xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision bc05aa6e)
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 /*
51  *	SOL_IP control messages.
52  */
53 
54 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
55 {
56 	struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
57 
58 	info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
59 
60 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
61 }
62 
63 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
64 {
65 	int ttl = ip_hdr(skb)->ttl;
66 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
67 }
68 
69 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
70 {
71 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
72 }
73 
74 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
75 {
76 	if (IPCB(skb)->opt.optlen == 0)
77 		return;
78 
79 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
80 		 ip_hdr(skb) + 1);
81 }
82 
83 
84 static void ip_cmsg_recv_retopts(struct net *net, struct msghdr *msg,
85 				 struct sk_buff *skb)
86 {
87 	unsigned char optbuf[sizeof(struct ip_options) + 40];
88 	struct ip_options *opt = (struct ip_options *)optbuf;
89 
90 	if (IPCB(skb)->opt.optlen == 0)
91 		return;
92 
93 	if (ip_options_echo(net, opt, skb)) {
94 		msg->msg_flags |= MSG_CTRUNC;
95 		return;
96 	}
97 	ip_options_undo(opt);
98 
99 	put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
100 }
101 
102 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb)
103 {
104 	int val;
105 
106 	if (IPCB(skb)->frag_max_size == 0)
107 		return;
108 
109 	val = IPCB(skb)->frag_max_size;
110 	put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
111 }
112 
113 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
114 				  int tlen, int offset)
115 {
116 	__wsum csum = skb->csum;
117 
118 	if (skb->ip_summed != CHECKSUM_COMPLETE)
119 		return;
120 
121 	if (offset != 0) {
122 		int tend_off = skb_transport_offset(skb) + tlen;
123 		csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
124 	}
125 
126 	put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
127 }
128 
129 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
130 {
131 	char *secdata;
132 	u32 seclen, secid;
133 	int err;
134 
135 	err = security_socket_getpeersec_dgram(NULL, skb, &secid);
136 	if (err)
137 		return;
138 
139 	err = security_secid_to_secctx(secid, &secdata, &seclen);
140 	if (err)
141 		return;
142 
143 	put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
144 	security_release_secctx(secdata, seclen);
145 }
146 
147 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
148 {
149 	struct sockaddr_in sin;
150 	const struct iphdr *iph = ip_hdr(skb);
151 	__be16 *ports = (__be16 *)skb_transport_header(skb);
152 
153 	if (skb_transport_offset(skb) + 4 > (int)skb->len)
154 		return;
155 
156 	/* All current transport protocols have the port numbers in the
157 	 * first four bytes of the transport header and this function is
158 	 * written with this assumption in mind.
159 	 */
160 
161 	sin.sin_family = AF_INET;
162 	sin.sin_addr.s_addr = iph->daddr;
163 	sin.sin_port = ports[1];
164 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
165 
166 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
167 }
168 
169 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
170 			 struct sk_buff *skb, int tlen, int offset)
171 {
172 	struct inet_sock *inet = inet_sk(sk);
173 	unsigned int flags = inet->cmsg_flags;
174 
175 	/* Ordered by supposed usage frequency */
176 	if (flags & IP_CMSG_PKTINFO) {
177 		ip_cmsg_recv_pktinfo(msg, skb);
178 
179 		flags &= ~IP_CMSG_PKTINFO;
180 		if (!flags)
181 			return;
182 	}
183 
184 	if (flags & IP_CMSG_TTL) {
185 		ip_cmsg_recv_ttl(msg, skb);
186 
187 		flags &= ~IP_CMSG_TTL;
188 		if (!flags)
189 			return;
190 	}
191 
192 	if (flags & IP_CMSG_TOS) {
193 		ip_cmsg_recv_tos(msg, skb);
194 
195 		flags &= ~IP_CMSG_TOS;
196 		if (!flags)
197 			return;
198 	}
199 
200 	if (flags & IP_CMSG_RECVOPTS) {
201 		ip_cmsg_recv_opts(msg, skb);
202 
203 		flags &= ~IP_CMSG_RECVOPTS;
204 		if (!flags)
205 			return;
206 	}
207 
208 	if (flags & IP_CMSG_RETOPTS) {
209 		ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
210 
211 		flags &= ~IP_CMSG_RETOPTS;
212 		if (!flags)
213 			return;
214 	}
215 
216 	if (flags & IP_CMSG_PASSSEC) {
217 		ip_cmsg_recv_security(msg, skb);
218 
219 		flags &= ~IP_CMSG_PASSSEC;
220 		if (!flags)
221 			return;
222 	}
223 
224 	if (flags & IP_CMSG_ORIGDSTADDR) {
225 		ip_cmsg_recv_dstaddr(msg, skb);
226 
227 		flags &= ~IP_CMSG_ORIGDSTADDR;
228 		if (!flags)
229 			return;
230 	}
231 
232 	if (flags & IP_CMSG_CHECKSUM)
233 		ip_cmsg_recv_checksum(msg, skb, tlen, offset);
234 
235 	if (flags & IP_CMSG_RECVFRAGSIZE)
236 		ip_cmsg_recv_fragsize(msg, skb);
237 }
238 EXPORT_SYMBOL(ip_cmsg_recv_offset);
239 
240 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
241 		 bool allow_ipv6)
242 {
243 	int err, val;
244 	struct cmsghdr *cmsg;
245 	struct net *net = sock_net(sk);
246 
247 	for_each_cmsghdr(cmsg, msg) {
248 		if (!CMSG_OK(msg, cmsg))
249 			return -EINVAL;
250 #if IS_ENABLED(CONFIG_IPV6)
251 		if (allow_ipv6 &&
252 		    cmsg->cmsg_level == SOL_IPV6 &&
253 		    cmsg->cmsg_type == IPV6_PKTINFO) {
254 			struct in6_pktinfo *src_info;
255 
256 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
257 				return -EINVAL;
258 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
259 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
260 				return -EINVAL;
261 			if (src_info->ipi6_ifindex)
262 				ipc->oif = src_info->ipi6_ifindex;
263 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
