xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 827634ad)
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 /* IPv4 supports cmsg on all imcp errors and some timestamps
436  *
437  * Timestamp code paths do not initialize the fields expected by cmsg:
438  * the PKTINFO fields in skb->cb[]. Fill those in here.
439  */
440 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
441 				       struct sk_buff *skb,
442 				       int ee_origin)
443 {
444 	struct in_pktinfo *info;
445 
446 	if (ee_origin == SO_EE_ORIGIN_ICMP)
447 		return true;
448 
449 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
450 		return false;
451 
452 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
453 	 * timestamp with egress dev. Not possible for packets without dev
454 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
455 	 */
456 	if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
457 	    (!skb->dev))
458 		return false;
459 
460 	info = PKTINFO_SKB_CB(skb);
461 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
462 	info->ipi_ifindex = skb->dev->ifindex;
463 	return true;
464 }
465 
466 /*
467  *	Handle MSG_ERRQUEUE
468  */
469 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
470 {
471 	struct sock_exterr_skb *serr;
472 	struct sk_buff *skb;
473 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
474 	struct {
475 		struct sock_extended_err ee;
476 		struct sockaddr_in	 offender;
477 	} errhdr;
478 	int err;
479 	int copied;
480 
481 	WARN_ON_ONCE(sk->sk_family == AF_INET6);
482 
483 	err = -EAGAIN;
484 	skb = sock_dequeue_err_skb(sk);
485 	if (!skb)
486 		goto out;
487 
488 	copied = skb->len;
489 	if (copied > len) {
490 		msg->msg_flags |= MSG_TRUNC;
491 		copied = len;
492 	}
493 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
494 	if (err)
495 		goto out_free_skb;
496 
497 	sock_recv_timestamp(msg, sk, skb);
498 
499 	serr = SKB_EXT_ERR(skb);
500 
501 	if (sin && serr->port) {
502 		sin->sin_family = AF_INET;
503 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
504 						   serr->addr_offset);
505 		sin->sin_port = serr->port;
506 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
507 		*addr_len = sizeof(*sin);
508 	}
509 
510 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
511 	sin = &errhdr.offender;
512 	memset(sin, 0, sizeof(*sin));
513 
514 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
515 		sin->sin_family = AF_INET;
516 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
517 		if (inet_sk(sk)->cmsg_flags)
518 			ip_cmsg_recv(msg, skb);
519 	}
520 
521 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
522 
523 	/* Now we could try to dump offended packet options */
524 
525 	msg->msg_flags |= MSG_ERRQUEUE;
526 	err = copied;
527 
528 out_free_skb:
529 	kfree_skb(skb);
530 out:
531 	return err;
532 }
533 
534 
535 /*
536  *	Socket option code for IP. This is the end of the line after any
537  *	TCP,UDP etc options on an IP socket.
538  */
539 static bool setsockopt_needs_rtnl(int optname)
540 {
541 	switch (optname) {
542 	case IP_ADD_MEMBERSHIP:
543 	case IP_ADD_SOURCE_MEMBERSHIP:
544 	case IP_BLOCK_SOURCE:
545 	case IP_DROP_MEMBERSHIP:
546 	case IP_DROP_SOURCE_MEMBERSHIP:
547 	case IP_MSFILTER:
548 	case IP_UNBLOCK_SOURCE:
549 	case MCAST_BLOCK_SOURCE:
550 	case MCAST_MSFILTER:
551 	case MCAST_JOIN_GROUP:
552 	case MCAST_JOIN_SOURCE_GROUP:
553 	case MCAST_LEAVE_GROUP:
554 	case MCAST_LEAVE_SOURCE_GROUP:
555 	case MCAST_UNBLOCK_SOURCE:
556 		return true;
557 	}
558 	return false;
559 }
560 
561 static int do_ip_setsockopt(struct sock *sk, int level,
562 			    int optname, char __user *optval, unsigned int optlen)
563 {
