xref: /linux/net/ipv4/ip_sockglue.c (revision b68fc09be48edbc47de1a0f3d42ef8adf6c0ac55)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		The IP to API glue.
8  *
9  * Authors:	see ip.c
10  *
11  * Fixes:
12  *		Many		:	Split from ip.c , see ip.c for history.
13  *		Martin Mares	:	TOS setting fixed.
14  *		Alan Cox	:	Fixed a couple of oopses in Martin's
15  *					TOS tweaks.
16  *		Mike McLagan	:	Routing by source
17  */
18 
19 #include <linux/module.h>
20 #include <linux/types.h>
21 #include <linux/mm.h>
22 #include <linux/skbuff.h>
23 #include <linux/ip.h>
24 #include <linux/icmp.h>
25 #include <linux/inetdevice.h>
26 #include <linux/netdevice.h>
27 #include <linux/slab.h>
28 #include <net/sock.h>
29 #include <net/ip.h>
30 #include <net/icmp.h>
31 #include <net/tcp_states.h>
32 #include <linux/udp.h>
33 #include <linux/igmp.h>
34 #include <linux/netfilter.h>
35 #include <linux/route.h>
36 #include <linux/mroute.h>
37 #include <net/inet_ecn.h>
38 #include <net/route.h>
39 #include <net/xfrm.h>
40 #include <net/compat.h>
41 #include <net/checksum.h>
42 #if IS_ENABLED(CONFIG_IPV6)
43 #include <net/transp_v6.h>
44 #endif
45 #include <net/ip_fib.h>
46 
47 #include <linux/errqueue.h>
48 #include <linux/uaccess.h>
49 
50 #include <linux/bpfilter.h>
51 
52 /*
53  *	SOL_IP control messages.
54  */
55 
56 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
57 {
58 	struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
59 
60 	info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
61 
62 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
63 }
64 
65 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
66 {
67 	int ttl = ip_hdr(skb)->ttl;
68 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
69 }
70 
71 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
72 {
73 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
74 }
75 
76 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
77 {
78 	if (IPCB(skb)->opt.optlen == 0)
79 		return;
80 
81 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
82 		 ip_hdr(skb) + 1);
83 }
84 
85 
86 static void ip_cmsg_recv_retopts(struct net *net, struct msghdr *msg,
87 				 struct sk_buff *skb)
88 {
89 	unsigned char optbuf[sizeof(struct ip_options) + 40];
90 	struct ip_options *opt = (struct ip_options *)optbuf;
91 
92 	if (IPCB(skb)->opt.optlen == 0)
93 		return;
94 
95 	if (ip_options_echo(net, opt, skb)) {
96 		msg->msg_flags |= MSG_CTRUNC;
97 		return;
98 	}
99 	ip_options_undo(opt);
100 
101 	put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
102 }
103 
104 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb)
105 {
106 	int val;
107 
108 	if (IPCB(skb)->frag_max_size == 0)
109 		return;
110 
111 	val = IPCB(skb)->frag_max_size;
112 	put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
113 }
114 
115 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
116 				  int tlen, int offset)
117 {
118 	__wsum csum = skb->csum;
119 
120 	if (skb->ip_summed != CHECKSUM_COMPLETE)
121 		return;
122 
123 	if (offset != 0) {
124 		int tend_off = skb_transport_offset(skb) + tlen;
125 		csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
126 	}
127 
128 	put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
129 }
130 
131 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
132 {
133 	char *secdata;
134 	u32 seclen, secid;
135 	int err;
136 
137 	err = security_socket_getpeersec_dgram(NULL, skb, &secid);
138 	if (err)
139 		return;
140 
141 	err = security_secid_to_secctx(secid, &secdata, &seclen);
142 	if (err)
143 		return;
144 
145 	put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
146 	security_release_secctx(secdata, seclen);
147 }
148 
149 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
150 {
151 	struct sockaddr_in sin;
152 	__be16 *ports;
153 	int end;
154 
155 	end = skb_transport_offset(skb) + 4;
156 	if (end > 0 && !pskb_may_pull(skb, end))
157 		return;
158 
159 	/* All current transport protocols have the port numbers in the
160 	 * first four bytes of the transport header and this function is
161 	 * written with this assumption in mind.
162 	 */
163 	ports = (__be16 *)skb_transport_header(skb);
164 
165 	sin.sin_family = AF_INET;
166 	sin.sin_addr.s_addr = ip_hdr(skb)->daddr;
167 	sin.sin_port = ports[1];
168 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
169 
170 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
171 }
172 
173 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
174 			 struct sk_buff *skb, int tlen, int offset)
175 {
176 	struct inet_sock *inet = inet_sk(sk);
177 	unsigned int flags = inet->cmsg_flags;
178 
179 	/* Ordered by supposed usage frequency */
180 	if (flags & IP_CMSG_PKTINFO) {
181 		ip_cmsg_recv_pktinfo(msg, skb);
182 
183 		flags &= ~IP_CMSG_PKTINFO;
184 		if (!flags)
185 			return;
186 	}
187 
188 	if (flags & IP_CMSG_TTL) {
189 		ip_cmsg_recv_ttl(msg, skb);
190 
191 		flags &= ~IP_CMSG_TTL;
192 		if (!flags)
193 			return;
194 	}
195 
196 	if (flags & IP_CMSG_TOS) {
197 		ip_cmsg_recv_tos(msg, skb);
198 
199 		flags &= ~IP_CMSG_TOS;
200 		if (!flags)
201 			return;
202 	}
203 
204 	if (flags & IP_CMSG_RECVOPTS) {
205 		ip_cmsg_recv_opts(msg, skb);
206 
207 		flags &= ~IP_CMSG_RECVOPTS;
208 		if (!flags)
209 			return;
210 	}
211 
212 	if (flags & IP_CMSG_RETOPTS) {
213 		ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
214 
215 		flags &= ~IP_CMSG_RETOPTS;
216 		if (!flags)
217 			return;
218 	}
219 
220 	if (flags & IP_CMSG_PASSSEC) {
221 		ip_cmsg_recv_security(msg, skb);
222 
223 		flags &= ~IP_CMSG_PASSSEC;
224 		if (!flags)
225 			return;
226 	}
227 
228 	if (flags & IP_CMSG_ORIGDSTADDR) {
229 		ip_cmsg_recv_dstaddr(msg, skb);
230 
231 		flags &= ~IP_CMSG_ORIGDSTADDR;
232 		if (!flags)
233 			return;
234 	}
235 
236 	if (flags & IP_CMSG_CHECKSUM)
237 		ip_cmsg_recv_checksum(msg, skb, tlen, offset);
238 
239 	if (flags & IP_CMSG_RECVFRAGSIZE)
240 		ip_cmsg_recv_fragsize(msg, skb);
241 }
242 EXPORT_SYMBOL(ip_cmsg_recv_offset);
243 
244 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
245 		 bool allow_ipv6)
246 {