264 			continue;
265 		}
266 #endif
267 		if (cmsg->cmsg_level == SOL_SOCKET) {
268 			err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
269 			if (err)
270 				return err;
271 			continue;
272 		}
273 
274 		if (cmsg->cmsg_level != SOL_IP)
275 			continue;
276 		switch (cmsg->cmsg_type) {
277 		case IP_RETOPTS:
278 			err = cmsg->cmsg_len - sizeof(struct cmsghdr);
279 
280 			/* Our caller is responsible for freeing ipc->opt */
281 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
282 					     err < 40 ? err : 40);
283 			if (err)
284 				return err;
285 			break;
286 		case IP_PKTINFO:
287 		{
288 			struct in_pktinfo *info;
289 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
290 				return -EINVAL;
291 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
292 			if (info->ipi_ifindex)
293 				ipc->oif = info->ipi_ifindex;
294 			ipc->addr = info->ipi_spec_dst.s_addr;
295 			break;
296 		}
297 		case IP_TTL:
298 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
299 				return -EINVAL;
300 			val = *(int *)CMSG_DATA(cmsg);
301 			if (val < 1 || val > 255)
302 				return -EINVAL;
303 			ipc->ttl = val;
304 			break;
305 		case IP_TOS:
306 			if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
307 				val = *(int *)CMSG_DATA(cmsg);
308 			else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
309 				val = *(u8 *)CMSG_DATA(cmsg);
310 			else
311 				return -EINVAL;
312 			if (val < 0 || val > 255)
313 				return -EINVAL;
314 			ipc->tos = val;
315 			ipc->priority = rt_tos2priority(ipc->tos);
316 			break;
317 
318 		default:
319 			return -EINVAL;
320 		}
321 	}
322 	return 0;
323 }
324 
325 
326 /* Special input handler for packets caught by router alert option.
327    They are selected only by protocol field, and then processed likely
328    local ones; but only if someone wants them! Otherwise, router
329    not running rsvpd will kill RSVP.
330 
331    It is user level problem, what it will make with them.
332    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
333    but receiver should be enough clever f.e. to forward mtrace requests,
334    sent to multicast group to reach destination designated router.
335  */
336 struct ip_ra_chain __rcu *ip_ra_chain;
337 
338 
339 static void ip_ra_destroy_rcu(struct rcu_head *head)
340 {
341 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
342 
343 	sock_put(ra->saved_sk);
344 	kfree(ra);
345 }
346 
347 int ip_ra_control(struct sock *sk, unsigned char on,
348 		  void (*destructor)(struct sock *))
349 {
350 	struct ip_ra_chain *ra, *new_ra;
351 	struct ip_ra_chain __rcu **rap;
352 
353 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
354 		return -EINVAL;
355 
356 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
357 
358 	for (rap = &ip_ra_chain;
359 	     (ra = rtnl_dereference(*rap)) != NULL;
360 	     rap = &ra->next) {
361 		if (ra->sk == sk) {
362 			if (on) {
363 				kfree(new_ra);
364 				return -EADDRINUSE;
365 			}
366 			/* dont let ip_call_ra_chain() use sk again */
367 			ra->sk = NULL;
368 			RCU_INIT_POINTER(*rap, ra->next);
369 
370 			if (ra->destructor)
371 				ra->destructor(sk);
372 			/*
373 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
374 			 * period. This guarantee ip_call_ra_chain() dont need
375 			 * to mess with socket refcounts.
376 			 */
377 			ra->saved_sk = sk;
378 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
379 			return 0;
380 		}
381 	}
382 	if (!new_ra)
383 		return -ENOBUFS;
384 	new_ra->sk = sk;
385 	new_ra->destructor = destructor;
386 
387 	RCU_INIT_POINTER(new_ra->next, ra);
388 	rcu_assign_pointer(*rap, new_ra);
389 	sock_hold(sk);
390 
391 	return 0;
392 }
393 
394 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
395 		   __be16 port, u32 info, u8 *payload)
396 {
397 	struct sock_exterr_skb *serr;
398 
399 	skb = skb_clone(skb, GFP_ATOMIC);
400 	if (!skb)
401 		return;
402 
403 	serr = SKB_EXT_ERR(skb);
404 	serr->ee.ee_errno = err;
405 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
406 	serr->ee.ee_type = icmp_hdr(skb)->type;
407 	serr->ee.ee_code = icmp_hdr(skb)->code;
408 	serr->ee.ee_pad = 0;
409 	serr->ee.ee_info = info;
410 	serr->ee.ee_data = 0;
411 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
412 				   skb_network_header(skb);
413 	serr->port = port;
414 
415 	if (skb_pull(skb, payload - skb->data)) {
416 		skb_reset_transport_header(skb);
417 		if (sock_queue_err_skb(sk, skb) == 0)
418 			return;
419 	}
420 	kfree_skb(skb);
421 }
422 
423 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
424 {
425 	struct inet_sock *inet = inet_sk(sk);
426 	struct sock_exterr_skb *serr;
427 	struct iphdr *iph;
428 	struct sk_buff *skb;
429 
430 	if (!inet->recverr)
431 		return;
432 
433 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
434 	if (!skb)
435 		return;
436 
437 	skb_put(skb, sizeof(struct iphdr));
438 	skb_reset_network_header(skb);
439 	iph = ip_hdr(skb);
440 	iph->daddr = daddr;
441 
442 	serr = SKB_EXT_ERR(skb);
443 	serr->ee.ee_errno = err;
444 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
445 	serr->ee.ee_type = 0;
446 	serr->ee.ee_code = 0;
447 	serr->ee.ee_pad = 0;
448 	serr->ee.ee_info = info;
449 	serr->ee.ee_data = 0;
450 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
451 	serr->port = port;
452 
453 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
454 	skb_reset_transport_header(skb);
455 
456 	if (sock_queue_err_skb(sk, skb))
457 		kfree_skb(skb);
458 }
459 
460 /* For some errors we have valid addr_offset even with zero payload and
461  * zero port. Also, addr_offset should be supported if port is set.
