xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 6db6b729)
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 	unsigned long flags = inet_cmsg_flags(inet_sk(sk));
175 
176 	if (!flags)
177 		return;
178 
179 	/* Ordered by supposed usage frequency */
180 	if (flags & IP_CMSG_PKTINFO) {
181 		ip_cmsg_recv_pktinfo(msg, skb);
182 
183 		flags &= ~IP_CMSG_PKTINFO;
184 		if (!flags)
185 			return;
186 	}
187 
188 	if (flags & IP_CMSG_TTL) {
189 		ip_cmsg_recv_ttl(msg, skb);
190 
191 		flags &= ~IP_CMSG_TTL;
192 		if (!flags)
193 			return;
194 	}
195 
196 	if (flags & IP_CMSG_TOS) {
197 		ip_cmsg_recv_tos(msg, skb);
198 
199 		flags &= ~IP_CMSG_TOS;
200 		if (!flags)
201 			return;
202 	}
203 
204 	if (flags & IP_CMSG_RECVOPTS) {
205 		ip_cmsg_recv_opts(msg, skb);
206 
207 		flags &= ~IP_CMSG_RECVOPTS;
208 		if (!flags)
209 			return;
210 	}
211 
212 	if (flags & IP_CMSG_RETOPTS) {
213 		ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
214 
215 		flags &= ~IP_CMSG_RETOPTS;
216 		if (!flags)
217 			return;
218 	}
219 
220 	if (flags & IP_CMSG_PASSSEC) {
221 		ip_cmsg_recv_security(msg, skb);
222 
223 		flags &= ~IP_CMSG_PASSSEC;
224 		if (!flags)
225 			return;
226 	}
227 
228 	if (flags & IP_CMSG_ORIGDSTADDR) {
229 		ip_cmsg_recv_dstaddr(msg, skb);
230 
231 		flags &= ~IP_CMSG_ORIGDSTADDR;
232 		if (!flags)
233 			return;
234 	}
235 
236 	if (flags & IP_CMSG_CHECKSUM)
237 		ip_cmsg_recv_checksum(msg, skb, tlen, offset);
238 
239 	if (flags & IP_CMSG_RECVFRAGSIZE)
240 		ip_cmsg_recv_fragsize(msg, skb);
241 }
242 EXPORT_SYMBOL(ip_cmsg_recv_offset);
243 
244 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
245 		 bool allow_ipv6)
246 {
247 	int err, val;
248 	struct cmsghdr *cmsg;
249 	struct net *net = sock_net(sk);
250 
251 	for_each_cmsghdr(cmsg, msg) {
252 		if (!CMSG_OK(msg, cmsg))
253 			return -EINVAL;
254 #if IS_ENABLED(CONFIG_IPV6)
255 		if (allow_ipv6 &&
256 		    cmsg->cmsg_level == SOL_IPV6 &&
257 		    cmsg->cmsg_type == IPV6_PKTINFO) {
258 			struct in6_pktinfo *src_info;
259 
260 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
261 				return -EINVAL;
262 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
263 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
264 				return -EINVAL;
265 			if (src_info->ipi6_ifindex)
266 				ipc->oif = src_info->ipi6_ifindex;
267 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
268 			continue;
269 		}
270 #endif
271 		if (cmsg->cmsg_level == SOL_SOCKET) {
272 			err = __sock_cmsg_send(sk, cmsg, &ipc->sockc);
273 			if (err)
274 				return err;
275 			continue;
276 		}
277 
278 		if (cmsg->cmsg_level != SOL_IP)
279 			continue;
280 		switch (cmsg->cmsg_type) {
281 		case IP_RETOPTS:
282 			err = cmsg->cmsg_len - sizeof(struct cmsghdr);
283 
284 			/* Our caller is responsible for freeing ipc->opt */
285 			err = ip_options_get(net, &ipc->opt,
286 					     KERNEL_SOCKPTR(CMSG_DATA(cmsg)),
287 					     err < 40 ? err : 40);
288 			if (err)
289 				return err;
290 			break;
291 		case IP_PKTINFO:
292 		{
293 			struct in_pktinfo *info;
294 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
295 				return -EINVAL;
296 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
297 			if (info->ipi_ifindex)
298 				ipc->oif = info->ipi_ifindex;
299 			ipc->addr = info->ipi_spec_dst.s_addr;
300 			break;
301 		}
302 		case IP_TTL:
303 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
304 				return -EINVAL;
305 			val = *(int *)CMSG_DATA(cmsg);
306 			if (val < 1 || val > 255)
307 				return -EINVAL;
308 			ipc->ttl = val;
309 			break;
310 		case IP_TOS:
311 			if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
312 				val = *(int *)CMSG_DATA(cmsg);
313 			else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
314 				val = *(u8 *)CMSG_DATA(cmsg);
315 			else
316 				return -EINVAL;
317 			if (val < 0 || val > 255)
318 				return -EINVAL;
319 			ipc->tos = val;
320 			ipc->priority = rt_tos2priority(ipc->tos);
321 			break;
322 		case IP_PROTOCOL:
323 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
324 				return -EINVAL;
325 			val = *(int *)CMSG_DATA(cmsg);
326 			if (val < 1 || val > 255)
327 				return -EINVAL;
328 			ipc->protocol = val;
329 			break;
330 		default:
331 			return -EINVAL;
332 		}
333 	}
334 	return 0;
335 }
336 
337 static void ip_ra_destroy_rcu(struct rcu_head *head)
338 {
339 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
340 
341 	sock_put(ra->saved_sk);
342 	kfree(ra);
343 }
344 
345 int ip_ra_control(struct sock *sk, unsigned char on,
346 		  void (*destructor)(struct sock *))
347 {
348 	struct ip_ra_chain *ra, *new_ra;
349 	struct ip_ra_chain __rcu **rap;
350 	struct net *net = sock_net(sk);
351 
352 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
353 		return -EINVAL;
354 
355 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
356 	if (on && !new_ra)
357 		return -ENOMEM;
358 
359 	mutex_lock(&net->ipv4.ra_mutex);
360 	for (rap = &net->ipv4.ra_chain;
361 	     (ra = rcu_dereference_protected(*rap,
362 			lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
363 	     rap = &ra->next) {
364 		if (ra->sk == sk) {
365 			if (on) {
366 				mutex_unlock(&net->ipv4.ra_mutex);
367 				kfree(new_ra);
368 				return -EADDRINUSE;
369 			}
370 			/* dont let ip_call_ra_chain() use sk again */
371 			ra->sk = NULL;
372 			RCU_INIT_POINTER(*rap, ra->next);
373 			mutex_unlock(&net->ipv4.ra_mutex);
374 
375 			if (ra->destructor)
376 				ra->destructor(sk);
377 			/*
378 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
379 			 * period. This guarantee ip_call_ra_chain() dont need
380 			 * to mess with socket refcounts.
