xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 930beb5a)
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 #define PKTINFO_SKB_CB(__skb) ((struct in_pktinfo *)((__skb)->cb))
60 
61 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
62 {
63 	struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
64 
65 	info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
66 
67 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
68 }
69 
70 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
71 {
72 	int ttl = ip_hdr(skb)->ttl;
73 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
74 }
75 
76 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
77 {
78 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
79 }
80 
81 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
82 {
83 	if (IPCB(skb)->opt.optlen == 0)
84 		return;
85 
86 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
87 		 ip_hdr(skb) + 1);
88 }
89 
90 
91 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
92 {
93 	unsigned char optbuf[sizeof(struct ip_options) + 40];
94 	struct ip_options *opt = (struct ip_options *)optbuf;
95 
96 	if (IPCB(skb)->opt.optlen == 0)
97 		return;
98 
99 	if (ip_options_echo(opt, skb)) {
100 		msg->msg_flags |= MSG_CTRUNC;
101 		return;
102 	}
103 	ip_options_undo(opt);
104 
105 	put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
106 }
107 
108 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
109 {
110 	char *secdata;
111 	u32 seclen, secid;
112 	int err;
113 
114 	err = security_socket_getpeersec_dgram(NULL, skb, &secid);
115 	if (err)
116 		return;
117 
118 	err = security_secid_to_secctx(secid, &secdata, &seclen);
119 	if (err)
120 		return;
121 
122 	put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
123 	security_release_secctx(secdata, seclen);
124 }
125 
126 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
127 {
128 	struct sockaddr_in sin;
129 	const struct iphdr *iph = ip_hdr(skb);
130 	__be16 *ports = (__be16 *)skb_transport_header(skb);
131 
132 	if (skb_transport_offset(skb) + 4 > skb->len)
133 		return;
134 
135 	/* All current transport protocols have the port numbers in the
136 	 * first four bytes of the transport header and this function is
137 	 * written with this assumption in mind.
138 	 */
139 
140 	sin.sin_family = AF_INET;
141 	sin.sin_addr.s_addr = iph->daddr;
142 	sin.sin_port = ports[1];
143 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
144 
145 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
146 }
147 
148 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
149 {
150 	struct inet_sock *inet = inet_sk(skb->sk);
151 	unsigned int flags = inet->cmsg_flags;
152 
153 	/* Ordered by supposed usage frequency */
154 	if (flags & 1)
155 		ip_cmsg_recv_pktinfo(msg, skb);
156 	if ((flags >>= 1) == 0)
157 		return;
158 
159 	if (flags & 1)
160 		ip_cmsg_recv_ttl(msg, skb);
161 	if ((flags >>= 1) == 0)
162 		return;
163 
164 	if (flags & 1)
165 		ip_cmsg_recv_tos(msg, skb);
166 	if ((flags >>= 1) == 0)
167 		return;
168 
169 	if (flags & 1)
170 		ip_cmsg_recv_opts(msg, skb);
171 	if ((flags >>= 1) == 0)
172 		return;
173 
174 	if (flags & 1)
175 		ip_cmsg_recv_retopts(msg, skb);
176 	if ((flags >>= 1) == 0)
177 		return;
178 
179 	if (flags & 1)
180 		ip_cmsg_recv_security(msg, skb);
181 
182 	if ((flags >>= 1) == 0)
183 		return;
184 	if (flags & 1)
185 		ip_cmsg_recv_dstaddr(msg, skb);
186 
187 }
188 EXPORT_SYMBOL(ip_cmsg_recv);
189 
190 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
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 (cmsg->cmsg_level != SOL_IP)
199 			continue;
200 		switch (cmsg->cmsg_type) {
201 		case IP_RETOPTS:
202 			err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
203 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
204 					     err < 40 ? err : 40);
205 			if (err)
206 				return err;
207 			break;
208 		case IP_PKTINFO:
209 		{
210 			struct in_pktinfo *info;
211 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
212 				return -EINVAL;
213 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
214 			ipc->oif = info->ipi_ifindex;
215 			ipc->addr = info->ipi_spec_dst.s_addr;
216 			break;
217 		}
218 		case IP_TTL:
219 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
220 				return -EINVAL;
221 			val = *(int *)CMSG_DATA(cmsg);
222 			if (val < 1 || val > 255)
223 				return -EINVAL;
224 			ipc->ttl = val;
225 			break;
226 		case IP_TOS:
227 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
228 				return -EINVAL;
229 			val = *(int *)CMSG_DATA(cmsg);
230 			if (val < 0 || val > 255)
231 				return -EINVAL;
232 			ipc->tos = val;
233 			ipc->priority = rt_tos2priority(ipc->tos);
234 			break;
235 
236 		default:
237 			return -EINVAL;
238 		}
239 	}
240 	return 0;
241 }
242 
243 
244 /* Special input handler for packets caught by router alert option.
245    They are selected only by protocol field, and then processed likely
246    local ones; but only if someone wants them! Otherwise, router
247    not running rsvpd will kill RSVP.
248 
249    It is user level problem, what it will make with them.
250    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
251    but receiver should be enough clever f.e. to forward mtrace requests,
252    sent to multicast group to reach destination designated router.
