xref: /openbmc/linux/net/ipv4/ip_sockglue.c (revision 9a5d500c)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		The IP to API glue.
8  *
9  * Authors:	see ip.c
10  *
11  * Fixes:
12  *		Many		:	Split from ip.c , see ip.c for history.
13  *		Martin Mares	:	TOS setting fixed.
14  *		Alan Cox	:	Fixed a couple of oopses in Martin's
15  *					TOS tweaks.
16  *		Mike McLagan	:	Routing by source
17  */
18 
19 #include <linux/module.h>
20 #include <linux/types.h>
21 #include <linux/mm.h>
22 #include <linux/skbuff.h>
23 #include <linux/ip.h>
24 #include <linux/icmp.h>
25 #include <linux/inetdevice.h>
26 #include <linux/netdevice.h>
27 #include <linux/slab.h>
28 #include <net/sock.h>
29 #include <net/ip.h>
30 #include <net/icmp.h>
31 #include <net/tcp_states.h>
32 #include <linux/udp.h>
33 #include <linux/igmp.h>
34 #include <linux/netfilter.h>
35 #include <linux/route.h>
36 #include <linux/mroute.h>
37 #include <net/inet_ecn.h>
38 #include <net/route.h>
39 #include <net/xfrm.h>
40 #include <net/compat.h>
41 #include <net/checksum.h>
42 #if IS_ENABLED(CONFIG_IPV6)
43 #include <net/transp_v6.h>
44 #endif
45 #include <net/ip_fib.h>
46 
47 #include <linux/errqueue.h>
48 #include <linux/uaccess.h>
49 
50 #include <linux/bpfilter.h>
51 
52 /*
53  *	SOL_IP control messages.
54  */
55 
56 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
57 {
58 	struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
59 
60 	info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
61 
62 	put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
63 }
64 
65 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
66 {
67 	int ttl = ip_hdr(skb)->ttl;
68 	put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
69 }
70 
71 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
72 {
73 	put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
74 }
75 
76 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
77 {
78 	if (IPCB(skb)->opt.optlen == 0)
79 		return;
80 
81 	put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
82 		 ip_hdr(skb) + 1);
83 }
84 
85 
86 static void ip_cmsg_recv_retopts(struct net *net, struct msghdr *msg,
87 				 struct sk_buff *skb)
88 {
89 	unsigned char optbuf[sizeof(struct ip_options) + 40];
90 	struct ip_options *opt = (struct ip_options *)optbuf;
91 
92 	if (IPCB(skb)->opt.optlen == 0)
93 		return;
94 
95 	if (ip_options_echo(net, opt, skb)) {
96 		msg->msg_flags |= MSG_CTRUNC;
97 		return;
98 	}
99 	ip_options_undo(opt);
100 
101 	put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
102 }
103 
104 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb)
105 {
106 	int val;
107 
108 	if (IPCB(skb)->frag_max_size == 0)
109 		return;
110 
111 	val = IPCB(skb)->frag_max_size;
112 	put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
113 }
114 
115 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
116 				  int tlen, int offset)
117 {
118 	__wsum csum = skb->csum;
119 
120 	if (skb->ip_summed != CHECKSUM_COMPLETE)
121 		return;
122 
123 	if (offset != 0) {
124 		int tend_off = skb_transport_offset(skb) + tlen;
125 		csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
126 	}
127 
128 	put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
129 }
130 
131 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
132 {
133 	char *secdata;
134 	u32 seclen, secid;
135 	int err;
136 
137 	err = security_socket_getpeersec_dgram(NULL, skb, &secid);
138 	if (err)
139 		return;
140 
141 	err = security_secid_to_secctx(secid, &secdata, &seclen);
142 	if (err)
143 		return;
144 
145 	put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
146 	security_release_secctx(secdata, seclen);
147 }
148 
149 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
150 {
151 	__be16 _ports[2], *ports;
152 	struct sockaddr_in sin;
153 
154 	/* All current transport protocols have the port numbers in the
155 	 * first four bytes of the transport header and this function is
156 	 * written with this assumption in mind.
157 	 */
158 	ports = skb_header_pointer(skb, skb_transport_offset(skb),
159 				   sizeof(_ports), &_ports);
160 	if (!ports)
161 		return;
162 
163 	sin.sin_family = AF_INET;
164 	sin.sin_addr.s_addr = ip_hdr(skb)->daddr;
165 	sin.sin_port = ports[1];
166 	memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
167 
168 	put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
169 }
170 
171 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
172 			 struct sk_buff *skb, int tlen, int offset)
173 {
174 	struct inet_sock *inet = inet_sk(sk);
175 	unsigned int flags = inet->cmsg_flags;
176 
177 	/* Ordered by supposed usage frequency */
178 	if (flags & IP_CMSG_PKTINFO) {
179 		ip_cmsg_recv_pktinfo(msg, skb);
180 
181 		flags &= ~IP_CMSG_PKTINFO;
182 		if (!flags)
183 			return;
184 	}
185 
186 	if (flags & IP_CMSG_TTL) {
187 		ip_cmsg_recv_ttl(msg, skb);
188 
189 		flags &= ~IP_CMSG_TTL;
190 		if (!flags)
191 			return;
192 	}
193 
194 	if (flags & IP_CMSG_TOS) {
195 		ip_cmsg_recv_tos(msg, skb);
196 
197 		flags &= ~IP_CMSG_TOS;
198 		if (!flags)
199 			return;
200 	}
201 
202 	if (flags & IP_CMSG_RECVOPTS) {
203 		ip_cmsg_recv_opts(msg, skb);
204 
205 		flags &= ~IP_CMSG_RECVOPTS;
206 		if (!flags)
207 			return;
208 	}
209 
210 	if (flags & IP_CMSG_RETOPTS) {
211 		ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
212 
213 		flags &= ~IP_CMSG_RETOPTS;
214 		if (!flags)
215 			return;
216 	}
217 
218 	if (flags & IP_CMSG_PASSSEC) {
219 		ip_cmsg_recv_security(msg, skb);
220 
221 		flags &= ~IP_CMSG_PASSSEC;
222 		if (!flags)
223 			return;
224 	}
225 
226 	if (flags & IP_CMSG_ORIGDSTADDR) {
227 		ip_cmsg_recv_dstaddr(msg, skb);
228 
229 		flags &= ~IP_CMSG_ORIGDSTADDR;
230 		if (!flags)
231 			return;
232 	}
233 
234 	if (flags & IP_CMSG_CHECKSUM)
235 		ip_cmsg_recv_checksum(msg, skb, tlen, offset);
236 
237 	if (flags & IP_CMSG_RECVFRAGSIZE)
238 		ip_cmsg_recv_fragsize(msg, skb);
239 }
240 EXPORT_SYMBOL(ip_cmsg_recv_offset);
241 
242 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
243 		 bool allow_ipv6)
244 {
245 	int err, val;
246 	struct cmsghdr *cmsg;
247 	struct net *net = sock_net(sk);
248 
249 	for_each_cmsghdr(cmsg, msg) {
250 		if (!CMSG_OK(msg, cmsg))
251 			return -EINVAL;
252 #if IS_ENABLED(CONFIG_IPV6)
253 		if (allow_ipv6 &&
254 		    cmsg->cmsg_level == SOL_IPV6 &&
255 		    cmsg->cmsg_type == IPV6_PKTINFO) {
256 			struct in6_pktinfo *src_info;
257 
258 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