381 			 */
382 			ra->saved_sk = sk;
383 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
384 			return 0;
385 		}
386 	}
387 	if (!new_ra) {
388 		mutex_unlock(&net->ipv4.ra_mutex);
389 		return -ENOBUFS;
390 	}
391 	new_ra->sk = sk;
392 	new_ra->destructor = destructor;
393 
394 	RCU_INIT_POINTER(new_ra->next, ra);
395 	rcu_assign_pointer(*rap, new_ra);
396 	sock_hold(sk);
397 	mutex_unlock(&net->ipv4.ra_mutex);
398 
399 	return 0;
400 }
401 
402 static void ipv4_icmp_error_rfc4884(const struct sk_buff *skb,
403 				    struct sock_ee_data_rfc4884 *out)
404 {
405 	switch (icmp_hdr(skb)->type) {
406 	case ICMP_DEST_UNREACH:
407 	case ICMP_TIME_EXCEEDED:
408 	case ICMP_PARAMETERPROB:
409 		ip_icmp_error_rfc4884(skb, out, sizeof(struct icmphdr),
410 				      icmp_hdr(skb)->un.reserved[1] * 4);
411 	}
412 }
413 
414 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
415 		   __be16 port, u32 info, u8 *payload)
416 {
417 	struct sock_exterr_skb *serr;
418 
419 	skb = skb_clone(skb, GFP_ATOMIC);
420 	if (!skb)
421 		return;
422 
423 	serr = SKB_EXT_ERR(skb);
424 	serr->ee.ee_errno = err;
425 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
426 	serr->ee.ee_type = icmp_hdr(skb)->type;
427 	serr->ee.ee_code = icmp_hdr(skb)->code;
428 	serr->ee.ee_pad = 0;
429 	serr->ee.ee_info = info;
430 	serr->ee.ee_data = 0;
431 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
432 				   skb_network_header(skb);
433 	serr->port = port;
434 
435 	if (skb_pull(skb, payload - skb->data)) {
436 		if (inet_test_bit(RECVERR_RFC4884, sk))
437 			ipv4_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
438 
439 		skb_reset_transport_header(skb);
440 		if (sock_queue_err_skb(sk, skb) == 0)
441 			return;
442 	}
443 	kfree_skb(skb);
444 }
445 EXPORT_SYMBOL_GPL(ip_icmp_error);
446 
447 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
448 {
449 	struct sock_exterr_skb *serr;
450 	struct iphdr *iph;
451 	struct sk_buff *skb;
452 
453 	if (!inet_test_bit(RECVERR, sk))
454 		return;
455 
456 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
457 	if (!skb)
458 		return;
459 
460 	skb_put(skb, sizeof(struct iphdr));
461 	skb_reset_network_header(skb);
462 	iph = ip_hdr(skb);
463 	iph->daddr = daddr;
464 
465 	serr = SKB_EXT_ERR(skb);
466 	serr->ee.ee_errno = err;
467 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
468 	serr->ee.ee_type = 0;
469 	serr->ee.ee_code = 0;
470 	serr->ee.ee_pad = 0;
471 	serr->ee.ee_info = info;
472 	serr->ee.ee_data = 0;
473 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
474 	serr->port = port;
475 
476 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
477 	skb_reset_transport_header(skb);
478 
479 	if (sock_queue_err_skb(sk, skb))
480 		kfree_skb(skb);
481 }
482 
483 /* For some errors we have valid addr_offset even with zero payload and
484  * zero port. Also, addr_offset should be supported if port is set.
485  */
486 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
487 {
488 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
489 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
490 }
491 
492 /* IPv4 supports cmsg on all imcp errors and some timestamps
493  *
494  * Timestamp code paths do not initialize the fields expected by cmsg:
495  * the PKTINFO fields in skb->cb[]. Fill those in here.
496  */
497 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
498 				       struct sk_buff *skb,
499 				       int ee_origin)
500 {
501 	struct in_pktinfo *info;
502 
503 	if (ee_origin == SO_EE_ORIGIN_ICMP)
504 		return true;
505 
506 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
507 		return false;
508 
509 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
510 	 * timestamp with egress dev. Not possible for packets without iif
511 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
512 	 */
513 	info = PKTINFO_SKB_CB(skb);
514 	if (!(READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_OPT_CMSG) ||
515 	    !info->ipi_ifindex)
516 		return false;
517 
518 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
519 	return true;
520 }
521 
522 /*
523  *	Handle MSG_ERRQUEUE
524  */
525 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
526 {
527 	struct sock_exterr_skb *serr;
528 	struct sk_buff *skb;
529 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
530 	struct {
531 		struct sock_extended_err ee;
532 		struct sockaddr_in	 offender;
533 	} errhdr;
534 	int err;
535 	int copied;
536 
537 	err = -EAGAIN;
538 	skb = sock_dequeue_err_skb(sk);
539 	if (!skb)
540 		goto out;
541 
542 	copied = skb->len;
543 	if (copied > len) {
544 		msg->msg_flags |= MSG_TRUNC;
545 		copied = len;
546 	}
547 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
548 	if (unlikely(err)) {
549 		kfree_skb(skb);
550 		return err;
551 	}
552 	sock_recv_timestamp(msg, sk, skb);
553 
554 	serr = SKB_EXT_ERR(skb);
555 
556 	if (sin && ipv4_datagram_support_addr(serr)) {
557 		sin->sin_family = AF_INET;
558 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
559 						   serr->addr_offset);
560 		sin->sin_port = serr->port;
561 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
562 		*addr_len = sizeof(*sin);
563 	}
564 
565 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
566 	sin = &errhdr.offender;
567 	memset(sin, 0, sizeof(*sin));
568 
569 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
570 		sin->sin_family = AF_INET;
571 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
572 		if (inet_cmsg_flags(inet_sk(sk)))
573 			ip_cmsg_recv(msg, skb);
574 	}
575 
576 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
577 
578 	/* Now we could try to dump offended packet options */
579 
580 	msg->msg_flags |= MSG_ERRQUEUE;
581 	err = copied;
582 
583 	consume_skb(skb);
584 out:
585 	return err;
586 }
587 
588 void __ip_sock_set_tos(struct sock *sk, int val)
589 {
590 	if (sk->sk_type == SOCK_STREAM) {
591 		val &= ~INET_ECN_MASK;
592 		val |= inet_sk(sk)->tos & INET_ECN_MASK;
593 	}
594 	if (inet_sk(sk)->tos != val) {
595 		inet_sk(sk)->tos = val;
596 		WRITE_ONCE(sk->sk_priority, rt_tos2priority(val));
597 		sk_dst_reset(sk);
598 	}
599 }
600 
601 void ip_sock_set_tos(struct sock *sk, int val)
602 {
603 	lock_sock(sk);
604 	__ip_sock_set_tos(sk, val);
605 	release_sock(sk);
606 }
607 EXPORT_SYMBOL(ip_sock_set_tos);
608 
609 void ip_sock_set_freebind(struct sock *sk)
610 {
611 	inet_set_bit(FREEBIND, sk);
612 }
613 EXPORT_SYMBOL(ip_sock_set_freebind);
614 
615 void ip_sock_set_recverr(struct sock *sk)
616 {
617 	inet_set_bit(RECVERR, sk);
618 }
619 EXPORT_SYMBOL(ip_sock_set_recverr);
620 
621 int ip_sock_set_mtu_discover(struct sock *sk, int val)
622 {
623 	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
624 		return -EINVAL;
625 	lock_sock(sk);
626 	inet_sk(sk)->pmtudisc = val;
627 	release_sock(sk);
628 	return 0;
629 }
630 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
631 
632 void ip_sock_set_pktinfo(struct sock *sk)
633 {
634 	inet_set_bit(PKTINFO, sk);
635 }
636 EXPORT_SYMBOL(ip_sock_set_pktinfo);
637 
638 /*
639  *	Socket option code for IP. This is the end of the line after any
640  *	TCP,UDP etc options on an IP socket.
