xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 7dd65feb)
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  * Authors:	see ip.c
9  *
10  * Fixes:
11  *		Many		:	Split from ip.c , see ip.c for history.
12  *		Martin Mares	:	TOS setting fixed.
13  *		Alan Cox	:	Fixed a couple of oopses in Martin's
14  *					TOS tweaks.
15  *		Mike McLagan	:	Routing by source
16  */
17 
18 #include <linux/module.h>
19 #include <linux/types.h>
20 #include <linux/mm.h>
21 #include <linux/skbuff.h>
22 #include <linux/ip.h>
23 #include <linux/icmp.h>
24 #include <linux/inetdevice.h>
25 #include <linux/netdevice.h>
26 #include <net/sock.h>
27 #include <net/ip.h>
28 #include <net/icmp.h>
29 #include <net/tcp_states.h>
30 #include <linux/udp.h>
31 #include <linux/igmp.h>
32 #include <linux/netfilter.h>
33 #include <linux/route.h>
34 #include <linux/mroute.h>
35 #include <net/route.h>
36 #include <net/xfrm.h>
37 #include <net/compat.h>
38 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
39 #include <net/transp_v6.h>
40 #endif
41 
42 #include <linux/errqueue.h>
43 #include <asm/uaccess.h>
44 
45 #define IP_CMSG_PKTINFO		1
46 #define IP_CMSG_TTL		2
47 #define IP_CMSG_TOS		4
48 #define IP_CMSG_RECVOPTS	8
49 #define IP_CMSG_RETOPTS		16
50 #define IP_CMSG_PASSSEC		32
51 #define IP_CMSG_ORIGDSTADDR     64
52 
53 /*
54  *	SOL_IP control messages.
55  */
56 
57 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
58 {
59 	struct in_pktinfo info;
60 	struct rtable *rt = skb_rtable(skb);
61 
62 	info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
63 	if (rt) {
64 		info.ipi_ifindex = rt->rt_iif;
65 		info.ipi_spec_dst.s_addr = rt->rt_spec_dst;
66 	} else {
67 		info.ipi_ifindex = 0;
68 		info.ipi_spec_dst.s_addr = 0;
69 	}
70 
71 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
72 }
73 
74 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
75 {
76 	int ttl = ip_hdr(skb)->ttl;
77 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
78 }
79 
80 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
81 {
82 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
83 }
84 
85 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
86 {
87 	if (IPCB(skb)->opt.optlen == 0)
88 		return;
89 
90 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
91 		 ip_hdr(skb) + 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 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
113 {
114 	char *secdata;
115 	u32 seclen, secid;
116 	int err;
117 
118 	err = security_socket_getpeersec_dgram(NULL, skb, &secid);
119 	if (err)
120 		return;
121 
122 	err = security_secid_to_secctx(secid, &secdata, &seclen);
123 	if (err)
124 		return;
125 
126 	put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
127 	security_release_secctx(secdata, seclen);
128 }
129 
130 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
131 {
132 	struct sockaddr_in sin;
133 	struct iphdr *iph = ip_hdr(skb);
134 	__be16 *ports = (__be16 *)skb_transport_header(skb);
135 
136 	if (skb_transport_offset(skb) + 4 > skb->len)
137 		return;
138 
139 	/* All current transport protocols have the port numbers in the
140 	 * first four bytes of the transport header and this function is
141 	 * written with this assumption in mind.
142 	 */
143 
144 	sin.sin_family = AF_INET;
145 	sin.sin_addr.s_addr = iph->daddr;
146 	sin.sin_port = ports[1];
147 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
148 
149 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
150 }
151 
152 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
153 {
154 	struct inet_sock *inet = inet_sk(skb->sk);
155 	unsigned flags = inet->cmsg_flags;
156 
157 	/* Ordered by supposed usage frequency */
158 	if (flags & 1)
159 		ip_cmsg_recv_pktinfo(msg, skb);
160 	if ((flags >>= 1) == 0)
161 		return;
162 
163 	if (flags & 1)
164 		ip_cmsg_recv_ttl(msg, skb);
165 	if ((flags >>= 1) == 0)
166 		return;
167 
168 	if (flags & 1)
169 		ip_cmsg_recv_tos(msg, skb);
170 	if ((flags >>= 1) == 0)
171 		return;
172 
173 	if (flags & 1)
174 		ip_cmsg_recv_opts(msg, skb);
175 	if ((flags >>= 1) == 0)
176 		return;
177 
178 	if (flags & 1)
179 		ip_cmsg_recv_retopts(msg, skb);
180 	if ((flags >>= 1) == 0)
181 		return;
182 
183 	if (flags & 1)
184 		ip_cmsg_recv_security(msg, skb);
185 
186 	if ((flags >>= 1) == 0)
187 		return;
188 	if (flags & 1)
189 		ip_cmsg_recv_dstaddr(msg, skb);
190 
191 }
192 EXPORT_SYMBOL(ip_cmsg_recv);
193 
194 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
195 {
196 	int err;
197 	struct cmsghdr *cmsg;
198 
199 	for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
200 		if (!CMSG_OK(msg, cmsg))
201 			return -EINVAL;
202 		if (cmsg->cmsg_level != SOL_IP)
203 			continue;
204 		switch (cmsg->cmsg_type) {
205 		case IP_RETOPTS:
206 			err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
207 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
208 					     err < 40 ? err : 40);
209 			if (err)
210 				return err;
211 			break;
212 		case IP_PKTINFO:
213 		{
214 			struct in_pktinfo *info;
215 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
216 				return -EINVAL;
217 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
218 			ipc->oif = info->ipi_ifindex;
219 			ipc->addr = info->ipi_spec_dst.s_addr;
220 			break;
221 		}
222 		default:
223 			return -EINVAL;
224 		}
225 	}
226 	return 0;
227 }
228 
229 
230 /* Special input handler for packets caught by router alert option.
