xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 3213486f)
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 	__be16 _ports[2], *ports;
152 	struct sockaddr_in sin;
153 
154 	/* All current transport protocols have the port numbers in the
155 	 * first four bytes of the transport header and this function is
156 	 * written with this assumption in mind.
157 	 */
158 	ports = skb_header_pointer(skb, skb_transport_offset(skb),
159 				   sizeof(_ports), &_ports);
160 	if (!ports)
161 		return;
162 
163 	sin.sin_family = AF_INET;
164 	sin.sin_addr.s_addr = ip_hdr(skb)->daddr;
165 	sin.sin_port = ports[1];
166 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
167 
168 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
169 }
170 
171 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
172 			 struct sk_buff *skb, int tlen, int offset)
173 {
174 	struct inet_sock *inet = inet_sk(sk);
175 	unsigned int flags = inet->cmsg_flags;
176 
177 	/* Ordered by supposed usage frequency */
178 	if (flags & IP_CMSG_PKTINFO) {
179 		ip_cmsg_recv_pktinfo(msg, skb);
180 
181 		flags &= ~IP_CMSG_PKTINFO;
182 		if (!flags)
183 			return;
184 	}
185 
186 	if (flags & IP_CMSG_TTL) {
187 		ip_cmsg_recv_ttl(msg, skb);
188 
189 		flags &= ~IP_CMSG_TTL;
190 		if (!flags)
191 			return;
192 	}
193 
194 	if (flags & IP_CMSG_TOS) {
195 		ip_cmsg_recv_tos(msg, skb);
196 
197 		flags &= ~IP_CMSG_TOS;
198 		if (!flags)
199 			return;
200 	}
201 
202 	if (flags & IP_CMSG_RECVOPTS) {
203 		ip_cmsg_recv_opts(msg, skb);
204 
205 		flags &= ~IP_CMSG_RECVOPTS;
206 		if (!flags)
207 			return;
208 	}
209 
210 	if (flags & IP_CMSG_RETOPTS) {
211 		ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
212 
213 		flags &= ~IP_CMSG_RETOPTS;
214 		if (!flags)
215 			return;
216 	}
217 
218 	if (flags & IP_CMSG_PASSSEC) {
219 		ip_cmsg_recv_security(msg, skb);
220 
221 		flags &= ~IP_CMSG_PASSSEC;
222 		if (!flags)
223 			return;
224 	}
225 
226 	if (flags & IP_CMSG_ORIGDSTADDR) {
227 		ip_cmsg_recv_dstaddr(msg, skb);
228 
229 		flags &= ~IP_CMSG_ORIGDSTADDR;
230 		if (!flags)
231 			return;
232 	}
233 
234 	if (flags & IP_CMSG_CHECKSUM)
235 		ip_cmsg_recv_checksum(msg, skb, tlen, offset);
236 
237 	if (flags & IP_CMSG_RECVFRAGSIZE)
238 		ip_cmsg_recv_fragsize(msg, skb);
239 }
240 EXPORT_SYMBOL(ip_cmsg_recv_offset);
241 
242 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
243 		 bool allow_ipv6)
244 {
245 	int err, val;
246 	struct cmsghdr *cmsg;
247 	struct net *net = sock_net(sk);
248 
249 	for_each_cmsghdr(cmsg, msg) {
250 		if (!CMSG_OK(msg, cmsg))
251 			return -EINVAL;
252 #if IS_ENABLED(CONFIG_IPV6)
253 		if (allow_ipv6 &&
254 		    cmsg->cmsg_level == SOL_IPV6 &&
255 		    cmsg->cmsg_type == IPV6_PKTINFO) {
256 			struct in6_pktinfo *src_info;
257 
258 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
259 				return -EINVAL;
260 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
261 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
262 				return -EINVAL;
263 			if (src_info->ipi6_ifindex)
264 				ipc->oif = src_info->ipi6_ifindex;
265 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
266 			continue;
267 		}
268 #endif
269 		if (cmsg->cmsg_level == SOL_SOCKET) {
270 			err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
271 			if (err)
272 				return err;
273 			continue;
274 		}
275 
276 		if (cmsg->cmsg_level != SOL_IP)
277 			continue;
278 		switch (cmsg->cmsg_type) {
279 		case IP_RETOPTS:
280 			err = cmsg->cmsg_len - sizeof(struct cmsghdr);
281 
282 			/* Our caller is responsible for freeing ipc->opt */
283 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
284 					     err < 40 ? err : 40);
285 			if (err)
286 				return err;
287 			break;
288 		case IP_PKTINFO:
289 		{
290 			struct in_pktinfo *info;
291 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
292 				return -EINVAL;
293 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
294 			if (info->ipi_ifindex)
295 				ipc->oif = info->ipi_ifindex;
296 			ipc->addr = info->ipi_spec_dst.s_addr;
297 			break;
298 		}
299 		case IP_TTL:
300 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
301 				return -EINVAL;
302 			val = *(int *)CMSG_DATA(cmsg);
303 			if (val < 1 || val > 255)
