xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 5ef12cb4a3a78ffb331c03a795a15eea4ae35155)
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 static void ip_ra_destroy_rcu(struct rcu_head *head)
326 {
327 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
328 
329 	sock_put(ra->saved_sk);
330 	kfree(ra);
331 }
332 
333 int ip_ra_control(struct sock *sk, unsigned char on,
334 		  void (*destructor)(struct sock *))
335 {
336 	struct ip_ra_chain *ra, *new_ra;
337 	struct ip_ra_chain __rcu **rap;
338 	struct net *net = sock_net(sk);
339 
340 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
341 		return -EINVAL;
342 
343 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
344 
345 	mutex_lock(&net->ipv4.ra_mutex);
346 	for (rap = &net->ipv4.ra_chain;
347 	     (ra = rcu_dereference_protected(*rap,
348 			lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
349 	     rap = &ra->next) {
350 		if (ra->sk == sk) {
351 			if (on) {
352 				mutex_unlock(&net->ipv4.ra_mutex);
353 				kfree(new_ra);
354 				return -EADDRINUSE;
355 			}
356 			/* dont let ip_call_ra_chain() use sk again */
357 			ra->sk = NULL;
358 			RCU_INIT_POINTER(*rap, ra->next);
359 			mutex_unlock(&net->ipv4.ra_mutex);
360 
361 			if (ra->destructor)
362 				ra->destructor(sk);
363 			/*
364 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
365 			 * period. This guarantee ip_call_ra_chain() dont need
366 			 * to mess with socket refcounts.
367 			 */
368 			ra->saved_sk = sk;
369 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
370 			return 0;
371 		}
372 	}
373 	if (!new_ra) {
374 		mutex_unlock(&net->ipv4.ra_mutex);
375 		return -ENOBUFS;
376 	}
377 	new_ra->sk = sk;
378 	new_ra->destructor = destructor;
379 
380 	RCU_INIT_POINTER(new_ra->next, ra);
381 	rcu_assign_pointer(*rap, new_ra);
382 	sock_hold(sk);
383 	mutex_unlock(&net->ipv4.ra_mutex);
384 
385 	return 0;
386 }
387 
388 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
389 		   __be16 port, u32 info, u8 *payload)
390 {
391 	struct sock_exterr_skb *serr;
392 
393 	skb = skb_clone(skb, GFP_ATOMIC);
394 	if (!skb)
395 		return;
396 
397 	serr = SKB_EXT_ERR(skb);
398 	serr->ee.ee_errno = err;
399 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
400 	serr->ee.ee_type = icmp_hdr(skb)->type;
401 	serr->ee.ee_code = icmp_hdr(skb)->code;
402 	serr->ee.ee_pad = 0;
403 	serr->ee.ee_info = info;
404 	serr->ee.ee_data = 0;
405 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
406 				   skb_network_header(skb);
407 	serr->port = port;
408 
409 	if (skb_pull(skb, payload - skb->data)) {
410 		skb_reset_transport_header(skb);
411 		if (sock_queue_err_skb(sk, skb) == 0)
412 			return;
413 	}
414 	kfree_skb(skb);
415 }
416 
417 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
418 {
419 	struct inet_sock *inet = inet_sk(sk);
420 	struct sock_exterr_skb *serr;
421 	struct iphdr *iph;
422 	struct sk_buff *skb;
423 
424 	if (!inet->recverr)
425 		return;
426 
427 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
428 	if (!skb)
429 		return;
430 
431 	skb_put(skb, sizeof(struct iphdr));
432 	skb_reset_network_header(skb);
433 	iph = ip_hdr(skb);
434 	iph->daddr = daddr;
435 
436 	serr = SKB_EXT_ERR(skb);
437 	serr->ee.ee_errno = err;
438 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
439 	serr->ee.ee_type = 0;
440 	serr->ee.ee_code = 0;
441 	serr->ee.ee_pad = 0;
442 	serr->ee.ee_info = info;
443 	serr->ee.ee_data = 0;
444 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
445 	serr->port = port;
446 
447 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
448 	skb_reset_transport_header(skb);
449 
450 	if (sock_queue_err_skb(sk, skb))
451 		kfree_skb(skb);
452 }
453 
454 /* For some errors we have valid addr_offset even with zero payload and
455  * zero port. Also, addr_offset should be supported if port is set.
456  */
457 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
458 {
459 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
460 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
461 }
462 
463 /* IPv4 supports cmsg on all imcp errors and some timestamps
464  *
465  * Timestamp code paths do not initialize the fields expected by cmsg:
466  * the PKTINFO fields in skb->cb[]. Fill those in here.
