xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 609e478b)
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 <linux/slab.h>
27 #include <net/sock.h>
28 #include <net/ip.h>
29 #include <net/icmp.h>
30 #include <net/tcp_states.h>
31 #include <linux/udp.h>
32 #include <linux/igmp.h>
33 #include <linux/netfilter.h>
34 #include <linux/route.h>
35 #include <linux/mroute.h>
36 #include <net/inet_ecn.h>
37 #include <net/route.h>
38 #include <net/xfrm.h>
39 #include <net/compat.h>
40 #if IS_ENABLED(CONFIG_IPV6)
41 #include <net/transp_v6.h>
42 #endif
43 #include <net/ip_fib.h>
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 #define IP_CMSG_PASSSEC		32
54 #define IP_CMSG_ORIGDSTADDR     64
55 
56 /*
57  *	SOL_IP control messages.
58  */
59 
60 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
61 {
62 	struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
63 
64 	info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
65 
66 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
67 }
68 
69 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
70 {
71 	int ttl = ip_hdr(skb)->ttl;
72 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
73 }
74 
75 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
76 {
77 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
78 }
79 
80 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
81 {
82 	if (IPCB(skb)->opt.optlen == 0)
83 		return;
84 
85 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
86 		 ip_hdr(skb) + 1);
87 }
88 
89 
90 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
91 {
92 	unsigned char optbuf[sizeof(struct ip_options) + 40];
93 	struct ip_options *opt = (struct ip_options *)optbuf;
94 
95 	if (IPCB(skb)->opt.optlen == 0)
96 		return;
97 
98 	if (ip_options_echo(opt, skb)) {
99 		msg->msg_flags |= MSG_CTRUNC;
100 		return;
101 	}
102 	ip_options_undo(opt);
103 
104 	put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
105 }
106 
107 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
108 {
109 	char *secdata;
110 	u32 seclen, secid;
111 	int err;
112 
113 	err = security_socket_getpeersec_dgram(NULL, skb, &secid);
114 	if (err)
115 		return;
116 
117 	err = security_secid_to_secctx(secid, &secdata, &seclen);
118 	if (err)
119 		return;
120 
121 	put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
122 	security_release_secctx(secdata, seclen);
123 }
124 
125 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
126 {
127 	struct sockaddr_in sin;
128 	const struct iphdr *iph = ip_hdr(skb);
129 	__be16 *ports = (__be16 *)skb_transport_header(skb);
130 
131 	if (skb_transport_offset(skb) + 4 > skb->len)
132 		return;
133 
134 	/* All current transport protocols have the port numbers in the
135 	 * first four bytes of the transport header and this function is
136 	 * written with this assumption in mind.
137 	 */
138 
139 	sin.sin_family = AF_INET;
140 	sin.sin_addr.s_addr = iph->daddr;
141 	sin.sin_port = ports[1];
142 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
143 
144 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
145 }
146 
147 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
148 {
149 	struct inet_sock *inet = inet_sk(skb->sk);
150 	unsigned int flags = inet->cmsg_flags;
151 
152 	/* Ordered by supposed usage frequency */
153 	if (flags & 1)
154 		ip_cmsg_recv_pktinfo(msg, skb);
155 	if ((flags >>= 1) == 0)
156 		return;
157 
158 	if (flags & 1)
159 		ip_cmsg_recv_ttl(msg, skb);
160 	if ((flags >>= 1) == 0)
161 		return;
162 
163 	if (flags & 1)
164 		ip_cmsg_recv_tos(msg, skb);
165 	if ((flags >>= 1) == 0)
166 		return;
167 
168 	if (flags & 1)
169 		ip_cmsg_recv_opts(msg, skb);
170 	if ((flags >>= 1) == 0)
171 		return;
172 
173 	if (flags & 1)
174 		ip_cmsg_recv_retopts(msg, skb);
175 	if ((flags >>= 1) == 0)
176 		return;
177 
178 	if (flags & 1)
179 		ip_cmsg_recv_security(msg, skb);
180 
181 	if ((flags >>= 1) == 0)
182 		return;
183 	if (flags & 1)
184 		ip_cmsg_recv_dstaddr(msg, skb);
185 
186 }
187 EXPORT_SYMBOL(ip_cmsg_recv);
188 
189 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc,
190 		 bool allow_ipv6)
191 {
192 	int err, val;
193 	struct cmsghdr *cmsg;
194 
195 	for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
196 		if (!CMSG_OK(msg, cmsg))
197 			return -EINVAL;
198 #if defined(CONFIG_IPV6)
199 		if (allow_ipv6 &&
200 		    cmsg->cmsg_level == SOL_IPV6 &&
201 		    cmsg->cmsg_type == IPV6_PKTINFO) {
202 			struct in6_pktinfo *src_info;
203 
204 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
205 				return -EINVAL;
206 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
207 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
208 				return -EINVAL;
209 			ipc->oif = src_info->ipi6_ifindex;
210 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
211 			continue;
212 		}
213 #endif
214 		if (cmsg->cmsg_level != SOL_IP)
215 			continue;
216 		switch (cmsg->cmsg_type) {
217 		case IP_RETOPTS:
218 			err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
219 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
220 					     err < 40 ? err : 40);
221 			if (err)
222 				return err;
223 			break;
224 		case IP_PKTINFO:
225 		{
226 			struct in_pktinfo *info;
227 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
228 				return -EINVAL;
229 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
230 			ipc->oif = info->ipi_ifindex;
231 			ipc->addr = info->ipi_spec_dst.s_addr;
232 			break;
233 		}
234 		case IP_TTL:
235 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
236 				return -EINVAL;
237 			val = *(int *)CMSG_DATA(cmsg);
238 			if (val < 1 || val > 255)
239 				return -EINVAL;
240 			ipc->ttl = val;
241 			break;
242 		case IP_TOS:
243 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
244 				return -EINVAL;
245 			val = *(int *)CMSG_DATA(cmsg);
246 			if (val < 0 || val > 255)
247 				return -EINVAL;
248 			ipc->tos = val;
249 			ipc->priority = rt_tos2priority(ipc->tos);
250 			break;
251 
252 		default:
253 			return -EINVAL;
254 		}
255 	}
256 	return 0;
257 }
258 
259 
260 /* Special input handler for packets caught by router alert option.