304 				return -EINVAL;
305 			ipc->ttl = val;
306 			break;
307 		case IP_TOS:
308 			if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
309 				val = *(int *)CMSG_DATA(cmsg);
310 			else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
311 				val = *(u8 *)CMSG_DATA(cmsg);
312 			else
313 				return -EINVAL;
314 			if (val < 0 || val > 255)
315 				return -EINVAL;
316 			ipc->tos = val;
317 			ipc->priority = rt_tos2priority(ipc->tos);
318 			break;
319 
320 		default:
321 			return -EINVAL;
322 		}
323 	}
324 	return 0;
325 }
326 
327 static void ip_ra_destroy_rcu(struct rcu_head *head)
328 {
329 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
330 
331 	sock_put(ra->saved_sk);
332 	kfree(ra);
333 }
334 
335 int ip_ra_control(struct sock *sk, unsigned char on,
336 		  void (*destructor)(struct sock *))
337 {
338 	struct ip_ra_chain *ra, *new_ra;
339 	struct ip_ra_chain __rcu **rap;
340 	struct net *net = sock_net(sk);
341 
342 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
343 		return -EINVAL;
344 
345 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
346 
347 	mutex_lock(&net->ipv4.ra_mutex);
348 	for (rap = &net->ipv4.ra_chain;
349 	     (ra = rcu_dereference_protected(*rap,
350 			lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
351 	     rap = &ra->next) {
352 		if (ra->sk == sk) {
353 			if (on) {
354 				mutex_unlock(&net->ipv4.ra_mutex);
355 				kfree(new_ra);
356 				return -EADDRINUSE;
357 			}
358 			/* dont let ip_call_ra_chain() use sk again */
359 			ra->sk = NULL;
360 			RCU_INIT_POINTER(*rap, ra->next);
361 			mutex_unlock(&net->ipv4.ra_mutex);
362 
363 			if (ra->destructor)
364 				ra->destructor(sk);
365 			/*
366 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
367 			 * period. This guarantee ip_call_ra_chain() dont need
368 			 * to mess with socket refcounts.
369 			 */
370 			ra->saved_sk = sk;
371 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
372 			return 0;
373 		}
374 	}
375 	if (!new_ra) {
376 		mutex_unlock(&net->ipv4.ra_mutex);
377 		return -ENOBUFS;
378 	}
379 	new_ra->sk = sk;
380 	new_ra->destructor = destructor;
381 
382 	RCU_INIT_POINTER(new_ra->next, ra);
383 	rcu_assign_pointer(*rap, new_ra);
384 	sock_hold(sk);
385 	mutex_unlock(&net->ipv4.ra_mutex);
386 
387 	return 0;
388 }
389 
390 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
391 		   __be16 port, u32 info, u8 *payload)
392 {
393 	struct sock_exterr_skb *serr;
394 
395 	skb = skb_clone(skb, GFP_ATOMIC);
396 	if (!skb)
397 		return;
398 
399 	serr = SKB_EXT_ERR(skb);
400 	serr->ee.ee_errno = err;
401 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
402 	serr->ee.ee_type = icmp_hdr(skb)->type;
403 	serr->ee.ee_code = icmp_hdr(skb)->code;
404 	serr->ee.ee_pad = 0;
405 	serr->ee.ee_info = info;
406 	serr->ee.ee_data = 0;
407 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
408 				   skb_network_header(skb);
409 	serr->port = port;
410 
411 	if (skb_pull(skb, payload - skb->data)) {
412 		skb_reset_transport_header(skb);
413 		if (sock_queue_err_skb(sk, skb) == 0)
414 			return;
415 	}
416 	kfree_skb(skb);
417 }
418 
419 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
420 {
421 	struct inet_sock *inet = inet_sk(sk);
422 	struct sock_exterr_skb *serr;
423 	struct iphdr *iph;
424 	struct sk_buff *skb;
425 
426 	if (!inet->recverr)
427 		return;
428 
429 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
430 	if (!skb)
431 		return;
432 
433 	skb_put(skb, sizeof(struct iphdr));
434 	skb_reset_network_header(skb);
435 	iph = ip_hdr(skb);
436 	iph->daddr = daddr;
437 
438 	serr = SKB_EXT_ERR(skb);
439 	serr->ee.ee_errno = err;
440 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
441 	serr->ee.ee_type = 0;
442 	serr->ee.ee_code = 0;
443 	serr->ee.ee_pad = 0;
444 	serr->ee.ee_info = info;
445 	serr->ee.ee_data = 0;
446 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
447 	serr->port = port;
448 
449 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
450 	skb_reset_transport_header(skb);
451 
452 	if (sock_queue_err_skb(sk, skb))
453 		kfree_skb(skb);
454 }
455 
456 /* For some errors we have valid addr_offset even with zero payload and
457  * zero port. Also, addr_offset should be supported if port is set.