462  */
463 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
464 {
465 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
466 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
467 }
468 
469 /* IPv4 supports cmsg on all imcp errors and some timestamps
470  *
471  * Timestamp code paths do not initialize the fields expected by cmsg:
472  * the PKTINFO fields in skb->cb[]. Fill those in here.
473  */
474 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
475 				       struct sk_buff *skb,
476 				       int ee_origin)
477 {
478 	struct in_pktinfo *info;
479 
480 	if (ee_origin == SO_EE_ORIGIN_ICMP)
481 		return true;
482 
483 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
484 		return false;
485 
486 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
487 	 * timestamp with egress dev. Not possible for packets without iif
488 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
489 	 */
490 	info = PKTINFO_SKB_CB(skb);
491 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
492 	    !info->ipi_ifindex)
493 		return false;
494 
495 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
496 	return true;
497 }
498 
499 /*
500  *	Handle MSG_ERRQUEUE
501  */
502 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
503 {
504 	struct sock_exterr_skb *serr;
505 	struct sk_buff *skb;
506 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
507 	struct {
508 		struct sock_extended_err ee;
509 		struct sockaddr_in	 offender;
510 	} errhdr;
511 	int err;
512 	int copied;
513 
514 	WARN_ON_ONCE(sk->sk_family == AF_INET6);
515 
516 	err = -EAGAIN;
517 	skb = sock_dequeue_err_skb(sk);
518 	if (!skb)
519 		goto out;
520 
521 	copied = skb->len;
522 	if (copied > len) {
523 		msg->msg_flags |= MSG_TRUNC;
524 		copied = len;
525 	}
526 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
527 	if (unlikely(err)) {
528 		kfree_skb(skb);
529 		return err;
530 	}
531 	sock_recv_timestamp(msg, sk, skb);
532 
533 	serr = SKB_EXT_ERR(skb);
534 
535 	if (sin && ipv4_datagram_support_addr(serr)) {
536 		sin->sin_family = AF_INET;
537 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
538 						   serr->addr_offset);
539 		sin->sin_port = serr->port;
540 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
541 		*addr_len = sizeof(*sin);
542 	}
543 
544 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
545 	sin = &errhdr.offender;
546 	memset(sin, 0, sizeof(*sin));
547 
548 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
549 		sin->sin_family = AF_INET;
550 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
551 		if (inet_sk(sk)->cmsg_flags)
552 			ip_cmsg_recv(msg, skb);
553 	}
554 
555 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
556 
557 	/* Now we could try to dump offended packet options */
558 
559 	msg->msg_flags |= MSG_ERRQUEUE;
560 	err = copied;
561 
562 	consume_skb(skb);
563 out:
564 	return err;
565 }
566 
567 
568 /*
569  *	Socket option code for IP. This is the end of the line after any
570  *	TCP,UDP etc options on an IP socket.
571  */
572 static bool setsockopt_needs_rtnl(int optname)
573 {
574 	switch (optname) {
575 	case IP_ADD_MEMBERSHIP:
576 	case IP_ADD_SOURCE_MEMBERSHIP:
577 	case IP_BLOCK_SOURCE:
578 	case IP_DROP_MEMBERSHIP:
579 	case IP_DROP_SOURCE_MEMBERSHIP:
580 	case IP_MSFILTER:
581 	case IP_UNBLOCK_SOURCE:
582 	case MCAST_BLOCK_SOURCE:
583 	case MCAST_MSFILTER:
584 	case MCAST_JOIN_GROUP:
585 	case MCAST_JOIN_SOURCE_GROUP:
586 	case MCAST_LEAVE_GROUP:
587 	case MCAST_LEAVE_SOURCE_GROUP:
588 	case MCAST_UNBLOCK_SOURCE:
589 	case IP_ROUTER_ALERT:
590 		return true;
591 	}
592 	return false;
593 }
594 
595 static int do_ip_setsockopt(struct sock *sk, int level,
596 			    int optname, char __user *optval, unsigned int optlen)
597 {
598 	struct inet_sock *inet = inet_sk(sk);
599 	struct net *net = sock_net(sk);
600 	int val = 0, err;
601 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
602 
603 	switch (optname) {
604 	case IP_PKTINFO:
605 	case IP_RECVTTL:
606 	case IP_RECVOPTS:
607 	case IP_RECVTOS:
608 	case IP_RETOPTS:
609 	case IP_TOS:
610 	case IP_TTL:
611 	case IP_HDRINCL:
612 	case IP_MTU_DISCOVER:
613 	case IP_RECVERR:
614 	case IP_ROUTER_ALERT:
615 	case IP_FREEBIND:
616 	case IP_PASSSEC:
617 	case IP_TRANSPARENT:
618 	case IP_MINTTL:
619 	case IP_NODEFRAG:
620 	case IP_BIND_ADDRESS_NO_PORT:
621 	case IP_UNICAST_IF:
622 	case IP_MULTICAST_TTL:
623 	case IP_MULTICAST_ALL:
624 	case IP_MULTICAST_LOOP:
625 	case IP_RECVORIGDSTADDR:
626 	case IP_CHECKSUM:
627 	case IP_RECVFRAGSIZE:
628 		if (optlen >= sizeof(int)) {
629 			if (get_user(val, (int __user *) optval))
630 				return -EFAULT;
631 		} else if (optlen >= sizeof(char)) {
632 			unsigned char ucval;
633 
634 			if (get_user(ucval, (unsigned char __user *) optval))
635 				return -EFAULT;
636 			val = (int) ucval;
637 		}
638 	}
639 
640 	/* If optlen==0, it is equivalent to val == 0 */
641 
642 	if (ip_mroute_opt(optname))
643 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
644 
645 	err = 0;
646 	if (needs_rtnl)
647 		rtnl_lock();
648 	lock_sock(sk);
649 
650 	switch (optname) {