259 				return -EINVAL;
260 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
261 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
262 				return -EINVAL;
263 			if (src_info->ipi6_ifindex)
264 				ipc->oif = src_info->ipi6_ifindex;
265 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
266 			continue;
267 		}
268 #endif
269 		if (cmsg->cmsg_level == SOL_SOCKET) {
270 			err = __sock_cmsg_send(sk, cmsg, &ipc->sockc);
271 			if (err)
272 				return err;
273 			continue;
274 		}
275 
276 		if (cmsg->cmsg_level != SOL_IP)
277 			continue;
278 		switch (cmsg->cmsg_type) {
279 		case IP_RETOPTS:
280 			err = cmsg->cmsg_len - sizeof(struct cmsghdr);
281 
282 			/* Our caller is responsible for freeing ipc->opt */
283 			err = ip_options_get(net, &ipc->opt,
284 					     KERNEL_SOCKPTR(CMSG_DATA(cmsg)),
285 					     err < 40 ? err : 40);
286 			if (err)
287 				return err;
288 			break;
289 		case IP_PKTINFO:
290 		{
291 			struct in_pktinfo *info;
292 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
293 				return -EINVAL;
294 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
295 			if (info->ipi_ifindex)
296 				ipc->oif = info->ipi_ifindex;
297 			ipc->addr = info->ipi_spec_dst.s_addr;
298 			break;
299 		}
300 		case IP_TTL:
301 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
302 				return -EINVAL;
303 			val = *(int *)CMSG_DATA(cmsg);
304 			if (val < 1 || val > 255)
305 				return -EINVAL;
306 			ipc->ttl = val;
307 			break;
308 		case IP_TOS:
309 			if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
310 				val = *(int *)CMSG_DATA(cmsg);
311 			else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
312 				val = *(u8 *)CMSG_DATA(cmsg);
313 			else
314 				return -EINVAL;
315 			if (val < 0 || val > 255)
316 				return -EINVAL;
317 			ipc->tos = val;
318 			ipc->priority = rt_tos2priority(ipc->tos);
319 			break;
320 		case IP_PROTOCOL:
321 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
322 				return -EINVAL;
323 			val = *(int *)CMSG_DATA(cmsg);
324 			if (val < 1 || val > 255)
325 				return -EINVAL;
326 			ipc->protocol = val;
327 			break;
328 		default:
329 			return -EINVAL;
330 		}
331 	}
332 	return 0;
333 }
334 
335 static void ip_ra_destroy_rcu(struct rcu_head *head)
336 {
337 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
338 
339 	sock_put(ra->saved_sk);
340 	kfree(ra);
341 }
342 
343 int ip_ra_control(struct sock *sk, unsigned char on,
344 		  void (*destructor)(struct sock *))
345 {
346 	struct ip_ra_chain *ra, *new_ra;
347 	struct ip_ra_chain __rcu **rap;
348 	struct net *net = sock_net(sk);
349 
350 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
351 		return -EINVAL;
352 
353 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
354 	if (on && !new_ra)
355 		return -ENOMEM;
356 
357 	mutex_lock(&net->ipv4.ra_mutex);
358 	for (rap = &net->ipv4.ra_chain;
359 	     (ra = rcu_dereference_protected(*rap,
360 			lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
361 	     rap = &ra->next) {
362 		if (ra->sk == sk) {
363 			if (on) {
364 				mutex_unlock(&net->ipv4.ra_mutex);
365 				kfree(new_ra);
366 				return -EADDRINUSE;
367 			}
368 			/* dont let ip_call_ra_chain() use sk again */
369 			ra->sk = NULL;
370 			RCU_INIT_POINTER(*rap, ra->next);
371 			mutex_unlock(&net->ipv4.ra_mutex);
372 
373 			if (ra->destructor)
374 				ra->destructor(sk);
375 			/*
376 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
377 			 * period. This guarantee ip_call_ra_chain() dont need
378 			 * to mess with socket refcounts.
379 			 */
380 			ra->saved_sk = sk;
381 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
382 			return 0;
383 		}
384 	}
385 	if (!new_ra) {
386 		mutex_unlock(&net->ipv4.ra_mutex);
387 		return -ENOBUFS;
388 	}
389 	new_ra->sk = sk;
390 	new_ra->destructor = destructor;
391 
392 	RCU_INIT_POINTER(new_ra->next, ra);
393 	rcu_assign_pointer(*rap, new_ra);
394 	sock_hold(sk);
395 	mutex_unlock(&net->ipv4.ra_mutex);
396 
397 	return 0;
398 }
399 
400 static void ipv4_icmp_error_rfc4884(const struct sk_buff *skb,
401 				    struct sock_ee_data_rfc4884 *out)
402 {
403 	switch (icmp_hdr(skb)->type) {
404 	case ICMP_DEST_UNREACH:
405 	case ICMP_TIME_EXCEEDED:
406 	case ICMP_PARAMETERPROB:
407 		ip_icmp_error_rfc4884(skb, out, sizeof(struct icmphdr),
408 				      icmp_hdr(skb)->un.reserved[1] * 4);
409 	}
410 }
411 
412 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
413 		   __be16 port, u32 info, u8 *payload)
414 {
415 	struct sock_exterr_skb *serr;
416 
417 	skb = skb_clone(skb, GFP_ATOMIC);
418 	if (!skb)
419 		return;
420 
421 	serr = SKB_EXT_ERR(skb);
422 	serr->ee.ee_errno = err;
423 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
424 	serr->ee.ee_type = icmp_hdr(skb)->type;
425 	serr->ee.ee_code = icmp_hdr(skb)->code;
426 	serr->ee.ee_pad = 0;
427 	serr->ee.ee_info = info;
428 	serr->ee.ee_data = 0;
429 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
430 				   skb_network_header(skb);
431 	serr->port = port;
432 
433 	if (skb_pull(skb, payload - skb->data)) {
434 		if (inet_sk(sk)->recverr_rfc4884)
435 			ipv4_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
436 
437 		skb_reset_transport_header(skb);
438 		if (sock_queue_err_skb(sk, skb) == 0)
439 			return;
440 	}
441 	kfree_skb(skb);
442 }
443 EXPORT_SYMBOL_GPL(ip_icmp_error);
444 
445 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
446 {
447 	struct inet_sock *inet = inet_sk(sk);
448 	struct sock_exterr_skb *serr;
449 	struct iphdr *iph;
450 	struct sk_buff *skb;
451 
452 	if (!inet->recverr)
453 		return;
454 
455 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
456 	if (!skb)
457 		return;
458 
459 	skb_put(skb, sizeof(struct iphdr));
460 	skb_reset_network_header(skb);
461 	iph = ip_hdr(skb);
462 	iph->daddr = daddr;
463 
464 	serr = SKB_EXT_ERR(skb);
465 	serr->ee.ee_errno = err;
466 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
467 	serr->ee.ee_type = 0;
468 	serr->ee.ee_code = 0;
469 	serr->ee.ee_pad = 0;
470 	serr->ee.ee_info = info;
471 	serr->ee.ee_data = 0;
472 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
473 	serr->port = port;
474 
475 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
476 	skb_reset_transport_header(skb);
477 
478 	if (sock_queue_err_skb(sk, skb))
479 		kfree_skb(skb);
480 }
481 
482 /* For some errors we have valid addr_offset even with zero payload and
483  * zero port. Also, addr_offset should be supported if port is set.