247 	int err, val;
248 	struct cmsghdr *cmsg;
249 	struct net *net = sock_net(sk);
250 
251 	for_each_cmsghdr(cmsg, msg) {
252 		if (!CMSG_OK(msg, cmsg))
253 			return -EINVAL;
254 #if IS_ENABLED(CONFIG_IPV6)
255 		if (allow_ipv6 &&
256 		    cmsg->cmsg_level == SOL_IPV6 &&
257 		    cmsg->cmsg_type == IPV6_PKTINFO) {
258 			struct in6_pktinfo *src_info;
259 
260 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
261 				return -EINVAL;
262 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
263 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
264 				return -EINVAL;
265 			if (src_info->ipi6_ifindex)
266 				ipc->oif = src_info->ipi6_ifindex;
267 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
268 			continue;
269 		}
270 #endif
271 		if (cmsg->cmsg_level == SOL_SOCKET) {
272 			err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
273 			if (err)
274 				return err;
275 			continue;
276 		}
277 
278 		if (cmsg->cmsg_level != SOL_IP)
279 			continue;
280 		switch (cmsg->cmsg_type) {
281 		case IP_RETOPTS:
282 			err = cmsg->cmsg_len - sizeof(struct cmsghdr);
283 
284 			/* Our caller is responsible for freeing ipc->opt */
285 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
286 					     err < 40 ? err : 40);
287 			if (err)
288 				return err;
289 			break;
290 		case IP_PKTINFO:
291 		{
292 			struct in_pktinfo *info;
293 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
294 				return -EINVAL;
295 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
296 			if (info->ipi_ifindex)
297 				ipc->oif = info->ipi_ifindex;
298 			ipc->addr = info->ipi_spec_dst.s_addr;
299 			break;
300 		}
301 		case IP_TTL:
302 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
303 				return -EINVAL;
304 			val = *(int *)CMSG_DATA(cmsg);
305 			if (val < 1 || val > 255)
306 				return -EINVAL;
307 			ipc->ttl = val;
308 			break;
309 		case IP_TOS:
310 			if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
311 				val = *(int *)CMSG_DATA(cmsg);
312 			else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
313 				val = *(u8 *)CMSG_DATA(cmsg);
314 			else
315 				return -EINVAL;
316 			if (val < 0 || val > 255)
317 				return -EINVAL;
318 			ipc->tos = val;
319 			ipc->priority = rt_tos2priority(ipc->tos);
320 			break;
321 
322 		default:
323 			return -EINVAL;
324 		}
325 	}
326 	return 0;
327 }
328 
329 static void ip_ra_destroy_rcu(struct rcu_head *head)
330 {
331 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
332 
333 	sock_put(ra->saved_sk);
334 	kfree(ra);
335 }
336 
337 int ip_ra_control(struct sock *sk, unsigned char on,
338 		  void (*destructor)(struct sock *))
339 {
340 	struct ip_ra_chain *ra, *new_ra;
341 	struct ip_ra_chain __rcu **rap;
342 	struct net *net = sock_net(sk);
343 
344 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
345 		return -EINVAL;
346 
347 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
348 
349 	mutex_lock(&net->ipv4.ra_mutex);
350 	for (rap = &net->ipv4.ra_chain;
351 	     (ra = rcu_dereference_protected(*rap,
352 			lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
353 	     rap = &ra->next) {
354 		if (ra->sk == sk) {
355 			if (on) {
356 				mutex_unlock(&net->ipv4.ra_mutex);
357 				kfree(new_ra);
358 				return -EADDRINUSE;
359 			}
360 			/* dont let ip_call_ra_chain() use sk again */
361 			ra->sk = NULL;
362 			RCU_INIT_POINTER(*rap, ra->next);
363 			mutex_unlock(&net->ipv4.ra_mutex);
364 
365 			if (ra->destructor)
366 				ra->destructor(sk);
367 			/*
368 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
369 			 * period. This guarantee ip_call_ra_chain() dont need
370 			 * to mess with socket refcounts.
371 			 */
372 			ra->saved_sk = sk;
373 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
374 			return 0;
375 		}
376 	}
377 	if (!new_ra) {
378 		mutex_unlock(&net->ipv4.ra_mutex);
379 		return -ENOBUFS;
380 	}
381 	new_ra->sk = sk;
382 	new_ra->destructor = destructor;
383 
384 	RCU_INIT_POINTER(new_ra->next, ra);
385 	rcu_assign_pointer(*rap, new_ra);
386 	sock_hold(sk);
387 	mutex_unlock(&net->ipv4.ra_mutex);
388 
389 	return 0;
390 }
391 
392 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
393 		   __be16 port, u32 info, u8 *payload)
394 {
395 	struct sock_exterr_skb *serr;
396 
397 	skb = skb_clone(skb, GFP_ATOMIC);
398 	if (!skb)
399 		return;
400 
401 	serr = SKB_EXT_ERR(skb);
402 	serr->ee.ee_errno = err;
403 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
404 	serr->ee.ee_type = icmp_hdr(skb)->type;
405 	serr->ee.ee_code = icmp_hdr(skb)->code;
406 	serr->ee.ee_pad = 0;
407 	serr->ee.ee_info = info;
408 	serr->ee.ee_data = 0;
409 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
410 				   skb_network_header(skb);
411 	serr->port = port;
412 
413 	if (skb_pull(skb, payload - skb->data)) {
414 		skb_reset_transport_header(skb);
415 		if (sock_queue_err_skb(sk, skb) == 0)
416 			return;
417 	}
418 	kfree_skb(skb);
419 }
420 
421 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
422 {
423 	struct inet_sock *inet = inet_sk(sk);
424 	struct sock_exterr_skb *serr;
425 	struct iphdr *iph;
426 	struct sk_buff *skb;
427 
428 	if (!inet->recverr)
429 		return;
430 
431 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
432 	if (!skb)
433 		return;
434 
435 	skb_put(skb, sizeof(struct iphdr));
436 	skb_reset_network_header(skb);
437 	iph = ip_hdr(skb);
438 	iph->daddr = daddr;
439 
440 	serr = SKB_EXT_ERR(skb);
441 	serr->ee.ee_errno = err;
442 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
443 	serr->ee.ee_type = 0;
444 	serr->ee.ee_code = 0;
445 	serr->ee.ee_pad = 0;
446 	serr->ee.ee_info = info;
447 	serr->ee.ee_data = 0;
448 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
449 	serr->port = port;
450 
451 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
452 	skb_reset_transport_header(skb);
453 
454 	if (sock_queue_err_skb(sk, skb))
455 		kfree_skb(skb);
456 }
457 
458 /* For some errors we have valid addr_offset even with zero payload and
459  * zero port. Also, addr_offset should be supported if port is set.