641  */
642 static bool setsockopt_needs_rtnl(int optname)
643 {
644 	switch (optname) {
645 	case IP_ADD_MEMBERSHIP:
646 	case IP_ADD_SOURCE_MEMBERSHIP:
647 	case IP_BLOCK_SOURCE:
648 	case IP_DROP_MEMBERSHIP:
649 	case IP_DROP_SOURCE_MEMBERSHIP:
650 	case IP_MSFILTER:
651 	case IP_UNBLOCK_SOURCE:
652 	case MCAST_BLOCK_SOURCE:
653 	case MCAST_MSFILTER:
654 	case MCAST_JOIN_GROUP:
655 	case MCAST_JOIN_SOURCE_GROUP:
656 	case MCAST_LEAVE_GROUP:
657 	case MCAST_LEAVE_SOURCE_GROUP:
658 	case MCAST_UNBLOCK_SOURCE:
659 		return true;
660 	}
661 	return false;
662 }
663 
664 static int set_mcast_msfilter(struct sock *sk, int ifindex,
665 			      int numsrc, int fmode,
666 			      struct sockaddr_storage *group,
667 			      struct sockaddr_storage *list)
668 {
669 	struct ip_msfilter *msf;
670 	struct sockaddr_in *psin;
671 	int err, i;
672 
673 	msf = kmalloc(IP_MSFILTER_SIZE(numsrc), GFP_KERNEL);
674 	if (!msf)
675 		return -ENOBUFS;
676 
677 	psin = (struct sockaddr_in *)group;
678 	if (psin->sin_family != AF_INET)
679 		goto Eaddrnotavail;
680 	msf->imsf_multiaddr = psin->sin_addr.s_addr;
681 	msf->imsf_interface = 0;
682 	msf->imsf_fmode = fmode;
683 	msf->imsf_numsrc = numsrc;
684 	for (i = 0; i < numsrc; ++i) {
685 		psin = (struct sockaddr_in *)&list[i];
686 
687 		if (psin->sin_family != AF_INET)
688 			goto Eaddrnotavail;
689 		msf->imsf_slist_flex[i] = psin->sin_addr.s_addr;
690 	}
691 	err = ip_mc_msfilter(sk, msf, ifindex);
692 	kfree(msf);
693 	return err;
694 
695 Eaddrnotavail:
696 	kfree(msf);
697 	return -EADDRNOTAVAIL;
698 }
699 
700 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
701 		sockptr_t optval, int optlen)
702 {
703 	if (in_compat_syscall()) {
704 		struct compat_group_source_req gr32;
705 
706 		if (optlen != sizeof(gr32))
707 			return -EINVAL;
708 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
709 			return -EFAULT;
710 		greqs->gsr_interface = gr32.gsr_interface;
711 		greqs->gsr_group = gr32.gsr_group;
712 		greqs->gsr_source = gr32.gsr_source;
713 	} else {
714 		if (optlen != sizeof(*greqs))
715 			return -EINVAL;
716 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
717 			return -EFAULT;
718 	}
719 
720 	return 0;
721 }
722 
723 static int do_mcast_group_source(struct sock *sk, int optname,
724 		sockptr_t optval, int optlen)
725 {
726 	struct group_source_req greqs;
727 	struct ip_mreq_source mreqs;
728 	struct sockaddr_in *psin;
729 	int omode, add, err;
730 
731 	err = copy_group_source_from_sockptr(&greqs, optval, optlen);
732 	if (err)
733 		return err;
734 
735 	if (greqs.gsr_group.ss_family != AF_INET ||
736 	    greqs.gsr_source.ss_family != AF_INET)
737 		return -EADDRNOTAVAIL;
738 
739 	psin = (struct sockaddr_in *)&greqs.gsr_group;
740 	mreqs.imr_multiaddr = psin->sin_addr.s_addr;
741 	psin = (struct sockaddr_in *)&greqs.gsr_source;
742 	mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
743 	mreqs.imr_interface = 0; /* use index for mc_source */
744 
745 	if (optname == MCAST_BLOCK_SOURCE) {
746 		omode = MCAST_EXCLUDE;
747 		add = 1;
748 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
749 		omode = MCAST_EXCLUDE;
750 		add = 0;
751 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
752 		struct ip_mreqn mreq;
753 
754 		psin = (struct sockaddr_in *)&greqs.gsr_group;
755 		mreq.imr_multiaddr = psin->sin_addr;
756 		mreq.imr_address.s_addr = 0;
757 		mreq.imr_ifindex = greqs.gsr_interface;
758 		err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
759 		if (err && err != -EADDRINUSE)
760 			return err;
761 		greqs.gsr_interface = mreq.imr_ifindex;
762 		omode = MCAST_INCLUDE;
763 		add = 1;
764 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
765 		omode = MCAST_INCLUDE;
766 		add = 0;
767 	}
768 	return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
769 }
770 
771 static int ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval, int optlen)
772 {
773 	struct group_filter *gsf = NULL;
774 	int err;
775 
776 	if (optlen < GROUP_FILTER_SIZE(0))
777 		return -EINVAL;
778 	if (optlen > READ_ONCE(sysctl_optmem_max))
779 		return -ENOBUFS;
780 
781 	gsf = memdup_sockptr(optval, optlen);
782 	if (IS_ERR(gsf))
783 		return PTR_ERR(gsf);
784 
785 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
786 	err = -ENOBUFS;
787 	if (gsf->gf_numsrc >= 0x1ffffff ||
788 	    gsf->gf_numsrc > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
789 		goto out_free_gsf;
790 
791 	err = -EINVAL;
792 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
793 		goto out_free_gsf;
794 
795 	err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
796 				 gsf->gf_fmode, &gsf->gf_group,
797 				 gsf->gf_slist_flex);
798 out_free_gsf:
799 	kfree(gsf);
800 	return err;
801 }
802 
803 static int compat_ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
804 		int optlen)
805 {
806 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
807 	struct compat_group_filter *gf32;
808 	unsigned int n;
809 	void *p;
810 	int err;
811 
812 	if (optlen < size0)
813 		return -EINVAL;
814 	if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
815 		return -ENOBUFS;
816 
817 	p = kmalloc(optlen + 4, GFP_KERNEL);
818 	if (!p)
819 		return -ENOMEM;
820 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
821 
822 	err = -EFAULT;
823 	if (copy_from_sockptr(gf32, optval, optlen))
824 		goto out_free_gsf;
825 
