xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 95e9fd10)
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;
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 		default:
219 			return -EINVAL;
220 		}
221 	}
222 	return 0;
223 }
224 
225 
226 /* Special input handler for packets caught by router alert option.
227    They are selected only by protocol field, and then processed likely
228    local ones; but only if someone wants them! Otherwise, router
229    not running rsvpd will kill RSVP.
230 
231    It is user level problem, what it will make with them.
232    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
233    but receiver should be enough clever f.e. to forward mtrace requests,
234    sent to multicast group to reach destination designated router.
235  */
236 struct ip_ra_chain __rcu *ip_ra_chain;
237 static DEFINE_SPINLOCK(ip_ra_lock);
238 
239 
240 static void ip_ra_destroy_rcu(struct rcu_head *head)
241 {
242 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
243 
244 	sock_put(ra->saved_sk);
245 	kfree(ra);
246 }
247 
248 int ip_ra_control(struct sock *sk, unsigned char on,
249 		  void (*destructor)(struct sock *))
250 {
251 	struct ip_ra_chain *ra, *new_ra;
252 	struct ip_ra_chain __rcu **rap;
253 
254 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
255 		return -EINVAL;
256 
257 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
258 
259 	spin_lock_bh(&ip_ra_lock);
260 	for (rap = &ip_ra_chain;
261 	     (ra = rcu_dereference_protected(*rap,
262 			lockdep_is_held(&ip_ra_lock))) != NULL;
263 	     rap = &ra->next) {
264 		if (ra->sk == sk) {
265 			if (on) {
266 				spin_unlock_bh(&ip_ra_lock);
267 				kfree(new_ra);
268 				return -EADDRINUSE;
269 			}
270 			/* dont let ip_call_ra_chain() use sk again */
271 			ra->sk = NULL;
272 			rcu_assign_pointer(*rap, ra->next);
273 			spin_unlock_bh(&ip_ra_lock);
274 
275 			if (ra->destructor)
276 				ra->destructor(sk);
277 			/*
278 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
279 			 * period. This guarantee ip_call_ra_chain() dont need
280 			 * to mess with socket refcounts.
281 			 */
282 			ra->saved_sk = sk;
283 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
284 			return 0;
285 		}
286 	}
287 	if (new_ra == NULL) {
288 		spin_unlock_bh(&ip_ra_lock);
289 		return -ENOBUFS;
290 	}
291 	new_ra->sk = sk;
292 	new_ra->destructor = destructor;
293 
294 	new_ra->next = ra;
295 	rcu_assign_pointer(*rap, new_ra);
296 	sock_hold(sk);
297 	spin_unlock_bh(&ip_ra_lock);
298 
299 	return 0;
300 }
301 
302 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
303 		   __be16 port, u32 info, u8 *payload)
304 {
305 	struct sock_exterr_skb *serr;
306 
307 	skb = skb_clone(skb, GFP_ATOMIC);
308 	if (!skb)
309 		return;
310 
311 	serr = SKB_EXT_ERR(skb);
312 	serr->ee.ee_errno = err;
313 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
314 	serr->ee.ee_type = icmp_hdr(skb)->type;
315 	serr->ee.ee_code = icmp_hdr(skb)->code;
316 	serr->ee.ee_pad = 0;
317 	serr->ee.ee_info = info;
318 	serr->ee.ee_data = 0;
319 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
320 				   skb_network_header(skb);
321 	serr->port = port;
322 
323 	if (skb_pull(skb, payload - skb->data) != NULL) {
324 		skb_reset_transport_header(skb);
325 		if (sock_queue_err_skb(sk, skb) == 0)
326 			return;
327 	}
328 	kfree_skb(skb);
329 }
330 
331 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
332 {
333 	struct inet_sock *inet = inet_sk(sk);
334 	struct sock_exterr_skb *serr;
335 	struct iphdr *iph;
336 	struct sk_buff *skb;
337 
338 	if (!inet->recverr)
339 		return;
340 
341 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
342 	if (!skb)
343 		return;
344 
345 	skb_put(skb, sizeof(struct iphdr));
346 	skb_reset_network_header(skb);
347 	iph = ip_hdr(skb);
348 	iph->daddr = daddr;
349 
350 	serr = SKB_EXT_ERR(skb);
351 	serr->ee.ee_errno = err;
352 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
353 	serr->ee.ee_type = 0;
354 	serr->ee.ee_code = 0;
