xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 79f08d9e)
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)
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 	}
427 
428 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
429 	sin = &errhdr.offender;
430 	sin->sin_family = AF_UNSPEC;
431 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
432 		struct inet_sock *inet = inet_sk(sk);
433 
434 		sin->sin_family = AF_INET;
435 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
436 		sin->sin_port = 0;
437 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
438 		if (inet->cmsg_flags)
439 			ip_cmsg_recv(msg, skb);
440 	}
441 
442 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
443 
444 	/* Now we could try to dump offended packet options */
445 
446 	msg->msg_flags |= MSG_ERRQUEUE;
447 	err = copied;
448 
449 	/* Reset and regenerate socket error */
450 	spin_lock_bh(&sk->sk_error_queue.lock);
451 	sk->sk_err = 0;
452 	skb2 = skb_peek(&sk->sk_error_queue);
453 	if (skb2 != NULL) {
454 		sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
455 		spin_unlock_bh(&sk->sk_error_queue.lock);
456 		sk->sk_error_report(sk);
457 	} else
458 		spin_unlock_bh(&sk->sk_error_queue.lock);
459 
460 out_free_skb:
461 	kfree_skb(skb);
462 out:
463 	return err;
464 }
465 
466 
467 /*
468  *	Socket option code for IP. This is the end of the line after any
469  *	TCP,UDP etc options on an IP socket.
470  */
471 
472 static int do_ip_setsockopt(struct sock *sk, int level,
473 			    int optname, char __user *optval, unsigned int optlen)
474 {
475 	struct inet_sock *inet = inet_sk(sk);
476 	int val = 0, err;
477 
478 	switch (optname) {
479 	case IP_PKTINFO:
480 	case IP_RECVTTL:
481 	case IP_RECVOPTS:
482 	case IP_RECVTOS:
483 	case IP_RETOPTS:
484 	case IP_TOS:
485 	case IP_TTL:
486 	case IP_HDRINCL:
487 	case IP_MTU_DISCOVER:
488 	case IP_RECVERR:
489 	case IP_ROUTER_ALERT:
490 	case IP_FREEBIND:
491 	case IP_PASSSEC:
492 	case IP_TRANSPARENT:
493 	case IP_MINTTL:
494 	case IP_NODEFRAG:
495 	case IP_UNICAST_IF:
496 	case IP_MULTICAST_TTL:
497 	case IP_MULTICAST_ALL:
498 	case IP_MULTICAST_LOOP:
499 	case IP_RECVORIGDSTADDR:
500 		if (optlen >= sizeof(int)) {
501 			if (get_user(val, (int __user *) optval))
502 				return -EFAULT;
503 		} else if (optlen >= sizeof(char)) {
504 			unsigned char ucval;
505 
506 			if (get_user(ucval, (unsigned char __user *) optval))
507 				return -EFAULT;
508 			val = (int) ucval;
509 		}
510 	}
511 
512 	/* If optlen==0, it is equivalent to val == 0 */
513 
514 	if (ip_mroute_opt(optname))
515 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
516 
517 	err = 0;
518 	lock_sock(sk);
519 
520 	switch (optname) {
521 	case IP_OPTIONS:
522 	{
523 		struct ip_options_rcu *old, *opt = NULL;
524 
525 		if (optlen > 40)
526 			goto e_inval;
527 		err = ip_options_get_from_user(sock_net(sk), &opt,
528 					       optval, optlen);
529 		if (err)
530 			break;
531 		old = rcu_dereference_protected(inet->inet_opt,
532 						sock_owned_by_user(sk));
533 		if (inet->is_icsk) {
534 			struct inet_connection_sock *icsk = inet_csk(sk);
535 #if IS_ENABLED(CONFIG_IPV6)
536 			if (sk->sk_family == PF_INET ||
537 			    (!((1 << sk->sk_state) &
538 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
539 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
540 #endif
541 				if (old)