467  */
468 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
469 				       struct sk_buff *skb,
470 				       int ee_origin)
471 {
472 	struct in_pktinfo *info;
473 
474 	if (ee_origin == SO_EE_ORIGIN_ICMP)
475 		return true;
476 
477 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
478 		return false;
479 
480 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
481 	 * timestamp with egress dev. Not possible for packets without iif
482 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
483 	 */
484 	info = PKTINFO_SKB_CB(skb);
485 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
486 	    !info->ipi_ifindex)
487 		return false;
488 
489 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
490 	return true;
491 }
492 
493 /*
494  *	Handle MSG_ERRQUEUE
495  */
496 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
497 {
498 	struct sock_exterr_skb *serr;
499 	struct sk_buff *skb;
500 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
501 	struct {
502 		struct sock_extended_err ee;
503 		struct sockaddr_in	 offender;
504 	} errhdr;
505 	int err;
506 	int copied;
507 
508 	WARN_ON_ONCE(sk->sk_family == AF_INET6);
509 
510 	err = -EAGAIN;
511 	skb = sock_dequeue_err_skb(sk);
512 	if (!skb)
513 		goto out;
514 
515 	copied = skb->len;
516 	if (copied > len) {
517 		msg->msg_flags |= MSG_TRUNC;
518 		copied = len;
519 	}
520 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
521 	if (unlikely(err)) {
522 		kfree_skb(skb);
523 		return err;
524 	}
525 	sock_recv_timestamp(msg, sk, skb);
526 
527 	serr = SKB_EXT_ERR(skb);
528 
529 	if (sin && ipv4_datagram_support_addr(serr)) {
530 		sin->sin_family = AF_INET;
531 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
532 						   serr->addr_offset);
533 		sin->sin_port = serr->port;
534 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
535 		*addr_len = sizeof(*sin);
536 	}
537 
538 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
539 	sin = &errhdr.offender;
540 	memset(sin, 0, sizeof(*sin));
541 
542 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
543 		sin->sin_family = AF_INET;
544 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
545 		if (inet_sk(sk)->cmsg_flags)
546 			ip_cmsg_recv(msg, skb);
547 	}
548 
549 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
550 
551 	/* Now we could try to dump offended packet options */
552 
553 	msg->msg_flags |= MSG_ERRQUEUE;
554 	err = copied;
555 
556 	consume_skb(skb);
557 out:
558 	return err;
559 }
560 
561 
562 /*
563  *	Socket option code for IP. This is the end of the line after any
564  *	TCP,UDP etc options on an IP socket.
565  */
566 static bool setsockopt_needs_rtnl(int optname)
567 {
568 	switch (optname) {
569 	case IP_ADD_MEMBERSHIP:
570 	case IP_ADD_SOURCE_MEMBERSHIP:
571 	case IP_BLOCK_SOURCE:
572 	case IP_DROP_MEMBERSHIP:
573 	case IP_DROP_SOURCE_MEMBERSHIP:
574 	case IP_MSFILTER:
575 	case IP_UNBLOCK_SOURCE:
576 	case MCAST_BLOCK_SOURCE:
577 	case MCAST_MSFILTER:
578 	case MCAST_JOIN_GROUP:
579 	case MCAST_JOIN_SOURCE_GROUP:
580 	case MCAST_LEAVE_GROUP:
581 	case MCAST_LEAVE_SOURCE_GROUP:
582 	case MCAST_UNBLOCK_SOURCE:
583 		return true;
584 	}
585 	return false;
586 }
587 
588 static int do_ip_setsockopt(struct sock *sk, int level,
589 			    int optname, char __user *optval, unsigned int optlen)
590 {
591 	struct inet_sock *inet = inet_sk(sk);
592 	struct net *net = sock_net(sk);
593 	int val = 0, err;
594 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
595 
596 	switch (optname) {
597 	case IP_PKTINFO:
598 	case IP_RECVTTL:
599 	case IP_RECVOPTS:
600 	case IP_RECVTOS:
601 	case IP_RETOPTS:
602 	case IP_TOS:
603 	case IP_TTL:
604 	case IP_HDRINCL:
605 	case IP_MTU_DISCOVER:
606 	case IP_RECVERR:
607 	case IP_ROUTER_ALERT:
608 	case IP_FREEBIND:
609 	case IP_PASSSEC:
610 	case IP_TRANSPARENT:
611 	case IP_MINTTL:
612 	case IP_NODEFRAG:
613 	case IP_BIND_ADDRESS_NO_PORT:
614 	case IP_UNICAST_IF:
615 	case IP_MULTICAST_TTL:
616 	case IP_MULTICAST_ALL:
617 	case IP_MULTICAST_LOOP:
618 	case IP_RECVORIGDSTADDR:
619 	case IP_CHECKSUM:
620 	case IP_RECVFRAGSIZE:
621 		if (optlen >= sizeof(int)) {
622 			if (get_user(val, (int __user *) optval))
623 				return -EFAULT;
624 		} else if (optlen >= sizeof(char)) {
625 			unsigned char ucval;
626 
627 			if (get_user(ucval, (unsigned char __user *) optval))
628 				return -EFAULT;
629 			val = (int) ucval;
630 		}
631 	}
632 
633 	/* If optlen==0, it is equivalent to val == 0 */
634 
635 	if (optname == IP_ROUTER_ALERT)
636 		return ip_ra_control(sk, val ? 1 : 0, NULL);
637 	if (ip_mroute_opt(optname))