651 	case IP_OPTIONS:
652 	{
653 		struct ip_options_rcu *old, *opt = NULL;
654 
655 		if (optlen > 40)
656 			goto e_inval;
657 		err = ip_options_get_from_user(sock_net(sk), &opt,
658 					       optval, optlen);
659 		if (err)
660 			break;
661 		old = rcu_dereference_protected(inet->inet_opt,
662 						lockdep_sock_is_held(sk));
663 		if (inet->is_icsk) {
664 			struct inet_connection_sock *icsk = inet_csk(sk);
665 #if IS_ENABLED(CONFIG_IPV6)
666 			if (sk->sk_family == PF_INET ||
667 			    (!((1 << sk->sk_state) &
668 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
669 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
670 #endif
671 				if (old)
672 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
673 				if (opt)
674 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
675 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
676 #if IS_ENABLED(CONFIG_IPV6)
677 			}
678 #endif
679 		}
680 		rcu_assign_pointer(inet->inet_opt, opt);
681 		if (old)
682 			kfree_rcu(old, rcu);
683 		break;
684 	}
685 	case IP_PKTINFO:
686 		if (val)
687 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
688 		else
689 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
690 		break;
691 	case IP_RECVTTL:
692 		if (val)
693 			inet->cmsg_flags |=  IP_CMSG_TTL;
694 		else
695 			inet->cmsg_flags &= ~IP_CMSG_TTL;
696 		break;
697 	case IP_RECVTOS:
698 		if (val)
699 			inet->cmsg_flags |=  IP_CMSG_TOS;
700 		else
701 			inet->cmsg_flags &= ~IP_CMSG_TOS;
702 		break;
703 	case IP_RECVOPTS:
704 		if (val)
705 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
706 		else
707 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
708 		break;
709 	case IP_RETOPTS:
710 		if (val)
711 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
712 		else
713 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
714 		break;
715 	case IP_PASSSEC:
716 		if (val)
717 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
718 		else
719 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
720 		break;
721 	case IP_RECVORIGDSTADDR:
722 		if (val)
723 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
724 		else
725 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
726 		break;
727 	case IP_CHECKSUM:
728 		if (val) {
729 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
730 				inet_inc_convert_csum(sk);
731 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
732 			}
733 		} else {
734 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
735 				inet_dec_convert_csum(sk);
736 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
737 			}
738 		}
739 		break;
740 	case IP_RECVFRAGSIZE:
741 		if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
742 			goto e_inval;
743 		if (val)
744 			inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
745 		else
746 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
747 		break;
748 	case IP_TOS:	/* This sets both TOS and Precedence */
749 		if (sk->sk_type == SOCK_STREAM) {
750 			val &= ~INET_ECN_MASK;
751 			val |= inet->tos & INET_ECN_MASK;
752 		}
753 		if (inet->tos != val) {
754 			inet->tos = val;
755 			sk->sk_priority = rt_tos2priority(val);
756 			sk_dst_reset(sk);
757 		}
758 		break;
759 	case IP_TTL:
760 		if (optlen < 1)
761 			goto e_inval;
762 		if (val != -1 && (val < 1 || val > 255))
763 			goto e_inval;
764 		inet->uc_ttl = val;
765 		break;
766 	case IP_HDRINCL:
767 		if (sk->sk_type != SOCK_RAW) {
768 			err = -ENOPROTOOPT;
769 			break;
770 		}
771 		inet->hdrincl = val ? 1 : 0;
772 		break;
773 	case IP_NODEFRAG:
774 		if (sk->sk_type != SOCK_RAW) {
775 			err = -ENOPROTOOPT;
776 			break;
777 		}
778 		inet->nodefrag = val ? 1 : 0;
779 		break;
780 	case IP_BIND_ADDRESS_NO_PORT:
781 		inet->bind_address_no_port = val ? 1 : 0;
782 		break;
783 	case IP_MTU_DISCOVER:
784 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
785 			goto e_inval;
786 		inet->pmtudisc = val;
787 		break;
788 	case IP_RECVERR:
789 		inet->recverr = !!val;
790 		if (!val)
791 			skb_queue_purge(&sk->sk_error_queue);
792 		break;
793 	case IP_MULTICAST_TTL:
794 		if (sk->sk_type == SOCK_STREAM)
795 			goto e_inval;
796 		if (optlen < 1)
797 			goto e_inval;
798 		if (val == -1)
799 			val = 1;
800 		if (val < 0 || val > 255)
801 			goto e_inval;
802 		inet->mc_ttl = val;
803 		break;
804 	case IP_MULTICAST_LOOP:
805 		if (optlen < 1)
806 			goto e_inval;
807 		inet->mc_loop = !!val;
808 		break;
809 	case IP_UNICAST_IF:
810 	{
811 		struct net_device *dev = NULL;
812 		int ifindex;
813 		int midx;
814 
815 		if (optlen != sizeof(int))
816 			goto e_inval;
817 
818 		ifindex = (__force int)ntohl((__force __be32)val);
819 		if (ifindex == 0) {
820 			inet->uc_index = 0;
821 			err = 0;
822 			break;
823 		}
824 
825 		dev = dev_get_by_index(sock_net(sk), ifindex);
826 		err = -EADDRNOTAVAIL;
827 		if (!dev)
828 			break;
829 
830 		midx = l3mdev_master_ifindex(dev);
831 		dev_put(dev);
832 
833 		err = -EINVAL;
834 		if (sk->sk_bound_dev_if &&
835 		    (!midx || midx != sk->sk_bound_dev_if))
836 			break;
837 
838 		inet->uc_index = ifindex;
839 		err = 0;
840 		break;
