xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision ecefa105)
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, 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,
284 					     KERNEL_SOCKPTR(CMSG_DATA(cmsg)),
285 					     err < 40 ? err : 40);
286 			if (err)
287 				return err;
288 			break;
289 		case IP_PKTINFO:
290 		{
291 			struct in_pktinfo *info;
292 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
293 				return -EINVAL;
294 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
295 			if (info->ipi_ifindex)
296 				ipc->oif = info->ipi_ifindex;
297 			ipc->addr = info->ipi_spec_dst.s_addr;
298 			break;
299 		}
300 		case IP_TTL:
301 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
302 				return -EINVAL;
303 			val = *(int *)CMSG_DATA(cmsg);
304 			if (val < 1 || val > 255)
305 				return -EINVAL;
306 			ipc->ttl = val;
307 			break;
308 		case IP_TOS:
309 			if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
310 				val = *(int *)CMSG_DATA(cmsg);
311 			else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
312 				val = *(u8 *)CMSG_DATA(cmsg);
313 			else
314 				return -EINVAL;
315 			if (val < 0 || val > 255)
316 				return -EINVAL;
317 			ipc->tos = val;
318 			ipc->priority = rt_tos2priority(ipc->tos);
319 			break;
320 
321 		default:
322 			return -EINVAL;
323 		}
324 	}
325 	return 0;
326 }
327 
328 static void ip_ra_destroy_rcu(struct rcu_head *head)
329 {
330 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
331 
332 	sock_put(ra->saved_sk);
333 	kfree(ra);
334 }
335 
336 int ip_ra_control(struct sock *sk, unsigned char on,
337 		  void (*destructor)(struct sock *))
338 {
339 	struct ip_ra_chain *ra, *new_ra;
340 	struct ip_ra_chain __rcu **rap;
341 	struct net *net = sock_net(sk);
342 
343 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
344 		return -EINVAL;
345 
346 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
347 	if (on && !new_ra)
348 		return -ENOMEM;
349 
350 	mutex_lock(&net->ipv4.ra_mutex);
351 	for (rap = &net->ipv4.ra_chain;
352 	     (ra = rcu_dereference_protected(*rap,
353 			lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
354 	     rap = &ra->next) {
355 		if (ra->sk == sk) {
356 			if (on) {
357 				mutex_unlock(&net->ipv4.ra_mutex);
358 				kfree(new_ra);
359 				return -EADDRINUSE;
360 			}
361 			/* dont let ip_call_ra_chain() use sk again */
362 			ra->sk = NULL;
363 			RCU_INIT_POINTER(*rap, ra->next);
364 			mutex_unlock(&net->ipv4.ra_mutex);
365 
366 			if (ra->destructor)
367 				ra->destructor(sk);
368 			/*
369 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
370 			 * period. This guarantee ip_call_ra_chain() dont need
371 			 * to mess with socket refcounts.
372 			 */
373 			ra->saved_sk = sk;
374 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
375 			return 0;
376 		}
377 	}
378 	if (!new_ra) {
379 		mutex_unlock(&net->ipv4.ra_mutex);
380 		return -ENOBUFS;
381 	}
382 	new_ra->sk = sk;
383 	new_ra->destructor = destructor;
384 
385 	RCU_INIT_POINTER(new_ra->next, ra);
386 	rcu_assign_pointer(*rap, new_ra);
387 	sock_hold(sk);
388 	mutex_unlock(&net->ipv4.ra_mutex);
389 
390 	return 0;
391 }
392 
393 static void ipv4_icmp_error_rfc4884(const struct sk_buff *skb,
394 				    struct sock_ee_data_rfc4884 *out)
395 {
396 	switch (icmp_hdr(skb)->type) {
397 	case ICMP_DEST_UNREACH:
398 	case ICMP_TIME_EXCEEDED:
399 	case ICMP_PARAMETERPROB:
400 		ip_icmp_error_rfc4884(skb, out, sizeof(struct icmphdr),
401 				      icmp_hdr(skb)->un.reserved[1] * 4);
402 	}
403 }
404 
405 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
406 		   __be16 port, u32 info, u8 *payload)
407 {
408 	struct sock_exterr_skb *serr;
409 
410 	skb = skb_clone(skb, GFP_ATOMIC);
411 	if (!skb)
412 		return;
413 
414 	serr = SKB_EXT_ERR(skb);
415 	serr->ee.ee_errno = err;
416 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
417 	serr->ee.ee_type = icmp_hdr(skb)->type;
418 	serr->ee.ee_code = icmp_hdr(skb)->code;
419 	serr->ee.ee_pad = 0;
420 	serr->ee.ee_info = info;
421 	serr->ee.ee_data = 0;
422 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
423 				   skb_network_header(skb);
424 	serr->port = port;
425 
426 	if (skb_pull(skb, payload - skb->data)) {
427 		if (inet_sk(sk)->recverr_rfc4884)
428 			ipv4_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
429 
430 		skb_reset_transport_header(skb);
431 		if (sock_queue_err_skb(sk, skb) == 0)
432 			return;
433 	}
434 	kfree_skb(skb);
435 }
436 EXPORT_SYMBOL_GPL(ip_icmp_error);
437 
438 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
439 {
440 	struct inet_sock *inet = inet_sk(sk);
441 	struct sock_exterr_skb *serr;
442 	struct iphdr *iph;
443 	struct sk_buff *skb;
444 
445 	if (!inet->recverr)
446 		return;
447 
448 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
449 	if (!skb)
450 		return;
451 
452 	skb_put(skb, sizeof(struct iphdr));
453 	skb_reset_network_header(skb);
454 	iph = ip_hdr(skb);
455 	iph->daddr = daddr;
456 
457 	serr = SKB_EXT_ERR(skb);
458 	serr->ee.ee_errno = err;
459 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
460 	serr->ee.ee_type = 0;
461 	serr->ee.ee_code = 0;
462 	serr->ee.ee_pad = 0;
463 	serr->ee.ee_info = info;
464 	serr->ee.ee_data = 0;
465 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
466 	serr->port = port;
467 
468 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
469 	skb_reset_transport_header(skb);
470 
471 	if (sock_queue_err_skb(sk, skb))
472 		kfree_skb(skb);
473 }
474 
475 /* For some errors we have valid addr_offset even with zero payload and
476  * zero port. Also, addr_offset should be supported if port is set.