253  */
254 struct ip_ra_chain __rcu *ip_ra_chain;
255 static DEFINE_SPINLOCK(ip_ra_lock);
256 
257 
258 static void ip_ra_destroy_rcu(struct rcu_head *head)
259 {
260 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
261 
262 	sock_put(ra->saved_sk);
263 	kfree(ra);
264 }
265 
266 int ip_ra_control(struct sock *sk, unsigned char on,
267 		  void (*destructor)(struct sock *))
268 {
269 	struct ip_ra_chain *ra, *new_ra;
270 	struct ip_ra_chain __rcu **rap;
271 
272 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
273 		return -EINVAL;
274 
275 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
276 
277 	spin_lock_bh(&ip_ra_lock);
278 	for (rap = &ip_ra_chain;
279 	     (ra = rcu_dereference_protected(*rap,
280 			lockdep_is_held(&ip_ra_lock))) != NULL;
281 	     rap = &ra->next) {
282 		if (ra->sk == sk) {
283 			if (on) {
284 				spin_unlock_bh(&ip_ra_lock);
285 				kfree(new_ra);
286 				return -EADDRINUSE;
287 			}
288 			/* dont let ip_call_ra_chain() use sk again */
289 			ra->sk = NULL;
290 			rcu_assign_pointer(*rap, ra->next);
291 			spin_unlock_bh(&ip_ra_lock);
292 
293 			if (ra->destructor)
294 				ra->destructor(sk);
295 			/*
296 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
297 			 * period. This guarantee ip_call_ra_chain() dont need
298 			 * to mess with socket refcounts.
299 			 */
300 			ra->saved_sk = sk;
301 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
302 			return 0;
303 		}
304 	}
305 	if (new_ra == NULL) {
306 		spin_unlock_bh(&ip_ra_lock);
307 		return -ENOBUFS;
308 	}
309 	new_ra->sk = sk;
310 	new_ra->destructor = destructor;
311 
312 	new_ra->next = ra;
313 	rcu_assign_pointer(*rap, new_ra);
314 	sock_hold(sk);
315 	spin_unlock_bh(&ip_ra_lock);
316 
317 	return 0;
318 }
319 
320 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
321 		   __be16 port, u32 info, u8 *payload)
322 {
323 	struct sock_exterr_skb *serr;
324 
325 	skb = skb_clone(skb, GFP_ATOMIC);
326 	if (!skb)
327 		return;
328 
329 	serr = SKB_EXT_ERR(skb);
330 	serr->ee.ee_errno = err;
331 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
332 	serr->ee.ee_type = icmp_hdr(skb)->type;
333 	serr->ee.ee_code = icmp_hdr(skb)->code;
334 	serr->ee.ee_pad = 0;
335 	serr->ee.ee_info = info;
336 	serr->ee.ee_data = 0;
337 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
338 				   skb_network_header(skb);
339 	serr->port = port;
340 
341 	if (skb_pull(skb, payload - skb->data) != NULL) {
342 		skb_reset_transport_header(skb);
343 		if (sock_queue_err_skb(sk, skb) == 0)
344 			return;
345 	}
346 	kfree_skb(skb);
347 }
348 
349 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
350 {
351 	struct inet_sock *inet = inet_sk(sk);
352 	struct sock_exterr_skb *serr;
353 	struct iphdr *iph;
354 	struct sk_buff *skb;
355 
356 	if (!inet->recverr)
357 		return;
358 
359 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
360 	if (!skb)
361 		return;
362 
363 	skb_put(skb, sizeof(struct iphdr));
364 	skb_reset_network_header(skb);
365 	iph = ip_hdr(skb);
366 	iph->daddr = daddr;
367 
368 	serr = SKB_EXT_ERR(skb);
369 	serr->ee.ee_errno = err;
370 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
371 	serr->ee.ee_type = 0;
372 	serr->ee.ee_code = 0;
373 	serr->ee.ee_pad = 0;
374 	serr->ee.ee_info = info;
375 	serr->ee.ee_data = 0;
376 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
377 	serr->port = port;
378 
379 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
380 	skb_reset_transport_header(skb);
381 
382 	if (sock_queue_err_skb(sk, skb))
383 		kfree_skb(skb);
384 }
385 
386 /*
387  *	Handle MSG_ERRQUEUE
388  */
389 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
390 {
391 	struct sock_exterr_skb *serr;
392 	struct sk_buff *skb, *skb2;
393 	struct sockaddr_in *sin;
394 	struct {
395 		struct sock_extended_err ee;
396 		struct sockaddr_in	 offender;
397 	} errhdr;
398 	int err;
399 	int copied;
400 
401 	err = -EAGAIN;
402 	skb = skb_dequeue(&sk->sk_error_queue);
403 	if (skb == NULL)
404 		goto out;
405 
406 	copied = skb->len;
407 	if (copied > len) {
408 		msg->msg_flags |= MSG_TRUNC;
409 		copied = len;
410 	}
411 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
412 	if (err)
413 		goto out_free_skb;
414 
415 	sock_recv_timestamp(msg, sk, skb);
416 
417 	serr = SKB_EXT_ERR(skb);
418 
419 	sin = (struct sockaddr_in *)msg->msg_name;
420 	if (sin) {
421 		sin->sin_family = AF_INET;
422 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
423 						   serr->addr_offset);
424 		sin->sin_port = serr->port;
425 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
426 		*addr_len = sizeof(*sin);
427 	}
428 
429 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
430 	sin = &errhdr.offender;
431 	sin->sin_family = AF_UNSPEC;
432 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
433 		struct inet_sock *inet = inet_sk(sk);
434 
435 		sin->sin_family = AF_INET;
436 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
437 		sin->sin_port = 0;
438 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
439 		if (inet->cmsg_flags)
440 			ip_cmsg_recv(msg, skb);
441 	}
442 
443 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
444 
445 	/* Now we could try to dump offended packet options */
446 
447 	msg->msg_flags |= MSG_ERRQUEUE;
448 	err = copied;
449 
450 	/* Reset and regenerate socket error */
451 	spin_lock_bh(&sk->sk_error_queue.lock);
452 	sk->sk_err = 0;
453 	skb2 = skb_peek(&sk->sk_error_queue);
454 	if (skb2 != NULL) {
455 		sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
456 		spin_unlock_bh(&sk->sk_error_queue.lock);
457 		sk->sk_error_report(sk);
458 	} else
459 		spin_unlock_bh(&sk->sk_error_queue.lock);
460 
461 out_free_skb:
462 	kfree_skb(skb);
463 out:
464 	return err;
465 }
466 
467 
468 /*
469  *	Socket option code for IP. This is the end of the line after any
470  *	TCP,UDP etc options on an IP socket.