638 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
639 
640 	err = 0;
641 	if (needs_rtnl)
642 		rtnl_lock();
643 	lock_sock(sk);
644 
645 	switch (optname) {
646 	case IP_OPTIONS:
647 	{
648 		struct ip_options_rcu *old, *opt = NULL;
649 
650 		if (optlen > 40)
651 			goto e_inval;
652 		err = ip_options_get_from_user(sock_net(sk), &opt,
653 					       optval, optlen);
654 		if (err)
655 			break;
656 		old = rcu_dereference_protected(inet->inet_opt,
657 						lockdep_sock_is_held(sk));
658 		if (inet->is_icsk) {
659 			struct inet_connection_sock *icsk = inet_csk(sk);
660 #if IS_ENABLED(CONFIG_IPV6)
661 			if (sk->sk_family == PF_INET ||
662 			    (!((1 << sk->sk_state) &
663 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
664 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
665 #endif
666 				if (old)
667 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
668 				if (opt)
669 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
670 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
671 #if IS_ENABLED(CONFIG_IPV6)
672 			}
673 #endif
674 		}
675 		rcu_assign_pointer(inet->inet_opt, opt);
676 		if (old)
677 			kfree_rcu(old, rcu);
678 		break;
679 	}
680 	case IP_PKTINFO:
681 		if (val)
682 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
683 		else
684 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
685 		break;
686 	case IP_RECVTTL:
687 		if (val)
688 			inet->cmsg_flags |=  IP_CMSG_TTL;
689 		else
690 			inet->cmsg_flags &= ~IP_CMSG_TTL;
691 		break;
692 	case IP_RECVTOS:
693 		if (val)
694 			inet->cmsg_flags |=  IP_CMSG_TOS;
695 		else
696 			inet->cmsg_flags &= ~IP_CMSG_TOS;
697 		break;
698 	case IP_RECVOPTS:
699 		if (val)
700 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
701 		else
702 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
703 		break;
704 	case IP_RETOPTS:
705 		if (val)
706 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
707 		else
708 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
709 		break;
710 	case IP_PASSSEC:
711 		if (val)
712 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
713 		else
714 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
715 		break;
716 	case IP_RECVORIGDSTADDR:
717 		if (val)
718 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
719 		else
720 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
721 		break;
722 	case IP_CHECKSUM:
723 		if (val) {
724 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
725 				inet_inc_convert_csum(sk);
726 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
727 			}
728 		} else {
729 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
730 				inet_dec_convert_csum(sk);
731 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
732 			}
733 		}
734 		break;
735 	case IP_RECVFRAGSIZE:
736 		if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
737 			goto e_inval;
738 		if (val)
739 			inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
740 		else
741 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
742 		break;
743 	case IP_TOS:	/* This sets both TOS and Precedence */
744 		if (sk->sk_type == SOCK_STREAM) {
745 			val &= ~INET_ECN_MASK;
746 			val |= inet->tos & INET_ECN_MASK;
747 		}
748 		if (inet->tos != val) {
749 			inet->tos = val;
750 			sk->sk_priority = rt_tos2priority(val);
751 			sk_dst_reset(sk);
752 		}
753 		break;
754 	case IP_TTL:
755 		if (optlen < 1)
756 			goto e_inval;
757 		if (val != -1 && (val < 1 || val > 255))
758 			goto e_inval;
759 		inet->uc_ttl = val;
760 		break;
761 	case IP_HDRINCL:
762 		if (sk->sk_type != SOCK_RAW) {
763 			err = -ENOPROTOOPT;
764 			break;
765 		}
766 		inet->hdrincl = val ? 1 : 0;
767 		break;
768 	case IP_NODEFRAG:
769 		if (sk->sk_type != SOCK_RAW) {
770 			err = -ENOPROTOOPT;
771 			break;
772 		}
773 		inet->nodefrag = val ? 1 : 0;
774 		break;
775 	case IP_BIND_ADDRESS_NO_PORT:
776 		inet->bind_address_no_port = val ? 1 : 0;
777 		break;
778 	case IP_MTU_DISCOVER:
779 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
780 			goto e_inval;
781 		inet->pmtudisc = val;
782 		break;
783 	case IP_RECVERR:
784 		inet->recverr = !!val;
785 		if (!val)
786 			skb_queue_purge(&sk->sk_error_queue);
787 		break;
788 	case IP_MULTICAST_TTL:
789 		if (sk->sk_type == SOCK_STREAM)
790 			goto e_inval;
791 		if (optlen < 1)
792 			goto e_inval;
793 		if (val == -1)
794 			val = 1;
795 		if (val < 0 || val > 255)
796 			goto e_inval;
797 		inet->mc_ttl = val;
798 		break;
799 	case IP_MULTICAST_LOOP:
800 		if (optlen < 1)
801 			goto e_inval;
802 		inet->mc_loop = !!val;