231    They are selected only by protocol field, and then processed likely
232    local ones; but only if someone wants them! Otherwise, router
233    not running rsvpd will kill RSVP.
234 
235    It is user level problem, what it will make with them.
236    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
237    but receiver should be enough clever f.e. to forward mtrace requests,
238    sent to multicast group to reach destination designated router.
239  */
240 struct ip_ra_chain *ip_ra_chain;
241 DEFINE_RWLOCK(ip_ra_lock);
242 
243 int ip_ra_control(struct sock *sk, unsigned char on,
244 		  void (*destructor)(struct sock *))
245 {
246 	struct ip_ra_chain *ra, *new_ra, **rap;
247 
248 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
249 		return -EINVAL;
250 
251 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
252 
253 	write_lock_bh(&ip_ra_lock);
254 	for (rap = &ip_ra_chain; (ra = *rap) != NULL; rap = &ra->next) {
255 		if (ra->sk == sk) {
256 			if (on) {
257 				write_unlock_bh(&ip_ra_lock);
258 				kfree(new_ra);
259 				return -EADDRINUSE;
260 			}
261 			*rap = ra->next;
262 			write_unlock_bh(&ip_ra_lock);
263 
264 			if (ra->destructor)
265 				ra->destructor(sk);
266 			sock_put(sk);
267 			kfree(ra);
268 			return 0;
269 		}
270 	}
271 	if (new_ra == NULL) {
272 		write_unlock_bh(&ip_ra_lock);
273 		return -ENOBUFS;
274 	}
275 	new_ra->sk = sk;
276 	new_ra->destructor = destructor;
277 
278 	new_ra->next = ra;
279 	*rap = new_ra;
280 	sock_hold(sk);
281 	write_unlock_bh(&ip_ra_lock);
282 
283 	return 0;
284 }
285 
286 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
287 		   __be16 port, u32 info, u8 *payload)
288 {
289 	struct inet_sock *inet = inet_sk(sk);
290 	struct sock_exterr_skb *serr;
291 
292 	if (!inet->recverr)
293 		return;
294 
295 	skb = skb_clone(skb, GFP_ATOMIC);
296 	if (!skb)
297 		return;
298 
299 	serr = SKB_EXT_ERR(skb);
300 	serr->ee.ee_errno = err;
301 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
302 	serr->ee.ee_type = icmp_hdr(skb)->type;
303 	serr->ee.ee_code = icmp_hdr(skb)->code;
304 	serr->ee.ee_pad = 0;
305 	serr->ee.ee_info = info;
306 	serr->ee.ee_data = 0;
307 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
308 				   skb_network_header(skb);
309 	serr->port = port;
310 
311 	if (skb_pull(skb, payload - skb->data) != NULL) {
312 		skb_reset_transport_header(skb);
313 		if (sock_queue_err_skb(sk, skb) == 0)
314 			return;
315 	}
316 	kfree_skb(skb);
317 }
318 
319 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
320 {
321 	struct inet_sock *inet = inet_sk(sk);
322 	struct sock_exterr_skb *serr;
323 	struct iphdr *iph;
324 	struct sk_buff *skb;
325 
326 	if (!inet->recverr)
327 		return;
328 
329 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
330 	if (!skb)
331 		return;
332 
333 	skb_put(skb, sizeof(struct iphdr));
334 	skb_reset_network_header(skb);
335 	iph = ip_hdr(skb);
336 	iph->daddr = daddr;
337 
338 	serr = SKB_EXT_ERR(skb);
339 	serr->ee.ee_errno = err;
340 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
341 	serr->ee.ee_type = 0;
342 	serr->ee.ee_code = 0;
343 	serr->ee.ee_pad = 0;
344 	serr->ee.ee_info = info;
345 	serr->ee.ee_data = 0;
346 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
347 	serr->port = port;
348 
349 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
350 	skb_reset_transport_header(skb);
351 
352 	if (sock_queue_err_skb(sk, skb))
353 		kfree_skb(skb);
354 }
355 
356 /*
357  *	Handle MSG_ERRQUEUE
358  */
359 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
360 {
361 	struct sock_exterr_skb *serr;
362 	struct sk_buff *skb, *skb2;
363 	struct sockaddr_in *sin;
364 	struct {
365 		struct sock_extended_err ee;
366 		struct sockaddr_in	 offender;
367 	} errhdr;
368 	int err;
369 	int copied;
370 
371 	err = -EAGAIN;
372 	skb = skb_dequeue(&sk->sk_error_queue);
373 	if (skb == NULL)
374 		goto out;
375 