564 	struct inet_sock *inet = inet_sk(sk);
565 	int val = 0, err;
566 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
567 
568 	switch (optname) {
569 	case IP_PKTINFO:
570 	case IP_RECVTTL:
571 	case IP_RECVOPTS:
572 	case IP_RECVTOS:
573 	case IP_RETOPTS:
574 	case IP_TOS:
575 	case IP_TTL:
576 	case IP_HDRINCL:
577 	case IP_MTU_DISCOVER:
578 	case IP_RECVERR:
579 	case IP_ROUTER_ALERT:
580 	case IP_FREEBIND:
581 	case IP_PASSSEC:
582 	case IP_TRANSPARENT:
583 	case IP_MINTTL:
584 	case IP_NODEFRAG:
585 	case IP_UNICAST_IF:
586 	case IP_MULTICAST_TTL:
587 	case IP_MULTICAST_ALL:
588 	case IP_MULTICAST_LOOP:
589 	case IP_RECVORIGDSTADDR:
590 	case IP_CHECKSUM:
591 		if (optlen >= sizeof(int)) {
592 			if (get_user(val, (int __user *) optval))
593 				return -EFAULT;
594 		} else if (optlen >= sizeof(char)) {
595 			unsigned char ucval;
596 
597 			if (get_user(ucval, (unsigned char __user *) optval))
598 				return -EFAULT;
599 			val = (int) ucval;
600 		}
601 	}
602 
603 	/* If optlen==0, it is equivalent to val == 0 */
604 
605 	if (ip_mroute_opt(optname))
606 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
607 
608 	err = 0;
609 	if (needs_rtnl)
610 		rtnl_lock();
611 	lock_sock(sk);
612 
613 	switch (optname) {
614 	case IP_OPTIONS:
615 	{
616 		struct ip_options_rcu *old, *opt = NULL;
617 
618 		if (optlen > 40)
619 			goto e_inval;
620 		err = ip_options_get_from_user(sock_net(sk), &opt,
621 					       optval, optlen);
622 		if (err)
623 			break;
624 		old = rcu_dereference_protected(inet->inet_opt,
625 						sock_owned_by_user(sk));
626 		if (inet->is_icsk) {
627 			struct inet_connection_sock *icsk = inet_csk(sk);
628 #if IS_ENABLED(CONFIG_IPV6)
629 			if (sk->sk_family == PF_INET ||
630 			    (!((1 << sk->sk_state) &
631 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
632 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
633 #endif
634 				if (old)
635 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
636 				if (opt)
637 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
638 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
639 #if IS_ENABLED(CONFIG_IPV6)
640 			}
641 #endif
642 		}
643 		rcu_assign_pointer(inet->inet_opt, opt);
644 		if (old)
645 			kfree_rcu(old, rcu);
646 		break;
647 	}
648 	case IP_PKTINFO:
649 		if (val)
650 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
651 		else
652 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
653 		break;
654 	case IP_RECVTTL:
655 		if (val)
656 			inet->cmsg_flags |=  IP_CMSG_TTL;
657 		else
658 			inet->cmsg_flags &= ~IP_CMSG_TTL;
659 		break;
660 	case IP_RECVTOS:
661 		if (val)
662 			inet->cmsg_flags |=  IP_CMSG_TOS;
663 		else
664 			inet->cmsg_flags &= ~IP_CMSG_TOS;
665 		break;
666 	case IP_RECVOPTS:
667 		if (val)
668 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
669 		else
670 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
671 		break;
672 	case IP_RETOPTS:
673 		if (val)
674 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
675 		else
676 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
677 		break;
678 	case IP_PASSSEC:
679 		if (val)
680 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
681 		else
682 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
683 		break;
684 	case IP_RECVORIGDSTADDR:
685 		if (val)
686 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
687 		else
688 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
689 		break;
690 	case IP_CHECKSUM:
691 		if (val) {
692 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
693 				inet_inc_convert_csum(sk);
694 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
695 			}
696 		} else {
