xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 92a2c6b2)
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 #include <net/checksum.h>
41 #if IS_ENABLED(CONFIG_IPV6)
42 #include <net/transp_v6.h>
43 #endif
44 #include <net/ip_fib.h>
45 
46 #include <linux/errqueue.h>
47 #include <asm/uaccess.h>
48 
49 /*
50  *	SOL_IP control messages.
51  */
52 
53 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
54 {
55 	struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
56 
57 	info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
58 
59 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
60 }
61 
62 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
63 {
64 	int ttl = ip_hdr(skb)->ttl;
65 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
66 }
67 
68 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
69 {
70 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
71 }
72 
73 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
74 {
75 	if (IPCB(skb)->opt.optlen == 0)
76 		return;
77 
78 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
79 		 ip_hdr(skb) + 1);
80 }
81 
82 
83 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
84 {
85 	unsigned char optbuf[sizeof(struct ip_options) + 40];
86 	struct ip_options *opt = (struct ip_options *)optbuf;
87 
88 	if (IPCB(skb)->opt.optlen == 0)
89 		return;
90 
91 	if (ip_options_echo(opt, skb)) {
92 		msg->msg_flags |= MSG_CTRUNC;
93 		return;
94 	}
95 	ip_options_undo(opt);
96 
97 	put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
98 }
99 
100 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
101 				  int offset)
102 {
103 	__wsum csum = skb->csum;
104 
105 	if (skb->ip_summed != CHECKSUM_COMPLETE)
106 		return;
107 
108 	if (offset != 0)
109 		csum = csum_sub(csum, csum_partial(skb->data, offset, 0));
110 
111 	put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
112 }
113 
114 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
115 {
116 	char *secdata;
117 	u32 seclen, secid;
118 	int err;
119 
120 	err = security_socket_getpeersec_dgram(NULL, skb, &secid);
121 	if (err)
122 		return;
123 
124 	err = security_secid_to_secctx(secid, &secdata, &seclen);
125 	if (err)
126 		return;
127 
128 	put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
129 	security_release_secctx(secdata, seclen);
130 }
131 
132 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
133 {
134 	struct sockaddr_in sin;
135 	const struct iphdr *iph = ip_hdr(skb);
136 	__be16 *ports = (__be16 *)skb_transport_header(skb);
137 
138 	if (skb_transport_offset(skb) + 4 > skb->len)
139 		return;
140 
141 	/* All current transport protocols have the port numbers in the
142 	 * first four bytes of the transport header and this function is
143 	 * written with this assumption in mind.
144 	 */
145 
146 	sin.sin_family = AF_INET;
147 	sin.sin_addr.s_addr = iph->daddr;
148 	sin.sin_port = ports[1];
149 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
150 
151 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
152 }
153 
154 void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb,
155 			 int offset)
156 {
157 	struct inet_sock *inet = inet_sk(skb->sk);
158 	unsigned int flags = inet->cmsg_flags;
159 
160 	/* Ordered by supposed usage frequency */
161 	if (flags & IP_CMSG_PKTINFO) {
162 		ip_cmsg_recv_pktinfo(msg, skb);
163 
164 		flags &= ~IP_CMSG_PKTINFO;
165 		if (!flags)
166 			return;
167 	}
168 
169 	if (flags & IP_CMSG_TTL) {
170 		ip_cmsg_recv_ttl(msg, skb);
171 
172 		flags &= ~IP_CMSG_TTL;
173 		if (!flags)
174 			return;
175 	}
176 
177 	if (flags & IP_CMSG_TOS) {
178 		ip_cmsg_recv_tos(msg, skb);
179 
180 		flags &= ~IP_CMSG_TOS;
181 		if (!flags)
182 			return;
183 	}
184 
185 	if (flags & IP_CMSG_RECVOPTS) {
186 		ip_cmsg_recv_opts(msg, skb);
187 
188 		flags &= ~IP_CMSG_RECVOPTS;
189 		if (!flags)
190 			return;
191 	}
192 
193 	if (flags & IP_CMSG_RETOPTS) {
194 		ip_cmsg_recv_retopts(msg, skb);
195 
196 		flags &= ~IP_CMSG_RETOPTS;
197 		if (!flags)
198 			return;
199 	}
200 
201 	if (flags & IP_CMSG_PASSSEC) {
202 		ip_cmsg_recv_security(msg, skb);
203 
204 		flags &= ~IP_CMSG_PASSSEC;
205 		if (!flags)
206 			return;
207 	}
208 
209 	if (flags & IP_CMSG_ORIGDSTADDR) {
210 		ip_cmsg_recv_dstaddr(msg, skb);
211 
212 		flags &= ~IP_CMSG_ORIGDSTADDR;
213 		if (!flags)
214 			return;
215 	}
216 
217 	if (flags & IP_CMSG_CHECKSUM)
218 		ip_cmsg_recv_checksum(msg, skb, offset);
219 }
220 EXPORT_SYMBOL(ip_cmsg_recv_offset);
221 
222 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc,
223 		 bool allow_ipv6)
224 {
225 	int err, val;
226 	struct cmsghdr *cmsg;
227 
228 	for_each_cmsghdr(cmsg, msg) {
229 		if (!CMSG_OK(msg, cmsg))
230 			return -EINVAL;
231 #if IS_ENABLED(CONFIG_IPV6)
232 		if (allow_ipv6 &&
233 		    cmsg->cmsg_level == SOL_IPV6 &&
234 		    cmsg->cmsg_type == IPV6_PKTINFO) {
235 			struct in6_pktinfo *src_info;
236 
237 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
238 				return -EINVAL;
239 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
240 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
241 				return -EINVAL;
242 			ipc->oif = src_info->ipi6_ifindex;
243 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
244 			continue;
245 		}
246 #endif
247 		if (cmsg->cmsg_level != SOL_IP)
248 			continue;
249 		switch (cmsg->cmsg_type) {
250 		case IP_RETOPTS:
251 			err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
252 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
253 					     err < 40 ? err : 40);
254 			if (err)
255 				return err;
256 			break;
257 		case IP_PKTINFO:
258 		{
259 			struct in_pktinfo *info;
260 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
261 				return -EINVAL;
262 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
263 			ipc->oif = info->ipi_ifindex;
264 			ipc->addr = info->ipi_spec_dst.s_addr;
265 			break;
266 		}
267 		case IP_TTL:
268 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
269 				return -EINVAL;
270 			val = *(int *)CMSG_DATA(cmsg);
271 			if (val < 1 || val > 255)
272 				return -EINVAL;
273 			ipc->ttl = val;
274 			break;
275 		case IP_TOS:
276 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
277 				return -EINVAL;
278 			val = *(int *)CMSG_DATA(cmsg);
279 			if (val < 0 || val > 255)
280 				return -EINVAL;
281 			ipc->tos = val;
282 			ipc->priority = rt_tos2priority(ipc->tos);
283 			break;
284 
285 		default:
286 			return -EINVAL;
287 		}
288 	}
289 	return 0;
290 }
291 
292 
293 /* Special input handler for packets caught by router alert option.
