xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision a2cce7a9)
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) {
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)) {
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 /* For some errors we have valid addr_offset even with zero payload and
436  * zero port. Also, addr_offset should be supported if port is set.
437  */
438 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
439 {
440 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
441 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
442 }
443 
444 /* IPv4 supports cmsg on all imcp errors and some timestamps
445  *
446  * Timestamp code paths do not initialize the fields expected by cmsg:
447  * the PKTINFO fields in skb->cb[]. Fill those in here.
448  */
449 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
450 				       struct sk_buff *skb,
451 				       int ee_origin)
452 {
453 	struct in_pktinfo *info;
454 
455 	if (ee_origin == SO_EE_ORIGIN_ICMP)
456 		return true;
457 
458 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
459 		return false;
460 
461 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
462 	 * timestamp with egress dev. Not possible for packets without dev
463 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
464 	 */
465 	if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
466 	    (!skb->dev))
467 		return false;
468 
469 	info = PKTINFO_SKB_CB(skb);
470 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
471 	info->ipi_ifindex = skb->dev->ifindex;
472 	return true;
473 }
474 
475 /*
476  *	Handle MSG_ERRQUEUE
477  */
478 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
479 {
480 	struct sock_exterr_skb *serr;
481 	struct sk_buff *skb;
482 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
483 	struct {
484 		struct sock_extended_err ee;
485 		struct sockaddr_in	 offender;
486 	} errhdr;
487 	int err;
488 	int copied;
489 
490 	WARN_ON_ONCE(sk->sk_family == AF_INET6);
491 
492 	err = -EAGAIN;
493 	skb = sock_dequeue_err_skb(sk);
494 	if (!skb)
495 		goto out;
496 
497 	copied = skb->len;
498 	if (copied > len) {
499 		msg->msg_flags |= MSG_TRUNC;
500 		copied = len;
501 	}
502 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
503 	if (err)
504 		goto out_free_skb;
505 
506 	sock_recv_timestamp(msg, sk, skb);
507 
508 	serr = SKB_EXT_ERR(skb);
509 
510 	if (sin && ipv4_datagram_support_addr(serr)) {
511 		sin->sin_family = AF_INET;
512 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
513 						   serr->addr_offset);
514 		sin->sin_port = serr->port;
515 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
516 		*addr_len = sizeof(*sin);
517 	}
518 
519 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
520 	sin = &errhdr.offender;
521 	memset(sin, 0, sizeof(*sin));
522 
523 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
524 		sin->sin_family = AF_INET;
525 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
526 		if (inet_sk(sk)->cmsg_flags)
527 			ip_cmsg_recv(msg, skb);
528 	}
529 
530 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
531 
532 	/* Now we could try to dump offended packet options */
533 
534 	msg->msg_flags |= MSG_ERRQUEUE;
535 	err = copied;
536 
537 out_free_skb:
538 	kfree_skb(skb);
539 out:
540 	return err;
541 }
542 
543 
544 /*
545  *	Socket option code for IP. This is the end of the line after any
546  *	TCP,UDP etc options on an IP socket.