460  */
461 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
462 {
463 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
464 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
465 }
466 
467 /* IPv4 supports cmsg on all imcp errors and some timestamps
468  *
469  * Timestamp code paths do not initialize the fields expected by cmsg:
470  * the PKTINFO fields in skb->cb[]. Fill those in here.
471  */
472 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
473 				       struct sk_buff *skb,
474 				       int ee_origin)
475 {
476 	struct in_pktinfo *info;
477 
478 	if (ee_origin == SO_EE_ORIGIN_ICMP)
479 		return true;
480 
481 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
482 		return false;
483 
484 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
485 	 * timestamp with egress dev. Not possible for packets without iif
486 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
487 	 */
488 	info = PKTINFO_SKB_CB(skb);
489 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
490 	    !info->ipi_ifindex)
491 		return false;
492 
493 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
494 	return true;
495 }
496 
497 /*
498  *	Handle MSG_ERRQUEUE
499  */
500 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
501 {
502 	struct sock_exterr_skb *serr;
503 	struct sk_buff *skb;
504 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
505 	struct {
506 		struct sock_extended_err ee;
507 		struct sockaddr_in	 offender;
508 	} errhdr;
509 	int err;
510 	int copied;
511 
512 	err = -EAGAIN;
513 	skb = sock_dequeue_err_skb(sk);
514 	if (!skb)
515 		goto out;
516 
517 	copied = skb->len;
518 	if (copied > len) {
519 		msg->msg_flags |= MSG_TRUNC;
520 		copied = len;
521 	}
522 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
523 	if (unlikely(err)) {
524 		kfree_skb(skb);
525 		return err;
526 	}
527 	sock_recv_timestamp(msg, sk, skb);
528 
529 	serr = SKB_EXT_ERR(skb);
530 
531 	if (sin && ipv4_datagram_support_addr(serr)) {
532 		sin->sin_family = AF_INET;
533 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
534 						   serr->addr_offset);
535 		sin->sin_port = serr->port;
536 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
537 		*addr_len = sizeof(*sin);
538 	}
539 
540 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
541 	sin = &errhdr.offender;
542 	memset(sin, 0, sizeof(*sin));
543 
544 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
545 		sin->sin_family = AF_INET;
546 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
547 		if (inet_sk(sk)->cmsg_flags)
548 			ip_cmsg_recv(msg, skb);
549 	}
550 
551 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
552 
553 	/* Now we could try to dump offended packet options */
554 
555 	msg->msg_flags |= MSG_ERRQUEUE;
556 	err = copied;
557 
558 	consume_skb(skb);
559 out:
560 	return err;
561 }
562 
563 
564 /*
565  *	Socket option code for IP. This is the end of the line after any
566  *	TCP,UDP etc options on an IP socket.
567  */
568 static bool setsockopt_needs_rtnl(int optname)
569 {
570 	switch (optname) {
571 	case IP_ADD_MEMBERSHIP:
572 	case IP_ADD_SOURCE_MEMBERSHIP:
573 	case IP_BLOCK_SOURCE:
574 	case IP_DROP_MEMBERSHIP:
575 	case IP_DROP_SOURCE_MEMBERSHIP:
576 	case IP_MSFILTER:
577 	case IP_UNBLOCK_SOURCE:
578 	case MCAST_BLOCK_SOURCE:
579 	case MCAST_MSFILTER:
580 	case MCAST_JOIN_GROUP:
581 	case MCAST_JOIN_SOURCE_GROUP:
582 	case MCAST_LEAVE_GROUP:
583 	case MCAST_LEAVE_SOURCE_GROUP:
584 	case MCAST_UNBLOCK_SOURCE:
585 		return true;
586 	}
587 	return false;
588 }
589 
590 static int do_ip_setsockopt(struct sock *sk, int level,
591 			    int optname, char __user *optval, unsigned int optlen)
592 {
593 	struct inet_sock *inet = inet_sk(sk);
594 	struct net *net = sock_net(sk);
595 	int val = 0, err;
596 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
597 
598 	switch (optname) {
599 	case IP_PKTINFO:
600 	case IP_RECVTTL:
601 	case IP_RECVOPTS:
602 	case IP_RECVTOS:
603 	case IP_RETOPTS:
604 	case IP_TOS:
605 	case IP_TTL:
606 	case IP_HDRINCL:
607 	case IP_MTU_DISCOVER:
608 	case IP_RECVERR:
609 	case IP_ROUTER_ALERT:
610 	case IP_FREEBIND:
611 	case IP_PASSSEC:
612 	case IP_TRANSPARENT:
613 	case IP_MINTTL:
614 	case IP_NODEFRAG:
615 	case IP_BIND_ADDRESS_NO_PORT:
616 	case IP_UNICAST_IF:
617 	case IP_MULTICAST_TTL:
618 	case IP_MULTICAST_ALL:
619 	case IP_MULTICAST_LOOP:
620 	case IP_RECVORIGDSTADDR:
621 	case IP_CHECKSUM:
622 	case IP_RECVFRAGSIZE:
623 		if (optlen >= sizeof(int)) {
624 			if (get_user(val, (int __user *) optval))
625 				return -EFAULT;
626 		} else if (optlen >= sizeof(char)) {
627 			unsigned char ucval;
628 
629 			if (get_user(ucval, (unsigned char __user *) optval))
630 				return -EFAULT;
631 			val = (int) ucval;
632 		}
633 	}
634 
635 	/* If optlen==0, it is equivalent to val == 0 */
636 
637 	if (optname == IP_ROUTER_ALERT)
638 		return ip_ra_control(sk, val ? 1 : 0, NULL);
639 	if (ip_mroute_opt(optname))
640 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
641 
642 	err = 0;
643 	if (needs_rtnl)
644 		rtnl_lock();
645 	lock_sock(sk);
646 