294    They are selected only by protocol field, and then processed likely
295    local ones; but only if someone wants them! Otherwise, router
296    not running rsvpd will kill RSVP.
297 
298    It is user level problem, what it will make with them.
299    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
300    but receiver should be enough clever f.e. to forward mtrace requests,
301    sent to multicast group to reach destination designated router.
302  */
303 struct ip_ra_chain __rcu *ip_ra_chain;
304 static DEFINE_SPINLOCK(ip_ra_lock);
305 
306 
307 static void ip_ra_destroy_rcu(struct rcu_head *head)
308 {
309 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
310 
311 	sock_put(ra->saved_sk);
312 	kfree(ra);
313 }
314 
315 int ip_ra_control(struct sock *sk, unsigned char on,
316 		  void (*destructor)(struct sock *))
317 {
318 	struct ip_ra_chain *ra, *new_ra;
319 	struct ip_ra_chain __rcu **rap;
320 
321 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
322 		return -EINVAL;
323 
324 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
325 
326 	spin_lock_bh(&ip_ra_lock);
327 	for (rap = &ip_ra_chain;
328 	     (ra = rcu_dereference_protected(*rap,
329 			lockdep_is_held(&ip_ra_lock))) != NULL;
330 	     rap = &ra->next) {
331 		if (ra->sk == sk) {
332 			if (on) {
333 				spin_unlock_bh(&ip_ra_lock);
334 				kfree(new_ra);
335 				return -EADDRINUSE;
336 			}
337 			/* dont let ip_call_ra_chain() use sk again */
338 			ra->sk = NULL;
339 			RCU_INIT_POINTER(*rap, ra->next);
340 			spin_unlock_bh(&ip_ra_lock);
341 
342 			if (ra->destructor)
343 				ra->destructor(sk);
344 			/*
345 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
346 			 * period. This guarantee ip_call_ra_chain() dont need
347 			 * to mess with socket refcounts.
348 			 */
349 			ra->saved_sk = sk;
350 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
351 			return 0;
352 		}
353 	}
354 	if (new_ra == NULL) {
355 		spin_unlock_bh(&ip_ra_lock);
356 		return -ENOBUFS;
357 	}
358 	new_ra->sk = sk;
359 	new_ra->destructor = destructor;
360 
361 	RCU_INIT_POINTER(new_ra->next, ra);
362 	rcu_assign_pointer(*rap, new_ra);
363 	sock_hold(sk);
364 	spin_unlock_bh(&ip_ra_lock);
365 
366 	return 0;
367 }
368 
369 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
370 		   __be16 port, u32 info, u8 *payload)
371 {
372 	struct sock_exterr_skb *serr;
373 
374 	skb = skb_clone(skb, GFP_ATOMIC);
375 	if (!skb)
376 		return;
377 
378 	serr = SKB_EXT_ERR(skb);
379 	serr->ee.ee_errno = err;
380 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
381 	serr->ee.ee_type = icmp_hdr(skb)->type;
382 	serr->ee.ee_code = icmp_hdr(skb)->code;
383 	serr->ee.ee_pad = 0;
384 	serr->ee.ee_info = info;
385 	serr->ee.ee_data = 0;
386 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
387 				   skb_network_header(skb);
388 	serr->port = port;
389 
390 	if (skb_pull(skb, payload - skb->data) != NULL) {
391 		skb_reset_transport_header(skb);
392 		if (sock_queue_err_skb(sk, skb) == 0)
393 			return;
394 	}
395 	kfree_skb(skb);
396 }
397 
398 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
399 {
400 	struct inet_sock *inet = inet_sk(sk);
401 	struct sock_exterr_skb *serr;
402 	struct iphdr *iph;
403 	struct sk_buff *skb;
404 
405 	if (!inet->recverr)
406 		return;
407 
408 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
409 	if (!skb)
410 		return;
411 
412 	skb_put(skb, sizeof(struct iphdr));
413 	skb_reset_network_header(skb);
414 	iph = ip_hdr(skb);
415 	iph->daddr = daddr;
416 
417 	serr = SKB_EXT_ERR(skb);
418 	serr->ee.ee_errno = err;
419 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
420 	serr->ee.ee_type = 0;
421 	serr->ee.ee_code = 0;
422 	serr->ee.ee_pad = 0;
423 	serr->ee.ee_info = info;
424 	serr->ee.ee_data = 0;
425 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
426 	serr->port = port;
427 
428 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
429 	skb_reset_transport_header(skb);
430 
431 	if (sock_queue_err_skb(sk, skb))
432 		kfree_skb(skb);
433 }
434 
435 static bool ipv4_pktinfo_prepare_errqueue(const struct sock *sk,
436 					  const struct sk_buff *skb,
437 					  int ee_origin)
438 {
439 	struct in_pktinfo *info = PKTINFO_SKB_CB(skb);
440 
441 	if ((ee_origin != SO_EE_ORIGIN_TIMESTAMPING) ||
442 	    (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
443 	    (!skb->dev))
444 		return false;
445 
446 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
447 	info->ipi_ifindex = skb->dev->ifindex;