484  */
485 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
486 {
487 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
488 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
489 }
490 
491 /* IPv4 supports cmsg on all imcp errors and some timestamps
492  *
493  * Timestamp code paths do not initialize the fields expected by cmsg:
494  * the PKTINFO fields in skb->cb[]. Fill those in here.
495  */
496 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
497 				       struct sk_buff *skb,
498 				       int ee_origin)
499 {
500 	struct in_pktinfo *info;
501 
502 	if (ee_origin == SO_EE_ORIGIN_ICMP)
503 		return true;
504 
505 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
506 		return false;
507 
508 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
509 	 * timestamp with egress dev. Not possible for packets without iif
510 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
511 	 */
512 	info = PKTINFO_SKB_CB(skb);
513 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
514 	    !info->ipi_ifindex)
515 		return false;
516 
517 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
518 	return true;
519 }
520 
521 /*
522  *	Handle MSG_ERRQUEUE
523  */
524 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
525 {
526 	struct sock_exterr_skb *serr;
527 	struct sk_buff *skb;
528 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
529 	struct {
530 		struct sock_extended_err ee;
531 		struct sockaddr_in	 offender;
532 	} errhdr;
533 	int err;
534 	int copied;
535 
536 	err = -EAGAIN;
537 	skb = sock_dequeue_err_skb(sk);
538 	if (!skb)
539 		goto out;
540 
541 	copied = skb->len;
542 	if (copied > len) {
543 		msg->msg_flags |= MSG_TRUNC;
544 		copied = len;
545 	}
546 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
547 	if (unlikely(err)) {
548 		kfree_skb(skb);
549 		return err;
550 	}
551 	sock_recv_timestamp(msg, sk, skb);
552 
553 	serr = SKB_EXT_ERR(skb);
554 
555 	if (sin && ipv4_datagram_support_addr(serr)) {
556 		sin->sin_family = AF_INET;
557 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
558 						   serr->addr_offset);
559 		sin->sin_port = serr->port;
560 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
561 		*addr_len = sizeof(*sin);
562 	}
563 
564 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
565 	sin = &errhdr.offender;
566 	memset(sin, 0, sizeof(*sin));
567 
568 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
569 		sin->sin_family = AF_INET;
570 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
571 		if (inet_sk(sk)->cmsg_flags)
572 			ip_cmsg_recv(msg, skb);
573 	}
574 
575 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
576 
577 	/* Now we could try to dump offended packet options */
578 
579 	msg->msg_flags |= MSG_ERRQUEUE;
580 	err = copied;
581 
582 	consume_skb(skb);
583 out:
584 	return err;
585 }
586 
587 void __ip_sock_set_tos(struct sock *sk, int val)
588 {
589 	if (sk->sk_type == SOCK_STREAM) {
590 		val &= ~INET_ECN_MASK;
591 		val |= inet_sk(sk)->tos & INET_ECN_MASK;
592 	}
593 	if (inet_sk(sk)->tos != val) {
594 		inet_sk(sk)->tos = val;
595 		sk->sk_priority = rt_tos2priority(val);
596 		sk_dst_reset(sk);
597 	}
598 }
599 
600 void ip_sock_set_tos(struct sock *sk, int val)
601 {
602 	lock_sock(sk);
603 	__ip_sock_set_tos(sk, val);
604 	release_sock(sk);
605 }
606 EXPORT_SYMBOL(ip_sock_set_tos);
607 
608 void ip_sock_set_freebind(struct sock *sk)
609 {
610 	lock_sock(sk);
611 	inet_sk(sk)->freebind = true;
612 	release_sock(sk);
613 }
614 EXPORT_SYMBOL(ip_sock_set_freebind);
615 
616 void ip_sock_set_recverr(struct sock *sk)
617 {
618 	lock_sock(sk);
619 	inet_sk(sk)->recverr = true;
620 	release_sock(sk);
621 }
622 EXPORT_SYMBOL(ip_sock_set_recverr);
623 
624 int ip_sock_set_mtu_discover(struct sock *sk, int val)
625 {
626 	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
627 		return -EINVAL;
628 	lock_sock(sk);
629 	inet_sk(sk)->pmtudisc = val;
630 	release_sock(sk);
631 	return 0;
632 }
633 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
634 
635 void ip_sock_set_pktinfo(struct sock *sk)
636 {
637 	lock_sock(sk);
638 	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
639 	release_sock(sk);
640 }
641 EXPORT_SYMBOL(ip_sock_set_pktinfo);
642 
643 /*
644  *	Socket option code for IP. This is the end of the line after any
645  *	TCP,UDP etc options on an IP socket.
646  */
647 static bool setsockopt_needs_rtnl(int optname)
648 {
649 	switch (optname) {
650 	case IP_ADD_MEMBERSHIP:
651 	case IP_ADD_SOURCE_MEMBERSHIP:
652 	case IP_BLOCK_SOURCE:
653 	case IP_DROP_MEMBERSHIP:
654 	case IP_DROP_SOURCE_MEMBERSHIP:
655 	case IP_MSFILTER:
656 	case IP_UNBLOCK_SOURCE:
657 	case MCAST_BLOCK_SOURCE:
658 	case MCAST_MSFILTER:
659 	case MCAST_JOIN_GROUP:
660 	case MCAST_JOIN_SOURCE_GROUP:
661 	case MCAST_LEAVE_GROUP:
662 	case MCAST_LEAVE_SOURCE_GROUP:
663 	case MCAST_UNBLOCK_SOURCE:
664 		return true;
665 	}
666 	return false;
667 }
668 
669 static int set_mcast_msfilter(struct sock *sk, int ifindex,
670 			      int numsrc, int fmode,
671 			      struct sockaddr_storage *group,
672 			      struct sockaddr_storage *list)
673 {
674 	struct ip_msfilter *msf;
675 	struct sockaddr_in *psin;
676 	int err, i;
677 
678 	msf = kmalloc(IP_MSFILTER_SIZE(numsrc), GFP_KERNEL);
679 	if (!msf)
680 		return -ENOBUFS;
681 
682 	psin = (struct sockaddr_in *)group;
683 	if (psin->sin_family != AF_INET)
684 		goto Eaddrnotavail;
685 	msf->imsf_multiaddr = psin->sin_addr.s_addr;
686 	msf->imsf_interface = 0;
687 	msf->imsf_fmode = fmode;
688 	msf->imsf_numsrc = numsrc;
689 	for (i = 0; i < numsrc; ++i) {
690 		psin = (struct sockaddr_in *)&list[i];
691 
692 		if (psin->sin_family != AF_INET)
693 			goto Eaddrnotavail;
694 		msf->imsf_slist_flex[i] = psin->sin_addr.s_addr;
695 	}
696 	err = ip_mc_msfilter(sk, msf, ifindex);
697 	kfree(msf);
698 	return err;
699 
700 Eaddrnotavail:
701 	kfree(msf);
702 	return -EADDRNOTAVAIL;
703 }
704 
705 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
706 		sockptr_t optval, int optlen)
707 {
708 	if (in_compat_syscall()) {
709 		struct compat_group_source_req gr32;
710 
711 		if (optlen != sizeof(gr32))
712 			return -EINVAL;