471  */
472 
473 static int do_ip_setsockopt(struct sock *sk, int level,
474 			    int optname, char __user *optval, unsigned int optlen)
475 {
476 	struct inet_sock *inet = inet_sk(sk);
477 	int val = 0, err;
478 
479 	switch (optname) {
480 	case IP_PKTINFO:
481 	case IP_RECVTTL:
482 	case IP_RECVOPTS:
483 	case IP_RECVTOS:
484 	case IP_RETOPTS:
485 	case IP_TOS:
486 	case IP_TTL:
487 	case IP_HDRINCL:
488 	case IP_MTU_DISCOVER:
489 	case IP_RECVERR:
490 	case IP_ROUTER_ALERT:
491 	case IP_FREEBIND:
492 	case IP_PASSSEC:
493 	case IP_TRANSPARENT:
494 	case IP_MINTTL:
495 	case IP_NODEFRAG:
496 	case IP_UNICAST_IF:
497 	case IP_MULTICAST_TTL:
498 	case IP_MULTICAST_ALL:
499 	case IP_MULTICAST_LOOP:
500 	case IP_RECVORIGDSTADDR:
501 		if (optlen >= sizeof(int)) {
502 			if (get_user(val, (int __user *) optval))
503 				return -EFAULT;
504 		} else if (optlen >= sizeof(char)) {
505 			unsigned char ucval;
506 
507 			if (get_user(ucval, (unsigned char __user *) optval))
508 				return -EFAULT;
509 			val = (int) ucval;
510 		}
511 	}
512 
513 	/* If optlen==0, it is equivalent to val == 0 */
514 
515 	if (ip_mroute_opt(optname))
516 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
517 
518 	err = 0;
519 	lock_sock(sk);
520 
521 	switch (optname) {
522 	case IP_OPTIONS:
523 	{
524 		struct ip_options_rcu *old, *opt = NULL;
525 
526 		if (optlen > 40)
527 			goto e_inval;
528 		err = ip_options_get_from_user(sock_net(sk), &opt,
529 					       optval, optlen);
530 		if (err)
531 			break;
532 		old = rcu_dereference_protected(inet->inet_opt,
533 						sock_owned_by_user(sk));
534 		if (inet->is_icsk) {
535 			struct inet_connection_sock *icsk = inet_csk(sk);
536 #if IS_ENABLED(CONFIG_IPV6)
537 			if (sk->sk_family == PF_INET ||
538 			    (!((1 << sk->sk_state) &
539 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
540 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
541 #endif
542 				if (old)
543 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
544 				if (opt)
545 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
546 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
547 #if IS_ENABLED(CONFIG_IPV6)
548 			}
549 #endif
550 		}
551 		rcu_assign_pointer(inet->inet_opt, opt);
552 		if (old)
553 			kfree_rcu(old, rcu);
554 		break;
555 	}
556 	case IP_PKTINFO:
557 		if (val)
558 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
559 		else
560 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
561 		break;
562 	case IP_RECVTTL:
563 		if (val)
564 			inet->cmsg_flags |=  IP_CMSG_TTL;
565 		else
566 			inet->cmsg_flags &= ~IP_CMSG_TTL;
567 		break;
568 	case IP_RECVTOS:
569 		if (val)
570 			inet->cmsg_flags |=  IP_CMSG_TOS;
571 		else
572 			inet->cmsg_flags &= ~IP_CMSG_TOS;
573 		break;
574 	case IP_RECVOPTS:
575 		if (val)
576 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
577 		else
578 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
579 		break;
580 	case IP_RETOPTS:
581 		if (val)
582 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
583 		else
584 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
585 		break;
586 	case IP_PASSSEC:
587 		if (val)
588 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
589 		else
590 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
591 		break;
592 	case IP_RECVORIGDSTADDR:
593 		if (val)
594 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
595 		else
596 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
597 		break;
598 	case IP_TOS:	/* This sets both TOS and Precedence */
599 		if (sk->sk_type == SOCK_STREAM) {
600 			val &= ~INET_ECN_MASK;
601 			val |= inet->tos & INET_ECN_MASK;
602 		}
603 		if (inet->tos != val) {
604 			inet->tos = val;
605 			sk->sk_priority = rt_tos2priority(val);
606 			sk_dst_reset(sk);
607 		}
608 		break;
609 	case IP_TTL:
610 		if (optlen < 1)
611 			goto e_inval;
