xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision b868a02e)
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,
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 
437 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
438 {
439 	struct inet_sock *inet = inet_sk(sk);
440 	struct sock_exterr_skb *serr;
441 	struct iphdr *iph;
442 	struct sk_buff *skb;
443 
444 	if (!inet->recverr)
445 		return;
446 
447 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
448 	if (!skb)
449 		return;
450 
451 	skb_put(skb, sizeof(struct iphdr));
452 	skb_reset_network_header(skb);
453 	iph = ip_hdr(skb);
454 	iph->daddr = daddr;
455 
456 	serr = SKB_EXT_ERR(skb);
457 	serr->ee.ee_errno = err;
458 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
459 	serr->ee.ee_type = 0;
460 	serr->ee.ee_code = 0;
461 	serr->ee.ee_pad = 0;
462 	serr->ee.ee_info = info;
463 	serr->ee.ee_data = 0;
464 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
465 	serr->port = port;
466 
467 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
468 	skb_reset_transport_header(skb);
469 
470 	if (sock_queue_err_skb(sk, skb))
471 		kfree_skb(skb);
472 }
473 
474 /* For some errors we have valid addr_offset even with zero payload and
475  * zero port. Also, addr_offset should be supported if port is set.
476  */
477 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
478 {
479 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
480 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
481 }
482 
483 /* IPv4 supports cmsg on all imcp errors and some timestamps
484  *
485  * Timestamp code paths do not initialize the fields expected by cmsg:
486  * the PKTINFO fields in skb->cb[]. Fill those in here.
487  */
488 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
489 				       struct sk_buff *skb,
490 				       int ee_origin)
491 {
492 	struct in_pktinfo *info;
493 
494 	if (ee_origin == SO_EE_ORIGIN_ICMP)
495 		return true;
496 
497 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
498 		return false;
499 
500 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
501 	 * timestamp with egress dev. Not possible for packets without iif
502 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
503 	 */
504 	info = PKTINFO_SKB_CB(skb);
505 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
506 	    !info->ipi_ifindex)
507 		return false;
508 
509 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
510 	return true;
511 }
512 
513 /*
514  *	Handle MSG_ERRQUEUE
515  */
516 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
517 {
518 	struct sock_exterr_skb *serr;
519 	struct sk_buff *skb;
520 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
521 	struct {
522 		struct sock_extended_err ee;
523 		struct sockaddr_in	 offender;
524 	} errhdr;
525 	int err;
526 	int copied;
527 
528 	err = -EAGAIN;
529 	skb = sock_dequeue_err_skb(sk);
530 	if (!skb)
531 		goto out;
532 
533 	copied = skb->len;
534 	if (copied > len) {
535 		msg->msg_flags |= MSG_TRUNC;
536 		copied = len;
537 	}
538 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
539 	if (unlikely(err)) {
540 		kfree_skb(skb);
541 		return err;
542 	}
543 	sock_recv_timestamp(msg, sk, skb);
544 
545 	serr = SKB_EXT_ERR(skb);
546 
547 	if (sin && ipv4_datagram_support_addr(serr)) {
548 		sin->sin_family = AF_INET;
549 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
550 						   serr->addr_offset);
551 		sin->sin_port = serr->port;
552 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
553 		*addr_len = sizeof(*sin);
554 	}
555 
556 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
557 	sin = &errhdr.offender;
558 	memset(sin, 0, sizeof(*sin));
559 
560 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
561 		sin->sin_family = AF_INET;
562 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
563 		if (inet_sk(sk)->cmsg_flags)
564 			ip_cmsg_recv(msg, skb);
565 	}
566 
567 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
568 
569 	/* Now we could try to dump offended packet options */
570 
571 	msg->msg_flags |= MSG_ERRQUEUE;
572 	err = copied;
573 
574 	consume_skb(skb);
575 out:
576 	return err;
577 }
578 
579 void __ip_sock_set_tos(struct sock *sk, int val)
580 {
581 	if (sk->sk_type == SOCK_STREAM) {
582 		val &= ~INET_ECN_MASK;
583 		val |= inet_sk(sk)->tos & INET_ECN_MASK;
584 	}
585 	if (inet_sk(sk)->tos != val) {
586 		inet_sk(sk)->tos = val;
587 		sk->sk_priority = rt_tos2priority(val);
588 		sk_dst_reset(sk);
589 	}
590 }
591 
592 void ip_sock_set_tos(struct sock *sk, int val)
593 {
594 	lock_sock(sk);
595 	__ip_sock_set_tos(sk, val);
596 	release_sock(sk);
597 }
598 EXPORT_SYMBOL(ip_sock_set_tos);
599 
600 void ip_sock_set_freebind(struct sock *sk)
601 {
602 	lock_sock(sk);
603 	inet_sk(sk)->freebind = true;
604 	release_sock(sk);
605 }
606 EXPORT_SYMBOL(ip_sock_set_freebind);
607 
608 void ip_sock_set_recverr(struct sock *sk)
609 {
610 	lock_sock(sk);
611 	inet_sk(sk)->recverr = true;
612 	release_sock(sk);
613 }
614 EXPORT_SYMBOL(ip_sock_set_recverr);
615 
616 int ip_sock_set_mtu_discover(struct sock *sk, int val)
617 {
618 	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
619 		return -EINVAL;
620 	lock_sock(sk);
621 	inet_sk(sk)->pmtudisc = val;
622 	release_sock(sk);
623 	return 0;
624 }
625 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
626 
627 void ip_sock_set_pktinfo(struct sock *sk)
628 {
629 	lock_sock(sk);
630 	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
631 	release_sock(sk);
632 }
633 EXPORT_SYMBOL(ip_sock_set_pktinfo);
634 
635 /*
636  *	Socket option code for IP. This is the end of the line after any
637  *	TCP,UDP etc options on an IP socket.
