct ip_options * opt = (struct ip_options*)optbuf;
99 
100 	if (IPCB(skb)->opt.optlen == 0)
101 		return;
102 
103 	if (ip_options_echo(opt, skb)) {
104 		msg->msg_flags |= MSG_CTRUNC;
105 		return;
106 	}
107 	ip_options_undo(opt);
108 
109 	put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
110 }
111 
112 
113 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
114 {
115 	struct inet_sock *inet = inet_sk(skb->sk);
116 	unsigned flags = inet->cmsg_flags;
117 
118 	/* Ordered by supposed usage frequency */
119 	if (flags & 1)
120 		ip_cmsg_recv_pktinfo(msg, skb);
121 	if ((flags>>=1) == 0)
122 		return;
123 
124 	if (flags & 1)
125 		ip_cmsg_recv_ttl(msg, skb);
126 	if ((flags>>=1) == 0)
127 		return;
128 
129 	if (flags & 1)
130 		ip_cmsg_recv_tos(msg, skb);
131 	if ((flags>>=1) == 0)
132 		return;
133 
134 	if (flags & 1)
135 		ip_cmsg_recv_opts(msg, skb);
136 	if ((flags>>=1) == 0)
137 		return;
138 
139 	if (flags & 1)
140 		ip_cmsg_recv_retopts(msg, skb);
141 }
142 
143 int ip_cmsg_send(struct msghdr *msg, struct ipcm_cookie *ipc)
144 {
145 	int err;
146 	struct cmsghdr *cmsg;
147 
148 	for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
149 		if (!CMSG_OK(msg, cmsg))
150 			return -EINVAL;
151 		if (cmsg->cmsg_level != SOL_IP)
152 			continue;
153 		switch (cmsg->cmsg_type) {
154 		case IP_RETOPTS:
155 			err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
156 			err = ip_options_get(&ipc->opt, CMSG_DATA(cmsg), err < 40 ? err : 40);
157 			if (err)
158 				return err;
159 			break;
160 		case IP_PKTINFO:
161 		{
162 			struct in_pktinfo *info;
163 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
164 				return -EINVAL;
165 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
166 			ipc->oif = info->ipi_ifindex;
167 			ipc->addr = info->ipi_spec_dst.s_addr;
168 			break;
169 		}
170 		default:
171 			return -EINVAL;
172 		}
173 	}
174 	return 0;
175 }
176 
177 
178 /* Special input handler for packets caught by router alert option.
179    They are selected only by protocol field, and then processed likely
180    local ones; but only if someone wants them! Otherwise, router
181    not running rsvpd will kill RSVP.
182 
183    It is user level problem, what it will make with them.
184    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
185    but receiver should be enough clever f.e. to forward mtrace requests,
186    sent to multicast group to reach destination designated router.
187  */
188 struct ip_ra_chain *ip_ra_chain;
189 DEFINE_RWLOCK(ip_ra_lock);
190 
191 int ip_ra_control(struct sock *sk, unsigned char on, void (*destructor)(struct sock *))
192 {
193 	struct ip_ra_chain *ra, *new_ra, **rap;
194 
195 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num == IPPROTO_RAW)
196 		return -EINVAL;
197 
198 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
199 
200 	write_lock_bh(&ip_ra_lock);
201 	for (rap = &ip_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
202 		if (ra->sk == sk) {
203 			if (on) {
204 				write_unlock_bh(&ip_ra_lock);
205 				if (new_ra)
206 					kfree(new_ra);
207 				return -EADDRINUSE;
208 			}
209 			*rap = ra->next;
210 			write_unlock_bh(&ip_ra_lock);
211 
212 			if (ra->destructor)
213 				ra->destructor(sk);
214 			sock_put(sk);
215 			kfree(ra);
216 			return 0;
217 		}
218 	}
219 	if (new_ra == NULL) {
220 		write_unlock_bh(&ip_ra_lock);
221 		return -ENOBUFS;
222 	}
223 	new_ra->sk = sk;
224 	new_ra->destructor = destructor;
225 
226 	new_ra->next = ra;
227 	*rap = new_ra;
228 	sock_hold(sk);
229 	write_unlock_bh(&ip_ra_lock);
230 
231 	return 0;
232 }
233 
234 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
235 		   u16 port, u32 info, u8 *payload)
236 {
237 	struct inet_sock *inet = inet_sk(sk);
238 	struct sock_exterr_skb *serr;
239 
240 	if (!inet->recverr)
241 		return;
242 