697 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
698 				inet_dec_convert_csum(sk);
699 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
700 			}
701 		}
702 		break;
703 	case IP_TOS:	/* This sets both TOS and Precedence */
704 		if (sk->sk_type == SOCK_STREAM) {
705 			val &= ~INET_ECN_MASK;
706 			val |= inet->tos & INET_ECN_MASK;
707 		}
708 		if (inet->tos != val) {
709 			inet->tos = val;
710 			sk->sk_priority = rt_tos2priority(val);
711 			sk_dst_reset(sk);
712 		}
713 		break;
714 	case IP_TTL:
715 		if (optlen < 1)
716 			goto e_inval;
717 		if (val != -1 && (val < 1 || val > 255))
718 			goto e_inval;
719 		inet->uc_ttl = val;
720 		break;
721 	case IP_HDRINCL:
722 		if (sk->sk_type != SOCK_RAW) {
723 			err = -ENOPROTOOPT;
724 			break;
725 		}
726 		inet->hdrincl = val ? 1 : 0;
727 		break;
728 	case IP_NODEFRAG:
729 		if (sk->sk_type != SOCK_RAW) {
730 			err = -ENOPROTOOPT;
731 			break;
732 		}
733 		inet->nodefrag = val ? 1 : 0;
734 		break;
735 	case IP_MTU_DISCOVER:
736 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
737 			goto e_inval;
738 		inet->pmtudisc = val;
739 		break;
740 	case IP_RECVERR:
741 		inet->recverr = !!val;
742 		if (!val)
743 			skb_queue_purge(&sk->sk_error_queue);
744 		break;
745 	case IP_MULTICAST_TTL:
746 		if (sk->sk_type == SOCK_STREAM)
747 			goto e_inval;
748 		if (optlen < 1)
749 			goto e_inval;
750 		if (val == -1)
751 			val = 1;
752 		if (val < 0 || val > 255)
753 			goto e_inval;
754 		inet->mc_ttl = val;
755 		break;
756 	case IP_MULTICAST_LOOP:
757 		if (optlen < 1)
758 			goto e_inval;
759 		inet->mc_loop = !!val;
760 		break;
761 	case IP_UNICAST_IF:
762 	{
763 		struct net_device *dev = NULL;
764 		int ifindex;
765 
766 		if (optlen != sizeof(int))
767 			goto e_inval;
768 
769 		ifindex = (__force int)ntohl((__force __be32)val);
770 		if (ifindex == 0) {
771 			inet->uc_index = 0;
772 			err = 0;
773 			break;
774 		}
775 
776 		dev = dev_get_by_index(sock_net(sk), ifindex);
777 		err = -EADDRNOTAVAIL;
778 		if (!dev)
779 			break;
780 		dev_put(dev);
781 
782 		err = -EINVAL;
783 		if (sk->sk_bound_dev_if)
784 			break;
785 
786 		inet->uc_index = ifindex;
787 		err = 0;
788 		break;
789 	}
790 	case IP_MULTICAST_IF:
791 	{
792 		struct ip_mreqn mreq;
793 		struct net_device *dev = NULL;
794 
795 		if (sk->sk_type == SOCK_STREAM)
796 			goto e_inval;
797 		/*
798 		 *	Check the arguments are allowable
799 		 */
800 
801 		if (optlen < sizeof(struct in_addr))
802 			goto e_inval;
803 
804 		err = -EFAULT;
805 		if (optlen >= sizeof(struct ip_mreqn)) {
806 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
807 				break;
808 		} else {
809 			memset(&mreq, 0, sizeof(mreq));
810 			if (optlen >= sizeof(struct ip_mreq)) {
811 				if (copy_from_user(&mreq, optval,
812 						   sizeof(struct ip_mreq)))
813 					break;
814 			} else if (optlen >= sizeof(struct in_addr)) {
815 				if (copy_from_user(&mreq.imr_address, optval,
816 						   sizeof(struct in_addr)))
817 					break;
818 			}
819 		}
820 
821 		if (!mreq.imr_ifindex) {
822 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
823 				inet->mc_index = 0;
824 				inet->mc_addr  = 0;
825 				err = 0;
826 				break;
827 			}
828 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
829 			if (dev)
830 				mreq.imr_ifindex = dev->ifindex;
831 		} else
832 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
833 
834 
835 		err = -EADDRNOTAVAIL;
836 		if (!dev)
837 			break;
838 		dev_put(dev);
839 
840 		err = -EINVAL;
841 		if (sk->sk_bound_dev_if &&
842 		    mreq.imr_ifindex != sk->sk_bound_dev_if)
843 			break;
844 
845 		inet->mc_index = mreq.imr_ifindex;