638  */
639 static bool setsockopt_needs_rtnl(int optname)
640 {
641 	switch (optname) {
642 	case IP_ADD_MEMBERSHIP:
643 	case IP_ADD_SOURCE_MEMBERSHIP:
644 	case IP_BLOCK_SOURCE:
645 	case IP_DROP_MEMBERSHIP:
646 	case IP_DROP_SOURCE_MEMBERSHIP:
647 	case IP_MSFILTER:
648 	case IP_UNBLOCK_SOURCE:
649 	case MCAST_BLOCK_SOURCE:
650 	case MCAST_MSFILTER:
651 	case MCAST_JOIN_GROUP:
652 	case MCAST_JOIN_SOURCE_GROUP:
653 	case MCAST_LEAVE_GROUP:
654 	case MCAST_LEAVE_SOURCE_GROUP:
655 	case MCAST_UNBLOCK_SOURCE:
656 		return true;
657 	}
658 	return false;
659 }
660 
661 static int set_mcast_msfilter(struct sock *sk, int ifindex,
662 			      int numsrc, int fmode,
663 			      struct sockaddr_storage *group,
664 			      struct sockaddr_storage *list)
665 {
666 	struct ip_msfilter *msf;
667 	struct sockaddr_in *psin;
668 	int err, i;
669 
670 	msf = kmalloc(IP_MSFILTER_SIZE(numsrc), GFP_KERNEL);
671 	if (!msf)
672 		return -ENOBUFS;
673 
674 	psin = (struct sockaddr_in *)group;
675 	if (psin->sin_family != AF_INET)
676 		goto Eaddrnotavail;
677 	msf->imsf_multiaddr = psin->sin_addr.s_addr;
678 	msf->imsf_interface = 0;
679 	msf->imsf_fmode = fmode;
680 	msf->imsf_numsrc = numsrc;
681 	for (i = 0; i < numsrc; ++i) {
682 		psin = (struct sockaddr_in *)&list[i];
683 
684 		if (psin->sin_family != AF_INET)
685 			goto Eaddrnotavail;
686 		msf->imsf_slist_flex[i] = psin->sin_addr.s_addr;
687 	}
688 	err = ip_mc_msfilter(sk, msf, ifindex);
689 	kfree(msf);
690 	return err;
691 
692 Eaddrnotavail:
693 	kfree(msf);
694 	return -EADDRNOTAVAIL;
695 }
696 
697 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
698 		sockptr_t optval, int optlen)
699 {
700 	if (in_compat_syscall()) {
701 		struct compat_group_source_req gr32;
702 
703 		if (optlen != sizeof(gr32))
704 			return -EINVAL;
705 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
706 			return -EFAULT;
707 		greqs->gsr_interface = gr32.gsr_interface;
708 		greqs->gsr_group = gr32.gsr_group;
709 		greqs->gsr_source = gr32.gsr_source;
710 	} else {
711 		if (optlen != sizeof(*greqs))
712 			return -EINVAL;
713 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
714 			return -EFAULT;
715 	}
716 
717 	return 0;
718 }
719 
720 static int do_mcast_group_source(struct sock *sk, int optname,
721 		sockptr_t optval, int optlen)
722 {
723 	struct group_source_req greqs;
724 	struct ip_mreq_source mreqs;
725 	struct sockaddr_in *psin;
726 	int omode, add, err;
727 
728 	err = copy_group_source_from_sockptr(&greqs, optval, optlen);
729 	if (err)
730 		return err;
731 
732 	if (greqs.gsr_group.ss_family != AF_INET ||
733 	    greqs.gsr_source.ss_family != AF_INET)
734 		return -EADDRNOTAVAIL;
735 
736 	psin = (struct sockaddr_in *)&greqs.gsr_group;
737 	mreqs.imr_multiaddr = psin->sin_addr.s_addr;
738 	psin = (struct sockaddr_in *)&greqs.gsr_source;
739 	mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
740 	mreqs.imr_interface = 0; /* use index for mc_source */
741 
742 	if (optname == MCAST_BLOCK_SOURCE) {
743 		omode = MCAST_EXCLUDE;
744 		add = 1;
745 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
746 		omode = MCAST_EXCLUDE;
747 		add = 0;
748 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
749 		struct ip_mreqn mreq;
750 
751 		psin = (struct sockaddr_in *)&greqs.gsr_group;
752 		mreq.imr_multiaddr = psin->sin_addr;
753 		mreq.imr_address.s_addr = 0;
754 		mreq.imr_ifindex = greqs.gsr_interface;
755 		err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
756 		if (err && err != -EADDRINUSE)
757 			return err;
758 		greqs.gsr_interface = mreq.imr_ifindex;
759 		omode = MCAST_INCLUDE;
760 		add = 1;
761 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
762 		omode = MCAST_INCLUDE;
763 		add = 0;
764 	}
765 	return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
766 }
767 
768 static int ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval, int optlen)
769 {
770 	struct group_filter *gsf = NULL;
771 	int err;
772 
773 	if (optlen < GROUP_FILTER_SIZE(0))
774 		return -EINVAL;
775 	if (optlen > READ_ONCE(sysctl_optmem_max))
776 		return -ENOBUFS;
777 
778 	gsf = memdup_sockptr(optval, optlen);
779 	if (IS_ERR(gsf))
780 		return PTR_ERR(gsf);
781 
782 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
783 	err = -ENOBUFS;
784 	if (gsf->gf_numsrc >= 0x1ffffff ||
785 	    gsf->gf_numsrc > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
786 		goto out_free_gsf;
787 
788 	err = -EINVAL;
789 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
790 		goto out_free_gsf;
791 
792 	err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
793 				 gsf->gf_fmode, &gsf->gf_group,
794 				 gsf->gf_slist_flex);