841 	}
842 	case IP_MULTICAST_IF:
843 	{
844 		struct ip_mreqn mreq;
845 		struct net_device *dev = NULL;
846 		int midx;
847 
848 		if (sk->sk_type == SOCK_STREAM)
849 			goto e_inval;
850 		/*
851 		 *	Check the arguments are allowable
852 		 */
853 
854 		if (optlen < sizeof(struct in_addr))
855 			goto e_inval;
856 
857 		err = -EFAULT;
858 		if (optlen >= sizeof(struct ip_mreqn)) {
859 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
860 				break;
861 		} else {
862 			memset(&mreq, 0, sizeof(mreq));
863 			if (optlen >= sizeof(struct ip_mreq)) {
864 				if (copy_from_user(&mreq, optval,
865 						   sizeof(struct ip_mreq)))
866 					break;
867 			} else if (optlen >= sizeof(struct in_addr)) {
868 				if (copy_from_user(&mreq.imr_address, optval,
869 						   sizeof(struct in_addr)))
870 					break;
871 			}
872 		}
873 
874 		if (!mreq.imr_ifindex) {
875 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
876 				inet->mc_index = 0;
877 				inet->mc_addr  = 0;
878 				err = 0;
879 				break;
880 			}
881 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
882 			if (dev)
883 				mreq.imr_ifindex = dev->ifindex;
884 		} else
885 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
886 
887 
888 		err = -EADDRNOTAVAIL;
889 		if (!dev)
890 			break;
891 
892 		midx = l3mdev_master_ifindex(dev);
893 
894 		dev_put(dev);
895 
896 		err = -EINVAL;
897 		if (sk->sk_bound_dev_if &&
898 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
899 		    (!midx || midx != sk->sk_bound_dev_if))
900 			break;
901 
902 		inet->mc_index = mreq.imr_ifindex;
903 		inet->mc_addr  = mreq.imr_address.s_addr;
904 		err = 0;
905 		break;
906 	}
907 
908 	case IP_ADD_MEMBERSHIP:
909 	case IP_DROP_MEMBERSHIP:
910 	{
911 		struct ip_mreqn mreq;
912 
913 		err = -EPROTO;
914 		if (inet_sk(sk)->is_icsk)
915 			break;
916 
917 		if (optlen < sizeof(struct ip_mreq))
918 			goto e_inval;
919 		err = -EFAULT;
920 		if (optlen >= sizeof(struct ip_mreqn)) {
921 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
922 				break;
923 		} else {
924 			memset(&mreq, 0, sizeof(mreq));
925 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
926 				break;
927 		}
928 
929 		if (optname == IP_ADD_MEMBERSHIP)
930 			err = ip_mc_join_group(sk, &mreq);
931 		else
932 			err = ip_mc_leave_group(sk, &mreq);
933 		break;
934 	}
935 	case IP_MSFILTER:
936 	{
937 		struct ip_msfilter *msf;
938 
939 		if (optlen < IP_MSFILTER_SIZE(0))
940 			goto e_inval;
941 		if (optlen > sysctl_optmem_max) {
942 			err = -ENOBUFS;
943 			break;
944 		}
945 		msf = memdup_user(optval, optlen);
946 		if (IS_ERR(msf)) {
947 			err = PTR_ERR(msf);
948 			break;
949 		}
950 		/* numsrc >= (1G-4) overflow in 32 bits */
951 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
952 		    msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
953 			kfree(msf);
954 			err = -ENOBUFS;
955 			break;
956 		}
957 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
958 			kfree(msf);
959 			err = -EINVAL;
960 			break;
961 		}
962 		err = ip_mc_msfilter(sk, msf, 0);
963 		kfree(msf);
964 		break;
965 	}
966 	case IP_BLOCK_SOURCE:
967 	case IP_UNBLOCK_SOURCE:
968 	case IP_ADD_SOURCE_MEMBERSHIP:
969 	case IP_DROP_SOURCE_MEMBERSHIP:
970 	{
971 		struct ip_mreq_source mreqs;
972 		int omode, add;
973 
974 		if (optlen != sizeof(struct ip_mreq_source))
975 			goto e_inval;
976 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
977 			err = -EFAULT;
978 			break;
979 		}
980 		if (optname == IP_BLOCK_SOURCE) {
981 			omode = MCAST_EXCLUDE;
982 			add = 1;
983 		} else if (optname == IP_UNBLOCK_SOURCE) {
984 			omode = MCAST_EXCLUDE;
985 			add = 0;
986 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
987 			struct ip_mreqn mreq;
988 
989 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
990 			mreq.imr_address.s_addr = mreqs.imr_interface;
991 			mreq.imr_ifindex = 0;
992 			err = ip_mc_join_group(sk, &mreq);
993 			if (err && err != -EADDRINUSE)
994 				break;
995 			omode = MCAST_INCLUDE;
996 			add = 1;
997 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
998 			omode = MCAST_INCLUDE;
999 			add = 0;
1000 		}
1001 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
1002 		break;
1003 	}
1004 	case MCAST_JOIN_GROUP:
1005 	case MCAST_LEAVE_GROUP:
1006 	{
1007 		struct group_req greq;
1008 		struct sockaddr_in *psin;
1009 		struct ip_mreqn mreq;
1010 
1011 		if (optlen < sizeof(struct group_req))
1012 			goto e_inval;
1013 		err = -EFAULT;
1014 		if (copy_from_user(&greq, optval, sizeof(greq)))
1015 			break;
1016 		psin = (struct sockaddr_in *)&greq.gr_group;
1017 		if (psin->sin_family != AF_INET)
1018 			goto e_inval;
1019 		memset(&mreq, 0, sizeof(mreq));
1020 		mreq.imr_multiaddr = psin->sin_addr;
1021 		mreq.imr_ifindex = greq.gr_interface;
1022 
1023 		if (optname == MCAST_JOIN_GROUP)
1024 			err = ip_mc_join_group(sk, &mreq);
1025 		else
1026 			err = ip_mc_leave_group(sk, &mreq);
1027 		break;
1028 	}
1029 	case MCAST_JOIN_SOURCE_GROUP:
1030 	case MCAST_LEAVE_SOURCE_GROUP:
1031 	case MCAST_BLOCK_SOURCE:
1032 	case MCAST_UNBLOCK_SOURCE:
1033 	{
