xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 2596e07a)
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 
253 			/* Our caller is responsible for freeing ipc->opt */
254 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
255 					     err < 40 ? err : 40);
256 			if (err)
257 				return err;
258 			break;
259 		case IP_PKTINFO:
260 		{
261 			struct in_pktinfo *info;
262 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
263 				return -EINVAL;
264 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
265 			ipc->oif = info->ipi_ifindex;
266 			ipc->addr = info->ipi_spec_dst.s_addr;
267 			break;
268 		}
269 		case IP_TTL:
270 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
271 				return -EINVAL;
272 			val = *(int *)CMSG_DATA(cmsg);
273 			if (val < 1 || val > 255)
274 				return -EINVAL;
275 			ipc->ttl = val;
276 			break;
277 		case IP_TOS:
278 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
279 				return -EINVAL;
280 			val = *(int *)CMSG_DATA(cmsg);
281 			if (val < 0 || val > 255)
282 				return -EINVAL;
283 			ipc->tos = val;
284 			ipc->priority = rt_tos2priority(ipc->tos);
285 			break;
286 
287 		default:
288 			return -EINVAL;
289 		}
290 	}
291 	return 0;
292 }
293 
294 
295 /* Special input handler for packets caught by router alert option.
296    They are selected only by protocol field, and then processed likely
297    local ones; but only if someone wants them! Otherwise, router
298    not running rsvpd will kill RSVP.
299 
300    It is user level problem, what it will make with them.
301    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
302    but receiver should be enough clever f.e. to forward mtrace requests,
303    sent to multicast group to reach destination designated router.
304  */
305 struct ip_ra_chain __rcu *ip_ra_chain;
306 static DEFINE_SPINLOCK(ip_ra_lock);
307 
308 
309 static void ip_ra_destroy_rcu(struct rcu_head *head)
310 {
311 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
312 
313 	sock_put(ra->saved_sk);
314 	kfree(ra);
315 }
316 
317 int ip_ra_control(struct sock *sk, unsigned char on,
318 		  void (*destructor)(struct sock *))
319 {
320 	struct ip_ra_chain *ra, *new_ra;
321 	struct ip_ra_chain __rcu **rap;
322 
323 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
324 		return -EINVAL;
325 
326 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
327 
328 	spin_lock_bh(&ip_ra_lock);
329 	for (rap = &ip_ra_chain;
330 	     (ra = rcu_dereference_protected(*rap,
331 			lockdep_is_held(&ip_ra_lock))) != NULL;
332 	     rap = &ra->next) {
333 		if (ra->sk == sk) {
334 			if (on) {
335 				spin_unlock_bh(&ip_ra_lock);
336 				kfree(new_ra);
337 				return -EADDRINUSE;
338 			}
339 			/* dont let ip_call_ra_chain() use sk again */
340 			ra->sk = NULL;
341 			RCU_INIT_POINTER(*rap, ra->next);
342 			spin_unlock_bh(&ip_ra_lock);
343 
344 			if (ra->destructor)
345 				ra->destructor(sk);
346 			/*
347 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
348 			 * period. This guarantee ip_call_ra_chain() dont need
349 			 * to mess with socket refcounts.
350 			 */
351 			ra->saved_sk = sk;
352 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
353 			return 0;
354 		}
355 	}
356 	if (!new_ra) {
357 		spin_unlock_bh(&ip_ra_lock);
358 		return -ENOBUFS;
359 	}
360 	new_ra->sk = sk;
361 	new_ra->destructor = destructor;
362 
363 	RCU_INIT_POINTER(new_ra->next, ra);
364 	rcu_assign_pointer(*rap, new_ra);
365 	sock_hold(sk);
366 	spin_unlock_bh(&ip_ra_lock);
367 
368 	return 0;
369 }
370 
371 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
372 		   __be16 port, u32 info, u8 *payload)
373 {
374 	struct sock_exterr_skb *serr;
375 
376 	skb = skb_clone(skb, GFP_ATOMIC);
377 	if (!skb)
378 		return;
379 
380 	serr = SKB_EXT_ERR(skb);
381 	serr->ee.ee_errno = err;
382 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
383 	serr->ee.ee_type = icmp_hdr(skb)->type;
384 	serr->ee.ee_code = icmp_hdr(skb)->code;
385 	serr->ee.ee_pad = 0;
386 	serr->ee.ee_info = info;
387 	serr->ee.ee_data = 0;
388 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
389 				   skb_network_header(skb);
390 	serr->port = port;
391 
392 	if (skb_pull(skb, payload - skb->data)) {
393 		skb_reset_transport_header(skb);
394 		if (sock_queue_err_skb(sk, skb) == 0)
395 			return;
396 	}
397 	kfree_skb(skb);
398 }
399 
400 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
401 {
402 	struct inet_sock *inet = inet_sk(sk);
403 	struct sock_exterr_skb *serr;
404 	struct iphdr *iph;
405 	struct sk_buff *skb;
406 
407 	if (!inet->recverr)
408 		return;
409 
410 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
411 	if (!skb)
412 		return;
413 
414 	skb_put(skb, sizeof(struct iphdr));
415 	skb_reset_network_header(skb);
416 	iph = ip_hdr(skb);
417 	iph->daddr = daddr;
418 
419 	serr = SKB_EXT_ERR(skb);
420 	serr->ee.ee_errno = err;
421 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
422 	serr->ee.ee_type = 0;
423 	serr->ee.ee_code = 0;
424 	serr->ee.ee_pad = 0;
425 	serr->ee.ee_info = info;
426 	serr->ee.ee_data = 0;
427 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
428 	serr->port = port;
429 
430 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
431 	skb_reset_transport_header(skb);
432 
433 	if (sock_queue_err_skb(sk, skb))
434 		kfree_skb(skb);
435 }
436 
437 /* For some errors we have valid addr_offset even with zero payload and
438  * zero port. Also, addr_offset should be supported if port is set.