542 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
543 				if (opt)
544 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
545 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
546 #if IS_ENABLED(CONFIG_IPV6)
547 			}
548 #endif
549 		}
550 		rcu_assign_pointer(inet->inet_opt, opt);
551 		if (old)
552 			kfree_rcu(old, rcu);
553 		break;
554 	}
555 	case IP_PKTINFO:
556 		if (val)
557 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
558 		else
559 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
560 		break;
561 	case IP_RECVTTL:
562 		if (val)
563 			inet->cmsg_flags |=  IP_CMSG_TTL;
564 		else
565 			inet->cmsg_flags &= ~IP_CMSG_TTL;
566 		break;
567 	case IP_RECVTOS:
568 		if (val)
569 			inet->cmsg_flags |=  IP_CMSG_TOS;
570 		else
571 			inet->cmsg_flags &= ~IP_CMSG_TOS;
572 		break;
573 	case IP_RECVOPTS:
574 		if (val)
575 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
576 		else
577 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
578 		break;
579 	case IP_RETOPTS:
580 		if (val)
581 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
582 		else
583 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
584 		break;
585 	case IP_PASSSEC:
586 		if (val)
587 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
588 		else
589 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
590 		break;
591 	case IP_RECVORIGDSTADDR:
592 		if (val)
593 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
594 		else
595 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
596 		break;
597 	case IP_TOS:	/* This sets both TOS and Precedence */
598 		if (sk->sk_type == SOCK_STREAM) {
599 			val &= ~INET_ECN_MASK;
600 			val |= inet->tos & INET_ECN_MASK;
601 		}
602 		if (inet->tos != val) {
603 			inet->tos = val;
604 			sk->sk_priority = rt_tos2priority(val);
605 			sk_dst_reset(sk);
606 		}
607 		break;
608 	case IP_TTL:
609 		if (optlen < 1)
610 			goto e_inval;
611 		if (val != -1 && (val < 1 || val > 255))
612 			goto e_inval;
613 		inet->uc_ttl = val;
614 		break;
615 	case IP_HDRINCL:
616 		if (sk->sk_type != SOCK_RAW) {
617 			err = -ENOPROTOOPT;
618 			break;
619 		}
620 		inet->hdrincl = val ? 1 : 0;
621 		break;
622 	case IP_NODEFRAG:
623 		if (sk->sk_type != SOCK_RAW) {
624 			err = -ENOPROTOOPT;
625 			break;
626 		}
627 		inet->nodefrag = val ? 1 : 0;
628 		break;
629 	case IP_MTU_DISCOVER:
630 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_INTERFACE)
631 			goto e_inval;
632 		inet->pmtudisc = val;
633 		break;
634 	case IP_RECVERR:
635 		inet->recverr = !!val;
636 		if (!val)
637 			skb_queue_purge(&sk->sk_error_queue);
638 		break;
639 	case IP_MULTICAST_TTL:
640 		if (sk->sk_type == SOCK_STREAM)
641 			goto e_inval;
642 		if (optlen < 1)
643 			goto e_inval;
644 		if (val == -1)
645 			val = 1;
646 		if (val < 0 || val > 255)
647 			goto e_inval;
648 		inet->mc_ttl = val;
649 		break;
650 	case IP_MULTICAST_LOOP:
651 		if (optlen < 1)
652 			goto e_inval;
653 		inet->mc_loop = !!val;
654 		break;
655 	case IP_UNICAST_IF:
656 	{
657 		struct net_device *dev = NULL;
658 		int ifindex;
659 
660 		if (optlen != sizeof(int))
661 			goto e_inval;
662 
663 		ifindex = (__force int)ntohl((__force __be32)val);
664 		if (ifindex == 0) {
665 			inet->uc_index = 0;
666 			err = 0;
667 			break;
668 		}
669 
670 		dev = dev_get_by_index(sock_net(sk), ifindex);
671 		err = -EADDRNOTAVAIL;