243 	skb = skb_clone(skb, GFP_ATOMIC);
244 	if (!skb)
245 		return;
246 
247 	serr = SKB_EXT_ERR(skb);
248 	serr->ee.ee_errno = err;
249 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
250 	serr->ee.ee_type = skb->h.icmph->type;
251 	serr->ee.ee_code = skb->h.icmph->code;
252 	serr->ee.ee_pad = 0;
253 	serr->ee.ee_info = info;
254 	serr->ee.ee_data = 0;
255 	serr->addr_offset = (u8*)&(((struct iphdr*)(skb->h.icmph+1))->daddr) - skb->nh.raw;
256 	serr->port = port;
257 
258 	skb->h.raw = payload;
259 	if (!skb_pull(skb, payload - skb->data) ||
260 	    sock_queue_err_skb(sk, skb))
261 		kfree_skb(skb);
262 }
263 
264 void ip_local_error(struct sock *sk, int err, u32 daddr, u16 port, u32 info)
265 {
266 	struct inet_sock *inet = inet_sk(sk);
267 	struct sock_exterr_skb *serr;
268 	struct iphdr *iph;
269 	struct sk_buff *skb;
270 
271 	if (!inet->recverr)
272 		return;
273 
274 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
275 	if (!skb)
276 		return;
277 
278 	iph = (struct iphdr*)skb_put(skb, sizeof(struct iphdr));
279 	skb->nh.iph = iph;
280 	iph->daddr = daddr;
281 
282 	serr = SKB_EXT_ERR(skb);
283 	serr->ee.ee_errno = err;
284 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
285 	serr->ee.ee_type = 0;
286 	serr->ee.ee_code = 0;
287 	serr->ee.ee_pad = 0;
288 	serr->ee.ee_info = info;
289 	serr->ee.ee_data = 0;
290 	serr->addr_offset = (u8*)&iph->daddr - skb->nh.raw;
291 	serr->port = port;
292 
293 	skb->h.raw = skb->tail;
294 	__skb_pull(skb, skb->tail - skb->data);
295 
296 	if (sock_queue_err_skb(sk, skb))
297 		kfree_skb(skb);
298 }
299 
300 /*
301  *	Handle MSG_ERRQUEUE
302  */
303 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
304 {
305 	struct sock_exterr_skb *serr;
306 	struct sk_buff *skb, *skb2;
307 	struct sockaddr_in *sin;
308 	struct {
309 		struct sock_extended_err ee;
310 		struct sockaddr_in	 offender;
311 	} errhdr;
312 	int err;
313 	int copied;
314 
315 	err = -EAGAIN;
316 	skb = skb_dequeue(&sk->sk_error_queue);
317 	if (skb == NULL)
318 		goto out;
319 
320 	copied = skb->len;
321 	if (copied > len) {
322 		msg->msg_flags |= MSG_TRUNC;
323 		copied = len;
324 	}
325 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
326 	if (err)
327 		goto out_free_skb;
328 
329 	sock_recv_timestamp(msg, sk, skb);
330 
331 	serr = SKB_EXT_ERR(skb);
332 
333 	sin = (struct sockaddr_in *)msg->msg_name;
334 	if (sin) {
335 		sin->sin_family = AF_INET;
336 		sin->sin_addr.s_addr = *(u32*)(skb->nh.raw + serr->addr_offset);
337 		sin->sin_port = serr->port;
338 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
339 	}
340 
341 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
342 	sin = &errhdr.offender;
343 	sin->sin_family = AF_UNSPEC;
344 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
345 		struct inet_sock *inet = inet_sk(sk);
346 
347 		sin->sin_family = AF_INET;
348 		sin->sin_addr.s_addr = skb->nh.iph->saddr;
349 		sin->sin_port = 0;
350 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
351 		if (inet->cmsg_flags)
352 			ip_cmsg_recv(msg, skb);
353 	}
354 
355 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
356 
357 	/* Now we could try to dump offended packet options */
358 
359 	msg->msg_flags |= MSG_ERRQUEUE;
360 	err = copied;
361 
362 	/* Reset and regenerate socket error */
363 	spin_lock_bh(&sk->sk_error_queue.lock);
364 	sk->sk_err = 0;
365 	if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
366 		sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
367 		spin_unlock_bh(&sk->sk_error_queue.lock);
368 		sk->sk_error_report(sk);
369 	} else
370 		spin_unlock_bh(&sk->sk_error_queue.lock);
371 
372 out_free_skb:
373 	kfree_skb(skb);
374 out:
375 	return err;
376 }
377 
378 
379 /*
380  *	Socket option code for IP. This is the end of the line after any TCP,UDP etc options on
381  *	an IP socket.