713 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
714 			return -EFAULT;
715 		greqs->gsr_interface = gr32.gsr_interface;
716 		greqs->gsr_group = gr32.gsr_group;
717 		greqs->gsr_source = gr32.gsr_source;
718 	} else {
719 		if (optlen != sizeof(*greqs))
720 			return -EINVAL;
721 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
722 			return -EFAULT;
723 	}
724 
725 	return 0;
726 }
727 
728 static int do_mcast_group_source(struct sock *sk, int optname,
729 		sockptr_t optval, int optlen)
730 {
731 	struct group_source_req greqs;
732 	struct ip_mreq_source mreqs;
733 	struct sockaddr_in *psin;
734 	int omode, add, err;
735 
736 	err = copy_group_source_from_sockptr(&greqs, optval, optlen);
737 	if (err)
738 		return err;
739 
740 	if (greqs.gsr_group.ss_family != AF_INET ||
741 	    greqs.gsr_source.ss_family != AF_INET)
742 		return -EADDRNOTAVAIL;
743 
744 	psin = (struct sockaddr_in *)&greqs.gsr_group;
745 	mreqs.imr_multiaddr = psin->sin_addr.s_addr;
746 	psin = (struct sockaddr_in *)&greqs.gsr_source;
747 	mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
748 	mreqs.imr_interface = 0; /* use index for mc_source */
749 
750 	if (optname == MCAST_BLOCK_SOURCE) {
751 		omode = MCAST_EXCLUDE;
752 		add = 1;
753 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
754 		omode = MCAST_EXCLUDE;
755 		add = 0;
756 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
757 		struct ip_mreqn mreq;
758 
759 		psin = (struct sockaddr_in *)&greqs.gsr_group;
760 		mreq.imr_multiaddr = psin->sin_addr;
761 		mreq.imr_address.s_addr = 0;
762 		mreq.imr_ifindex = greqs.gsr_interface;
763 		err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
764 		if (err && err != -EADDRINUSE)
765 			return err;
766 		greqs.gsr_interface = mreq.imr_ifindex;
767 		omode = MCAST_INCLUDE;
768 		add = 1;
769 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
770 		omode = MCAST_INCLUDE;
771 		add = 0;
772 	}
773 	return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
774 }
775 
776 static int ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval, int optlen)
777 {
778 	struct group_filter *gsf = NULL;
779 	int err;
780 
781 	if (optlen < GROUP_FILTER_SIZE(0))
782 		return -EINVAL;
783 	if (optlen > READ_ONCE(sysctl_optmem_max))
784 		return -ENOBUFS;
785 
786 	gsf = memdup_sockptr(optval, optlen);
787 	if (IS_ERR(gsf))
788 		return PTR_ERR(gsf);
789 
790 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
791 	err = -ENOBUFS;
792 	if (gsf->gf_numsrc >= 0x1ffffff ||
793 	    gsf->gf_numsrc > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
794 		goto out_free_gsf;
795 
796 	err = -EINVAL;
797 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
798 		goto out_free_gsf;
799 
800 	err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
801 				 gsf->gf_fmode, &gsf->gf_group,
802 				 gsf->gf_slist_flex);
803 out_free_gsf:
804 	kfree(gsf);
805 	return err;
806 }
807 
808 static int compat_ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
809 		int optlen)
810 {
811 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
812 	struct compat_group_filter *gf32;
813 	unsigned int n;
814 	void *p;
815 	int err;
816 
817 	if (optlen < size0)
818 		return -EINVAL;
819 	if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
820 		return -ENOBUFS;
821 
822 	p = kmalloc(optlen + 4, GFP_KERNEL);
823 	if (!p)
824 		return -ENOMEM;
825 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
826 
827 	err = -EFAULT;
828 	if (copy_from_sockptr(gf32, optval, optlen))
829 		goto out_free_gsf;
830 
831 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
832 	n = gf32->gf_numsrc;
833 	err = -ENOBUFS;
834 	if (n >= 0x1ffffff)
835 		goto out_free_gsf;
836 
837 	err = -EINVAL;
838 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
839 		goto out_free_gsf;
840 
841 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
842 	err = -ENOBUFS;
843 	if (n > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
844 		goto out_free_gsf;
845 	err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
846 				 &gf32->gf_group, gf32->gf_slist_flex);
847 out_free_gsf:
848 	kfree(p);
849 	return err;
850 }
851 
852 static int ip_mcast_join_leave(struct sock *sk, int optname,
853 		sockptr_t optval, int optlen)
854 {
855 	struct ip_mreqn mreq = { };
856 	struct sockaddr_in *psin;
857 	struct group_req greq;
858 
859 	if (optlen < sizeof(struct group_req))
860 		return -EINVAL;
861 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
862 		return -EFAULT;
863 
864 	psin = (struct sockaddr_in *)&greq.gr_group;
865 	if (psin->sin_family != AF_INET)
866 		return -EINVAL;
867 	mreq.imr_multiaddr = psin->sin_addr;
868 	mreq.imr_ifindex = greq.gr_interface;
869 	if (optname == MCAST_JOIN_GROUP)
870 		return ip_mc_join_group(sk, &mreq);
871 	return ip_mc_leave_group(sk, &mreq);
872 }
873 
874 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
875 		sockptr_t optval, int optlen)
876 {
877 	struct compat_group_req greq;
878 	struct ip_mreqn mreq = { };
879 	struct sockaddr_in *psin;
880 
881 	if (optlen < sizeof(struct compat_group_req))
882 		return -EINVAL;
883 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
884 		return -EFAULT;
885 
886 	psin = (struct sockaddr_in *)&greq.gr_group;
887 	if (psin->sin_family != AF_INET)
888 		return -EINVAL;
889 	mreq.imr_multiaddr = psin->sin_addr;
890 	mreq.imr_ifindex = greq.gr_interface;
891 
892 	if (optname == MCAST_JOIN_GROUP)
893 		return ip_mc_join_group(sk, &mreq);
894 	return ip_mc_leave_group(sk, &mreq);
895 }
896 
897 DEFINE_STATIC_KEY_FALSE(ip4_min_ttl);
898 
899 int do_ip_setsockopt(struct sock *sk, int level, int optname,
900 		     sockptr_t optval, unsigned int optlen)
901 {
902 	struct inet_sock *inet = inet_sk(sk);
903 	struct net *net = sock_net(sk);
904 	int val = 0, err;
905 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
906 
907 	switch (optname) {
908 	case IP_PKTINFO:
909 	case IP_RECVTTL:
910 	case IP_RECVOPTS:
911 	case IP_RECVTOS:
912 	case IP_RETOPTS:
913 	case IP_TOS:
914 	case IP_TTL:
915 	case IP_HDRINCL:
916 	case IP_MTU_DISCOVER:
917 	case IP_RECVERR:
918 	case IP_ROUTER_ALERT:
919 	case IP_FREEBIND:
920 	case IP_PASSSEC:
921 	case IP_TRANSPARENT:
922 	case IP_MINTTL:
923 	case IP_NODEFRAG:
924 	case IP_BIND_ADDRESS_NO_PORT:
925 	case IP_UNICAST_IF:
926 	case IP_MULTICAST_TTL:
927 	case IP_MULTICAST_ALL:
928 	case IP_MULTICAST_LOOP:
929 	case IP_RECVORIGDSTADDR:
930 	case IP_CHECKSUM:
931 	case IP_RECVFRAGSIZE:
932 	case IP_RECVERR_RFC4884:
933 	case IP_LOCAL_PORT_RANGE:
934 		if (optlen >= sizeof(int)) {
935 			if (copy_from_sockptr(&val, optval, sizeof(val)))
936 				return -EFAULT;
937 		} else if (optlen >= sizeof(char)) {
938 			unsigned char ucval;
939 
940 			if (copy_from_sockptr(&ucval, optval, sizeof(ucval)))
941 				return -EFAULT;
942 			val = (int) ucval;
943 		}
944 	}
945 
946 	/* If optlen==0, it is equivalent to val == 0 */
947 
948 	if (optname == IP_ROUTER_ALERT)
949 		return ip_ra_control(sk, val ? 1 : 0, NULL);
950 	if (ip_mroute_opt(optname))
951 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
952 
953 	err = 0;
954 	if (needs_rtnl)
955 		rtnl_lock();
956 	sockopt_lock_sock(sk);
957 
958 	switch (optname) {
959 	case IP_OPTIONS:
960 	{
961 		struct ip_options_rcu *old, *opt = NULL;
962 
963 		if (optlen > 40)
964 			goto e_inval;
965 		err = ip_options_get(sock_net(sk), &opt, optval, optlen);
966 		if (err)
967 			break;
968 		old = rcu_dereference_protected(inet->inet_opt,
969 						lockdep_sock_is_held(sk));
970 		if (inet->is_icsk) {
971 			struct inet_connection_sock *icsk = inet_csk(sk);
972 #if IS_ENABLED(CONFIG_IPV6)
973 			if (sk->sk_family == PF_INET ||
974 			    (!((1 << sk->sk_state) &
975 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
976 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
977 #endif
978 				if (old)
979 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
980 				if (opt)
981 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
982 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
983 #if IS_ENABLED(CONFIG_IPV6)
984 			}
985 #endif
986 		}
987 		rcu_assign_pointer(inet->inet_opt, opt);
988 		if (old)
989 			kfree_rcu(old, rcu);
990 		break;
991 	}
992 	case IP_PKTINFO:
993 		if (val)
994 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
995 		else
996 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
997 		break;
998 	case IP_RECVTTL:
999 		if (val)
1000 			inet->cmsg_flags |=  IP_CMSG_TTL;
1001 		else
1002 			inet->cmsg_flags &= ~IP_CMSG_TTL;
1003 		break;
1004 	case IP_RECVTOS:
1005 		if (val)
1006 			inet->cmsg_flags |=  IP_CMSG_TOS;
1007 		else
1008 			inet->cmsg_flags &= ~IP_CMSG_TOS;
1009 		break;
1010 	case IP_RECVOPTS:
1011 		if (val)
1012 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
1013 		else
1014 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
1015 		break;
1016 	case IP_RETOPTS:
1017 		if (val)
1018 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
1019 		else
1020 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
1021 		break;
1022 	case IP_PASSSEC:
1023 		if (val)
1024 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
1025 		else
1026 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
1027 		break;
1028 	case IP_RECVORIGDSTADDR:
1029 		if (val)
1030 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
1031 		else
1032 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
1033 		break;
1034 	case IP_CHECKSUM:
1035 		if (val) {
1036 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
1037 				inet_inc_convert_csum(sk);
1038 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
1039 			}
1040 		} else {
1041 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
1042 				inet_dec_convert_csum(sk);
1043 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
1044 			}
1045 		}
1046 		break;
1047 	case IP_RECVFRAGSIZE:
1048 		if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
1049 			goto e_inval;
1050 		if (val)
1051 			inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
1052 		else
1053 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
1054 		break;
1055 	case IP_TOS:	/* This sets both TOS and Precedence */
1056 		__ip_sock_set_tos(sk, val);
1057 		break;
1058 	case IP_TTL:
1059 		if (optlen < 1)
1060 			goto e_inval;
1061 		if (val != -1 && (val < 1 || val > 255))
1062 			goto e_inval;
1063 		inet->uc_ttl = val;
1064 		break;
1065 	case IP_HDRINCL:
1066 		if (sk->sk_type != SOCK_RAW) {
1067 			err = -ENOPROTOOPT;
1068 			break;
1069 		}
1070 		inet->hdrincl = val ? 1 : 0;
1071 		break;
1072 	case IP_NODEFRAG:
1073 		if (sk->sk_type != SOCK_RAW) {
1074 			err = -ENOPROTOOPT;
1075 			break;
1076 		}
1077 		inet->nodefrag = val ? 1 : 0;
1078 		break;
1079 	case IP_BIND_ADDRESS_NO_PORT:
1080 		inet->bind_address_no_port = val ? 1 : 0;
1081 		break;
1082 	case IP_MTU_DISCOVER:
1083 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
1084 			goto e_inval;
1085 		inet->pmtudisc = val;
1086 		break;
1087 	case IP_RECVERR:
1088 		inet->recverr = !!val;
1089 		if (!val)
1090 			skb_queue_purge(&sk->sk_error_queue);
1091 		break;
1092 	case IP_RECVERR_RFC4884:
1093 		if (val < 0 || val > 1)
1094 			goto e_inval;
1095 		inet->recverr_rfc4884 = !!val;
1096 		break;
1097 	case IP_MULTICAST_TTL:
1098 		if (sk->sk_type == SOCK_STREAM)
1099 			goto e_inval;
1100 		if (optlen < 1)
1101 			goto e_inval;
1102 		if (val == -1)
1103 			val = 1;
1104 		if (val < 0 || val > 255)
1105 			goto e_inval;
1106 		inet->mc_ttl = val;
1107 		break;
1108 	case IP_MULTICAST_LOOP:
1109 		if (optlen < 1)
1110 			goto e_inval;
1111 		inet->mc_loop = !!val;
1112 		break;
1113 	case IP_UNICAST_IF:
1114 	{
1115 		struct net_device *dev = NULL;
1116 		int ifindex;
1117 		int midx;
1118 
1119 		if (optlen != sizeof(int))
1120 			goto e_inval;
1121 
1122 		ifindex = (__force int)ntohl((__force __be32)val);
1123 		if (ifindex == 0) {
1124 			inet->uc_index = 0;
1125 			err = 0;
1126 			break;
1127 		}
1128 
1129 		dev = dev_get_by_index(sock_net(sk), ifindex);
1130 		err = -EADDRNOTAVAIL;
1131 		if (!dev)
1132 			break;
1133 
1134 		midx = l3mdev_master_ifindex(dev);
1135 		dev_put(dev);
1136 
1137 		err = -EINVAL;
1138 		if (sk->sk_bound_dev_if && midx != sk->sk_bound_dev_if)
1139 			break;
1140 
1141 		inet->uc_index = ifindex;
1142 		err = 0;
1143 		break;
1144 	}
1145 	case IP_MULTICAST_IF:
1146 	{
1147 		struct ip_mreqn mreq;