477  */
478 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
479 {
480 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
481 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
482 }
483 
484 /* IPv4 supports cmsg on all imcp errors and some timestamps
485  *
486  * Timestamp code paths do not initialize the fields expected by cmsg:
487  * the PKTINFO fields in skb->cb[]. Fill those in here.
488  */
489 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
490 				       struct sk_buff *skb,
491 				       int ee_origin)
492 {
493 	struct in_pktinfo *info;
494 
495 	if (ee_origin == SO_EE_ORIGIN_ICMP)
496 		return true;
497 
498 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
499 		return false;
500 
501 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
502 	 * timestamp with egress dev. Not possible for packets without iif
503 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
504 	 */
505 	info = PKTINFO_SKB_CB(skb);
506 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
507 	    !info->ipi_ifindex)
508 		return false;
509 
510 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
511 	return true;
512 }
513 
514 /*
515  *	Handle MSG_ERRQUEUE
516  */
517 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
518 {
519 	struct sock_exterr_skb *serr;
520 	struct sk_buff *skb;
521 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
522 	struct {
523 		struct sock_extended_err ee;
524 		struct sockaddr_in	 offender;
525 	} errhdr;
526 	int err;
527 	int copied;
528 
529 	err = -EAGAIN;
530 	skb = sock_dequeue_err_skb(sk);
531 	if (!skb)
532 		goto out;
533 
534 	copied = skb->len;
535 	if (copied > len) {
536 		msg->msg_flags |= MSG_TRUNC;
537 		copied = len;
538 	}
539 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
540 	if (unlikely(err)) {
541 		kfree_skb(skb);
542 		return err;
543 	}
544 	sock_recv_timestamp(msg, sk, skb);
545 
546 	serr = SKB_EXT_ERR(skb);
547 
548 	if (sin && ipv4_datagram_support_addr(serr)) {
549 		sin->sin_family = AF_INET;
550 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
551 						   serr->addr_offset);
552 		sin->sin_port = serr->port;
553 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
554 		*addr_len = sizeof(*sin);
555 	}
556 
557 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
558 	sin = &errhdr.offender;
559 	memset(sin, 0, sizeof(*sin));
560 
561 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
562 		sin->sin_family = AF_INET;
563 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
564 		if (inet_sk(sk)->cmsg_flags)
565 			ip_cmsg_recv(msg, skb);
566 	}
567 
568 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
569 
570 	/* Now we could try to dump offended packet options */
571 
572 	msg->msg_flags |= MSG_ERRQUEUE;
573 	err = copied;
574 
575 	consume_skb(skb);
576 out:
577 	return err;
578 }
579 
580 void __ip_sock_set_tos(struct sock *sk, int val)
581 {
582 	if (sk->sk_type == SOCK_STREAM) {
583 		val &= ~INET_ECN_MASK;
584 		val |= inet_sk(sk)->tos & INET_ECN_MASK;
585 	}
586 	if (inet_sk(sk)->tos != val) {
587 		inet_sk(sk)->tos = val;
588 		sk->sk_priority = rt_tos2priority(val);
589 		sk_dst_reset(sk);
590 	}
591 }
592 
593 void ip_sock_set_tos(struct sock *sk, int val)
594 {
595 	lock_sock(sk);
596 	__ip_sock_set_tos(sk, val);
597 	release_sock(sk);
598 }
599 EXPORT_SYMBOL(ip_sock_set_tos);
600 
601 void ip_sock_set_freebind(struct sock *sk)
602 {
603 	lock_sock(sk);
604 	inet_sk(sk)->freebind = true;
605 	release_sock(sk);
606 }
607 EXPORT_SYMBOL(ip_sock_set_freebind);
608 
609 void ip_sock_set_recverr(struct sock *sk)
610 {
611 	lock_sock(sk);
612 	inet_sk(sk)->recverr = true;
613 	release_sock(sk);
614 }
615 EXPORT_SYMBOL(ip_sock_set_recverr);
616 
617 int ip_sock_set_mtu_discover(struct sock *sk, int val)
618 {
619 	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
620 		return -EINVAL;
621 	lock_sock(sk);
622 	inet_sk(sk)->pmtudisc = val;
623 	release_sock(sk);
624 	return 0;
625 }
626 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
627 
628 void ip_sock_set_pktinfo(struct sock *sk)
629 {
630 	lock_sock(sk);
631 	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
632 	release_sock(sk);
633 }
634 EXPORT_SYMBOL(ip_sock_set_pktinfo);
635 
636 /*
637  *	Socket option code for IP. This is the end of the line after any
638  *	TCP,UDP etc options on an IP socket.