355 	serr->ee.ee_pad = 0;
356 	serr->ee.ee_info = info;
357 	serr->ee.ee_data = 0;
358 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
359 	serr->port = port;
360 
361 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
362 	skb_reset_transport_header(skb);
363 
364 	if (sock_queue_err_skb(sk, skb))
365 		kfree_skb(skb);
366 }
367 
368 /*
369  *	Handle MSG_ERRQUEUE
370  */
371 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
372 {
373 	struct sock_exterr_skb *serr;
374 	struct sk_buff *skb, *skb2;
375 	struct sockaddr_in *sin;
376 	struct {
377 		struct sock_extended_err ee;
378 		struct sockaddr_in	 offender;
379 	} errhdr;
380 	int err;
381 	int copied;
382 
383 	err = -EAGAIN;
384 	skb = skb_dequeue(&sk->sk_error_queue);
385 	if (skb == NULL)
386 		goto out;
387 
388 	copied = skb->len;
389 	if (copied > len) {
390 		msg->msg_flags |= MSG_TRUNC;
391 		copied = len;
392 	}
393 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
394 	if (err)
395 		goto out_free_skb;
396 
397 	sock_recv_timestamp(msg, sk, skb);
398 
399 	serr = SKB_EXT_ERR(skb);
400 
401 	sin = (struct sockaddr_in *)msg->msg_name;
402 	if (sin) {
403 		sin->sin_family = AF_INET;
404 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
405 						   serr->addr_offset);
406 		sin->sin_port = serr->port;
407 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
408 	}
409 
410 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
411 	sin = &errhdr.offender;
412 	sin->sin_family = AF_UNSPEC;
413 	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
414 		struct inet_sock *inet = inet_sk(sk);
415 
416 		sin->sin_family = AF_INET;
417 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
418 		sin->sin_port = 0;
419 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
420 		if (inet->cmsg_flags)
421 			ip_cmsg_recv(msg, skb);
422 	}
423 
424 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
425 
426 	/* Now we could try to dump offended packet options */
427 
428 	msg->msg_flags |= MSG_ERRQUEUE;
429 	err = copied;
430 
431 	/* Reset and regenerate socket error */
432 	spin_lock_bh(&sk->sk_error_queue.lock);
433 	sk->sk_err = 0;
434 	skb2 = skb_peek(&sk->sk_error_queue);
435 	if (skb2 != NULL) {
436 		sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
437 		spin_unlock_bh(&sk->sk_error_queue.lock);
438 		sk->sk_error_report(sk);
439 	} else
440 		spin_unlock_bh(&sk->sk_error_queue.lock);
441 
442 out_free_skb:
443 	kfree_skb(skb);
444 out:
445 	return err;
446 }
447 
448 
449 /*
450  *	Socket option code for IP. This is the end of the line after any
451  *	TCP,UDP etc options on an IP socket.
452  */
453 
454 static int do_ip_setsockopt(struct sock *sk, int level,
455 			    int optname, char __user *optval, unsigned int optlen)
456 {
457 	struct inet_sock *inet = inet_sk(sk);
458 	int val = 0, err;
459 
460 	if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
461 			     (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
462 			     (1<<IP_RETOPTS) | (1<<IP_TOS) |
463 			     (1<<IP_TTL) | (1<<IP_HDRINCL) |
464 			     (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
465 			     (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) |
466 			     (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) |
467 			     (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) ||
468 	    optname == IP_UNICAST_IF ||
469 	    optname == IP_MULTICAST_TTL ||
470 	    optname == IP_MULTICAST_ALL ||
471 	    optname == IP_MULTICAST_LOOP ||
472 	    optname == IP_RECVORIGDSTADDR) {
473 		if (optlen >= sizeof(int)) {
474 			if (get_user(val, (int __user *) optval))
475 				return -EFAULT;
476 		} else if (optlen >= sizeof(char)) {
477 			unsigned char ucval;
478 
479 			if (get_user(ucval, (unsigned char __user *) optval))
480 				return -EFAULT;
481 			val = (int) ucval;
482 		}
483 	}
484 
485 	/* If optlen==0, it is equivalent to val == 0 */
486 
487 	if (ip_mroute_opt(optname))
488 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
489 
490 	err = 0;