376 	copied = skb->len;
377 	if (copied > len) {
378 		msg->msg_flags |= MSG_TRUNC;
379 		copied = len;
380 	}
381 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
382 	if (err)
383 		goto out_free_skb;
384 
385 	sock_recv_timestamp(msg, sk, skb);
386 
387 	serr = SKB_EXT_ERR(skb);
388 
389 	sin = (struct sockaddr_in *)msg->msg_name;
390 	if (sin) {
391 		sin->sin_family = AF_INET;
392 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
393 						   serr->addr_offset);
394 		sin->sin_port = serr->port;
395 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
396 	}
397 
398 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
399 	sin = &errhdr.offender;
400 	sin->sin_family = AF_UNSPEC;
401 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
402 		struct inet_sock *inet = inet_sk(sk);
403 
404 		sin->sin_family = AF_INET;
405 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
406 		sin->sin_port = 0;
407 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
408 		if (inet->cmsg_flags)
409 			ip_cmsg_recv(msg, skb);
410 	}
411 
412 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
413 
414 	/* Now we could try to dump offended packet options */
415 
416 	msg->msg_flags |= MSG_ERRQUEUE;
417 	err = copied;
418 
419 	/* Reset and regenerate socket error */
420 	spin_lock_bh(&sk->sk_error_queue.lock);
421 	sk->sk_err = 0;
422 	skb2 = skb_peek(&sk->sk_error_queue);
423 	if (skb2 != NULL) {
424 		sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
425 		spin_unlock_bh(&sk->sk_error_queue.lock);
426 		sk->sk_error_report(sk);
427 	} else
428 		spin_unlock_bh(&sk->sk_error_queue.lock);
429 
430 out_free_skb:
431 	kfree_skb(skb);
432 out:
433 	return err;
434 }
435 
436 
437 /*
438  *	Socket option code for IP. This is the end of the line after any
439  *	TCP,UDP etc options on an IP socket.
440  */
441 
442 static int do_ip_setsockopt(struct sock *sk, int level,
443 			    int optname, char __user *optval, unsigned int optlen)
444 {
445 	struct inet_sock *inet = inet_sk(sk);
446 	int val = 0, err;
447 
448 	if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
449 			     (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
450 			     (1<<IP_RETOPTS) | (1<<IP_TOS) |
451 			     (1<<IP_TTL) | (1<<IP_HDRINCL) |
452 			     (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
453 			     (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) |
454 			     (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT))) ||
455 	    optname == IP_MULTICAST_TTL ||
456 	    optname == IP_MULTICAST_ALL ||
457 	    optname == IP_MULTICAST_LOOP ||
458 	    optname == IP_RECVORIGDSTADDR) {
459 		if (optlen >= sizeof(int)) {
460 			if (get_user(val, (int __user *) optval))
461 				return -EFAULT;
462 		} else if (optlen >= sizeof(char)) {
463 			unsigned char ucval;
464 
465 			if (get_user(ucval, (unsigned char __user *) optval))
466 				return -EFAULT;
467 			val = (int) ucval;
468 		}
469 	}
470 
471 	/* If optlen==0, it is equivalent to val == 0 */
472 
473 	if (ip_mroute_opt(optname))
474 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
475 
476 	err = 0;
477 	lock_sock(sk);
478 
479 	switch (optname) {
480 	case IP_OPTIONS:
481 	{
482 		struct ip_options *opt = NULL;
483 		if (optlen > 40)
484 			goto e_inval;
485 		err = ip_options_get_from_user(sock_net(sk), &opt,
486 					       optval, optlen);
487 		if (err)
488 			break;
489 		if (inet->is_icsk) {
490 			struct inet_connection_sock *icsk = inet_csk(sk);
491 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
492 			if (sk->sk_family == PF_INET ||
493 			    (!((1 << sk->sk_state) &
494 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
495 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
496 #endif
497 				if (inet->opt)
498 					icsk->icsk_ext_hdr_len -= inet->opt->optlen;
499 				if (opt)
500 					icsk->icsk_ext_hdr_len += opt->optlen;