439  */
440 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
441 {
442 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
443 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
444 }
445 
446 /* IPv4 supports cmsg on all imcp errors and some timestamps
447  *
448  * Timestamp code paths do not initialize the fields expected by cmsg:
449  * the PKTINFO fields in skb->cb[]. Fill those in here.
450  */
451 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
452 				       struct sk_buff *skb,
453 				       int ee_origin)
454 {
455 	struct in_pktinfo *info;
456 
457 	if (ee_origin == SO_EE_ORIGIN_ICMP)
458 		return true;
459 
460 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
461 		return false;
462 
463 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
464 	 * timestamp with egress dev. Not possible for packets without dev
465 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
466 	 */
467 	if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
468 	    (!skb->dev))
469 		return false;
470 
471 	info = PKTINFO_SKB_CB(skb);
472 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
473 	info->ipi_ifindex = skb->dev->ifindex;
474 	return true;
475 }
476 
477 /*
478  *	Handle MSG_ERRQUEUE
479  */
480 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
481 {
482 	struct sock_exterr_skb *serr;
483 	struct sk_buff *skb;
484 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
485 	struct {
486 		struct sock_extended_err ee;
487 		struct sockaddr_in	 offender;
488 	} errhdr;
489 	int err;
490 	int copied;
491 
492 	WARN_ON_ONCE(sk->sk_family == AF_INET6);
493 
494 	err = -EAGAIN;
495 	skb = sock_dequeue_err_skb(sk);
496 	if (!skb)
497 		goto out;
498 
499 	copied = skb->len;
500 	if (copied > len) {
501 		msg->msg_flags |= MSG_TRUNC;
502 		copied = len;
503 	}
504 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
505 	if (err)
506 		goto out_free_skb;
507 
508 	sock_recv_timestamp(msg, sk, skb);
509 
510 	serr = SKB_EXT_ERR(skb);
511 
512 	if (sin && ipv4_datagram_support_addr(serr)) {
513 		sin->sin_family = AF_INET;
514 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
515 						   serr->addr_offset);
516 		sin->sin_port = serr->port;
517 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
518 		*addr_len = sizeof(*sin);
519 	}
520 
521 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
522 	sin = &errhdr.offender;
523 	memset(sin, 0, sizeof(*sin));
524 
525 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
526 		sin->sin_family = AF_INET;
527 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
528 		if (inet_sk(sk)->cmsg_flags)
529 			ip_cmsg_recv(msg, skb);
530 	}
531 
532 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
533 
534 	/* Now we could try to dump offended packet options */
535 
536 	msg->msg_flags |= MSG_ERRQUEUE;
537 	err = copied;
538 
539 out_free_skb:
540 	kfree_skb(skb);
541 out:
542 	return err;
543 }
544 
545 
546 /*
547  *	Socket option code for IP. This is the end of the line after any
548  *	TCP,UDP etc options on an IP socket.