1034 		struct group_source_req greqs;
1035 		struct ip_mreq_source mreqs;
1036 		struct sockaddr_in *psin;
1037 		int omode, add;
1038 
1039 		if (optlen != sizeof(struct group_source_req))
1040 			goto e_inval;
1041 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1042 			err = -EFAULT;
1043 			break;
1044 		}
1045 		if (greqs.gsr_group.ss_family != AF_INET ||
1046 		    greqs.gsr_source.ss_family != AF_INET) {
1047 			err = -EADDRNOTAVAIL;
1048 			break;
1049 		}
1050 		psin = (struct sockaddr_in *)&greqs.gsr_group;
1051 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1052 		psin = (struct sockaddr_in *)&greqs.gsr_source;
1053 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1054 		mreqs.imr_interface = 0; /* use index for mc_source */
1055 
1056 		if (optname == MCAST_BLOCK_SOURCE) {
1057 			omode = MCAST_EXCLUDE;
1058 			add = 1;
1059 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
1060 			omode = MCAST_EXCLUDE;
1061 			add = 0;
1062 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1063 			struct ip_mreqn mreq;
1064 
1065 			psin = (struct sockaddr_in *)&greqs.gsr_group;
1066 			mreq.imr_multiaddr = psin->sin_addr;
1067 			mreq.imr_address.s_addr = 0;
1068 			mreq.imr_ifindex = greqs.gsr_interface;
1069 			err = ip_mc_join_group(sk, &mreq);
1070 			if (err && err != -EADDRINUSE)
1071 				break;
1072 			greqs.gsr_interface = mreq.imr_ifindex;
1073 			omode = MCAST_INCLUDE;
1074 			add = 1;
1075 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
1076 			omode = MCAST_INCLUDE;
1077 			add = 0;
1078 		}
1079 		err = ip_mc_source(add, omode, sk, &mreqs,
1080 				   greqs.gsr_interface);
1081 		break;
1082 	}
1083 	case MCAST_MSFILTER:
1084 	{
1085 		struct sockaddr_in *psin;
1086 		struct ip_msfilter *msf = NULL;
1087 		struct group_filter *gsf = NULL;
1088 		int msize, i, ifindex;
1089 
1090 		if (optlen < GROUP_FILTER_SIZE(0))
1091 			goto e_inval;
1092 		if (optlen > sysctl_optmem_max) {
1093 			err = -ENOBUFS;
1094 			break;
1095 		}
1096 		gsf = memdup_user(optval, optlen);
1097 		if (IS_ERR(gsf)) {
1098 			err = PTR_ERR(gsf);
1099 			break;
1100 		}
1101 
1102 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1103 		if (gsf->gf_numsrc >= 0x1ffffff ||
1104 		    gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1105 			err = -ENOBUFS;
1106 			goto mc_msf_out;
1107 		}
1108 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1109 			err = -EINVAL;
1110 			goto mc_msf_out;
1111 		}
1112 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1113 		msf = kmalloc(msize, GFP_KERNEL);
1114 		if (!msf) {
1115 			err = -ENOBUFS;
1116 			goto mc_msf_out;
1117 		}
1118 		ifindex = gsf->gf_interface;
1119 		psin = (struct sockaddr_in *)&gsf->gf_group;
1120 		if (psin->sin_family != AF_INET) {
1121 			err = -EADDRNOTAVAIL;
1122 			goto mc_msf_out;
1123 		}
1124 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1125 		msf->imsf_interface = 0;
1126 		msf->imsf_fmode = gsf->gf_fmode;
1127 		msf->imsf_numsrc = gsf->gf_numsrc;
1128 		err = -EADDRNOTAVAIL;
1129 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1130 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1131 
1132 			if (psin->sin_family != AF_INET)
1133 				goto mc_msf_out;
1134 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1135 		}
1136 		kfree(gsf);
1137 		gsf = NULL;
1138 
1139 		err = ip_mc_msfilter(sk, msf, ifindex);
1140 mc_msf_out:
1141 		kfree(msf);
1142 		kfree(gsf);
1143 		break;
1144 	}
1145 	case IP_MULTICAST_ALL:
1146 		if (optlen < 1)
1147 			goto e_inval;
1148 		if (val != 0 && val != 1)
1149 			goto e_inval;
1150 		inet->mc_all = val;
1151 		break;
1152 	case IP_ROUTER_ALERT:
1153 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1154 		break;
1155 
1156 	case IP_FREEBIND:
1157 		if (optlen < 1)
1158 			goto e_inval;
1159 		inet->freebind = !!val;
1160 		break;
1161 
1162 	case IP_IPSEC_POLICY:
1163 	case IP_XFRM_POLICY:
1164 		err = -EPERM;
1165 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1166 			break;
1167 		err = xfrm_user_policy(sk, optname, optval, optlen);
1168 		break;
1169 
1170 	case IP_TRANSPARENT:
1171 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1172 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1173 			err = -EPERM;
1174 			break;
1175 		}
1176 		if (optlen < 1)
1177 			goto e_inval;
1178 		inet->transparent = !!val;
1179 		break;
1180 
1181 	case IP_MINTTL:
1182 		if (optlen < 1)
1183 			goto e_inval;
1184 		if (val < 0 || val > 255)
1185 			goto e_inval;
1186 		inet->min_ttl = val;
1187 		break;
1188 
1189 	default:
1190 		err = -ENOPROTOOPT;
1191 		break;
1192 	}
1193 	release_sock(sk);
1194 	if (needs_rtnl)
1195 		rtnl_unlock();
1196 	return err;
1197 
1198 e_inval:
1199 	release_sock(sk);
1200 	if (needs_rtnl)
1201 		rtnl_unlock();
1202 	return -EINVAL;
1203 }
1204 
1205 /**
1206  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1207  * @sk: socket
1208  * @skb: buffer
1209  *
1210  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1211  * destination in skb->cb[] before dst drop.
1212  * This way, receiver doesn't make cache line misses to read rtable.