846 		inet->mc_addr  = mreq.imr_address.s_addr;
847 		err = 0;
848 		break;
849 	}
850 
851 	case IP_ADD_MEMBERSHIP:
852 	case IP_DROP_MEMBERSHIP:
853 	{
854 		struct ip_mreqn mreq;
855 
856 		err = -EPROTO;
857 		if (inet_sk(sk)->is_icsk)
858 			break;
859 
860 		if (optlen < sizeof(struct ip_mreq))
861 			goto e_inval;
862 		err = -EFAULT;
863 		if (optlen >= sizeof(struct ip_mreqn)) {
864 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
865 				break;
866 		} else {
867 			memset(&mreq, 0, sizeof(mreq));
868 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
869 				break;
870 		}
871 
872 		if (optname == IP_ADD_MEMBERSHIP)
873 			err = ip_mc_join_group(sk, &mreq);
874 		else
875 			err = ip_mc_leave_group(sk, &mreq);
876 		break;
877 	}
878 	case IP_MSFILTER:
879 	{
880 		struct ip_msfilter *msf;
881 
882 		if (optlen < IP_MSFILTER_SIZE(0))
883 			goto e_inval;
884 		if (optlen > sysctl_optmem_max) {
885 			err = -ENOBUFS;
886 			break;
887 		}
888 		msf = kmalloc(optlen, GFP_KERNEL);
889 		if (!msf) {
890 			err = -ENOBUFS;
891 			break;
892 		}
893 		err = -EFAULT;
894 		if (copy_from_user(msf, optval, optlen)) {
895 			kfree(msf);
896 			break;
897 		}
898 		/* numsrc >= (1G-4) overflow in 32 bits */
899 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
900 		    msf->imsf_numsrc > sysctl_igmp_max_msf) {
901 			kfree(msf);
902 			err = -ENOBUFS;
903 			break;
904 		}
905 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
906 			kfree(msf);
907 			err = -EINVAL;
908 			break;
909 		}
910 		err = ip_mc_msfilter(sk, msf, 0);
911 		kfree(msf);
912 		break;
913 	}
914 	case IP_BLOCK_SOURCE:
915 	case IP_UNBLOCK_SOURCE:
916 	case IP_ADD_SOURCE_MEMBERSHIP:
917 	case IP_DROP_SOURCE_MEMBERSHIP:
918 	{
919 		struct ip_mreq_source mreqs;
920 		int omode, add;
921 
922 		if (optlen != sizeof(struct ip_mreq_source))
923 			goto e_inval;
924 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
925 			err = -EFAULT;
926 			break;
927 		}
928 		if (optname == IP_BLOCK_SOURCE) {
929 			omode = MCAST_EXCLUDE;
930 			add = 1;
931 		} else if (optname == IP_UNBLOCK_SOURCE) {
932 			omode = MCAST_EXCLUDE;
933 			add = 0;
934 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
935 			struct ip_mreqn mreq;
936 
937 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
938 			mreq.imr_address.s_addr = mreqs.imr_interface;
939 			mreq.imr_ifindex = 0;
940 			err = ip_mc_join_group(sk, &mreq);
941 			if (err && err != -EADDRINUSE)
942 				break;
943 			omode = MCAST_INCLUDE;
944 			add = 1;
945 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
946 			omode = MCAST_INCLUDE;
947 			add = 0;
948 		}
949 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
950 		break;
951 	}
952 	case MCAST_JOIN_GROUP:
953 	case MCAST_LEAVE_GROUP:
954 	{
955 		struct group_req greq;
956 		struct sockaddr_in *psin;
957 		struct ip_mreqn mreq;
958 
959 		if (optlen < sizeof(struct group_req))
960 			goto e_inval;
961 		err = -EFAULT;
962 		if (copy_from_user(&greq, optval, sizeof(greq)))
963 			break;
964 		psin = (struct sockaddr_in *)&greq.gr_group;
965 		if (psin->sin_family != AF_INET)
966 			goto e_inval;
967 		memset(&mreq, 0, sizeof(mreq));
968 		mreq.imr_multiaddr = psin->sin_addr;
969 		mreq.imr_ifindex = greq.gr_interface;
970 
971 		if (optname == MCAST_JOIN_GROUP)
972 			err = ip_mc_join_group(sk, &mreq);
973 		else
974 			err = ip_mc_leave_group(sk, &mreq);
975 		break;
976 	}
977 	case MCAST_JOIN_SOURCE_GROUP:
978 	case MCAST_LEAVE_SOURCE_GROUP:
979 	case MCAST_BLOCK_SOURCE:
980 	case MCAST_UNBLOCK_SOURCE:
981 	{
982 		struct group_source_req greqs;
983 		struct ip_mreq_source mreqs;
984 		struct sockaddr_in *psin;