501 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
502 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
503 			}
504 #endif
505 		}
506 		opt = xchg(&inet->opt, opt);
507 		kfree(opt);
508 		break;
509 	}
510 	case IP_PKTINFO:
511 		if (val)
512 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
513 		else
514 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
515 		break;
516 	case IP_RECVTTL:
517 		if (val)
518 			inet->cmsg_flags |=  IP_CMSG_TTL;
519 		else
520 			inet->cmsg_flags &= ~IP_CMSG_TTL;
521 		break;
522 	case IP_RECVTOS:
523 		if (val)
524 			inet->cmsg_flags |=  IP_CMSG_TOS;
525 		else
526 			inet->cmsg_flags &= ~IP_CMSG_TOS;
527 		break;
528 	case IP_RECVOPTS:
529 		if (val)
530 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
531 		else
532 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
533 		break;
534 	case IP_RETOPTS:
535 		if (val)
536 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
537 		else
538 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
539 		break;
540 	case IP_PASSSEC:
541 		if (val)
542 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
543 		else
544 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
545 		break;
546 	case IP_RECVORIGDSTADDR:
547 		if (val)
548 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
549 		else
550 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
551 		break;
552 	case IP_TOS:	/* This sets both TOS and Precedence */
553 		if (sk->sk_type == SOCK_STREAM) {
554 			val &= ~3;
555 			val |= inet->tos & 3;
556 		}
557 		if (inet->tos != val) {
558 			inet->tos = val;
559 			sk->sk_priority = rt_tos2priority(val);
560 			sk_dst_reset(sk);
561 		}
562 		break;
563 	case IP_TTL:
564 		if (optlen < 1)
565 			goto e_inval;
566 		if (val != -1 && (val < 0 || val > 255))
567 			goto e_inval;
568 		inet->uc_ttl = val;
569 		break;
570 	case IP_HDRINCL:
571 		if (sk->sk_type != SOCK_RAW) {
572 			err = -ENOPROTOOPT;
573 			break;
574 		}
575 		inet->hdrincl = val ? 1 : 0;
576 		break;
577 	case IP_MTU_DISCOVER:
578 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
579 			goto e_inval;
580 		inet->pmtudisc = val;
581 		break;
582 	case IP_RECVERR:
583 		inet->recverr = !!val;
584 		if (!val)
585 			skb_queue_purge(&sk->sk_error_queue);
586 		break;
587 	case IP_MULTICAST_TTL:
588 		if (sk->sk_type == SOCK_STREAM)
589 			goto e_inval;
590 		if (optlen < 1)
591 			goto e_inval;
592 		if (val == -1)
593 			val = 1;
594 		if (val < 0 || val > 255)
595 			goto e_inval;
596 		inet->mc_ttl = val;
597 		break;
598 	case IP_MULTICAST_LOOP:
599 		if (optlen < 1)
600 			goto e_inval;
601 		inet->mc_loop = !!val;
602 		break;
603 	case IP_MULTICAST_IF:
604 	{
605 		struct ip_mreqn mreq;
606 		struct net_device *dev = NULL;
607 
608 		if (sk->sk_type == SOCK_STREAM)
609 			goto e_inval;
610 		/*
611 		 *	Check the arguments are allowable
612 		 */
613 
614 		if (optlen < sizeof(struct in_addr))
615 			goto e_inval;
616 
617 		err = -EFAULT;
618 		if (optlen >= sizeof(struct ip_mreqn)) {
619 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
620 				break;
621 		} else {
622 			memset(&mreq, 0, sizeof(mreq));
623 			if (optlen >= sizeof(struct in_addr) &&
624 			    copy_from_user(&mreq.imr_address, optval,
625 					   sizeof(struct in_addr)))
626 				break;
627 		}
628 
629 		if (!mreq.imr_ifindex) {
630 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
631 				inet->mc_index = 0;
632 				inet->mc_addr  = 0;
633 				err = 0;
634 				break;
635 			}
636 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
637 			if (dev)
638 				mreq.imr_ifindex = dev->ifindex;
639 		} else
640 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
641 
642 
643 		err = -EADDRNOTAVAIL;
644 		if (!dev)
645 			break;
646 		dev_put(dev);
647 
648 		err = -EINVAL;
649 		if (sk->sk_bound_dev_if &&
650 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
651 			break;
652 
