• 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 	struct sockaddr_in sin;
137 	const struct iphdr *iph = ip_hdr(skb);
138 	__be16 *ports = (__be16 *)skb_transport_header(skb);
139 
140 	if (skb_transport_offset(skb) + 4 > (int)skb->len)
141 		return;
142 
143 	/* All current transport protocols have the port numbers in the
144 	 * first four bytes of the transport header and this function is
145 	 * written with this assumption in mind.
146 	 */
147 
148 	sin.sin_family = AF_INET;
149 	sin.sin_addr.s_addr = iph->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 sock * sk,struct msghdr * msg,struct ipcm_cookie * ipc,bool allow_ipv6)224 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
225 		 bool allow_ipv6)
226 {
227 	int err, val;
228 	struct cmsghdr *cmsg;
229 	struct net *net = sock_net(sk);
230 
231 	for_each_cmsghdr(cmsg, msg) {
232 		if (!CMSG_OK(msg, cmsg))
233 			return -EINVAL;
234 #if IS_ENABLED(CONFIG_IPV6)
235 		if (allow_ipv6 &&
236 		    cmsg->cmsg_level == SOL_IPV6 &&
237 		    cmsg->cmsg_type == IPV6_PKTINFO) {
238 			struct in6_pktinfo *src_info;
239 
240 			if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
241 				return -EINVAL;
242 			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
243 			if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
244 				return -EINVAL;
245 			if (src_info->ipi6_ifindex)
246 				ipc->oif = src_info->ipi6_ifindex;
247 			ipc->addr = src_info->ipi6_addr.s6_addr32[3];
248 			continue;
249 		}
250 #endif
251 		if (cmsg->cmsg_level == SOL_SOCKET) {
252 			err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
253 			if (err)
254 				return err;
255 			continue;
256 		}
257 
258 		if (cmsg->cmsg_level != SOL_IP)
259 			continue;
260 		switch (cmsg->cmsg_type) {
261 		case IP_RETOPTS:
262 			err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
263 
264 			/* Our caller is responsible for freeing ipc->opt */
265 			err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
266 					     err < 40 ? err : 40);
267 			if (err)
268 				return err;
269 			break;
270 		case IP_PKTINFO:
271 		{
272 			struct in_pktinfo *info;
273 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
274 				return -EINVAL;
275 			info = (struct in_pktinfo *)CMSG_DATA(cmsg);
276 			if (info->ipi_ifindex)
277 				ipc->oif = info->ipi_ifindex;
278 			ipc->addr = info->ipi_spec_dst.s_addr;
279 			break;
280 		}
281 		case IP_TTL:
282 			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
283 				return -EINVAL;
284 			val = *(int *)CMSG_DATA(cmsg);
285 			if (val < 1 || val > 255)
286 				return -EINVAL;
287 			ipc->ttl = val;
288 			break;
289 		case IP_TOS:
290 			if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
291 				val = *(int *)CMSG_DATA(cmsg);
292 			else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
293 				val = *(u8 *)CMSG_DATA(cmsg);
294 			else
295 				return -EINVAL;
296 			if (val < 0 || val > 255)
297 				return -EINVAL;
298 			ipc->tos = val;
299 			ipc->priority = rt_tos2priority(ipc->tos);
300 			break;
301 
302 		default:
303 			return -EINVAL;
304 		}
305 	}
306 	return 0;
307 }
308 
309 
310 /* Special input handler for packets caught by router alert option.
311    They are selected only by protocol field, and then processed likely
312    local ones; but only if someone wants them! Otherwise, router
313    not running rsvpd will kill RSVP.
314 
315    It is user level problem, what it will make with them.
316    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
317    but receiver should be enough clever f.e. to forward mtrace requests,
318    sent to multicast group to reach destination designated router.