491 	lock_sock(sk);
492 
493 	switch (optname) {
494 	case IP_OPTIONS:
495 	{
496 		struct ip_options_rcu *old, *opt = NULL;
497 
498 		if (optlen > 40)
499 			goto e_inval;
500 		err = ip_options_get_from_user(sock_net(sk), &opt,
501 					       optval, optlen);
502 		if (err)
503 			break;
504 		old = rcu_dereference_protected(inet->inet_opt,
505 						sock_owned_by_user(sk));
506 		if (inet->is_icsk) {
507 			struct inet_connection_sock *icsk = inet_csk(sk);
508 #if IS_ENABLED(CONFIG_IPV6)
509 			if (sk->sk_family == PF_INET ||
510 			    (!((1 << sk->sk_state) &
511 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
512 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
513 #endif
514 				if (old)
515 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
516 				if (opt)
517 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
518 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
519 #if IS_ENABLED(CONFIG_IPV6)
520 			}
521 #endif
522 		}
523 		rcu_assign_pointer(inet->inet_opt, opt);
524 		if (old)
525 			kfree_rcu(old, rcu);
526 		break;
527 	}
528 	case IP_PKTINFO:
529 		if (val)
530 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
531 		else
532 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
533 		break;
534 	case IP_RECVTTL:
535 		if (val)
536 			inet->cmsg_flags |=  IP_CMSG_TTL;
537 		else
538 			inet->cmsg_flags &= ~IP_CMSG_TTL;
539 		break;
540 	case IP_RECVTOS:
541 		if (val)
542 			inet->cmsg_flags |=  IP_CMSG_TOS;
543 		else
544 			inet->cmsg_flags &= ~IP_CMSG_TOS;
545 		break;
546 	case IP_RECVOPTS:
547 		if (val)
548 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
549 		else
550 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
551 		break;
552 	case IP_RETOPTS:
553 		if (val)
554 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
555 		else
556 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
557 		break;
558 	case IP_PASSSEC:
559 		if (val)
560 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
561 		else
562 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
563 		break;
564 	case IP_RECVORIGDSTADDR:
565 		if (val)
566 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
567 		else
568 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
569 		break;
570 	case IP_TOS:	/* This sets both TOS and Precedence */
571 		if (sk->sk_type == SOCK_STREAM) {
572 			val &= ~INET_ECN_MASK;
573 			val |= inet->tos & INET_ECN_MASK;
574 		}
575 		if (inet->tos != val) {
576 			inet->tos = val;
577 			sk->sk_priority = rt_tos2priority(val);
578 			sk_dst_reset(sk);
579 		}
580 		break;
581 	case IP_TTL:
582 		if (optlen < 1)
583 			goto e_inval;
584 		if (val != -1 && (val < 0 || val > 255))
585 			goto e_inval;
586 		inet->uc_ttl = val;
587 		break;
588 	case IP_HDRINCL:
589 		if (sk->sk_type != SOCK_RAW) {
590 			err = -ENOPROTOOPT;
591 			break;
592 		}
593 		inet->hdrincl = val ? 1 : 0;
594 		break;
595 	case IP_NODEFRAG:
596 		if (sk->sk_type != SOCK_RAW) {
597 			err = -ENOPROTOOPT;
598 			break;
599 		}
600 		inet->nodefrag = val ? 1 : 0;
601 		break;
602 	case IP_MTU_DISCOVER:
603 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
604 			goto e_inval;
605 		inet->pmtudisc = val;
606 		break;
607 	case IP_RECVERR:
608 		inet->recverr = !!val;
609 		if (!val)
610 			skb_queue_purge(&sk->sk_error_queue);
611 		break;
612 	case IP_MULTICAST_TTL:
613 		if (sk->sk_type == SOCK_STREAM)
614 			goto e_inval;
615 		if (optlen < 1)
616 			goto e_inval;
617 		if (val == -1)
618 			val = 1;
619 		if (val < 0 || val > 255)
620 			goto e_inval;
621 		inet->mc_ttl = val;
622 		break;
623 	case IP_MULTICAST_LOOP:
624 		if (optlen < 1)
625 			goto e_inval;
626 		inet->mc_loop = !!val;
627 		break;
628 	case IP_UNICAST_IF:
629 	{
630 		struct net_device *dev = NULL;
631 		int ifindex;
632 
633 		if (optlen != sizeof(int))
634 			goto e_inval;
635 
636 		ifindex = (__force int)ntohl((__force __be32)val);
637 		if (ifindex == 0) {
638 			inet->uc_index = 0;
639 			err = 0;
640 			break;
641 		}
642 