653 		inet->mc_index = mreq.imr_ifindex;
654 		inet->mc_addr  = mreq.imr_address.s_addr;
655 		err = 0;
656 		break;
657 	}
658 
659 	case IP_ADD_MEMBERSHIP:
660 	case IP_DROP_MEMBERSHIP:
661 	{
662 		struct ip_mreqn mreq;
663 
664 		err = -EPROTO;
665 		if (inet_sk(sk)->is_icsk)
666 			break;
667 
668 		if (optlen < sizeof(struct ip_mreq))
669 			goto e_inval;
670 		err = -EFAULT;
671 		if (optlen >= sizeof(struct ip_mreqn)) {
672 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
673 				break;
674 		} else {
675 			memset(&mreq, 0, sizeof(mreq));
676 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
677 				break;
678 		}
679 
680 		if (optname == IP_ADD_MEMBERSHIP)
681 			err = ip_mc_join_group(sk, &mreq);
682 		else
683 			err = ip_mc_leave_group(sk, &mreq);
684 		break;
685 	}
686 	case IP_MSFILTER:
687 	{
688 		struct ip_msfilter *msf;
689 
690 		if (optlen < IP_MSFILTER_SIZE(0))
691 			goto e_inval;
692 		if (optlen > sysctl_optmem_max) {
693 			err = -ENOBUFS;
694 			break;
695 		}
696 		msf = kmalloc(optlen, GFP_KERNEL);
697 		if (!msf) {
698 			err = -ENOBUFS;
699 			break;
700 		}
701 		err = -EFAULT;
702 		if (copy_from_user(msf, optval, optlen)) {
703 			kfree(msf);
704 			break;
705 		}
706 		/* numsrc >= (1G-4) overflow in 32 bits */
707 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
708 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
709 			kfree(msf);
710 			err = -ENOBUFS;
711 			break;
712 		}
713 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
714 			kfree(msf);
715 			err = -EINVAL;
716 			break;
717 		}
718 		err = ip_mc_msfilter(sk, msf, 0);
719 		kfree(msf);
720 		break;
721 	}
722 	case IP_BLOCK_SOURCE:
723 	case IP_UNBLOCK_SOURCE:
724 	case IP_ADD_SOURCE_MEMBERSHIP:
725 	case IP_DROP_SOURCE_MEMBERSHIP:
726 	{
727 		struct ip_mreq_source mreqs;
728 		int omode, add;
729 
730 		if (optlen != sizeof(struct ip_mreq_source))
731 			goto e_inval;
732 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
733 			err = -EFAULT;
734 			break;
735 		}
736 		if (optname == IP_BLOCK_SOURCE) {
737 			omode = MCAST_EXCLUDE;
738 			add = 1;
739 		} else if (optname == IP_UNBLOCK_SOURCE) {
740 			omode = MCAST_EXCLUDE;
741 			add = 0;
742 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
743 			struct ip_mreqn mreq;
744 
745 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
746 			mreq.imr_address.s_addr = mreqs.imr_interface;
747 			mreq.imr_ifindex = 0;
748 			err = ip_mc_join_group(sk, &mreq);
749 			if (err && err != -EADDRINUSE)
750 				break;
751 			omode = MCAST_INCLUDE;
752 			add = 1;
753 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
754 			omode = MCAST_INCLUDE;
755 			add = 0;
756 		}
757 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
758 		break;
759 	}
760 	case MCAST_JOIN_GROUP:
761 	case MCAST_LEAVE_GROUP:
762 	{
763 		struct group_req greq;
764 		struct sockaddr_in *psin;
765 		struct ip_mreqn mreq;
766 
767 		if (optlen < sizeof(struct group_req))
768 			goto e_inval;
769 		err = -EFAULT;
770 		if (copy_from_user(&greq, optval, sizeof(greq)))
771 			break;
772 		psin = (struct sockaddr_in *)&greq.gr_group;
773 		if (psin->sin_family != AF_INET)
774 			goto e_inval;
775 		memset(&mreq, 0, sizeof(mreq));
776 		mreq.imr_multiaddr = psin->sin_addr;
777 		mreq.imr_ifindex = greq.gr_interface;
778 
779 		if (optname == MCAST_JOIN_GROUP)
780 			err = ip_mc_join_group(sk, &mreq);
781 		else
782 			err = ip_mc_leave_group(sk, &mreq);
783 		break;
784 	}
785 	case MCAST_JOIN_SOURCE_GROUP:
786 	case MCAST_LEAVE_SOURCE_GROUP:
787 	case MCAST_BLOCK_SOURCE:
788 	case MCAST_UNBLOCK_SOURCE:
789 	{
790 		struct group_source_req greqs;
791 		struct ip_mreq_source mreqs;
792 		struct sockaddr_in *psin;
793 		int omode, add;
794 
795 		if (optlen != sizeof(struct group_source_req))
796 			goto e_inval;
797 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
798 			err = -EFAULT;