795 out_free_gsf:
796 	kfree(gsf);
797 	return err;
798 }
799 
800 static int compat_ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
801 		int optlen)
802 {
803 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
804 	struct compat_group_filter *gf32;
805 	unsigned int n;
806 	void *p;
807 	int err;
808 
809 	if (optlen < size0)
810 		return -EINVAL;
811 	if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
812 		return -ENOBUFS;
813 
814 	p = kmalloc(optlen + 4, GFP_KERNEL);
815 	if (!p)
816 		return -ENOMEM;
817 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
818 
819 	err = -EFAULT;
820 	if (copy_from_sockptr(gf32, optval, optlen))
821 		goto out_free_gsf;
822 
823 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
824 	n = gf32->gf_numsrc;
825 	err = -ENOBUFS;
826 	if (n >= 0x1ffffff)
827 		goto out_free_gsf;
828 
829 	err = -EINVAL;
830 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
831 		goto out_free_gsf;
832 
833 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
834 	err = -ENOBUFS;
835 	if (n > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
836 		goto out_free_gsf;
837 	err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
838 				 &gf32->gf_group, gf32->gf_slist_flex);
839 out_free_gsf:
840 	kfree(p);
841 	return err;
842 }
843 
844 static int ip_mcast_join_leave(struct sock *sk, int optname,
845 		sockptr_t optval, int optlen)
846 {
847 	struct ip_mreqn mreq = { };
848 	struct sockaddr_in *psin;
849 	struct group_req greq;
850 
851 	if (optlen < sizeof(struct group_req))
852 		return -EINVAL;
853 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
854 		return -EFAULT;
855 
856 	psin = (struct sockaddr_in *)&greq.gr_group;
857 	if (psin->sin_family != AF_INET)
858 		return -EINVAL;
859 	mreq.imr_multiaddr = psin->sin_addr;
860 	mreq.imr_ifindex = greq.gr_interface;
861 	if (optname == MCAST_JOIN_GROUP)
862 		return ip_mc_join_group(sk, &mreq);
863 	return ip_mc_leave_group(sk, &mreq);
864 }
865 
866 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
867 		sockptr_t optval, int optlen)
868 {
869 	struct compat_group_req greq;
870 	struct ip_mreqn mreq = { };
871 	struct sockaddr_in *psin;
872 
873 	if (optlen < sizeof(struct compat_group_req))
874 		return -EINVAL;
875 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
876 		return -EFAULT;
877 
878 	psin = (struct sockaddr_in *)&greq.gr_group;
879 	if (psin->sin_family != AF_INET)
880 		return -EINVAL;
881 	mreq.imr_multiaddr = psin->sin_addr;
882 	mreq.imr_ifindex = greq.gr_interface;
883 
884 	if (optname == MCAST_JOIN_GROUP)
885 		return ip_mc_join_group(sk, &mreq);
886 	return ip_mc_leave_group(sk, &mreq);
887 }
888 
889 DEFINE_STATIC_KEY_FALSE(ip4_min_ttl);
890 
891 int do_ip_setsockopt(struct sock *sk, int level, int optname,
892 		     sockptr_t optval, unsigned int optlen)
893 {
894 	struct inet_sock *inet = inet_sk(sk);
895 	struct net *net = sock_net(sk);
896 	int val = 0, err;
897 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
898 
899 	switch (optname) {
900 	case IP_PKTINFO:
901 	case IP_RECVTTL:
902 	case IP_RECVOPTS:
903 	case IP_RECVTOS:
904 	case IP_RETOPTS:
905 	case IP_TOS:
906 	case IP_TTL:
907 	case IP_HDRINCL:
908 	case IP_MTU_DISCOVER:
909 	case IP_RECVERR:
910 	case IP_ROUTER_ALERT:
911 	case IP_FREEBIND:
912 	case IP_PASSSEC:
913 	case IP_TRANSPARENT:
914 	case IP_MINTTL:
915 	case IP_NODEFRAG:
916 	case IP_BIND_ADDRESS_NO_PORT:
917 	case IP_UNICAST_IF:
918 	case IP_MULTICAST_TTL:
919 	case IP_MULTICAST_ALL:
920 	case IP_MULTICAST_LOOP:
921 	case IP_RECVORIGDSTADDR:
922 	case IP_CHECKSUM:
923 	case IP_RECVFRAGSIZE:
924 	case IP_RECVERR_RFC4884:
925 		if (optlen >= sizeof(int)) {
926 			if (copy_from_sockptr(&val, optval, sizeof(val)))
927 				return -EFAULT;
928 		} else if (optlen >= sizeof(char)) {
929 			unsigned char ucval;
930 
931 			if (copy_from_sockptr(&ucval, optval, sizeof(ucval)))
932 				return -EFAULT;
933 			val = (int) ucval;
934 		}
935 	}
936 
937 	/* If optlen==0, it is equivalent to val == 0 */
938 
939 	if (optname == IP_ROUTER_ALERT)
940 		return ip_ra_control(sk, val ? 1 : 0, NULL);
941 	if (ip_mroute_opt(optname))
942 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
943 
944 	err = 0;
945 	if (needs_rtnl)
946 		rtnl_lock();
947 	sockopt_lock_sock(sk);
948 
949 	switch (optname) {
950 	case IP_OPTIONS:
951 	{
952 		struct ip_options_rcu *old, *opt = NULL;
953 
954 		if (optlen > 40)
955 			goto e_inval;
956 		err = ip_options_get(sock_net(sk), &opt, optval, optlen);
957 		if (err)
958 			break;
959 		old = rcu_dereference_protected(inet->inet_opt,
960 						lockdep_sock_is_held(sk));
961 		if (inet->is_icsk) {