448 	return true;
449 }
450 
451 /*
452  *	Handle MSG_ERRQUEUE
453  */
454 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
455 {
456 	struct sock_exterr_skb *serr;
457 	struct sk_buff *skb;
458 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
459 	struct {
460 		struct sock_extended_err ee;
461 		struct sockaddr_in	 offender;
462 	} errhdr;
463 	int err;
464 	int copied;
465 
466 	WARN_ON_ONCE(sk->sk_family == AF_INET6);
467 
468 	err = -EAGAIN;
469 	skb = sock_dequeue_err_skb(sk);
470 	if (skb == NULL)
471 		goto out;
472 
473 	copied = skb->len;
474 	if (copied > len) {
475 		msg->msg_flags |= MSG_TRUNC;
476 		copied = len;
477 	}
478 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
479 	if (err)
480 		goto out_free_skb;
481 
482 	sock_recv_timestamp(msg, sk, skb);
483 
484 	serr = SKB_EXT_ERR(skb);
485 
486 	if (sin && skb->len) {
487 		sin->sin_family = AF_INET;
488 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
489 						   serr->addr_offset);
490 		sin->sin_port = serr->port;
491 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
492 		*addr_len = sizeof(*sin);
493 	}
494 
495 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
496 	sin = &errhdr.offender;
497 	memset(sin, 0, sizeof(*sin));
498 
499 	if (skb->len &&
500 	    (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
501 	     ipv4_pktinfo_prepare_errqueue(sk, skb, serr->ee.ee_origin))) {
502 		sin->sin_family = AF_INET;
503 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
504 		if (inet_sk(sk)->cmsg_flags)
505 			ip_cmsg_recv(msg, skb);
506 	}
507 
508 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
509 
510 	/* Now we could try to dump offended packet options */
511 
512 	msg->msg_flags |= MSG_ERRQUEUE;
513 	err = copied;
514 
515 out_free_skb:
516 	kfree_skb(skb);
517 out:
518 	return err;
519 }
520 
521 
522 /*
523  *	Socket option code for IP. This is the end of the line after any
524  *	TCP,UDP etc options on an IP socket.
525  */
526 
527 static int do_ip_setsockopt(struct sock *sk, int level,
528 			    int optname, char __user *optval, unsigned int optlen)
529 {
530 	struct inet_sock *inet = inet_sk(sk);
531 	int val = 0, err;
532 
533 	switch (optname) {
534 	case IP_PKTINFO:
535 	case IP_RECVTTL:
536 	case IP_RECVOPTS:
537 	case IP_RECVTOS:
538 	case IP_RETOPTS:
539 	case IP_TOS:
540 	case IP_TTL:
541 	case IP_HDRINCL:
542 	case IP_MTU_DISCOVER:
543 	case IP_RECVERR:
544 	case IP_ROUTER_ALERT:
545 	case IP_FREEBIND:
546 	case IP_PASSSEC:
547 	case IP_TRANSPARENT:
548 	case IP_MINTTL:
549 	case IP_NODEFRAG:
550 	case IP_UNICAST_IF:
551 	case IP_MULTICAST_TTL:
552 	case IP_MULTICAST_ALL:
553 	case IP_MULTICAST_LOOP:
554 	case IP_RECVORIGDSTADDR:
555 	case IP_CHECKSUM:
556 		if (optlen >= sizeof(int)) {
557 			if (get_user(val, (int __user *) optval))
558 				return -EFAULT;
559 		} else if (optlen >= sizeof(char)) {
560 			unsigned char ucval;
561 
562 			if (get_user(ucval, (unsigned char __user *) optval))
563 				return -EFAULT;
564 			val = (int) ucval;
565 		}
566 	}
567 
568 	/* If optlen==0, it is equivalent to val == 0 */
569 
570 	if (ip_mroute_opt(optname))
571 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
572 
573 	err = 0;
574 	lock_sock(sk);
575 
576 	switch (optname) {
577 	case IP_OPTIONS:
578 	{
579 		struct ip_options_rcu *old, *opt = NULL;
580 
581 		if (optlen > 40)
582 			goto e_inval;
583 		err = ip_options_get_from_user(sock_net(sk), &opt,
584 					       optval, optlen);
585 		if (err)
586 			break;
587 		old = rcu_dereference_protected(inet->inet_opt,
588 						sock_owned_by_user(sk));
589 		if (inet->is_icsk) {
590 			struct inet_connection_sock *icsk = inet_csk(sk);
591 #if IS_ENABLED(CONFIG_IPV6)
592 			if (sk->sk_family == PF_INET ||
593 			    (!((1 << sk->sk_state) &
594 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
595 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
596 #endif
597 				if (old)
598 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
599 				if (opt)
600 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
601 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
602 #if IS_ENABLED(CONFIG_IPV6)
603 			}
604 #endif
605 		}
606 		rcu_assign_pointer(inet->inet_opt, opt);
607 		if (old)
608 			kfree_rcu(old, rcu);
609 		break;
610 	}
611 	case IP_PKTINFO:
612 		if (val)
613 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
614 		else
615 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
616 		break;
617 	case IP_RECVTTL:
618 		if (val)
619 			inet->cmsg_flags |=  IP_CMSG_TTL;
620 		else
621 			inet->cmsg_flags &= ~IP_CMSG_TTL;
622 		break;
623 	case IP_RECVTOS:
624 		if (val)
625 			inet->cmsg_flags |=  IP_CMSG_TOS;
626 		else
627 			inet->cmsg_flags &= ~IP_CMSG_TOS;
628 		break;
629 	case IP_RECVOPTS:
630 		if (val)
631 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
632 		else
633 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
634 		break;
635 	case IP_RETOPTS:
636 		if (val)
637 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
638 		else
639 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
640 		break;
641 	case IP_PASSSEC:
642 		if (val)
643 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
644 		else
645 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
646 		break;
647 	case IP_RECVORIGDSTADDR:
648 		if (val)
649 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
650 		else
651 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
652 		break;
653 	case IP_CHECKSUM:
654 		if (val) {
655 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
656 				inet_inc_convert_csum(sk);
657 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
658 			}
659 		} else {
660 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
661 				inet_dec_convert_csum(sk);
662 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
663 			}
664 		}
665 		break;
666 	case IP_TOS:	/* This sets both TOS and Precedence */
667 		if (sk->sk_type == SOCK_STREAM) {
668 			val &= ~INET_ECN_MASK;
669 			val |= inet->tos & INET_ECN_MASK;
670 		}
671 		if (inet->tos != val) {
672 			inet->tos = val;
673 			sk->sk_priority = rt_tos2priority(val);
674 			sk_dst_reset(sk);
675 		}
676 		break;
677 	case IP_TTL:
678 		if (optlen < 1)
679 			goto e_inval;
680 		if (val != -1 && (val < 1 || val > 255))
681 			goto e_inval;
682 		inet->uc_ttl = val;
683 		break;
684 	case IP_HDRINCL:
685 		if (sk->sk_type != SOCK_RAW) {
686 			err = -ENOPROTOOPT;
687 			break;
688 		}
689 		inet->hdrincl = val ? 1 : 0;
690 		break;
691 	case IP_NODEFRAG:
692 		if (sk->sk_type != SOCK_RAW) {
693 			err = -ENOPROTOOPT;
694 			break;
695 		}
696 		inet->nodefrag = val ? 1 : 0;
697 		break;
698 	case IP_MTU_DISCOVER:
699 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
700 			goto e_inval;
701 		inet->pmtudisc = val;
702 		break;
703 	case IP_RECVERR:
704 		inet->recverr = !!val;
705 		if (!val)
706 			skb_queue_purge(&sk->sk_error_queue);
707 		break;
708 	case IP_MULTICAST_TTL:
709 		if (sk->sk_type == SOCK_STREAM)
710 			goto e_inval;
711 		if (optlen < 1)
712 			goto e_inval;
713 		if (val == -1)
714 			val = 1;
715 		if (val < 0 || val > 255)
716 			goto e_inval;
717 		inet->mc_ttl = val;
718 		break;
719 	case IP_MULTICAST_LOOP:
720 		if (optlen < 1)
721 			goto e_inval;
722 		inet->mc_loop = !!val;
723 		break;
724 	case IP_UNICAST_IF:
725 	{
726 		struct net_device *dev = NULL;
727 		int ifindex;
728 
729 		if (optlen != sizeof(int))
730 			goto e_inval;
731 
732 		ifindex = (__force int)ntohl((__force __be32)val);
733 		if (ifindex == 0) {
734 			inet->uc_index = 0;
735 			err = 0;
736 			break;
737 		}
738 
739 		dev = dev_get_by_index(sock_net(sk), ifindex);
740 		err = -EADDRNOTAVAIL;
741 		if (!dev)
742 			break;
743 		dev_put(dev);
744 
745 		err = -EINVAL;
746 		if (sk->sk_bound_dev_if)
747 			break;
748 
749 		inet->uc_index = ifindex;
750 		err = 0;
751 		break;
752 	}
753 	case IP_MULTICAST_IF:
754 	{
755 		struct ip_mreqn mreq;
756 		struct net_device *dev = NULL;
757 
758 		if (sk->sk_type == SOCK_STREAM)
759 			goto e_inval;
760 		/*
761 		 *	Check the arguments are allowable
762 		 */
763 
764 		if (optlen < sizeof(struct in_addr))
765 			goto e_inval;
766 
767 		err = -EFAULT;
768 		if (optlen >= sizeof(struct ip_mreqn)) {
769 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
770 				break;
771 		} else {
772 			memset(&mreq, 0, sizeof(mreq));
773 			if (optlen >= sizeof(struct ip_mreq)) {
774 				if (copy_from_user(&mreq, optval,
775 						   sizeof(struct ip_mreq)))
776 					break;
777 			} else if (optlen >= sizeof(struct in_addr)) {
778 				if (copy_from_user(&mreq.imr_address, optval,
779 						   sizeof(struct in_addr)))
780 					break;
781 			}
782 		}
783 
784 		if (!mreq.imr_ifindex) {
785 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
786 				inet->mc_index = 0;
787 				inet->mc_addr  = 0;
788 				err = 0;
789 				break;
790 			}
791 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
792 			if (dev)