458  */
459 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
460 {
461 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
462 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
463 }
464 
465 /* IPv4 supports cmsg on all imcp errors and some timestamps
466  *
467  * Timestamp code paths do not initialize the fields expected by cmsg:
468  * the PKTINFO fields in skb->cb[]. Fill those in here.
469  */
470 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
471 				       struct sk_buff *skb,
472 				       int ee_origin)
473 {
474 	struct in_pktinfo *info;
475 
476 	if (ee_origin == SO_EE_ORIGIN_ICMP)
477 		return true;
478 
479 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
480 		return false;
481 
482 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
483 	 * timestamp with egress dev. Not possible for packets without iif
484 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
485 	 */
486 	info = PKTINFO_SKB_CB(skb);
487 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
488 	    !info->ipi_ifindex)
489 		return false;
490 
491 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
492 	return true;
493 }
494 
495 /*
496  *	Handle MSG_ERRQUEUE
497  */
498 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
499 {
500 	struct sock_exterr_skb *serr;
501 	struct sk_buff *skb;
502 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
503 	struct {
504 		struct sock_extended_err ee;
505 		struct sockaddr_in	 offender;
506 	} errhdr;
507 	int err;
508 	int copied;
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_ssm(sk, &mreq, MCAST_INCLUDE);
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_ssm(sk, &mreq, MCAST_INCLUDE);
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 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1248 	if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1249 	    optname < BPFILTER_IPT_SET_MAX)
1250 		err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1251 #endif
1252 #ifdef CONFIG_NETFILTER
1253 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1254 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1255 			optname != IP_IPSEC_POLICY &&
1256 			optname != IP_XFRM_POLICY &&
1257 			!ip_mroute_opt(optname))
1258 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1259 #endif
1260 	return err;
1261 }
1262 EXPORT_SYMBOL(ip_setsockopt);
1263 
1264 #ifdef CONFIG_COMPAT
1265 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1266 			 char __user *optval, unsigned int optlen)
1267 {
1268 	int err;
1269 
1270 	if (level != SOL_IP)
1271 		return -ENOPROTOOPT;
1272 
1273 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1274 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1275 			ip_setsockopt);
1276 
1277 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1278 #ifdef CONFIG_NETFILTER
1279 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1280 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1281 			optname != IP_IPSEC_POLICY &&
1282 			optname != IP_XFRM_POLICY &&
1283 			!ip_mroute_opt(optname))
1284 		err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1285 					   optlen);
1286 #endif
1287 	return err;
1288 }
1289 EXPORT_SYMBOL(compat_ip_setsockopt);
1290 #endif
1291 
1292 /*
1293  *	Get the options. Note for future reference. The GET of IP options gets
1294  *	the _received_ ones. The set sets the _sent_ ones.