547  */
548 static bool setsockopt_needs_rtnl(int optname)
549 {
550 	switch (optname) {
551 	case IP_ADD_MEMBERSHIP:
552 	case IP_ADD_SOURCE_MEMBERSHIP:
553 	case IP_BLOCK_SOURCE:
554 	case IP_DROP_MEMBERSHIP:
555 	case IP_DROP_SOURCE_MEMBERSHIP:
556 	case IP_MSFILTER:
557 	case IP_UNBLOCK_SOURCE:
558 	case MCAST_BLOCK_SOURCE:
559 	case MCAST_MSFILTER:
560 	case MCAST_JOIN_GROUP:
561 	case MCAST_JOIN_SOURCE_GROUP:
562 	case MCAST_LEAVE_GROUP:
563 	case MCAST_LEAVE_SOURCE_GROUP:
564 	case MCAST_UNBLOCK_SOURCE:
565 		return true;
566 	}
567 	return false;
568 }
569 
570 static int do_ip_setsockopt(struct sock *sk, int level,
571 			    int optname, char __user *optval, unsigned int optlen)
572 {
573 	struct inet_sock *inet = inet_sk(sk);
574 	int val = 0, err;
575 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
576 
577 	switch (optname) {
578 	case IP_PKTINFO:
579 	case IP_RECVTTL:
580 	case IP_RECVOPTS:
581 	case IP_RECVTOS:
582 	case IP_RETOPTS:
583 	case IP_TOS:
584 	case IP_TTL:
585 	case IP_HDRINCL:
586 	case IP_MTU_DISCOVER:
587 	case IP_RECVERR:
588 	case IP_ROUTER_ALERT:
589 	case IP_FREEBIND:
590 	case IP_PASSSEC:
591 	case IP_TRANSPARENT:
592 	case IP_MINTTL:
593 	case IP_NODEFRAG:
594 	case IP_BIND_ADDRESS_NO_PORT:
595 	case IP_UNICAST_IF:
596 	case IP_MULTICAST_TTL:
597 	case IP_MULTICAST_ALL:
598 	case IP_MULTICAST_LOOP:
599 	case IP_RECVORIGDSTADDR:
600 	case IP_CHECKSUM:
601 		if (optlen >= sizeof(int)) {
602 			if (get_user(val, (int __user *) optval))
603 				return -EFAULT;
604 		} else if (optlen >= sizeof(char)) {
605 			unsigned char ucval;
606 
607 			if (get_user(ucval, (unsigned char __user *) optval))
608 				return -EFAULT;
609 			val = (int) ucval;
610 		}
611 	}
612 
613 	/* If optlen==0, it is equivalent to val == 0 */
614 
615 	if (ip_mroute_opt(optname))
616 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
617 
618 	err = 0;
619 	if (needs_rtnl)
620 		rtnl_lock();
621 	lock_sock(sk);
622 
623 	switch (optname) {
624 	case IP_OPTIONS:
625 	{
626 		struct ip_options_rcu *old, *opt = NULL;
627 
628 		if (optlen > 40)
629 			goto e_inval;
630 		err = ip_options_get_from_user(sock_net(sk), &opt,
631 					       optval, optlen);
632 		if (err)
633 			break;
634 		old = rcu_dereference_protected(inet->inet_opt,
635 						sock_owned_by_user(sk));
636 		if (inet->is_icsk) {
637 			struct inet_connection_sock *icsk = inet_csk(sk);
638 #if IS_ENABLED(CONFIG_IPV6)
639 			if (sk->sk_family == PF_INET ||
640 			    (!((1 << sk->sk_state) &
641 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
642 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
643 #endif
644 				if (old)
645 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
646 				if (opt)
647 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
648 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
649 #if IS_ENABLED(CONFIG_IPV6)
650 			}
651 #endif
652 		}
653 		rcu_assign_pointer(inet->inet_opt, opt);
654 		if (old)
655 			kfree_rcu(old, rcu);
656 		break;
657 	}
658 	case IP_PKTINFO:
659 		if (val)
660 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
661 		else
662 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
663 		break;
664 	case IP_RECVTTL:
665 		if (val)
666 			inet->cmsg_flags |=  IP_CMSG_TTL;
667 		else
668 			inet->cmsg_flags &= ~IP_CMSG_TTL;
669 		break;
670 	case IP_RECVTOS:
671 		if (val)
672 			inet->cmsg_flags |=  IP_CMSG_TOS;
673 		else
674 			inet->cmsg_flags &= ~IP_CMSG_TOS;
675 		break;
676 	case IP_RECVOPTS:
677 		if (val)
678 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
679 		else
680 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
681 		break;
682 	case IP_RETOPTS:
683 		if (val)
684 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
685 		else
686 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
687 		break;