826 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
827 	n = gf32->gf_numsrc;
828 	err = -ENOBUFS;
829 	if (n >= 0x1ffffff)
830 		goto out_free_gsf;
831 
832 	err = -EINVAL;
833 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
834 		goto out_free_gsf;
835 
836 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
837 	err = -ENOBUFS;
838 	if (n > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
839 		goto out_free_gsf;
840 	err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
841 				 &gf32->gf_group, gf32->gf_slist_flex);
842 out_free_gsf:
843 	kfree(p);
844 	return err;
845 }
846 
847 static int ip_mcast_join_leave(struct sock *sk, int optname,
848 		sockptr_t optval, int optlen)
849 {
850 	struct ip_mreqn mreq = { };
851 	struct sockaddr_in *psin;
852 	struct group_req greq;
853 
854 	if (optlen < sizeof(struct group_req))
855 		return -EINVAL;
856 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
857 		return -EFAULT;
858 
859 	psin = (struct sockaddr_in *)&greq.gr_group;
860 	if (psin->sin_family != AF_INET)
861 		return -EINVAL;
862 	mreq.imr_multiaddr = psin->sin_addr;
863 	mreq.imr_ifindex = greq.gr_interface;
864 	if (optname == MCAST_JOIN_GROUP)
865 		return ip_mc_join_group(sk, &mreq);
866 	return ip_mc_leave_group(sk, &mreq);
867 }
868 
869 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
870 		sockptr_t optval, int optlen)
871 {
872 	struct compat_group_req greq;
873 	struct ip_mreqn mreq = { };
874 	struct sockaddr_in *psin;
875 
876 	if (optlen < sizeof(struct compat_group_req))
877 		return -EINVAL;
878 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
879 		return -EFAULT;
880 
881 	psin = (struct sockaddr_in *)&greq.gr_group;
882 	if (psin->sin_family != AF_INET)
883 		return -EINVAL;
884 	mreq.imr_multiaddr = psin->sin_addr;
885 	mreq.imr_ifindex = greq.gr_interface;
886 
887 	if (optname == MCAST_JOIN_GROUP)
888 		return ip_mc_join_group(sk, &mreq);
889 	return ip_mc_leave_group(sk, &mreq);
890 }
891 
892 DEFINE_STATIC_KEY_FALSE(ip4_min_ttl);
893 
894 int do_ip_setsockopt(struct sock *sk, int level, int optname,
895 		     sockptr_t optval, unsigned int optlen)
896 {
897 	struct inet_sock *inet = inet_sk(sk);
898 	struct net *net = sock_net(sk);
899 	int val = 0, err;
900 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
901 
902 	switch (optname) {
903 	case IP_PKTINFO:
904 	case IP_RECVTTL:
905 	case IP_RECVOPTS:
906 	case IP_RECVTOS:
907 	case IP_RETOPTS:
908 	case IP_TOS:
909 	case IP_TTL:
910 	case IP_HDRINCL:
911 	case IP_MTU_DISCOVER:
912 	case IP_RECVERR:
913 	case IP_ROUTER_ALERT:
914 	case IP_FREEBIND:
915 	case IP_PASSSEC:
916 	case IP_TRANSPARENT:
917 	case IP_MINTTL:
918 	case IP_NODEFRAG:
919 	case IP_BIND_ADDRESS_NO_PORT:
920 	case IP_UNICAST_IF:
921 	case IP_MULTICAST_TTL:
922 	case IP_MULTICAST_ALL:
923 	case IP_MULTICAST_LOOP:
924 	case IP_RECVORIGDSTADDR:
925 	case IP_CHECKSUM:
926 	case IP_RECVFRAGSIZE:
927 	case IP_RECVERR_RFC4884:
928 	case IP_LOCAL_PORT_RANGE:
929 		if (optlen >= sizeof(int)) {
930 			if (copy_from_sockptr(&val, optval, sizeof(val)))
931 				return -EFAULT;
932 		} else if (optlen >= sizeof(char)) {
933 			unsigned char ucval;
934 
935 			if (copy_from_sockptr(&ucval, optval, sizeof(ucval)))
936 				return -EFAULT;
937 			val = (int) ucval;
938 		}
939 	}
940 
941 	/* If optlen==0, it is equivalent to val == 0 */
942 
943 	if (optname == IP_ROUTER_ALERT)
944 		return ip_ra_control(sk, val ? 1 : 0, NULL);
945 	if (ip_mroute_opt(optname))
946 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
947 
948 	/* Handle options that can be set without locking the socket. */
949 	switch (optname) {
950 	case IP_PKTINFO:
951 		inet_assign_bit(PKTINFO, sk, val);
952 		return 0;
953 	case IP_RECVTTL:
954 		inet_assign_bit(TTL, sk, val);
955 		return 0;
956 	case IP_RECVTOS:
957 		inet_assign_bit(TOS, sk, val);
958 		return 0;
959 	case IP_RECVOPTS:
960 		inet_assign_bit(RECVOPTS, sk, val);
961 		return 0;
962 	case IP_RETOPTS:
963 		inet_assign_bit(RETOPTS, sk, val);
964 		return 0;
965 	case IP_PASSSEC:
966 		inet_assign_bit(PASSSEC, sk, val);
967 		return 0;
968 	case IP_RECVORIGDSTADDR:
969 		inet_assign_bit(ORIGDSTADDR, sk, val);
970 		return 0;
971 	case IP_RECVFRAGSIZE:
972 		if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
973 			return -EINVAL;
974 		inet_assign_bit(RECVFRAGSIZE, sk, val);
975 		return 0;
976 	case IP_RECVERR:
977 		inet_assign_bit(RECVERR, sk, val);
978 		if (!val)
979 			skb_errqueue_purge(&sk->sk_error_queue);
980 		return 0;
981 	case IP_RECVERR_RFC4884:
982 		if (val < 0 || val > 1)
983 			return -EINVAL;
984 		inet_assign_bit(RECVERR_RFC4884, sk, val);
985 		return 0;
986 	case IP_FREEBIND:
987 		if (optlen < 1)
988 			return -EINVAL;
989 		inet_assign_bit(FREEBIND, sk, val);
990 		return 0;
991 	case IP_HDRINCL:
992 		if (sk->sk_type != SOCK_RAW)
993 			return -ENOPROTOOPT;
994 		inet_assign_bit(HDRINCL, sk, val);
995 		return 0;
996 	case IP_MULTICAST_LOOP:
997 		if (optlen < 1)
998 			return -EINVAL;
999 		inet_assign_bit(MC_LOOP, sk, val);
1000 		return 0;
1001 	case IP_MULTICAST_ALL:
1002 		if (optlen < 1)
1003 			return -EINVAL;
1004 		if (val != 0 && val != 1)
1005 			return -EINVAL;
1006 		inet_assign_bit(MC_ALL, sk, val);
1007 		return 0;
1008 	case IP_TRANSPARENT:
1009 		if (!!val && !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1010 		    !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1011 			return -EPERM;
1012 		if (optlen < 1)
1013 			return -EINVAL;
1014 		inet_assign_bit(TRANSPARENT, sk, val);
1015 		return 0;
1016 	case IP_NODEFRAG:
1017 		if (sk->sk_type != SOCK_RAW)
1018 			return -ENOPROTOOPT;
1019 		inet_assign_bit(NODEFRAG, sk, val);
1020 		return 0;
1021 	case IP_BIND_ADDRESS_NO_PORT:
1022 		inet_assign_bit(BIND_ADDRESS_NO_PORT, sk, val);
1023 		return 0;
1024 	case IP_TTL:
1025 		if (optlen < 1)
1026 			return -EINVAL;
1027 		if (val != -1 && (val < 1 || val > 255))
1028 			return -EINVAL;
1029 		WRITE_ONCE(inet->uc_ttl, val);
1030 		return 0;
1031 	case IP_MINTTL:
1032 		if (optlen < 1)
1033 			return -EINVAL;
1034 		if (val < 0 || val > 255)
1035 			return -EINVAL;
1036 
1037 		if (val)
1038 			static_branch_enable(&ip4_min_ttl);
1039 
1040 		WRITE_ONCE(inet->min_ttl, val);
1041 		return 0;
1042 	}
1043 
1044 	err = 0;
1045 	if (needs_rtnl)
1046 		rtnl_lock();
1047 	sockopt_lock_sock(sk);
1048 
1049 	switch (optname) {
1050 	case IP_OPTIONS:
1051 	{
1052 		struct ip_options_rcu *old, *opt = NULL;
1053 
1054 		if (optlen > 40)
1055 			goto e_inval;
1056 		err = ip_options_get(sock_net(sk), &opt, optval, optlen);
1057 		if (err)
1058 			break;
1059 		old = rcu_dereference_protected(inet->inet_opt,
1060 						lockdep_sock_is_held(sk));
1061 		if (inet_test_bit(IS_ICSK, sk)) {
1062 			struct inet_connection_sock *icsk = inet_csk(sk);
1063 #if IS_ENABLED(CONFIG_IPV6)
1064 			if (sk->sk_family == PF_INET ||
1065 			    (!((1 << sk->sk_state) &
1066 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
1067 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
1068 #endif
1069 				if (old)
1070 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
1071 				if (opt)
1072 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
1073 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
1074 #if IS_ENABLED(CONFIG_IPV6)
1075 			}
1076 #endif
1077 		}
1078 		rcu_assign_pointer(inet->inet_opt, opt);
1079 		if (old)
1080 			kfree_rcu(old, rcu);
1081 		break;
1082 	}
1083 	case IP_CHECKSUM:
1084 		if (val) {
1085 			if (!(inet_test_bit(CHECKSUM, sk))) {
1086 				inet_inc_convert_csum(sk);
1087 				inet_set_bit(CHECKSUM, sk);
1088 			}
1089 		} else {
1090 			if (inet_test_bit(CHECKSUM, sk)) {
1091 				inet_dec_convert_csum(sk);
1092 				inet_clear_bit(CHECKSUM, sk);
1093 			}
1094 		}
1095 		break;
1096 	case IP_TOS:	/* This sets both TOS and Precedence */
1097 		__ip_sock_set_tos(sk, val);
1098 		break;
1099 	case IP_MTU_DISCOVER:
1100 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
1101 			goto e_inval;
1102 		inet->pmtudisc = val;
1103 		break;
1104 	case IP_MULTICAST_TTL:
1105 		if (sk->sk_type == SOCK_STREAM)
1106 			goto e_inval;
1107 		if (optlen < 1)
1108 			goto e_inval;
1109 		if (val == -1)
1110 			val = 1;
1111 		if (val < 0 || val > 255)
1112 			goto e_inval;
1113 		inet->mc_ttl = val;
1114 		break;
1115 	case IP_UNICAST_IF:
1116 	{
1117 		struct net_device *dev = NULL;
1118 		int ifindex;
1119 		int midx;
1120 
1121 		if (optlen != sizeof(int))
1122 			goto e_inval;
1123 
1124 		ifindex = (__force int)ntohl((__force __be32)val);
1125 		if (ifindex == 0) {
1126 			inet->uc_index = 0;
1127 			err = 0;
1128 			break;
1129 		}
1130 
1131 		dev = dev_get_by_index(sock_net(sk), ifindex);
1132 		err = -EADDRNOTAVAIL;
1133 		if (!dev)
1134 			break;
1135 
1136 		midx = l3mdev_master_ifindex(dev);
1137 		dev_put(dev);
1138 
1139 		err = -EINVAL;
1140 		if (sk->sk_bound_dev_if && midx != sk->sk_bound_dev_if)
1141 			break;
1142 
1143 		inet->uc_index = ifindex;
1144 		err = 0;
1145 		break;
1146 	}
1147 	case IP_MULTICAST_IF:
1148 	{
1149 		struct ip_mreqn mreq;
1150 		struct net_device *dev = NULL;
1151 		int midx;
1152 
1153 		if (sk->sk_type == SOCK_STREAM)
1154 			goto e_inval;
1155 		/*
1156 		 *	Check the arguments are allowable
1157 		 */
1158 
1159 		if (optlen < sizeof(struct in_addr))
1160 			goto e_inval;
1161 
1162 		err = -EFAULT;
1163 		if (optlen >= sizeof(struct ip_mreqn)) {
1164 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1165 				break;
1166 		} else {
1167 			memset(&mreq, 0, sizeof(mreq));
1168 			if (optlen >= sizeof(struct ip_mreq)) {
1169 				if (copy_from_sockptr(&mreq, optval,
1170 						      sizeof(struct ip_mreq)))
1171 					break;
1172 			} else if (optlen >= sizeof(struct in_addr)) {
1173 				if (copy_from_sockptr(&mreq.imr_address, optval,
1174 						      sizeof(struct in_addr)))
1175 					break;
1176 			}
1177 		}
1178 
1179 		if (!mreq.imr_ifindex) {
1180 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1181 				inet->mc_index = 0;
1182 				inet->mc_addr  = 0;
1183 				err = 0;
1184 				break;
1185 			}
1186 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1187 			if (dev)
1188 				mreq.imr_ifindex = dev->ifindex;
1189 		} else
1190 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1191 
1192 
1193 		err = -EADDRNOTAVAIL;
1194 		if (!dev)
1195 			break;
1196 
1197 		midx = l3mdev_master_ifindex(dev);
1198 
1199 		dev_put(dev);
1200 
1201 		err = -EINVAL;
1202 		if (sk->sk_bound_dev_if &&
1203 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
1204 		    midx != sk->sk_bound_dev_if)
1205 			break;
1206 
1207 		inet->mc_index = mreq.imr_ifindex;
1208 		inet->mc_addr  = mreq.imr_address.s_addr;
1209 		err = 0;
1210 		break;