382  */
383 
384 int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval, int optlen)
385 {
386 	struct inet_sock *inet = inet_sk(sk);
387 	int val=0,err;
388 
389 	if (level != SOL_IP)
390 		return -ENOPROTOOPT;
391 
392 	if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
393 			    (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
394 			    (1<<IP_RETOPTS) | (1<<IP_TOS) |
395 			    (1<<IP_TTL) | (1<<IP_HDRINCL) |
396 			    (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
397 			    (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND))) ||
398 				optname == IP_MULTICAST_TTL ||
399 				optname == IP_MULTICAST_LOOP) {
400 		if (optlen >= sizeof(int)) {
401 			if (get_user(val, (int __user *) optval))
402 				return -EFAULT;
403 		} else if (optlen >= sizeof(char)) {
404 			unsigned char ucval;
405 
406 			if (get_user(ucval, (unsigned char __user *) optval))
407 				return -EFAULT;
408 			val = (int) ucval;
409 		}
410 	}
411 
412 	/* If optlen==0, it is equivalent to val == 0 */
413 
414 #ifdef CONFIG_IP_MROUTE
415 	if (optname >= MRT_BASE && optname <= (MRT_BASE + 10))
416 		return ip_mroute_setsockopt(sk,optname,optval,optlen);
417 #endif
418 
419 	err = 0;
420 	lock_sock(sk);
421 
422 	switch (optname) {
423 		case IP_OPTIONS:
424 		{
425 			struct ip_options * opt = NULL;
426 			if (optlen > 40 || optlen < 0)
427 				goto e_inval;
428 			err = ip_options_get_from_user(&opt, optval, optlen);
429 			if (err)
430 				break;
431 			if (sk->sk_type == SOCK_STREAM) {
432 				struct tcp_sock *tp = tcp_sk(sk);
433 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
434 				if (sk->sk_family == PF_INET ||
435 				    (!((1 << sk->sk_state) &
436 				       (TCPF_LISTEN | TCPF_CLOSE)) &&
437 				     inet->daddr != LOOPBACK4_IPV6)) {
438 #endif
439 					if (inet->opt)
440 						tp->ext_header_len -= inet->opt->optlen;
441 					if (opt)
442 						tp->ext_header_len += opt->optlen;
443 					tcp_sync_mss(sk, tp->pmtu_cookie);
444 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
445 				}
446 #endif
447 			}
448 			opt = xchg(&inet->opt, opt);
449 			if (opt)
450 				kfree(opt);
451 			break;
452 		}
453 		case IP_PKTINFO:
454 			if (val)
455 				inet->cmsg_flags |= IP_CMSG_PKTINFO;
456 			else
457 				inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
458 			break;
459 		case IP_RECVTTL:
460 			if (val)
461 				inet->cmsg_flags |=  IP_CMSG_TTL;
462 			else
463 				inet->cmsg_flags &= ~IP_CMSG_TTL;
464 			break;
465 		case IP_RECVTOS:
466 			if (val)
467 				inet->cmsg_flags |=  IP_CMSG_TOS;
468 			else
469 				inet->cmsg_flags &= ~IP_CMSG_TOS;
470 			break;
471 		case IP_RECVOPTS:
472 			if (val)
473 				inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
474 			else
475 				inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
476 			break;
477 		case IP_RETOPTS:
478 			if (val)
479 				inet->cmsg_flags |= IP_CMSG_RETOPTS;
480 			else
481 				inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
482 			break;
483 		case IP_TOS:	/* This sets both TOS and Precedence */
484 			if (sk->sk_type == SOCK_STREAM) {
485 				val &= ~3;
486 				val |= inet->tos & 3;