261    They are selected only by protocol field, and then processed likely
262    local ones; but only if someone wants them! Otherwise, router
263    not running rsvpd will kill RSVP.
264 
265    It is user level problem, what it will make with them.
266    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
267    but receiver should be enough clever f.e. to forward mtrace requests,
268    sent to multicast group to reach destination designated router.
269  */
270 struct ip_ra_chain __rcu *ip_ra_chain;
271 static DEFINE_SPINLOCK(ip_ra_lock);
272 
273 
274 static void ip_ra_destroy_rcu(struct rcu_head *head)
275 {
276 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
277 
278 	sock_put(ra->saved_sk);
279 	kfree(ra);
280 }
281 
282 int ip_ra_control(struct sock *sk, unsigned char on,
283 		  void (*destructor)(struct sock *))
284 {
285 	struct ip_ra_chain *ra, *new_ra;
286 	struct ip_ra_chain __rcu **rap;
287 
288 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
289 		return -EINVAL;
290 
291 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
292 
293 	spin_lock_bh(&ip_ra_lock);
294 	for (rap = &ip_ra_chain;
295 	     (ra = rcu_dereference_protected(*rap,
296 			lockdep_is_held(&ip_ra_lock))) != NULL;
297 	     rap = &ra->next) {
298 		if (ra->sk == sk) {
299 			if (on) {
300 				spin_unlock_bh(&ip_ra_lock);
301 				kfree(new_ra);
302 				return -EADDRINUSE;
303 			}
304 			/* dont let ip_call_ra_chain() use sk again */
305 			ra->sk = NULL;
306 			RCU_INIT_POINTER(*rap, ra->next);
307 			spin_unlock_bh(&ip_ra_lock);
308 
309 			if (ra->destructor)
310 				ra->destructor(sk);
311 			/*
312 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
313 			 * period. This guarantee ip_call_ra_chain() dont need
314 			 * to mess with socket refcounts.
315 			 */
316 			ra->saved_sk = sk;
317 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
318 			return 0;
319 		}
320 	}
321 	if (new_ra == NULL) {
322 		spin_unlock_bh(&ip_ra_lock);
323 		return -ENOBUFS;
324 	}
325 	new_ra->sk = sk;
326 	new_ra->destructor = destructor;
327 
328 	RCU_INIT_POINTER(new_ra->next, ra);
329 	rcu_assign_pointer(*rap, new_ra);
330 	sock_hold(sk);
331 	spin_unlock_bh(&ip_ra_lock);
332 
333 	return 0;
334 }
335 
336 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
337 		   __be16 port, u32 info, u8 *payload)
338 {
339 	struct sock_exterr_skb *serr;
340 
341 	skb = skb_clone(skb, GFP_ATOMIC);
342 	if (!skb)
343 		return;
344 
345 	serr = SKB_EXT_ERR(skb);
346 	serr->ee.ee_errno = err;
347 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
348 	serr->ee.ee_type = icmp_hdr(skb)->type;
349 	serr->ee.ee_code = icmp_hdr(skb)->code;
350 	serr->ee.ee_pad = 0;
351 	serr->ee.ee_info = info;
352 	serr->ee.ee_data = 0;
353 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
354 				   skb_network_header(skb);
355 	serr->port = port;
356 
357 	if (skb_pull(skb, payload - skb->data) != NULL) {
358 		skb_reset_transport_header(skb);
359 		if (sock_queue_err_skb(sk, skb) == 0)
360 			return;
361 	}
362 	kfree_skb(skb);
363 }
364 
365 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
366 {
367 	struct inet_sock *inet = inet_sk(sk);
368 	struct sock_exterr_skb *serr;
369 	struct iphdr *iph;
370 	struct sk_buff *skb;
371 
372 	if (!inet->recverr)
373 		return;
374 
375 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
376 	if (!skb)
377 		return;
378 
379 	skb_put(skb, sizeof(struct iphdr));
380 	skb_reset_network_header(skb);
381 	iph = ip_hdr(skb);
382 	iph->daddr = daddr;
383 
384 	serr = SKB_EXT_ERR(skb);
385 	serr->ee.ee_errno = err;
386 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
387 	serr->ee.ee_type = 0;
388 	serr->ee.ee_code = 0;
389 	serr->ee.ee_pad = 0;
390 	serr->ee.ee_info = info;
391 	serr->ee.ee_data = 0;
392 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
393 	serr->port = port;
394 
395 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
396 	skb_reset_transport_header(skb);
397 
398 	if (sock_queue_err_skb(sk, skb))
399 		kfree_skb(skb);
400 }
401 
402 /*
403  *	Handle MSG_ERRQUEUE
404  */
405 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
406 {
407 	struct sock_exterr_skb *serr;
408 	struct sk_buff *skb;
409 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
410 	struct {