639  */
640 static bool setsockopt_needs_rtnl(int optname)
641 {
642 	switch (optname) {
643 	case IP_ADD_MEMBERSHIP:
644 	case IP_ADD_SOURCE_MEMBERSHIP:
645 	case IP_BLOCK_SOURCE:
646 	case IP_DROP_MEMBERSHIP:
647 	case IP_DROP_SOURCE_MEMBERSHIP:
648 	case IP_MSFILTER:
649 	case IP_UNBLOCK_SOURCE:
650 	case MCAST_BLOCK_SOURCE:
651 	case MCAST_MSFILTER:
652 	case MCAST_JOIN_GROUP:
653 	case MCAST_JOIN_SOURCE_GROUP:
654 	case MCAST_LEAVE_GROUP:
655 	case MCAST_LEAVE_SOURCE_GROUP:
656 	case MCAST_UNBLOCK_SOURCE:
657 		return true;
658 	}
659 	return false;
660 }
661 
662 static int set_mcast_msfilter(struct sock *sk, int ifindex,
663 			      int numsrc, int fmode,
664 			      struct sockaddr_storage *group,
665 			      struct sockaddr_storage *list)
666 {
667 	struct ip_msfilter *msf;
668 	struct sockaddr_in *psin;
669 	int err, i;
670 
671 	msf = kmalloc(IP_MSFILTER_SIZE(numsrc), GFP_KERNEL);
672 	if (!msf)
673 		return -ENOBUFS;
674 
675 	psin = (struct sockaddr_in *)group;
676 	if (psin->sin_family != AF_INET)
677 		goto Eaddrnotavail;
678 	msf->imsf_multiaddr = psin->sin_addr.s_addr;
679 	msf->imsf_interface = 0;
680 	msf->imsf_fmode = fmode;
681 	msf->imsf_numsrc = numsrc;
682 	for (i = 0; i < numsrc; ++i) {
683 		psin = (struct sockaddr_in *)&list[i];
684 
685 		if (psin->sin_family != AF_INET)
686 			goto Eaddrnotavail;
687 		msf->imsf_slist_flex[i] = psin->sin_addr.s_addr;
688 	}
689 	err = ip_mc_msfilter(sk, msf, ifindex);
690 	kfree(msf);
691 	return err;
692 
693 Eaddrnotavail:
694 	kfree(msf);
695 	return -EADDRNOTAVAIL;
696 }
697 
698 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
699 		sockptr_t optval, int optlen)
700 {
701 	if (in_compat_syscall()) {
702 		struct compat_group_source_req gr32;
703 
704 		if (optlen != sizeof(gr32))
705 			return -EINVAL;
706 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
707 			return -EFAULT;
708 		greqs->gsr_interface = gr32.gsr_interface;
709 		greqs->gsr_group = gr32.gsr_group;
710 		greqs->gsr_source = gr32.gsr_source;
711 	} else {
712 		if (optlen != sizeof(*greqs))
713 			return -EINVAL;
714 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
715 			return -EFAULT;
716 	}
717 
718 	return 0;
719 }
720 
721 static int do_mcast_group_source(struct sock *sk, int optname,
722 		sockptr_t optval, int optlen)
723 {
724 	struct group_source_req greqs;
725 	struct ip_mreq_source mreqs;
726 	struct sockaddr_in *psin;
727 	int omode, add, err;
728 
729 	err = copy_group_source_from_sockptr(&greqs, optval, optlen);
730 	if (err)
731 		return err;
732 
733 	if (greqs.gsr_group.ss_family != AF_INET ||
734 	    greqs.gsr_source.ss_family != AF_INET)
735 		return -EADDRNOTAVAIL;
736 
737 	psin = (struct sockaddr_in *)&greqs.gsr_group;
738 	mreqs.imr_multiaddr = psin->sin_addr.s_addr;
739 	psin = (struct sockaddr_in *)&greqs.gsr_source;
740 	mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
741 	mreqs.imr_interface = 0; /* use index for mc_source */
742 
743 	if (optname == MCAST_BLOCK_SOURCE) {
744 		omode = MCAST_EXCLUDE;
745 		add = 1;
746 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
747 		omode = MCAST_EXCLUDE;
748 		add = 0;
749 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
750 		struct ip_mreqn mreq;
751 
752 		psin = (struct sockaddr_in *)&greqs.gsr_group;
753 		mreq.imr_multiaddr = psin->sin_addr;
754 		mreq.imr_address.s_addr = 0;
755 		mreq.imr_ifindex = greqs.gsr_interface;
756 		err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
757 		if (err && err != -EADDRINUSE)
758 			return err;
759 		greqs.gsr_interface = mreq.imr_ifindex;
760 		omode = MCAST_INCLUDE;
761 		add = 1;
762 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
763 		omode = MCAST_INCLUDE;
764 		add = 0;
765 	}
766 	return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
767 }
768 
769 static int ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval, int optlen)
770 {
771 	struct group_filter *gsf = NULL;
772 	int err;
773 
774 	if (optlen < GROUP_FILTER_SIZE(0))
775 		return -EINVAL;
776 	if (optlen > READ_ONCE(sysctl_optmem_max))
777 		return -ENOBUFS;
778 
779 	gsf = memdup_sockptr(optval, optlen);
780 	if (IS_ERR(gsf))
781 		return PTR_ERR(gsf);
782 
783 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
784 	err = -ENOBUFS;
785 	if (gsf->gf_numsrc >= 0x1ffffff ||
786 	    gsf->gf_numsrc > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
787 		goto out_free_gsf;
788 
789 	err = -EINVAL;
790 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
791 		goto out_free_gsf;
792 
793 	err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
794 				 gsf->gf_fmode, &gsf->gf_group,
795 				 gsf->gf_slist_flex);
796 out_free_gsf:
797 	kfree(gsf);
798 	return err;
799 }
800 
801 static int compat_ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
802 		int optlen)
803 {
804 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
805 	struct compat_group_filter *gf32;
806 	unsigned int n;
807 	void *p;
808 	int err;
809 
810 	if (optlen < size0)
811 		return -EINVAL;
812 	if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
813 		return -ENOBUFS;
814 
815 	p = kmalloc(optlen + 4, GFP_KERNEL);
816 	if (!p)
817 		return -ENOMEM;
818 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
819 
820 	err = -EFAULT;
821 	if (copy_from_sockptr(gf32, optval, optlen))
822 		goto out_free_gsf;
823 
824 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
825 	n = gf32->gf_numsrc;
826 	err = -ENOBUFS;
827 	if (n >= 0x1ffffff)
828 		goto out_free_gsf;
829 
830 	err = -EINVAL;