962 			struct inet_connection_sock *icsk = inet_csk(sk);
963 #if IS_ENABLED(CONFIG_IPV6)
964 			if (sk->sk_family == PF_INET ||
965 			    (!((1 << sk->sk_state) &
966 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
967 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
968 #endif
969 				if (old)
970 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
971 				if (opt)
972 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
973 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
974 #if IS_ENABLED(CONFIG_IPV6)
975 			}
976 #endif
977 		}
978 		rcu_assign_pointer(inet->inet_opt, opt);
979 		if (old)
980 			kfree_rcu(old, rcu);
981 		break;
982 	}
983 	case IP_PKTINFO:
984 		if (val)
985 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
986 		else
987 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
988 		break;
989 	case IP_RECVTTL:
990 		if (val)
991 			inet->cmsg_flags |=  IP_CMSG_TTL;
992 		else
993 			inet->cmsg_flags &= ~IP_CMSG_TTL;
994 		break;
995 	case IP_RECVTOS:
996 		if (val)
997 			inet->cmsg_flags |=  IP_CMSG_TOS;
998 		else
999 			inet->cmsg_flags &= ~IP_CMSG_TOS;
1000 		break;
1001 	case IP_RECVOPTS:
1002 		if (val)
1003 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
1004 		else
1005 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
1006 		break;
1007 	case IP_RETOPTS:
1008 		if (val)
1009 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
1010 		else
1011 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
1012 		break;
1013 	case IP_PASSSEC:
1014 		if (val)
1015 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
1016 		else
1017 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
1018 		break;
1019 	case IP_RECVORIGDSTADDR:
1020 		if (val)
1021 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
1022 		else
1023 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
1024 		break;
1025 	case IP_CHECKSUM:
1026 		if (val) {
1027 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
1028 				inet_inc_convert_csum(sk);
1029 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
1030 			}
1031 		} else {
1032 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
1033 				inet_dec_convert_csum(sk);
1034 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
1035 			}
1036 		}
1037 		break;
1038 	case IP_RECVFRAGSIZE:
1039 		if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
1040 			goto e_inval;
1041 		if (val)
1042 			inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
1043 		else
1044 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
1045 		break;
1046 	case IP_TOS:	/* This sets both TOS and Precedence */
1047 		__ip_sock_set_tos(sk, val);
1048 		break;
1049 	case IP_TTL:
1050 		if (optlen < 1)
1051 			goto e_inval;
1052 		if (val != -1 && (val < 1 || val > 255))
1053 			goto e_inval;
1054 		inet->uc_ttl = val;
1055 		break;
1056 	case IP_HDRINCL:
1057 		if (sk->sk_type != SOCK_RAW) {
1058 			err = -ENOPROTOOPT;
1059 			break;
1060 		}
1061 		inet->hdrincl = val ? 1 : 0;
1062 		break;
1063 	case IP_NODEFRAG:
1064 		if (sk->sk_type != SOCK_RAW) {
1065 			err = -ENOPROTOOPT;
1066 			break;
1067 		}
1068 		inet->nodefrag = val ? 1 : 0;
1069 		break;
1070 	case IP_BIND_ADDRESS_NO_PORT:
1071 		inet->bind_address_no_port = val ? 1 : 0;
1072 		break;
1073 	case IP_MTU_DISCOVER:
1074 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
1075 			goto e_inval;
1076 		inet->pmtudisc = val;
1077 		break;
1078 	case IP_RECVERR:
1079 		inet->recverr = !!val;
1080 		if (!val)
1081 			skb_queue_purge(&sk->sk_error_queue);
1082 		break;
1083 	case IP_RECVERR_RFC4884:
1084 		if (val < 0 || val > 1)
1085 			goto e_inval;
1086 		inet->recverr_rfc4884 = !!val;
1087 		break;
1088 	case IP_MULTICAST_TTL:
1089 		if (sk->sk_type == SOCK_STREAM)
1090 			goto e_inval;
1091 		if (optlen < 1)
1092 			goto e_inval;
1093 		if (val == -1)
1094 			val = 1;
1095 		if (val < 0 || val > 255)
1096 			goto e_inval;
1097 		inet->mc_ttl = val;
1098 		break;
1099 	case IP_MULTICAST_LOOP:
1100 		if (optlen < 1)
1101 			goto e_inval;
1102 		inet->mc_loop = !!val;
1103 		break;
1104 	case IP_UNICAST_IF:
1105 	{
1106 		struct net_device *dev = NULL;
1107 		int ifindex;
1108 		int midx;
1109 
1110 		if (optlen != sizeof(int))
1111 			goto e_inval;
1112 
1113 		ifindex = (__force int)ntohl((__force __be32)val);
1114 		if (ifindex == 0) {
1115 			inet->uc_index = 0;
1116 			err = 0;
1117 			break;
1118 		}
1119 
1120 		dev = dev_get_by_index(sock_net(sk), ifindex);
1121 		err = -EADDRNOTAVAIL;
1122 		if (!dev)
1123 			break;
1124 
1125 		midx = l3mdev_master_ifindex(dev);
1126 		dev_put(dev);
1127 
1128 		err = -EINVAL;
1129 		if (sk->sk_bound_dev_if && midx != sk->sk_bound_dev_if)