411 		struct sock_extended_err ee;
412 		struct sockaddr_in	 offender;
413 	} errhdr;
414 	int err;
415 	int copied;
416 
417 	err = -EAGAIN;
418 	skb = sock_dequeue_err_skb(sk);
419 	if (skb == NULL)
420 		goto out;
421 
422 	copied = skb->len;
423 	if (copied > len) {
424 		msg->msg_flags |= MSG_TRUNC;
425 		copied = len;
426 	}
427 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
428 	if (err)
429 		goto out_free_skb;
430 
431 	sock_recv_timestamp(msg, sk, skb);
432 
433 	serr = SKB_EXT_ERR(skb);
434 
435 	if (sin) {
436 		sin->sin_family = AF_INET;
437 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
438 						   serr->addr_offset);
439 		sin->sin_port = serr->port;
440 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
441 		*addr_len = sizeof(*sin);
442 	}
443 
444 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
445 	sin = &errhdr.offender;
446 	sin->sin_family = AF_UNSPEC;
447 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
448 		struct inet_sock *inet = inet_sk(sk);
449 
450 		sin->sin_family = AF_INET;
451 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
452 		sin->sin_port = 0;
453 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
454 		if (inet->cmsg_flags)
455 			ip_cmsg_recv(msg, skb);
456 	}
457 
458 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
459 
460 	/* Now we could try to dump offended packet options */
461 
462 	msg->msg_flags |= MSG_ERRQUEUE;
463 	err = copied;
464 
465 out_free_skb:
466 	kfree_skb(skb);
467 out:
468 	return err;
469 }
470 
471 
472 /*
473  *	Socket option code for IP. This is the end of the line after any
474  *	TCP,UDP etc options on an IP socket.
475  */
476 
477 static int do_ip_setsockopt(struct sock *sk, int level,
478 			    int optname, char __user *optval, unsigned int optlen)
479 {
480 	struct inet_sock *inet = inet_sk(sk);
481 	int val = 0, err;
482 
483 	switch (optname) {
484 	case IP_PKTINFO:
485 	case IP_RECVTTL:
486 	case IP_RECVOPTS:
487 	case IP_RECVTOS:
488 	case IP_RETOPTS:
489 	case IP_TOS:
490 	case IP_TTL:
491 	case IP_HDRINCL:
492 	case IP_MTU_DISCOVER:
493 	case IP_RECVERR:
494 	case IP_ROUTER_ALERT:
495 	case IP_FREEBIND:
496 	case IP_PASSSEC:
497 	case IP_TRANSPARENT:
498 	case IP_MINTTL:
499 	case IP_NODEFRAG:
500 	case IP_UNICAST_IF:
501 	case IP_MULTICAST_TTL:
502 	case IP_MULTICAST_ALL:
503 	case IP_MULTICAST_LOOP:
504 	case IP_RECVORIGDSTADDR:
505 		if (optlen >= sizeof(int)) {
506 			if (get_user(val, (int __user *) optval))
507 				return -EFAULT;
508 		} else if (optlen >= sizeof(char)) {
509 			unsigned char ucval;
510 
511 			if (get_user(ucval, (unsigned char __user *) optval))
512 				return -EFAULT;
513 			val = (int) ucval;
514 		}
515 	}
516 
517 	/* If optlen==0, it is equivalent to val == 0 */
518 
519 	if (ip_mroute_opt(optname))
520 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
521 
522 	err = 0;
523 	lock_sock(sk);
524 
525 	switch (optname) {
526 	case IP_OPTIONS:
527 	{
528 		struct ip_options_rcu *old, *opt = NULL;
529 
530 		if (optlen > 40)
531 			goto e_inval;
532 		err = ip_options_get_from_user(sock_net(sk), &opt,
533 					       optval, optlen);
534 		if (err)
535 			break;
536 		old = rcu_dereference_protected(inet->inet_opt,
537 						sock_owned_by_user(sk));
538 		if (inet->is_icsk) {
539 			struct inet_connection_sock *icsk = inet_csk(sk);
540 #if IS_ENABLED(CONFIG_IPV6)
541 			if (sk->sk_family == PF_INET ||
542 			    (!((1 << sk->sk_state) &
543 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
544 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
545 #endif
546 				if (old)
547 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
548 				if (opt)
549 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
550 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
551 #if IS_ENABLED(CONFIG_IPV6)
552 			}
553 #endif
554 		}
555 		rcu_assign_pointer(inet->inet_opt, opt);
556 		if (old)
557 			kfree_rcu(old, rcu);
558 		break;
559 	}
560 	case IP_PKTINFO:
561 		if (val)
562 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
563 		else
564 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
565 		break;
566 	case IP_RECVTTL:
567 		if (val)
568 			inet->cmsg_flags |=  IP_CMSG_TTL;
569 		else
570 			inet->cmsg_flags &= ~IP_CMSG_TTL;
571 		break;
572 	case IP_RECVTOS:
573 		if (val)
574 			inet->cmsg_flags |=  IP_CMSG_TOS;
575 		else
576 			inet->cmsg_flags &= ~IP_CMSG_TOS;
577 		break;
578 	case IP_RECVOPTS:
579 		if (val)
580 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
581 		else
582 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
583 		break;
584 	case IP_RETOPTS:
585 		if (val)
586 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
587 		else
588 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
589 		break;
590 	case IP_PASSSEC:
591 		if (val)
592 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
593 		else
594 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
595 		break;
596 	case IP_RECVORIGDSTADDR:
597 		if (val)
598 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
599 		else
600 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
601 		break;
602 	case IP_TOS:	/* This sets both TOS and Precedence */
603 		if (sk->sk_type == SOCK_STREAM) {
604 			val &= ~INET_ECN_MASK;
605 			val |= inet->tos & INET_ECN_MASK;
606 		}
607 		if (inet->tos != val) {
608 			inet->tos = val;
609 			sk->sk_priority = rt_tos2priority(val);
610 			sk_dst_reset(sk);
611 		}
612 		break;
613 	case IP_TTL:
614 		if (optlen < 1)
615 			goto e_inval;
616 		if (val != -1 && (val < 1 || val > 255))
617 			goto e_inval;
618 		inet->uc_ttl = val;
619 		break;
620 	case IP_HDRINCL:
621 		if (sk->sk_type != SOCK_RAW) {
622 			err = -ENOPROTOOPT;
623 			break;
624 		}
625 		inet->hdrincl = val ? 1 : 0;
626 		break;
627 	case IP_NODEFRAG:
628 		if (sk->sk_type != SOCK_RAW) {
629 			err = -ENOPROTOOPT;
630 			break;
631 		}
632 		inet->nodefrag = val ? 1 : 0;
633 		break;
634 	case IP_MTU_DISCOVER:
635 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
636 			goto e_inval;
637 		inet->pmtudisc = val;
638 		break;
639 	case IP_RECVERR:
640 		inet->recverr = !!val;
641 		if (!val)
642 			skb_queue_purge(&sk->sk_error_queue);
643 		break;
644 	case IP_MULTICAST_TTL:
645 		if (sk->sk_type == SOCK_STREAM)
646 			goto e_inval;
647 		if (optlen < 1)
648 			goto e_inval;
649 		if (val == -1)
650 			val = 1;
651 		if (val < 0 || val > 255)
652 			goto e_inval;
653 		inet->mc_ttl = val;
654 		break;
655 	case IP_MULTICAST_LOOP:
656 		if (optlen < 1)
657 			goto e_inval;
658 		inet->mc_loop = !!val;
659 		break;
660 	case IP_UNICAST_IF:
661 	{
662 		struct net_device *dev = NULL;
663 		int ifindex;
664 
665 		if (optlen != sizeof(int))
666 			goto e_inval;
667 
668 		ifindex = (__force int)ntohl((__force __be32)val);
669 		if (ifindex == 0) {
670 			inet->uc_index = 0;
671 			err = 0;
672 			break;
673 		}
674 
675 		dev = dev_get_by_index(sock_net(sk), ifindex);
676 		err = -EADDRNOTAVAIL;
677 		if (!dev)
678 			break;
679 		dev_put(dev);
680 
681 		err = -EINVAL;
682 		if (sk->sk_bound_dev_if)
683 			break;
684 
685 		inet->uc_index = ifindex;
686 		err = 0;
687 		break;
688 	}
689 	case IP_MULTICAST_IF:
690 	{
691 		struct ip_mreqn mreq;
692 		struct net_device *dev = NULL;
693 
694 		if (sk->sk_type == SOCK_STREAM)
695 			goto e_inval;
696 		/*
697 		 *	Check the arguments are allowable
698 		 */
699 
700 		if (optlen < sizeof(struct in_addr))
701 			goto e_inval;
702 
703 		err = -EFAULT;
704 		if (optlen >= sizeof(struct ip_mreqn)) {
705 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
706 				break;
707 		} else {
708 			memset(&mreq, 0, sizeof(mreq));
709 			if (optlen >= sizeof(struct ip_mreq)) {
710 				if (copy_from_user(&mreq, optval,
711 						   sizeof(struct ip_mreq)))
712 					break;
713 			} else if (optlen >= sizeof(struct in_addr)) {
714 				if (copy_from_user(&mreq.imr_address, optval,
715 						   sizeof(struct in_addr)))
716 					break;
717 			}
718 		}
719 
720 		if (!mreq.imr_ifindex) {
721 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
722 				inet->mc_index = 0;
723 				inet->mc_addr  = 0;
724 				err = 0;
725 				break;
726 			}
727 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
728 			if (dev)
729 				mreq.imr_ifindex = dev->ifindex;
730 		} else
731 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
732 
733 
734 		err = -EADDRNOTAVAIL;
735 		if (!dev)
736 			break;
737 		dev_put(dev);
738 
739 		err = -EINVAL;
740 		if (sk->sk_bound_dev_if &&
741 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
742 			break;
743 