831 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
832 		goto out_free_gsf;
833 
834 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
835 	err = -ENOBUFS;
836 	if (n > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
837 		goto out_free_gsf;
838 	err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
839 				 &gf32->gf_group, gf32->gf_slist_flex);
840 out_free_gsf:
841 	kfree(p);
842 	return err;
843 }
844 
845 static int ip_mcast_join_leave(struct sock *sk, int optname,
846 		sockptr_t optval, int optlen)
847 {
848 	struct ip_mreqn mreq = { };
849 	struct sockaddr_in *psin;
850 	struct group_req greq;
851 
852 	if (optlen < sizeof(struct group_req))
853 		return -EINVAL;
854 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
855 		return -EFAULT;
856 
857 	psin = (struct sockaddr_in *)&greq.gr_group;
858 	if (psin->sin_family != AF_INET)
859 		return -EINVAL;
860 	mreq.imr_multiaddr = psin->sin_addr;
861 	mreq.imr_ifindex = greq.gr_interface;
862 	if (optname == MCAST_JOIN_GROUP)
863 		return ip_mc_join_group(sk, &mreq);
864 	return ip_mc_leave_group(sk, &mreq);
865 }
866 
867 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
868 		sockptr_t optval, int optlen)
869 {
870 	struct compat_group_req greq;
871 	struct ip_mreqn mreq = { };
872 	struct sockaddr_in *psin;
873 
874 	if (optlen < sizeof(struct compat_group_req))
875 		return -EINVAL;
876 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
877 		return -EFAULT;
878 
879 	psin = (struct sockaddr_in *)&greq.gr_group;
880 	if (psin->sin_family != AF_INET)
881 		return -EINVAL;
882 	mreq.imr_multiaddr = psin->sin_addr;
883 	mreq.imr_ifindex = greq.gr_interface;
884 
885 	if (optname == MCAST_JOIN_GROUP)
886 		return ip_mc_join_group(sk, &mreq);
887 	return ip_mc_leave_group(sk, &mreq);
888 }
889 
890 DEFINE_STATIC_KEY_FALSE(ip4_min_ttl);
891 
892 int do_ip_setsockopt(struct sock *sk, int level, int optname,
893 		     sockptr_t optval, unsigned int optlen)
894 {
895 	struct inet_sock *inet = inet_sk(sk);
896 	struct net *net = sock_net(sk);
897 	int val = 0, err;
898 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
899 
900 	switch (optname) {
901 	case IP_PKTINFO:
902 	case IP_RECVTTL:
903 	case IP_RECVOPTS:
904 	case IP_RECVTOS:
905 	case IP_RETOPTS:
906 	case IP_TOS:
907 	case IP_TTL:
908 	case IP_HDRINCL:
909 	case IP_MTU_DISCOVER:
910 	case IP_RECVERR:
911 	case IP_ROUTER_ALERT:
912 	case IP_FREEBIND:
913 	case IP_PASSSEC:
914 	case IP_TRANSPARENT:
915 	case IP_MINTTL:
916 	case IP_NODEFRAG:
917 	case IP_BIND_ADDRESS_NO_PORT:
918 	case IP_UNICAST_IF:
919 	case IP_MULTICAST_TTL:
920 	case IP_MULTICAST_ALL:
921 	case IP_MULTICAST_LOOP:
922 	case IP_RECVORIGDSTADDR:
923 	case IP_CHECKSUM:
924 	case IP_RECVFRAGSIZE:
925 	case IP_RECVERR_RFC4884:
926 	case IP_LOCAL_PORT_RANGE:
927 		if (optlen >= sizeof(int)) {
928 			if (copy_from_sockptr(&val, optval, sizeof(val)))
929 				return -EFAULT;
930 		} else if (optlen >= sizeof(char)) {
931 			unsigned char ucval;
932 
933 			if (copy_from_sockptr(&ucval, optval, sizeof(ucval)))
934 				return -EFAULT;
935 			val = (int) ucval;
936 		}
937 	}
938 
939 	/* If optlen==0, it is equivalent to val == 0 */
940 
941 	if (optname == IP_ROUTER_ALERT)
942 		return ip_ra_control(sk, val ? 1 : 0, NULL);
943 	if (ip_mroute_opt(optname))
944 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
945 
946 	err = 0;
947 	if (needs_rtnl)
948 		rtnl_lock();
949 	sockopt_lock_sock(sk);
950 
951 	switch (optname) {
952 	case IP_OPTIONS:
953 	{
954 		struct ip_options_rcu *old, *opt = NULL;
955 
956 		if (optlen > 40)
957 			goto e_inval;
958 		err = ip_options_get(sock_net(sk), &opt, optval, optlen);
959 		if (err)
960 			break;
961 		old = rcu_dereference_protected(inet->inet_opt,
962 						lockdep_sock_is_held(sk));
963 		if (inet->is_icsk) {
964 			struct inet_connection_sock *icsk = inet_csk(sk);
965 #if IS_ENABLED(CONFIG_IPV6)
966 			if (sk->sk_family == PF_INET ||
967 			    (!((1 << sk->sk_state) &
968 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
969 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
970 #endif
971 				if (old)
972 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
973 				if (opt)
974 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
975 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
976 #if IS_ENABLED(CONFIG_IPV6)
977 			}
978 #endif
979 		}
980 		rcu_assign_pointer(inet->inet_opt, opt);
981 		if (old)
982 			kfree_rcu(old, rcu);
983 		break;
984 	}
985 	case IP_PKTINFO:
986 		if (val)
987 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
988 		else
989 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
990 		break;
991 	case IP_RECVTTL:
992 		if (val)
993 			inet->cmsg_flags |=  IP_CMSG_TTL;
994 		else
995 			inet->cmsg_flags &= ~IP_CMSG_TTL;
996 		break;
997 	case IP_RECVTOS:
998 		if (val)
999 			inet->cmsg_flags |=  IP_CMSG_TOS;
1000 		else
1001 			inet->cmsg_flags &= ~IP_CMSG_TOS;
1002 		break;
1003 	case IP_RECVOPTS:
1004 		if (val)
1005 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
1006 		else
1007 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
1008 		break;
1009 	case IP_RETOPTS:
1010 		if (val)
1011 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
1012 		else
1013 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
1014 		break;
1015 	case IP_PASSSEC:
1016 		if (val)
1017 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
1018 		else
1019 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
1020 		break;
1021 	case IP_RECVORIGDSTADDR:
1022 		if (val)