319  */
320 struct ip_ra_chain __rcu *ip_ra_chain;
321 static DEFINE_SPINLOCK(ip_ra_lock);
322 
323 
ip_ra_destroy_rcu(struct rcu_head * head)324 static void ip_ra_destroy_rcu(struct rcu_head *head)
325 {
326 	struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
327 
328 	sock_put(ra->saved_sk);
329 	kfree(ra);
330 }
331 
ip_ra_control(struct sock * sk,unsigned char on,void (* destructor)(struct sock *))332 int ip_ra_control(struct sock *sk, unsigned char on,
333 		  void (*destructor)(struct sock *))
334 {
335 	struct ip_ra_chain *ra, *new_ra;
336 	struct ip_ra_chain __rcu **rap;
337 
338 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
339 		return -EINVAL;
340 
341 	new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
342 
343 	spin_lock_bh(&ip_ra_lock);
344 	for (rap = &ip_ra_chain;
345 	     (ra = rcu_dereference_protected(*rap,
346 			lockdep_is_held(&ip_ra_lock))) != NULL;
347 	     rap = &ra->next) {
348 		if (ra->sk == sk) {
349 			if (on) {
350 				spin_unlock_bh(&ip_ra_lock);
351 				kfree(new_ra);
352 				return -EADDRINUSE;
353 			}
354 			/* dont let ip_call_ra_chain() use sk again */
355 			ra->sk = NULL;
356 			RCU_INIT_POINTER(*rap, ra->next);
357 			spin_unlock_bh(&ip_ra_lock);
358 
359 			if (ra->destructor)
360 				ra->destructor(sk);
361 			/*
362 			 * Delay sock_put(sk) and kfree(ra) after one rcu grace
363 			 * period. This guarantee ip_call_ra_chain() dont need
364 			 * to mess with socket refcounts.
365 			 */
366 			ra->saved_sk = sk;
367 			call_rcu(&ra->rcu, ip_ra_destroy_rcu);
368 			return 0;
369 		}
370 	}
371 	if (!new_ra) {
372 		spin_unlock_bh(&ip_ra_lock);
373 		return -ENOBUFS;
374 	}
375 	new_ra->sk = sk;
376 	new_ra->destructor = destructor;
377 
378 	RCU_INIT_POINTER(new_ra->next, ra);
379 	rcu_assign_pointer(*rap, new_ra);
380 	sock_hold(sk);
381 	spin_unlock_bh(&ip_ra_lock);
382 
383 	return 0;
384 }
385 
ip_icmp_error(struct sock * sk,struct sk_buff * skb,int err,__be16 port,u32 info,u8 * payload)386 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
387 		   __be16 port, u32 info, u8 *payload)
388 {
389 	struct sock_exterr_skb *serr;
390 
391 	skb = skb_clone(skb, GFP_ATOMIC);
392 	if (!skb)
393 		return;
394 
395 	serr = SKB_EXT_ERR(skb);
396 	serr->ee.ee_errno = err;
397 	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
398 	serr->ee.ee_type = icmp_hdr(skb)->type;
399 	serr->ee.ee_code = icmp_hdr(skb)->code;
400 	serr->ee.ee_pad = 0;
401 	serr->ee.ee_info = info;
402 	serr->ee.ee_data = 0;
403 	serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
404 				   skb_network_header(skb);
405 	serr->port = port;
406 
407 	if (skb_pull(skb, payload - skb->data)) {
408 		skb_reset_transport_header(skb);
409 		if (sock_queue_err_skb(sk, skb) == 0)
410 			return;
411 	}
412 	kfree_skb(skb);
413 }
414 
ip_local_error(struct sock * sk,int err,__be32 daddr,__be16 port,u32 info)415 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
416 {
417 	struct inet_sock *inet = inet_sk(sk);
418 	struct sock_exterr_skb *serr;
419 	struct iphdr *iph;
420 	struct sk_buff *skb;
421 
422 	if (!inet->recverr)
423 		return;
424 
425 	skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
426 	if (!skb)
427 		return;
428 
429 	skb_put(skb, sizeof(struct iphdr));
430 	skb_reset_network_header(skb);
431 	iph = ip_hdr(skb);
432 	iph->daddr = daddr;
433 
434 	serr = SKB_EXT_ERR(skb);
435 	serr->ee.ee_errno = err;
436 	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
437 	serr->ee.ee_type = 0;
438 	serr->ee.ee_code = 0;
439 	serr->ee.ee_pad = 0;
440 	serr->ee.ee_info = info;
441 	serr->ee.ee_data = 0;
442 	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
443 	serr->port = port;
444 
445 	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
446 	skb_reset_transport_header(skb);
447 
448 	if (sock_queue_err_skb(sk, skb))
449 		kfree_skb(skb);
450 }
451 
452 /* For some errors we have valid addr_offset even with zero payload and
453  * zero port. Also, addr_offset should be supported if port is set.