688 	case IP_PASSSEC:
689 		if (val)
690 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
691 		else
692 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
693 		break;
694 	case IP_RECVORIGDSTADDR:
695 		if (val)
696 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
697 		else
698 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
699 		break;
700 	case IP_CHECKSUM:
701 		if (val) {
702 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
703 				inet_inc_convert_csum(sk);
704 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
705 			}
706 		} else {
707 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
708 				inet_dec_convert_csum(sk);
709 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
710 			}
711 		}
712 		break;
713 	case IP_TOS:	/* This sets both TOS and Precedence */
714 		if (sk->sk_type == SOCK_STREAM) {
715 			val &= ~INET_ECN_MASK;
716 			val |= inet->tos & INET_ECN_MASK;
717 		}
718 		if (inet->tos != val) {
719 			inet->tos = val;
720 			sk->sk_priority = rt_tos2priority(val);
721 			sk_dst_reset(sk);
722 		}
723 		break;
724 	case IP_TTL:
725 		if (optlen < 1)
726 			goto e_inval;
727 		if (val != -1 && (val < 1 || val > 255))
728 			goto e_inval;
729 		inet->uc_ttl = val;
730 		break;
731 	case IP_HDRINCL:
732 		if (sk->sk_type != SOCK_RAW) {
733 			err = -ENOPROTOOPT;
734 			break;
735 		}
736 		inet->hdrincl = val ? 1 : 0;
737 		break;
738 	case IP_NODEFRAG:
739 		if (sk->sk_type != SOCK_RAW) {
740 			err = -ENOPROTOOPT;
741 			break;
742 		}
743 		inet->nodefrag = val ? 1 : 0;
744 		break;
745 	case IP_BIND_ADDRESS_NO_PORT:
746 		inet->bind_address_no_port = val ? 1 : 0;
747 		break;
748 	case IP_MTU_DISCOVER:
749 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
750 			goto e_inval;
751 		inet->pmtudisc = val;
752 		break;
753 	case IP_RECVERR:
754 		inet->recverr = !!val;
755 		if (!val)
756 			skb_queue_purge(&sk->sk_error_queue);
757 		break;
758 	case IP_MULTICAST_TTL:
759 		if (sk->sk_type == SOCK_STREAM)
760 			goto e_inval;
761 		if (optlen < 1)
762 			goto e_inval;
763 		if (val == -1)
764 			val = 1;
765 		if (val < 0 || val > 255)
766 			goto e_inval;
767 		inet->mc_ttl = val;
768 		break;
769 	case IP_MULTICAST_LOOP:
770 		if (optlen < 1)
771 			goto e_inval;
772 		inet->mc_loop = !!val;
773 		break;
774 	case IP_UNICAST_IF:
775 	{
776 		struct net_device *dev = NULL;
777 		int ifindex;
778 
779 		if (optlen != sizeof(int))
780 			goto e_inval;
781 
782 		ifindex = (__force int)ntohl((__force __be32)val);
783 		if (ifindex == 0) {
784 			inet->uc_index = 0;
785 			err = 0;
786 			break;
787 		}
788 
789 		dev = dev_get_by_index(sock_net(sk), ifindex);
790 		err = -EADDRNOTAVAIL;
791 		if (!dev)
792 			break;
793 		dev_put(dev);
794 
795 		err = -EINVAL;
796 		if (sk->sk_bound_dev_if)
797 			break;
798 
799 		inet->uc_index = ifindex;
800 		err = 0;
801 		break;
802 	}
803 	case IP_MULTICAST_IF:
804 	{
805 		struct ip_mreqn mreq;
806 		struct net_device *dev = NULL;
807 
808 		if (sk->sk_type == SOCK_STREAM)
809 			goto e_inval;
810 		/*
811 		 *	Check the arguments are allowable
812 		 */
813 
814 		if (optlen < sizeof(struct in_addr))
815 			goto e_inval;
816 
817 		err = -EFAULT;
818 		if (optlen >= sizeof(struct ip_mreqn)) {
819 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
820 				break;
821 		} else {
822 			memset(&mreq, 0, sizeof(mreq));
823 			if (optlen >= sizeof(struct ip_mreq)) {
824 				if (copy_from_user(&mreq, optval,
825 						   sizeof(struct ip_mreq)))
826 					break;
827 			} else if (optlen >= sizeof(struct in_addr)) {
828 				if (copy_from_user(&mreq.imr_address, optval,
829 						   sizeof(struct in_addr)))
830 					break;
831 			}
832 		}
833 
834 		if (!mreq.imr_ifindex) {
835 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
836 				inet->mc_index = 0;
837 				inet->mc_addr  = 0;