612 		if (val != -1 && (val < 1 || val > 255))
613 			goto e_inval;
614 		inet->uc_ttl = val;
615 		break;
616 	case IP_HDRINCL:
617 		if (sk->sk_type != SOCK_RAW) {
618 			err = -ENOPROTOOPT;
619 			break;
620 		}
621 		inet->hdrincl = val ? 1 : 0;
622 		break;
623 	case IP_NODEFRAG:
624 		if (sk->sk_type != SOCK_RAW) {
625 			err = -ENOPROTOOPT;
626 			break;
627 		}
628 		inet->nodefrag = val ? 1 : 0;
629 		break;
630 	case IP_MTU_DISCOVER:
631 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_INTERFACE)
632 			goto e_inval;
633 		inet->pmtudisc = val;
634 		break;
635 	case IP_RECVERR:
636 		inet->recverr = !!val;
637 		if (!val)
638 			skb_queue_purge(&sk->sk_error_queue);
639 		break;
640 	case IP_MULTICAST_TTL:
641 		if (sk->sk_type == SOCK_STREAM)
642 			goto e_inval;
643 		if (optlen < 1)
644 			goto e_inval;
645 		if (val == -1)
646 			val = 1;
647 		if (val < 0 || val > 255)
648 			goto e_inval;
649 		inet->mc_ttl = val;
650 		break;
651 	case IP_MULTICAST_LOOP:
652 		if (optlen < 1)
653 			goto e_inval;
654 		inet->mc_loop = !!val;
655 		break;
656 	case IP_UNICAST_IF:
657 	{
658 		struct net_device *dev = NULL;
659 		int ifindex;
660 
661 		if (optlen != sizeof(int))
662 			goto e_inval;
663 
664 		ifindex = (__force int)ntohl((__force __be32)val);
665 		if (ifindex == 0) {
666 			inet->uc_index = 0;
667 			err = 0;
668 			break;
669 		}
670 
671 		dev = dev_get_by_index(sock_net(sk), ifindex);
672 		err = -EADDRNOTAVAIL;
673 		if (!dev)
674 			break;
675 		dev_put(dev);
676 
677 		err = -EINVAL;
678 		if (sk->sk_bound_dev_if)
679 			break;
680 
681 		inet->uc_index = ifindex;
682 		err = 0;
683 		break;
684 	}
685 	case IP_MULTICAST_IF:
686 	{
687 		struct ip_mreqn mreq;
688 		struct net_device *dev = NULL;
689 
690 		if (sk->sk_type == SOCK_STREAM)
691 			goto e_inval;
692 		/*
693 		 *	Check the arguments are allowable
694 		 */
695 
696 		if (optlen < sizeof(struct in_addr))
697 			goto e_inval;
698 
699 		err = -EFAULT;
700 		if (optlen >= sizeof(struct ip_mreqn)) {
701 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
702 				break;
703 		} else {
704 			memset(&mreq, 0, sizeof(mreq));
705 			if (optlen >= sizeof(struct ip_mreq)) {
706 				if (copy_from_user(&mreq, optval,
707 						   sizeof(struct ip_mreq)))
708 					break;
709 			} else if (optlen >= sizeof(struct in_addr)) {
710 				if (copy_from_user(&mreq.imr_address, optval,
711 						   sizeof(struct in_addr)))
712 					break;
713 			}
714 		}
715 
716 		if (!mreq.imr_ifindex) {
717 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
718 				inet->mc_index = 0;
719 				inet->mc_addr  = 0;
720 				err = 0;
721 				break;
722 			}
723 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
724 			if (dev)
725 				mreq.imr_ifindex = dev->ifindex;
726 		} else
727 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
728 
729 
730 		err = -EADDRNOTAVAIL;
731 		if (!dev)
732 			break;
733 		dev_put(dev);
734 
735 		err = -EINVAL;
736 		if (sk->sk_bound_dev_if &&
737 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
738 			break;
739 
740 		inet->mc_index = mreq.imr_ifindex;
741 		inet->mc_addr  = mreq.imr_address.s_addr;
742 		err = 0;
743 		break;
744 	}
745 
746 	case IP_ADD_MEMBERSHIP:
747 	case IP_DROP_MEMBERSHIP:
748 	{
749 		struct ip_mreqn mreq;
750 
751 		err = -EPROTO;
752 		if (inet_sk(sk)->is_icsk)
753 			break;
754 
755 		if (optlen < sizeof(struct ip_mreq))
756 			goto e_inval;
757 		err = -EFAULT;
758 		if (optlen >= sizeof(struct ip_mreqn)) {
759 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
760 				break;
761 		} else {
762 			memset(&mreq, 0, sizeof(mreq));
763 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
764 				break;
765 		}
766 
767 		if (optname == IP_ADD_MEMBERSHIP)
768 			err = ip_mc_join_group(sk, &mreq);
769 		else
770 			err = ip_mc_leave_group(sk, &mreq);
771 		break;
772 	}
773 	case IP_MSFILTER:
774 	{