744 		inet->mc_index = mreq.imr_ifindex;
745 		inet->mc_addr  = mreq.imr_address.s_addr;
746 		err = 0;
747 		break;
748 	}
749 
750 	case IP_ADD_MEMBERSHIP:
751 	case IP_DROP_MEMBERSHIP:
752 	{
753 		struct ip_mreqn mreq;
754 
755 		err = -EPROTO;
756 		if (inet_sk(sk)->is_icsk)
757 			break;
758 
759 		if (optlen < sizeof(struct ip_mreq))
760 			goto e_inval;
761 		err = -EFAULT;
762 		if (optlen >= sizeof(struct ip_mreqn)) {
763 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
764 				break;
765 		} else {
766 			memset(&mreq, 0, sizeof(mreq));
767 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
768 				break;
769 		}
770 
771 		if (optname == IP_ADD_MEMBERSHIP)
772 			err = ip_mc_join_group(sk, &mreq);
773 		else
774 			err = ip_mc_leave_group(sk, &mreq);
775 		break;
776 	}
777 	case IP_MSFILTER:
778 	{
779 		struct ip_msfilter *msf;
780 
781 		if (optlen < IP_MSFILTER_SIZE(0))
782 			goto e_inval;
783 		if (optlen > sysctl_optmem_max) {
784 			err = -ENOBUFS;
785 			break;
786 		}
787 		msf = kmalloc(optlen, GFP_KERNEL);
788 		if (!msf) {
789 			err = -ENOBUFS;
790 			break;
791 		}
792 		err = -EFAULT;
793 		if (copy_from_user(msf, optval, optlen)) {
794 			kfree(msf);
795 			break;
796 		}
797 		/* numsrc >= (1G-4) overflow in 32 bits */
798 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
799 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
800 			kfree(msf);
801 			err = -ENOBUFS;
802 			break;
803 		}
804 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
805 			kfree(msf);
806 			err = -EINVAL;
807 			break;
808 		}
809 		err = ip_mc_msfilter(sk, msf, 0);
810 		kfree(msf);
811 		break;
812 	}
813 	case IP_BLOCK_SOURCE:
814 	case IP_UNBLOCK_SOURCE:
815 	case IP_ADD_SOURCE_MEMBERSHIP:
816 	case IP_DROP_SOURCE_MEMBERSHIP:
817 	{
818 		struct ip_mreq_source mreqs;
819 		int omode, add;
820 
821 		if (optlen != sizeof(struct ip_mreq_source))
822 			goto e_inval;
823 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
824 			err = -EFAULT;
825 			break;
826 		}
827 		if (optname == IP_BLOCK_SOURCE) {
828 			omode = MCAST_EXCLUDE;
829 			add = 1;
830 		} else if (optname == IP_UNBLOCK_SOURCE) {
831 			omode = MCAST_EXCLUDE;
832 			add = 0;
833 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
834 			struct ip_mreqn mreq;
835 
836 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
837 			mreq.imr_address.s_addr = mreqs.imr_interface;
838 			mreq.imr_ifindex = 0;
839 			err = ip_mc_join_group(sk, &mreq);
840 			if (err && err != -EADDRINUSE)
841 				break;
842 			omode = MCAST_INCLUDE;
843 			add = 1;
844 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
845 			omode = MCAST_INCLUDE;
846 			add = 0;
847 		}
848 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
849 		break;
850 	}
851 	case MCAST_JOIN_GROUP:
852 	case MCAST_LEAVE_GROUP:
853 	{
854 		struct group_req greq;
855 		struct sockaddr_in *psin;
856 		struct ip_mreqn mreq;
857 
858 		if (optlen < sizeof(struct group_req))
859 			goto e_inval;
860 		err = -EFAULT;
861 		if (copy_from_user(&greq, optval, sizeof(greq)))
862 			break;
863 		psin = (struct sockaddr_in *)&greq.gr_group;
864 		if (psin->sin_family != AF_INET)
865 			goto e_inval;
866 		memset(&mreq, 0, sizeof(mreq));
867 		mreq.imr_multiaddr = psin->sin_addr;
868 		mreq.imr_ifindex = greq.gr_interface;
869 
870 		if (optname == MCAST_JOIN_GROUP)
871 			err = ip_mc_join_group(sk, &mreq);
872 		else
873 			err = ip_mc_leave_group(sk, &mreq);
874 		break;
875 	}
876 	case MCAST_JOIN_SOURCE_GROUP:
877 	case MCAST_LEAVE_SOURCE_GROUP:
878 	case MCAST_BLOCK_SOURCE:
879 	case MCAST_UNBLOCK_SOURCE:
880 	{
881 		struct group_source_req greqs;
882 		struct ip_mreq_source mreqs;
883 		struct sockaddr_in *psin;
884 		int omode, add;
885 
886 		if (optlen != sizeof(struct group_source_req))
887 			goto e_inval;
888 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
889 			err = -EFAULT;
890 			break;
891 		}
892 		if (greqs.gsr_group.ss_family != AF_INET ||
893 		    greqs.gsr_source.ss_family != AF_INET) {
894 			err = -EADDRNOTAVAIL;
895 			break;
896 		}
897 		psin = (struct sockaddr_in *)&greqs.gsr_group;
898 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
899 		psin = (struct sockaddr_in *)&greqs.gsr_source;
900 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