803 		break;
804 	case IP_UNICAST_IF:
805 	{
806 		struct net_device *dev = NULL;
807 		int ifindex;
808 		int midx;
809 
810 		if (optlen != sizeof(int))
811 			goto e_inval;
812 
813 		ifindex = (__force int)ntohl((__force __be32)val);
814 		if (ifindex == 0) {
815 			inet->uc_index = 0;
816 			err = 0;
817 			break;
818 		}
819 
820 		dev = dev_get_by_index(sock_net(sk), ifindex);
821 		err = -EADDRNOTAVAIL;
822 		if (!dev)
823 			break;
824 
825 		midx = l3mdev_master_ifindex(dev);
826 		dev_put(dev);
827 
828 		err = -EINVAL;
829 		if (sk->sk_bound_dev_if &&
830 		    (!midx || midx != sk->sk_bound_dev_if))
831 			break;
832 
833 		inet->uc_index = ifindex;
834 		err = 0;
835 		break;
836 	}
837 	case IP_MULTICAST_IF:
838 	{
839 		struct ip_mreqn mreq;
840 		struct net_device *dev = NULL;
841 		int midx;
842 
843 		if (sk->sk_type == SOCK_STREAM)
844 			goto e_inval;
845 		/*
846 		 *	Check the arguments are allowable
847 		 */
848 
849 		if (optlen < sizeof(struct in_addr))
850 			goto e_inval;
851 
852 		err = -EFAULT;
853 		if (optlen >= sizeof(struct ip_mreqn)) {
854 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
855 				break;
856 		} else {
857 			memset(&mreq, 0, sizeof(mreq));
858 			if (optlen >= sizeof(struct ip_mreq)) {
859 				if (copy_from_user(&mreq, optval,
860 						   sizeof(struct ip_mreq)))
861 					break;
862 			} else if (optlen >= sizeof(struct in_addr)) {
863 				if (copy_from_user(&mreq.imr_address, optval,
864 						   sizeof(struct in_addr)))
865 					break;
866 			}
867 		}
868 
869 		if (!mreq.imr_ifindex) {
870 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
871 				inet->mc_index = 0;
872 				inet->mc_addr  = 0;
873 				err = 0;
874 				break;
875 			}
876 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
877 			if (dev)
878 				mreq.imr_ifindex = dev->ifindex;
879 		} else
880 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
881 
882 
883 		err = -EADDRNOTAVAIL;
884 		if (!dev)
885 			break;
886 
887 		midx = l3mdev_master_ifindex(dev);
888 
889 		dev_put(dev);
890 
891 		err = -EINVAL;
892 		if (sk->sk_bound_dev_if &&
893 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
894 		    (!midx || midx != sk->sk_bound_dev_if))
895 			break;
896 
897 		inet->mc_index = mreq.imr_ifindex;
898 		inet->mc_addr  = mreq.imr_address.s_addr;
899 		err = 0;
900 		break;
901 	}
902 
903 	case IP_ADD_MEMBERSHIP:
904 	case IP_DROP_MEMBERSHIP:
905 	{
906 		struct ip_mreqn mreq;
907 
908 		err = -EPROTO;
909 		if (inet_sk(sk)->is_icsk)
910 			break;
911 
912 		if (optlen < sizeof(struct ip_mreq))
913 			goto e_inval;
914 		err = -EFAULT;
915 		if (optlen >= sizeof(struct ip_mreqn)) {
916 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
917 				break;
918 		} else {
919 			memset(&mreq, 0, sizeof(mreq));
920 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
921 				break;
922 		}
923 
924 		if (optname == IP_ADD_MEMBERSHIP)
925 			err = ip_mc_join_group(sk, &mreq);
926 		else
927 			err = ip_mc_leave_group(sk, &mreq);
928 		break;
929 	}
930 	case IP_MSFILTER:
931 	{
932 		struct ip_msfilter *msf;
933 
934 		if (optlen < IP_MSFILTER_SIZE(0))
935 			goto e_inval;
936 		if (optlen > sysctl_optmem_max) {
937 			err = -ENOBUFS;
938 			break;
939 		}
940 		msf = memdup_user(optval, optlen);
941 		if (IS_ERR(msf)) {
942 			err = PTR_ERR(msf);
943 			break;
944 		}
945 		/* numsrc >= (1G-4) overflow in 32 bits */
946 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
947 		    msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
948 			kfree(msf);
949 			err = -ENOBUFS;
950 			break;
951 		}
952 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
953 			kfree(msf);
954 			err = -EINVAL;
955 			break;
956 		}
957 		err = ip_mc_msfilter(sk, msf, 0);
958 		kfree(msf);
959 		break;
960 	}
961 	case IP_BLOCK_SOURCE:
962 	case IP_UNBLOCK_SOURCE:
963 	case IP_ADD_SOURCE_MEMBERSHIP:
964 	case IP_DROP_SOURCE_MEMBERSHIP:
965 	{
966 		struct ip_mreq_source mreqs;
967 		int omode, add;
968 
969 		if (optlen != sizeof(struct ip_mreq_source))
970 			goto e_inval;
971 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
972 			err = -EFAULT;
973 			break;
974 		}
975 		if (optname == IP_BLOCK_SOURCE) {
976 			omode = MCAST_EXCLUDE;
977 			add = 1;
978 		} else if (optname == IP_UNBLOCK_SOURCE) {
979 			omode = MCAST_EXCLUDE;
980 			add = 0;
981 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
982 			struct ip_mreqn mreq;
983 
984 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
985 			mreq.imr_address.s_addr = mreqs.imr_interface;
986 			mreq.imr_ifindex = 0;