775 		struct ip_msfilter *msf;
776 
777 		if (optlen < IP_MSFILTER_SIZE(0))
778 			goto e_inval;
779 		if (optlen > sysctl_optmem_max) {
780 			err = -ENOBUFS;
781 			break;
782 		}
783 		msf = kmalloc(optlen, GFP_KERNEL);
784 		if (!msf) {
785 			err = -ENOBUFS;
786 			break;
787 		}
788 		err = -EFAULT;
789 		if (copy_from_user(msf, optval, optlen)) {
790 			kfree(msf);
791 			break;
792 		}
793 		/* numsrc >= (1G-4) overflow in 32 bits */
794 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
795 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
796 			kfree(msf);
797 			err = -ENOBUFS;
798 			break;
799 		}
800 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
801 			kfree(msf);
802 			err = -EINVAL;
803 			break;
804 		}
805 		err = ip_mc_msfilter(sk, msf, 0);
806 		kfree(msf);
807 		break;
808 	}
809 	case IP_BLOCK_SOURCE:
810 	case IP_UNBLOCK_SOURCE:
811 	case IP_ADD_SOURCE_MEMBERSHIP:
812 	case IP_DROP_SOURCE_MEMBERSHIP:
813 	{
814 		struct ip_mreq_source mreqs;
815 		int omode, add;
816 
817 		if (optlen != sizeof(struct ip_mreq_source))
818 			goto e_inval;
819 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
820 			err = -EFAULT;
821 			break;
822 		}
823 		if (optname == IP_BLOCK_SOURCE) {
824 			omode = MCAST_EXCLUDE;
825 			add = 1;
826 		} else if (optname == IP_UNBLOCK_SOURCE) {
827 			omode = MCAST_EXCLUDE;
828 			add = 0;
829 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
830 			struct ip_mreqn mreq;
831 
832 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
833 			mreq.imr_address.s_addr = mreqs.imr_interface;
834 			mreq.imr_ifindex = 0;
835 			err = ip_mc_join_group(sk, &mreq);
836 			if (err && err != -EADDRINUSE)
837 				break;
838 			omode = MCAST_INCLUDE;
839 			add = 1;
840 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
841 			omode = MCAST_INCLUDE;
842 			add = 0;
843 		}
844 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
845 		break;
846 	}
847 	case MCAST_JOIN_GROUP:
848 	case MCAST_LEAVE_GROUP:
849 	{
850 		struct group_req greq;
851 		struct sockaddr_in *psin;
852 		struct ip_mreqn mreq;
853 
854 		if (optlen < sizeof(struct group_req))
855 			goto e_inval;
856 		err = -EFAULT;
857 		if (copy_from_user(&greq, optval, sizeof(greq)))
858 			break;
859 		psin = (struct sockaddr_in *)&greq.gr_group;
860 		if (psin->sin_family != AF_INET)
861 			goto e_inval;
862 		memset(&mreq, 0, sizeof(mreq));
863 		mreq.imr_multiaddr = psin->sin_addr;
864 		mreq.imr_ifindex = greq.gr_interface;
865 
866 		if (optname == MCAST_JOIN_GROUP)
867 			err = ip_mc_join_group(sk, &mreq);
868 		else
869 			err = ip_mc_leave_group(sk, &mreq);
870 		break;
871 	}
872 	case MCAST_JOIN_SOURCE_GROUP:
873 	case MCAST_LEAVE_SOURCE_GROUP:
874 	case MCAST_BLOCK_SOURCE:
875 	case MCAST_UNBLOCK_SOURCE:
876 	{
877 		struct group_source_req greqs;
878 		struct ip_mreq_source mreqs;
879 		struct sockaddr_in *psin;
880 		int omode, add;
881 
882 		if (optlen != sizeof(struct group_source_req))
883 			goto e_inval;
884 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
885 			err = -EFAULT;
886 			break;
887 		}
888 		if (greqs.gsr_group.ss_family != AF_INET ||
889 		    greqs.gsr_source.ss_family != AF_INET) {
890 			err = -EADDRNOTAVAIL;
891 			break;
892 		}
893 		psin = (struct sockaddr_in *)&greqs.gsr_group;
894 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
895 		psin = (struct sockaddr_in *)&greqs.gsr_source;
896 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
897 		mreqs.imr_interface = 0; /* use index for mc_source */
898 
899 		if (optname == MCAST_BLOCK_SOURCE) {
900 			omode = MCAST_EXCLUDE;
901 			add = 1;
902 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
903 			omode = MCAST_EXCLUDE;
904 			add = 0;
905 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
906 			struct ip_mreqn mreq;
907 
908 			psin = (struct sockaddr_in *)&greqs.gsr_group;