549  */
550 static bool setsockopt_needs_rtnl(int optname)
551 {
552 	switch (optname) {
553 	case IP_ADD_MEMBERSHIP:
554 	case IP_ADD_SOURCE_MEMBERSHIP:
555 	case IP_BLOCK_SOURCE:
556 	case IP_DROP_MEMBERSHIP:
557 	case IP_DROP_SOURCE_MEMBERSHIP:
558 	case IP_MSFILTER:
559 	case IP_UNBLOCK_SOURCE:
560 	case MCAST_BLOCK_SOURCE:
561 	case MCAST_MSFILTER:
562 	case MCAST_JOIN_GROUP:
563 	case MCAST_JOIN_SOURCE_GROUP:
564 	case MCAST_LEAVE_GROUP:
565 	case MCAST_LEAVE_SOURCE_GROUP:
566 	case MCAST_UNBLOCK_SOURCE:
567 		return true;
568 	}
569 	return false;
570 }
571 
572 static int do_ip_setsockopt(struct sock *sk, int level,
573 			    int optname, char __user *optval, unsigned int optlen)
574 {
575 	struct inet_sock *inet = inet_sk(sk);
576 	int val = 0, err;
577 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
578 
579 	switch (optname) {
580 	case IP_PKTINFO:
581 	case IP_RECVTTL:
582 	case IP_RECVOPTS:
583 	case IP_RECVTOS:
584 	case IP_RETOPTS:
585 	case IP_TOS:
586 	case IP_TTL:
587 	case IP_HDRINCL:
588 	case IP_MTU_DISCOVER:
589 	case IP_RECVERR:
590 	case IP_ROUTER_ALERT:
591 	case IP_FREEBIND:
592 	case IP_PASSSEC:
593 	case IP_TRANSPARENT:
594 	case IP_MINTTL:
595 	case IP_NODEFRAG:
596 	case IP_BIND_ADDRESS_NO_PORT:
597 	case IP_UNICAST_IF:
598 	case IP_MULTICAST_TTL:
599 	case IP_MULTICAST_ALL:
600 	case IP_MULTICAST_LOOP:
601 	case IP_RECVORIGDSTADDR:
602 	case IP_CHECKSUM:
603 		if (optlen >= sizeof(int)) {
604 			if (get_user(val, (int __user *) optval))
605 				return -EFAULT;
606 		} else if (optlen >= sizeof(char)) {
607 			unsigned char ucval;
608 
609 			if (get_user(ucval, (unsigned char __user *) optval))
610 				return -EFAULT;
611 			val = (int) ucval;
612 		}
613 	}
614 
615 	/* If optlen==0, it is equivalent to val == 0 */
616 
617 	if (ip_mroute_opt(optname))
618 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
619 
620 	err = 0;
621 	if (needs_rtnl)
622 		rtnl_lock();
623 	lock_sock(sk);
624 
625 	switch (optname) {
626 	case IP_OPTIONS:
627 	{
628 		struct ip_options_rcu *old, *opt = NULL;
629 
630 		if (optlen > 40)
631 			goto e_inval;
632 		err = ip_options_get_from_user(sock_net(sk), &opt,
633 					       optval, optlen);
634 		if (err)
635 			break;
636 		old = rcu_dereference_protected(inet->inet_opt,
637 						sock_owned_by_user(sk));
638 		if (inet->is_icsk) {
639 			struct inet_connection_sock *icsk = inet_csk(sk);
640 #if IS_ENABLED(CONFIG_IPV6)
641 			if (sk->sk_family == PF_INET ||
642 			    (!((1 << sk->sk_state) &
643 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
644 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
645 #endif
646 				if (old)
647 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
648 				if (opt)
649 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
650 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
651 #if IS_ENABLED(CONFIG_IPV6)
652 			}
653 #endif
654 		}
655 		rcu_assign_pointer(inet->inet_opt, opt);
656 		if (old)
657 			kfree_rcu(old, rcu);
658 		break;
659 	}
660 	case IP_PKTINFO:
661 		if (val)
662 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
663 		else
664 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
665 		break;
666 	case IP_RECVTTL:
667 		if (val)
668 			inet->cmsg_flags |=  IP_CMSG_TTL;
669 		else
670 			inet->cmsg_flags &= ~IP_CMSG_TTL;
671 		break;
672 	case IP_RECVTOS:
673 		if (val)
674 			inet->cmsg_flags |=  IP_CMSG_TOS;
675 		else
676 			inet->cmsg_flags &= ~IP_CMSG_TOS;
677 		break;
678 	case IP_RECVOPTS:
679 		if (val)
680 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
681 		else
682 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
683 		break;
684 	case IP_RETOPTS:
685 		if (val)
686 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
687 		else
688 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
689 		break;
690 	case IP_PASSSEC:
691 		if (val)
692 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
693 		else
694 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
695 		break;
696 	case IP_RECVORIGDSTADDR:
697 		if (val)
698 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
699 		else
700 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
701 		break;
702 	case IP_CHECKSUM:
703 		if (val) {
704 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
705 				inet_inc_convert_csum(sk);
706 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
707 			}
708 		} else {
709 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
710 				inet_dec_convert_csum(sk);
711 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
712 			}
713 		}
714 		break;
715 	case IP_TOS:	/* This sets both TOS and Precedence */
716 		if (sk->sk_type == SOCK_STREAM) {
717 			val &= ~INET_ECN_MASK;
718 			val |= inet->tos & INET_ECN_MASK;
719 		}
720 		if (inet->tos != val) {