799 			break;
800 		}
801 		if (greqs.gsr_group.ss_family != AF_INET ||
802 		    greqs.gsr_source.ss_family != AF_INET) {
803 			err = -EADDRNOTAVAIL;
804 			break;
805 		}
806 		psin = (struct sockaddr_in *)&greqs.gsr_group;
807 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
808 		psin = (struct sockaddr_in *)&greqs.gsr_source;
809 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
810 		mreqs.imr_interface = 0; /* use index for mc_source */
811 
812 		if (optname == MCAST_BLOCK_SOURCE) {
813 			omode = MCAST_EXCLUDE;
814 			add = 1;
815 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
816 			omode = MCAST_EXCLUDE;
817 			add = 0;
818 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
819 			struct ip_mreqn mreq;
820 
821 			psin = (struct sockaddr_in *)&greqs.gsr_group;
822 			mreq.imr_multiaddr = psin->sin_addr;
823 			mreq.imr_address.s_addr = 0;
824 			mreq.imr_ifindex = greqs.gsr_interface;
825 			err = ip_mc_join_group(sk, &mreq);
826 			if (err && err != -EADDRINUSE)
827 				break;
828 			greqs.gsr_interface = mreq.imr_ifindex;
829 			omode = MCAST_INCLUDE;
830 			add = 1;
831 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
832 			omode = MCAST_INCLUDE;
833 			add = 0;
834 		}
835 		err = ip_mc_source(add, omode, sk, &mreqs,
836 				   greqs.gsr_interface);
837 		break;
838 	}
839 	case MCAST_MSFILTER:
840 	{
841 		struct sockaddr_in *psin;
842 		struct ip_msfilter *msf = NULL;
843 		struct group_filter *gsf = NULL;
844 		int msize, i, ifindex;
845 
846 		if (optlen < GROUP_FILTER_SIZE(0))
847 			goto e_inval;
848 		if (optlen > sysctl_optmem_max) {
849 			err = -ENOBUFS;
850 			break;
851 		}
852 		gsf = kmalloc(optlen, GFP_KERNEL);
853 		if (!gsf) {
854 			err = -ENOBUFS;
855 			break;
856 		}
857 		err = -EFAULT;
858 		if (copy_from_user(gsf, optval, optlen))
859 			goto mc_msf_out;
860 
861 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
862 		if (gsf->gf_numsrc >= 0x1ffffff ||
863 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
864 			err = -ENOBUFS;
865 			goto mc_msf_out;
866 		}
867 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
868 			err = -EINVAL;
869 			goto mc_msf_out;
870 		}
871 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
872 		msf = kmalloc(msize, GFP_KERNEL);
873 		if (!msf) {
874 			err = -ENOBUFS;
875 			goto mc_msf_out;
876 		}
877 		ifindex = gsf->gf_interface;
878 		psin = (struct sockaddr_in *)&gsf->gf_group;
879 		if (psin->sin_family != AF_INET) {
880 			err = -EADDRNOTAVAIL;
881 			goto mc_msf_out;
882 		}
883 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
884 		msf->imsf_interface = 0;
885 		msf->imsf_fmode = gsf->gf_fmode;
886 		msf->imsf_numsrc = gsf->gf_numsrc;
887 		err = -EADDRNOTAVAIL;
888 		for (i = 0; i < gsf->gf_numsrc; ++i) {
889 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
890 
891 			if (psin->sin_family != AF_INET)
892 				goto mc_msf_out;
893 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
894 		}
895 		kfree(gsf);
896 		gsf = NULL;
897 
898 		err = ip_mc_msfilter(sk, msf, ifindex);
899 mc_msf_out:
900 		kfree(msf);
901 		kfree(gsf);
902 		break;
903 	}
904 	case IP_MULTICAST_ALL:
905 		if (optlen < 1)
906 			goto e_inval;
907 		if (val != 0 && val != 1)
908 			goto e_inval;
909 		inet->mc_all = val;
910 		break;
911 	case IP_ROUTER_ALERT:
912 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
913 		break;
914 
915 	case IP_FREEBIND:
916 		if (optlen < 1)
917 			goto e_inval;
918 		inet->freebind = !!val;
919 		break;
920 
921 	case IP_IPSEC_POLICY:
922 	case IP_XFRM_POLICY:
923 		err = -EPERM;
924 		if (!capable(CAP_NET_ADMIN))
925 			break;
926 		err = xfrm_user_policy(sk, optname, optval, optlen);
927 		break;
928 
929 	case IP_TRANSPARENT:
930 		if (!capable(CAP_NET_ADMIN)) {
931 			err = -EPERM;
932 			break;
933 		}
934 		if (optlen < 1)
935 			goto e_inval;
936 		inet->transparent = !!val;
937 		break;
938 