1130 			break;
1131 
1132 		inet->uc_index = ifindex;
1133 		err = 0;
1134 		break;
1135 	}
1136 	case IP_MULTICAST_IF:
1137 	{
1138 		struct ip_mreqn mreq;
1139 		struct net_device *dev = NULL;
1140 		int midx;
1141 
1142 		if (sk->sk_type == SOCK_STREAM)
1143 			goto e_inval;
1144 		/*
1145 		 *	Check the arguments are allowable
1146 		 */
1147 
1148 		if (optlen < sizeof(struct in_addr))
1149 			goto e_inval;
1150 
1151 		err = -EFAULT;
1152 		if (optlen >= sizeof(struct ip_mreqn)) {
1153 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1154 				break;
1155 		} else {
1156 			memset(&mreq, 0, sizeof(mreq));
1157 			if (optlen >= sizeof(struct ip_mreq)) {
1158 				if (copy_from_sockptr(&mreq, optval,
1159 						      sizeof(struct ip_mreq)))
1160 					break;
1161 			} else if (optlen >= sizeof(struct in_addr)) {
1162 				if (copy_from_sockptr(&mreq.imr_address, optval,
1163 						      sizeof(struct in_addr)))
1164 					break;
1165 			}
1166 		}
1167 
1168 		if (!mreq.imr_ifindex) {
1169 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1170 				inet->mc_index = 0;
1171 				inet->mc_addr  = 0;
1172 				err = 0;
1173 				break;
1174 			}
1175 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1176 			if (dev)
1177 				mreq.imr_ifindex = dev->ifindex;
1178 		} else
1179 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1180 
1181 
1182 		err = -EADDRNOTAVAIL;
1183 		if (!dev)
1184 			break;
1185 
1186 		midx = l3mdev_master_ifindex(dev);
1187 
1188 		dev_put(dev);
1189 
1190 		err = -EINVAL;
1191 		if (sk->sk_bound_dev_if &&
1192 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
1193 		    midx != sk->sk_bound_dev_if)
1194 			break;
1195 
1196 		inet->mc_index = mreq.imr_ifindex;
1197 		inet->mc_addr  = mreq.imr_address.s_addr;
1198 		err = 0;
1199 		break;
1200 	}
1201 
1202 	case IP_ADD_MEMBERSHIP:
1203 	case IP_DROP_MEMBERSHIP:
1204 	{
1205 		struct ip_mreqn mreq;
1206 
1207 		err = -EPROTO;
1208 		if (inet_sk(sk)->is_icsk)
1209 			break;
1210 
1211 		if (optlen < sizeof(struct ip_mreq))
1212 			goto e_inval;
1213 		err = -EFAULT;
1214 		if (optlen >= sizeof(struct ip_mreqn)) {
1215 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1216 				break;
1217 		} else {
1218 			memset(&mreq, 0, sizeof(mreq));
1219 			if (copy_from_sockptr(&mreq, optval,
1220 					      sizeof(struct ip_mreq)))
1221 				break;
1222 		}
1223 
1224 		if (optname == IP_ADD_MEMBERSHIP)
1225 			err = ip_mc_join_group(sk, &mreq);
1226 		else
1227 			err = ip_mc_leave_group(sk, &mreq);
1228 		break;
1229 	}
1230 	case IP_MSFILTER:
1231 	{
1232 		struct ip_msfilter *msf;
1233 
1234 		if (optlen < IP_MSFILTER_SIZE(0))
1235 			goto e_inval;
1236 		if (optlen > READ_ONCE(sysctl_optmem_max)) {
1237 			err = -ENOBUFS;
1238 			break;
1239 		}
1240 		msf = memdup_sockptr(optval, optlen);
1241 		if (IS_ERR(msf)) {
1242 			err = PTR_ERR(msf);
1243 			break;
1244 		}
1245 		/* numsrc >= (1G-4) overflow in 32 bits */
1246 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
1247 		    msf->imsf_numsrc > READ_ONCE(net->ipv4.sysctl_igmp_max_msf)) {
1248 			kfree(msf);
1249 			err = -ENOBUFS;
1250 			break;
1251 		}
1252 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1253 			kfree(msf);
1254 			err = -EINVAL;
1255 			break;
1256 		}
1257 		err = ip_mc_msfilter(sk, msf, 0);
1258 		kfree(msf);
1259 		break;
1260 	}
1261 	case IP_BLOCK_SOURCE:
1262 	case IP_UNBLOCK_SOURCE:
1263 	case IP_ADD_SOURCE_MEMBERSHIP:
1264 	case IP_DROP_SOURCE_MEMBERSHIP:
1265 	{
1266 		struct ip_mreq_source mreqs;
1267 		int omode, add;
1268 
1269 		if (optlen != sizeof(struct ip_mreq_source))
1270 			goto e_inval;
1271 		if (copy_from_sockptr(&mreqs, optval, sizeof(mreqs))) {
1272 			err = -EFAULT;
1273 			break;
1274 		}
1275 		if (optname == IP_BLOCK_SOURCE) {
1276 			omode = MCAST_EXCLUDE;
1277 			add = 1;
1278 		} else if (optname == IP_UNBLOCK_SOURCE) {
1279 			omode = MCAST_EXCLUDE;
1280 			add = 0;
1281 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1282 			struct ip_mreqn mreq;
1283 
1284 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1285 			mreq.imr_address.s_addr = mreqs.imr_interface;
1286 			mreq.imr_ifindex = 0;
1287 			err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1288 			if (err && err != -EADDRINUSE)
1289 				break;
1290 			omode = MCAST_INCLUDE;
1291 			add = 1;
1292 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1293 			omode = MCAST_INCLUDE;
1294 			add = 0;
1295 		}
1296 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
1297 		break;
1298 	}
1299 	case MCAST_JOIN_GROUP:
1300 	case MCAST_LEAVE_GROUP:
1301 		if (in_compat_syscall())