643 		dev = dev_get_by_index(sock_net(sk), ifindex);
644 		err = -EADDRNOTAVAIL;
645 		if (!dev)
646 			break;
647 		dev_put(dev);
648 
649 		err = -EINVAL;
650 		if (sk->sk_bound_dev_if)
651 			break;
652 
653 		inet->uc_index = ifindex;
654 		err = 0;
655 		break;
656 	}
657 	case IP_MULTICAST_IF:
658 	{
659 		struct ip_mreqn mreq;
660 		struct net_device *dev = NULL;
661 
662 		if (sk->sk_type == SOCK_STREAM)
663 			goto e_inval;
664 		/*
665 		 *	Check the arguments are allowable
666 		 */
667 
668 		if (optlen < sizeof(struct in_addr))
669 			goto e_inval;
670 
671 		err = -EFAULT;
672 		if (optlen >= sizeof(struct ip_mreqn)) {
673 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
674 				break;
675 		} else {
676 			memset(&mreq, 0, sizeof(mreq));
677 			if (optlen >= sizeof(struct ip_mreq)) {
678 				if (copy_from_user(&mreq, optval,
679 						   sizeof(struct ip_mreq)))
680 					break;
681 			} else if (optlen >= sizeof(struct in_addr)) {
682 				if (copy_from_user(&mreq.imr_address, optval,
683 						   sizeof(struct in_addr)))
684 					break;
685 			}
686 		}
687 
688 		if (!mreq.imr_ifindex) {
689 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
690 				inet->mc_index = 0;
691 				inet->mc_addr  = 0;
692 				err = 0;
693 				break;
694 			}
695 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
696 			if (dev)
697 				mreq.imr_ifindex = dev->ifindex;
698 		} else
699 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
700 
701 
702 		err = -EADDRNOTAVAIL;
703 		if (!dev)
704 			break;
705 		dev_put(dev);
706 
707 		err = -EINVAL;
708 		if (sk->sk_bound_dev_if &&
709 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
710 			break;
711 
712 		inet->mc_index = mreq.imr_ifindex;
713 		inet->mc_addr  = mreq.imr_address.s_addr;
714 		err = 0;
715 		break;
716 	}
717 
718 	case IP_ADD_MEMBERSHIP:
719 	case IP_DROP_MEMBERSHIP:
720 	{
721 		struct ip_mreqn mreq;
722 
723 		err = -EPROTO;
724 		if (inet_sk(sk)->is_icsk)
725 			break;
726 
727 		if (optlen < sizeof(struct ip_mreq))
728 			goto e_inval;
729 		err = -EFAULT;
730 		if (optlen >= sizeof(struct ip_mreqn)) {
731 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
732 				break;
733 		} else {
734 			memset(&mreq, 0, sizeof(mreq));
735 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
736 				break;
737 		}
738 
739 		if (optname == IP_ADD_MEMBERSHIP)
740 			err = ip_mc_join_group(sk, &mreq);
741 		else
742 			err = ip_mc_leave_group(sk, &mreq);
743 		break;
744 	}
745 	case IP_MSFILTER:
746 	{
747 		struct ip_msfilter *msf;
748 
749 		if (optlen < IP_MSFILTER_SIZE(0))
750 			goto e_inval;
751 		if (optlen > sysctl_optmem_max) {
752 			err = -ENOBUFS;
753 			break;
754 		}
755 		msf = kmalloc(optlen, GFP_KERNEL);
756 		if (!msf) {
757 			err = -ENOBUFS;
758 			break;
759 		}
760 		err = -EFAULT;
761 		if (copy_from_user(msf, optval, optlen)) {
762 			kfree(msf);
763 			break;
764 		}
765 		/* numsrc >= (1G-4) overflow in 32 bits */
766 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
767 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
768 			kfree(msf);
769 			err = -ENOBUFS;
770 			break;
771 		}
772 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
773 			kfree(msf);
774 			err = -EINVAL;
775 			break;
776 		}
777 		err = ip_mc_msfilter(sk, msf, 0);
778 		kfree(msf);
779 		break;
780 	}
781 	case IP_BLOCK_SOURCE:
782 	case IP_UNBLOCK_SOURCE:
783 	case IP_ADD_SOURCE_MEMBERSHIP:
784 	case IP_DROP_SOURCE_MEMBERSHIP:
785 	{
786 		struct ip_mreq_source mreqs;
787 		int omode, add;
788 
789 		if (optlen != sizeof(struct ip_mreq_source))
790 			goto e_inval;
791 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
792 			err = -EFAULT;
793 			break;
794 		}
795 		if (optname == IP_BLOCK_SOURCE) {
796 			omode = MCAST_EXCLUDE;
797 			add = 1;
798 		} else if (optname == IP_UNBLOCK_SOURCE) {
799 			omode = MCAST_EXCLUDE;
800 			add = 0;
801 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
802 			struct ip_mreqn mreq;
803 
804 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