454  */
ipv4_datagram_support_addr(struct sock_exterr_skb * serr)455 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
456 {
457 	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
458 	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
459 }
460 
461 /* IPv4 supports cmsg on all imcp errors and some timestamps
462  *
463  * Timestamp code paths do not initialize the fields expected by cmsg:
464  * the PKTINFO fields in skb->cb[]. Fill those in here.
465  */
ipv4_datagram_support_cmsg(const struct sock * sk,struct sk_buff * skb,int ee_origin)466 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
467 				       struct sk_buff *skb,
468 				       int ee_origin)
469 {
470 	struct in_pktinfo *info;
471 
472 	if (ee_origin == SO_EE_ORIGIN_ICMP)
473 		return true;
474 
475 	if (ee_origin == SO_EE_ORIGIN_LOCAL)
476 		return false;
477 
478 	/* Support IP_PKTINFO on tstamp packets if requested, to correlate
479 	 * timestamp with egress dev. Not possible for packets without iif
480 	 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
481 	 */
482 	info = PKTINFO_SKB_CB(skb);
483 	if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
484 	    !info->ipi_ifindex)
485 		return false;
486 
487 	info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
488 	return true;
489 }
490 
491 /*
492  *	Handle MSG_ERRQUEUE
493  */
ip_recv_error(struct sock * sk,struct msghdr * msg,int len,int * addr_len)494 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
495 {
496 	struct sock_exterr_skb *serr;
497 	struct sk_buff *skb;
498 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
499 	struct {
500 		struct sock_extended_err ee;
501 		struct sockaddr_in	 offender;
502 	} errhdr;
503 	int err;
504 	int copied;
505 
506 	WARN_ON_ONCE(sk->sk_family == AF_INET6);
507 
508 	err = -EAGAIN;
509 	skb = sock_dequeue_err_skb(sk);
510 	if (!skb)
511 		goto out;
512 
513 	copied = skb->len;
514 	if (copied > len) {
515 		msg->msg_flags |= MSG_TRUNC;
516 		copied = len;
517 	}
518 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
519 	if (unlikely(err)) {
520 		kfree_skb(skb);
521 		return err;
522 	}
523 	sock_recv_timestamp(msg, sk, skb);
524 
525 	serr = SKB_EXT_ERR(skb);
526 
527 	if (sin && ipv4_datagram_support_addr(serr)) {
528 		sin->sin_family = AF_INET;
529 		sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
530 						   serr->addr_offset);
531 		sin->sin_port = serr->port;
532 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
533 		*addr_len = sizeof(*sin);
534 	}
535 
536 	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
537 	sin = &errhdr.offender;
538 	memset(sin, 0, sizeof(*sin));
539 
540 	if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
541 		sin->sin_family = AF_INET;
542 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
543 		if (inet_sk(sk)->cmsg_flags)
544 			ip_cmsg_recv(msg, skb);
545 	}
546 
547 	put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
548 
549 	/* Now we could try to dump offended packet options */
550 
551 	msg->msg_flags |= MSG_ERRQUEUE;
552 	err = copied;
553 
554 	consume_skb(skb);
555 out:
556 	return err;
557 }
558 
559 
560 /*
561  *	Socket option code for IP. This is the end of the line after any
562  *	TCP,UDP etc options on an IP socket.