1302 			err = compat_ip_mcast_join_leave(sk, optname, optval,
1303 							 optlen);
1304 		else
1305 			err = ip_mcast_join_leave(sk, optname, optval, optlen);
1306 		break;
1307 	case MCAST_JOIN_SOURCE_GROUP:
1308 	case MCAST_LEAVE_SOURCE_GROUP:
1309 	case MCAST_BLOCK_SOURCE:
1310 	case MCAST_UNBLOCK_SOURCE:
1311 		err = do_mcast_group_source(sk, optname, optval, optlen);
1312 		break;
1313 	case MCAST_MSFILTER:
1314 		if (in_compat_syscall())
1315 			err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1316 		else
1317 			err = ip_set_mcast_msfilter(sk, optval, optlen);
1318 		break;
1319 	case IP_MULTICAST_ALL:
1320 		if (optlen < 1)
1321 			goto e_inval;
1322 		if (val != 0 && val != 1)
1323 			goto e_inval;
1324 		inet->mc_all = val;
1325 		break;
1326 
1327 	case IP_FREEBIND:
1328 		if (optlen < 1)
1329 			goto e_inval;
1330 		inet->freebind = !!val;
1331 		break;
1332 
1333 	case IP_IPSEC_POLICY:
1334 	case IP_XFRM_POLICY:
1335 		err = -EPERM;
1336 		if (!sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1337 			break;
1338 		err = xfrm_user_policy(sk, optname, optval, optlen);
1339 		break;
1340 
1341 	case IP_TRANSPARENT:
1342 		if (!!val && !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1343 		    !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1344 			err = -EPERM;
1345 			break;
1346 		}
1347 		if (optlen < 1)
1348 			goto e_inval;
1349 		inet->transparent = !!val;
1350 		break;
1351 
1352 	case IP_MINTTL:
1353 		if (optlen < 1)
1354 			goto e_inval;
1355 		if (val < 0 || val > 255)
1356 			goto e_inval;
1357 
1358 		if (val)
1359 			static_branch_enable(&ip4_min_ttl);
1360 
1361 		/* tcp_v4_err() and tcp_v4_rcv() might read min_ttl
1362 		 * while we are changint it.
1363 		 */
1364 		WRITE_ONCE(inet->min_ttl, val);
1365 		break;
1366 
1367 	default:
1368 		err = -ENOPROTOOPT;
1369 		break;
1370 	}
1371 	sockopt_release_sock(sk);
1372 	if (needs_rtnl)
1373 		rtnl_unlock();
1374 	return err;
1375 
1376 e_inval:
1377 	sockopt_release_sock(sk);
1378 	if (needs_rtnl)
1379 		rtnl_unlock();
1380 	return -EINVAL;
1381 }
1382 
1383 /**
1384  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1385  * @sk: socket
1386  * @skb: buffer
1387  *
1388  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1389  * destination in skb->cb[] before dst drop.
1390  * This way, receiver doesn't make cache line misses to read rtable.
1391  */
1392 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1393 {
1394 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1395 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1396 		       ipv6_sk_rxinfo(sk);
1397 
1398 	if (prepare && skb_rtable(skb)) {
1399 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1400 		 * which has interface index (iif) as the first member of the
1401 		 * underlying inet{6}_skb_parm struct. This code then overlays
1402 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1403 		 * element so the iif is picked up from the prior IPCB. If iif
1404 		 * is the loopback interface, then return the sending interface
1405 		 * (e.g., process binds socket to eth0 for Tx which is
1406 		 * redirected to loopback in the rtable/dst).
1407 		 */
1408 		struct rtable *rt = skb_rtable(skb);
1409 		bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1410 
1411 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1412 			pktinfo->ipi_ifindex = inet_iif(skb);
1413 		else if (l3slave && rt && rt->rt_iif)
1414 			pktinfo->ipi_ifindex = rt->rt_iif;
1415 
1416 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1417 	} else {
1418 		pktinfo->ipi_ifindex = 0;
1419 		pktinfo->ipi_spec_dst.s_addr = 0;
1420 	}
1421 	skb_dst_drop(skb);
1422 }
1423 
1424 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1425 		unsigned int optlen)
1426 {
1427 	int err;
1428 
1429 	if (level != SOL_IP)
1430 		return -ENOPROTOOPT;
1431 
1432 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1433 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1434 	if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1435 	    optname < BPFILTER_IPT_SET_MAX)
1436 		err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1437 #endif
1438 #ifdef CONFIG_NETFILTER
1439 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1440 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1441 			optname != IP_IPSEC_POLICY &&
1442 			optname != IP_XFRM_POLICY &&
1443 			!ip_mroute_opt(optname))
1444 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1445 #endif
1446 	return err;
1447 }
1448 EXPORT_SYMBOL(ip_setsockopt);
1449 
1450 /*
1451  *	Get the options. Note for future reference. The GET of IP options gets
1452  *	the _received_ ones. The set sets the _sent_ ones.