987 			err = ip_mc_join_group(sk, &mreq);
988 			if (err && err != -EADDRINUSE)
989 				break;
990 			omode = MCAST_INCLUDE;
991 			add = 1;
992 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
993 			omode = MCAST_INCLUDE;
994 			add = 0;
995 		}
996 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
997 		break;
998 	}
999 	case MCAST_JOIN_GROUP:
1000 	case MCAST_LEAVE_GROUP:
1001 	{
1002 		struct group_req greq;
1003 		struct sockaddr_in *psin;
1004 		struct ip_mreqn mreq;
1005 
1006 		if (optlen < sizeof(struct group_req))
1007 			goto e_inval;
1008 		err = -EFAULT;
1009 		if (copy_from_user(&greq, optval, sizeof(greq)))
1010 			break;
1011 		psin = (struct sockaddr_in *)&greq.gr_group;
1012 		if (psin->sin_family != AF_INET)
1013 			goto e_inval;
1014 		memset(&mreq, 0, sizeof(mreq));
1015 		mreq.imr_multiaddr = psin->sin_addr;
1016 		mreq.imr_ifindex = greq.gr_interface;
1017 
1018 		if (optname == MCAST_JOIN_GROUP)
1019 			err = ip_mc_join_group(sk, &mreq);
1020 		else
1021 			err = ip_mc_leave_group(sk, &mreq);
1022 		break;
1023 	}
1024 	case MCAST_JOIN_SOURCE_GROUP:
1025 	case MCAST_LEAVE_SOURCE_GROUP:
1026 	case MCAST_BLOCK_SOURCE:
1027 	case MCAST_UNBLOCK_SOURCE:
1028 	{
1029 		struct group_source_req greqs;
1030 		struct ip_mreq_source mreqs;
1031 		struct sockaddr_in *psin;
1032 		int omode, add;
1033 
1034 		if (optlen != sizeof(struct group_source_req))
1035 			goto e_inval;
1036 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1037 			err = -EFAULT;
1038 			break;
1039 		}
1040 		if (greqs.gsr_group.ss_family != AF_INET ||
1041 		    greqs.gsr_source.ss_family != AF_INET) {
1042 			err = -EADDRNOTAVAIL;
1043 			break;
1044 		}
1045 		psin = (struct sockaddr_in *)&greqs.gsr_group;
1046 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1047 		psin = (struct sockaddr_in *)&greqs.gsr_source;
1048 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1049 		mreqs.imr_interface = 0; /* use index for mc_source */
1050 
1051 		if (optname == MCAST_BLOCK_SOURCE) {
1052 			omode = MCAST_EXCLUDE;
1053 			add = 1;
1054 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
1055 			omode = MCAST_EXCLUDE;
1056 			add = 0;
1057 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1058 			struct ip_mreqn mreq;
1059 
1060 			psin = (struct sockaddr_in *)&greqs.gsr_group;
1061 			mreq.imr_multiaddr = psin->sin_addr;
1062 			mreq.imr_address.s_addr = 0;
1063 			mreq.imr_ifindex = greqs.gsr_interface;
1064 			err = ip_mc_join_group(sk, &mreq);
1065 			if (err && err != -EADDRINUSE)
1066 				break;
1067 			greqs.gsr_interface = mreq.imr_ifindex;
1068 			omode = MCAST_INCLUDE;
1069 			add = 1;
1070 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
1071 			omode = MCAST_INCLUDE;
1072 			add = 0;
1073 		}
1074 		err = ip_mc_source(add, omode, sk, &mreqs,
1075 				   greqs.gsr_interface);
1076 		break;
1077 	}
1078 	case MCAST_MSFILTER:
1079 	{
1080 		struct sockaddr_in *psin;
1081 		struct ip_msfilter *msf = NULL;
1082 		struct group_filter *gsf = NULL;
1083 		int msize, i, ifindex;
1084 
1085 		if (optlen < GROUP_FILTER_SIZE(0))
1086 			goto e_inval;
1087 		if (optlen > sysctl_optmem_max) {
1088 			err = -ENOBUFS;
1089 			break;
1090 		}
1091 		gsf = memdup_user(optval, optlen);
1092 		if (IS_ERR(gsf)) {
1093 			err = PTR_ERR(gsf);
1094 			break;
1095 		}
1096 
1097 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1098 		if (gsf->gf_numsrc >= 0x1ffffff ||
1099 		    gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1100 			err = -ENOBUFS;
1101 			goto mc_msf_out;
1102 		}
1103 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1104 			err = -EINVAL;
1105 			goto mc_msf_out;
1106 		}
1107 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1108 		msf = kmalloc(msize, GFP_KERNEL);
1109 		if (!msf) {
1110 			err = -ENOBUFS;
1111 			goto mc_msf_out;
1112 		}
1113 		ifindex = gsf->gf_interface;
1114 		psin = (struct sockaddr_in *)&gsf->gf_group;
1115 		if (psin->sin_family != AF_INET) {
1116 			err = -EADDRNOTAVAIL;
1117 			goto mc_msf_out;
1118 		}
1119 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1120 		msf->imsf_interface = 0;
1121 		msf->imsf_fmode = gsf->gf_fmode;
1122 		msf->imsf_numsrc = gsf->gf_numsrc;
1123 		err = -EADDRNOTAVAIL;
1124 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1125 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1126 
1127 			if (psin->sin_family != AF_INET)
1128 				goto mc_msf_out;
1129 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1130 		}
1131 		kfree(gsf);
1132 		gsf = NULL;
1133 
1134 		err = ip_mc_msfilter(sk, msf, ifindex);