563  */
setsockopt_needs_rtnl(int optname)564 static bool setsockopt_needs_rtnl(int optname)
565 {
566 	switch (optname) {
567 	case IP_ADD_MEMBERSHIP:
568 	case IP_ADD_SOURCE_MEMBERSHIP:
569 	case IP_BLOCK_SOURCE:
570 	case IP_DROP_MEMBERSHIP:
571 	case IP_DROP_SOURCE_MEMBERSHIP:
572 	case IP_MSFILTER:
573 	case IP_UNBLOCK_SOURCE:
574 	case MCAST_BLOCK_SOURCE:
575 	case MCAST_MSFILTER:
576 	case MCAST_JOIN_GROUP:
577 	case MCAST_JOIN_SOURCE_GROUP:
578 	case MCAST_LEAVE_GROUP:
579 	case MCAST_LEAVE_SOURCE_GROUP:
580 	case MCAST_UNBLOCK_SOURCE:
581 		return true;
582 	}
583 	return false;
584 }
585 
do_ip_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)586 static int do_ip_setsockopt(struct sock *sk, int level,
587 			    int optname, char __user *optval, unsigned int optlen)
588 {
589 	struct inet_sock *inet = inet_sk(sk);
590 	struct net *net = sock_net(sk);
591 	int val = 0, err;
592 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
593 
594 	switch (optname) {
595 	case IP_PKTINFO:
596 	case IP_RECVTTL:
597 	case IP_RECVOPTS:
598 	case IP_RECVTOS:
599 	case IP_RETOPTS:
600 	case IP_TOS:
601 	case IP_TTL:
602 	case IP_HDRINCL:
603 	case IP_MTU_DISCOVER:
604 	case IP_RECVERR:
605 	case IP_ROUTER_ALERT:
606 	case IP_FREEBIND:
607 	case IP_PASSSEC:
608 	case IP_TRANSPARENT:
609 	case IP_MINTTL:
610 	case IP_NODEFRAG:
611 	case IP_BIND_ADDRESS_NO_PORT:
612 	case IP_UNICAST_IF:
613 	case IP_MULTICAST_TTL:
614 	case IP_MULTICAST_ALL:
615 	case IP_MULTICAST_LOOP:
616 	case IP_RECVORIGDSTADDR:
617 	case IP_CHECKSUM:
618 		if (optlen >= sizeof(int)) {
619 			if (get_user(val, (int __user *) optval))
620 				return -EFAULT;
621 		} else if (optlen >= sizeof(char)) {
622 			unsigned char ucval;
623 
624 			if (get_user(ucval, (unsigned char __user *) optval))
625 				return -EFAULT;
626 			val = (int) ucval;
627 		}
628 	}
629 
630 	/* If optlen==0, it is equivalent to val == 0 */
631 
632 	if (ip_mroute_opt(optname))
633 		return ip_mroute_setsockopt(sk, optname, optval, optlen);
634 
635 	err = 0;
636 	if (needs_rtnl)
637 		rtnl_lock();
638 	lock_sock(sk);
639 
640 	switch (optname) {
641 	case IP_OPTIONS:
642 	{
643 		struct ip_options_rcu *old, *opt = NULL;
644 
645 		if (optlen > 40)
646 			goto e_inval;
647 		err = ip_options_get_from_user(sock_net(sk), &opt,
648 					       optval, optlen);
649 		if (err)
650 			break;
651 		old = rcu_dereference_protected(inet->inet_opt,
652 						lockdep_sock_is_held(sk));
653 		if (inet->is_icsk) {
654 			struct inet_connection_sock *icsk = inet_csk(sk);
655 #if IS_ENABLED(CONFIG_IPV6)
656 			if (sk->sk_family == PF_INET ||
657 			    (!((1 << sk->sk_state) &
658 			       (TCPF_LISTEN | TCPF_CLOSE)) &&
659 			     inet->inet_daddr != LOOPBACK4_IPV6)) {
660 #endif
661 				if (old)
662 					icsk->icsk_ext_hdr_len -= old->opt.optlen;
663 				if (opt)
664 					icsk->icsk_ext_hdr_len += opt->opt.optlen;
665 				icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
666 #if IS_ENABLED(CONFIG_IPV6)
667 			}
668 #endif
669 		}
670 		rcu_assign_pointer(inet->inet_opt, opt);
671 		if (old)
672 			kfree_rcu(old, rcu);
673 		break;
674 	}
675 	case IP_PKTINFO:
676 		if (val)
677 			inet->cmsg_flags |= IP_CMSG_PKTINFO;
678 		else
679 			inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
680 		break;
681 	case IP_RECVTTL:
682 		if (val)
683 			inet->cmsg_flags |=  IP_CMSG_TTL;
684 		else
685 			inet->cmsg_flags &= ~IP_CMSG_TTL;
686 		break;
687 	case IP_RECVTOS:
688 		if (val)
689 			inet->cmsg_flags |=  IP_CMSG_TOS;