1211 	}
1212 
1213 	case IP_ADD_MEMBERSHIP:
1214 	case IP_DROP_MEMBERSHIP:
1215 	{
1216 		struct ip_mreqn mreq;
1217 
1218 		err = -EPROTO;
1219 		if (inet_test_bit(IS_ICSK, sk))
1220 			break;
1221 
1222 		if (optlen < sizeof(struct ip_mreq))
1223 			goto e_inval;
1224 		err = -EFAULT;
1225 		if (optlen >= sizeof(struct ip_mreqn)) {
1226 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1227 				break;
1228 		} else {
1229 			memset(&mreq, 0, sizeof(mreq));
1230 			if (copy_from_sockptr(&mreq, optval,
1231 					      sizeof(struct ip_mreq)))
1232 				break;
1233 		}
1234 
1235 		if (optname == IP_ADD_MEMBERSHIP)
1236 			err = ip_mc_join_group(sk, &mreq);
1237 		else
1238 			err = ip_mc_leave_group(sk, &mreq);
1239 		break;
1240 	}
1241 	case IP_MSFILTER:
1242 	{
1243 		struct ip_msfilter *msf;
1244 
1245 		if (optlen < IP_MSFILTER_SIZE(0))
1246 			goto e_inval;
1247 		if (optlen > READ_ONCE(sysctl_optmem_max)) {
1248 			err = -ENOBUFS;
1249 			break;
1250 		}
1251 		msf = memdup_sockptr(optval, optlen);
1252 		if (IS_ERR(msf)) {
1253 			err = PTR_ERR(msf);
1254 			break;
1255 		}
1256 		/* numsrc >= (1G-4) overflow in 32 bits */
1257 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
1258 		    msf->imsf_numsrc > READ_ONCE(net->ipv4.sysctl_igmp_max_msf)) {
1259 			kfree(msf);
1260 			err = -ENOBUFS;
1261 			break;
1262 		}
1263 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1264 			kfree(msf);
1265 			err = -EINVAL;
1266 			break;
1267 		}
1268 		err = ip_mc_msfilter(sk, msf, 0);
1269 		kfree(msf);
1270 		break;
1271 	}
1272 	case IP_BLOCK_SOURCE:
1273 	case IP_UNBLOCK_SOURCE:
1274 	case IP_ADD_SOURCE_MEMBERSHIP:
1275 	case IP_DROP_SOURCE_MEMBERSHIP:
1276 	{
1277 		struct ip_mreq_source mreqs;
1278 		int omode, add;
1279 
1280 		if (optlen != sizeof(struct ip_mreq_source))
1281 			goto e_inval;
1282 		if (copy_from_sockptr(&mreqs, optval, sizeof(mreqs))) {
1283 			err = -EFAULT;
1284 			break;
1285 		}
1286 		if (optname == IP_BLOCK_SOURCE) {
1287 			omode = MCAST_EXCLUDE;
1288 			add = 1;
1289 		} else if (optname == IP_UNBLOCK_SOURCE) {
1290 			omode = MCAST_EXCLUDE;
1291 			add = 0;
1292 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1293 			struct ip_mreqn mreq;
1294 
1295 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1296 			mreq.imr_address.s_addr = mreqs.imr_interface;
1297 			mreq.imr_ifindex = 0;
1298 			err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1299 			if (err && err != -EADDRINUSE)
1300 				break;
1301 			omode = MCAST_INCLUDE;
1302 			add = 1;
1303 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1304 			omode = MCAST_INCLUDE;
1305 			add = 0;
1306 		}
1307 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
1308 		break;
1309 	}
1310 	case MCAST_JOIN_GROUP:
1311 	case MCAST_LEAVE_GROUP:
1312 		if (in_compat_syscall())
1313 			err = compat_ip_mcast_join_leave(sk, optname, optval,
1314 							 optlen);
1315 		else
1316 			err = ip_mcast_join_leave(sk, optname, optval, optlen);
1317 		break;
1318 	case MCAST_JOIN_SOURCE_GROUP:
1319 	case MCAST_LEAVE_SOURCE_GROUP:
1320 	case MCAST_BLOCK_SOURCE:
1321 	case MCAST_UNBLOCK_SOURCE:
1322 		err = do_mcast_group_source(sk, optname, optval, optlen);
1323 		break;
1324 	case MCAST_MSFILTER:
1325 		if (in_compat_syscall())
1326 			err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1327 		else
1328 			err = ip_set_mcast_msfilter(sk, optval, optlen);
1329 		break;
1330 	case IP_IPSEC_POLICY:
1331 	case IP_XFRM_POLICY:
1332 		err = -EPERM;
1333 		if (!sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1334 			break;
1335 		err = xfrm_user_policy(sk, optname, optval, optlen);
1336 		break;
1337 
1338 	case IP_LOCAL_PORT_RANGE:
1339 	{
1340 		const __u16 lo = val;
1341 		const __u16 hi = val >> 16;
1342 
1343 		if (optlen != sizeof(__u32))
1344 			goto e_inval;
1345 		if (lo != 0 && hi != 0 && lo > hi)
1346 			goto e_inval;
1347 
1348 		inet->local_port_range.lo = lo;
1349 		inet->local_port_range.hi = hi;
1350 		break;
1351 	}
1352 	default:
1353 		err = -ENOPROTOOPT;
1354 		break;
1355 	}
1356 	sockopt_release_sock(sk);
1357 	if (needs_rtnl)
1358 		rtnl_unlock();
1359 	return err;
1360 
1361 e_inval:
1362 	sockopt_release_sock(sk);
1363 	if (needs_rtnl)
1364 		rtnl_unlock();
1365 	return -EINVAL;
1366 }
1367 
1368 /**
1369  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1370  * @sk: socket
1371  * @skb: buffer
1372  *
1373  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1374  * destination in skb->cb[] before dst drop.
1375  * This way, receiver doesn't make cache line misses to read rtable.
1376  */
1377 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1378 {
1379 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1380 	bool prepare = inet_test_bit(PKTINFO, sk) ||
1381 		       ipv6_sk_rxinfo(sk);
1382 
1383 	if (prepare && skb_rtable(skb)) {
1384 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1385 		 * which has interface index (iif) as the first member of the
1386 		 * underlying inet{6}_skb_parm struct. This code then overlays
1387 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1388 		 * element so the iif is picked up from the prior IPCB. If iif
1389 		 * is the loopback interface, then return the sending interface
1390 		 * (e.g., process binds socket to eth0 for Tx which is
1391 		 * redirected to loopback in the rtable/dst).