672 		if (!dev)
673 			break;
674 		dev_put(dev);
675 
676 		err = -EINVAL;
677 		if (sk->sk_bound_dev_if)
678 			break;
679 
680 		inet->uc_index = ifindex;
681 		err = 0;
682 		break;
683 	}
684 	case IP_MULTICAST_IF:
685 	{
686 		struct ip_mreqn mreq;
687 		struct net_device *dev = NULL;
688 
689 		if (sk->sk_type == SOCK_STREAM)
690 			goto e_inval;
691 		/*
692 		 *	Check the arguments are allowable
693 		 */
694 
695 		if (optlen < sizeof(struct in_addr))
696 			goto e_inval;
697 
698 		err = -EFAULT;
699 		if (optlen >= sizeof(struct ip_mreqn)) {
700 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
701 				break;
702 		} else {
703 			memset(&mreq, 0, sizeof(mreq));
704 			if (optlen >= sizeof(struct ip_mreq)) {
705 				if (copy_from_user(&mreq, optval,
706 						   sizeof(struct ip_mreq)))
707 					break;
708 			} else if (optlen >= sizeof(struct in_addr)) {
709 				if (copy_from_user(&mreq.imr_address, optval,
710 						   sizeof(struct in_addr)))
711 					break;
712 			}
713 		}
714 
715 		if (!mreq.imr_ifindex) {
716 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
717 				inet->mc_index = 0;
718 				inet->mc_addr  = 0;
719 				err = 0;
720 				break;
721 			}
722 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
723 			if (dev)
724 				mreq.imr_ifindex = dev->ifindex;
725 		} else
726 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
727 
728 
729 		err = -EADDRNOTAVAIL;
730 		if (!dev)
731 			break;
732 		dev_put(dev);
733 
734 		err = -EINVAL;
735 		if (sk->sk_bound_dev_if &&
736 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
737 			break;
738 
739 		inet->mc_index = mreq.imr_ifindex;
740 		inet->mc_addr  = mreq.imr_address.s_addr;
741 		err = 0;
742 		break;
743 	}
744 
745 	case IP_ADD_MEMBERSHIP:
746 	case IP_DROP_MEMBERSHIP:
747 	{
748 		struct ip_mreqn mreq;
749 
750 		err = -EPROTO;
751 		if (inet_sk(sk)->is_icsk)
752 			break;
753 
754 		if (optlen < sizeof(struct ip_mreq))
755 			goto e_inval;
756 		err = -EFAULT;
757 		if (optlen >= sizeof(struct ip_mreqn)) {
758 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
759 				break;
760 		} else {
761 			memset(&mreq, 0, sizeof(mreq));
762 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
763 				break;
764 		}
765 
766 		if (optname == IP_ADD_MEMBERSHIP)
767 			err = ip_mc_join_group(sk, &mreq);
768 		else
769 			err = ip_mc_leave_group(sk, &mreq);
770 		break;
771 	}
772 	case IP_MSFILTER:
773 	{
774 		struct ip_msfilter *msf;
775 
776 		if (optlen < IP_MSFILTER_SIZE(0))
777 			goto e_inval;
778 		if (optlen > sysctl_optmem_max) {
779 			err = -ENOBUFS;
780 			break;
781 		}
782 		msf = kmalloc(optlen, GFP_KERNEL);
783 		if (!msf) {
784 			err = -ENOBUFS;
785 			break;
786 		}
787 		err = -EFAULT;
788 		if (copy_from_user(msf, optval, optlen)) {
789 			kfree(msf);
790 			break;
791 		}
792 		/* numsrc >= (1G-4) overflow in 32 bits */
793 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
794 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
795 			kfree(msf);
796 			err = -ENOBUFS;
797 			break;
798 		}
799 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
800 			kfree(msf);
801 			err = -EINVAL;
802 			break;
803 		}
804 		err = ip_mc_msfilter(sk, msf, 0);
805 		kfree(msf);
806 		break;
807 	}
808 	case IP_BLOCK_SOURCE:
809 	case IP_UNBLOCK_SOURCE:
810 	case IP_ADD_SOURCE_MEMBERSHIP:
811 	case IP_DROP_SOURCE_MEMBERSHIP:
812 	{
813 		struct ip_mreq_source mreqs;
814 		int omode, add;
815 
816 		if (optlen != sizeof(struct ip_mreq_source))
817 			goto e_inval;
818 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
819 			err = -EFAULT;
820 			break;
821 		}
822 		if (optname == IP_BLOCK_SOURCE) {
823 			omode = MCAST_EXCLUDE;
824 			add = 1;
825 		} else if (optname == IP_UNBLOCK_SOURCE) {
826 			omode = MCAST_EXCLUDE;
827 			add = 0;
828 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
829 			struct ip_mreqn mreq;
830 
831 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
832 			mreq.imr_address.s_addr = mreqs.imr_interface;
833 			mreq.imr_ifindex = 0;
834 			err = ip_mc_join_group(sk, &mreq);
835 			if (err && err != -EADDRINUSE)
836 				break;
837 			omode = MCAST_INCLUDE;
838 			add = 1;
839 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
840 			omode = MCAST_INCLUDE;
841 			add = 0;
842 		}
843 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
844 		break;
845 	}
846 	case MCAST_JOIN_GROUP:
847 	case MCAST_LEAVE_GROUP:
848 	{
849 		struct group_req greq;
850 		struct sockaddr_in *psin;
851 		struct ip_mreqn mreq;
852 
853 		if (optlen < sizeof(struct group_req))
854 			goto e_inval;
855 		err = -EFAULT;
856 		if (copy_from_user(&greq, optval, sizeof(greq)))
857 			break;
858 		psin = (struct sockaddr_in *)&greq.gr_group;
859 		if (psin->sin_family != AF_INET)
860 			goto e_inval;
861 		memset(&mreq, 0, sizeof(mreq));
862 		mreq.imr_multiaddr = psin->sin_addr;
863 		mreq.imr_ifindex = greq.gr_interface;
864 
865 		if (optname == MCAST_JOIN_GROUP)
866 			err = ip_mc_join_group(sk, &mreq);
867 		else
868 			err = ip_mc_leave_group(sk, &mreq);
869 		break;
870 	}
871 	case MCAST_JOIN_SOURCE_GROUP:
872 	case MCAST_LEAVE_SOURCE_GROUP:
873 	case MCAST_BLOCK_SOURCE:
874 	case MCAST_UNBLOCK_SOURCE:
875 	{
876 		struct group_source_req greqs;
877 		struct ip_mreq_source mreqs;
878 		struct sockaddr_in *psin;
879 		int omode, add;
880 
881 		if (optlen != sizeof(struct group_source_req))
882 			goto e_inval;
883 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
884 			err = -EFAULT;
885 			break;
886 		}
887 		if (greqs.gsr_group.ss_family != AF_INET ||
888 		    greqs.gsr_source.ss_family != AF_INET) {
889 			err = -EADDRNOTAVAIL;
890 			break;
891 		}
892 		psin = (struct sockaddr_in *)&greqs.gsr_group;
893 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
894 		psin = (struct sockaddr_in *)&greqs.gsr_source;
895 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
896 		mreqs.imr_interface = 0; /* use index for mc_source */
897 
898 		if (optname == MCAST_BLOCK_SOURCE) {
899 			omode = MCAST_EXCLUDE;
900 			add = 1;
901 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
902 			omode = MCAST_EXCLUDE;
903 			add = 0;
904 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
905 			struct ip_mreqn mreq;
906 
907 			psin = (struct sockaddr_in *)&greqs.gsr_group;
908 			mreq.imr_multiaddr = psin->sin_addr;
909 			mreq.imr_address.s_addr = 0;
910 			mreq.imr_ifindex = greqs.gsr_interface;
911 			err = ip_mc_join_group(sk, &mreq);
912 			if (err && err != -EADDRINUSE)
913 				break;
914 			greqs.gsr_interface = mreq.imr_ifindex;
915 			omode = MCAST_INCLUDE;
916 			add = 1;
917 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
918 			omode = MCAST_INCLUDE;
919 			add = 0;
920 		}