1023 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
1024 		else
1025 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
1026 		break;
1027 	case IP_CHECKSUM:
1028 		if (val) {
1029 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
1030 				inet_inc_convert_csum(sk);
1031 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
1032 			}
1033 		} else {
1034 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
1035 				inet_dec_convert_csum(sk);
1036 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
1037 			}
1038 		}
1039 		break;
1040 	case IP_RECVFRAGSIZE:
1041 		if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
1042 			goto e_inval;
1043 		if (val)
1044 			inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
1045 		else
1046 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
1047 		break;
1048 	case IP_TOS:	/* This sets both TOS and Precedence */
1049 		__ip_sock_set_tos(sk, val);
1050 		break;
1051 	case IP_TTL:
1052 		if (optlen < 1)
1053 			goto e_inval;
1054 		if (val != -1 && (val < 1 || val > 255))
1055 			goto e_inval;
1056 		inet->uc_ttl = val;
1057 		break;
1058 	case IP_HDRINCL:
1059 		if (sk->sk_type != SOCK_RAW) {
1060 			err = -ENOPROTOOPT;
1061 			break;
1062 		}
1063 		inet->hdrincl = val ? 1 : 0;
1064 		break;
1065 	case IP_NODEFRAG:
1066 		if (sk->sk_type != SOCK_RAW) {
1067 			err = -ENOPROTOOPT;
1068 			break;
1069 		}
1070 		inet->nodefrag = val ? 1 : 0;
1071 		break;
1072 	case IP_BIND_ADDRESS_NO_PORT:
1073 		inet->bind_address_no_port = val ? 1 : 0;
1074 		break;
1075 	case IP_MTU_DISCOVER:
1076 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
1077 			goto e_inval;
1078 		inet->pmtudisc = val;
1079 		break;
1080 	case IP_RECVERR:
1081 		inet->recverr = !!val;
1082 		if (!val)
1083 			skb_queue_purge(&sk->sk_error_queue);
1084 		break;
1085 	case IP_RECVERR_RFC4884:
1086 		if (val < 0 || val > 1)
1087 			goto e_inval;
1088 		inet->recverr_rfc4884 = !!val;
1089 		break;
1090 	case IP_MULTICAST_TTL:
1091 		if (sk->sk_type == SOCK_STREAM)
1092 			goto e_inval;
1093 		if (optlen < 1)
1094 			goto e_inval;
1095 		if (val == -1)
1096 			val = 1;
1097 		if (val < 0 || val > 255)
1098 			goto e_inval;
1099 		inet->mc_ttl = val;
1100 		break;
1101 	case IP_MULTICAST_LOOP:
1102 		if (optlen < 1)
1103 			goto e_inval;
1104 		inet->mc_loop = !!val;
1105 		break;
1106 	case IP_UNICAST_IF:
1107 	{
1108 		struct net_device *dev = NULL;
1109 		int ifindex;
1110 		int midx;
1111 
1112 		if (optlen != sizeof(int))
1113 			goto e_inval;
1114 
1115 		ifindex = (__force int)ntohl((__force __be32)val);
1116 		if (ifindex == 0) {
1117 			inet->uc_index = 0;
1118 			err = 0;
1119 			break;
1120 		}
1121 
1122 		dev = dev_get_by_index(sock_net(sk), ifindex);
1123 		err = -EADDRNOTAVAIL;
1124 		if (!dev)
1125 			break;
1126 
1127 		midx = l3mdev_master_ifindex(dev);
1128 		dev_put(dev);
1129 
1130 		err = -EINVAL;
1131 		if (sk->sk_bound_dev_if && midx != sk->sk_bound_dev_if)
1132 			break;
1133 
1134 		inet->uc_index = ifindex;
1135 		err = 0;
1136 		break;
1137 	}
1138 	case IP_MULTICAST_IF:
1139 	{
1140 		struct ip_mreqn mreq;
1141 		struct net_device *dev = NULL;
1142 		int midx;
1143 
1144 		if (sk->sk_type == SOCK_STREAM)
1145 			goto e_inval;
1146 		/*
1147 		 *	Check the arguments are allowable
1148 		 */
1149 
1150 		if (optlen < sizeof(struct in_addr))
1151 			goto e_inval;
1152 
1153 		err = -EFAULT;
1154 		if (optlen >= sizeof(struct ip_mreqn)) {
1155 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1156 				break;
1157 		} else {
1158 			memset(&mreq, 0, sizeof(mreq));
1159 			if (optlen >= sizeof(struct ip_mreq)) {
1160 				if (copy_from_sockptr(&mreq, optval,
1161 						      sizeof(struct ip_mreq)))
1162 					break;
1163 			} else if (optlen >= sizeof(struct in_addr)) {
1164 				if (copy_from_sockptr(&mreq.imr_address, optval,
1165 						      sizeof(struct in_addr)))
1166 					break;
1167 			}
1168 		}
1169 
1170 		if (!mreq.imr_ifindex) {
1171 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1172 				inet->mc_index = 0;
1173 				inet->mc_addr  = 0;
1174 				err = 0;
1175 				break;
1176 			}
1177 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1178 			if (dev)
1179 				mreq.imr_ifindex = dev->ifindex;
1180 		} else
1181 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1182 
1183 
1184 		err = -EADDRNOTAVAIL;
1185 		if (!dev)
1186 			break;
1187 
1188 		midx = l3mdev_master_ifindex(dev);
1189 
1190 		dev_put(dev);
1191 
1192 		err = -EINVAL;
1193 		if (sk->sk_bound_dev_if &&
1194 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
1195 		    midx != sk->sk_bound_dev_if)
1196 			break;
1197 
1198 		inet->mc_index = mreq.imr_ifindex;
1199 		inet->mc_addr  = mreq.imr_address.s_addr;
1200 		err = 0;
1201 		break;
1202 	}
1203 
1204 	case IP_ADD_MEMBERSHIP:
1205 	case IP_DROP_MEMBERSHIP:
1206 	{
1207 		struct ip_mreqn mreq;
1208 
1209 		err = -EPROTO;
1210 		if (inet_sk(sk)->is_icsk)
1211 			break;
1212 
1213 		if (optlen < sizeof(struct ip_mreq))
1214 			goto e_inval;
1215 		err = -EFAULT;
1216 		if (optlen >= sizeof(struct ip_mreqn)) {
1217 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1218 				break;
1219 		} else {
1220 			memset(&mreq, 0, sizeof(mreq));
1221 			if (copy_from_sockptr(&mreq, optval,
1222 					      sizeof(struct ip_mreq)))
1223 				break;
1224 		}
1225 
1226 		if (optname == IP_ADD_MEMBERSHIP)
1227 			err = ip_mc_join_group(sk, &mreq);
1228 		else
1229 			err = ip_mc_leave_group(sk, &mreq);
1230 		break;
1231 	}
1232 	case IP_MSFILTER:
1233 	{