1453  */
1454 
1455 static bool getsockopt_needs_rtnl(int optname)
1456 {
1457 	switch (optname) {
1458 	case IP_MSFILTER:
1459 	case MCAST_MSFILTER:
1460 		return true;
1461 	}
1462 	return false;
1463 }
1464 
1465 static int ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1466 				 sockptr_t optlen, int len)
1467 {
1468 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1469 	struct group_filter gsf;
1470 	int num, gsf_size;
1471 	int err;
1472 
1473 	if (len < size0)
1474 		return -EINVAL;
1475 	if (copy_from_sockptr(&gsf, optval, size0))
1476 		return -EFAULT;
1477 
1478 	num = gsf.gf_numsrc;
1479 	err = ip_mc_gsfget(sk, &gsf, optval,
1480 			   offsetof(struct group_filter, gf_slist_flex));
1481 	if (err)
1482 		return err;
1483 	if (gsf.gf_numsrc < num)
1484 		num = gsf.gf_numsrc;
1485 	gsf_size = GROUP_FILTER_SIZE(num);
1486 	if (copy_to_sockptr(optlen, &gsf_size, sizeof(int)) ||
1487 	    copy_to_sockptr(optval, &gsf, size0))
1488 		return -EFAULT;
1489 	return 0;
1490 }
1491 
1492 static int compat_ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1493 					sockptr_t optlen, int len)
1494 {
1495 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1496 	struct compat_group_filter gf32;
1497 	struct group_filter gf;
1498 	int num;
1499 	int err;
1500 
1501 	if (len < size0)
1502 		return -EINVAL;
1503 	if (copy_from_sockptr(&gf32, optval, size0))
1504 		return -EFAULT;
1505 
1506 	gf.gf_interface = gf32.gf_interface;
1507 	gf.gf_fmode = gf32.gf_fmode;
1508 	num = gf.gf_numsrc = gf32.gf_numsrc;
1509 	gf.gf_group = gf32.gf_group;
1510 
1511 	err = ip_mc_gsfget(sk, &gf, optval,
1512 			   offsetof(struct compat_group_filter, gf_slist_flex));
1513 	if (err)
1514 		return err;
1515 	if (gf.gf_numsrc < num)
1516 		num = gf.gf_numsrc;
1517 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1518 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1519 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1520 				   &gf.gf_fmode, sizeof(gf.gf_fmode)) ||
1521 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1522 				   &gf.gf_numsrc, sizeof(gf.gf_numsrc)))
1523 		return -EFAULT;
1524 	return 0;
1525 }
1526 
1527 int do_ip_getsockopt(struct sock *sk, int level, int optname,
1528 		     sockptr_t optval, sockptr_t optlen)
1529 {
1530 	struct inet_sock *inet = inet_sk(sk);
1531 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1532 	int val, err = 0;
1533 	int len;
1534 
1535 	if (level != SOL_IP)
1536 		return -EOPNOTSUPP;
1537 
1538 	if (ip_mroute_opt(optname))
1539 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1540 
1541 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1542 		return -EFAULT;
1543 	if (len < 0)
1544 		return -EINVAL;
1545 
1546 	if (needs_rtnl)
1547 		rtnl_lock();
1548 	sockopt_lock_sock(sk);
1549 
1550 	switch (optname) {
1551 	case IP_OPTIONS:
1552 	{
1553 		unsigned char optbuf[sizeof(struct ip_options)+40];
1554 		struct ip_options *opt = (struct ip_options *)optbuf;
1555 		struct ip_options_rcu *inet_opt;
1556 
1557 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1558 						     lockdep_sock_is_held(sk));
1559 		opt->optlen = 0;
1560 		if (inet_opt)
1561 			memcpy(optbuf, &inet_opt->opt,
1562 			       sizeof(struct ip_options) +
1563 			       inet_opt->opt.optlen);
1564 		sockopt_release_sock(sk);
1565 
1566 		if (opt->optlen == 0) {
1567 			len = 0;
1568 			return copy_to_sockptr(optlen, &len, sizeof(int));
1569 		}
1570 
1571 		ip_options_undo(opt);
1572 
1573 		len = min_t(unsigned int, len, opt->optlen);
1574 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1575 			return -EFAULT;
1576 		if (copy_to_sockptr(optval, opt->__data, len))
1577 			return -EFAULT;
1578 		return 0;
1579 	}
1580 	case IP_PKTINFO:
1581 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1582 		break;
1583 	case IP_RECVTTL:
1584 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1585 		break;
1586 	case IP_RECVTOS:
1587 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1588 		break;
1589 	case IP_RECVOPTS:
1590 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1591 		break;
1592 	case IP_RETOPTS:
1593 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1594 		break;
1595 	case IP_PASSSEC:
1596 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1597 		break;
1598 	case IP_RECVORIGDSTADDR:
1599 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1600 		break;
1601 	case IP_CHECKSUM:
1602 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1603 		break;
1604 	case IP_RECVFRAGSIZE:
1605 		val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1606 		break;
1607 	case IP_TOS:
1608 		val = inet->tos;
1609 		break;
1610 	case IP_TTL:
1611 	{
1612 		struct net *net = sock_net(sk);
1613 		val = (inet->uc_ttl == -1 ?