647 	switch (optname) {
648 	case IP_OPTIONS:
649 	{
650 		struct ip_options_rcu *old, *opt = NULL;
651 
652 		if (optlen > 40)
653 			goto e_inval;
654 		err = ip_options_get_from_user(sock_net(sk), &opt,
655 					       optval, optlen);
656 		if (err)
657 			break;
658 		old = rcu_dereference_protected(inet->inet_opt,
659 						lockdep_sock_is_held(sk));
660 		if (inet->is_icsk) {
661 			struct inet_connection_sock *icsk = inet_csk(sk);
662 #if IS_ENABLED(CONFIG_IPV6)
663 			if (sk->sk_family == PF_INET ||
664 			    (!((1 << sk->sk_state) &
665 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
666 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
667 #endif
668 				if (old)
669 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
670 				if (opt)
671 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
672 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
673 #if IS_ENABLED(CONFIG_IPV6)
674 			}
675 #endif
676 		}
677 		rcu_assign_pointer(inet->inet_opt, opt);
678 		if (old)
679 			kfree_rcu(old, rcu);
680 		break;
681 	}
682 	case IP_PKTINFO:
683 		if (val)
684 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
685 		else
686 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
687 		break;
688 	case IP_RECVTTL:
689 		if (val)
690 			inet->cmsg_flags |=  IP_CMSG_TTL;
691 		else
692 			inet->cmsg_flags &= ~IP_CMSG_TTL;
693 		break;
694 	case IP_RECVTOS:
695 		if (val)
696 			inet->cmsg_flags |=  IP_CMSG_TOS;
697 		else
698 			inet->cmsg_flags &= ~IP_CMSG_TOS;
699 		break;
700 	case IP_RECVOPTS:
701 		if (val)
702 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
703 		else
704 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
705 		break;
706 	case IP_RETOPTS:
707 		if (val)
708 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
709 		else
710 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
711 		break;
712 	case IP_PASSSEC:
713 		if (val)
714 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
715 		else
716 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
717 		break;
718 	case IP_RECVORIGDSTADDR:
719 		if (val)
720 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
721 		else
722 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
723 		break;
724 	case IP_CHECKSUM:
725 		if (val) {
726 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
727 				inet_inc_convert_csum(sk);
728 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
729 			}
730 		} else {
731 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
732 				inet_dec_convert_csum(sk);
733 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
734 			}
735 		}
736 		break;
737 	case IP_RECVFRAGSIZE:
738 		if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
739 			goto e_inval;
740 		if (val)
741 			inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
742 		else
743 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
744 		break;
745 	case IP_TOS:	/* This sets both TOS and Precedence */
746 		if (sk->sk_type == SOCK_STREAM) {
747 			val &= ~INET_ECN_MASK;
748 			val |= inet->tos & INET_ECN_MASK;
749 		}
750 		if (inet->tos != val) {
751 			inet->tos = val;
752 			sk->sk_priority = rt_tos2priority(val);
753 			sk_dst_reset(sk);
754 		}
755 		break;
756 	case IP_TTL:
757 		if (optlen < 1)
758 			goto e_inval;
759 		if (val != -1 && (val < 1 || val > 255))
760 			goto e_inval;
761 		inet->uc_ttl = val;
762 		break;
763 	case IP_HDRINCL:
764 		if (sk->sk_type != SOCK_RAW) {
765 			err = -ENOPROTOOPT;
766 			break;
767 		}
768 		inet->hdrincl = val ? 1 : 0;
769 		break;
770 	case IP_NODEFRAG:
771 		if (sk->sk_type != SOCK_RAW) {
772 			err = -ENOPROTOOPT;
773 			break;
774 		}
775 		inet->nodefrag = val ? 1 : 0;
776 		break;
777 	case IP_BIND_ADDRESS_NO_PORT:
778 		inet->bind_address_no_port = val ? 1 : 0;
779 		break;
780 	case IP_MTU_DISCOVER:
781 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
782 			goto e_inval;
783 		inet->pmtudisc = val;
784 		break;
785 	case IP_RECVERR:
786 		inet->recverr = !!val;
787 		if (!val)
788 			skb_queue_purge(&sk->sk_error_queue);
789 		break;
790 	case IP_MULTICAST_TTL:
791 		if (sk->sk_type == SOCK_STREAM)
792 			goto e_inval;
793 		if (optlen < 1)
794 			goto e_inval;
795 		if (val == -1)
796 			val = 1;
797 		if (val < 0 || val > 255)
798 			goto e_inval;
799 		inet->mc_ttl = val;
800 		break;
801 	case IP_MULTICAST_LOOP:
802 		if (optlen < 1)
803 			goto e_inval;
804 		inet->mc_loop = !!val;
805 		break;
806 	case IP_UNICAST_IF:
807 	{
808 		struct net_device *dev = NULL;
809 		int ifindex;
810 		int midx;
811 
812 		if (optlen != sizeof(int))
813 			goto e_inval;
814 
815 		ifindex = (__force int)ntohl((__force __be32)val);
816 		if (ifindex == 0) {
817 			inet->uc_index = 0;
818 			err = 0;
819 			break;
820 		}
821 
822 		dev = dev_get_by_index(sock_net(sk), ifindex);
823 		err = -EADDRNOTAVAIL;
824 		if (!dev)
825 			break;
826 
827 		midx = l3mdev_master_ifindex(dev);
828 		dev_put(dev);
829 
830 		err = -EINVAL;
831 		if (sk->sk_bound_dev_if &&
832 		    (!midx || midx != sk->sk_bound_dev_if))
833 			break;
834 
835 		inet->uc_index = ifindex;