721 			inet->tos = val;
722 			sk->sk_priority = rt_tos2priority(val);
723 			sk_dst_reset(sk);
724 		}
725 		break;
726 	case IP_TTL:
727 		if (optlen < 1)
728 			goto e_inval;
729 		if (val != -1 && (val < 1 || val > 255))
730 			goto e_inval;
731 		inet->uc_ttl = val;
732 		break;
733 	case IP_HDRINCL:
734 		if (sk->sk_type != SOCK_RAW) {
735 			err = -ENOPROTOOPT;
736 			break;
737 		}
738 		inet->hdrincl = val ? 1 : 0;
739 		break;
740 	case IP_NODEFRAG:
741 		if (sk->sk_type != SOCK_RAW) {
742 			err = -ENOPROTOOPT;
743 			break;
744 		}
745 		inet->nodefrag = val ? 1 : 0;
746 		break;
747 	case IP_BIND_ADDRESS_NO_PORT:
748 		inet->bind_address_no_port = val ? 1 : 0;
749 		break;
750 	case IP_MTU_DISCOVER:
751 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
752 			goto e_inval;
753 		inet->pmtudisc = val;
754 		break;
755 	case IP_RECVERR:
756 		inet->recverr = !!val;
757 		if (!val)
758 			skb_queue_purge(&sk->sk_error_queue);
759 		break;
760 	case IP_MULTICAST_TTL:
761 		if (sk->sk_type == SOCK_STREAM)
762 			goto e_inval;
763 		if (optlen < 1)
764 			goto e_inval;
765 		if (val == -1)
766 			val = 1;
767 		if (val < 0 || val > 255)
768 			goto e_inval;
769 		inet->mc_ttl = val;
770 		break;
771 	case IP_MULTICAST_LOOP:
772 		if (optlen < 1)
773 			goto e_inval;
774 		inet->mc_loop = !!val;
775 		break;
776 	case IP_UNICAST_IF:
777 	{
778 		struct net_device *dev = NULL;
779 		int ifindex;
780 
781 		if (optlen != sizeof(int))
782 			goto e_inval;
783 
784 		ifindex = (__force int)ntohl((__force __be32)val);
785 		if (ifindex == 0) {
786 			inet->uc_index = 0;
787 			err = 0;
788 			break;
789 		}
790 
791 		dev = dev_get_by_index(sock_net(sk), ifindex);
792 		err = -EADDRNOTAVAIL;
793 		if (!dev)
794 			break;
795 		dev_put(dev);
796 
797 		err = -EINVAL;
798 		if (sk->sk_bound_dev_if)
799 			break;
800 
801 		inet->uc_index = ifindex;
802 		err = 0;
803 		break;
804 	}
805 	case IP_MULTICAST_IF:
806 	{
807 		struct ip_mreqn mreq;
808 		struct net_device *dev = NULL;
809 
810 		if (sk->sk_type == SOCK_STREAM)
811 			goto e_inval;
812 		/*
813 		 *	Check the arguments are allowable
814 		 */
815 
816 		if (optlen < sizeof(struct in_addr))
817 			goto e_inval;
818 
819 		err = -EFAULT;
820 		if (optlen >= sizeof(struct ip_mreqn)) {
821 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
822 				break;
823 		} else {
824 			memset(&mreq, 0, sizeof(mreq));
825 			if (optlen >= sizeof(struct ip_mreq)) {
826 				if (copy_from_user(&mreq, optval,
827 						   sizeof(struct ip_mreq)))
828 					break;
829 			} else if (optlen >= sizeof(struct in_addr)) {
830 				if (copy_from_user(&mreq.imr_address, optval,
831 						   sizeof(struct in_addr)))
832 					break;
833 			}
834 		}
835 
836 		if (!mreq.imr_ifindex) {
837 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
838 				inet->mc_index = 0;
839 				inet->mc_addr  = 0;
840 				err = 0;
841 				break;
842 			}
843 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
844 			if (dev)
845 				mreq.imr_ifindex = dev->ifindex;
846 		} else
847 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
848 
849 
850 		err = -EADDRNOTAVAIL;
851 		if (!dev)
852 			break;
853 		dev_put(dev);
854 
855 		err = -EINVAL;
856 		if (sk->sk_bound_dev_if &&
857 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
858 			break;
859 
860 		inet->mc_index = mreq.imr_ifindex;
861 		inet->mc_addr  = mreq.imr_address.s_addr;
862 		err = 0;
863 		break;
864 	}
865 
866 	case IP_ADD_MEMBERSHIP:
867 	case IP_DROP_MEMBERSHIP:
868 	{
869 		struct ip_mreqn mreq;
870 
871 		err = -EPROTO;
872 		if (inet_sk(sk)->is_icsk)
873 			break;
874 
875 		if (optlen < sizeof(struct ip_mreq))
876 			goto e_inval;
877 		err = -EFAULT;
878 		if (optlen >= sizeof(struct ip_mreqn)) {
879 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
880 				break;
881 		} else {
882 			memset(&mreq, 0, sizeof(mreq));
883 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
884 				break;
885 		}
886 
887 		if (optname == IP_ADD_MEMBERSHIP)
888 			err = ip_mc_join_group(sk, &mreq);
889 		else
890 			err = ip_mc_leave_group(sk, &mreq);
891 		break;
892 	}
893 	case IP_MSFILTER:
894 	{
895 		struct ip_msfilter *msf;
896 
897 		if (optlen < IP_MSFILTER_SIZE(0))
898 			goto e_inval;
899 		if (optlen > sysctl_optmem_max) {
900 			err = -ENOBUFS;
901 			break;
902 		}
903 		msf = kmalloc(optlen, GFP_KERNEL);
904 		if (!msf) {
905 			err = -ENOBUFS;
906 			break;
907 		}
908 		err = -EFAULT;
909 		if (copy_from_user(msf, optval, optlen)) {
910 			kfree(msf);
911 			break;
912 		}
913 		/* numsrc >= (1G-4) overflow in 32 bits */
914 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
915 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
916 			kfree(msf);
917 			err = -ENOBUFS;
918 			break;
919 		}