901 		mreqs.imr_interface = 0; /* use index for mc_source */
902 
903 		if (optname == MCAST_BLOCK_SOURCE) {
904 			omode = MCAST_EXCLUDE;
905 			add = 1;
906 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
907 			omode = MCAST_EXCLUDE;
908 			add = 0;
909 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
910 			struct ip_mreqn mreq;
911 
912 			psin = (struct sockaddr_in *)&greqs.gsr_group;
913 			mreq.imr_multiaddr = psin->sin_addr;
914 			mreq.imr_address.s_addr = 0;
915 			mreq.imr_ifindex = greqs.gsr_interface;
916 			err = ip_mc_join_group(sk, &mreq);
917 			if (err && err != -EADDRINUSE)
918 				break;
919 			greqs.gsr_interface = mreq.imr_ifindex;
920 			omode = MCAST_INCLUDE;
921 			add = 1;
922 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
923 			omode = MCAST_INCLUDE;
924 			add = 0;
925 		}
926 		err = ip_mc_source(add, omode, sk, &mreqs,
927 				   greqs.gsr_interface);
928 		break;
929 	}
930 	case MCAST_MSFILTER:
931 	{
932 		struct sockaddr_in *psin;
933 		struct ip_msfilter *msf = NULL;
934 		struct group_filter *gsf = NULL;
935 		int msize, i, ifindex;
936 
937 		if (optlen < GROUP_FILTER_SIZE(0))
938 			goto e_inval;
939 		if (optlen > sysctl_optmem_max) {
940 			err = -ENOBUFS;
941 			break;
942 		}
943 		gsf = kmalloc(optlen, GFP_KERNEL);
944 		if (!gsf) {
945 			err = -ENOBUFS;
946 			break;
947 		}
948 		err = -EFAULT;
949 		if (copy_from_user(gsf, optval, optlen))
950 			goto mc_msf_out;
951 
952 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
953 		if (gsf->gf_numsrc >= 0x1ffffff ||
954 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
955 			err = -ENOBUFS;
956 			goto mc_msf_out;
957 		}
958 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
959 			err = -EINVAL;
960 			goto mc_msf_out;
961 		}
962 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
963 		msf = kmalloc(msize, GFP_KERNEL);
964 		if (!msf) {
965 			err = -ENOBUFS;
966 			goto mc_msf_out;
967 		}
968 		ifindex = gsf->gf_interface;
969 		psin = (struct sockaddr_in *)&gsf->gf_group;
970 		if (psin->sin_family != AF_INET) {
971 			err = -EADDRNOTAVAIL;
972 			goto mc_msf_out;
973 		}
974 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
975 		msf->imsf_interface = 0;
976 		msf->imsf_fmode = gsf->gf_fmode;
977 		msf->imsf_numsrc = gsf->gf_numsrc;
978 		err = -EADDRNOTAVAIL;
979 		for (i = 0; i < gsf->gf_numsrc; ++i) {
980 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
981 
982 			if (psin->sin_family != AF_INET)
983 				goto mc_msf_out;
984 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
985 		}
986 		kfree(gsf);
987 		gsf = NULL;
988 
989 		err = ip_mc_msfilter(sk, msf, ifindex);
990 mc_msf_out:
991 		kfree(msf);
992 		kfree(gsf);
993 		break;
994 	}
995 	case IP_MULTICAST_ALL:
996 		if (optlen < 1)
997 			goto e_inval;
998 		if (val != 0 && val != 1)
999 			goto e_inval;
1000 		inet->mc_all = val;
1001 		break;
1002 	case IP_ROUTER_ALERT:
1003 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1004 		break;
1005 
1006 	case IP_FREEBIND:
1007 		if (optlen < 1)
1008 			goto e_inval;
1009 		inet->freebind = !!val;
1010 		break;
1011 
1012 	case IP_IPSEC_POLICY:
1013 	case IP_XFRM_POLICY:
1014 		err = -EPERM;
1015 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1016 			break;
1017 		err = xfrm_user_policy(sk, optname, optval, optlen);
1018 		break;
1019 
1020 	case IP_TRANSPARENT:
1021 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1022 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1023 			err = -EPERM;
1024 			break;
1025 		}
1026 		if (optlen < 1)
1027 			goto e_inval;
1028 		inet->transparent = !!val;
1029 		break;
1030 
1031 	case IP_MINTTL:
1032 		if (optlen < 1)
1033 			goto e_inval;
1034 		if (val < 0 || val > 255)
1035 			goto e_inval;
1036 		inet->min_ttl = val;
1037 		break;
1038 
1039 	default:
1040 		err = -ENOPROTOOPT;
1041 		break;
1042 	}
1043 	release_sock(sk);
1044 	return err;
1045 
1046 e_inval:
1047 	release_sock(sk);
1048 	return -EINVAL;
1049 }
1050 
1051 /**
1052  * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1053  * @sk: socket
1054  * @skb: buffer
1055  *
1056  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1057  * destination in skb->cb[] before dst drop.
1058  * This way, receiver doesn't make cache line misses to read rtable.