487 			}
488 			if (IPTOS_PREC(val) >= IPTOS_PREC_CRITIC_ECP &&
489 			    !capable(CAP_NET_ADMIN)) {
490 				err = -EPERM;
491 				break;
492 			}
493 			if (inet->tos != val) {
494 				inet->tos = val;
495 				sk->sk_priority = rt_tos2priority(val);
496 				sk_dst_reset(sk);
497 			}
498 			break;
499 		case IP_TTL:
500 			if (optlen<1)
501 				goto e_inval;
502 			if (val != -1 && (val < 1 || val>255))
503 				goto e_inval;
504 			inet->uc_ttl = val;
505 			break;
506 		case IP_HDRINCL:
507 			if (sk->sk_type != SOCK_RAW) {
508 				err = -ENOPROTOOPT;
509 				break;
510 			}
511 			inet->hdrincl = val ? 1 : 0;
512 			break;
513 		case IP_MTU_DISCOVER:
514 			if (val<0 || val>2)
515 				goto e_inval;
516 			inet->pmtudisc = val;
517 			break;
518 		case IP_RECVERR:
519 			inet->recverr = !!val;
520 			if (!val)
521 				skb_queue_purge(&sk->sk_error_queue);
522 			break;
523 		case IP_MULTICAST_TTL:
524 			if (sk->sk_type == SOCK_STREAM)
525 				goto e_inval;
526 			if (optlen<1)
527 				goto e_inval;
528 			if (val==-1)
529 				val = 1;
530 			if (val < 0 || val > 255)
531 				goto e_inval;
532 			inet->mc_ttl = val;
533 	                break;
534 		case IP_MULTICAST_LOOP:
535 			if (optlen<1)
536 				goto e_inval;
537 			inet->mc_loop = !!val;
538 	                break;
539 		case IP_MULTICAST_IF:
540 		{
541 			struct ip_mreqn mreq;
542 			struct net_device *dev = NULL;
543 
544 			if (sk->sk_type == SOCK_STREAM)
545 				goto e_inval;
546 			/*
547 			 *	Check the arguments are allowable
548 			 */
549 
550 			err = -EFAULT;
551 			if (optlen >= sizeof(struct ip_mreqn)) {
552 				if (copy_from_user(&mreq,optval,sizeof(mreq)))
553 					break;
554 			} else {
555 				memset(&mreq, 0, sizeof(mreq));
556 				if (optlen >= sizeof(struct in_addr) &&
557 				    copy_from_user(&mreq.imr_address,optval,sizeof(struct in_addr)))
558 					break;
559 			}
560 
561 			if (!mreq.imr_ifindex) {
562 				if (mreq.imr_address.s_addr == INADDR_ANY) {
563 					inet->mc_index = 0;
564 					inet->mc_addr  = 0;
565 					err = 0;
566 					break;
567 				}
568 				dev = ip_dev_find(mreq.imr_address.s_addr);
569 				if (dev) {
570 					mreq.imr_ifindex = dev->ifindex;
571 					dev_put(dev);
572 				}
573 			} else
574 				dev = __dev_get_by_index(mreq.imr_ifindex);
575 
576 
577 			err = -EADDRNOTAVAIL;
578 			if (!dev)
579 				break;
580 
581 			err = -EINVAL;
582 			if (sk->sk_bound_dev_if &&
583 			    mreq.imr_ifindex != sk->sk_bound_dev_if)
584 				break;
585 
586 			inet->mc_index = mreq.imr_ifindex;
587 			inet->mc_addr  = mreq.imr_address.s_addr;
588 			err = 0;
589 			break;
590 		}
591 
592 		case IP_ADD_MEMBERSHIP:
593 		case IP_DROP_MEMBERSHIP:
594 		{
595 			struct ip_mreqn mreq;
596 
597 			if (optlen < sizeof(struct ip_mreq))
598 				goto e_inval;
599 			err = -EFAULT;
600 			if (optlen >= sizeof(struct ip_mreqn)) {
601 				if(copy_from_user(&mreq,optval,sizeof(mreq)))
602 					break;
603 			} else {
604 				memset(&mreq, 0, sizeof(mreq));
605 				if (copy_from_user(&mreq,optval,sizeof(struct ip_mreq)))
606 					break;
607 			}
608 
609 			if (optname == IP_ADD_MEMBERSHIP)