838 				err = 0;
839 				break;
840 			}
841 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
842 			if (dev)
843 				mreq.imr_ifindex = dev->ifindex;
844 		} else
845 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
846 
847 
848 		err = -EADDRNOTAVAIL;
849 		if (!dev)
850 			break;
851 		dev_put(dev);
852 
853 		err = -EINVAL;
854 		if (sk->sk_bound_dev_if &&
855 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
856 			break;
857 
858 		inet->mc_index = mreq.imr_ifindex;
859 		inet->mc_addr  = mreq.imr_address.s_addr;
860 		err = 0;
861 		break;
862 	}
863 
864 	case IP_ADD_MEMBERSHIP:
865 	case IP_DROP_MEMBERSHIP:
866 	{
867 		struct ip_mreqn mreq;
868 
869 		err = -EPROTO;
870 		if (inet_sk(sk)->is_icsk)
871 			break;
872 
873 		if (optlen < sizeof(struct ip_mreq))
874 			goto e_inval;
875 		err = -EFAULT;
876 		if (optlen >= sizeof(struct ip_mreqn)) {
877 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
878 				break;
879 		} else {
880 			memset(&mreq, 0, sizeof(mreq));
881 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
882 				break;
883 		}
884 
885 		if (optname == IP_ADD_MEMBERSHIP)
886 			err = ip_mc_join_group(sk, &mreq);
887 		else
888 			err = ip_mc_leave_group(sk, &mreq);
889 		break;
890 	}
891 	case IP_MSFILTER:
892 	{
893 		struct ip_msfilter *msf;
894 
895 		if (optlen < IP_MSFILTER_SIZE(0))
896 			goto e_inval;
897 		if (optlen > sysctl_optmem_max) {
898 			err = -ENOBUFS;
899 			break;
900 		}
901 		msf = kmalloc(optlen, GFP_KERNEL);
902 		if (!msf) {
903 			err = -ENOBUFS;
904 			break;
905 		}
906 		err = -EFAULT;
907 		if (copy_from_user(msf, optval, optlen)) {
908 			kfree(msf);
909 			break;
910 		}
911 		/* numsrc >= (1G-4) overflow in 32 bits */
912 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
913 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
914 			kfree(msf);
915 			err = -ENOBUFS;
916 			break;
917 		}
918 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
919 			kfree(msf);
920 			err = -EINVAL;
921 			break;
922 		}
923 		err = ip_mc_msfilter(sk, msf, 0);
924 		kfree(msf);
925 		break;
926 	}
927 	case IP_BLOCK_SOURCE:
928 	case IP_UNBLOCK_SOURCE:
929 	case IP_ADD_SOURCE_MEMBERSHIP:
930 	case IP_DROP_SOURCE_MEMBERSHIP:
931 	{
932 		struct ip_mreq_source mreqs;
933 		int omode, add;
934 
935 		if (optlen != sizeof(struct ip_mreq_source))
936 			goto e_inval;
937 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
938 			err = -EFAULT;
939 			break;
940 		}
941 		if (optname == IP_BLOCK_SOURCE) {
942 			omode = MCAST_EXCLUDE;
943 			add = 1;
944 		} else if (optname == IP_UNBLOCK_SOURCE) {
945 			omode = MCAST_EXCLUDE;
946 			add = 0;
947 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
948 			struct ip_mreqn mreq;
949 
950 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
951 			mreq.imr_address.s_addr = mreqs.imr_interface;
952 			mreq.imr_ifindex = 0;
953 			err = ip_mc_join_group(sk, &mreq);
954 			if (err && err != -EADDRINUSE)
955 				break;
956 			omode = MCAST_INCLUDE;
957 			add = 1;
958 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
959 			omode = MCAST_INCLUDE;
960 			add = 0;
961 		}
962 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
963 		break;
964 	}
965 	case MCAST_JOIN_GROUP:
966 	case MCAST_LEAVE_GROUP:
967 	{
968 		struct group_req greq;
969 		struct sockaddr_in *psin;
970 		struct ip_mreqn mreq;
971 
972 		if (optlen < sizeof(struct group_req))
973 			goto e_inval;
974 		err = -EFAULT;
975 		if (copy_from_user(&greq, optval, sizeof(greq)))
976 			break;
977 		psin = (struct sockaddr_in *)&greq.gr_group;
978 		if (psin->sin_family != AF_INET)
979 			goto e_inval;
980 		memset(&mreq, 0, sizeof(mreq));
981 		mreq.imr_multiaddr = psin->sin_addr;
982 		mreq.imr_ifindex = greq.gr_interface;
983 
984 		if (optname == MCAST_JOIN_GROUP)
985 			err = ip_mc_join_group(sk, &mreq);