920 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
921 			kfree(msf);
922 			err = -EINVAL;
923 			break;
924 		}
925 		err = ip_mc_msfilter(sk, msf, 0);
926 		kfree(msf);
927 		break;
928 	}
929 	case IP_BLOCK_SOURCE:
930 	case IP_UNBLOCK_SOURCE:
931 	case IP_ADD_SOURCE_MEMBERSHIP:
932 	case IP_DROP_SOURCE_MEMBERSHIP:
933 	{
934 		struct ip_mreq_source mreqs;
935 		int omode, add;
936 
937 		if (optlen != sizeof(struct ip_mreq_source))
938 			goto e_inval;
939 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
940 			err = -EFAULT;
941 			break;
942 		}
943 		if (optname == IP_BLOCK_SOURCE) {
944 			omode = MCAST_EXCLUDE;
945 			add = 1;
946 		} else if (optname == IP_UNBLOCK_SOURCE) {
947 			omode = MCAST_EXCLUDE;
948 			add = 0;
949 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
950 			struct ip_mreqn mreq;
951 
952 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
953 			mreq.imr_address.s_addr = mreqs.imr_interface;
954 			mreq.imr_ifindex = 0;
955 			err = ip_mc_join_group(sk, &mreq);
956 			if (err && err != -EADDRINUSE)
957 				break;
958 			omode = MCAST_INCLUDE;
959 			add = 1;
960 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
961 			omode = MCAST_INCLUDE;
962 			add = 0;
963 		}
964 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
965 		break;
966 	}
967 	case MCAST_JOIN_GROUP:
968 	case MCAST_LEAVE_GROUP:
969 	{
970 		struct group_req greq;
971 		struct sockaddr_in *psin;
972 		struct ip_mreqn mreq;
973 
974 		if (optlen < sizeof(struct group_req))
975 			goto e_inval;
976 		err = -EFAULT;
977 		if (copy_from_user(&greq, optval, sizeof(greq)))
978 			break;
979 		psin = (struct sockaddr_in *)&greq.gr_group;
980 		if (psin->sin_family != AF_INET)
981 			goto e_inval;
982 		memset(&mreq, 0, sizeof(mreq));
983 		mreq.imr_multiaddr = psin->sin_addr;
984 		mreq.imr_ifindex = greq.gr_interface;
985 
986 		if (optname == MCAST_JOIN_GROUP)
987 			err = ip_mc_join_group(sk, &mreq);
988 		else
989 			err = ip_mc_leave_group(sk, &mreq);
990 		break;
991 	}
992 	case MCAST_JOIN_SOURCE_GROUP:
993 	case MCAST_LEAVE_SOURCE_GROUP:
994 	case MCAST_BLOCK_SOURCE:
995 	case MCAST_UNBLOCK_SOURCE:
996 	{
997 		struct group_source_req greqs;
998 		struct ip_mreq_source mreqs;
999 		struct sockaddr_in *psin;
1000 		int omode, add;
1001 
1002 		if (optlen != sizeof(struct group_source_req))
1003 			goto e_inval;
1004 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1005 			err = -EFAULT;
1006 			break;
1007 		}
1008 		if (greqs.gsr_group.ss_family != AF_INET ||
1009 		    greqs.gsr_source.ss_family != AF_INET) {
1010 			err = -EADDRNOTAVAIL;
1011 			break;
1012 		}
1013 		psin = (struct sockaddr_in *)&greqs.gsr_group;
1014 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1015 		psin = (struct sockaddr_in *)&greqs.gsr_source;
1016 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1017 		mreqs.imr_interface = 0; /* use index for mc_source */
1018 
1019 		if (optname == MCAST_BLOCK_SOURCE) {
1020 			omode = MCAST_EXCLUDE;
1021 			add = 1;
1022 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
1023 			omode = MCAST_EXCLUDE;
1024 			add = 0;
1025 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1026 			struct ip_mreqn mreq;
1027 
1028 			psin = (struct sockaddr_in *)&greqs.gsr_group;
1029 			mreq.imr_multiaddr = psin->sin_addr;
1030 			mreq.imr_address.s_addr = 0;
1031 			mreq.imr_ifindex = greqs.gsr_interface;
1032 			err = ip_mc_join_group(sk, &mreq);
1033 			if (err && err != -EADDRINUSE)
1034 				break;
1035 			greqs.gsr_interface = mreq.imr_ifindex;
1036 			omode = MCAST_INCLUDE;
1037 			add = 1;
1038 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
1039 			omode = MCAST_INCLUDE;
1040 			add = 0;
1041 		}
1042 		err = ip_mc_source(add, omode, sk, &mreqs,
1043 				   greqs.gsr_interface);
1044 		break;
1045 	}
1046 	case MCAST_MSFILTER:
1047 	{
1048 		struct sockaddr_in *psin;
1049 		struct ip_msfilter *msf = NULL;
1050 		struct group_filter *gsf = NULL;
1051 		int msize, i, ifindex;
1052 
1053 		if (optlen < GROUP_FILTER_SIZE(0))
1054 			goto e_inval;
1055 		if (optlen > sysctl_optmem_max) {
1056 			err = -ENOBUFS;
1057 			break;
1058 		}
1059 		gsf = kmalloc(optlen, GFP_KERNEL);
1060 		if (!gsf) {
1061 			err = -ENOBUFS;
1062 			break;
1063 		}
1064 		err = -EFAULT;
1065 		if (copy_from_user(gsf, optval, optlen))
1066 			goto mc_msf_out;
1067 
1068 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1069 		if (gsf->gf_numsrc >= 0x1ffffff ||
1070 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
1071 			err = -ENOBUFS;
1072 			goto mc_msf_out;
1073 		}
1074 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1075 			err = -EINVAL;
1076 			goto mc_msf_out;
1077 		}
1078 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1079 		msf = kmalloc(msize, GFP_KERNEL);
1080 		if (!msf) {