1148 		struct net_device *dev = NULL;
1149 		int midx;
1150 
1151 		if (sk->sk_type == SOCK_STREAM)
1152 			goto e_inval;
1153 		/*
1154 		 *	Check the arguments are allowable
1155 		 */
1156 
1157 		if (optlen < sizeof(struct in_addr))
1158 			goto e_inval;
1159 
1160 		err = -EFAULT;
1161 		if (optlen >= sizeof(struct ip_mreqn)) {
1162 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1163 				break;
1164 		} else {
1165 			memset(&mreq, 0, sizeof(mreq));
1166 			if (optlen >= sizeof(struct ip_mreq)) {
1167 				if (copy_from_sockptr(&mreq, optval,
1168 						      sizeof(struct ip_mreq)))
1169 					break;
1170 			} else if (optlen >= sizeof(struct in_addr)) {
1171 				if (copy_from_sockptr(&mreq.imr_address, optval,
1172 						      sizeof(struct in_addr)))
1173 					break;
1174 			}
1175 		}
1176 
1177 		if (!mreq.imr_ifindex) {
1178 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1179 				inet->mc_index = 0;
1180 				inet->mc_addr  = 0;
1181 				err = 0;
1182 				break;
1183 			}
1184 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1185 			if (dev)
1186 				mreq.imr_ifindex = dev->ifindex;
1187 		} else
1188 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1189 
1190 
1191 		err = -EADDRNOTAVAIL;
1192 		if (!dev)
1193 			break;
1194 
1195 		midx = l3mdev_master_ifindex(dev);
1196 
1197 		dev_put(dev);
1198 
1199 		err = -EINVAL;
1200 		if (sk->sk_bound_dev_if &&
1201 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
1202 		    midx != sk->sk_bound_dev_if)
1203 			break;
1204 
1205 		inet->mc_index = mreq.imr_ifindex;
1206 		inet->mc_addr  = mreq.imr_address.s_addr;
1207 		err = 0;
1208 		break;
1209 	}
1210 
1211 	case IP_ADD_MEMBERSHIP:
1212 	case IP_DROP_MEMBERSHIP:
1213 	{
1214 		struct ip_mreqn mreq;
1215 
1216 		err = -EPROTO;
1217 		if (inet_sk(sk)->is_icsk)
1218 			break;
1219 
1220 		if (optlen < sizeof(struct ip_mreq))
1221 			goto e_inval;
1222 		err = -EFAULT;
1223 		if (optlen >= sizeof(struct ip_mreqn)) {
1224 			if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1225 				break;
1226 		} else {
1227 			memset(&mreq, 0, sizeof(mreq));
1228 			if (copy_from_sockptr(&mreq, optval,
1229 					      sizeof(struct ip_mreq)))
1230 				break;
1231 		}
1232 
1233 		if (optname == IP_ADD_MEMBERSHIP)
1234 			err = ip_mc_join_group(sk, &mreq);
1235 		else
1236 			err = ip_mc_leave_group(sk, &mreq);
1237 		break;
1238 	}
1239 	case IP_MSFILTER:
1240 	{
1241 		struct ip_msfilter *msf;
1242 
1243 		if (optlen < IP_MSFILTER_SIZE(0))
1244 			goto e_inval;
1245 		if (optlen > READ_ONCE(sysctl_optmem_max)) {
1246 			err = -ENOBUFS;
1247 			break;
1248 		}
1249 		msf = memdup_sockptr(optval, optlen);
1250 		if (IS_ERR(msf)) {
1251 			err = PTR_ERR(msf);
1252 			break;
1253 		}
1254 		/* numsrc >= (1G-4) overflow in 32 bits */
1255 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
1256 		    msf->imsf_numsrc > READ_ONCE(net->ipv4.sysctl_igmp_max_msf)) {
1257 			kfree(msf);
1258 			err = -ENOBUFS;
1259 			break;
1260 		}
1261 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1262 			kfree(msf);
1263 			err = -EINVAL;
1264 			break;
1265 		}
1266 		err = ip_mc_msfilter(sk, msf, 0);
1267 		kfree(msf);
1268 		break;
1269 	}
1270 	case IP_BLOCK_SOURCE:
1271 	case IP_UNBLOCK_SOURCE:
1272 	case IP_ADD_SOURCE_MEMBERSHIP:
1273 	case IP_DROP_SOURCE_MEMBERSHIP:
1274 	{
1275 		struct ip_mreq_source mreqs;
1276 		int omode, add;
1277 
1278 		if (optlen != sizeof(struct ip_mreq_source))
1279 			goto e_inval;
1280 		if (copy_from_sockptr(&mreqs, optval, sizeof(mreqs))) {
1281 			err = -EFAULT;
1282 			break;
1283 		}
1284 		if (optname == IP_BLOCK_SOURCE) {
1285 			omode = MCAST_EXCLUDE;
1286 			add = 1;
1287 		} else if (optname == IP_UNBLOCK_SOURCE) {
1288 			omode = MCAST_EXCLUDE;
1289 			add = 0;
1290 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1291 			struct ip_mreqn mreq;
1292 
1293 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1294 			mreq.imr_address.s_addr = mreqs.imr_interface;
1295 			mreq.imr_ifindex = 0;
1296 			err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1297 			if (err && err != -EADDRINUSE)
1298 				break;
1299 			omode = MCAST_INCLUDE;
1300 			add = 1;
1301 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1302 			omode = MCAST_INCLUDE;
1303 			add = 0;
1304 		}
1305 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
1306 		break;
1307 	}
1308 	case MCAST_JOIN_GROUP:
1309 	case MCAST_LEAVE_GROUP:
1310 		if (in_compat_syscall())
1311 			err = compat_ip_mcast_join_leave(sk, optname, optval,
1312 							 optlen);
1313 		else
1314 			err = ip_mcast_join_leave(sk, optname, optval, optlen);
1315 		break;
1316 	case MCAST_JOIN_SOURCE_GROUP:
1317 	case MCAST_LEAVE_SOURCE_GROUP:
1318 	case MCAST_BLOCK_SOURCE:
1319 	case MCAST_UNBLOCK_SOURCE:
1320 		err = do_mcast_group_source(sk, optname, optval, optlen);
1321 		break;
1322 	case MCAST_MSFILTER:
1323 		if (in_compat_syscall())
1324 			err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1325 		else
1326 			err = ip_set_mcast_msfilter(sk, optval, optlen);
1327 		break;
1328 	case IP_MULTICAST_ALL:
1329 		if (optlen < 1)
1330 			goto e_inval;
1331 		if (val != 0 && val != 1)
1332 			goto e_inval;
1333 		inet->mc_all = val;
1334 		break;
1335 
1336 	case IP_FREEBIND:
1337 		if (optlen < 1)
1338 			goto e_inval;
1339 		inet->freebind = !!val;
1340 		break;
1341 
1342 	case IP_IPSEC_POLICY:
1343 	case IP_XFRM_POLICY:
1344 		err = -EPERM;
1345 		if (!sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1346 			break;
1347 		err = xfrm_user_policy(sk, optname, optval, optlen);
1348 		break;
1349 
1350 	case IP_TRANSPARENT:
1351 		if (!!val && !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1352 		    !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1353 			err = -EPERM;
1354 			break;
1355 		}
1356 		if (optlen < 1)
1357 			goto e_inval;
1358 		inet->transparent = !!val;
1359 		break;
1360 
1361 	case IP_MINTTL:
1362 		if (optlen < 1)
1363 			goto e_inval;
1364 		if (val < 0 || val > 255)
1365 			goto e_inval;
1366 
1367 		if (val)
1368 			static_branch_enable(&ip4_min_ttl);
1369 
1370 		/* tcp_v4_err() and tcp_v4_rcv() might read min_ttl
1371 		 * while we are changint it.