610 				err = ip_mc_join_group(sk, &mreq);
611 			else
612 				err = ip_mc_leave_group(sk, &mreq);
613 			break;
614 		}
615 		case IP_MSFILTER:
616 		{
617 			extern int sysctl_igmp_max_msf;
618 			struct ip_msfilter *msf;
619 
620 			if (optlen < IP_MSFILTER_SIZE(0))
621 				goto e_inval;
622 			if (optlen > sysctl_optmem_max) {
623 				err = -ENOBUFS;
624 				break;
625 			}
626 			msf = (struct ip_msfilter *)kmalloc(optlen, GFP_KERNEL);
627 			if (msf == 0) {
628 				err = -ENOBUFS;
629 				break;
630 			}
631 			err = -EFAULT;
632 			if (copy_from_user(msf, optval, optlen)) {
633 				kfree(msf);
634 				break;
635 			}
636 			/* numsrc >= (1G-4) overflow in 32 bits */
637 			if (msf->imsf_numsrc >= 0x3ffffffcU ||
638 			    msf->imsf_numsrc > sysctl_igmp_max_msf) {
639 				kfree(msf);
640 				err = -ENOBUFS;
641 				break;
642 			}
643 			if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
644 				kfree(msf);
645 				err = -EINVAL;
646 				break;
647 			}
648 			err = ip_mc_msfilter(sk, msf, 0);
649 			kfree(msf);
650 			break;
651 		}
652 		case IP_BLOCK_SOURCE:
653 		case IP_UNBLOCK_SOURCE:
654 		case IP_ADD_SOURCE_MEMBERSHIP:
655 		case IP_DROP_SOURCE_MEMBERSHIP:
656 		{
657 			struct ip_mreq_source mreqs;
658 			int omode, add;
659 
660 			if (optlen != sizeof(struct ip_mreq_source))
661 				goto e_inval;
662 			if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
663 				err = -EFAULT;
664 				break;
665 			}
666 			if (optname == IP_BLOCK_SOURCE) {
667 				omode = MCAST_EXCLUDE;
668 				add = 1;
669 			} else if (optname == IP_UNBLOCK_SOURCE) {
670 				omode = MCAST_EXCLUDE;
671 				add = 0;
672 			} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
673 				struct ip_mreqn mreq;
674 
675 				mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
676 				mreq.imr_address.s_addr = mreqs.imr_interface;
677 				mreq.imr_ifindex = 0;
678 				err = ip_mc_join_group(sk, &mreq);
679 				if (err && err != -EADDRINUSE)
680 					break;
681 				omode = MCAST_INCLUDE;
682 				add = 1;
683 			} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
684 				omode = MCAST_INCLUDE;
685 				add = 0;
686 			}
687 			err = ip_mc_source(add, omode, sk, &mreqs, 0);
688 			break;
689 		}
690 		case MCAST_JOIN_GROUP:
691 		case MCAST_LEAVE_GROUP:
692 		{
693 			struct group_req greq;
694 			struct sockaddr_in *psin;
695 			struct ip_mreqn mreq;
696 
697 			if (optlen < sizeof(struct group_req))
698 				goto e_inval;
699 			err = -EFAULT;
700 			if(copy_from_user(&greq, optval, sizeof(greq)))
701 				break;
702 			psin = (struct sockaddr_in *)&greq.gr_group;
703 			if (psin->sin_family != AF_INET)
704 				goto e_inval;
705 			memset(&mreq, 0, sizeof(mreq));
706 			mreq.imr_multiaddr = psin->sin_addr;
707 			mreq.imr_ifindex = greq.gr_interface;
708 
709 			if (optname == MCAST_JOIN_GROUP)
710 				err = ip_mc_join_group(sk, &mreq);
711 			else
712 				err = ip_mc_leave_group(sk, &mreq);
713 			break;
714 		}
715 		case MCAST_JOIN_SOURCE_GROUP:
716 		case MCAST_LEAVE_SOURCE_GROUP:
717 		case MCAST_BLOCK_SOURCE:
718 		case MCAST_UNBLOCK_SOURCE:
719 		{
720 			struct group_source_req greqs;