1081 			err = -ENOBUFS;
1082 			goto mc_msf_out;
1083 		}
1084 		ifindex = gsf->gf_interface;
1085 		psin = (struct sockaddr_in *)&gsf->gf_group;
1086 		if (psin->sin_family != AF_INET) {
1087 			err = -EADDRNOTAVAIL;
1088 			goto mc_msf_out;
1089 		}
1090 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1091 		msf->imsf_interface = 0;
1092 		msf->imsf_fmode = gsf->gf_fmode;
1093 		msf->imsf_numsrc = gsf->gf_numsrc;
1094 		err = -EADDRNOTAVAIL;
1095 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1096 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1097 
1098 			if (psin->sin_family != AF_INET)
1099 				goto mc_msf_out;
1100 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1101 		}
1102 		kfree(gsf);
1103 		gsf = NULL;
1104 
1105 		err = ip_mc_msfilter(sk, msf, ifindex);
1106 mc_msf_out:
1107 		kfree(msf);
1108 		kfree(gsf);
1109 		break;
1110 	}
1111 	case IP_MULTICAST_ALL:
1112 		if (optlen < 1)
1113 			goto e_inval;
1114 		if (val != 0 && val != 1)
1115 			goto e_inval;
1116 		inet->mc_all = val;
1117 		break;
1118 	case IP_ROUTER_ALERT:
1119 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1120 		break;
1121 
1122 	case IP_FREEBIND:
1123 		if (optlen < 1)
1124 			goto e_inval;
1125 		inet->freebind = !!val;
1126 		break;
1127 
1128 	case IP_IPSEC_POLICY:
1129 	case IP_XFRM_POLICY:
1130 		err = -EPERM;
1131 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1132 			break;
1133 		err = xfrm_user_policy(sk, optname, optval, optlen);
1134 		break;
1135 
1136 	case IP_TRANSPARENT:
1137 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1138 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1139 			err = -EPERM;
1140 			break;
1141 		}
1142 		if (optlen < 1)
1143 			goto e_inval;
1144 		inet->transparent = !!val;
1145 		break;
1146 
1147 	case IP_MINTTL:
1148 		if (optlen < 1)
1149 			goto e_inval;
1150 		if (val < 0 || val > 255)
1151 			goto e_inval;
1152 		inet->min_ttl = val;
1153 		break;
1154 
1155 	default:
1156 		err = -ENOPROTOOPT;
1157 		break;
1158 	}
1159 	release_sock(sk);
1160 	if (needs_rtnl)
1161 		rtnl_unlock();
1162 	return err;
1163 
1164 e_inval:
1165 	release_sock(sk);
1166 	if (needs_rtnl)
1167 		rtnl_unlock();
1168 	return -EINVAL;
1169 }
1170 
1171 /**
1172  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1173  * @sk: socket
1174  * @skb: buffer
1175  *
1176  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1177  * destination in skb->cb[] before dst drop.
1178  * This way, receiver doesn't make cache line misses to read rtable.
1179  */
1180 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1181 {
1182 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1183 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1184 		       ipv6_sk_rxinfo(sk);
1185 
1186 	if (prepare && skb_rtable(skb)) {
1187 		pktinfo->ipi_ifindex = inet_iif(skb);
1188 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1189 	} else {
1190 		pktinfo->ipi_ifindex = 0;
1191 		pktinfo->ipi_spec_dst.s_addr = 0;
1192 	}
1193 	skb_dst_drop(skb);
1194 }
1195 
1196 int ip_setsockopt(struct sock *sk, int level,
1197 		int optname, char __user *optval, unsigned int optlen)
1198 {
1199 	int err;
1200 
1201 	if (level != SOL_IP)
1202 		return -ENOPROTOOPT;
1203 
1204 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1205 #ifdef CONFIG_NETFILTER
1206 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1207 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1208 			optname != IP_IPSEC_POLICY &&
1209 			optname != IP_XFRM_POLICY &&
1210 			!ip_mroute_opt(optname)) {
1211 		lock_sock(sk);
1212 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1213 		release_sock(sk);
1214 	}
1215 #endif
1216 	return err;
1217 }
1218 EXPORT_SYMBOL(ip_setsockopt);
1219 
1220 #ifdef CONFIG_COMPAT
1221 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1222 			 char __user *optval, unsigned int optlen)
1223 {
1224 	int err;
1225 
1226 	if (level != SOL_IP)
1227 		return -ENOPROTOOPT;
1228 
1229 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1230 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1231 			ip_setsockopt);
1232 
1233 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1234 #ifdef CONFIG_NETFILTER
1235 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1236 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1237 			optname != IP_IPSEC_POLICY &&
1238 			optname != IP_XFRM_POLICY &&
1239 			!ip_mroute_opt(optname)) {
1240 		lock_sock(sk);
1241 		err = compat_nf_setsockopt(sk, PF_INET, optname,
1242 					   optval, optlen);
1243 		release_sock(sk);
1244 	}
1245 #endif
1246 	return err;
1247 }
1248 EXPORT_SYMBOL(compat_ip_setsockopt);
1249 #endif
1250 
1251 /*
1252  *	Get the options. Note for future reference. The GET of IP options gets
1253  *	the _received_ ones. The set sets the _sent_ ones.