1295  */
1296 
1297 static bool getsockopt_needs_rtnl(int optname)
1298 {
1299 	switch (optname) {
1300 	case IP_MSFILTER:
1301 	case MCAST_MSFILTER:
1302 		return true;
1303 	}
1304 	return false;
1305 }
1306 
1307 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1308 			    char __user *optval, int __user *optlen, unsigned int flags)
1309 {
1310 	struct inet_sock *inet = inet_sk(sk);
1311 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1312 	int val, err = 0;
1313 	int len;
1314 
1315 	if (level != SOL_IP)
1316 		return -EOPNOTSUPP;
1317 
1318 	if (ip_mroute_opt(optname))
1319 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1320 
1321 	if (get_user(len, optlen))
1322 		return -EFAULT;
1323 	if (len < 0)
1324 		return -EINVAL;
1325 
1326 	if (needs_rtnl)
1327 		rtnl_lock();
1328 	lock_sock(sk);
1329 
1330 	switch (optname) {
1331 	case IP_OPTIONS:
1332 	{
1333 		unsigned char optbuf[sizeof(struct ip_options)+40];
1334 		struct ip_options *opt = (struct ip_options *)optbuf;
1335 		struct ip_options_rcu *inet_opt;
1336 
1337 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1338 						     lockdep_sock_is_held(sk));
1339 		opt->optlen = 0;
1340 		if (inet_opt)
1341 			memcpy(optbuf, &inet_opt->opt,
1342 			       sizeof(struct ip_options) +
1343 			       inet_opt->opt.optlen);
1344 		release_sock(sk);
1345 
1346 		if (opt->optlen == 0)
1347 			return put_user(0, optlen);
1348 
1349 		ip_options_undo(opt);
1350 
1351 		len = min_t(unsigned int, len, opt->optlen);
1352 		if (put_user(len, optlen))
1353 			return -EFAULT;
1354 		if (copy_to_user(optval, opt->__data, len))
1355 			return -EFAULT;
1356 		return 0;
1357 	}
1358 	case IP_PKTINFO:
1359 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1360 		break;
1361 	case IP_RECVTTL:
1362 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1363 		break;
1364 	case IP_RECVTOS:
1365 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1366 		break;
1367 	case IP_RECVOPTS:
1368 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1369 		break;
1370 	case IP_RETOPTS:
1371 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1372 		break;
1373 	case IP_PASSSEC:
1374 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1375 		break;
1376 	case IP_RECVORIGDSTADDR:
1377 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1378 		break;
1379 	case IP_CHECKSUM:
1380 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1381 		break;
1382 	case IP_RECVFRAGSIZE:
1383 		val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1384 		break;
1385 	case IP_TOS:
1386 		val = inet->tos;
1387 		break;
1388 	case IP_TTL:
1389 	{
1390 		struct net *net = sock_net(sk);
1391 		val = (inet->uc_ttl == -1 ?
1392 		       net->ipv4.sysctl_ip_default_ttl :
1393 		       inet->uc_ttl);
1394 		break;
1395 	}
1396 	case IP_HDRINCL:
1397 		val = inet->hdrincl;
1398 		break;
1399 	case IP_NODEFRAG:
1400 		val = inet->nodefrag;
1401 		break;
1402 	case IP_BIND_ADDRESS_NO_PORT:
1403 		val = inet->bind_address_no_port;
1404 		break;
1405 	case IP_MTU_DISCOVER:
1406 		val = inet->pmtudisc;
1407 		break;
1408 	case IP_MTU:
1409 	{
1410 		struct dst_entry *dst;
1411 		val = 0;
1412 		dst = sk_dst_get(sk);
1413 		if (dst) {
1414 			val = dst_mtu(dst);
1415 			dst_release(dst);
1416 		}
1417 		if (!val) {
1418 			release_sock(sk);
1419 			return -ENOTCONN;
1420 		}
1421 		break;
1422 	}
1423 	case IP_RECVERR:
1424 		val = inet->recverr;
1425 		break;
1426 	case IP_MULTICAST_TTL:
1427 		val = inet->mc_ttl;
1428 		break;
1429 	case IP_MULTICAST_LOOP:
1430 		val = inet->mc_loop;
1431 		break;
1432 	case IP_UNICAST_IF:
1433 		val = (__force int)htonl((__u32) inet->uc_index);
1434 		break;
1435 	case IP_MULTICAST_IF:
1436 	{
1437 		struct in_addr addr;
1438 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1439 		addr.s_addr = inet->mc_addr;
1440 		release_sock(sk);
1441 
1442 		if (put_user(len, optlen))
1443 			return -EFAULT;
1444 		if (copy_to_user(optval, &addr, len))
1445 			return -EFAULT;
1446 		return 0;
1447 	}
1448 	case IP_MSFILTER:
1449 	{
1450 		struct ip_msfilter msf;
1451 
1452 		if (len < IP_MSFILTER_SIZE(0)) {
1453 			err = -EINVAL;
1454 			goto out;
1455 		}
1456 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1457 			err = -EFAULT;
1458 			goto out;
1459 		}
1460 		err = ip_mc_msfget(sk, &msf,
1461 				   (struct ip_msfilter __user *)optval, optlen);
1462 		goto out;