1213  */
1214 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1215 {
1216 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1217 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1218 		       ipv6_sk_rxinfo(sk);
1219 
1220 	if (prepare && skb_rtable(skb)) {
1221 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1222 		 * which has interface index (iif) as the first member of the
1223 		 * underlying inet{6}_skb_parm struct. This code then overlays
1224 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1225 		 * element so the iif is picked up from the prior IPCB. If iif
1226 		 * is the loopback interface, then return the sending interface
1227 		 * (e.g., process binds socket to eth0 for Tx which is
1228 		 * redirected to loopback in the rtable/dst).
1229 		 */
1230 		struct rtable *rt = skb_rtable(skb);
1231 		bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1232 
1233 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1234 			pktinfo->ipi_ifindex = inet_iif(skb);
1235 		else if (l3slave && rt && rt->rt_iif)
1236 			pktinfo->ipi_ifindex = rt->rt_iif;
1237 
1238 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1239 	} else {
1240 		pktinfo->ipi_ifindex = 0;
1241 		pktinfo->ipi_spec_dst.s_addr = 0;
1242 	}
1243 	skb_dst_drop(skb);
1244 }
1245 
1246 int ip_setsockopt(struct sock *sk, int level,
1247 		int optname, char __user *optval, unsigned int optlen)
1248 {
1249 	int err;
1250 
1251 	if (level != SOL_IP)
1252 		return -ENOPROTOOPT;
1253 
1254 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1255 #ifdef CONFIG_NETFILTER
1256 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1257 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1258 			optname != IP_IPSEC_POLICY &&
1259 			optname != IP_XFRM_POLICY &&
1260 			!ip_mroute_opt(optname))
1261 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1262 #endif
1263 	return err;
1264 }
1265 EXPORT_SYMBOL(ip_setsockopt);
1266 
1267 #ifdef CONFIG_COMPAT
1268 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1269 			 char __user *optval, unsigned int optlen)
1270 {
1271 	int err;
1272 
1273 	if (level != SOL_IP)
1274 		return -ENOPROTOOPT;
1275 
1276 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1277 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1278 			ip_setsockopt);
1279 
1280 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1281 #ifdef CONFIG_NETFILTER
1282 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1283 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1284 			optname != IP_IPSEC_POLICY &&
1285 			optname != IP_XFRM_POLICY &&
1286 			!ip_mroute_opt(optname))
1287 		err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1288 					   optlen);
1289 #endif
1290 	return err;
1291 }
1292 EXPORT_SYMBOL(compat_ip_setsockopt);
1293 #endif
1294 
1295 /*
1296  *	Get the options. Note for future reference. The GET of IP options gets
1297  *	the _received_ ones. The set sets the _sent_ ones.
1298  */
1299 
1300 static bool getsockopt_needs_rtnl(int optname)
1301 {
1302 	switch (optname) {
1303 	case IP_MSFILTER:
1304 	case MCAST_MSFILTER:
1305 		return true;
1306 	}
1307 	return false;
1308 }
1309 
1310 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1311 			    char __user *optval, int __user *optlen, unsigned int flags)
1312 {
1313 	struct inet_sock *inet = inet_sk(sk);
1314 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1315 	int val, err = 0;
1316 	int len;
1317 
1318 	if (level != SOL_IP)
1319 		return -EOPNOTSUPP;
1320 
1321 	if (ip_mroute_opt(optname))
1322 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1323 
1324 	if (get_user(len, optlen))
1325 		return -EFAULT;
1326 	if (len < 0)
1327 		return -EINVAL;
1328 
1329 	if (needs_rtnl)
1330 		rtnl_lock();
1331 	lock_sock(sk);
1332 
1333 	switch (optname) {
1334 	case IP_OPTIONS:
1335 	{
1336 		unsigned char optbuf[sizeof(struct ip_options)+40];
1337 		struct ip_options *opt = (struct ip_options *)optbuf;
1338 		struct ip_options_rcu *inet_opt;
1339 
1340 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1341 						     lockdep_sock_is_held(sk));
1342 		opt->optlen = 0;
1343 		if (inet_opt)
1344 			memcpy(optbuf, &inet_opt->opt,
1345 			       sizeof(struct ip_options) +
1346 			       inet_opt->opt.optlen);
1347 		release_sock(sk);
1348 
1349 		if (opt->optlen == 0)
1350 			return put_user(0, optlen);
1351 
1352 		ip_options_undo(opt);
1353 
1354 		len = min_t(unsigned int, len, opt->optlen);
1355 		if (put_user(len, optlen))
1356 			return -EFAULT;
1357 		if (copy_to_user(optval, opt->__data, len))
1358 			return -EFAULT;
1359 		return 0;
1360 	}
1361 	case IP_PKTINFO:
1362 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1363 		break;
1364 	case IP_RECVTTL:
1365 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1366 		break;
1367 	case IP_RECVTOS:
1368 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1369 		break;
1370 	case IP_RECVOPTS:
1371 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1372 		break;
1373 	case IP_RETOPTS:
1374 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1375 		break;
1376 	case IP_PASSSEC:
1377 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1378 		break;
1379 	case IP_RECVORIGDSTADDR:
1380 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1381 		break;
1382 	case IP_CHECKSUM:
1383 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1384 		break;
1385 	case IP_RECVFRAGSIZE:
1386 		val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1387 		break;
1388 	case IP_TOS:
1389 		val = inet->tos;
1390 		break;
1391 	case IP_TTL:
1392 	{
1393 		struct net *net = sock_net(sk);
1394 		val = (inet->uc_ttl == -1 ?