1614 		       READ_ONCE(net->ipv4.sysctl_ip_default_ttl) :
1615 		       inet->uc_ttl);
1616 		break;
1617 	}
1618 	case IP_HDRINCL:
1619 		val = inet->hdrincl;
1620 		break;
1621 	case IP_NODEFRAG:
1622 		val = inet->nodefrag;
1623 		break;
1624 	case IP_BIND_ADDRESS_NO_PORT:
1625 		val = inet->bind_address_no_port;
1626 		break;
1627 	case IP_MTU_DISCOVER:
1628 		val = inet->pmtudisc;
1629 		break;
1630 	case IP_MTU:
1631 	{
1632 		struct dst_entry *dst;
1633 		val = 0;
1634 		dst = sk_dst_get(sk);
1635 		if (dst) {
1636 			val = dst_mtu(dst);
1637 			dst_release(dst);
1638 		}
1639 		if (!val) {
1640 			sockopt_release_sock(sk);
1641 			return -ENOTCONN;
1642 		}
1643 		break;
1644 	}
1645 	case IP_RECVERR:
1646 		val = inet->recverr;
1647 		break;
1648 	case IP_RECVERR_RFC4884:
1649 		val = inet->recverr_rfc4884;
1650 		break;
1651 	case IP_MULTICAST_TTL:
1652 		val = inet->mc_ttl;
1653 		break;
1654 	case IP_MULTICAST_LOOP:
1655 		val = inet->mc_loop;
1656 		break;
1657 	case IP_UNICAST_IF:
1658 		val = (__force int)htonl((__u32) inet->uc_index);
1659 		break;
1660 	case IP_MULTICAST_IF:
1661 	{
1662 		struct in_addr addr;
1663 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1664 		addr.s_addr = inet->mc_addr;
1665 		sockopt_release_sock(sk);
1666 
1667 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1668 			return -EFAULT;
1669 		if (copy_to_sockptr(optval, &addr, len))
1670 			return -EFAULT;
1671 		return 0;
1672 	}
1673 	case IP_MSFILTER:
1674 	{
1675 		struct ip_msfilter msf;
1676 
1677 		if (len < IP_MSFILTER_SIZE(0)) {
1678 			err = -EINVAL;
1679 			goto out;
1680 		}
1681 		if (copy_from_sockptr(&msf, optval, IP_MSFILTER_SIZE(0))) {
1682 			err = -EFAULT;
1683 			goto out;
1684 		}
1685 		err = ip_mc_msfget(sk, &msf, optval, optlen);
1686 		goto out;
1687 	}
1688 	case MCAST_MSFILTER:
1689 		if (in_compat_syscall())
1690 			err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1691 							   len);
1692 		else
1693 			err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1694 		goto out;
1695 	case IP_MULTICAST_ALL:
1696 		val = inet->mc_all;
1697 		break;
1698 	case IP_PKTOPTIONS:
1699 	{
1700 		struct msghdr msg;
1701 
1702 		sockopt_release_sock(sk);
1703 
1704 		if (sk->sk_type != SOCK_STREAM)
1705 			return -ENOPROTOOPT;
1706 
1707 		if (optval.is_kernel) {
1708 			msg.msg_control_is_user = false;
1709 			msg.msg_control = optval.kernel;
1710 		} else {
1711 			msg.msg_control_is_user = true;
1712 			msg.msg_control_user = optval.user;
1713 		}
1714 		msg.msg_controllen = len;
1715 		msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1716 
1717 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1718 			struct in_pktinfo info;
1719 
1720 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1721 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1722 			info.ipi_ifindex = inet->mc_index;
1723 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1724 		}
1725 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1726 			int hlim = inet->mc_ttl;
1727 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1728 		}
1729 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1730 			int tos = inet->rcv_tos;
1731 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1732 		}
1733 		len -= msg.msg_controllen;
1734 		return copy_to_sockptr(optlen, &len, sizeof(int));
1735 	}
1736 	case IP_FREEBIND:
1737 		val = inet->freebind;
1738 		break;
1739 	case IP_TRANSPARENT:
1740 		val = inet->transparent;
1741 		break;
1742 	case IP_MINTTL:
1743 		val = inet->min_ttl;
1744 		break;
1745 	default:
1746 		sockopt_release_sock(sk);
1747 		return -ENOPROTOOPT;
1748 	}
1749 	sockopt_release_sock(sk);
1750 
1751 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1752 		unsigned char ucval = (unsigned char)val;
1753 		len = 1;
1754 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1755 			return -EFAULT;
1756 		if (copy_to_sockptr(optval, &ucval, 1))
1757 			return -EFAULT;
1758 	} else {
1759 		len = min_t(unsigned int, sizeof(int), len);
1760 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1761 			return -EFAULT;
1762 		if (copy_to_sockptr(optval, &val, len))
1763 			return -EFAULT;
1764 	}
1765 	return 0;
1766 
1767 out:
1768 	sockopt_release_sock(sk);
1769 	if (needs_rtnl)
1770 		rtnl_unlock();
1771 	return err;
1772 }
1773 
1774 int ip_getsockopt(struct sock *sk, int level,
1775 		  int optname, char __user *optval, int __user *optlen)
1776 {
1777 	int err;
1778 
1779 	err = do_ip_getsockopt(sk, level, optname,
1780 			       USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1781 
1782 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1783 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1784 	    optname < BPFILTER_IPT_GET_MAX)
1785 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1786 #endif
1787 #ifdef CONFIG_NETFILTER
1788 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1789 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1790 			!ip_mroute_opt(optname)) {
1791 		int len;
1792 
1793 		if (get_user(len, optlen))
1794 			return -EFAULT;
1795 
1796 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1797 		if (err >= 0)
1798 			err = put_user(len, optlen);
1799 		return err;
1800 	}
1801 #endif
1802 	return err;
1803 }
1804 EXPORT_SYMBOL(ip_getsockopt);
1805