985 		int omode, add;
986 
987 		if (optlen != sizeof(struct group_source_req))
988 			goto e_inval;
989 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
990 			err = -EFAULT;
991 			break;
992 		}
993 		if (greqs.gsr_group.ss_family != AF_INET ||
994 		    greqs.gsr_source.ss_family != AF_INET) {
995 			err = -EADDRNOTAVAIL;
996 			break;
997 		}
998 		psin = (struct sockaddr_in *)&greqs.gsr_group;
999 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1000 		psin = (struct sockaddr_in *)&greqs.gsr_source;
1001 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1002 		mreqs.imr_interface = 0; /* use index for mc_source */
1003 
1004 		if (optname == MCAST_BLOCK_SOURCE) {
1005 			omode = MCAST_EXCLUDE;
1006 			add = 1;
1007 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
1008 			omode = MCAST_EXCLUDE;
1009 			add = 0;
1010 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1011 			struct ip_mreqn mreq;
1012 
1013 			psin = (struct sockaddr_in *)&greqs.gsr_group;
1014 			mreq.imr_multiaddr = psin->sin_addr;
1015 			mreq.imr_address.s_addr = 0;
1016 			mreq.imr_ifindex = greqs.gsr_interface;
1017 			err = ip_mc_join_group(sk, &mreq);
1018 			if (err && err != -EADDRINUSE)
1019 				break;
1020 			greqs.gsr_interface = mreq.imr_ifindex;
1021 			omode = MCAST_INCLUDE;
1022 			add = 1;
1023 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
1024 			omode = MCAST_INCLUDE;
1025 			add = 0;
1026 		}
1027 		err = ip_mc_source(add, omode, sk, &mreqs,
1028 				   greqs.gsr_interface);
1029 		break;
1030 	}
1031 	case MCAST_MSFILTER:
1032 	{
1033 		struct sockaddr_in *psin;
1034 		struct ip_msfilter *msf = NULL;
1035 		struct group_filter *gsf = NULL;
1036 		int msize, i, ifindex;
1037 
1038 		if (optlen < GROUP_FILTER_SIZE(0))
1039 			goto e_inval;
1040 		if (optlen > sysctl_optmem_max) {
1041 			err = -ENOBUFS;
1042 			break;
1043 		}
1044 		gsf = kmalloc(optlen, GFP_KERNEL);
1045 		if (!gsf) {
1046 			err = -ENOBUFS;
1047 			break;
1048 		}
1049 		err = -EFAULT;
1050 		if (copy_from_user(gsf, optval, optlen))
1051 			goto mc_msf_out;
1052 
1053 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1054 		if (gsf->gf_numsrc >= 0x1ffffff ||
1055 		    gsf->gf_numsrc > sysctl_igmp_max_msf) {
1056 			err = -ENOBUFS;
1057 			goto mc_msf_out;
1058 		}
1059 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1060 			err = -EINVAL;
1061 			goto mc_msf_out;
1062 		}
1063 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1064 		msf = kmalloc(msize, GFP_KERNEL);
1065 		if (!msf) {
1066 			err = -ENOBUFS;
1067 			goto mc_msf_out;
1068 		}
1069 		ifindex = gsf->gf_interface;
1070 		psin = (struct sockaddr_in *)&gsf->gf_group;
1071 		if (psin->sin_family != AF_INET) {
1072 			err = -EADDRNOTAVAIL;
1073 			goto mc_msf_out;
1074 		}
1075 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1076 		msf->imsf_interface = 0;
1077 		msf->imsf_fmode = gsf->gf_fmode;
1078 		msf->imsf_numsrc = gsf->gf_numsrc;
1079 		err = -EADDRNOTAVAIL;
1080 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1081 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1082 
1083 			if (psin->sin_family != AF_INET)
1084 				goto mc_msf_out;
1085 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1086 		}
1087 		kfree(gsf);
1088 		gsf = NULL;
1089 
1090 		err = ip_mc_msfilter(sk, msf, ifindex);
1091 mc_msf_out:
1092 		kfree(msf);
1093 		kfree(gsf);
1094 		break;
1095 	}
1096 	case IP_MULTICAST_ALL:
1097 		if (optlen < 1)
1098 			goto e_inval;
1099 		if (val != 0 && val != 1)
1100 			goto e_inval;
1101 		inet->mc_all = val;
1102 		break;
1103 	case IP_ROUTER_ALERT:
1104 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1105 		break;