721 			struct ip_mreq_source mreqs;
722 			struct sockaddr_in *psin;
723 			int omode, add;
724 
725 			if (optlen != sizeof(struct group_source_req))
726 				goto e_inval;
727 			if (copy_from_user(&greqs, optval, sizeof(greqs))) {
728 				err = -EFAULT;
729 				break;
730 			}
731 			if (greqs.gsr_group.ss_family != AF_INET ||
732 			    greqs.gsr_source.ss_family != AF_INET) {
733 				err = -EADDRNOTAVAIL;
734 				break;
735 			}
736 			psin = (struct sockaddr_in *)&greqs.gsr_group;
737 			mreqs.imr_multiaddr = psin->sin_addr.s_addr;
738 			psin = (struct sockaddr_in *)&greqs.gsr_source;
739 			mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
740 			mreqs.imr_interface = 0; /* use index for mc_source */
741 
742 			if (optname == MCAST_BLOCK_SOURCE) {
743 				omode = MCAST_EXCLUDE;
744 				add = 1;
745 			} else if (optname == MCAST_UNBLOCK_SOURCE) {
746 				omode = MCAST_EXCLUDE;
747 				add = 0;
748 			} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
749 				struct ip_mreqn mreq;
750 
751 				psin = (struct sockaddr_in *)&greqs.gsr_group;
752 				mreq.imr_multiaddr = psin->sin_addr;
753 				mreq.imr_address.s_addr = 0;
754 				mreq.imr_ifindex = greqs.gsr_interface;
755 				err = ip_mc_join_group(sk, &mreq);
756 				if (err && err != -EADDRINUSE)
757 					break;
758 				greqs.gsr_interface = mreq.imr_ifindex;
759 				omode = MCAST_INCLUDE;
760 				add = 1;
761 			} else /* MCAST_LEAVE_SOURCE_GROUP */ {
762 				omode = MCAST_INCLUDE;
763 				add = 0;
764 			}
765 			err = ip_mc_source(add, omode, sk, &mreqs,
766 				greqs.gsr_interface);
767 			break;
768 		}
769 		case MCAST_MSFILTER:
770 		{
771 			extern int sysctl_igmp_max_msf;
772 			struct sockaddr_in *psin;
773 			struct ip_msfilter *msf = NULL;
774 			struct group_filter *gsf = NULL;
775 			int msize, i, ifindex;
776 
777 			if (optlen < GROUP_FILTER_SIZE(0))
778 				goto e_inval;
779 			if (optlen > sysctl_optmem_max) {
780 				err = -ENOBUFS;
781 				break;
782 			}
783 			gsf = (struct group_filter *)kmalloc(optlen,GFP_KERNEL);
784 			if (gsf == 0) {
785 				err = -ENOBUFS;
786 				break;
787 			}
788 			err = -EFAULT;
789 			if (copy_from_user(gsf, optval, optlen)) {
790 				goto mc_msf_out;
791 			}
792 			/* numsrc >= (4G-140)/128 overflow in 32 bits */
793 			if (gsf->gf_numsrc >= 0x1ffffff ||
794 			    gsf->gf_numsrc > sysctl_igmp_max_msf) {
795 				err = -ENOBUFS;
796 				goto mc_msf_out;
797 			}
798 			if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
799 				err = -EINVAL;
800 				goto mc_msf_out;
801 			}
802 			msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
803 			msf = (struct ip_msfilter *)kmalloc(msize,GFP_KERNEL);
804 			if (msf == 0) {
805 				err = -ENOBUFS;
806 				goto mc_msf_out;
807 			}
808 			ifindex = gsf->gf_interface;
809 			psin = (struct sockaddr_in *)&gsf->gf_group;
810 			if (psin->sin_family != AF_INET) {
811 				err = -EADDRNOTAVAIL;
812 				goto mc_msf_out;
813 			}
814 			msf->imsf_multiaddr = psin->sin_addr.s_addr;
815 			msf->imsf_interface = 0;
816 			msf->imsf_fmode = gsf->gf_fmode;
817 			msf->imsf_numsrc = gsf->gf_numsrc;
818 			err = -EADDRNOTAVAIL;