921 		err = ip_mc_source(add, omode, sk, &mreqs,
922 				   greqs.gsr_interface);
923 		break;
924 	}
925 	case MCAST_MSFILTER:
926 	{
927 		struct sockaddr_in *psin;
928 		struct ip_msfilter *msf = NULL;
929 		struct group_filter *gsf = NULL;
930 		int msize, i, ifindex;
931 
932 		if (optlen < GROUP_FILTER_SIZE(0))
933 			goto e_inval;
934 		if (optlen > sysctl_optmem_max) {
935 			err = -ENOBUFS;
936 			break;
937 		}
938 		gsf = kmalloc(optlen, GFP_KERNEL);
939 		if (!gsf) {
940 			err = -ENOBUFS;
941 			break;
942 		}
943 		err = -EFAULT;
944 		if (copy_from_user(gsf, optval, optlen))
945 			goto mc_msf_out;
946 
947 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
948 		if (gsf->gf_numsrc >= 0x1ffffff ||
949 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
950 			err = -ENOBUFS;
951 			goto mc_msf_out;
952 		}
953 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
954 			err = -EINVAL;
955 			goto mc_msf_out;
956 		}
957 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
958 		msf = kmalloc(msize, GFP_KERNEL);
959 		if (!msf) {
960 			err = -ENOBUFS;
961 			goto mc_msf_out;
962 		}
963 		ifindex = gsf->gf_interface;
964 		psin = (struct sockaddr_in *)&gsf->gf_group;
965 		if (psin->sin_family != AF_INET) {
966 			err = -EADDRNOTAVAIL;
967 			goto mc_msf_out;
968 		}
969 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
970 		msf->imsf_interface = 0;
971 		msf->imsf_fmode = gsf->gf_fmode;
972 		msf->imsf_numsrc = gsf->gf_numsrc;
973 		err = -EADDRNOTAVAIL;
974 		for (i = 0; i < gsf->gf_numsrc; ++i) {
975 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
976 
977 			if (psin->sin_family != AF_INET)
978 				goto mc_msf_out;
979 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
980 		}
981 		kfree(gsf);
982 		gsf = NULL;
983 
984 		err = ip_mc_msfilter(sk, msf, ifindex);
985 mc_msf_out:
986 		kfree(msf);
987 		kfree(gsf);
988 		break;
989 	}
990 	case IP_MULTICAST_ALL:
991 		if (optlen < 1)
992 			goto e_inval;
993 		if (val != 0 && val != 1)
994 			goto e_inval;
995 		inet->mc_all = val;
996 		break;
997 	case IP_ROUTER_ALERT:
998 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
999 		break;
1000 
1001 	case IP_FREEBIND:
1002 		if (optlen < 1)
1003 			goto e_inval;
1004 		inet->freebind = !!val;
1005 		break;
1006 
1007 	case IP_IPSEC_POLICY:
1008 	case IP_XFRM_POLICY:
1009 		err = -EPERM;
1010 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1011 			break;
1012 		err = xfrm_user_policy(sk, optname, optval, optlen);
1013 		break;
1014 
1015 	case IP_TRANSPARENT:
1016 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1017 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1018 			err = -EPERM;
1019 			break;
1020 		}
1021 		if (optlen < 1)
1022 			goto e_inval;
1023 		inet->transparent = !!val;
1024 		break;
1025 
1026 	case IP_MINTTL:
1027 		if (optlen < 1)
1028 			goto e_inval;
1029 		if (val < 0 || val > 255)
1030 			goto e_inval;
1031 		inet->min_ttl = val;
1032 		break;
1033 
1034 	default:
1035 		err = -ENOPROTOOPT;
1036 		break;
1037 	}
1038 	release_sock(sk);
1039 	return err;
1040 
1041 e_inval:
1042 	release_sock(sk);
1043 	return -EINVAL;
1044 }
1045 
1046 /**
1047  * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1048  * @sk: socket
1049  * @skb: buffer
1050  *
1051  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1052  * destination in skb->cb[] before dst drop.