986 		else
987 			err = ip_mc_leave_group(sk, &mreq);
988 		break;
989 	}
990 	case MCAST_JOIN_SOURCE_GROUP:
991 	case MCAST_LEAVE_SOURCE_GROUP:
992 	case MCAST_BLOCK_SOURCE:
993 	case MCAST_UNBLOCK_SOURCE:
994 	{
995 		struct group_source_req greqs;
996 		struct ip_mreq_source mreqs;
997 		struct sockaddr_in *psin;
998 		int omode, add;
999 
1000 		if (optlen != sizeof(struct group_source_req))
1001 			goto e_inval;
1002 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1003 			err = -EFAULT;
1004 			break;
1005 		}
1006 		if (greqs.gsr_group.ss_family != AF_INET ||
1007 		    greqs.gsr_source.ss_family != AF_INET) {
1008 			err = -EADDRNOTAVAIL;
1009 			break;
1010 		}
1011 		psin = (struct sockaddr_in *)&greqs.gsr_group;
1012 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1013 		psin = (struct sockaddr_in *)&greqs.gsr_source;
1014 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1015 		mreqs.imr_interface = 0; /* use index for mc_source */
1016 
1017 		if (optname == MCAST_BLOCK_SOURCE) {
1018 			omode = MCAST_EXCLUDE;
1019 			add = 1;
1020 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
1021 			omode = MCAST_EXCLUDE;
1022 			add = 0;
1023 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1024 			struct ip_mreqn mreq;
1025 
1026 			psin = (struct sockaddr_in *)&greqs.gsr_group;
1027 			mreq.imr_multiaddr = psin->sin_addr;
1028 			mreq.imr_address.s_addr = 0;
1029 			mreq.imr_ifindex = greqs.gsr_interface;
1030 			err = ip_mc_join_group(sk, &mreq);
1031 			if (err && err != -EADDRINUSE)
1032 				break;
1033 			greqs.gsr_interface = mreq.imr_ifindex;
1034 			omode = MCAST_INCLUDE;
1035 			add = 1;
1036 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
1037 			omode = MCAST_INCLUDE;
1038 			add = 0;
1039 		}
1040 		err = ip_mc_source(add, omode, sk, &mreqs,
1041 				   greqs.gsr_interface);
1042 		break;
1043 	}
1044 	case MCAST_MSFILTER:
1045 	{
1046 		struct sockaddr_in *psin;
1047 		struct ip_msfilter *msf = NULL;
1048 		struct group_filter *gsf = NULL;
1049 		int msize, i, ifindex;
1050 
1051 		if (optlen < GROUP_FILTER_SIZE(0))
1052 			goto e_inval;
1053 		if (optlen > sysctl_optmem_max) {
1054 			err = -ENOBUFS;
1055 			break;
1056 		}
1057 		gsf = kmalloc(optlen, GFP_KERNEL);
1058 		if (!gsf) {
1059 			err = -ENOBUFS;
1060 			break;
1061 		}
1062 		err = -EFAULT;
1063 		if (copy_from_user(gsf, optval, optlen))
1064 			goto mc_msf_out;
1065 
1066 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1067 		if (gsf->gf_numsrc >= 0x1ffffff ||
1068 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
1069 			err = -ENOBUFS;
1070 			goto mc_msf_out;
1071 		}
1072 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1073 			err = -EINVAL;
1074 			goto mc_msf_out;
1075 		}
1076 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1077 		msf = kmalloc(msize, GFP_KERNEL);
1078 		if (!msf) {
1079 			err = -ENOBUFS;
1080 			goto mc_msf_out;
1081 		}
1082 		ifindex = gsf->gf_interface;
1083 		psin = (struct sockaddr_in *)&gsf->gf_group;
1084 		if (psin->sin_family != AF_INET) {
1085 			err = -EADDRNOTAVAIL;
1086 			goto mc_msf_out;
1087 		}
1088 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1089 		msf->imsf_interface = 0;
1090 		msf->imsf_fmode = gsf->gf_fmode;
1091 		msf->imsf_numsrc = gsf->gf_numsrc;
1092 		err = -EADDRNOTAVAIL;
1093 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1094 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1095 
1096 			if (psin->sin_family != AF_INET)
1097 				goto mc_msf_out;
1098 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1099 		}
1100 		kfree(gsf);
1101 		gsf = NULL;
1102 
1103 		err = ip_mc_msfilter(sk, msf, ifindex);
1104 mc_msf_out:
1105 		kfree(msf);
1106 		kfree(gsf);
1107 		break;
1108 	}
1109 	case IP_MULTICAST_ALL:
1110 		if (optlen < 1)
1111 			goto e_inval;
1112 		if (val != 0 && val != 1)