1106 
1107 	case IP_FREEBIND:
1108 		if (optlen < 1)
1109 			goto e_inval;
1110 		inet->freebind = !!val;
1111 		break;
1112 
1113 	case IP_IPSEC_POLICY:
1114 	case IP_XFRM_POLICY:
1115 		err = -EPERM;
1116 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1117 			break;
1118 		err = xfrm_user_policy(sk, optname, optval, optlen);
1119 		break;
1120 
1121 	case IP_TRANSPARENT:
1122 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1123 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1124 			err = -EPERM;
1125 			break;
1126 		}
1127 		if (optlen < 1)
1128 			goto e_inval;
1129 		inet->transparent = !!val;
1130 		break;
1131 
1132 	case IP_MINTTL:
1133 		if (optlen < 1)
1134 			goto e_inval;
1135 		if (val < 0 || val > 255)
1136 			goto e_inval;
1137 		inet->min_ttl = val;
1138 		break;
1139 
1140 	default:
1141 		err = -ENOPROTOOPT;
1142 		break;
1143 	}
1144 	release_sock(sk);
1145 	if (needs_rtnl)
1146 		rtnl_unlock();
1147 	return err;
1148 
1149 e_inval:
1150 	release_sock(sk);
1151 	if (needs_rtnl)
1152 		rtnl_unlock();
1153 	return -EINVAL;
1154 }
1155 
1156 /**
1157  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1158  * @sk: socket
1159  * @skb: buffer
1160  *
1161  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1162  * destination in skb->cb[] before dst drop.
1163  * This way, receiver doesn't make cache line misses to read rtable.
1164  */
1165 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1166 {
1167 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1168 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1169 		       ipv6_sk_rxinfo(sk);
1170 
1171 	if (prepare && skb_rtable(skb)) {
1172 		pktinfo->ipi_ifindex = inet_iif(skb);
1173 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1174 	} else {
1175 		pktinfo->ipi_ifindex = 0;
1176 		pktinfo->ipi_spec_dst.s_addr = 0;
1177 	}
1178 	skb_dst_drop(skb);
1179 }
1180 
1181 int ip_setsockopt(struct sock *sk, int level,
1182 		int optname, char __user *optval, unsigned int optlen)
1183 {
1184 	int err;
1185 
1186 	if (level != SOL_IP)
1187 		return -ENOPROTOOPT;
1188 
1189 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1190 #ifdef CONFIG_NETFILTER
1191 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1192 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1193 			optname != IP_IPSEC_POLICY &&
1194 			optname != IP_XFRM_POLICY &&
1195 			!ip_mroute_opt(optname)) {
1196 		lock_sock(sk);
1197 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1198 		release_sock(sk);
1199 	}
1200 #endif
1201 	return err;
1202 }
1203 EXPORT_SYMBOL(ip_setsockopt);
1204 
1205 #ifdef CONFIG_COMPAT
1206 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1207 			 char __user *optval, unsigned int optlen)
1208 {
1209 	int err;
1210 
1211 	if (level != SOL_IP)
1212 		return -ENOPROTOOPT;
1213 
1214 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1215 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1216 			ip_setsockopt);
1217 
1218 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1219 #ifdef CONFIG_NETFILTER
1220 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1221 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1222 			optname != IP_IPSEC_POLICY &&
1223 			optname != IP_XFRM_POLICY &&
1224 			!ip_mroute_opt(optname)) {
1225 		lock_sock(sk);
1226 		err = compat_nf_setsockopt(sk, PF_INET, optname,
1227 					   optval, optlen);
1228 		release_sock(sk);
1229 	}
1230 #endif
1231 	return err;
1232 }
1233 EXPORT_SYMBOL(compat_ip_setsockopt);
1234 #endif
1235 
1236 /*
1237  *	Get the options. Note for future reference. The GET of IP options gets
1238  *	the _received_ ones. The set sets the _sent_ ones.