1392 		 */
1393 		struct rtable *rt = skb_rtable(skb);
1394 		bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1395 
1396 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1397 			pktinfo->ipi_ifindex = inet_iif(skb);
1398 		else if (l3slave && rt && rt->rt_iif)
1399 			pktinfo->ipi_ifindex = rt->rt_iif;
1400 
1401 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1402 	} else {
1403 		pktinfo->ipi_ifindex = 0;
1404 		pktinfo->ipi_spec_dst.s_addr = 0;
1405 	}
1406 	skb_dst_drop(skb);
1407 }
1408 
1409 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1410 		unsigned int optlen)
1411 {
1412 	int err;
1413 
1414 	if (level != SOL_IP)
1415 		return -ENOPROTOOPT;
1416 
1417 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1418 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1419 	if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1420 	    optname < BPFILTER_IPT_SET_MAX)
1421 		err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1422 #endif
1423 #ifdef CONFIG_NETFILTER
1424 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1425 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1426 			optname != IP_IPSEC_POLICY &&
1427 			optname != IP_XFRM_POLICY &&
1428 			!ip_mroute_opt(optname))
1429 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1430 #endif
1431 	return err;
1432 }
1433 EXPORT_SYMBOL(ip_setsockopt);
1434 
1435 /*
1436  *	Get the options. Note for future reference. The GET of IP options gets
1437  *	the _received_ ones. The set sets the _sent_ ones.
1438  */
1439 
1440 static bool getsockopt_needs_rtnl(int optname)
1441 {
1442 	switch (optname) {
1443 	case IP_MSFILTER:
1444 	case MCAST_MSFILTER:
1445 		return true;
1446 	}
1447 	return false;
1448 }
1449 
1450 static int ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1451 				 sockptr_t optlen, int len)
1452 {
1453 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1454 	struct group_filter gsf;
1455 	int num, gsf_size;
1456 	int err;
1457 
1458 	if (len < size0)
1459 		return -EINVAL;
1460 	if (copy_from_sockptr(&gsf, optval, size0))
1461 		return -EFAULT;
1462 
1463 	num = gsf.gf_numsrc;
1464 	err = ip_mc_gsfget(sk, &gsf, optval,
1465 			   offsetof(struct group_filter, gf_slist_flex));
1466 	if (err)
1467 		return err;
1468 	if (gsf.gf_numsrc < num)
1469 		num = gsf.gf_numsrc;
1470 	gsf_size = GROUP_FILTER_SIZE(num);
1471 	if (copy_to_sockptr(optlen, &gsf_size, sizeof(int)) ||
1472 	    copy_to_sockptr(optval, &gsf, size0))
1473 		return -EFAULT;
1474 	return 0;
1475 }
1476 
1477 static int compat_ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1478 					sockptr_t optlen, int len)
1479 {
1480 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1481 	struct compat_group_filter gf32;
1482 	struct group_filter gf;
1483 	int num;
1484 	int err;
1485 
1486 	if (len < size0)
1487 		return -EINVAL;
1488 	if (copy_from_sockptr(&gf32, optval, size0))
1489 		return -EFAULT;
1490 
1491 	gf.gf_interface = gf32.gf_interface;
1492 	gf.gf_fmode = gf32.gf_fmode;
1493 	num = gf.gf_numsrc = gf32.gf_numsrc;
1494 	gf.gf_group = gf32.gf_group;
1495 
1496 	err = ip_mc_gsfget(sk, &gf, optval,
1497 			   offsetof(struct compat_group_filter, gf_slist_flex));
1498 	if (err)
1499 		return err;
1500 	if (gf.gf_numsrc < num)
1501 		num = gf.gf_numsrc;
1502 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1503 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1504 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1505 				   &gf.gf_fmode, sizeof(gf.gf_fmode)) ||
1506 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1507 				   &gf.gf_numsrc, sizeof(gf.gf_numsrc)))
1508 		return -EFAULT;
1509 	return 0;
1510 }
1511 
1512 int do_ip_getsockopt(struct sock *sk, int level, int optname,
1513 		     sockptr_t optval, sockptr_t optlen)
1514 {
1515 	struct inet_sock *inet = inet_sk(sk);
1516 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1517 	int val, err = 0;
1518 	int len;
1519 
1520 	if (level != SOL_IP)
1521 		return -EOPNOTSUPP;
1522 
1523 	if (ip_mroute_opt(optname))
1524 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1525 
1526 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1527 		return -EFAULT;
1528 	if (len < 0)
1529 		return -EINVAL;
1530 
1531 	/* Handle options that can be read without locking the socket. */
1532 	switch (optname) {
1533 	case IP_PKTINFO:
1534 		val = inet_test_bit(PKTINFO, sk);
1535 		goto copyval;
1536 	case IP_RECVTTL:
1537 		val = inet_test_bit(TTL, sk);
1538 		goto copyval;
1539 	case IP_RECVTOS:
1540 		val = inet_test_bit(TOS, sk);
1541 		goto copyval;
1542 	case IP_RECVOPTS:
1543 		val = inet_test_bit(RECVOPTS, sk);
1544 		goto copyval;
1545 	case IP_RETOPTS:
1546 		val = inet_test_bit(RETOPTS, sk);
1547 		goto copyval;
1548 	case IP_PASSSEC:
1549 		val = inet_test_bit(PASSSEC, sk);
1550 		goto copyval;
1551 	case IP_RECVORIGDSTADDR:
1552 		val = inet_test_bit(ORIGDSTADDR, sk);
1553 		goto copyval;
1554 	case IP_CHECKSUM:
1555 		val = inet_test_bit(CHECKSUM, sk);
1556 		goto copyval;
1557 	case IP_RECVFRAGSIZE:
1558 		val = inet_test_bit(RECVFRAGSIZE, sk);
1559 		goto copyval;
1560 	case IP_RECVERR:
1561 		val = inet_test_bit(RECVERR, sk);
1562 		goto copyval;
1563 	case IP_RECVERR_RFC4884:
1564 		val = inet_test_bit(RECVERR_RFC4884, sk);
1565 		goto copyval;
1566 	case IP_FREEBIND:
1567 		val = inet_test_bit(FREEBIND, sk);
1568 		goto copyval;
1569 	case IP_HDRINCL:
1570 		val = inet_test_bit(HDRINCL, sk);
1571 		goto copyval;
1572 	case IP_MULTICAST_LOOP:
1573 		val = inet_test_bit(MC_LOOP, sk);
1574 		goto copyval;
1575 	case IP_MULTICAST_ALL:
1576 		val = inet_test_bit(MC_ALL, sk);
1577 		goto copyval;
1578 	case IP_TRANSPARENT:
1579 		val = inet_test_bit(TRANSPARENT, sk);
1580 		goto copyval;
1581 	case IP_NODEFRAG:
1582 		val = inet_test_bit(NODEFRAG, sk);
1583 		goto copyval;
1584 	case IP_BIND_ADDRESS_NO_PORT:
1585 		val = inet_test_bit(BIND_ADDRESS_NO_PORT, sk);
1586 		goto copyval;