793 				mreq.imr_ifindex = dev->ifindex;
794 		} else
795 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
796 
797 
798 		err = -EADDRNOTAVAIL;
799 		if (!dev)
800 			break;
801 		dev_put(dev);
802 
803 		err = -EINVAL;
804 		if (sk->sk_bound_dev_if &&
805 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
806 			break;
807 
808 		inet->mc_index = mreq.imr_ifindex;
809 		inet->mc_addr  = mreq.imr_address.s_addr;
810 		err = 0;
811 		break;
812 	}
813 
814 	case IP_ADD_MEMBERSHIP:
815 	case IP_DROP_MEMBERSHIP:
816 	{
817 		struct ip_mreqn mreq;
818 
819 		err = -EPROTO;
820 		if (inet_sk(sk)->is_icsk)
821 			break;
822 
823 		if (optlen < sizeof(struct ip_mreq))
824 			goto e_inval;
825 		err = -EFAULT;
826 		if (optlen >= sizeof(struct ip_mreqn)) {
827 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
828 				break;
829 		} else {
830 			memset(&mreq, 0, sizeof(mreq));
831 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
832 				break;
833 		}
834 
835 		if (optname == IP_ADD_MEMBERSHIP)
836 			err = ip_mc_join_group(sk, &mreq);
837 		else
838 			err = ip_mc_leave_group(sk, &mreq);
839 		break;
840 	}
841 	case IP_MSFILTER:
842 	{
843 		struct ip_msfilter *msf;
844 
845 		if (optlen < IP_MSFILTER_SIZE(0))
846 			goto e_inval;
847 		if (optlen > sysctl_optmem_max) {
848 			err = -ENOBUFS;
849 			break;
850 		}
851 		msf = kmalloc(optlen, GFP_KERNEL);
852 		if (!msf) {
853 			err = -ENOBUFS;
854 			break;
855 		}
856 		err = -EFAULT;
857 		if (copy_from_user(msf, optval, optlen)) {
858 			kfree(msf);
859 			break;
860 		}
861 		/* numsrc >= (1G-4) overflow in 32 bits */
862 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
863 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
864 			kfree(msf);
865 			err = -ENOBUFS;
866 			break;
867 		}
868 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
869 			kfree(msf);
870 			err = -EINVAL;
871 			break;
872 		}
873 		err = ip_mc_msfilter(sk, msf, 0);
874 		kfree(msf);
875 		break;
876 	}
877 	case IP_BLOCK_SOURCE:
878 	case IP_UNBLOCK_SOURCE:
879 	case IP_ADD_SOURCE_MEMBERSHIP:
880 	case IP_DROP_SOURCE_MEMBERSHIP:
881 	{
882 		struct ip_mreq_source mreqs;
883 		int omode, add;
884 
885 		if (optlen != sizeof(struct ip_mreq_source))
886 			goto e_inval;
887 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
888 			err = -EFAULT;
889 			break;
890 		}
891 		if (optname == IP_BLOCK_SOURCE) {
892 			omode = MCAST_EXCLUDE;
893 			add = 1;
894 		} else if (optname == IP_UNBLOCK_SOURCE) {
895 			omode = MCAST_EXCLUDE;
896 			add = 0;
897 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
898 			struct ip_mreqn mreq;
899 
900 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
901 			mreq.imr_address.s_addr = mreqs.imr_interface;
902 			mreq.imr_ifindex = 0;
903 			err = ip_mc_join_group(sk, &mreq);
904 			if (err && err != -EADDRINUSE)
905 				break;
906 			omode = MCAST_INCLUDE;
907 			add = 1;
908 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
909 			omode = MCAST_INCLUDE;
910 			add = 0;
911 		}
912 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
913 		break;
914 	}
915 	case MCAST_JOIN_GROUP:
916 	case MCAST_LEAVE_GROUP:
917 	{
918 		struct group_req greq;
919 		struct sockaddr_in *psin;
920 		struct ip_mreqn mreq;
921 
922 		if (optlen < sizeof(struct group_req))
923 			goto e_inval;
924 		err = -EFAULT;
925 		if (copy_from_user(&greq, optval, sizeof(greq)))
926 			break;
927 		psin = (struct sockaddr_in *)&greq.gr_group;
928 		if (psin->sin_family != AF_INET)
929 			goto e_inval;
930 		memset(&mreq, 0, sizeof(mreq));
931 		mreq.imr_multiaddr = psin->sin_addr;
932 		mreq.imr_ifindex = greq.gr_interface;
933 
934 		if (optname == MCAST_JOIN_GROUP)
935 			err = ip_mc_join_group(sk, &mreq);
936 		else
937 			err = ip_mc_leave_group(sk, &mreq);
938 		break;
939 	}
940 	case MCAST_JOIN_SOURCE_GROUP:
941 	case MCAST_LEAVE_SOURCE_GROUP:
942 	case MCAST_BLOCK_SOURCE:
943 	case MCAST_UNBLOCK_SOURCE:
944 	{
945 		struct group_source_req greqs;
946 		struct ip_mreq_source mreqs;
947 		struct sockaddr_in *psin;
948 		int omode, add;
949 
950 		if (optlen != sizeof(struct group_source_req))
951 			goto e_inval;
952 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
953 			err = -EFAULT;
954 			break;
955 		}
956 		if (greqs.gsr_group.ss_family != AF_INET ||
957 		    greqs.gsr_source.ss_family != AF_INET) {
958 			err = -EADDRNOTAVAIL;
959 			break;
960 		}
961 		psin = (struct sockaddr_in *)&greqs.gsr_group;
962 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