1239  */
1240 
1241 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1242 			    char __user *optval, int __user *optlen, unsigned int flags)
1243 {
1244 	struct inet_sock *inet = inet_sk(sk);
1245 	int val;
1246 	int len;
1247 
1248 	if (level != SOL_IP)
1249 		return -EOPNOTSUPP;
1250 
1251 	if (ip_mroute_opt(optname))
1252 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1253 
1254 	if (get_user(len, optlen))
1255 		return -EFAULT;
1256 	if (len < 0)
1257 		return -EINVAL;
1258 
1259 	lock_sock(sk);
1260 
1261 	switch (optname) {
1262 	case IP_OPTIONS:
1263 	{
1264 		unsigned char optbuf[sizeof(struct ip_options)+40];
1265 		struct ip_options *opt = (struct ip_options *)optbuf;
1266 		struct ip_options_rcu *inet_opt;
1267 
1268 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1269 						     sock_owned_by_user(sk));
1270 		opt->optlen = 0;
1271 		if (inet_opt)
1272 			memcpy(optbuf, &inet_opt->opt,
1273 			       sizeof(struct ip_options) +
1274 			       inet_opt->opt.optlen);
1275 		release_sock(sk);
1276 
1277 		if (opt->optlen == 0)
1278 			return put_user(0, optlen);
1279 
1280 		ip_options_undo(opt);
1281 
1282 		len = min_t(unsigned int, len, opt->optlen);
1283 		if (put_user(len, optlen))
1284 			return -EFAULT;
1285 		if (copy_to_user(optval, opt->__data, len))
1286 			return -EFAULT;
1287 		return 0;
1288 	}
1289 	case IP_PKTINFO:
1290 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1291 		break;
1292 	case IP_RECVTTL:
1293 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1294 		break;
1295 	case IP_RECVTOS:
1296 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1297 		break;
1298 	case IP_RECVOPTS:
1299 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1300 		break;
1301 	case IP_RETOPTS:
1302 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1303 		break;
1304 	case IP_PASSSEC:
1305 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1306 		break;
1307 	case IP_RECVORIGDSTADDR:
1308 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1309 		break;
1310 	case IP_CHECKSUM:
1311 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1312 		break;
1313 	case IP_TOS:
1314 		val = inet->tos;
1315 		break;
1316 	case IP_TTL:
1317 		val = (inet->uc_ttl == -1 ?
1318 		       sysctl_ip_default_ttl :
1319 		       inet->uc_ttl);
1320 		break;
1321 	case IP_HDRINCL:
1322 		val = inet->hdrincl;
1323 		break;
1324 	case IP_NODEFRAG:
1325 		val = inet->nodefrag;
1326 		break;
1327 	case IP_MTU_DISCOVER:
1328 		val = inet->pmtudisc;
1329 		break;
1330 	case IP_MTU:
1331 	{
1332 		struct dst_entry *dst;
1333 		val = 0;
1334 		dst = sk_dst_get(sk);
1335 		if (dst) {
1336 			val = dst_mtu(dst);
1337 			dst_release(dst);
1338 		}
1339 		if (!val) {
1340 			release_sock(sk);
1341 			return -ENOTCONN;
1342 		}
1343 		break;
1344 	}
1345 	case IP_RECVERR:
1346 		val = inet->recverr;
1347 		break;
1348 	case IP_MULTICAST_TTL:
1349 		val = inet->mc_ttl;
1350 		break;
1351 	case IP_MULTICAST_LOOP:
1352 		val = inet->mc_loop;
1353 		break;
1354 	case IP_UNICAST_IF:
1355 		val = (__force int)htonl((__u32) inet->uc_index);
1356 		break;
1357 	case IP_MULTICAST_IF:
1358 	{
1359 		struct in_addr addr;
1360 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1361 		addr.s_addr = inet->mc_addr;
1362 		release_sock(sk);
1363 
1364 		if (put_user(len, optlen))
1365 			return -EFAULT;
1366 		if (copy_to_user(optval, &addr, len))
1367 			return -EFAULT;
1368 		return 0;
1369 	}
1370 	case IP_MSFILTER:
1371 	{
1372 		struct ip_msfilter msf;
1373 		int err;
1374 
1375 		if (len < IP_MSFILTER_SIZE(0)) {
1376 			release_sock(sk);
1377 			return -EINVAL;
1378 		}
1379 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1380 			release_sock(sk);
1381 			return -EFAULT;
1382 		}
1383 		err = ip_mc_msfget(sk, &msf,
1384 				   (struct ip_msfilter __user *)optval, optlen);
1385 		release_sock(sk);
1386 		return err;
1387 	}
1388 	case MCAST_MSFILTER:
1389 	{
1390 		struct group_filter gsf;