1372 		 */
1373 		WRITE_ONCE(inet->min_ttl, val);
1374 		break;
1375 
1376 	case IP_LOCAL_PORT_RANGE:
1377 	{
1378 		const __u16 lo = val;
1379 		const __u16 hi = val >> 16;
1380 
1381 		if (optlen != sizeof(__u32))
1382 			goto e_inval;
1383 		if (lo != 0 && hi != 0 && lo > hi)
1384 			goto e_inval;
1385 
1386 		inet->local_port_range.lo = lo;
1387 		inet->local_port_range.hi = hi;
1388 		break;
1389 	}
1390 	default:
1391 		err = -ENOPROTOOPT;
1392 		break;
1393 	}
1394 	sockopt_release_sock(sk);
1395 	if (needs_rtnl)
1396 		rtnl_unlock();
1397 	return err;
1398 
1399 e_inval:
1400 	sockopt_release_sock(sk);
1401 	if (needs_rtnl)
1402 		rtnl_unlock();
1403 	return -EINVAL;
1404 }
1405 
1406 /**
1407  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1408  * @sk: socket
1409  * @skb: buffer
1410  *
1411  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1412  * destination in skb->cb[] before dst drop.
1413  * This way, receiver doesn't make cache line misses to read rtable.
1414  */
1415 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1416 {
1417 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1418 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1419 		       ipv6_sk_rxinfo(sk);
1420 
1421 	if (prepare && skb_rtable(skb)) {
1422 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1423 		 * which has interface index (iif) as the first member of the
1424 		 * underlying inet{6}_skb_parm struct. This code then overlays
1425 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1426 		 * element so the iif is picked up from the prior IPCB. If iif
1427 		 * is the loopback interface, then return the sending interface
1428 		 * (e.g., process binds socket to eth0 for Tx which is
1429 		 * redirected to loopback in the rtable/dst).
1430 		 */
1431 		struct rtable *rt = skb_rtable(skb);
1432 		bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1433 
1434 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1435 			pktinfo->ipi_ifindex = inet_iif(skb);
1436 		else if (l3slave && rt && rt->rt_iif)
1437 			pktinfo->ipi_ifindex = rt->rt_iif;
1438 
1439 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1440 	} else {
1441 		pktinfo->ipi_ifindex = 0;
1442 		pktinfo->ipi_spec_dst.s_addr = 0;
1443 	}
1444 	skb_dst_drop(skb);
1445 }
1446 
1447 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1448 		unsigned int optlen)
1449 {
1450 	int err;
1451 
1452 	if (level != SOL_IP)
1453 		return -ENOPROTOOPT;
1454 
1455 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1456 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1457 	if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1458 	    optname < BPFILTER_IPT_SET_MAX)
1459 		err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1460 #endif
1461 #ifdef CONFIG_NETFILTER
1462 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1463 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1464 			optname != IP_IPSEC_POLICY &&
1465 			optname != IP_XFRM_POLICY &&
1466 			!ip_mroute_opt(optname))
1467 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1468 #endif
1469 	return err;
1470 }
1471 EXPORT_SYMBOL(ip_setsockopt);
1472 
1473 /*
1474  *	Get the options. Note for future reference. The GET of IP options gets
1475  *	the _received_ ones. The set sets the _sent_ ones.
1476  */
1477 
1478 static bool getsockopt_needs_rtnl(int optname)
1479 {
1480 	switch (optname) {
1481 	case IP_MSFILTER:
1482 	case MCAST_MSFILTER:
1483 		return true;
1484 	}
1485 	return false;
1486 }
1487 
1488 static int ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1489 				 sockptr_t optlen, int len)
1490 {
1491 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1492 	struct group_filter gsf;
1493 	int num, gsf_size;
1494 	int err;
1495 
1496 	if (len < size0)
1497 		return -EINVAL;
1498 	if (copy_from_sockptr(&gsf, optval, size0))
1499 		return -EFAULT;
1500 
1501 	num = gsf.gf_numsrc;
1502 	err = ip_mc_gsfget(sk, &gsf, optval,
1503 			   offsetof(struct group_filter, gf_slist_flex));
1504 	if (err)
1505 		return err;
1506 	if (gsf.gf_numsrc < num)
1507 		num = gsf.gf_numsrc;
1508 	gsf_size = GROUP_FILTER_SIZE(num);
1509 	if (copy_to_sockptr(optlen, &gsf_size, sizeof(int)) ||
1510 	    copy_to_sockptr(optval, &gsf, size0))
1511 		return -EFAULT;
1512 	return 0;
1513 }
1514 
1515 static int compat_ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1516 					sockptr_t optlen, int len)
1517 {
1518 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1519 	struct compat_group_filter gf32;
1520 	struct group_filter gf;
1521 	int num;
1522 	int err;
1523 
1524 	if (len < size0)
1525 		return -EINVAL;
1526 	if (copy_from_sockptr(&gf32, optval, size0))
1527 		return -EFAULT;
1528 
1529 	gf.gf_interface = gf32.gf_interface;
1530 	gf.gf_fmode = gf32.gf_fmode;
1531 	num = gf.gf_numsrc = gf32.gf_numsrc;
1532 	gf.gf_group = gf32.gf_group;
1533 
1534 	err = ip_mc_gsfget(sk, &gf, optval,
1535 			   offsetof(struct compat_group_filter, gf_slist_flex));
1536 	if (err)
1537 		return err;
1538 	if (gf.gf_numsrc < num)
1539 		num = gf.gf_numsrc;
1540 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1541 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1542 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1543 				   &gf.gf_fmode, sizeof(gf.gf_fmode)) ||
1544 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1545 				   &gf.gf_numsrc, sizeof(gf.gf_numsrc)))
1546 		return -EFAULT;
1547 	return 0;
1548 }
1549 
1550 int do_ip_getsockopt(struct sock *sk, int level, int optname,
1551 		     sockptr_t optval, sockptr_t optlen)
1552 {
1553 	struct inet_sock *inet = inet_sk(sk);
1554 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1555 	int val, err = 0;
1556 	int len;
1557 
1558 	if (level != SOL_IP)
1559 		return -EOPNOTSUPP;
1560 
1561 	if (ip_mroute_opt(optname))
1562 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1563 
1564 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1565 		return -EFAULT;
1566 	if (len < 0)
1567 		return -EINVAL;
1568 
1569 	if (needs_rtnl)
1570 		rtnl_lock();
1571 	sockopt_lock_sock(sk);
1572 
1573 	switch (optname) {
1574 	case IP_OPTIONS:
1575 	{
1576 		unsigned char optbuf[sizeof(struct ip_options)+40];
1577 		struct ip_options *opt = (struct ip_options *)optbuf;
1578 		struct ip_options_rcu *inet_opt;
1579 
1580 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1581 						     lockdep_sock_is_held(sk));
1582 		opt->optlen = 0;
1583 		if (inet_opt)
1584 			memcpy(optbuf, &inet_opt->opt,
1585 			       sizeof(struct ip_options) +
1586 			       inet_opt->opt.optlen);
1587 		sockopt_release_sock(sk);
1588 
1589 		if (opt->optlen == 0) {
1590 			len = 0;
1591 			return copy_to_sockptr(optlen, &len, sizeof(int));
1592 		}
1593 
1594 		ip_options_undo(opt);
1595 
1596 		len = min_t(unsigned int, len, opt->optlen);
1597 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1598 			return -EFAULT;
1599 		if (copy_to_sockptr(optval, opt->__data, len))
1600 			return -EFAULT;
1601 		return 0;
1602 	}
1603 	case IP_PKTINFO:
1604 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1605 		break;
1606 	case IP_RECVTTL:
1607 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1608 		break;
1609 	case IP_RECVTOS:
1610 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1611 		break;
1612 	case IP_RECVOPTS:
1613 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1614 		break;
1615 	case IP_RETOPTS:
1616 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1617 		break;
1618 	case IP_PASSSEC:
1619 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1620 		break;
1621 	case IP_RECVORIGDSTADDR:
1622 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1623 		break;
1624 	case IP_CHECKSUM:
1625 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1626 		break;
1627 	case IP_RECVFRAGSIZE:
1628 		val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1629 		break;
1630 	case IP_TOS:
1631 		val = inet->tos;
1632 		break;
1633 	case IP_TTL:
1634 	{
1635 		struct net *net = sock_net(sk);
1636 		val = (inet->uc_ttl == -1 ?