1587 	case IP_TTL:
1588 		val = READ_ONCE(inet->uc_ttl);
1589 		if (val < 0)
1590 			val = READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_default_ttl);
1591 		goto copyval;
1592 	case IP_MINTTL:
1593 		val = READ_ONCE(inet->min_ttl);
1594 		goto copyval;
1595 	}
1596 
1597 	if (needs_rtnl)
1598 		rtnl_lock();
1599 	sockopt_lock_sock(sk);
1600 
1601 	switch (optname) {
1602 	case IP_OPTIONS:
1603 	{
1604 		unsigned char optbuf[sizeof(struct ip_options)+40];
1605 		struct ip_options *opt = (struct ip_options *)optbuf;
1606 		struct ip_options_rcu *inet_opt;
1607 
1608 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1609 						     lockdep_sock_is_held(sk));
1610 		opt->optlen = 0;
1611 		if (inet_opt)
1612 			memcpy(optbuf, &inet_opt->opt,
1613 			       sizeof(struct ip_options) +
1614 			       inet_opt->opt.optlen);
1615 		sockopt_release_sock(sk);
1616 
1617 		if (opt->optlen == 0) {
1618 			len = 0;
1619 			return copy_to_sockptr(optlen, &len, sizeof(int));
1620 		}
1621 
1622 		ip_options_undo(opt);
1623 
1624 		len = min_t(unsigned int, len, opt->optlen);
1625 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1626 			return -EFAULT;
1627 		if (copy_to_sockptr(optval, opt->__data, len))
1628 			return -EFAULT;
1629 		return 0;
1630 	}
1631 	case IP_TOS:
1632 		val = inet->tos;
1633 		break;
1634 	case IP_MTU_DISCOVER:
1635 		val = inet->pmtudisc;
1636 		break;
1637 	case IP_MTU:
1638 	{
1639 		struct dst_entry *dst;
1640 		val = 0;
1641 		dst = sk_dst_get(sk);
1642 		if (dst) {
1643 			val = dst_mtu(dst);
1644 			dst_release(dst);
1645 		}
1646 		if (!val) {
1647 			sockopt_release_sock(sk);
1648 			return -ENOTCONN;
1649 		}
1650 		break;
1651 	}
1652 	case IP_MULTICAST_TTL:
1653 		val = inet->mc_ttl;
1654 		break;
1655 	case IP_UNICAST_IF:
1656 		val = (__force int)htonl((__u32) inet->uc_index);
1657 		break;
1658 	case IP_MULTICAST_IF:
1659 	{
1660 		struct in_addr addr;
1661 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1662 		addr.s_addr = inet->mc_addr;
1663 		sockopt_release_sock(sk);
1664 
1665 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1666 			return -EFAULT;
1667 		if (copy_to_sockptr(optval, &addr, len))
1668 			return -EFAULT;
1669 		return 0;
1670 	}
1671 	case IP_MSFILTER:
1672 	{
1673 		struct ip_msfilter msf;
1674 
1675 		if (len < IP_MSFILTER_SIZE(0)) {
1676 			err = -EINVAL;
1677 			goto out;
1678 		}
1679 		if (copy_from_sockptr(&msf, optval, IP_MSFILTER_SIZE(0))) {
1680 			err = -EFAULT;
1681 			goto out;
1682 		}
1683 		err = ip_mc_msfget(sk, &msf, optval, optlen);
1684 		goto out;
1685 	}
1686 	case MCAST_MSFILTER:
1687 		if (in_compat_syscall())
1688 			err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1689 							   len);
1690 		else
1691 			err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1692 		goto out;
1693 	case IP_PKTOPTIONS:
1694 	{
1695 		struct msghdr msg;
1696 
1697 		sockopt_release_sock(sk);
1698 
1699 		if (sk->sk_type != SOCK_STREAM)
1700 			return -ENOPROTOOPT;
1701 
1702 		if (optval.is_kernel) {
1703 			msg.msg_control_is_user = false;
1704 			msg.msg_control = optval.kernel;
1705 		} else {
1706 			msg.msg_control_is_user = true;
1707 			msg.msg_control_user = optval.user;
1708 		}
1709 		msg.msg_controllen = len;
1710 		msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1711 
1712 		if (inet_test_bit(PKTINFO, sk)) {
1713 			struct in_pktinfo info;
1714 
1715 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1716 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1717 			info.ipi_ifindex = inet->mc_index;
1718 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1719 		}
1720 		if (inet_test_bit(TTL, sk)) {
1721 			int hlim = inet->mc_ttl;
1722 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1723 		}
1724 		if (inet_test_bit(TOS, sk)) {
1725 			int tos = inet->rcv_tos;
1726 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1727 		}
1728 		len -= msg.msg_controllen;
1729 		return copy_to_sockptr(optlen, &len, sizeof(int));
1730 	}
1731 	case IP_LOCAL_PORT_RANGE:
1732 		val = inet->local_port_range.hi << 16 | inet->local_port_range.lo;
1733 		break;
1734 	case IP_PROTOCOL:
1735 		val = inet_sk(sk)->inet_num;
1736 		break;
1737 	default:
1738 		sockopt_release_sock(sk);
1739 		return -ENOPROTOOPT;
1740 	}
1741 	sockopt_release_sock(sk);
1742 copyval:
1743 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1744 		unsigned char ucval = (unsigned char)val;
1745 		len = 1;
1746 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1747 			return -EFAULT;
1748 		if (copy_to_sockptr(optval, &ucval, 1))
1749 			return -EFAULT;
1750 	} else {
1751 		len = min_t(unsigned int, sizeof(int), len);
1752 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1753 			return -EFAULT;
1754 		if (copy_to_sockptr(optval, &val, len))
1755 			return -EFAULT;
1756 	}
1757 	return 0;
1758 
1759 out:
1760 	sockopt_release_sock(sk);
1761 	if (needs_rtnl)
1762 		rtnl_unlock();
1763 	return err;
1764 }
1765 
1766 int ip_getsockopt(struct sock *sk, int level,
1767 		  int optname, char __user *optval, int __user *optlen)
1768 {
1769 	int err;
1770 
1771 	err = do_ip_getsockopt(sk, level, optname,
1772 			       USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1773 
1774 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1775 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1776 	    optname < BPFILTER_IPT_GET_MAX)
1777 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1778 #endif
1779 #ifdef CONFIG_NETFILTER
1780 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1781 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1782 			!ip_mroute_opt(optname)) {
1783 		int len;
1784 
1785 		if (get_user(len, optlen))
1786 			return -EFAULT;
1787 
1788 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1789 		if (err >= 0)
1790 			err = put_user(len, optlen);
1791 		return err;
1792 	}
1793 #endif
1794 	return err;
1795 }
1796 EXPORT_SYMBOL(ip_getsockopt);
1797