1234 		struct ip_msfilter *msf;
1235 
1236 		if (optlen < IP_MSFILTER_SIZE(0))
1237 			goto e_inval;
1238 		if (optlen > READ_ONCE(sysctl_optmem_max)) {
1239 			err = -ENOBUFS;
1240 			break;
1241 		}
1242 		msf = memdup_sockptr(optval, optlen);
1243 		if (IS_ERR(msf)) {
1244 			err = PTR_ERR(msf);
1245 			break;
1246 		}
1247 		/* numsrc >= (1G-4) overflow in 32 bits */
1248 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
1249 		    msf->imsf_numsrc > READ_ONCE(net->ipv4.sysctl_igmp_max_msf)) {
1250 			kfree(msf);
1251 			err = -ENOBUFS;
1252 			break;
1253 		}
1254 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1255 			kfree(msf);
1256 			err = -EINVAL;
1257 			break;
1258 		}
1259 		err = ip_mc_msfilter(sk, msf, 0);
1260 		kfree(msf);
1261 		break;
1262 	}
1263 	case IP_BLOCK_SOURCE:
1264 	case IP_UNBLOCK_SOURCE:
1265 	case IP_ADD_SOURCE_MEMBERSHIP:
1266 	case IP_DROP_SOURCE_MEMBERSHIP:
1267 	{
1268 		struct ip_mreq_source mreqs;
1269 		int omode, add;
1270 
1271 		if (optlen != sizeof(struct ip_mreq_source))
1272 			goto e_inval;
1273 		if (copy_from_sockptr(&mreqs, optval, sizeof(mreqs))) {
1274 			err = -EFAULT;
1275 			break;
1276 		}
1277 		if (optname == IP_BLOCK_SOURCE) {
1278 			omode = MCAST_EXCLUDE;
1279 			add = 1;
1280 		} else if (optname == IP_UNBLOCK_SOURCE) {
1281 			omode = MCAST_EXCLUDE;
1282 			add = 0;
1283 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1284 			struct ip_mreqn mreq;
1285 
1286 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1287 			mreq.imr_address.s_addr = mreqs.imr_interface;
1288 			mreq.imr_ifindex = 0;
1289 			err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1290 			if (err && err != -EADDRINUSE)
1291 				break;
1292 			omode = MCAST_INCLUDE;
1293 			add = 1;
1294 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1295 			omode = MCAST_INCLUDE;
1296 			add = 0;
1297 		}
1298 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
1299 		break;
1300 	}
1301 	case MCAST_JOIN_GROUP:
1302 	case MCAST_LEAVE_GROUP:
1303 		if (in_compat_syscall())
1304 			err = compat_ip_mcast_join_leave(sk, optname, optval,
1305 							 optlen);
1306 		else
1307 			err = ip_mcast_join_leave(sk, optname, optval, optlen);
1308 		break;
1309 	case MCAST_JOIN_SOURCE_GROUP:
1310 	case MCAST_LEAVE_SOURCE_GROUP:
1311 	case MCAST_BLOCK_SOURCE:
1312 	case MCAST_UNBLOCK_SOURCE:
1313 		err = do_mcast_group_source(sk, optname, optval, optlen);
1314 		break;
1315 	case MCAST_MSFILTER:
1316 		if (in_compat_syscall())
1317 			err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1318 		else
1319 			err = ip_set_mcast_msfilter(sk, optval, optlen);
1320 		break;
1321 	case IP_MULTICAST_ALL:
1322 		if (optlen < 1)
1323 			goto e_inval;
1324 		if (val != 0 && val != 1)
1325 			goto e_inval;
1326 		inet->mc_all = val;
1327 		break;
1328 
1329 	case IP_FREEBIND:
1330 		if (optlen < 1)
1331 			goto e_inval;
1332 		inet->freebind = !!val;
1333 		break;
1334 
1335 	case IP_IPSEC_POLICY:
1336 	case IP_XFRM_POLICY:
1337 		err = -EPERM;
1338 		if (!sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1339 			break;
1340 		err = xfrm_user_policy(sk, optname, optval, optlen);
1341 		break;
1342 
1343 	case IP_TRANSPARENT:
1344 		if (!!val && !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1345 		    !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1346 			err = -EPERM;
1347 			break;
1348 		}
1349 		if (optlen < 1)
1350 			goto e_inval;
1351 		inet->transparent = !!val;
1352 		break;
1353 
1354 	case IP_MINTTL:
1355 		if (optlen < 1)
1356 			goto e_inval;
1357 		if (val < 0 || val > 255)
1358 			goto e_inval;
1359 
1360 		if (val)
1361 			static_branch_enable(&ip4_min_ttl);
1362 
1363 		/* tcp_v4_err() and tcp_v4_rcv() might read min_ttl
1364 		 * while we are changint it.
1365 		 */
1366 		WRITE_ONCE(inet->min_ttl, val);
1367 		break;
1368 
1369 	case IP_LOCAL_PORT_RANGE:
1370 	{
1371 		const __u16 lo = val;
1372 		const __u16 hi = val >> 16;
1373 
1374 		if (optlen != sizeof(__u32))
1375 			goto e_inval;
1376 		if (lo != 0 && hi != 0 && lo > hi)
1377 			goto e_inval;
1378 
1379 		inet->local_port_range.lo = lo;
1380 		inet->local_port_range.hi = hi;
1381 		break;
1382 	}
1383 	default:
1384 		err = -ENOPROTOOPT;
1385 		break;
1386 	}
1387 	sockopt_release_sock(sk);
1388 	if (needs_rtnl)
1389 		rtnl_unlock();
1390 	return err;
1391 
1392 e_inval:
1393 	sockopt_release_sock(sk);
1394 	if (needs_rtnl)
1395 		rtnl_unlock();
1396 	return -EINVAL;
1397 }
1398 
1399 /**
1400  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1401  * @sk: socket
1402  * @skb: buffer
1403  *
1404  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1405  * destination in skb->cb[] before dst drop.
1406  * This way, receiver doesn't make cache line misses to read rtable.
1407  */
1408 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1409 {
1410 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1411 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1412 		       ipv6_sk_rxinfo(sk);
1413 
1414 	if (prepare && skb_rtable(skb)) {
1415 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1416 		 * which has interface index (iif) as the first member of the
1417 		 * underlying inet{6}_skb_parm struct. This code then overlays
1418 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1419 		 * element so the iif is picked up from the prior IPCB. If iif
1420 		 * is the loopback interface, then return the sending interface
1421 		 * (e.g., process binds socket to eth0 for Tx which is
1422 		 * redirected to loopback in the rtable/dst).