909 			mreq.imr_multiaddr = psin->sin_addr;
910 			mreq.imr_address.s_addr = 0;
911 			mreq.imr_ifindex = greqs.gsr_interface;
912 			err = ip_mc_join_group(sk, &mreq);
913 			if (err && err != -EADDRINUSE)
914 				break;
915 			greqs.gsr_interface = mreq.imr_ifindex;
916 			omode = MCAST_INCLUDE;
917 			add = 1;
918 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
919 			omode = MCAST_INCLUDE;
920 			add = 0;
921 		}
922 		err = ip_mc_source(add, omode, sk, &mreqs,
923 				   greqs.gsr_interface);
924 		break;
925 	}
926 	case MCAST_MSFILTER:
927 	{
928 		struct sockaddr_in *psin;
929 		struct ip_msfilter *msf = NULL;
930 		struct group_filter *gsf = NULL;
931 		int msize, i, ifindex;
932 
933 		if (optlen < GROUP_FILTER_SIZE(0))
934 			goto e_inval;
935 		if (optlen > sysctl_optmem_max) {
936 			err = -ENOBUFS;
937 			break;
938 		}
939 		gsf = kmalloc(optlen, GFP_KERNEL);
940 		if (!gsf) {
941 			err = -ENOBUFS;
942 			break;
943 		}
944 		err = -EFAULT;
945 		if (copy_from_user(gsf, optval, optlen))
946 			goto mc_msf_out;
947 
948 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
949 		if (gsf->gf_numsrc >= 0x1ffffff ||
950 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
951 			err = -ENOBUFS;
952 			goto mc_msf_out;
953 		}
954 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
955 			err = -EINVAL;
956 			goto mc_msf_out;
957 		}
958 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
959 		msf = kmalloc(msize, GFP_KERNEL);
960 		if (!msf) {
961 			err = -ENOBUFS;
962 			goto mc_msf_out;
963 		}
964 		ifindex = gsf->gf_interface;
965 		psin = (struct sockaddr_in *)&gsf->gf_group;
966 		if (psin->sin_family != AF_INET) {
967 			err = -EADDRNOTAVAIL;
968 			goto mc_msf_out;
969 		}
970 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
971 		msf->imsf_interface = 0;
972 		msf->imsf_fmode = gsf->gf_fmode;
973 		msf->imsf_numsrc = gsf->gf_numsrc;
974 		err = -EADDRNOTAVAIL;
975 		for (i = 0; i < gsf->gf_numsrc; ++i) {
976 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
977 
978 			if (psin->sin_family != AF_INET)
979 				goto mc_msf_out;
980 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
981 		}
982 		kfree(gsf);
983 		gsf = NULL;
984 
985 		err = ip_mc_msfilter(sk, msf, ifindex);
986 mc_msf_out:
987 		kfree(msf);
988 		kfree(gsf);
989 		break;
990 	}
991 	case IP_MULTICAST_ALL:
992 		if (optlen < 1)
993 			goto e_inval;
994 		if (val != 0 && val != 1)
995 			goto e_inval;
996 		inet->mc_all = val;
997 		break;
998 	case IP_ROUTER_ALERT:
999 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1000 		break;
1001 
1002 	case IP_FREEBIND:
1003 		if (optlen < 1)
1004 			goto e_inval;
1005 		inet->freebind = !!val;
1006 		break;
1007 
1008 	case IP_IPSEC_POLICY:
1009 	case IP_XFRM_POLICY:
1010 		err = -EPERM;
1011 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1012 			break;
1013 		err = xfrm_user_policy(sk, optname, optval, optlen);
1014 		break;
1015 
1016 	case IP_TRANSPARENT:
1017 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1018 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1019 			err = -EPERM;
1020 			break;
1021 		}
1022 		if (optlen < 1)
1023 			goto e_inval;
1024 		inet->transparent = !!val;
1025 		break;
1026 
1027 	case IP_MINTTL:
1028 		if (optlen < 1)
1029 			goto e_inval;
1030 		if (val < 0 || val > 255)
1031 			goto e_inval;
1032 		inet->min_ttl = val;
1033 		break;
1034 
1035 	default:
1036 		err = -ENOPROTOOPT;
1037 		break;
1038 	}
1039 	release_sock(sk);
1040 	return err;
1041 
1042 e_inval:
1043 	release_sock(sk);
1044 	return -EINVAL;
1045 }
1046 
1047 /**
1048  * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1049  * @sk: socket
1050  * @skb: buffer
1051  *
1052  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1053  * destination in skb->cb[] before dst drop.
1054  * This way, receiver doesnt make cache line misses to read rtable.
