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