1113 			goto e_inval;
1114 		inet->mc_all = val;
1115 		break;
1116 	case IP_ROUTER_ALERT:
1117 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1118 		break;
1119 
1120 	case IP_FREEBIND:
1121 		if (optlen < 1)
1122 			goto e_inval;
1123 		inet->freebind = !!val;
1124 		break;
1125 
1126 	case IP_IPSEC_POLICY:
1127 	case IP_XFRM_POLICY:
1128 		err = -EPERM;
1129 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1130 			break;
1131 		err = xfrm_user_policy(sk, optname, optval, optlen);
1132 		break;
1133 
1134 	case IP_TRANSPARENT:
1135 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1136 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1137 			err = -EPERM;
1138 			break;
1139 		}
1140 		if (optlen < 1)
1141 			goto e_inval;
1142 		inet->transparent = !!val;
1143 		break;
1144 
1145 	case IP_MINTTL:
1146 		if (optlen < 1)
1147 			goto e_inval;
1148 		if (val < 0 || val > 255)
1149 			goto e_inval;
1150 		inet->min_ttl = val;
1151 		break;
1152 
1153 	default:
1154 		err = -ENOPROTOOPT;
1155 		break;
1156 	}
1157 	release_sock(sk);
1158 	if (needs_rtnl)
1159 		rtnl_unlock();
1160 	return err;
1161 
1162 e_inval:
1163 	release_sock(sk);
1164 	if (needs_rtnl)
1165 		rtnl_unlock();
1166 	return -EINVAL;
1167 }
1168 
1169 /**
1170  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1171  * @sk: socket
1172  * @skb: buffer
1173  *
1174  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1175  * destination in skb->cb[] before dst drop.
1176  * This way, receiver doesn't make cache line misses to read rtable.
1177  */
1178 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1179 {
1180 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1181 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1182 		       ipv6_sk_rxinfo(sk);
1183 
1184 	if (prepare && skb_rtable(skb)) {
1185 		pktinfo->ipi_ifindex = inet_iif(skb);
1186 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1187 	} else {
1188 		pktinfo->ipi_ifindex = 0;
1189 		pktinfo->ipi_spec_dst.s_addr = 0;
1190 	}
1191 	skb_dst_drop(skb);
1192 }
1193 
1194 int ip_setsockopt(struct sock *sk, int level,
1195 		int optname, char __user *optval, unsigned int optlen)
1196 {
1197 	int err;
1198 
1199 	if (level != SOL_IP)
1200 		return -ENOPROTOOPT;
1201 
1202 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1203 #ifdef CONFIG_NETFILTER
1204 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1205 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1206 			optname != IP_IPSEC_POLICY &&
1207 			optname != IP_XFRM_POLICY &&
1208 			!ip_mroute_opt(optname)) {
1209 		lock_sock(sk);
1210 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1211 		release_sock(sk);
1212 	}
1213 #endif
1214 	return err;
1215 }
1216 EXPORT_SYMBOL(ip_setsockopt);
1217 
1218 #ifdef CONFIG_COMPAT
1219 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1220 			 char __user *optval, unsigned int optlen)
1221 {
1222 	int err;
1223 
1224 	if (level != SOL_IP)
1225 		return -ENOPROTOOPT;
1226 
1227 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1228 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1229 			ip_setsockopt);
1230 
1231 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1232 #ifdef CONFIG_NETFILTER
1233 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1234 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1235 			optname != IP_IPSEC_POLICY &&
1236 			optname != IP_XFRM_POLICY &&
1237 			!ip_mroute_opt(optname)) {
1238 		lock_sock(sk);
1239 		err = compat_nf_setsockopt(sk, PF_INET, optname,
1240 					   optval, optlen);
1241 		release_sock(sk);
1242 	}
1243 #endif
1244 	return err;
1245 }
1246 EXPORT_SYMBOL(compat_ip_setsockopt);
1247 #endif
1248 
1249 /*
1250  *	Get the options. Note for future reference. The GET of IP options gets
1251  *	the _received_ ones. The set sets the _sent_ ones.