1053  * This way, receiver doesnt make cache line misses to read rtable.
1054  */
1055 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1056 {
1057 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1058 
1059 	if ((inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) &&
1060 	    skb_rtable(skb)) {
1061 		pktinfo->ipi_ifindex = inet_iif(skb);
1062 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1063 	} else {
1064 		pktinfo->ipi_ifindex = 0;
1065 		pktinfo->ipi_spec_dst.s_addr = 0;
1066 	}
1067 	skb_dst_drop(skb);
1068 }
1069 
1070 int ip_setsockopt(struct sock *sk, int level,
1071 		int optname, char __user *optval, unsigned int optlen)
1072 {
1073 	int err;
1074 
1075 	if (level != SOL_IP)
1076 		return -ENOPROTOOPT;
1077 
1078 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1079 #ifdef CONFIG_NETFILTER
1080 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1081 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1082 			optname != IP_IPSEC_POLICY &&
1083 			optname != IP_XFRM_POLICY &&
1084 			!ip_mroute_opt(optname)) {
1085 		lock_sock(sk);
1086 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1087 		release_sock(sk);
1088 	}
1089 #endif
1090 	return err;
1091 }
1092 EXPORT_SYMBOL(ip_setsockopt);
1093 
1094 #ifdef CONFIG_COMPAT
1095 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1096 			 char __user *optval, unsigned int optlen)
1097 {
1098 	int err;
1099 
1100 	if (level != SOL_IP)
1101 		return -ENOPROTOOPT;
1102 
1103 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1104 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1105 			ip_setsockopt);
1106 
1107 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1108 #ifdef CONFIG_NETFILTER
1109 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1110 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1111 			optname != IP_IPSEC_POLICY &&
1112 			optname != IP_XFRM_POLICY &&
1113 			!ip_mroute_opt(optname)) {
1114 		lock_sock(sk);
1115 		err = compat_nf_setsockopt(sk, PF_INET, optname,
1116 					   optval, optlen);
1117 		release_sock(sk);
1118 	}
1119 #endif
1120 	return err;
1121 }
1122 EXPORT_SYMBOL(compat_ip_setsockopt);
1123 #endif
1124 
1125 /*
1126  *	Get the options. Note for future reference. The GET of IP options gets
1127  *	the _received_ ones. The set sets the _sent_ ones.