1055  */
1056 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1057 {
1058 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1059 
1060 	if ((inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) &&
1061 	    skb_rtable(skb)) {
1062 		pktinfo->ipi_ifindex = inet_iif(skb);
1063 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1064 	} else {
1065 		pktinfo->ipi_ifindex = 0;
1066 		pktinfo->ipi_spec_dst.s_addr = 0;
1067 	}
1068 	skb_dst_drop(skb);
1069 }
1070 
1071 int ip_setsockopt(struct sock *sk, int level,
1072 		int optname, char __user *optval, unsigned int optlen)
1073 {
1074 	int err;
1075 
1076 	if (level != SOL_IP)
1077 		return -ENOPROTOOPT;
1078 
1079 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1080 #ifdef CONFIG_NETFILTER
1081 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1082 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1083 			optname != IP_IPSEC_POLICY &&
1084 			optname != IP_XFRM_POLICY &&
1085 			!ip_mroute_opt(optname)) {
1086 		lock_sock(sk);
1087 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1088 		release_sock(sk);
1089 	}
1090 #endif
1091 	return err;
1092 }
1093 EXPORT_SYMBOL(ip_setsockopt);
1094 
1095 #ifdef CONFIG_COMPAT
1096 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1097 			 char __user *optval, unsigned int optlen)
1098 {
1099 	int err;
1100 
1101 	if (level != SOL_IP)
1102 		return -ENOPROTOOPT;
1103 
1104 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1105 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1106 			ip_setsockopt);
1107 
1108 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1109 #ifdef CONFIG_NETFILTER
1110 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1111 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1112 			optname != IP_IPSEC_POLICY &&
1113 			optname != IP_XFRM_POLICY &&
1114 			!ip_mroute_opt(optname)) {
1115 		lock_sock(sk);
1116 		err = compat_nf_setsockopt(sk, PF_INET, optname,
1117 					   optval, optlen);
1118 		release_sock(sk);
1119 	}
1120 #endif
1121 	return err;
1122 }
1123 EXPORT_SYMBOL(compat_ip_setsockopt);
1124 #endif
1125 
1126 /*
1127  *	Get the options. Note for future reference. The GET of IP options gets
1128  *	the _received_ ones. The set sets the _sent_ ones.
1129  */
1130 
1131 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1132 			    char __user *optval, int __user *optlen, unsigned int flags)
1133 {
1134 	struct inet_sock *inet = inet_sk(sk);
1135 	int val;
1136 	int len;
1137 
1138 	if (level != SOL_IP)
1139 		return -EOPNOTSUPP;
1140 
1141 	if (ip_mroute_opt(optname))
1142 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1143 
1144 	if (get_user(len, optlen))
1145 		return -EFAULT;
1146 	if (len < 0)
1147 		return -EINVAL;
1148 
1149 	lock_sock(sk);
1150 
1151 	switch (optname) {
1152 	case IP_OPTIONS:
1153 	{
1154 		unsigned char optbuf[sizeof(struct ip_options)+40];
1155 		struct ip_options *opt = (struct ip_options *)optbuf;
1156 		struct ip_options_rcu *inet_opt;
1157 
1158 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1159 						     sock_owned_by_user(sk));
1160 		opt->optlen = 0;
1161 		if (inet_opt)
1162 			memcpy(optbuf, &inet_opt->opt,
1163 			       sizeof(struct ip_options) +
1164 			       inet_opt->opt.optlen);
1165 		release_sock(sk);
1166 
1167 		if (opt->optlen == 0)
1168 			return put_user(0, optlen);
1169 
1170 		ip_options_undo(opt);
1171 
1172 		len = min_t(unsigned int, len, opt->optlen);
1173 		if (put_user(len, optlen))
1174 			return -EFAULT;
1175 		if (copy_to_user(optval, opt->__data, len))
1176 			return -EFAULT;
1177 		return 0;
1178 	}
1179 	case IP_PKTINFO:
1180 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1181 		break;
1182 	case IP_RECVTTL:
1183 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1184 		break;
1185 	case IP_RECVTOS:
1186 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1187 		break;
1188 	case IP_RECVOPTS:
1189 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1190 		break;
1191 	case IP_RETOPTS:
1192 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1193 		break;
1194 	case IP_PASSSEC:
1195 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1196 		break;
1197 	case IP_RECVORIGDSTADDR:
1198 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1199 		break;
1200 	case IP_TOS:
1201 		val = inet->tos;
1202 		break;
1203 	case IP_TTL:
1204 		val = (inet->uc_ttl == -1 ?