963 		psin = (struct sockaddr_in *)&greqs.gsr_source;
964 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
965 		mreqs.imr_interface = 0; /* use index for mc_source */
966 
967 		if (optname == MCAST_BLOCK_SOURCE) {
968 			omode = MCAST_EXCLUDE;
969 			add = 1;
970 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
971 			omode = MCAST_EXCLUDE;
972 			add = 0;
973 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
974 			struct ip_mreqn mreq;
975 
976 			psin = (struct sockaddr_in *)&greqs.gsr_group;
977 			mreq.imr_multiaddr = psin->sin_addr;
978 			mreq.imr_address.s_addr = 0;
979 			mreq.imr_ifindex = greqs.gsr_interface;
980 			err = ip_mc_join_group(sk, &mreq);
981 			if (err && err != -EADDRINUSE)
982 				break;
983 			greqs.gsr_interface = mreq.imr_ifindex;
984 			omode = MCAST_INCLUDE;
985 			add = 1;
986 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
987 			omode = MCAST_INCLUDE;
988 			add = 0;
989 		}
990 		err = ip_mc_source(add, omode, sk, &mreqs,
991 				   greqs.gsr_interface);
992 		break;
993 	}
994 	case MCAST_MSFILTER:
995 	{
996 		struct sockaddr_in *psin;
997 		struct ip_msfilter *msf = NULL;
998 		struct group_filter *gsf = NULL;
999 		int msize, i, ifindex;
1000 
1001 		if (optlen < GROUP_FILTER_SIZE(0))
1002 			goto e_inval;
1003 		if (optlen > sysctl_optmem_max) {
1004 			err = -ENOBUFS;
1005 			break;
1006 		}
1007 		gsf = kmalloc(optlen, GFP_KERNEL);
1008 		if (!gsf) {
1009 			err = -ENOBUFS;
1010 			break;
1011 		}
1012 		err = -EFAULT;
1013 		if (copy_from_user(gsf, optval, optlen))
1014 			goto mc_msf_out;
1015 
1016 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1017 		if (gsf->gf_numsrc >= 0x1ffffff ||
1018 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
1019 			err = -ENOBUFS;
1020 			goto mc_msf_out;
1021 		}
1022 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1023 			err = -EINVAL;
1024 			goto mc_msf_out;
1025 		}
1026 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1027 		msf = kmalloc(msize, GFP_KERNEL);
1028 		if (!msf) {
1029 			err = -ENOBUFS;
1030 			goto mc_msf_out;
1031 		}
1032 		ifindex = gsf->gf_interface;
1033 		psin = (struct sockaddr_in *)&gsf->gf_group;
1034 		if (psin->sin_family != AF_INET) {
1035 			err = -EADDRNOTAVAIL;
1036 			goto mc_msf_out;
1037 		}
1038 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1039 		msf->imsf_interface = 0;
1040 		msf->imsf_fmode = gsf->gf_fmode;
1041 		msf->imsf_numsrc = gsf->gf_numsrc;
1042 		err = -EADDRNOTAVAIL;
1043 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1044 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1045 
1046 			if (psin->sin_family != AF_INET)
1047 				goto mc_msf_out;
1048 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1049 		}
1050 		kfree(gsf);
1051 		gsf = NULL;
1052 
1053 		err = ip_mc_msfilter(sk, msf, ifindex);
1054 mc_msf_out:
1055 		kfree(msf);
1056 		kfree(gsf);
1057 		break;
1058 	}
1059 	case IP_MULTICAST_ALL:
1060 		if (optlen < 1)
1061 			goto e_inval;
1062 		if (val != 0 && val != 1)
1063 			goto e_inval;
1064 		inet->mc_all = val;
1065 		break;
1066 	case IP_ROUTER_ALERT:
1067 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1068 		break;
1069 
1070 	case IP_FREEBIND:
1071 		if (optlen < 1)
1072 			goto e_inval;
1073 		inet->freebind = !!val;
1074 		break;
1075 
1076 	case IP_IPSEC_POLICY:
1077 	case IP_XFRM_POLICY:
1078 		err = -EPERM;
1079 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1080 			break;
1081 		err = xfrm_user_policy(sk, optname, optval, optlen);
1082 		break;
1083 
1084 	case IP_TRANSPARENT:
1085 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1086 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1087 			err = -EPERM;
1088 			break;
1089 		}
1090 		if (optlen < 1)
1091 			goto e_inval;
1092 		inet->transparent = !!val;
1093 		break;
1094 
1095 	case IP_MINTTL:
1096 		if (optlen < 1)
1097 			goto e_inval;
1098 		if (val < 0 || val > 255)
1099 			goto e_inval;
1100 		inet->min_ttl = val;
1101 		break;
1102 
1103 	default:
1104 		err = -ENOPROTOOPT;
1105 		break;
1106 	}
1107 	release_sock(sk);
1108 	return err;
1109 
1110 e_inval:
1111 	release_sock(sk);
1112 	return -EINVAL;
1113 }
1114 
1115 /**
1116  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1117  * @sk: socket
1118  * @skb: buffer
1119  *
1120  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1121  * destination in skb->cb[] before dst drop.
1122  * This way, receiver doesn't make cache line misses to read rtable.