1135 mc_msf_out:
1136 		kfree(msf);
1137 		kfree(gsf);
1138 		break;
1139 	}
1140 	case IP_MULTICAST_ALL:
1141 		if (optlen < 1)
1142 			goto e_inval;
1143 		if (val != 0 && val != 1)
1144 			goto e_inval;
1145 		inet->mc_all = val;
1146 		break;
1147 
1148 	case IP_FREEBIND:
1149 		if (optlen < 1)
1150 			goto e_inval;
1151 		inet->freebind = !!val;
1152 		break;
1153 
1154 	case IP_IPSEC_POLICY:
1155 	case IP_XFRM_POLICY:
1156 		err = -EPERM;
1157 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1158 			break;
1159 		err = xfrm_user_policy(sk, optname, optval, optlen);
1160 		break;
1161 
1162 	case IP_TRANSPARENT:
1163 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1164 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1165 			err = -EPERM;
1166 			break;
1167 		}
1168 		if (optlen < 1)
1169 			goto e_inval;
1170 		inet->transparent = !!val;
1171 		break;
1172 
1173 	case IP_MINTTL:
1174 		if (optlen < 1)
1175 			goto e_inval;
1176 		if (val < 0 || val > 255)
1177 			goto e_inval;
1178 		inet->min_ttl = val;
1179 		break;
1180 
1181 	default:
1182 		err = -ENOPROTOOPT;
1183 		break;
1184 	}
1185 	release_sock(sk);
1186 	if (needs_rtnl)
1187 		rtnl_unlock();
1188 	return err;
1189 
1190 e_inval:
1191 	release_sock(sk);
1192 	if (needs_rtnl)
1193 		rtnl_unlock();
1194 	return -EINVAL;
1195 }
1196 
1197 /**
1198  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1199  * @sk: socket
1200  * @skb: buffer
1201  *
1202  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1203  * destination in skb->cb[] before dst drop.
1204  * This way, receiver doesn't make cache line misses to read rtable.
1205  */
1206 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1207 {
1208 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1209 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1210 		       ipv6_sk_rxinfo(sk);
1211 
1212 	if (prepare && skb_rtable(skb)) {
1213 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1214 		 * which has interface index (iif) as the first member of the
1215 		 * underlying inet{6}_skb_parm struct. This code then overlays
1216 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1217 		 * element so the iif is picked up from the prior IPCB. If iif
1218 		 * is the loopback interface, then return the sending interface
1219 		 * (e.g., process binds socket to eth0 for Tx which is
1220 		 * redirected to loopback in the rtable/dst).
1221 		 */
1222 		struct rtable *rt = skb_rtable(skb);
1223 		bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1224 
1225 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1226 			pktinfo->ipi_ifindex = inet_iif(skb);
1227 		else if (l3slave && rt && rt->rt_iif)
1228 			pktinfo->ipi_ifindex = rt->rt_iif;
1229 
1230 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1231 	} else {
1232 		pktinfo->ipi_ifindex = 0;
1233 		pktinfo->ipi_spec_dst.s_addr = 0;
1234 	}
1235 	skb_dst_drop(skb);
1236 }
1237 
1238 int ip_setsockopt(struct sock *sk, int level,
1239 		int optname, char __user *optval, unsigned int optlen)
1240 {
1241 	int err;
1242 
1243 	if (level != SOL_IP)
1244 		return -ENOPROTOOPT;
1245 
1246 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1247 #ifdef CONFIG_NETFILTER
1248 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1249 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1250 			optname != IP_IPSEC_POLICY &&
1251 			optname != IP_XFRM_POLICY &&
1252 			!ip_mroute_opt(optname))
1253 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1254 #endif
1255 	return err;
1256 }
1257 EXPORT_SYMBOL(ip_setsockopt);
1258 
1259 #ifdef CONFIG_COMPAT
1260 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1261 			 char __user *optval, unsigned int optlen)
1262 {
1263 	int err;
1264 
1265 	if (level != SOL_IP)
1266 		return -ENOPROTOOPT;
1267 
1268 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1269 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1270 			ip_setsockopt);
1271 
1272 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1273 #ifdef CONFIG_NETFILTER
1274 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1275 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1276 			optname != IP_IPSEC_POLICY &&
1277 			optname != IP_XFRM_POLICY &&
1278 			!ip_mroute_opt(optname))
1279 		err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1280 					   optlen);
1281 #endif
1282 	return err;
1283 }
1284 EXPORT_SYMBOL(compat_ip_setsockopt);
1285 #endif
1286 
1287 /*
1288  *	Get the options. Note for future reference. The GET of IP options gets
1289  *	the _received_ ones. The set sets the _sent_ ones.