1252  */
1253 
1254 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1255 			    char __user *optval, int __user *optlen, unsigned int flags)
1256 {
1257 	struct inet_sock *inet = inet_sk(sk);
1258 	int val;
1259 	int len;
1260 
1261 	if (level != SOL_IP)
1262 		return -EOPNOTSUPP;
1263 
1264 	if (ip_mroute_opt(optname))
1265 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1266 
1267 	if (get_user(len, optlen))
1268 		return -EFAULT;
1269 	if (len < 0)
1270 		return -EINVAL;
1271 
1272 	lock_sock(sk);
1273 
1274 	switch (optname) {
1275 	case IP_OPTIONS:
1276 	{
1277 		unsigned char optbuf[sizeof(struct ip_options)+40];
1278 		struct ip_options *opt = (struct ip_options *)optbuf;
1279 		struct ip_options_rcu *inet_opt;
1280 
1281 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1282 						     sock_owned_by_user(sk));
1283 		opt->optlen = 0;
1284 		if (inet_opt)
1285 			memcpy(optbuf, &inet_opt->opt,
1286 			       sizeof(struct ip_options) +
1287 			       inet_opt->opt.optlen);
1288 		release_sock(sk);
1289 
1290 		if (opt->optlen == 0)
1291 			return put_user(0, optlen);
1292 
1293 		ip_options_undo(opt);
1294 
1295 		len = min_t(unsigned int, len, opt->optlen);
1296 		if (put_user(len, optlen))
1297 			return -EFAULT;
1298 		if (copy_to_user(optval, opt->__data, len))
1299 			return -EFAULT;
1300 		return 0;
1301 	}
1302 	case IP_PKTINFO:
1303 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1304 		break;
1305 	case IP_RECVTTL:
1306 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1307 		break;
1308 	case IP_RECVTOS:
1309 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1310 		break;
1311 	case IP_RECVOPTS:
1312 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1313 		break;
1314 	case IP_RETOPTS:
1315 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1316 		break;
1317 	case IP_PASSSEC:
1318 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1319 		break;
1320 	case IP_RECVORIGDSTADDR:
1321 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1322 		break;
1323 	case IP_CHECKSUM:
1324 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1325 		break;
1326 	case IP_TOS:
1327 		val = inet->tos;
1328 		break;
1329 	case IP_TTL:
1330 		val = (inet->uc_ttl == -1 ?
1331 		       sysctl_ip_default_ttl :
1332 		       inet->uc_ttl);
1333 		break;
1334 	case IP_HDRINCL:
1335 		val = inet->hdrincl;
1336 		break;
1337 	case IP_NODEFRAG:
1338 		val = inet->nodefrag;
1339 		break;
1340 	case IP_BIND_ADDRESS_NO_PORT:
1341 		val = inet->bind_address_no_port;
1342 		break;
1343 	case IP_MTU_DISCOVER:
1344 		val = inet->pmtudisc;
1345 		break;
1346 	case IP_MTU:
1347 	{
1348 		struct dst_entry *dst;
1349 		val = 0;
1350 		dst = sk_dst_get(sk);
1351 		if (dst) {
1352 			val = dst_mtu(dst);
1353 			dst_release(dst);
1354 		}
1355 		if (!val) {
1356 			release_sock(sk);
1357 			return -ENOTCONN;
1358 		}
1359 		break;
1360 	}
1361 	case IP_RECVERR:
1362 		val = inet->recverr;
1363 		break;
1364 	case IP_MULTICAST_TTL:
1365 		val = inet->mc_ttl;
1366 		break;
1367 	case IP_MULTICAST_LOOP:
1368 		val = inet->mc_loop;
1369 		break;
1370 	case IP_UNICAST_IF:
1371 		val = (__force int)htonl((__u32) inet->uc_index);
1372 		break;
1373 	case IP_MULTICAST_IF:
1374 	{
1375 		struct in_addr addr;
1376 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1377 		addr.s_addr = inet->mc_addr;
1378 		release_sock(sk);
1379 
1380 		if (put_user(len, optlen))
1381 			return -EFAULT;
1382 		if (copy_to_user(optval, &addr, len))
1383 			return -EFAULT;
1384 		return 0;
1385 	}
1386 	case IP_MSFILTER:
1387 	{
1388 		struct ip_msfilter msf;
1389 		int err;
1390 
1391 		if (len < IP_MSFILTER_SIZE(0)) {
1392 			release_sock(sk);
1393 			return -EINVAL;
1394 		}
1395 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1396 			release_sock(sk);
1397 			return -EFAULT;
1398 		}
1399 		err = ip_mc_msfget(sk, &msf,
1400 				   (struct ip_msfilter __user *)optval, optlen);
1401 		release_sock(sk);
1402 		return err;
1403 	}