836 		err = 0;
837 		break;
838 	}
839 	case IP_MULTICAST_IF:
840 	{
841 		struct ip_mreqn mreq;
842 		struct net_device *dev = NULL;
843 		int midx;
844 
845 		if (sk->sk_type == SOCK_STREAM)
846 			goto e_inval;
847 		/*
848 		 *	Check the arguments are allowable
849 		 */
850 
851 		if (optlen < sizeof(struct in_addr))
852 			goto e_inval;
853 
854 		err = -EFAULT;
855 		if (optlen >= sizeof(struct ip_mreqn)) {
856 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
857 				break;
858 		} else {
859 			memset(&mreq, 0, sizeof(mreq));
860 			if (optlen >= sizeof(struct ip_mreq)) {
861 				if (copy_from_user(&mreq, optval,
862 						   sizeof(struct ip_mreq)))
863 					break;
864 			} else if (optlen >= sizeof(struct in_addr)) {
865 				if (copy_from_user(&mreq.imr_address, optval,
866 						   sizeof(struct in_addr)))
867 					break;
868 			}
869 		}
870 
871 		if (!mreq.imr_ifindex) {
872 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
873 				inet->mc_index = 0;
874 				inet->mc_addr  = 0;
875 				err = 0;
876 				break;
877 			}
878 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
879 			if (dev)
880 				mreq.imr_ifindex = dev->ifindex;
881 		} else
882 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
883 
884 
885 		err = -EADDRNOTAVAIL;
886 		if (!dev)
887 			break;
888 
889 		midx = l3mdev_master_ifindex(dev);
890 
891 		dev_put(dev);
892 
893 		err = -EINVAL;
894 		if (sk->sk_bound_dev_if &&
895 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
896 		    (!midx || midx != sk->sk_bound_dev_if))
897 			break;
898 
899 		inet->mc_index = mreq.imr_ifindex;
900 		inet->mc_addr  = mreq.imr_address.s_addr;
901 		err = 0;
902 		break;
903 	}
904 
905 	case IP_ADD_MEMBERSHIP:
906 	case IP_DROP_MEMBERSHIP:
907 	{
908 		struct ip_mreqn mreq;
909 
910 		err = -EPROTO;
911 		if (inet_sk(sk)->is_icsk)
912 			break;
913 
914 		if (optlen < sizeof(struct ip_mreq))
915 			goto e_inval;
916 		err = -EFAULT;
917 		if (optlen >= sizeof(struct ip_mreqn)) {
918 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
919 				break;
920 		} else {
921 			memset(&mreq, 0, sizeof(mreq));
922 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
923 				break;
924 		}
925 
926 		if (optname == IP_ADD_MEMBERSHIP)
927 			err = ip_mc_join_group(sk, &mreq);
928 		else
929 			err = ip_mc_leave_group(sk, &mreq);
930 		break;
931 	}
932 	case IP_MSFILTER:
933 	{
934 		struct ip_msfilter *msf;
935 
936 		if (optlen < IP_MSFILTER_SIZE(0))
937 			goto e_inval;
938 		if (optlen > sysctl_optmem_max) {
939 			err = -ENOBUFS;
940 			break;
941 		}
942 		msf = memdup_user(optval, optlen);
943 		if (IS_ERR(msf)) {
944 			err = PTR_ERR(msf);
945 			break;
946 		}
947 		/* numsrc >= (1G-4) overflow in 32 bits */
948 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
949 		    msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
950 			kfree(msf);
951 			err = -ENOBUFS;
952 			break;
953 		}
954 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
955 			kfree(msf);
956 			err = -EINVAL;
957 			break;
958 		}
959 		err = ip_mc_msfilter(sk, msf, 0);
960 		kfree(msf);
961 		break;
962 	}
963 	case IP_BLOCK_SOURCE:
964 	case IP_UNBLOCK_SOURCE:
965 	case IP_ADD_SOURCE_MEMBERSHIP:
966 	case IP_DROP_SOURCE_MEMBERSHIP:
967 	{
968 		struct ip_mreq_source mreqs;
969 		int omode, add;
970 
971 		if (optlen != sizeof(struct ip_mreq_source))
972 			goto e_inval;
973 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
974 			err = -EFAULT;
975 			break;
976 		}
977 		if (optname == IP_BLOCK_SOURCE) {
978 			omode = MCAST_EXCLUDE;
979 			add = 1;
980 		} else if (optname == IP_UNBLOCK_SOURCE) {
981 			omode = MCAST_EXCLUDE;
982 			add = 0;
983 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
984 			struct ip_mreqn mreq;
985 
986 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
987 			mreq.imr_address.s_addr = mreqs.imr_interface;
988 			mreq.imr_ifindex = 0;
989 			err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
990 			if (err && err != -EADDRINUSE)
991 				break;
992 			omode = MCAST_INCLUDE;
993 			add = 1;
994 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
995 			omode = MCAST_INCLUDE;
996 			add = 0;
997 		}
998 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
999 		break;
1000 	}
1001 	case MCAST_JOIN_GROUP:
1002 	case MCAST_LEAVE_GROUP:
1003 	{
1004 		struct group_req greq;
1005 		struct sockaddr_in *psin;
1006 		struct ip_mreqn mreq;
1007 
1008 		if (optlen < sizeof(struct group_req))
1009 			goto e_inval;
1010 		err = -EFAULT;
1011 		if (copy_from_user(&greq, optval, sizeof(greq)))
1012 			break;
1013 		psin = (struct sockaddr_in *)&greq.gr_group;
1014 		if (psin->sin_family != AF_INET)
1015 			goto e_inval;
1016 		memset(&mreq, 0, sizeof(mreq));
1017 		mreq.imr_multiaddr = psin->sin_addr;
1018 		mreq.imr_ifindex = greq.gr_interface;
1019 
1020 		if (optname == MCAST_JOIN_GROUP)
1021 			err = ip_mc_join_group(sk, &mreq);
1022 		else
1023 			err = ip_mc_leave_group(sk, &mreq);
1024 		break;
1025 	}
1026 	case MCAST_JOIN_SOURCE_GROUP:
1027 	case MCAST_LEAVE_SOURCE_GROUP:
1028 	case MCAST_BLOCK_SOURCE:
1029 	case MCAST_UNBLOCK_SOURCE:
1030 	{
1031 		struct group_source_req greqs;
1032 		struct ip_mreq_source mreqs;
1033 		struct sockaddr_in *psin;
1034 		int omode, add;
1035 
1036 		if (optlen != sizeof(struct group_source_req))
1037 			goto e_inval;
1038 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1039 			err = -EFAULT;
1040 			break;
1041 		}
1042 		if (greqs.gsr_group.ss_family != AF_INET ||
1043 		    greqs.gsr_source.ss_family != AF_INET) {
1044 			err = -EADDRNOTAVAIL;
1045 			break;
1046 		}
1047 		psin = (struct sockaddr_in *)&greqs.gsr_group;
1048 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1049 		psin = (struct sockaddr_in *)&greqs.gsr_source;
1050 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1051 		mreqs.imr_interface = 0; /* use index for mc_source */
1052 
1053 		if (optname == MCAST_BLOCK_SOURCE) {
1054 			omode = MCAST_EXCLUDE;
1055 			add = 1;
1056 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
1057 			omode = MCAST_EXCLUDE;
1058 			add = 0;
1059 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1060 			struct ip_mreqn mreq;
1061 
1062 			psin = (struct sockaddr_in *)&greqs.gsr_group;
1063 			mreq.imr_multiaddr = psin->sin_addr;
1064 			mreq.imr_address.s_addr = 0;
1065 			mreq.imr_ifindex = greqs.gsr_interface;
1066 			err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1067 			if (err && err != -EADDRINUSE)
1068 				break;
1069 			greqs.gsr_interface = mreq.imr_ifindex;
1070 			omode = MCAST_INCLUDE;
1071 			add = 1;
1072 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
1073 			omode = MCAST_INCLUDE;
1074 			add = 0;
1075 		}
1076 		err = ip_mc_source(add, omode, sk, &mreqs,
1077 				   greqs.gsr_interface);
1078 		break;
1079 	}
1080 	case MCAST_MSFILTER:
1081 	{
1082 		struct sockaddr_in *psin;
1083 		struct ip_msfilter *msf = NULL;
1084 		struct group_filter *gsf = NULL;
1085 		int msize, i, ifindex;
1086 
1087 		if (optlen < GROUP_FILTER_SIZE(0))
1088 			goto e_inval;
1089 		if (optlen > sysctl_optmem_max) {
1090 			err = -ENOBUFS;
1091 			break;
1092 		}
1093 		gsf = memdup_user(optval, optlen);
1094 		if (IS_ERR(gsf)) {
1095 			err = PTR_ERR(gsf);
1096 			break;
1097 		}
1098 
1099 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1100 		if (gsf->gf_numsrc >= 0x1ffffff ||
1101 		    gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1102 			err = -ENOBUFS;
1103 			goto mc_msf_out;
1104 		}
1105 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1106 			err = -EINVAL;
1107 			goto mc_msf_out;
1108 		}
1109 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1110 		msf = kmalloc(msize, GFP_KERNEL);
1111 		if (!msf) {
1112 			err = -ENOBUFS;
1113 			goto mc_msf_out;
1114 		}
1115 		ifindex = gsf->gf_interface;
1116 		psin = (struct sockaddr_in *)&gsf->gf_group;
1117 		if (psin->sin_family != AF_INET) {
1118 			err = -EADDRNOTAVAIL;
1119 			goto mc_msf_out;
1120 		}
1121 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1122 		msf->imsf_interface = 0;
1123 		msf->imsf_fmode = gsf->gf_fmode;
1124 		msf->imsf_numsrc = gsf->gf_numsrc;
1125 		err = -EADDRNOTAVAIL;
1126 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1127 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1128 
1129 			if (psin->sin_family != AF_INET)
1130 				goto mc_msf_out;
1131 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1132 		}
1133 		kfree(gsf);
1134 		gsf = NULL;
1135 
1136 		err = ip_mc_msfilter(sk, msf, ifindex);
1137 mc_msf_out:
1138 		kfree(msf);
1139 		kfree(gsf);
1140 		break;
1141 	}
1142 	case IP_MULTICAST_ALL:
1143 		if (optlen < 1)
1144 			goto e_inval;
1145 		if (val != 0 && val != 1)
1146 			goto e_inval;
1147 		inet->mc_all = val;
1148 		break;
1149 
1150 	case IP_FREEBIND:
1151 		if (optlen < 1)
1152 			goto e_inval;
1153 		inet->freebind = !!val;
1154 		break;
1155 
1156 	case IP_IPSEC_POLICY:
1157 	case IP_XFRM_POLICY:
1158 		err = -EPERM;
1159 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1160 			break;
1161 		err = xfrm_user_policy(sk, optname, optval, optlen);
1162 		break;
1163 
1164 	case IP_TRANSPARENT:
1165 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1166 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1167 			err = -EPERM;
1168 			break;
1169 		}
1170 		if (optlen < 1)
1171 			goto e_inval;
1172 		inet->transparent = !!val;
1173 		break;
1174 
1175 	case IP_MINTTL:
1176 		if (optlen < 1)
1177 			goto e_inval;
1178 		if (val < 0 || val > 255)
1179 			goto e_inval;
1180 		inet->min_ttl = val;
1181 		break;
1182 
1183 	default:
1184 		err = -ENOPROTOOPT;
1185 		break;
1186 	}
1187 	release_sock(sk);
1188 	if (needs_rtnl)
1189 		rtnl_unlock();
1190 	return err;
1191 
1192 e_inval:
1193 	release_sock(sk);
1194 	if (needs_rtnl)
1195 		rtnl_unlock();
1196 	return -EINVAL;
1197 }
1198 
1199 /**
1200  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1201  * @sk: socket
1202  * @skb: buffer
1203  *
1204  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1205  * destination in skb->cb[] before dst drop.
1206  * This way, receiver doesn't make cache line misses to read rtable.