805 			mreq.imr_address.s_addr = mreqs.imr_interface;
806 			mreq.imr_ifindex = 0;
807 			err = ip_mc_join_group(sk, &mreq);
808 			if (err && err != -EADDRINUSE)
809 				break;
810 			omode = MCAST_INCLUDE;
811 			add = 1;
812 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
813 			omode = MCAST_INCLUDE;
814 			add = 0;
815 		}
816 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
817 		break;
818 	}
819 	case MCAST_JOIN_GROUP:
820 	case MCAST_LEAVE_GROUP:
821 	{
822 		struct group_req greq;
823 		struct sockaddr_in *psin;
824 		struct ip_mreqn mreq;
825 
826 		if (optlen < sizeof(struct group_req))
827 			goto e_inval;
828 		err = -EFAULT;
829 		if (copy_from_user(&greq, optval, sizeof(greq)))
830 			break;
831 		psin = (struct sockaddr_in *)&greq.gr_group;
832 		if (psin->sin_family != AF_INET)
833 			goto e_inval;
834 		memset(&mreq, 0, sizeof(mreq));
835 		mreq.imr_multiaddr = psin->sin_addr;
836 		mreq.imr_ifindex = greq.gr_interface;
837 
838 		if (optname == MCAST_JOIN_GROUP)
839 			err = ip_mc_join_group(sk, &mreq);
840 		else
841 			err = ip_mc_leave_group(sk, &mreq);
842 		break;
843 	}
844 	case MCAST_JOIN_SOURCE_GROUP:
845 	case MCAST_LEAVE_SOURCE_GROUP:
846 	case MCAST_BLOCK_SOURCE:
847 	case MCAST_UNBLOCK_SOURCE:
848 	{
849 		struct group_source_req greqs;
850 		struct ip_mreq_source mreqs;
851 		struct sockaddr_in *psin;
852 		int omode, add;
853 
854 		if (optlen != sizeof(struct group_source_req))
855 			goto e_inval;
856 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
857 			err = -EFAULT;
858 			break;
859 		}
860 		if (greqs.gsr_group.ss_family != AF_INET ||
861 		    greqs.gsr_source.ss_family != AF_INET) {
862 			err = -EADDRNOTAVAIL;
863 			break;
864 		}
865 		psin = (struct sockaddr_in *)&greqs.gsr_group;
866 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
867 		psin = (struct sockaddr_in *)&greqs.gsr_source;
868 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
869 		mreqs.imr_interface = 0; /* use index for mc_source */
870 
871 		if (optname == MCAST_BLOCK_SOURCE) {
872 			omode = MCAST_EXCLUDE;
873 			add = 1;
874 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
875 			omode = MCAST_EXCLUDE;
876 			add = 0;
877 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
878 			struct ip_mreqn mreq;
879 
880 			psin = (struct sockaddr_in *)&greqs.gsr_group;
881 			mreq.imr_multiaddr = psin->sin_addr;
882 			mreq.imr_address.s_addr = 0;
883 			mreq.imr_ifindex = greqs.gsr_interface;
884 			err = ip_mc_join_group(sk, &mreq);
885 			if (err && err != -EADDRINUSE)
886 				break;
887 			greqs.gsr_interface = mreq.imr_ifindex;
888 			omode = MCAST_INCLUDE;
889 			add = 1;
890 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
891 			omode = MCAST_INCLUDE;
892 			add = 0;
893 		}
894 		err = ip_mc_source(add, omode, sk, &mreqs,
895 				   greqs.gsr_interface);
896 		break;
897 	}
898 	case MCAST_MSFILTER:
899 	{
900 		struct sockaddr_in *psin;
901 		struct ip_msfilter *msf = NULL;
902 		struct group_filter *gsf = NULL;
903 		int msize, i, ifindex;
904 
905 		if (optlen < GROUP_FILTER_SIZE(0))
906 			goto e_inval;
907 		if (optlen > sysctl_optmem_max) {
908 			err = -ENOBUFS;
909 			break;
910 		}
911 		gsf = kmalloc(optlen, GFP_KERNEL);
912 		if (!gsf) {
913 			err = -ENOBUFS;
914 			break;
915 		}
916 		err = -EFAULT;
917 		if (copy_from_user(gsf, optval, optlen))
918 			goto mc_msf_out;
919 
920 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
921 		if (gsf->gf_numsrc >= 0x1ffffff ||
922 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
923 			err = -ENOBUFS;
924 			goto mc_msf_out;
925 		}
926 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
927 			err = -EINVAL;
928 			goto mc_msf_out;
929 		}
930 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
931 		msf = kmalloc(msize, GFP_KERNEL);
932 		if (!msf) {
933 			err = -ENOBUFS;
934 			goto mc_msf_out;
935 		}
936 		ifindex = gsf->gf_interface;
937 		psin = (struct sockaddr_in *)&gsf->gf_group;