819 			for (i=0; i<gsf->gf_numsrc; ++i) {
820 				psin = (struct sockaddr_in *)&gsf->gf_slist[i];
821 
822 				if (psin->sin_family != AF_INET)
823 					goto mc_msf_out;
824 				msf->imsf_slist[i] = psin->sin_addr.s_addr;
825 			}
826 			kfree(gsf);
827 			gsf = NULL;
828 
829 			err = ip_mc_msfilter(sk, msf, ifindex);
830 mc_msf_out:
831 			if (msf)
832 				kfree(msf);
833 			if (gsf)
834 				kfree(gsf);
835 			break;
836 		}
837 		case IP_ROUTER_ALERT:
838 			err = ip_ra_control(sk, val ? 1 : 0, NULL);
839 			break;
840 
841 		case IP_FREEBIND:
842 			if (optlen<1)
843 				goto e_inval;
844 			inet->freebind = !!val;
845 	                break;
846 
847 		case IP_IPSEC_POLICY:
848 		case IP_XFRM_POLICY:
849 			err = -EPERM;
850 			if (!capable(CAP_NET_ADMIN))
851 				break;
852 			err = xfrm_user_policy(sk, optname, optval, optlen);
853 			break;
854 
855 		default:
856 #ifdef CONFIG_NETFILTER
857 			err = nf_setsockopt(sk, PF_INET, optname, optval,
858 					    optlen);
859 #else
860 			err = -ENOPROTOOPT;
861 #endif
862 			break;
863 	}
864 	release_sock(sk);
865 	return err;
866 
867 e_inval:
868 	release_sock(sk);
869 	return -EINVAL;
870 }
871 
872 /*
873  *	Get the options. Note for future reference. The GET of IP options gets the
874  *	_received_ ones. The set sets the _sent_ ones.
875  */
876 
877 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval, int __user *optlen)
878 {
879 	struct inet_sock *inet = inet_sk(sk);
880 	int val;
881 	int len;
882 
883 	if(level!=SOL_IP)
884 		return -EOPNOTSUPP;
885 
886 #ifdef CONFIG_IP_MROUTE
887 	if(optname>=MRT_BASE && optname <=MRT_BASE+10)
888 	{
889 		return ip_mroute_getsockopt(sk,optname,optval,optlen);
890 	}
891 #endif
892 
893 	if(get_user(len,optlen))
894 		return -EFAULT;
895 	if(len < 0)
896 		return -EINVAL;
897 
898 	lock_sock(sk);
899 
900 	switch(optname)	{
901 		case IP_OPTIONS:
902 			{
903 				unsigned char optbuf[sizeof(struct ip_options)+40];
904 				struct ip_options * opt = (struct ip_options*)optbuf;
905 				opt->optlen = 0;
906 				if (inet->opt)
907 					memcpy(optbuf, inet->opt,
908 					       sizeof(struct ip_options)+
909 					       inet->opt->optlen);
910 				release_sock(sk);
911 
912 				if (opt->optlen == 0)
913 					return put_user(0, optlen);
914 
915 				ip_options_undo(opt);
916 
917 				len = min_t(unsigned int, len, opt->optlen);
918 				if(put_user(len, optlen))
919 					return -EFAULT;
920 				if(copy_to_user(optval, opt->__data, len))
921 					return -EFAULT;
922 				return 0;
923 			}
924 		case IP_PKTINFO:
925 			val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
926 			break;
927 		case IP_RECVTTL:
928 			val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
929 			break;
930 		case IP_RECVTOS:
931 			val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
932 			break;
933 		case IP_RECVOPTS:
934 			val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
935 			break;
936 		case IP_RETOPTS:
937 			val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
938 			break;
939 		case IP_TOS:
940 			val = inet->tos;
941 			break;
942 		case IP_TTL:
943 			val = (inet->uc_ttl == -1 ?