1207  */
1208 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1209 {
1210 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1211 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1212 		       ipv6_sk_rxinfo(sk);
1213 
1214 	if (prepare && skb_rtable(skb)) {
1215 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1216 		 * which has interface index (iif) as the first member of the
1217 		 * underlying inet{6}_skb_parm struct. This code then overlays
1218 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1219 		 * element so the iif is picked up from the prior IPCB. If iif
1220 		 * is the loopback interface, then return the sending interface
1221 		 * (e.g., process binds socket to eth0 for Tx which is
1222 		 * redirected to loopback in the rtable/dst).
1223 		 */
1224 		struct rtable *rt = skb_rtable(skb);
1225 		bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1226 
1227 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1228 			pktinfo->ipi_ifindex = inet_iif(skb);
1229 		else if (l3slave && rt && rt->rt_iif)
1230 			pktinfo->ipi_ifindex = rt->rt_iif;
1231 
1232 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1233 	} else {
1234 		pktinfo->ipi_ifindex = 0;
1235 		pktinfo->ipi_spec_dst.s_addr = 0;
1236 	}
1237 	skb_dst_drop(skb);
1238 }
1239 
1240 int ip_setsockopt(struct sock *sk, int level,
1241 		int optname, char __user *optval, unsigned int optlen)
1242 {
1243 	int err;
1244 
1245 	if (level != SOL_IP)
1246 		return -ENOPROTOOPT;
1247 
1248 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1249 #ifdef CONFIG_BPFILTER
1250 	if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1251 	    optname < BPFILTER_IPT_SET_MAX)
1252 		err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1253 #endif
1254 #ifdef CONFIG_NETFILTER
1255 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1256 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1257 			optname != IP_IPSEC_POLICY &&
1258 			optname != IP_XFRM_POLICY &&
1259 			!ip_mroute_opt(optname))
1260 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1261 #endif
1262 	return err;
1263 }
1264 EXPORT_SYMBOL(ip_setsockopt);
1265 
1266 #ifdef CONFIG_COMPAT
1267 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1268 			 char __user *optval, unsigned int optlen)
1269 {
1270 	int err;
1271 
1272 	if (level != SOL_IP)
1273 		return -ENOPROTOOPT;
1274 
1275 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1276 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1277 			ip_setsockopt);
1278 
1279 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1280 #ifdef CONFIG_NETFILTER
1281 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1282 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1283 			optname != IP_IPSEC_POLICY &&
1284 			optname != IP_XFRM_POLICY &&
1285 			!ip_mroute_opt(optname))
1286 		err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1287 					   optlen);
1288 #endif
1289 	return err;
1290 }
1291 EXPORT_SYMBOL(compat_ip_setsockopt);
1292 #endif
1293 
1294 /*
1295  *	Get the options. Note for future reference. The GET of IP options gets
1296  *	the _received_ ones. The set sets the _sent_ ones.
1297  */
1298 
1299 static bool getsockopt_needs_rtnl(int optname)
1300 {
1301 	switch (optname) {
1302 	case IP_MSFILTER:
1303 	case MCAST_MSFILTER:
1304 		return true;
1305 	}
1306 	return false;
1307 }
1308 
1309 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1310 			    char __user *optval, int __user *optlen, unsigned int flags)
1311 {
1312 	struct inet_sock *inet = inet_sk(sk);
1313 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1314 	int val, err = 0;
1315 	int len;
1316 
1317 	if (level != SOL_IP)
1318 		return -EOPNOTSUPP;
1319 
1320 	if (ip_mroute_opt(optname))
1321 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1322 
1323 	if (get_user(len, optlen))
1324 		return -EFAULT;
1325 	if (len < 0)
1326 		return -EINVAL;
1327 
1328 	if (needs_rtnl)
1329 		rtnl_lock();
1330 	lock_sock(sk);
1331 
1332 	switch (optname) {
1333 	case IP_OPTIONS:
1334 	{
1335 		unsigned char optbuf[sizeof(struct ip_options)+40];
1336 		struct ip_options *opt = (struct ip_options *)optbuf;
1337 		struct ip_options_rcu *inet_opt;
1338 
1339 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1340 						     lockdep_sock_is_held(sk));
1341 		opt->optlen = 0;
1342 		if (inet_opt)
1343 			memcpy(optbuf, &inet_opt->opt,
1344 			       sizeof(struct ip_options) +
1345 			       inet_opt->opt.optlen);
1346 		release_sock(sk);
1347 
1348 		if (opt->optlen == 0)
1349 			return put_user(0, optlen);
1350 
1351 		ip_options_undo(opt);
1352 
1353 		len = min_t(unsigned int, len, opt->optlen);
1354 		if (put_user(len, optlen))
1355 			return -EFAULT;
1356 		if (copy_to_user(optval, opt->__data, len))
1357 			return -EFAULT;
1358 		return 0;
1359 	}
1360 	case IP_PKTINFO:
1361 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1362 		break;
1363 	case IP_RECVTTL:
1364 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1365 		break;
1366 	case IP_RECVTOS:
1367 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1368 		break;
1369 	case IP_RECVOPTS:
1370 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1371 		break;
1372 	case IP_RETOPTS:
1373 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1374 		break;
1375 	case IP_PASSSEC:
1376 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1377 		break;
1378 	case IP_RECVORIGDSTADDR:
1379 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1380 		break;
1381 	case IP_CHECKSUM:
1382 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1383 		break;
1384 	case IP_RECVFRAGSIZE:
1385 		val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1386 		break;
1387 	case IP_TOS:
1388 		val = inet->tos;
1389 		break;
1390 	case IP_TTL:
1391 	{
1392 		struct net *net = sock_net(sk);
1393 		val = (inet->uc_ttl == -1 ?