938 		if (psin->sin_family != AF_INET) {
939 			err = -EADDRNOTAVAIL;
940 			goto mc_msf_out;
941 		}
942 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
943 		msf->imsf_interface = 0;
944 		msf->imsf_fmode = gsf->gf_fmode;
945 		msf->imsf_numsrc = gsf->gf_numsrc;
946 		err = -EADDRNOTAVAIL;
947 		for (i = 0; i < gsf->gf_numsrc; ++i) {
948 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
949 
950 			if (psin->sin_family != AF_INET)
951 				goto mc_msf_out;
952 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
953 		}
954 		kfree(gsf);
955 		gsf = NULL;
956 
957 		err = ip_mc_msfilter(sk, msf, ifindex);
958 mc_msf_out:
959 		kfree(msf);
960 		kfree(gsf);
961 		break;
962 	}
963 	case IP_MULTICAST_ALL:
964 		if (optlen < 1)
965 			goto e_inval;
966 		if (val != 0 && val != 1)
967 			goto e_inval;
968 		inet->mc_all = val;
969 		break;
970 	case IP_ROUTER_ALERT:
971 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
972 		break;
973 
974 	case IP_FREEBIND:
975 		if (optlen < 1)
976 			goto e_inval;
977 		inet->freebind = !!val;
978 		break;
979 
980 	case IP_IPSEC_POLICY:
981 	case IP_XFRM_POLICY:
982 		err = -EPERM;
983 		if (!capable(CAP_NET_ADMIN))
984 			break;
985 		err = xfrm_user_policy(sk, optname, optval, optlen);
986 		break;
987 
988 	case IP_TRANSPARENT:
989 		if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) {
990 			err = -EPERM;
991 			break;
992 		}
993 		if (optlen < 1)
994 			goto e_inval;
995 		inet->transparent = !!val;
996 		break;
997 
998 	case IP_MINTTL:
999 		if (optlen < 1)
1000 			goto e_inval;
1001 		if (val < 0 || val > 255)
1002 			goto e_inval;
1003 		inet->min_ttl = val;
1004 		break;
1005 
1006 	default:
1007 		err = -ENOPROTOOPT;
1008 		break;
1009 	}
1010 	release_sock(sk);
1011 	return err;
1012 
1013 e_inval:
1014 	release_sock(sk);
1015 	return -EINVAL;
1016 }
1017 
1018 /**
1019  * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1020  * @sk: socket
1021  * @skb: buffer
1022  *
1023  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1024  * destination in skb->cb[] before dst drop.
1025  * This way, receiver doesnt make cache line misses to read rtable.
1026  */
1027 void ipv4_pktinfo_prepare(struct sk_buff *skb)
1028 {
1029 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1030 
1031 	if (skb_rtable(skb)) {
1032 		pktinfo->ipi_ifindex = inet_iif(skb);
1033 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1034 	} else {
1035 		pktinfo->ipi_ifindex = 0;
1036 		pktinfo->ipi_spec_dst.s_addr = 0;
1037 	}
1038 	skb_dst_drop(skb);
1039 }
1040 
1041 int ip_setsockopt(struct sock *sk, int level,
1042 		int optname, char __user *optval, unsigned int optlen)
1043 {
1044 	int err;
1045 
1046 	if (level != SOL_IP)
1047 		return -ENOPROTOOPT;
1048 
1049 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1050 #ifdef CONFIG_NETFILTER
1051 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1052 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1053 			optname != IP_IPSEC_POLICY &&
1054 			optname != IP_XFRM_POLICY &&
1055 			!ip_mroute_opt(optname)) {
1056 		lock_sock(sk);
1057 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1058 		release_sock(sk);
1059 	}
1060 #endif
1061 	return err;
1062 }
1063 EXPORT_SYMBOL(ip_setsockopt);
1064 
1065 #ifdef CONFIG_COMPAT
1066 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1067 			 char __user *optval, unsigned int optlen)
1068 {
1069 	int err;
1070 
1071 	if (level != SOL_IP)
1072 		return -ENOPROTOOPT;
1073 
1074 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1075 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1076 			ip_setsockopt);
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 = compat_nf_setsockopt(sk, PF_INET, optname,
1087 					   optval, optlen);
1088 		release_sock(sk);
1089 	}
1090 #endif
1091 	return err;
1092 }
1093 EXPORT_SYMBOL(compat_ip_setsockopt);
1094 #endif
1095 
1096 /*
1097  *	Get the options. Note for future reference. The GET of IP options gets
1098  *	the _received_ ones. The set sets the _sent_ ones.