1205 		       sysctl_ip_default_ttl :
1206 		       inet->uc_ttl);
1207 		break;
1208 	case IP_HDRINCL:
1209 		val = inet->hdrincl;
1210 		break;
1211 	case IP_NODEFRAG:
1212 		val = inet->nodefrag;
1213 		break;
1214 	case IP_MTU_DISCOVER:
1215 		val = inet->pmtudisc;
1216 		break;
1217 	case IP_MTU:
1218 	{
1219 		struct dst_entry *dst;
1220 		val = 0;
1221 		dst = sk_dst_get(sk);
1222 		if (dst) {
1223 			val = dst_mtu(dst);
1224 			dst_release(dst);
1225 		}
1226 		if (!val) {
1227 			release_sock(sk);
1228 			return -ENOTCONN;
1229 		}
1230 		break;
1231 	}
1232 	case IP_RECVERR:
1233 		val = inet->recverr;
1234 		break;
1235 	case IP_MULTICAST_TTL:
1236 		val = inet->mc_ttl;
1237 		break;
1238 	case IP_MULTICAST_LOOP:
1239 		val = inet->mc_loop;
1240 		break;
1241 	case IP_UNICAST_IF:
1242 		val = (__force int)htonl((__u32) inet->uc_index);
1243 		break;
1244 	case IP_MULTICAST_IF:
1245 	{
1246 		struct in_addr addr;
1247 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1248 		addr.s_addr = inet->mc_addr;
1249 		release_sock(sk);
1250 
1251 		if (put_user(len, optlen))
1252 			return -EFAULT;
1253 		if (copy_to_user(optval, &addr, len))
1254 			return -EFAULT;
1255 		return 0;
1256 	}
1257 	case IP_MSFILTER:
1258 	{
1259 		struct ip_msfilter msf;
1260 		int err;
1261 
1262 		if (len < IP_MSFILTER_SIZE(0)) {
1263 			release_sock(sk);
1264 			return -EINVAL;
1265 		}
1266 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1267 			release_sock(sk);
1268 			return -EFAULT;
1269 		}
1270 		err = ip_mc_msfget(sk, &msf,
1271 				   (struct ip_msfilter __user *)optval, optlen);
1272 		release_sock(sk);
1273 		return err;
1274 	}
1275 	case MCAST_MSFILTER:
1276 	{
1277 		struct group_filter gsf;
1278 		int err;
1279 
1280 		if (len < GROUP_FILTER_SIZE(0)) {
1281 			release_sock(sk);
1282 			return -EINVAL;
1283 		}
1284 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1285 			release_sock(sk);
1286 			return -EFAULT;
1287 		}
1288 		err = ip_mc_gsfget(sk, &gsf,
1289 				   (struct group_filter __user *)optval,
1290 				   optlen);
1291 		release_sock(sk);
1292 		return err;
1293 	}
1294 	case IP_MULTICAST_ALL:
1295 		val = inet->mc_all;
1296 		break;
1297 	case IP_PKTOPTIONS:
1298 	{
1299 		struct msghdr msg;
1300 
1301 		release_sock(sk);
1302 
1303 		if (sk->sk_type != SOCK_STREAM)
1304 			return -ENOPROTOOPT;
1305 
1306 		msg.msg_control = optval;
1307 		msg.msg_controllen = len;
1308 		msg.msg_flags = flags;
1309 
1310 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1311 			struct in_pktinfo info;
1312 
1313 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1314 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1315 			info.ipi_ifindex = inet->mc_index;
1316 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1317 		}
1318 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1319 			int hlim = inet->mc_ttl;
1320 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1321 		}
1322 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1323 			int tos = inet->rcv_tos;
1324 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1325 		}
1326 		len -= msg.msg_controllen;
1327 		return put_user(len, optlen);
1328 	}
1329 	case IP_FREEBIND:
1330 		val = inet->freebind;
1331 		break;
1332 	case IP_TRANSPARENT:
1333 		val = inet->transparent;
1334 		break;
1335 	case IP_MINTTL:
1336 		val = inet->min_ttl;
1337 		break;
1338 	default:
1339 		release_sock(sk);
1340 		return -ENOPROTOOPT;
1341 	}
1342 	release_sock(sk);
1343 
1344 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1345 		unsigned char ucval = (unsigned char)val;
1346 		len = 1;
1347 		if (put_user(len, optlen))
1348 			return -EFAULT;
1349 		if (copy_to_user(optval, &ucval, 1))
1350 			return -EFAULT;
1351 	} else {
1352 		len = min_t(unsigned int, sizeof(int), len);
1353 		if (put_user(len, optlen))
1354 			return -EFAULT;
1355 		if (copy_to_user(optval, &val, len))
1356 			return -EFAULT;
1357 	}
1358 	return 0;
1359 }
1360 
1361 int ip_getsockopt(struct sock *sk, int level,
1362 		  int optname, char __user *optval, int __user *optlen)
1363 {
1364 	int err;
1365 
1366 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1367 #ifdef CONFIG_NETFILTER
1368 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1369 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1370 			!ip_mroute_opt(optname)) {
1371 		int len;
1372 
1373 		if (get_user(len, optlen))
1374 			return -EFAULT;
1375 
1376 		lock_sock(sk);
1377 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1378 				&len);
1379 		release_sock(sk);
1380 		if (err >= 0)
1381 			err = put_user(len, optlen);
1382 		return err;
1383 	}
1384 #endif
1385 	return err;
1386 }
1387 EXPORT_SYMBOL(ip_getsockopt);
1388 
1389 #ifdef CONFIG_COMPAT
1390 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1391 			 char __user *optval, int __user *optlen)
1392 {
1393 	int err;
1394 
1395 	if (optname == MCAST_MSFILTER)
1396 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1397 			ip_getsockopt);
1398 
1399 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1400 		MSG_CMSG_COMPAT);
1401 
1402 #ifdef CONFIG_NETFILTER
1403 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1404 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1405 			!ip_mroute_opt(optname)) {
1406 		int len;
1407 
1408 		if (get_user(len, optlen))
1409 			return -EFAULT;
1410 
1411 		lock_sock(sk);
1412 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1413 		release_sock(sk);
1414 		if (err >= 0)
1415 			err = put_user(len, optlen);
1416 		return err;
1417 	}
1418 #endif
1419 	return err;
1420 }
1421 EXPORT_SYMBOL(compat_ip_getsockopt);
1422 #endif
1423