1394 		       net->ipv4.sysctl_ip_default_ttl :
1395 		       inet->uc_ttl);
1396 		break;
1397 	}
1398 	case IP_HDRINCL:
1399 		val = inet->hdrincl;
1400 		break;
1401 	case IP_NODEFRAG:
1402 		val = inet->nodefrag;
1403 		break;
1404 	case IP_BIND_ADDRESS_NO_PORT:
1405 		val = inet->bind_address_no_port;
1406 		break;
1407 	case IP_MTU_DISCOVER:
1408 		val = inet->pmtudisc;
1409 		break;
1410 	case IP_MTU:
1411 	{
1412 		struct dst_entry *dst;
1413 		val = 0;
1414 		dst = sk_dst_get(sk);
1415 		if (dst) {
1416 			val = dst_mtu(dst);
1417 			dst_release(dst);
1418 		}
1419 		if (!val) {
1420 			release_sock(sk);
1421 			return -ENOTCONN;
1422 		}
1423 		break;
1424 	}
1425 	case IP_RECVERR:
1426 		val = inet->recverr;
1427 		break;
1428 	case IP_MULTICAST_TTL:
1429 		val = inet->mc_ttl;
1430 		break;
1431 	case IP_MULTICAST_LOOP:
1432 		val = inet->mc_loop;
1433 		break;
1434 	case IP_UNICAST_IF:
1435 		val = (__force int)htonl((__u32) inet->uc_index);
1436 		break;
1437 	case IP_MULTICAST_IF:
1438 	{
1439 		struct in_addr addr;
1440 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1441 		addr.s_addr = inet->mc_addr;
1442 		release_sock(sk);
1443 
1444 		if (put_user(len, optlen))
1445 			return -EFAULT;
1446 		if (copy_to_user(optval, &addr, len))
1447 			return -EFAULT;
1448 		return 0;
1449 	}
1450 	case IP_MSFILTER:
1451 	{
1452 		struct ip_msfilter msf;
1453 
1454 		if (len < IP_MSFILTER_SIZE(0)) {
1455 			err = -EINVAL;
1456 			goto out;
1457 		}
1458 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1459 			err = -EFAULT;
1460 			goto out;
1461 		}
1462 		err = ip_mc_msfget(sk, &msf,
1463 				   (struct ip_msfilter __user *)optval, optlen);
1464 		goto out;
1465 	}
1466 	case MCAST_MSFILTER:
1467 	{
1468 		struct group_filter gsf;
1469 
1470 		if (len < GROUP_FILTER_SIZE(0)) {
1471 			err = -EINVAL;
1472 			goto out;
1473 		}
1474 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1475 			err = -EFAULT;
1476 			goto out;
1477 		}
1478 		err = ip_mc_gsfget(sk, &gsf,
1479 				   (struct group_filter __user *)optval,
1480 				   optlen);
1481 		goto out;
1482 	}
1483 	case IP_MULTICAST_ALL:
1484 		val = inet->mc_all;
1485 		break;
1486 	case IP_PKTOPTIONS:
1487 	{
1488 		struct msghdr msg;
1489 
1490 		release_sock(sk);
1491 
1492 		if (sk->sk_type != SOCK_STREAM)
1493 			return -ENOPROTOOPT;
1494 
1495 		msg.msg_control = (__force void *) optval;
1496 		msg.msg_controllen = len;
1497 		msg.msg_flags = flags;
1498 
1499 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1500 			struct in_pktinfo info;
1501 
1502 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1503 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1504 			info.ipi_ifindex = inet->mc_index;
1505 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1506 		}
1507 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1508 			int hlim = inet->mc_ttl;
1509 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1510 		}
1511 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1512 			int tos = inet->rcv_tos;
1513 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1514 		}
1515 		len -= msg.msg_controllen;
1516 		return put_user(len, optlen);
1517 	}
1518 	case IP_FREEBIND:
1519 		val = inet->freebind;
1520 		break;
1521 	case IP_TRANSPARENT:
1522 		val = inet->transparent;
1523 		break;
1524 	case IP_MINTTL:
1525 		val = inet->min_ttl;
1526 		break;
1527 	default:
1528 		release_sock(sk);
1529 		return -ENOPROTOOPT;
1530 	}
1531 	release_sock(sk);
1532 
1533 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1534 		unsigned char ucval = (unsigned char)val;
1535 		len = 1;
1536 		if (put_user(len, optlen))
1537 			return -EFAULT;
1538 		if (copy_to_user(optval, &ucval, 1))
1539 			return -EFAULT;
1540 	} else {
1541 		len = min_t(unsigned int, sizeof(int), len);
1542 		if (put_user(len, optlen))
1543 			return -EFAULT;
1544 		if (copy_to_user(optval, &val, len))
1545 			return -EFAULT;
1546 	}
1547 	return 0;
1548 
1549 out:
1550 	release_sock(sk);
1551 	if (needs_rtnl)
1552 		rtnl_unlock();
1553 	return err;
1554 }
1555 
1556 int ip_getsockopt(struct sock *sk, int level,
1557 		  int optname, char __user *optval, int __user *optlen)
1558 {
1559 	int err;
1560 
1561 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1562 #ifdef CONFIG_BPFILTER
1563 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1564 	    optname < BPFILTER_IPT_GET_MAX)
1565 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1566 #endif
1567 #ifdef CONFIG_NETFILTER
1568 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1569 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1570 			!ip_mroute_opt(optname)) {
1571 		int len;
1572 
1573 		if (get_user(len, optlen))
1574 			return -EFAULT;
1575 
1576 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1577 		if (err >= 0)
1578 			err = put_user(len, optlen);
1579 		return err;
1580 	}
1581 #endif
1582 	return err;
1583 }
1584 EXPORT_SYMBOL(ip_getsockopt);
1585 
1586 #ifdef CONFIG_COMPAT
1587 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1588 			 char __user *optval, int __user *optlen)
1589 {
1590 	int err;
1591 
1592 	if (optname == MCAST_MSFILTER)
1593 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1594 			ip_getsockopt);
1595 
1596 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1597 		MSG_CMSG_COMPAT);
1598 
1599 #ifdef CONFIG_BPFILTER
1600 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1601 	    optname < BPFILTER_IPT_GET_MAX)
1602 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1603 #endif
1604 #ifdef CONFIG_NETFILTER
1605 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1606 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1607 			!ip_mroute_opt(optname)) {
1608 		int len;
1609 
1610 		if (get_user(len, optlen))
1611 			return -EFAULT;
1612 
1613 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1614 		if (err >= 0)
1615 			err = put_user(len, optlen);
1616 		return err;
1617 	}
1618 #endif
1619 	return err;
1620 }
1621 EXPORT_SYMBOL(compat_ip_getsockopt);
1622 #endif
1623