1099  */
1100 
1101 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1102 			    char __user *optval, int __user *optlen, unsigned int flags)
1103 {
1104 	struct inet_sock *inet = inet_sk(sk);
1105 	int val;
1106 	int len;
1107 
1108 	if (level != SOL_IP)
1109 		return -EOPNOTSUPP;
1110 
1111 	if (ip_mroute_opt(optname))
1112 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1113 
1114 	if (get_user(len, optlen))
1115 		return -EFAULT;
1116 	if (len < 0)
1117 		return -EINVAL;
1118 
1119 	lock_sock(sk);
1120 
1121 	switch (optname) {
1122 	case IP_OPTIONS:
1123 	{
1124 		unsigned char optbuf[sizeof(struct ip_options)+40];
1125 		struct ip_options *opt = (struct ip_options *)optbuf;
1126 		struct ip_options_rcu *inet_opt;
1127 
1128 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1129 						     sock_owned_by_user(sk));
1130 		opt->optlen = 0;
1131 		if (inet_opt)
1132 			memcpy(optbuf, &inet_opt->opt,
1133 			       sizeof(struct ip_options) +
1134 			       inet_opt->opt.optlen);
1135 		release_sock(sk);
1136 
1137 		if (opt->optlen == 0)
1138 			return put_user(0, optlen);
1139 
1140 		ip_options_undo(opt);
1141 
1142 		len = min_t(unsigned int, len, opt->optlen);
1143 		if (put_user(len, optlen))
1144 			return -EFAULT;
1145 		if (copy_to_user(optval, opt->__data, len))
1146 			return -EFAULT;
1147 		return 0;
1148 	}
1149 	case IP_PKTINFO:
1150 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1151 		break;
1152 	case IP_RECVTTL:
1153 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1154 		break;
1155 	case IP_RECVTOS:
1156 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1157 		break;
1158 	case IP_RECVOPTS:
1159 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1160 		break;
1161 	case IP_RETOPTS:
1162 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1163 		break;
1164 	case IP_PASSSEC:
1165 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1166 		break;
1167 	case IP_RECVORIGDSTADDR:
1168 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1169 		break;
1170 	case IP_TOS:
1171 		val = inet->tos;
1172 		break;
1173 	case IP_TTL:
1174 		val = (inet->uc_ttl == -1 ?
1175 		       sysctl_ip_default_ttl :
1176 		       inet->uc_ttl);
1177 		break;
1178 	case IP_HDRINCL:
1179 		val = inet->hdrincl;
1180 		break;
1181 	case IP_NODEFRAG:
1182 		val = inet->nodefrag;
1183 		break;
1184 	case IP_MTU_DISCOVER:
1185 		val = inet->pmtudisc;
1186 		break;
1187 	case IP_MTU:
1188 	{
1189 		struct dst_entry *dst;
1190 		val = 0;
1191 		dst = sk_dst_get(sk);
1192 		if (dst) {
1193 			val = dst_mtu(dst);
1194 			dst_release(dst);
1195 		}
1196 		if (!val) {
1197 			release_sock(sk);
1198 			return -ENOTCONN;
1199 		}
1200 		break;
1201 	}
1202 	case IP_RECVERR:
1203 		val = inet->recverr;
1204 		break;
1205 	case IP_MULTICAST_TTL:
1206 		val = inet->mc_ttl;
1207 		break;
1208 	case IP_MULTICAST_LOOP:
1209 		val = inet->mc_loop;
1210 		break;
1211 	case IP_UNICAST_IF:
1212 		val = (__force int)htonl((__u32) inet->uc_index);
1213 		break;
1214 	case IP_MULTICAST_IF:
1215 	{
1216 		struct in_addr addr;
1217 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1218 		addr.s_addr = inet->mc_addr;
1219 		release_sock(sk);
1220 
1221 		if (put_user(len, optlen))
1222 			return -EFAULT;
1223 		if (copy_to_user(optval, &addr, len))
1224 			return -EFAULT;
1225 		return 0;
1226 	}
1227 	case IP_MSFILTER:
1228 	{
1229 		struct ip_msfilter msf;
1230 		int err;
1231 
1232 		if (len < IP_MSFILTER_SIZE(0)) {
1233 			release_sock(sk);
1234 			return -EINVAL;
1235 		}
1236 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1237 			release_sock(sk);
1238 			return -EFAULT;
1239 		}
1240 		err = ip_mc_msfget(sk, &msf,
1241 				   (struct ip_msfilter __user *)optval, optlen);
1242 		release_sock(sk);