944 			       sysctl_ip_default_ttl :
945 			       inet->uc_ttl);
946 			break;
947 		case IP_HDRINCL:
948 			val = inet->hdrincl;
949 			break;
950 		case IP_MTU_DISCOVER:
951 			val = inet->pmtudisc;
952 			break;
953 		case IP_MTU:
954 		{
955 			struct dst_entry *dst;
956 			val = 0;
957 			dst = sk_dst_get(sk);
958 			if (dst) {
959 				val = dst_mtu(dst);
960 				dst_release(dst);
961 			}
962 			if (!val) {
963 				release_sock(sk);
964 				return -ENOTCONN;
965 			}
966 			break;
967 		}
968 		case IP_RECVERR:
969 			val = inet->recverr;
970 			break;
971 		case IP_MULTICAST_TTL:
972 			val = inet->mc_ttl;
973 			break;
974 		case IP_MULTICAST_LOOP:
975 			val = inet->mc_loop;
976 			break;
977 		case IP_MULTICAST_IF:
978 		{
979 			struct in_addr addr;
980 			len = min_t(unsigned int, len, sizeof(struct in_addr));
981 			addr.s_addr = inet->mc_addr;
982 			release_sock(sk);
983 
984   			if(put_user(len, optlen))
985   				return -EFAULT;
986 			if(copy_to_user(optval, &addr, len))
987 				return -EFAULT;
988 			return 0;
989 		}
990 		case IP_MSFILTER:
991 		{
992 			struct ip_msfilter msf;
993 			int err;
994 
995 			if (len < IP_MSFILTER_SIZE(0)) {
996 				release_sock(sk);
997 				return -EINVAL;
998 			}
999 			if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1000 				release_sock(sk);
1001 				return -EFAULT;
1002 			}
1003 			err = ip_mc_msfget(sk, &msf,
1004 				(struct ip_msfilter __user *)optval, optlen);
1005 			release_sock(sk);
1006 			return err;
1007 		}
1008 		case MCAST_MSFILTER:
1009 		{
1010 			struct group_filter gsf;
1011 			int err;
1012 
1013 			if (len < GROUP_FILTER_SIZE(0)) {
1014 				release_sock(sk);
1015 				return -EINVAL;
1016 			}
1017 			if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1018 				release_sock(sk);
1019 				return -EFAULT;
1020 			}
1021 			err = ip_mc_gsfget(sk, &gsf,
1022 				(struct group_filter __user *)optval, optlen);
1023 			release_sock(sk);
1024 			return err;
1025 		}
1026 		case IP_PKTOPTIONS:
1027 		{
1028 			struct msghdr msg;
1029 
1030 			release_sock(sk);
1031 
1032 			if (sk->sk_type != SOCK_STREAM)
1033 				return -ENOPROTOOPT;
1034 
1035 			msg.msg_control = optval;
1036 			msg.msg_controllen = len;
1037 			msg.msg_flags = 0;
1038 
1039 			if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1040 				struct in_pktinfo info;
1041 
1042 				info.ipi_addr.s_addr = inet->rcv_saddr;
1043 				info.ipi_spec_dst.s_addr = inet->rcv_saddr;
1044 				info.ipi_ifindex = inet->mc_index;
1045 				put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1046 			}
1047 			if (inet->cmsg_flags & IP_CMSG_TTL) {
1048 				int hlim = inet->mc_ttl;
1049 				put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1050 			}
1051 			len -= msg.msg_controllen;
1052 			return put_user(len, optlen);
1053 		}
1054 		case IP_FREEBIND:
1055 			val = inet->freebind;
1056 			break;
1057 		default:
1058 #ifdef CONFIG_NETFILTER
1059 			val = nf_getsockopt(sk, PF_INET, optname, optval,
1060 					    &len);
1061 			release_sock(sk);
1062 			if (val >= 0)
1063 				val = put_user(len, optlen);
1064 			return val;
1065 #else
1066 			release_sock(sk);
1067 			return -ENOPROTOOPT;
1068 #endif
1069 	}
1070 	release_sock(sk);
1071 
1072 	if (len < sizeof(int) && len > 0 && val>=0 && val<255) {
1073 		unsigned char ucval = (unsigned char)val;
1074 		len = 1;
1075 		if(put_user(len, optlen))
1076 			return -EFAULT;
1077 		if(copy_to_user(optval,&ucval,1))
1078 			return -EFAULT;
1079 	} else {
1080 		len = min_t(unsigned int, sizeof(int), len);
1081 		if(put_user(len, optlen))
1082 			return -EFAULT;
1083 		if(copy_to_user(optval,&val,len))
1084 			return -EFAULT;
1085 	}
1086 	return 0;
1087 }
1088 
1089 EXPORT_SYMBOL(ip_cmsg_recv);
1090 
1091 EXPORT_SYMBOL(ip_getsockopt);
1092 EXPORT_SYMBOL(ip_setsockopt);
1093