1254  */
1255 
1256 static bool getsockopt_needs_rtnl(int optname)
1257 {
1258 	switch (optname) {
1259 	case IP_MSFILTER:
1260 	case MCAST_MSFILTER:
1261 		return true;
1262 	}
1263 	return false;
1264 }
1265 
1266 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1267 			    char __user *optval, int __user *optlen, unsigned int flags)
1268 {
1269 	struct inet_sock *inet = inet_sk(sk);
1270 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1271 	int val, err = 0;
1272 	int len;
1273 
1274 	if (level != SOL_IP)
1275 		return -EOPNOTSUPP;
1276 
1277 	if (ip_mroute_opt(optname))
1278 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1279 
1280 	if (get_user(len, optlen))
1281 		return -EFAULT;
1282 	if (len < 0)
1283 		return -EINVAL;
1284 
1285 	if (needs_rtnl)
1286 		rtnl_lock();
1287 	lock_sock(sk);
1288 
1289 	switch (optname) {
1290 	case IP_OPTIONS:
1291 	{
1292 		unsigned char optbuf[sizeof(struct ip_options)+40];
1293 		struct ip_options *opt = (struct ip_options *)optbuf;
1294 		struct ip_options_rcu *inet_opt;
1295 
1296 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1297 						     sock_owned_by_user(sk));
1298 		opt->optlen = 0;
1299 		if (inet_opt)
1300 			memcpy(optbuf, &inet_opt->opt,
1301 			       sizeof(struct ip_options) +
1302 			       inet_opt->opt.optlen);
1303 		release_sock(sk);
1304 
1305 		if (opt->optlen == 0)
1306 			return put_user(0, optlen);
1307 
1308 		ip_options_undo(opt);
1309 
1310 		len = min_t(unsigned int, len, opt->optlen);
1311 		if (put_user(len, optlen))
1312 			return -EFAULT;
1313 		if (copy_to_user(optval, opt->__data, len))
1314 			return -EFAULT;
1315 		return 0;
1316 	}
1317 	case IP_PKTINFO:
1318 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1319 		break;
1320 	case IP_RECVTTL:
1321 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1322 		break;
1323 	case IP_RECVTOS:
1324 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1325 		break;
1326 	case IP_RECVOPTS:
1327 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1328 		break;
1329 	case IP_RETOPTS:
1330 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1331 		break;
1332 	case IP_PASSSEC:
1333 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1334 		break;
1335 	case IP_RECVORIGDSTADDR:
1336 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1337 		break;
1338 	case IP_CHECKSUM:
1339 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1340 		break;
1341 	case IP_TOS:
1342 		val = inet->tos;
1343 		break;
1344 	case IP_TTL:
1345 		val = (inet->uc_ttl == -1 ?