939 	default:
940 		err = -ENOPROTOOPT;
941 		break;
942 	}
943 	release_sock(sk);
944 	return err;
945 
946 e_inval:
947 	release_sock(sk);
948 	return -EINVAL;
949 }
950 
951 int ip_setsockopt(struct sock *sk, int level,
952 		int optname, char __user *optval, unsigned int optlen)
953 {
954 	int err;
955 
956 	if (level != SOL_IP)
957 		return -ENOPROTOOPT;
958 
959 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
960 #ifdef CONFIG_NETFILTER
961 	/* we need to exclude all possible ENOPROTOOPTs except default case */
962 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
963 			optname != IP_IPSEC_POLICY &&
964 			optname != IP_XFRM_POLICY &&
965 			!ip_mroute_opt(optname)) {
966 		lock_sock(sk);
967 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
968 		release_sock(sk);
969 	}
970 #endif
971 	return err;
972 }
973 EXPORT_SYMBOL(ip_setsockopt);
974 
975 #ifdef CONFIG_COMPAT
976 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
977 			 char __user *optval, unsigned int optlen)
978 {
979 	int err;
980 
981 	if (level != SOL_IP)
982 		return -ENOPROTOOPT;
983 
984 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
985 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
986 			ip_setsockopt);
987 
988 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
989 #ifdef CONFIG_NETFILTER
990 	/* we need to exclude all possible ENOPROTOOPTs except default case */
991 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
992 			optname != IP_IPSEC_POLICY &&
993 			optname != IP_XFRM_POLICY &&
994 			!ip_mroute_opt(optname)) {
995 		lock_sock(sk);
996 		err = compat_nf_setsockopt(sk, PF_INET, optname,
997 					   optval, optlen);
998 		release_sock(sk);
999 	}
1000 #endif
1001 	return err;
1002 }
1003 EXPORT_SYMBOL(compat_ip_setsockopt);
1004 #endif
1005 
1006 /*
1007  *	Get the options. Note for future reference. The GET of IP options gets
1008  *	the _received_ ones. The set sets the _sent_ ones.
1009  */
1010 
1011 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1012 			    char __user *optval, int __user *optlen)
1013 {
1014 	struct inet_sock *inet = inet_sk(sk);
1015 	int val;
1016 	int len;
1017 
1018 	if (level != SOL_IP)
1019 		return -EOPNOTSUPP;
1020 
1021 	if (ip_mroute_opt(optname))
1022 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1023 
1024 	if (get_user(len, optlen))
1025 		return -EFAULT;
1026 	if (len < 0)
1027 		return -EINVAL;
1028 
1029 	lock_sock(sk);
1030 
1031 	switch (optname) {
1032 	case IP_OPTIONS:
1033 	{
1034 		unsigned char optbuf[sizeof(struct ip_options)+40];
1035 		struct ip_options * opt = (struct ip_options *)optbuf;
1036 		opt->optlen = 0;
1037 		if (inet->opt)
1038 			memcpy(optbuf, inet->opt,
1039 			       sizeof(struct ip_options)+
1040 			       inet->opt->optlen);
1041 		release_sock(sk);
1042 
1043 		if (opt->optlen == 0)
1044 			return put_user(0, optlen);
1045 
1046 		ip_options_undo(opt);
1047 
1048 		len = min_t(unsigned int, len, opt->optlen);
1049 		if (put_user(len, optlen))
1050 			return -EFAULT;
1051 		if (copy_to_user(optval, opt->__data, len))
1052 			return -EFAULT;
1053 		return 0;
1054 	}
1055 	case IP_PKTINFO:
1056 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1057 		break;
1058 	case IP_RECVTTL:
1059 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1060 		break;
1061 	case IP_RECVTOS:
1062 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1063 		break;
1064 	case IP_RECVOPTS:
1065 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1066 		break;
1067 	case IP_RETOPTS:
1068 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1069 		break;
1070 	case IP_PASSSEC:
1071 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1072 		break;
1073 	case IP_RECVORIGDSTADDR:
1074 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1075 		break;
1076 	case IP_TOS:
1077 		val = inet->tos;
1078 		break;
1079 	case IP_TTL:
1080 		val = (inet->uc_ttl == -1 ?