1404 	case MCAST_MSFILTER:
1405 	{
1406 		struct group_filter gsf;
1407 		int err;
1408 
1409 		if (len < GROUP_FILTER_SIZE(0)) {
1410 			release_sock(sk);
1411 			return -EINVAL;
1412 		}
1413 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1414 			release_sock(sk);
1415 			return -EFAULT;
1416 		}
1417 		err = ip_mc_gsfget(sk, &gsf,
1418 				   (struct group_filter __user *)optval,
1419 				   optlen);
1420 		release_sock(sk);
1421 		return err;
1422 	}
1423 	case IP_MULTICAST_ALL:
1424 		val = inet->mc_all;
1425 		break;
1426 	case IP_PKTOPTIONS:
1427 	{
1428 		struct msghdr msg;
1429 
1430 		release_sock(sk);
1431 
1432 		if (sk->sk_type != SOCK_STREAM)
1433 			return -ENOPROTOOPT;
1434 
1435 		msg.msg_control = (__force void *) optval;
1436 		msg.msg_controllen = len;
1437 		msg.msg_flags = flags;
1438 
1439 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1440 			struct in_pktinfo info;
1441 
1442 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1443 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1444 			info.ipi_ifindex = inet->mc_index;
1445 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1446 		}
1447 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1448 			int hlim = inet->mc_ttl;
1449 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1450 		}
1451 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1452 			int tos = inet->rcv_tos;
1453 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1454 		}
1455 		len -= msg.msg_controllen;
1456 		return put_user(len, optlen);
1457 	}
1458 	case IP_FREEBIND:
1459 		val = inet->freebind;
1460 		break;
1461 	case IP_TRANSPARENT:
1462 		val = inet->transparent;
1463 		break;
1464 	case IP_MINTTL:
1465 		val = inet->min_ttl;
1466 		break;
1467 	default:
1468 		release_sock(sk);
1469 		return -ENOPROTOOPT;
1470 	}
1471 	release_sock(sk);
1472 
1473 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1474 		unsigned char ucval = (unsigned char)val;
1475 		len = 1;
1476 		if (put_user(len, optlen))
1477 			return -EFAULT;
1478 		if (copy_to_user(optval, &ucval, 1))
1479 			return -EFAULT;
1480 	} else {
1481 		len = min_t(unsigned int, sizeof(int), len);
1482 		if (put_user(len, optlen))
1483 			return -EFAULT;
1484 		if (copy_to_user(optval, &val, len))
1485 			return -EFAULT;
1486 	}
1487 	return 0;
1488 }
1489 
1490 int ip_getsockopt(struct sock *sk, int level,
1491 		  int optname, char __user *optval, int __user *optlen)
1492 {
1493 	int err;
1494 
1495 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1496 #ifdef CONFIG_NETFILTER
1497 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1498 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1499 			!ip_mroute_opt(optname)) {
1500 		int len;
1501 
1502 		if (get_user(len, optlen))
1503 			return -EFAULT;
1504 
1505 		lock_sock(sk);
1506 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1507 				&len);
1508 		release_sock(sk);
1509 		if (err >= 0)
1510 			err = put_user(len, optlen);
1511 		return err;
1512 	}
1513 #endif
1514 	return err;
1515 }
1516 EXPORT_SYMBOL(ip_getsockopt);
1517 
1518 #ifdef CONFIG_COMPAT
1519 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1520 			 char __user *optval, int __user *optlen)
1521 {
1522 	int err;
1523 
1524 	if (optname == MCAST_MSFILTER)
1525 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1526 			ip_getsockopt);
1527 
1528 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1529 		MSG_CMSG_COMPAT);
1530 
1531 #ifdef CONFIG_NETFILTER
1532 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1533 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1534 			!ip_mroute_opt(optname)) {
1535 		int len;
1536 
1537 		if (get_user(len, optlen))
1538 			return -EFAULT;
1539 
1540 		lock_sock(sk);
1541 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1542 		release_sock(sk);
1543 		if (err >= 0)
1544 			err = put_user(len, optlen);
1545 		return err;
1546 	}
1547 #endif
1548 	return err;
1549 }
1550 EXPORT_SYMBOL(compat_ip_getsockopt);
1551 #endif
1552