1395 		       net->ipv4.sysctl_ip_default_ttl :
1396 		       inet->uc_ttl);
1397 		break;
1398 	}
1399 	case IP_HDRINCL:
1400 		val = inet->hdrincl;
1401 		break;
1402 	case IP_NODEFRAG:
1403 		val = inet->nodefrag;
1404 		break;
1405 	case IP_BIND_ADDRESS_NO_PORT:
1406 		val = inet->bind_address_no_port;
1407 		break;
1408 	case IP_MTU_DISCOVER:
1409 		val = inet->pmtudisc;
1410 		break;
1411 	case IP_MTU:
1412 	{
1413 		struct dst_entry *dst;
1414 		val = 0;
1415 		dst = sk_dst_get(sk);
1416 		if (dst) {
1417 			val = dst_mtu(dst);
1418 			dst_release(dst);
1419 		}
1420 		if (!val) {
1421 			release_sock(sk);
1422 			return -ENOTCONN;
1423 		}
1424 		break;
1425 	}
1426 	case IP_RECVERR:
1427 		val = inet->recverr;
1428 		break;
1429 	case IP_MULTICAST_TTL:
1430 		val = inet->mc_ttl;
1431 		break;
1432 	case IP_MULTICAST_LOOP:
1433 		val = inet->mc_loop;
1434 		break;
1435 	case IP_UNICAST_IF:
1436 		val = (__force int)htonl((__u32) inet->uc_index);
1437 		break;
1438 	case IP_MULTICAST_IF:
1439 	{
1440 		struct in_addr addr;
1441 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1442 		addr.s_addr = inet->mc_addr;
1443 		release_sock(sk);
1444 
1445 		if (put_user(len, optlen))
1446 			return -EFAULT;
1447 		if (copy_to_user(optval, &addr, len))
1448 			return -EFAULT;
1449 		return 0;
1450 	}
1451 	case IP_MSFILTER:
1452 	{
1453 		struct ip_msfilter msf;
1454 
1455 		if (len < IP_MSFILTER_SIZE(0)) {
1456 			err = -EINVAL;
1457 			goto out;
1458 		}
1459 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1460 			err = -EFAULT;
1461 			goto out;
1462 		}
1463 		err = ip_mc_msfget(sk, &msf,
1464 				   (struct ip_msfilter __user *)optval, optlen);
1465 		goto out;
1466 	}
1467 	case MCAST_MSFILTER:
1468 	{
1469 		struct group_filter gsf;
1470 
1471 		if (len < GROUP_FILTER_SIZE(0)) {
1472 			err = -EINVAL;
1473 			goto out;
1474 		}
1475 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1476 			err = -EFAULT;
1477 			goto out;
1478 		}
1479 		err = ip_mc_gsfget(sk, &gsf,
1480 				   (struct group_filter __user *)optval,
1481 				   optlen);
1482 		goto out;
1483 	}
1484 	case IP_MULTICAST_ALL:
1485 		val = inet->mc_all;
1486 		break;
1487 	case IP_PKTOPTIONS:
1488 	{
1489 		struct msghdr msg;
1490 
1491 		release_sock(sk);
1492 
1493 		if (sk->sk_type != SOCK_STREAM)
1494 			return -ENOPROTOOPT;
1495 
1496 		msg.msg_control = (__force void *) optval;
1497 		msg.msg_controllen = len;
1498 		msg.msg_flags = flags;
1499 
1500 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1501 			struct in_pktinfo info;
1502 
1503 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1504 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1505 			info.ipi_ifindex = inet->mc_index;
1506 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1507 		}
1508 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1509 			int hlim = inet->mc_ttl;
1510 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1511 		}
1512 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1513 			int tos = inet->rcv_tos;
1514 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1515 		}
1516 		len -= msg.msg_controllen;
1517 		return put_user(len, optlen);
1518 	}
1519 	case IP_FREEBIND:
1520 		val = inet->freebind;
1521 		break;
1522 	case IP_TRANSPARENT:
1523 		val = inet->transparent;
1524 		break;
1525 	case IP_MINTTL:
1526 		val = inet->min_ttl;
1527 		break;
1528 	default:
1529 		release_sock(sk);
1530 		return -ENOPROTOOPT;
1531 	}
1532 	release_sock(sk);
1533 
1534 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1535 		unsigned char ucval = (unsigned char)val;
1536 		len = 1;
1537 		if (put_user(len, optlen))
1538 			return -EFAULT;
1539 		if (copy_to_user(optval, &ucval, 1))
1540 			return -EFAULT;
1541 	} else {
1542 		len = min_t(unsigned int, sizeof(int), len);
1543 		if (put_user(len, optlen))
1544 			return -EFAULT;
1545 		if (copy_to_user(optval, &val, len))
1546 			return -EFAULT;
1547 	}
1548 	return 0;
1549 
1550 out:
1551 	release_sock(sk);
1552 	if (needs_rtnl)
1553 		rtnl_unlock();
1554 	return err;
1555 }
1556 
1557 int ip_getsockopt(struct sock *sk, int level,
1558 		  int optname, char __user *optval, int __user *optlen)
1559 {
1560 	int err;
1561 
1562 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1563 #ifdef CONFIG_NETFILTER
1564 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1565 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1566 			!ip_mroute_opt(optname)) {
1567 		int len;
1568 
1569 		if (get_user(len, optlen))
1570 			return -EFAULT;
1571 
1572 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1573 		if (err >= 0)
1574 			err = put_user(len, optlen);
1575 		return err;
1576 	}
1577 #endif
1578 	return err;
1579 }
1580 EXPORT_SYMBOL(ip_getsockopt);
1581 
1582 #ifdef CONFIG_COMPAT
1583 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1584 			 char __user *optval, int __user *optlen)
1585 {
1586 	int err;
1587 
1588 	if (optname == MCAST_MSFILTER)
1589 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1590 			ip_getsockopt);
1591 
1592 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1593 		MSG_CMSG_COMPAT);
1594 
1595 #ifdef CONFIG_NETFILTER
1596 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1597 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1598 			!ip_mroute_opt(optname)) {
1599 		int len;
1600 
1601 		if (get_user(len, optlen))
1602 			return -EFAULT;
1603 
1604 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1605 		if (err >= 0)
1606 			err = put_user(len, optlen);
1607 		return err;
1608 	}
1609 #endif
1610 	return err;
1611 }
1612 EXPORT_SYMBOL(compat_ip_getsockopt);
1613 #endif
1614