1346 		       sysctl_ip_default_ttl :
1347 		       inet->uc_ttl);
1348 		break;
1349 	case IP_HDRINCL:
1350 		val = inet->hdrincl;
1351 		break;
1352 	case IP_NODEFRAG:
1353 		val = inet->nodefrag;
1354 		break;
1355 	case IP_BIND_ADDRESS_NO_PORT:
1356 		val = inet->bind_address_no_port;
1357 		break;
1358 	case IP_MTU_DISCOVER:
1359 		val = inet->pmtudisc;
1360 		break;
1361 	case IP_MTU:
1362 	{
1363 		struct dst_entry *dst;
1364 		val = 0;
1365 		dst = sk_dst_get(sk);
1366 		if (dst) {
1367 			val = dst_mtu(dst);
1368 			dst_release(dst);
1369 		}
1370 		if (!val) {
1371 			release_sock(sk);
1372 			return -ENOTCONN;
1373 		}
1374 		break;
1375 	}
1376 	case IP_RECVERR:
1377 		val = inet->recverr;
1378 		break;
1379 	case IP_MULTICAST_TTL:
1380 		val = inet->mc_ttl;
1381 		break;
1382 	case IP_MULTICAST_LOOP:
1383 		val = inet->mc_loop;
1384 		break;
1385 	case IP_UNICAST_IF:
1386 		val = (__force int)htonl((__u32) inet->uc_index);
1387 		break;
1388 	case IP_MULTICAST_IF:
1389 	{
1390 		struct in_addr addr;
1391 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1392 		addr.s_addr = inet->mc_addr;
1393 		release_sock(sk);
1394 
1395 		if (put_user(len, optlen))
1396 			return -EFAULT;
1397 		if (copy_to_user(optval, &addr, len))
1398 			return -EFAULT;
1399 		return 0;
1400 	}
1401 	case IP_MSFILTER:
1402 	{
1403 		struct ip_msfilter msf;
1404 
1405 		if (len < IP_MSFILTER_SIZE(0)) {
1406 			err = -EINVAL;
1407 			goto out;
1408 		}
1409 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1410 			err = -EFAULT;
1411 			goto out;
1412 		}
1413 		err = ip_mc_msfget(sk, &msf,
1414 				   (struct ip_msfilter __user *)optval, optlen);
1415 		goto out;
1416 	}
1417 	case MCAST_MSFILTER:
1418 	{
1419 		struct group_filter gsf;
1420 
1421 		if (len < GROUP_FILTER_SIZE(0)) {
1422 			err = -EINVAL;
1423 			goto out;
1424 		}
1425 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1426 			err = -EFAULT;
1427 			goto out;
1428 		}
1429 		err = ip_mc_gsfget(sk, &gsf,
1430 				   (struct group_filter __user *)optval,
1431 				   optlen);
1432 		goto out;
1433 	}
1434 	case IP_MULTICAST_ALL:
1435 		val = inet->mc_all;
1436 		break;
1437 	case IP_PKTOPTIONS:
1438 	{
1439 		struct msghdr msg;
1440 
1441 		release_sock(sk);
1442 
1443 		if (sk->sk_type != SOCK_STREAM)
1444 			return -ENOPROTOOPT;
1445 
1446 		msg.msg_control = (__force void *) optval;
1447 		msg.msg_controllen = len;
1448 		msg.msg_flags = flags;
1449 
1450 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1451 			struct in_pktinfo info;
1452 
1453 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1454 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1455 			info.ipi_ifindex = inet->mc_index;
1456 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1457 		}
1458 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1459 			int hlim = inet->mc_ttl;
1460 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1461 		}
1462 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1463 			int tos = inet->rcv_tos;
1464 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1465 		}
1466 		len -= msg.msg_controllen;
1467 		return put_user(len, optlen);
1468 	}
1469 	case IP_FREEBIND:
1470 		val = inet->freebind;
1471 		break;
1472 	case IP_TRANSPARENT:
1473 		val = inet->transparent;
1474 		break;
1475 	case IP_MINTTL:
1476 		val = inet->min_ttl;
1477 		break;
1478 	default:
1479 		release_sock(sk);
1480 		return -ENOPROTOOPT;
1481 	}
1482 	release_sock(sk);
1483 
1484 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1485 		unsigned char ucval = (unsigned char)val;
1486 		len = 1;
1487 		if (put_user(len, optlen))
1488 			return -EFAULT;
1489 		if (copy_to_user(optval, &ucval, 1))
1490 			return -EFAULT;
1491 	} else {
1492 		len = min_t(unsigned int, sizeof(int), len);
1493 		if (put_user(len, optlen))
1494 			return -EFAULT;
1495 		if (copy_to_user(optval, &val, len))
1496 			return -EFAULT;
1497 	}
1498 	return 0;
1499 
1500 out:
1501 	release_sock(sk);
1502 	if (needs_rtnl)
1503 		rtnl_unlock();
1504 	return err;
1505 }
1506 
1507 int ip_getsockopt(struct sock *sk, int level,
1508 		  int optname, char __user *optval, int __user *optlen)
1509 {
1510 	int err;
1511 
1512 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1513 #ifdef CONFIG_NETFILTER
1514 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1515 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1516 			!ip_mroute_opt(optname)) {
1517 		int len;
1518 
1519 		if (get_user(len, optlen))
1520 			return -EFAULT;
1521 
1522 		lock_sock(sk);
1523 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1524 				&len);
1525 		release_sock(sk);
1526 		if (err >= 0)
1527 			err = put_user(len, optlen);
1528 		return err;
1529 	}
1530 #endif
1531 	return err;
1532 }
1533 EXPORT_SYMBOL(ip_getsockopt);
1534 
1535 #ifdef CONFIG_COMPAT
1536 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1537 			 char __user *optval, int __user *optlen)
1538 {
1539 	int err;
1540 
1541 	if (optname == MCAST_MSFILTER)
1542 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1543 			ip_getsockopt);
1544 
1545 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1546 		MSG_CMSG_COMPAT);
1547 
1548 #ifdef CONFIG_NETFILTER
1549 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1550 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1551 			!ip_mroute_opt(optname)) {
1552 		int len;
1553 
1554 		if (get_user(len, optlen))
1555 			return -EFAULT;
1556 
1557 		lock_sock(sk);
1558 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1559 		release_sock(sk);
1560 		if (err >= 0)
1561 			err = put_user(len, optlen);
1562 		return err;
1563 	}
1564 #endif
1565 	return err;
1566 }
1567 EXPORT_SYMBOL(compat_ip_getsockopt);
1568 #endif
1569