1637 		       READ_ONCE(net->ipv4.sysctl_ip_default_ttl) :
1638 		       inet->uc_ttl);
1639 		break;
1640 	}
1641 	case IP_HDRINCL:
1642 		val = inet->hdrincl;
1643 		break;
1644 	case IP_NODEFRAG:
1645 		val = inet->nodefrag;
1646 		break;
1647 	case IP_BIND_ADDRESS_NO_PORT:
1648 		val = inet->bind_address_no_port;
1649 		break;
1650 	case IP_MTU_DISCOVER:
1651 		val = inet->pmtudisc;
1652 		break;
1653 	case IP_MTU:
1654 	{
1655 		struct dst_entry *dst;
1656 		val = 0;
1657 		dst = sk_dst_get(sk);
1658 		if (dst) {
1659 			val = dst_mtu(dst);
1660 			dst_release(dst);
1661 		}
1662 		if (!val) {
1663 			sockopt_release_sock(sk);
1664 			return -ENOTCONN;
1665 		}
1666 		break;
1667 	}
1668 	case IP_RECVERR:
1669 		val = inet->recverr;
1670 		break;
1671 	case IP_RECVERR_RFC4884:
1672 		val = inet->recverr_rfc4884;
1673 		break;
1674 	case IP_MULTICAST_TTL:
1675 		val = inet->mc_ttl;
1676 		break;
1677 	case IP_MULTICAST_LOOP:
1678 		val = inet->mc_loop;
1679 		break;
1680 	case IP_UNICAST_IF:
1681 		val = (__force int)htonl((__u32) inet->uc_index);
1682 		break;
1683 	case IP_MULTICAST_IF:
1684 	{
1685 		struct in_addr addr;
1686 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1687 		addr.s_addr = inet->mc_addr;
1688 		sockopt_release_sock(sk);
1689 
1690 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1691 			return -EFAULT;
1692 		if (copy_to_sockptr(optval, &addr, len))
1693 			return -EFAULT;
1694 		return 0;
1695 	}
1696 	case IP_MSFILTER:
1697 	{
1698 		struct ip_msfilter msf;
1699 
1700 		if (len < IP_MSFILTER_SIZE(0)) {
1701 			err = -EINVAL;
1702 			goto out;
1703 		}
1704 		if (copy_from_sockptr(&msf, optval, IP_MSFILTER_SIZE(0))) {
1705 			err = -EFAULT;
1706 			goto out;
1707 		}
1708 		err = ip_mc_msfget(sk, &msf, optval, optlen);
1709 		goto out;
1710 	}
1711 	case MCAST_MSFILTER:
1712 		if (in_compat_syscall())
1713 			err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1714 							   len);
1715 		else
1716 			err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1717 		goto out;
1718 	case IP_MULTICAST_ALL:
1719 		val = inet->mc_all;
1720 		break;
1721 	case IP_PKTOPTIONS:
1722 	{
1723 		struct msghdr msg;
1724 
1725 		sockopt_release_sock(sk);
1726 
1727 		if (sk->sk_type != SOCK_STREAM)
1728 			return -ENOPROTOOPT;
1729 
1730 		if (optval.is_kernel) {
1731 			msg.msg_control_is_user = false;
1732 			msg.msg_control = optval.kernel;
1733 		} else {
1734 			msg.msg_control_is_user = true;
1735 			msg.msg_control_user = optval.user;
1736 		}
1737 		msg.msg_controllen = len;
1738 		msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1739 
1740 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1741 			struct in_pktinfo info;
1742 
1743 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1744 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1745 			info.ipi_ifindex = inet->mc_index;
1746 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1747 		}
1748 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1749 			int hlim = inet->mc_ttl;
1750 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1751 		}
1752 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1753 			int tos = inet->rcv_tos;
1754 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1755 		}
1756 		len -= msg.msg_controllen;
1757 		return copy_to_sockptr(optlen, &len, sizeof(int));
1758 	}
1759 	case IP_FREEBIND:
1760 		val = inet->freebind;
1761 		break;
1762 	case IP_TRANSPARENT:
1763 		val = inet->transparent;
1764 		break;
1765 	case IP_MINTTL:
1766 		val = inet->min_ttl;
1767 		break;
1768 	case IP_LOCAL_PORT_RANGE:
1769 		val = inet->local_port_range.hi << 16 | inet->local_port_range.lo;
1770 		break;
1771 	case IP_PROTOCOL:
1772 		val = inet_sk(sk)->inet_num;
1773 		break;
1774 	default:
1775 		sockopt_release_sock(sk);
1776 		return -ENOPROTOOPT;
1777 	}
1778 	sockopt_release_sock(sk);
1779 
1780 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1781 		unsigned char ucval = (unsigned char)val;
1782 		len = 1;
1783 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1784 			return -EFAULT;
1785 		if (copy_to_sockptr(optval, &ucval, 1))
1786 			return -EFAULT;
1787 	} else {
1788 		len = min_t(unsigned int, sizeof(int), len);
1789 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1790 			return -EFAULT;
1791 		if (copy_to_sockptr(optval, &val, len))
1792 			return -EFAULT;
1793 	}
1794 	return 0;
1795 
1796 out:
1797 	sockopt_release_sock(sk);
1798 	if (needs_rtnl)
1799 		rtnl_unlock();
1800 	return err;
1801 }
1802 
1803 int ip_getsockopt(struct sock *sk, int level,
1804 		  int optname, char __user *optval, int __user *optlen)
1805 {
1806 	int err;
1807 
1808 	err = do_ip_getsockopt(sk, level, optname,
1809 			       USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1810 
1811 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1812 	if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1813 	    optname < BPFILTER_IPT_GET_MAX)
1814 		err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1815 #endif
1816 #ifdef CONFIG_NETFILTER
1817 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1818 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1819 			!ip_mroute_opt(optname)) {
1820 		int len;
1821 
1822 		if (get_user(len, optlen))
1823 			return -EFAULT;
1824 
1825 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1826 		if (err >= 0)
1827 			err = put_user(len, optlen);
1828 		return err;
1829 	}
1830 #endif
1831 	return err;
1832 }
1833 EXPORT_SYMBOL(ip_getsockopt);
1834