1423 		 */
1424 		struct rtable *rt = skb_rtable(skb);
1425 		bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1426 
1427 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1428 			pktinfo->ipi_ifindex = inet_iif(skb);
1429 		else if (l3slave && rt && rt->rt_iif)
1430 			pktinfo->ipi_ifindex = rt->rt_iif;
1431 
1432 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1433 	} else {
1434 		pktinfo->ipi_ifindex = 0;
1435 		pktinfo->ipi_spec_dst.s_addr = 0;
1436 	}
1437 	skb_dst_drop(skb);
1438 }
1439 
1440 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1441 		unsigned int optlen)
1442 {
1443 	int err;
1444 
1445 	if (level != SOL_IP)
1446 		return -ENOPROTOOPT;
1447 
1448 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1449 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1450 	if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1451 	    optname < BPFILTER_IPT_SET_MAX)
1452 		err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1453 #endif
1454 #ifdef CONFIG_NETFILTER
1455 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1456 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1457 			optname != IP_IPSEC_POLICY &&
1458 			optname != IP_XFRM_POLICY &&
1459 			!ip_mroute_opt(optname))
1460 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1461 #endif
1462 	return err;
1463 }
1464 EXPORT_SYMBOL(ip_setsockopt);
1465 
1466 /*
1467  *	Get the options. Note for future reference. The GET of IP options gets
1468  *	the _received_ ones. The set sets the _sent_ ones.
1469  */
1470 
1471 static bool getsockopt_needs_rtnl(int optname)
1472 {
1473 	switch (optname) {
1474 	case IP_MSFILTER:
1475 	case MCAST_MSFILTER:
1476 		return true;
1477 	}
1478 	return false;
1479 }
1480 
1481 static int ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1482 				 sockptr_t optlen, int len)
1483 {
1484 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1485 	struct group_filter gsf;
1486 	int num, gsf_size;
1487 	int err;
1488 
1489 	if (len < size0)
1490 		return -EINVAL;
1491 	if (copy_from_sockptr(&gsf, optval, size0))
1492 		return -EFAULT;
1493 
1494 	num = gsf.gf_numsrc;
1495 	err = ip_mc_gsfget(sk, &gsf, optval,
1496 			   offsetof(struct group_filter, gf_slist_flex));
1497 	if (err)
1498 		return err;
1499 	if (gsf.gf_numsrc < num)
1500 		num = gsf.gf_numsrc;
1501 	gsf_size = GROUP_FILTER_SIZE(num);
1502 	if (copy_to_sockptr(optlen, &gsf_size, sizeof(int)) ||
1503 	    copy_to_sockptr(optval, &gsf, size0))
1504 		return -EFAULT;
1505 	return 0;
1506 }
1507 
1508 static int compat_ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1509 					sockptr_t optlen, int len)
1510 {
1511 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1512 	struct compat_group_filter gf32;
1513 	struct group_filter gf;
1514 	int num;
1515 	int err;
1516 
1517 	if (len < size0)
1518 		return -EINVAL;
1519 	if (copy_from_sockptr(&gf32, optval, size0))
1520 		return -EFAULT;
1521 
1522 	gf.gf_interface = gf32.gf_interface;
1523 	gf.gf_fmode = gf32.gf_fmode;
1524 	num = gf.gf_numsrc = gf32.gf_numsrc;
1525 	gf.gf_group = gf32.gf_group;
1526 
1527 	err = ip_mc_gsfget(sk, &gf, optval,
1528 			   offsetof(struct compat_group_filter, gf_slist_flex));
1529 	if (err)
1530 		return err;
1531 	if (gf.gf_numsrc < num)
1532 		num = gf.gf_numsrc;
1533 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1534 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1535 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1536 				   &gf.gf_fmode, sizeof(gf.gf_fmode)) ||
1537 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1538 				   &gf.gf_numsrc, sizeof(gf.gf_numsrc)))
1539 		return -EFAULT;
1540 	return 0;
1541 }
1542 
1543 int do_ip_getsockopt(struct sock *sk, int level, int optname,
1544 		     sockptr_t optval, sockptr_t optlen)
1545 {
1546 	struct inet_sock *inet = inet_sk(sk);
1547 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1548 	int val, err = 0;
1549 	int len;
1550 
1551 	if (level != SOL_IP)
1552 		return -EOPNOTSUPP;
1553 
1554 	if (ip_mroute_opt(optname))
1555 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1556 
1557 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1558 		return -EFAULT;
1559 	if (len < 0)
1560 		return -EINVAL;
1561 
1562 	if (needs_rtnl)
1563 		rtnl_lock();
1564 	sockopt_lock_sock(sk);
1565 
1566 	switch (optname) {
1567 	case IP_OPTIONS:
1568 	{
1569 		unsigned char optbuf[sizeof(struct ip_options)+40];
1570 		struct ip_options *opt = (struct ip_options *)optbuf;
1571 		struct ip_options_rcu *inet_opt;
1572 
1573 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1574 						     lockdep_sock_is_held(sk));
1575 		opt->optlen = 0;
1576 		if (inet_opt)
1577 			memcpy(optbuf, &inet_opt->opt,
1578 			       sizeof(struct ip_options) +
1579 			       inet_opt->opt.optlen);
1580 		sockopt_release_sock(sk);
1581 
1582 		if (opt->optlen == 0) {
1583 			len = 0;
1584 			return copy_to_sockptr(optlen, &len, sizeof(int));
1585 		}
1586 
1587 		ip_options_undo(opt);
1588 
1589 		len = min_t(unsigned int, len, opt->optlen);
1590 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1591 			return -EFAULT;
1592 		if (copy_to_sockptr(optval, opt->__data, len))
1593 			return -EFAULT;
1594 		return 0;
1595 	}
1596 	case IP_PKTINFO:
1597 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1598 		break;
1599 	case IP_RECVTTL:
1600 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1601 		break;
1602 	case IP_RECVTOS:
1603 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1604 		break;
1605 	case IP_RECVOPTS:
1606 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1607 		break;
1608 	case IP_RETOPTS:
1609 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1610 		break;
1611 	case IP_PASSSEC:
1612 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1613 		break;
1614 	case IP_RECVORIGDSTADDR:
1615 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1616 		break;
1617 	case IP_CHECKSUM:
1618 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1619 		break;
1620 	case IP_RECVFRAGSIZE:
1621 		val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1622 		break;
1623 	case IP_TOS:
1624 		val = inet->tos;
1625 		break;
1626 	case IP_TTL:
1627 	{
1628 		struct net *net = sock_net(sk);
1629 		val = (inet->uc_ttl == -1 ?