1463 	}
1464 	case MCAST_MSFILTER:
1465 	{
1466 		struct group_filter gsf;
1467 
1468 		if (len < GROUP_FILTER_SIZE(0)) {
1469 			err = -EINVAL;
1470 			goto out;
1471 		}
1472 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1473 			err = -EFAULT;
1474 			goto out;
1475 		}
1476 		err = ip_mc_gsfget(sk, &gsf,
1477 				   (struct group_filter __user *)optval,
1478 				   optlen);
1479 		goto out;
1480 	}
1481 	case IP_MULTICAST_ALL:
1482 		val = inet->mc_all;
1483 		break;
1484 	case IP_PKTOPTIONS:
1485 	{
1486 		struct msghdr msg;
1487 
1488 		release_sock(sk);
1489 
1490 		if (sk->sk_type != SOCK_STREAM)
1491 			return -ENOPROTOOPT;
1492 
1493 		msg.msg_control = (__force void *) optval;
1494 		msg.msg_controllen = len;
1495 		msg.msg_flags = flags;
1496 
1497 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1498 			struct in_pktinfo info;
1499 
1500 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1501 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1502 			info.ipi_ifindex = inet->mc_index;
1503 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1504 		}
1505 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1506 			int hlim = inet->mc_ttl;
1507 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1508 		}
1509 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1510 			int tos = inet->rcv_tos;
1511 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1512 		}
1513 		len -= msg.msg_controllen;
1514 		return put_user(len, optlen);
1515 	}
1516 	case IP_FREEBIND:
1517 		val = inet->freebind;
1518 		break;
1519 	case IP_TRANSPARENT:
1520 		val = inet->transparent;
1521 		break;
1522 	case IP_MINTTL:
1523 		val = inet->min_ttl;
1524 		break;
1525 	default:
1526 		release_sock(sk);
1527 		return -ENOPROTOOPT;
1528 	}
1529 	release_sock(sk);
1530 
1531 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1532 		unsigned char ucval = (unsigned char)val;
1533 		len = 1;
1534 		if (put_user(len, optlen))
1535 			return -EFAULT;
1536 		if (copy_to_user(optval, &ucval, 1))
1537 			return -EFAULT;
1538 	} else {
1539 		len = min_t(unsigned int, sizeof(int), len);
1540 		if (put_user(len, optlen))
1541 			return -EFAULT;
1542 		if (copy_to_user(optval, &val, len))
1543 			return -EFAULT;
1544 	}
1545 	return 0;
1546 
1547 out:
1548 	release_sock(sk);
1549 	if (needs_rtnl)
1550 		rtnl_unlock();
1551 	return err;
1552 }
1553 
1554 int ip_getsockopt(struct sock *sk, int level,
1555 		  int optname, char __user *optval, int __user *optlen)
1556 {
1557 	int err;
1558 
1559 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1560 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1561 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1562 	    optname < BPFILTER_IPT_GET_MAX)
1563 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1564 #endif
1565 #ifdef CONFIG_NETFILTER
1566 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1567 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1568 			!ip_mroute_opt(optname)) {
1569 		int len;
1570 
1571 		if (get_user(len, optlen))
1572 			return -EFAULT;
1573 
1574 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1575 		if (err >= 0)
1576 			err = put_user(len, optlen);
1577 		return err;
1578 	}
1579 #endif
1580 	return err;
1581 }
1582 EXPORT_SYMBOL(ip_getsockopt);
1583 
1584 #ifdef CONFIG_COMPAT
1585 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1586 			 char __user *optval, int __user *optlen)
1587 {
1588 	int err;
1589 
1590 	if (optname == MCAST_MSFILTER)
1591 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1592 			ip_getsockopt);
1593 
1594 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1595 		MSG_CMSG_COMPAT);
1596 
1597 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1598 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1599 	    optname < BPFILTER_IPT_GET_MAX)
1600 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1601 #endif
1602 #ifdef CONFIG_NETFILTER
1603 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1604 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1605 			!ip_mroute_opt(optname)) {
1606 		int len;
1607 
1608 		if (get_user(len, optlen))
1609 			return -EFAULT;
1610 
1611 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1612 		if (err >= 0)
1613 			err = put_user(len, optlen);
1614 		return err;
1615 	}
1616 #endif
1617 	return err;
1618 }
1619 EXPORT_SYMBOL(compat_ip_getsockopt);
1620 #endif
1621