1243 		return err;
1244 	}
1245 	case MCAST_MSFILTER:
1246 	{
1247 		struct group_filter gsf;
1248 		int err;
1249 
1250 		if (len < GROUP_FILTER_SIZE(0)) {
1251 			release_sock(sk);
1252 			return -EINVAL;
1253 		}
1254 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1255 			release_sock(sk);
1256 			return -EFAULT;
1257 		}
1258 		err = ip_mc_gsfget(sk, &gsf,
1259 				   (struct group_filter __user *)optval,
1260 				   optlen);
1261 		release_sock(sk);
1262 		return err;
1263 	}
1264 	case IP_MULTICAST_ALL:
1265 		val = inet->mc_all;
1266 		break;
1267 	case IP_PKTOPTIONS:
1268 	{
1269 		struct msghdr msg;
1270 
1271 		release_sock(sk);
1272 
1273 		if (sk->sk_type != SOCK_STREAM)
1274 			return -ENOPROTOOPT;
1275 
1276 		msg.msg_control = optval;
1277 		msg.msg_controllen = len;
1278 		msg.msg_flags = flags;
1279 
1280 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1281 			struct in_pktinfo info;
1282 
1283 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1284 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1285 			info.ipi_ifindex = inet->mc_index;
1286 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1287 		}
1288 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1289 			int hlim = inet->mc_ttl;
1290 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1291 		}
1292 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1293 			int tos = inet->rcv_tos;
1294 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1295 		}
1296 		len -= msg.msg_controllen;
1297 		return put_user(len, optlen);
1298 	}
1299 	case IP_FREEBIND:
1300 		val = inet->freebind;
1301 		break;
1302 	case IP_TRANSPARENT:
1303 		val = inet->transparent;
1304 		break;
1305 	case IP_MINTTL:
1306 		val = inet->min_ttl;
1307 		break;
1308 	default:
1309 		release_sock(sk);
1310 		return -ENOPROTOOPT;
1311 	}
1312 	release_sock(sk);
1313 
1314 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1315 		unsigned char ucval = (unsigned char)val;
1316 		len = 1;
1317 		if (put_user(len, optlen))
1318 			return -EFAULT;
1319 		if (copy_to_user(optval, &ucval, 1))
1320 			return -EFAULT;
1321 	} else {
1322 		len = min_t(unsigned int, sizeof(int), len);
1323 		if (put_user(len, optlen))
1324 			return -EFAULT;
1325 		if (copy_to_user(optval, &val, len))
1326 			return -EFAULT;
1327 	}
1328 	return 0;
1329 }
1330 
1331 int ip_getsockopt(struct sock *sk, int level,
1332 		  int optname, char __user *optval, int __user *optlen)
1333 {
1334 	int err;
1335 
1336 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1337 #ifdef CONFIG_NETFILTER
1338 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1339 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1340 			!ip_mroute_opt(optname)) {
1341 		int len;
1342 
1343 		if (get_user(len, optlen))
1344 			return -EFAULT;
1345 
1346 		lock_sock(sk);
1347 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1348 				&len);
1349 		release_sock(sk);
1350 		if (err >= 0)
1351 			err = put_user(len, optlen);
1352 		return err;
1353 	}
1354 #endif
1355 	return err;
1356 }
1357 EXPORT_SYMBOL(ip_getsockopt);
1358 
1359 #ifdef CONFIG_COMPAT
1360 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1361 			 char __user *optval, int __user *optlen)
1362 {
1363 	int err;
1364 
1365 	if (optname == MCAST_MSFILTER)
1366 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1367 			ip_getsockopt);
1368 
1369 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1370 		MSG_CMSG_COMPAT);
1371 
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 = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1383 		release_sock(sk);
1384 		if (err >= 0)
1385 			err = put_user(len, optlen);
1386 		return err;
1387 	}
1388 #endif
1389 	return err;
1390 }
1391 EXPORT_SYMBOL(compat_ip_getsockopt);
1392 #endif
1393