1128  */
1129 
1130 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1131 			    char __user *optval, int __user *optlen, unsigned int flags)
1132 {
1133 	struct inet_sock *inet = inet_sk(sk);
1134 	int val;
1135 	int len;
1136 
1137 	if (level != SOL_IP)
1138 		return -EOPNOTSUPP;
1139 
1140 	if (ip_mroute_opt(optname))
1141 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1142 
1143 	if (get_user(len, optlen))
1144 		return -EFAULT;
1145 	if (len < 0)
1146 		return -EINVAL;
1147 
1148 	lock_sock(sk);
1149 
1150 	switch (optname) {
1151 	case IP_OPTIONS:
1152 	{
1153 		unsigned char optbuf[sizeof(struct ip_options)+40];
1154 		struct ip_options *opt = (struct ip_options *)optbuf;
1155 		struct ip_options_rcu *inet_opt;
1156 
1157 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1158 						     sock_owned_by_user(sk));
1159 		opt->optlen = 0;
1160 		if (inet_opt)
1161 			memcpy(optbuf, &inet_opt->opt,
1162 			       sizeof(struct ip_options) +
1163 			       inet_opt->opt.optlen);
1164 		release_sock(sk);
1165 
1166 		if (opt->optlen == 0)
1167 			return put_user(0, optlen);
1168 
1169 		ip_options_undo(opt);
1170 
1171 		len = min_t(unsigned int, len, opt->optlen);
1172 		if (put_user(len, optlen))
1173 			return -EFAULT;
1174 		if (copy_to_user(optval, opt->__data, len))
1175 			return -EFAULT;
1176 		return 0;
1177 	}
1178 	case IP_PKTINFO:
1179 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1180 		break;
1181 	case IP_RECVTTL:
1182 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1183 		break;
1184 	case IP_RECVTOS:
1185 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1186 		break;
1187 	case IP_RECVOPTS:
1188 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1189 		break;
1190 	case IP_RETOPTS:
1191 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1192 		break;
1193 	case IP_PASSSEC:
1194 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1195 		break;
1196 	case IP_RECVORIGDSTADDR:
1197 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1198 		break;
1199 	case IP_TOS:
1200 		val = inet->tos;
1201 		break;
1202 	case IP_TTL:
1203 		val = (inet->uc_ttl == -1 ?
1204 		       sysctl_ip_default_ttl :
1205 		       inet->uc_ttl);
1206 		break;
1207 	case IP_HDRINCL:
1208 		val = inet->hdrincl;
1209 		break;
1210 	case IP_NODEFRAG:
1211 		val = inet->nodefrag;
1212 		break;
1213 	case IP_MTU_DISCOVER:
1214 		val = inet->pmtudisc;
1215 		break;
1216 	case IP_MTU:
1217 	{
1218 		struct dst_entry *dst;
1219 		val = 0;
1220 		dst = sk_dst_get(sk);
1221 		if (dst) {
1222 			val = dst_mtu(dst);
1223 			dst_release(dst);
1224 		}
1225 		if (!val) {
1226 			release_sock(sk);
1227 			return -ENOTCONN;
1228 		}
1229 		break;
1230 	}
1231 	case IP_RECVERR:
1232 		val = inet->recverr;
1233 		break;
1234 	case IP_MULTICAST_TTL:
1235 		val = inet->mc_ttl;
1236 		break;
1237 	case IP_MULTICAST_LOOP:
1238 		val = inet->mc_loop;
1239 		break;
1240 	case IP_UNICAST_IF:
1241 		val = (__force int)htonl((__u32) inet->uc_index);
1242 		break;
1243 	case IP_MULTICAST_IF:
1244 	{
1245 		struct in_addr addr;
1246 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1247 		addr.s_addr = inet->mc_addr;
1248 		release_sock(sk);
1249 
1250 		if (put_user(len, optlen))
1251 			return -EFAULT;
1252 		if (copy_to_user(optval, &addr, len))
1253 			return -EFAULT;
1254 		return 0;
1255 	}
1256 	case IP_MSFILTER:
1257 	{
1258 		struct ip_msfilter msf;
1259 		int err;
1260 
1261 		if (len < IP_MSFILTER_SIZE(0)) {
1262 			release_sock(sk);
1263 			return -EINVAL;
1264 		}
1265 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1266 			release_sock(sk);
1267 			return -EFAULT;
1268 		}
1269 		err = ip_mc_msfget(sk, &msf,
1270 				   (struct ip_msfilter __user *)optval, optlen);
1271 		release_sock(sk);
1272 		return err;
1273 	}
1274 	case MCAST_MSFILTER:
1275 	{
1276 		struct group_filter gsf;
1277 		int err;
1278 