1391 		int err;
1392 
1393 		if (len < GROUP_FILTER_SIZE(0)) {
1394 			release_sock(sk);
1395 			return -EINVAL;
1396 		}
1397 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1398 			release_sock(sk);
1399 			return -EFAULT;
1400 		}
1401 		err = ip_mc_gsfget(sk, &gsf,
1402 				   (struct group_filter __user *)optval,
1403 				   optlen);
1404 		release_sock(sk);
1405 		return err;
1406 	}
1407 	case IP_MULTICAST_ALL:
1408 		val = inet->mc_all;
1409 		break;
1410 	case IP_PKTOPTIONS:
1411 	{
1412 		struct msghdr msg;
1413 
1414 		release_sock(sk);
1415 
1416 		if (sk->sk_type != SOCK_STREAM)
1417 			return -ENOPROTOOPT;
1418 
1419 		msg.msg_control = (__force void *) optval;
1420 		msg.msg_controllen = len;
1421 		msg.msg_flags = flags;
1422 
1423 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1424 			struct in_pktinfo info;
1425 
1426 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1427 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1428 			info.ipi_ifindex = inet->mc_index;
1429 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1430 		}
1431 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1432 			int hlim = inet->mc_ttl;
1433 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1434 		}
1435 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1436 			int tos = inet->rcv_tos;
1437 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1438 		}
1439 		len -= msg.msg_controllen;
1440 		return put_user(len, optlen);
1441 	}
1442 	case IP_FREEBIND:
1443 		val = inet->freebind;
1444 		break;
1445 	case IP_TRANSPARENT:
1446 		val = inet->transparent;
1447 		break;
1448 	case IP_MINTTL:
1449 		val = inet->min_ttl;
1450 		break;
1451 	default:
1452 		release_sock(sk);
1453 		return -ENOPROTOOPT;
1454 	}
1455 	release_sock(sk);
1456 
1457 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1458 		unsigned char ucval = (unsigned char)val;
1459 		len = 1;
1460 		if (put_user(len, optlen))
1461 			return -EFAULT;
1462 		if (copy_to_user(optval, &ucval, 1))
1463 			return -EFAULT;
1464 	} else {
1465 		len = min_t(unsigned int, sizeof(int), len);
1466 		if (put_user(len, optlen))
1467 			return -EFAULT;
1468 		if (copy_to_user(optval, &val, len))
1469 			return -EFAULT;
1470 	}
1471 	return 0;
1472 }
1473 
1474 int ip_getsockopt(struct sock *sk, int level,
1475 		  int optname, char __user *optval, int __user *optlen)
1476 {
1477 	int err;
1478 
1479 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1480 #ifdef CONFIG_NETFILTER
1481 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1482 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1483 			!ip_mroute_opt(optname)) {
1484 		int len;
1485 
1486 		if (get_user(len, optlen))
1487 			return -EFAULT;
1488 
1489 		lock_sock(sk);
1490 		err = nf_getsockopt(sk, PF_INET, optname, optval,
1491 				&len);
1492 		release_sock(sk);
1493 		if (err >= 0)
1494 			err = put_user(len, optlen);
1495 		return err;
1496 	}
1497 #endif
1498 	return err;
1499 }
1500 EXPORT_SYMBOL(ip_getsockopt);
1501 
1502 #ifdef CONFIG_COMPAT
1503 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1504 			 char __user *optval, int __user *optlen)
1505 {
1506 	int err;
1507 
1508 	if (optname == MCAST_MSFILTER)
1509 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1510 			ip_getsockopt);
1511 
1512 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1513 		MSG_CMSG_COMPAT);
1514 
1515 #ifdef CONFIG_NETFILTER
1516 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1517 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1518 			!ip_mroute_opt(optname)) {
1519 		int len;
1520 
1521 		if (get_user(len, optlen))
1522 			return -EFAULT;
1523 
1524 		lock_sock(sk);
1525 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1526 		release_sock(sk);
1527 		if (err >= 0)
1528 			err = put_user(len, optlen);
1529 		return err;
1530 	}
1531 #endif
1532 	return err;
1533 }
1534 EXPORT_SYMBOL(compat_ip_getsockopt);
1535 #endif
1536