1059  */
1060 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1061 {
1062 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1063 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1064 		       ipv6_sk_rxinfo(sk);
1065 
1066 	if (prepare && skb_rtable(skb)) {
1067 		pktinfo->ipi_ifindex = inet_iif(skb);
1068 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1069 	} else {
1070 		pktinfo->ipi_ifindex = 0;
1071 		pktinfo->ipi_spec_dst.s_addr = 0;
1072 	}
1073 	skb_dst_drop(skb);
1074 }
1075 
1076 int ip_setsockopt(struct sock *sk, int level,
1077 		int optname, char __user *optval, unsigned int optlen)
1078 {
1079 	int err;
1080 
1081 	if (level != SOL_IP)
1082 		return -ENOPROTOOPT;
1083 
1084 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1085 #ifdef CONFIG_NETFILTER
1086 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1087 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1088 			optname != IP_IPSEC_POLICY &&
1089 			optname != IP_XFRM_POLICY &&
1090 			!ip_mroute_opt(optname)) {
1091 		lock_sock(sk);
1092 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1093 		release_sock(sk);
1094 	}
1095 #endif
1096 	return err;
1097 }
1098 EXPORT_SYMBOL(ip_setsockopt);
1099 
1100 #ifdef CONFIG_COMPAT
1101 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1102 			 char __user *optval, unsigned int optlen)
1103 {
1104 	int err;
1105 
1106 	if (level != SOL_IP)
1107 		return -ENOPROTOOPT;
1108 
1109 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1110 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1111 			ip_setsockopt);
1112 
1113 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1114 #ifdef CONFIG_NETFILTER
1115 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1116 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1117 			optname != IP_IPSEC_POLICY &&
1118 			optname != IP_XFRM_POLICY &&
1119 			!ip_mroute_opt(optname)) {
1120 		lock_sock(sk);
1121 		err = compat_nf_setsockopt(sk, PF_INET, optname,
1122 					   optval, optlen);
1123 		release_sock(sk);
1124 	}
1125 #endif
1126 	return err;
1127 }
1128 EXPORT_SYMBOL(compat_ip_setsockopt);
1129 #endif
1130 
1131 /*
1132  *	Get the options. Note for future reference. The GET of IP options gets
1133  *	the _received_ ones. The set sets the _sent_ ones.
1134  */
1135 
1136 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1137 			    char __user *optval, int __user *optlen, unsigned int flags)
1138 {
1139 	struct inet_sock *inet = inet_sk(sk);
1140 	int val;
1141 	int len;
1142 
1143 	if (level != SOL_IP)
1144 		return -EOPNOTSUPP;
1145 
1146 	if (ip_mroute_opt(optname))
1147 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1148 
1149 	if (get_user(len, optlen))
1150 		return -EFAULT;
1151 	if (len < 0)
1152 		return -EINVAL;
1153 
1154 	lock_sock(sk);
1155 
1156 	switch (optname) {
1157 	case IP_OPTIONS:
1158 	{
1159 		unsigned char optbuf[sizeof(struct ip_options)+40];
1160 		struct ip_options *opt = (struct ip_options *)optbuf;
1161 		struct ip_options_rcu *inet_opt;
1162 
1163 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1164 						     sock_owned_by_user(sk));
1165 		opt->optlen = 0;
1166 		if (inet_opt)
1167 			memcpy(optbuf, &inet_opt->opt,
1168 			       sizeof(struct ip_options) +
1169 			       inet_opt->opt.optlen);
1170 		release_sock(sk);
1171 
1172 		if (opt->optlen == 0)
1173 			return put_user(0, optlen);
1174 
1175 		ip_options_undo(opt);
1176 
1177 		len = min_t(unsigned int, len, opt->optlen);
1178 		if (put_user(len, optlen))
1179 			return -EFAULT;
1180 		if (copy_to_user(optval, opt->__data, len))
1181 			return -EFAULT;
1182 		return 0;
1183 	}
1184 	case IP_PKTINFO:
1185 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1186 		break;
1187 	case IP_RECVTTL:
1188 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1189 		break;
1190 	case IP_RECVTOS:
1191 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1192 		break;
1193 	case IP_RECVOPTS:
1194 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1195 		break;
1196 	case IP_RETOPTS:
1197 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1198 		break;
1199 	case IP_PASSSEC:
1200 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1201 		break;
1202 	case IP_RECVORIGDSTADDR:
1203 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1204 		break;
1205 	case IP_TOS:
1206 		val = inet->tos;
1207 		break;
1208 	case IP_TTL:
1209 		val = (inet->uc_ttl == -1 ?