1279 		if (len < GROUP_FILTER_SIZE(0)) {
1280 			release_sock(sk);
1281 			return -EINVAL;
1282 		}
1283 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1284 			release_sock(sk);
1285 			return -EFAULT;
1286 		}
1287 		err = ip_mc_gsfget(sk, &gsf,
1288 				   (struct group_filter __user *)optval,
1289 				   optlen);
1290 		release_sock(sk);
1291 		return err;
1292 	}
1293 	case IP_MULTICAST_ALL:
1294 		val = inet->mc_all;
1295 		break;
1296 	case IP_PKTOPTIONS:
1297 	{
1298 		struct msghdr msg;
1299 
1300 		release_sock(sk);
1301 
1302 		if (sk->sk_type != SOCK_STREAM)
1303 			return -ENOPROTOOPT;
1304 
1305 		msg.msg_control = optval;
1306 		msg.msg_controllen = len;
1307 		msg.msg_flags = flags;
1308 
1309 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1310 			struct in_pktinfo info;
1311 
1312 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1313 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1314 			info.ipi_ifindex = inet->mc_index;
1315 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1316 		}
1317 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1318 			int hlim = inet->mc_ttl;
1319 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1320 		}
1321 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1322 			int tos = inet->rcv_tos;
1323 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1324 		}
1325 		len -= msg.msg_controllen;
1326 		return put_user(len, optlen);
1327 	}
1328 	case IP_FREEBIND:
1329 		val = inet->freebind;
1330 		break;
1331 	case IP_TRANSPARENT:
1332 		val = inet->transparent;
1333 		break;
1334 	case IP_MINTTL:
1335 		val = inet->min_ttl;
1336 		break;
1337 	default:
1338 		release_sock(sk);
1339 		return -ENOPROTOOPT;
1340 	}
1341 	release_sock(sk);
1342 
1343 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1344 		unsigned char ucval = (unsigned char)val;
1345 		len = 1;
1346 		if (put_user(len, optlen))
1347 			return -EFAULT;
1348 		if (copy_to_user(optval, &ucval, 1))
1349 			return -EFAULT;
1350 	} else {
1351 		len = min_t(unsigned int, sizeof(int), len);
1352 		if (put_user(len, optlen))
1353 			return -EFAULT;
1354 		if (copy_to_user(optval, &val, len))
1355 			return -EFAULT;
1356 	}
1357 	return 0;
1358 }
1359 
1360 int ip_getsockopt(struct sock *sk, int level,
1361 		  int optname, char __user *optval, int __user *optlen)
1362 {
1363 	int err;
1364 
1365 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1366 #ifdef CONFIG_NETFILTER
1367 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1368 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1369 			!ip_mroute_opt(optname)) {
1370 		int len;
1371 
1372 		if (get_user(len, optlen))
1373 			return -EFAULT;
1374 
1375 		lock_sock(sk);
1376 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1377 				&len);
1378 		release_sock(sk);
1379 		if (err >= 0)
1380 			err = put_user(len, optlen);
1381 		return err;
1382 	}
1383 #endif
1384 	return err;
1385 }
1386 EXPORT_SYMBOL(ip_getsockopt);
1387 
1388 #ifdef CONFIG_COMPAT
1389 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1390 			 char __user *optval, int __user *optlen)
1391 {
1392 	int err;
1393 
1394 	if (optname == MCAST_MSFILTER)
1395 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1396 			ip_getsockopt);
1397 
1398 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1399 		MSG_CMSG_COMPAT);
1400 
1401 #ifdef CONFIG_NETFILTER
1402 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1403 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1404 			!ip_mroute_opt(optname)) {
1405 		int len;
1406 
1407 		if (get_user(len, optlen))
1408 			return -EFAULT;
1409 
1410 		lock_sock(sk);
1411 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1412 		release_sock(sk);
1413 		if (err >= 0)
1414 			err = put_user(len, optlen);
1415 		return err;
1416 	}
1417 #endif
1418 	return err;
1419 }
1420 EXPORT_SYMBOL(compat_ip_getsockopt);
1421 #endif
1422