1290  */
1291 
1292 static bool getsockopt_needs_rtnl(int optname)
1293 {
1294 	switch (optname) {
1295 	case IP_MSFILTER:
1296 	case MCAST_MSFILTER:
1297 		return true;
1298 	}
1299 	return false;
1300 }
1301 
1302 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1303 			    char __user *optval, int __user *optlen, unsigned int flags)
1304 {
1305 	struct inet_sock *inet = inet_sk(sk);
1306 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1307 	int val, err = 0;
1308 	int len;
1309 
1310 	if (level != SOL_IP)
1311 		return -EOPNOTSUPP;
1312 
1313 	if (ip_mroute_opt(optname))
1314 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1315 
1316 	if (get_user(len, optlen))
1317 		return -EFAULT;
1318 	if (len < 0)
1319 		return -EINVAL;
1320 
1321 	if (needs_rtnl)
1322 		rtnl_lock();
1323 	lock_sock(sk);
1324 
1325 	switch (optname) {
1326 	case IP_OPTIONS:
1327 	{
1328 		unsigned char optbuf[sizeof(struct ip_options)+40];
1329 		struct ip_options *opt = (struct ip_options *)optbuf;
1330 		struct ip_options_rcu *inet_opt;
1331 
1332 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1333 						     lockdep_sock_is_held(sk));
1334 		opt->optlen = 0;
1335 		if (inet_opt)
1336 			memcpy(optbuf, &inet_opt->opt,
1337 			       sizeof(struct ip_options) +
1338 			       inet_opt->opt.optlen);
1339 		release_sock(sk);
1340 
1341 		if (opt->optlen == 0)
1342 			return put_user(0, optlen);
1343 
1344 		ip_options_undo(opt);
1345 
1346 		len = min_t(unsigned int, len, opt->optlen);
1347 		if (put_user(len, optlen))
1348 			return -EFAULT;
1349 		if (copy_to_user(optval, opt->__data, len))
1350 			return -EFAULT;
1351 		return 0;
1352 	}
1353 	case IP_PKTINFO:
1354 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1355 		break;
1356 	case IP_RECVTTL:
1357 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1358 		break;
1359 	case IP_RECVTOS:
1360 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1361 		break;
1362 	case IP_RECVOPTS:
1363 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1364 		break;
1365 	case IP_RETOPTS:
1366 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1367 		break;
1368 	case IP_PASSSEC:
1369 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1370 		break;
1371 	case IP_RECVORIGDSTADDR:
1372 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1373 		break;
1374 	case IP_CHECKSUM:
1375 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1376 		break;
1377 	case IP_RECVFRAGSIZE:
1378 		val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1379 		break;
1380 	case IP_TOS:
1381 		val = inet->tos;
1382 		break;
1383 	case IP_TTL:
1384 	{
1385 		struct net *net = sock_net(sk);
1386 		val = (inet->uc_ttl == -1 ?