1210 		       sysctl_ip_default_ttl :
1211 		       inet->uc_ttl);
1212 		break;
1213 	case IP_HDRINCL:
1214 		val = inet->hdrincl;
1215 		break;
1216 	case IP_NODEFRAG:
1217 		val = inet->nodefrag;
1218 		break;
1219 	case IP_MTU_DISCOVER:
1220 		val = inet->pmtudisc;
1221 		break;
1222 	case IP_MTU:
1223 	{
1224 		struct dst_entry *dst;
1225 		val = 0;
1226 		dst = sk_dst_get(sk);
1227 		if (dst) {
1228 			val = dst_mtu(dst);
1229 			dst_release(dst);
1230 		}
1231 		if (!val) {
1232 			release_sock(sk);
1233 			return -ENOTCONN;
1234 		}
1235 		break;
1236 	}
1237 	case IP_RECVERR:
1238 		val = inet->recverr;
1239 		break;
1240 	case IP_MULTICAST_TTL:
1241 		val = inet->mc_ttl;
1242 		break;
1243 	case IP_MULTICAST_LOOP:
1244 		val = inet->mc_loop;
1245 		break;
1246 	case IP_UNICAST_IF:
1247 		val = (__force int)htonl((__u32) inet->uc_index);
1248 		break;
1249 	case IP_MULTICAST_IF:
1250 	{
1251 		struct in_addr addr;
1252 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1253 		addr.s_addr = inet->mc_addr;
1254 		release_sock(sk);
1255 
1256 		if (put_user(len, optlen))
1257 			return -EFAULT;
1258 		if (copy_to_user(optval, &addr, len))
1259 			return -EFAULT;
1260 		return 0;
1261 	}
1262 	case IP_MSFILTER:
1263 	{
1264 		struct ip_msfilter msf;
1265 		int err;
1266 
1267 		if (len < IP_MSFILTER_SIZE(0)) {
1268 			release_sock(sk);
1269 			return -EINVAL;
1270 		}
1271 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1272 			release_sock(sk);
1273 			return -EFAULT;
1274 		}
1275 		err = ip_mc_msfget(sk, &msf,
1276 				   (struct ip_msfilter __user *)optval, optlen);
1277 		release_sock(sk);
1278 		return err;
1279 	}
1280 	case MCAST_MSFILTER:
1281 	{
1282 		struct group_filter gsf;
1283 		int err;
1284 
1285 		if (len < GROUP_FILTER_SIZE(0)) {
1286 			release_sock(sk);
1287 			return -EINVAL;
1288 		}
1289 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1290 			release_sock(sk);
1291 			return -EFAULT;
1292 		}
1293 		err = ip_mc_gsfget(sk, &gsf,
1294 				   (struct group_filter __user *)optval,
1295 				   optlen);
1296 		release_sock(sk);
1297 		return err;
1298 	}
1299 	case IP_MULTICAST_ALL:
1300 		val = inet->mc_all;
1301 		break;
1302 	case IP_PKTOPTIONS:
1303 	{
1304 		struct msghdr msg;
1305 
1306 		release_sock(sk);
1307 
1308 		if (sk->sk_type != SOCK_STREAM)
1309 			return -ENOPROTOOPT;
1310 
1311 		msg.msg_control = (__force void *) optval;
1312 		msg.msg_controllen = len;
1313 		msg.msg_flags = flags;
1314 
1315 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1316 			struct in_pktinfo info;
1317 
1318 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1319 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1320 			info.ipi_ifindex = inet->mc_index;
1321 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1322 		}
1323 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1324 			int hlim = inet->mc_ttl;
1325 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1326 		}
1327 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1328 			int tos = inet->rcv_tos;
1329 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1330 		}
1331 		len -= msg.msg_controllen;
1332 		return put_user(len, optlen);
1333 	}
1334 	case IP_FREEBIND:
1335 		val = inet->freebind;
1336 		break;
1337 	case IP_TRANSPARENT:
1338 		val = inet->transparent;
1339 		break;
1340 	case IP_MINTTL:
1341 		val = inet->min_ttl;
1342 		break;
1343 	default:
1344 		release_sock(sk);
1345 		return -ENOPROTOOPT;
1346 	}
1347 	release_sock(sk);
1348 
1349 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1350 		unsigned char ucval = (unsigned char)val;
1351 		len = 1;
1352 		if (put_user(len, optlen))
1353 			return -EFAULT;
1354 		if (copy_to_user(optval, &ucval, 1))
1355 			return -EFAULT;
1356 	} else {
1357 		len = min_t(unsigned int, sizeof(int), len);
1358 		if (put_user(len, optlen))
1359 			return -EFAULT;
1360 		if (copy_to_user(optval, &val, len))
1361 			return -EFAULT;
1362 	}
1363 	return 0;
1364 }
1365 
1366 int ip_getsockopt(struct sock *sk, int level,
1367 		  int optname, char __user *optval, int __user *optlen)
1368 {
1369 	int err;
1370 
1371 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1372 #ifdef CONFIG_NETFILTER
1373 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1374 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1375 			!ip_mroute_opt(optname)) {
1376 		int len;
1377 
1378 		if (get_user(len, optlen))
1379 			return -EFAULT;
1380 
1381 		lock_sock(sk);
1382 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1383 				&len);
1384 		release_sock(sk);
1385 		if (err >= 0)
1386 			err = put_user(len, optlen);
1387 		return err;
1388 	}
1389 #endif
1390 	return err;
1391 }
1392 EXPORT_SYMBOL(ip_getsockopt);
1393 
1394 #ifdef CONFIG_COMPAT
1395 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1396 			 char __user *optval, int __user *optlen)
1397 {
1398 	int err;
1399 
1400 	if (optname == MCAST_MSFILTER)
1401 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1402 			ip_getsockopt);
1403 
1404 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1405 		MSG_CMSG_COMPAT);
1406 
1407 #ifdef CONFIG_NETFILTER
1408 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1409 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1410 			!ip_mroute_opt(optname)) {
1411 		int len;
1412 
1413 		if (get_user(len, optlen))
1414 			return -EFAULT;
1415 
1416 		lock_sock(sk);
1417 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1418 		release_sock(sk);
1419 		if (err >= 0)
1420 			err = put_user(len, optlen);
1421 		return err;
1422 	}
1423 #endif
1424 	return err;
1425 }
1426 EXPORT_SYMBOL(compat_ip_getsockopt);
1427 #endif
1428