1630 		       READ_ONCE(net->ipv4.sysctl_ip_default_ttl) :
1631 		       inet->uc_ttl);
1632 		break;
1633 	}
1634 	case IP_HDRINCL:
1635 		val = inet->hdrincl;
1636 		break;
1637 	case IP_NODEFRAG:
1638 		val = inet->nodefrag;
1639 		break;
1640 	case IP_BIND_ADDRESS_NO_PORT:
1641 		val = inet->bind_address_no_port;
1642 		break;
1643 	case IP_MTU_DISCOVER:
1644 		val = inet->pmtudisc;
1645 		break;
1646 	case IP_MTU:
1647 	{
1648 		struct dst_entry *dst;
1649 		val = 0;
1650 		dst = sk_dst_get(sk);
1651 		if (dst) {
1652 			val = dst_mtu(dst);
1653 			dst_release(dst);
1654 		}
1655 		if (!val) {
1656 			sockopt_release_sock(sk);
1657 			return -ENOTCONN;
1658 		}
1659 		break;
1660 	}
1661 	case IP_RECVERR:
1662 		val = inet->recverr;
1663 		break;
1664 	case IP_RECVERR_RFC4884:
1665 		val = inet->recverr_rfc4884;
1666 		break;
1667 	case IP_MULTICAST_TTL:
1668 		val = inet->mc_ttl;
1669 		break;
1670 	case IP_MULTICAST_LOOP:
1671 		val = inet->mc_loop;
1672 		break;
1673 	case IP_UNICAST_IF:
1674 		val = (__force int)htonl((__u32) inet->uc_index);
1675 		break;
1676 	case IP_MULTICAST_IF:
1677 	{
1678 		struct in_addr addr;
1679 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1680 		addr.s_addr = inet->mc_addr;
1681 		sockopt_release_sock(sk);
1682 
1683 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1684 			return -EFAULT;
1685 		if (copy_to_sockptr(optval, &addr, len))
1686 			return -EFAULT;
1687 		return 0;
1688 	}
1689 	case IP_MSFILTER:
1690 	{
1691 		struct ip_msfilter msf;
1692 
1693 		if (len < IP_MSFILTER_SIZE(0)) {
1694 			err = -EINVAL;
1695 			goto out;
1696 		}
1697 		if (copy_from_sockptr(&msf, optval, IP_MSFILTER_SIZE(0))) {
1698 			err = -EFAULT;
1699 			goto out;
1700 		}
1701 		err = ip_mc_msfget(sk, &msf, optval, optlen);
1702 		goto out;
1703 	}
1704 	case MCAST_MSFILTER:
1705 		if (in_compat_syscall())
1706 			err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1707 							   len);
1708 		else
1709 			err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1710 		goto out;
1711 	case IP_MULTICAST_ALL:
1712 		val = inet->mc_all;
1713 		break;
1714 	case IP_PKTOPTIONS:
1715 	{
1716 		struct msghdr msg;
1717 
1718 		sockopt_release_sock(sk);
1719 
1720 		if (sk->sk_type != SOCK_STREAM)
1721 			return -ENOPROTOOPT;
1722 
1723 		if (optval.is_kernel) {
1724 			msg.msg_control_is_user = false;
1725 			msg.msg_control = optval.kernel;
1726 		} else {
1727 			msg.msg_control_is_user = true;
1728 			msg.msg_control_user = optval.user;
1729 		}
1730 		msg.msg_controllen = len;
1731 		msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1732 
1733 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1734 			struct in_pktinfo info;
1735 
1736 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1737 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1738 			info.ipi_ifindex = inet->mc_index;
1739 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1740 		}
1741 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1742 			int hlim = inet->mc_ttl;
1743 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1744 		}
1745 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1746 			int tos = inet->rcv_tos;
1747 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1748 		}
1749 		len -= msg.msg_controllen;
1750 		return copy_to_sockptr(optlen, &len, sizeof(int));
1751 	}
1752 	case IP_FREEBIND:
1753 		val = inet->freebind;
1754 		break;
1755 	case IP_TRANSPARENT:
1756 		val = inet->transparent;
1757 		break;
1758 	case IP_MINTTL:
1759 		val = inet->min_ttl;
1760 		break;
1761 	case IP_LOCAL_PORT_RANGE:
1762 		val = inet->local_port_range.hi << 16 | inet->local_port_range.lo;
1763 		break;
1764 	default:
1765 		sockopt_release_sock(sk);
1766 		return -ENOPROTOOPT;
1767 	}
1768 	sockopt_release_sock(sk);
1769 
1770 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1771 		unsigned char ucval = (unsigned char)val;
1772 		len = 1;
1773 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1774 			return -EFAULT;
1775 		if (copy_to_sockptr(optval, &ucval, 1))
1776 			return -EFAULT;
1777 	} else {
1778 		len = min_t(unsigned int, sizeof(int), len);
1779 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1780 			return -EFAULT;
1781 		if (copy_to_sockptr(optval, &val, len))
1782 			return -EFAULT;
1783 	}
1784 	return 0;
1785 
1786 out:
1787 	sockopt_release_sock(sk);
1788 	if (needs_rtnl)
1789 		rtnl_unlock();
1790 	return err;
1791 }
1792 
1793 int ip_getsockopt(struct sock *sk, int level,
1794 		  int optname, char __user *optval, int __user *optlen)
1795 {
1796 	int err;
1797 
1798 	err = do_ip_getsockopt(sk, level, optname,
1799 			       USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1800 
1801 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1802 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1803 	    optname < BPFILTER_IPT_GET_MAX)
1804 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1805 #endif
1806 #ifdef CONFIG_NETFILTER
1807 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1808 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1809 			!ip_mroute_opt(optname)) {
1810 		int len;
1811 
1812 		if (get_user(len, optlen))
1813 			return -EFAULT;
1814 
1815 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1816 		if (err >= 0)
1817 			err = put_user(len, optlen);
1818 		return err;
1819 	}
1820 #endif
1821 	return err;
1822 }
1823 EXPORT_SYMBOL(ip_getsockopt);
1824