1387 		       net->ipv4.sysctl_ip_default_ttl :
1388 		       inet->uc_ttl);
1389 		break;
1390 	}
1391 	case IP_HDRINCL:
1392 		val = inet->hdrincl;
1393 		break;
1394 	case IP_NODEFRAG:
1395 		val = inet->nodefrag;
1396 		break;
1397 	case IP_BIND_ADDRESS_NO_PORT:
1398 		val = inet->bind_address_no_port;
1399 		break;
1400 	case IP_MTU_DISCOVER:
1401 		val = inet->pmtudisc;
1402 		break;
1403 	case IP_MTU:
1404 	{
1405 		struct dst_entry *dst;
1406 		val = 0;
1407 		dst = sk_dst_get(sk);
1408 		if (dst) {
1409 			val = dst_mtu(dst);
1410 			dst_release(dst);
1411 		}
1412 		if (!val) {
1413 			release_sock(sk);
1414 			return -ENOTCONN;
1415 		}
1416 		break;
1417 	}
1418 	case IP_RECVERR:
1419 		val = inet->recverr;
1420 		break;
1421 	case IP_MULTICAST_TTL:
1422 		val = inet->mc_ttl;
1423 		break;
1424 	case IP_MULTICAST_LOOP:
1425 		val = inet->mc_loop;
1426 		break;
1427 	case IP_UNICAST_IF:
1428 		val = (__force int)htonl((__u32) inet->uc_index);
1429 		break;
1430 	case IP_MULTICAST_IF:
1431 	{
1432 		struct in_addr addr;
1433 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1434 		addr.s_addr = inet->mc_addr;
1435 		release_sock(sk);
1436 
1437 		if (put_user(len, optlen))
1438 			return -EFAULT;
1439 		if (copy_to_user(optval, &addr, len))
1440 			return -EFAULT;
1441 		return 0;
1442 	}
1443 	case IP_MSFILTER:
1444 	{
1445 		struct ip_msfilter msf;
1446 
1447 		if (len < IP_MSFILTER_SIZE(0)) {
1448 			err = -EINVAL;
1449 			goto out;
1450 		}
1451 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1452 			err = -EFAULT;
1453 			goto out;
1454 		}
1455 		err = ip_mc_msfget(sk, &msf,
1456 				   (struct ip_msfilter __user *)optval, optlen);
1457 		goto out;
1458 	}
1459 	case MCAST_MSFILTER:
1460 	{
1461 		struct group_filter gsf;
1462 
1463 		if (len < GROUP_FILTER_SIZE(0)) {
1464 			err = -EINVAL;
1465 			goto out;
1466 		}
1467 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1468 			err = -EFAULT;
1469 			goto out;
1470 		}
1471 		err = ip_mc_gsfget(sk, &gsf,
1472 				   (struct group_filter __user *)optval,
1473 				   optlen);
1474 		goto out;
1475 	}
1476 	case IP_MULTICAST_ALL:
1477 		val = inet->mc_all;
1478 		break;
1479 	case IP_PKTOPTIONS:
1480 	{
1481 		struct msghdr msg;
1482 
1483 		release_sock(sk);
1484 
1485 		if (sk->sk_type != SOCK_STREAM)
1486 			return -ENOPROTOOPT;
1487 
1488 		msg.msg_control = (__force void *) optval;
1489 		msg.msg_controllen = len;
1490 		msg.msg_flags = flags;
1491 
1492 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1493 			struct in_pktinfo info;
1494 
1495 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1496 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1497 			info.ipi_ifindex = inet->mc_index;
1498 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1499 		}
1500 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1501 			int hlim = inet->mc_ttl;
1502 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1503 		}
1504 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1505 			int tos = inet->rcv_tos;
1506 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1507 		}
1508 		len -= msg.msg_controllen;
1509 		return put_user(len, optlen);
1510 	}
1511 	case IP_FREEBIND:
1512 		val = inet->freebind;
1513 		break;
1514 	case IP_TRANSPARENT:
1515 		val = inet->transparent;
1516 		break;
1517 	case IP_MINTTL:
1518 		val = inet->min_ttl;
1519 		break;
1520 	default:
1521 		release_sock(sk);
1522 		return -ENOPROTOOPT;
1523 	}
1524 	release_sock(sk);
1525 
1526 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1527 		unsigned char ucval = (unsigned char)val;
1528 		len = 1;
1529 		if (put_user(len, optlen))
1530 			return -EFAULT;
1531 		if (copy_to_user(optval, &ucval, 1))
1532 			return -EFAULT;
1533 	} else {
1534 		len = min_t(unsigned int, sizeof(int), len);
1535 		if (put_user(len, optlen))
1536 			return -EFAULT;
1537 		if (copy_to_user(optval, &val, len))
1538 			return -EFAULT;
1539 	}
1540 	return 0;
1541 
1542 out:
1543 	release_sock(sk);
1544 	if (needs_rtnl)
1545 		rtnl_unlock();
1546 	return err;
1547 }
1548 
1549 int ip_getsockopt(struct sock *sk, int level,
1550 		  int optname, char __user *optval, int __user *optlen)
1551 {
1552 	int err;
1553 
1554 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1555 #ifdef CONFIG_NETFILTER
1556 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1557 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1558 			!ip_mroute_opt(optname)) {
1559 		int len;
1560 
1561 		if (get_user(len, optlen))
1562 			return -EFAULT;
1563 
1564 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1565 		if (err >= 0)
1566 			err = put_user(len, optlen);
1567 		return err;
1568 	}
1569 #endif
1570 	return err;
1571 }
1572 EXPORT_SYMBOL(ip_getsockopt);
1573 
1574 #ifdef CONFIG_COMPAT
1575 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1576 			 char __user *optval, int __user *optlen)
1577 {
1578 	int err;
1579 
1580 	if (optname == MCAST_MSFILTER)
1581 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1582 			ip_getsockopt);
1583 
1584 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1585 		MSG_CMSG_COMPAT);
1586 
1587 #ifdef CONFIG_NETFILTER
1588 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1589 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1590 			!ip_mroute_opt(optname)) {
1591 		int len;
1592 
1593 		if (get_user(len, optlen))
1594 			return -EFAULT;
1595 
1596 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1597 		if (err >= 0)
1598 			err = put_user(len, optlen);
1599 		return err;
1600 	}
1601 #endif
1602 	return err;
1603 }
1604 EXPORT_SYMBOL(compat_ip_getsockopt);
1605 #endif
1606