1123  */
1124 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1125 {
1126 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1127 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1128 		       ipv6_sk_rxinfo(sk);
1129 
1130 	if (prepare && skb_rtable(skb)) {
1131 		pktinfo->ipi_ifindex = inet_iif(skb);
1132 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1133 	} else {
1134 		pktinfo->ipi_ifindex = 0;
1135 		pktinfo->ipi_spec_dst.s_addr = 0;
1136 	}
1137 	skb_dst_drop(skb);
1138 }
1139 
1140 int ip_setsockopt(struct sock *sk, int level,
1141 		int optname, char __user *optval, unsigned int optlen)
1142 {
1143 	int err;
1144 
1145 	if (level != SOL_IP)
1146 		return -ENOPROTOOPT;
1147 
1148 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1149 #ifdef CONFIG_NETFILTER
1150 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1151 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1152 			optname != IP_IPSEC_POLICY &&
1153 			optname != IP_XFRM_POLICY &&
1154 			!ip_mroute_opt(optname)) {
1155 		lock_sock(sk);
1156 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1157 		release_sock(sk);
1158 	}
1159 #endif
1160 	return err;
1161 }
1162 EXPORT_SYMBOL(ip_setsockopt);
1163 
1164 #ifdef CONFIG_COMPAT
1165 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1166 			 char __user *optval, unsigned int optlen)
1167 {
1168 	int err;
1169 
1170 	if (level != SOL_IP)
1171 		return -ENOPROTOOPT;
1172 
1173 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1174 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1175 			ip_setsockopt);
1176 
1177 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1178 #ifdef CONFIG_NETFILTER
1179 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1180 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1181 			optname != IP_IPSEC_POLICY &&
1182 			optname != IP_XFRM_POLICY &&
1183 			!ip_mroute_opt(optname)) {
1184 		lock_sock(sk);
1185 		err = compat_nf_setsockopt(sk, PF_INET, optname,
1186 					   optval, optlen);
1187 		release_sock(sk);
1188 	}
1189 #endif
1190 	return err;
1191 }
1192 EXPORT_SYMBOL(compat_ip_setsockopt);
1193 #endif
1194 
1195 /*
1196  *	Get the options. Note for future reference. The GET of IP options gets
1197  *	the _received_ ones. The set sets the _sent_ ones.
1198  */
1199 
1200 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1201 			    char __user *optval, int __user *optlen, unsigned int flags)
1202 {
1203 	struct inet_sock *inet = inet_sk(sk);
1204 	int val;
1205 	int len;
1206 
1207 	if (level != SOL_IP)
1208 		return -EOPNOTSUPP;
1209 
1210 	if (ip_mroute_opt(optname))
1211 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1212 
1213 	if (get_user(len, optlen))
1214 		return -EFAULT;
1215 	if (len < 0)
1216 		return -EINVAL;
1217 
1218 	lock_sock(sk);
1219 
1220 	switch (optname) {
1221 	case IP_OPTIONS:
1222 	{
1223 		unsigned char optbuf[sizeof(struct ip_options)+40];
1224 		struct ip_options *opt = (struct ip_options *)optbuf;
1225 		struct ip_options_rcu *inet_opt;
1226 
1227 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1228 						     sock_owned_by_user(sk));
1229 		opt->optlen = 0;
1230 		if (inet_opt)
1231 			memcpy(optbuf, &inet_opt->opt,
1232 			       sizeof(struct ip_options) +
1233 			       inet_opt->opt.optlen);
1234 		release_sock(sk);
1235 
1236 		if (opt->optlen == 0)
1237 			return put_user(0, optlen);
1238 
1239 		ip_options_undo(opt);
1240 
1241 		len = min_t(unsigned int, len, opt->optlen);
1242 		if (put_user(len, optlen))
1243 			return -EFAULT;
1244 		if (copy_to_user(optval, opt->__data, len))
1245 			return -EFAULT;
1246 		return 0;
1247 	}
1248 	case IP_PKTINFO:
1249 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1250 		break;
1251 	case IP_RECVTTL:
1252 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1253 		break;
1254 	case IP_RECVTOS:
1255 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1256 		break;
1257 	case IP_RECVOPTS:
1258 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1259 		break;
1260 	case IP_RETOPTS:
1261 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1262 		break;
1263 	case IP_PASSSEC:
1264 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1265 		break;
1266 	case IP_RECVORIGDSTADDR:
1267 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1268 		break;
1269 	case IP_CHECKSUM:
1270 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1271 		break;
1272 	case IP_TOS:
1273 		val = inet->tos;
1274 		break;
1275 	case IP_TTL:
1276 		val = (inet->uc_ttl == -1 ?