1081 		       sysctl_ip_default_ttl :
1082 		       inet->uc_ttl);
1083 		break;
1084 	case IP_HDRINCL:
1085 		val = inet->hdrincl;
1086 		break;
1087 	case IP_MTU_DISCOVER:
1088 		val = inet->pmtudisc;
1089 		break;
1090 	case IP_MTU:
1091 	{
1092 		struct dst_entry *dst;
1093 		val = 0;
1094 		dst = sk_dst_get(sk);
1095 		if (dst) {
1096 			val = dst_mtu(dst);
1097 			dst_release(dst);
1098 		}
1099 		if (!val) {
1100 			release_sock(sk);
1101 			return -ENOTCONN;
1102 		}
1103 		break;
1104 	}
1105 	case IP_RECVERR:
1106 		val = inet->recverr;
1107 		break;
1108 	case IP_MULTICAST_TTL:
1109 		val = inet->mc_ttl;
1110 		break;
1111 	case IP_MULTICAST_LOOP:
1112 		val = inet->mc_loop;
1113 		break;
1114 	case IP_MULTICAST_IF:
1115 	{
1116 		struct in_addr addr;
1117 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1118 		addr.s_addr = inet->mc_addr;
1119 		release_sock(sk);
1120 
1121 		if (put_user(len, optlen))
1122 			return -EFAULT;
1123 		if (copy_to_user(optval, &addr, len))
1124 			return -EFAULT;
1125 		return 0;
1126 	}
1127 	case IP_MSFILTER:
1128 	{
1129 		struct ip_msfilter msf;
1130 		int err;
1131 
1132 		if (len < IP_MSFILTER_SIZE(0)) {
1133 			release_sock(sk);
1134 			return -EINVAL;
1135 		}
1136 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1137 			release_sock(sk);
1138 			return -EFAULT;
1139 		}
1140 		err = ip_mc_msfget(sk, &msf,
1141 				   (struct ip_msfilter __user *)optval, optlen);
1142 		release_sock(sk);
1143 		return err;
1144 	}
1145 	case MCAST_MSFILTER:
1146 	{
1147 		struct group_filter gsf;
1148 		int err;
1149 
1150 		if (len < GROUP_FILTER_SIZE(0)) {
1151 			release_sock(sk);
1152 			return -EINVAL;
1153 		}
1154 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1155 			release_sock(sk);
1156 			return -EFAULT;
1157 		}
1158 		err = ip_mc_gsfget(sk, &gsf,
1159 				   (struct group_filter __user *)optval,
1160 				   optlen);
1161 		release_sock(sk);
1162 		return err;
1163 	}
1164 	case IP_MULTICAST_ALL:
1165 		val = inet->mc_all;
1166 		break;
1167 	case IP_PKTOPTIONS:
1168 	{
1169 		struct msghdr msg;
1170 
1171 		release_sock(sk);
1172 
1173 		if (sk->sk_type != SOCK_STREAM)
1174 			return -ENOPROTOOPT;
1175 
1176 		msg.msg_control = optval;
1177 		msg.msg_controllen = len;
1178 		msg.msg_flags = 0;
1179 
1180 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1181 			struct in_pktinfo info;
1182 
1183 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1184 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1185 			info.ipi_ifindex = inet->mc_index;
1186 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1187 		}
1188 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1189 			int hlim = inet->mc_ttl;
1190 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1191 		}
1192 		len -= msg.msg_controllen;
1193 		return put_user(len, optlen);
1194 	}
1195 	case IP_FREEBIND:
1196 		val = inet->freebind;
1197 		break;
1198 	case IP_TRANSPARENT:
1199 		val = inet->transparent;
1200 		break;
1201 	default:
1202 		release_sock(sk);
1203 		return -ENOPROTOOPT;
1204 	}
1205 	release_sock(sk);
1206 
1207 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1208 		unsigned char ucval = (unsigned char)val;
1209 		len = 1;
1210 		if (put_user(len, optlen))
1211 			return -EFAULT;
1212 		if (copy_to_user(optval, &ucval, 1))
1213 			return -EFAULT;
1214 	} else {
1215 		len = min_t(unsigned int, sizeof(int), len);
1216 		if (put_user(len, optlen))
1217 			return -EFAULT;
1218 		if (copy_to_user(optval, &val, len))
1219 			return -EFAULT;
1220 	}
1221 	return 0;
1222 }
1223 
1224 int ip_getsockopt(struct sock *sk, int level,
1225 		  int optname, char __user *optval, int __user *optlen)
1226 {
1227 	int err;
1228 
1229 	err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1230 #ifdef CONFIG_NETFILTER
1231 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1232 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1233 			!ip_mroute_opt(optname)) {
1234 		int len;
1235 
1236 		if (get_user(len, optlen))
1237 			return -EFAULT;
1238 
1239 		lock_sock(sk);
1240 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1241 				&len);
1242 		release_sock(sk);
1243 		if (err >= 0)
1244 			err = put_user(len, optlen);
1245 		return err;
1246 	}
1247 #endif
1248 	return err;
1249 }
1250 EXPORT_SYMBOL(ip_getsockopt);
1251 
1252 #ifdef CONFIG_COMPAT
1253 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1254 			 char __user *optval, int __user *optlen)
1255 {
1256 	int err;
1257 
1258 	if (optname == MCAST_MSFILTER)
1259 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1260 			ip_getsockopt);
1261 
1262 	err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1263 
1264 #ifdef CONFIG_NETFILTER
1265 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1266 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1267 			!ip_mroute_opt(optname)) {
1268 		int len;
1269 
1270 		if (get_user(len, optlen))
1271 			return -EFAULT;
1272 
1273 		lock_sock(sk);
1274 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1275 		release_sock(sk);
1276 		if (err >= 0)
1277 			err = put_user(len, optlen);
1278 		return err;
1279 	}
1280 #endif
1281 	return err;
1282 }
1283 EXPORT_SYMBOL(compat_ip_getsockopt);
1284 #endif
1285