1277 		       sysctl_ip_default_ttl :
1278 		       inet->uc_ttl);
1279 		break;
1280 	case IP_HDRINCL:
1281 		val = inet->hdrincl;
1282 		break;
1283 	case IP_NODEFRAG:
1284 		val = inet->nodefrag;
1285 		break;
1286 	case IP_MTU_DISCOVER:
1287 		val = inet->pmtudisc;
1288 		break;
1289 	case IP_MTU:
1290 	{
1291 		struct dst_entry *dst;
1292 		val = 0;
1293 		dst = sk_dst_get(sk);
1294 		if (dst) {
1295 			val = dst_mtu(dst);
1296 			dst_release(dst);
1297 		}
1298 		if (!val) {
1299 			release_sock(sk);
1300 			return -ENOTCONN;
1301 		}
1302 		break;
1303 	}
1304 	case IP_RECVERR:
1305 		val = inet->recverr;
1306 		break;
1307 	case IP_MULTICAST_TTL:
1308 		val = inet->mc_ttl;
1309 		break;
1310 	case IP_MULTICAST_LOOP:
1311 		val = inet->mc_loop;
1312 		break;
1313 	case IP_UNICAST_IF:
1314 		val = (__force int)htonl((__u32) inet->uc_index);
1315 		break;
1316 	case IP_MULTICAST_IF:
1317 	{
1318 		struct in_addr addr;
1319 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1320 		addr.s_addr = inet->mc_addr;
1321 		release_sock(sk);
1322 
1323 		if (put_user(len, optlen))
1324 			return -EFAULT;
1325 		if (copy_to_user(optval, &addr, len))
1326 			return -EFAULT;
1327 		return 0;
1328 	}
1329 	case IP_MSFILTER:
1330 	{
1331 		struct ip_msfilter msf;
1332 		int err;
1333 
1334 		if (len < IP_MSFILTER_SIZE(0)) {
1335 			release_sock(sk);
1336 			return -EINVAL;
1337 		}
1338 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1339 			release_sock(sk);
1340 			return -EFAULT;
1341 		}
1342 		err = ip_mc_msfget(sk, &msf,
1343 				   (struct ip_msfilter __user *)optval, optlen);
1344 		release_sock(sk);
1345 		return err;
1346 	}
1347 	case MCAST_MSFILTER:
1348 	{
1349 		struct group_filter gsf;
1350 		int err;
1351 
1352 		if (len < GROUP_FILTER_SIZE(0)) {
1353 			release_sock(sk);
1354 			return -EINVAL;
1355 		}
1356 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1357 			release_sock(sk);
1358 			return -EFAULT;
1359 		}
1360 		err = ip_mc_gsfget(sk, &gsf,
1361 				   (struct group_filter __user *)optval,
1362 				   optlen);
1363 		release_sock(sk);
1364 		return err;
1365 	}
1366 	case IP_MULTICAST_ALL:
1367 		val = inet->mc_all;
1368 		break;
1369 	case IP_PKTOPTIONS:
1370 	{
1371 		struct msghdr msg;
1372 
1373 		release_sock(sk);
1374 
1375 		if (sk->sk_type != SOCK_STREAM)
1376 			return -ENOPROTOOPT;
1377 
1378 		msg.msg_control = (__force void *) optval;
1379 		msg.msg_controllen = len;
1380 		msg.msg_flags = flags;
1381 
1382 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1383 			struct in_pktinfo info;
1384 
1385 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1386 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1387 			info.ipi_ifindex = inet->mc_index;
1388 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1389 		}
1390 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1391 			int hlim = inet->mc_ttl;
1392 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1393 		}
1394 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1395 			int tos = inet->rcv_tos;
1396 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1397 		}
1398 		len -= msg.msg_controllen;
1399 		return put_user(len, optlen);
1400 	}
1401 	case IP_FREEBIND:
1402 		val = inet->freebind;
1403 		break;
1404 	case IP_TRANSPARENT:
1405 		val = inet->transparent;
1406 		break;
1407 	case IP_MINTTL:
1408 		val = inet->min_ttl;
1409 		break;
1410 	default:
1411 		release_sock(sk);
1412 		return -ENOPROTOOPT;
1413 	}
1414 	release_sock(sk);
1415 
1416 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1417 		unsigned char ucval = (unsigned char)val;
1418 		len = 1;
1419 		if (put_user(len, optlen))
1420 			return -EFAULT;
1421 		if (copy_to_user(optval, &ucval, 1))
1422 			return -EFAULT;
1423 	} else {
1424 		len = min_t(unsigned int, sizeof(int), len);
1425 		if (put_user(len, optlen))
1426 			return -EFAULT;
1427 		if (copy_to_user(optval, &val, len))
1428 			return -EFAULT;
1429 	}
1430 	return 0;
1431 }
1432 
1433 int ip_getsockopt(struct sock *sk, int level,
1434 		  int optname, char __user *optval, int __user *optlen)
1435 {
1436 	int err;
1437 
1438 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1439 #ifdef CONFIG_NETFILTER
1440 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1441 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1442 			!ip_mroute_opt(optname)) {
1443 		int len;
1444 
1445 		if (get_user(len, optlen))
1446 			return -EFAULT;
1447 
1448 		lock_sock(sk);
1449 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1450 				&len);
1451 		release_sock(sk);
1452 		if (err >= 0)
1453 			err = put_user(len, optlen);
1454 		return err;
1455 	}
1456 #endif
1457 	return err;
1458 }
1459 EXPORT_SYMBOL(ip_getsockopt);
1460 
1461 #ifdef CONFIG_COMPAT
1462 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1463 			 char __user *optval, int __user *optlen)
1464 {
1465 	int err;
1466 
1467 	if (optname == MCAST_MSFILTER)
1468 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1469 			ip_getsockopt);
1470 
1471 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1472 		MSG_CMSG_COMPAT);
1473 
1474 #ifdef CONFIG_NETFILTER
1475 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1476 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1477 			!ip_mroute_opt(optname)) {
1478 		int len;
1479 
1480 		if (get_user(len, optlen))
1481 			return -EFAULT;
1482 
1483 		lock_sock(sk);
1484 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1485 		release_sock(sk);
1486 		if (err >= 0)
1487 			err = put_user(len, optlen);
1488 		return err;
1489 	}
1490 #endif
1491 	return err;
1492 }
1493 EXPORT_SYMBOL(compat_ip_getsockopt);
1494 #endif
1495