690 		else
691 			inet->cmsg_flags &= ~IP_CMSG_TOS;
692 		break;
693 	case IP_RECVOPTS:
694 		if (val)
695 			inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
696 		else
697 			inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
698 		break;
699 	case IP_RETOPTS:
700 		if (val)
701 			inet->cmsg_flags |= IP_CMSG_RETOPTS;
702 		else
703 			inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
704 		break;
705 	case IP_PASSSEC:
706 		if (val)
707 			inet->cmsg_flags |= IP_CMSG_PASSSEC;
708 		else
709 			inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
710 		break;
711 	case IP_RECVORIGDSTADDR:
712 		if (val)
713 			inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
714 		else
715 			inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
716 		break;
717 	case IP_CHECKSUM:
718 		if (val) {
719 			if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
720 				inet_inc_convert_csum(sk);
721 				inet->cmsg_flags |= IP_CMSG_CHECKSUM;
722 			}
723 		} else {
724 			if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
725 				inet_dec_convert_csum(sk);
726 				inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
727 			}
728 		}
729 		break;
730 	case IP_TOS:	/* This sets both TOS and Precedence */
731 		if (sk->sk_type == SOCK_STREAM) {
732 			val &= ~INET_ECN_MASK;
733 			val |= inet->tos & INET_ECN_MASK;
734 		}
735 		if (inet->tos != val) {
736 			inet->tos = val;
737 			sk->sk_priority = rt_tos2priority(val);
738 			sk_dst_reset(sk);
739 		}
740 		break;
741 	case IP_TTL:
742 		if (optlen < 1)
743 			goto e_inval;
744 		if (val != -1 && (val < 1 || val > 255))
745 			goto e_inval;
746 		inet->uc_ttl = val;
747 		break;
748 	case IP_HDRINCL:
749 		if (sk->sk_type != SOCK_RAW) {
750 			err = -ENOPROTOOPT;
751 			break;
752 		}
753 		inet->hdrincl = val ? 1 : 0;
754 		break;
755 	case IP_NODEFRAG:
756 		if (sk->sk_type != SOCK_RAW) {
757 			err = -ENOPROTOOPT;
758 			break;
759 		}
760 		inet->nodefrag = val ? 1 : 0;
761 		break;
762 	case IP_BIND_ADDRESS_NO_PORT:
763 		inet->bind_address_no_port = val ? 1 : 0;
764 		break;
765 	case IP_MTU_DISCOVER:
766 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
767 			goto e_inval;
768 		inet->pmtudisc = val;
769 		break;
770 	case IP_RECVERR:
771 		inet->recverr = !!val;
772 		if (!val)
773 			skb_queue_purge(&sk->sk_error_queue);
774 		break;
775 	case IP_MULTICAST_TTL:
776 		if (sk->sk_type == SOCK_STREAM)
777 			goto e_inval;
778 		if (optlen < 1)
779 			goto e_inval;
780 		if (val == -1)
781 			val = 1;
782 		if (val < 0 || val > 255)
783 			goto e_inval;
784 		inet->mc_ttl = val;
785 		break;
786 	case IP_MULTICAST_LOOP:
787 		if (optlen < 1)
788 			goto e_inval;
789 		inet->mc_loop = !!val;
790 		break;
791 	case IP_UNICAST_IF:
792 	{
793 		struct net_device *dev = NULL;
794 		int ifindex;
795 
796 		if (optlen != sizeof(int))
797 			goto e_inval;
798 
799 		ifindex = (__force int)ntohl((__force __be32)val);
800 		if (ifindex == 0) {
801 			inet->uc_index = 0;
802 			err = 0;
803 			break;
804 		}
805 
806 		dev = dev_get_by_index(sock_net(sk), ifindex);
807 		err = -EADDRNOTAVAIL;
808 		if (!dev)
809 			break;
810 		dev_put(dev);
811 
812 		err = -EINVAL;
813 		if (sk->sk_bound_dev_if)
814 			break;
815 
816 		inet->uc_index = ifindex;
817 		err = 0;
818 		break;
819 	}
820 	case IP_MULTICAST_IF:
821 	{
822 		struct ip_mreqn mreq;
823 		struct net_device *dev = NULL;
824 		int midx;
825 
826 		if (sk->sk_type == SOCK_STREAM)
827 			goto e_inval;
828 		/*
829 		 *	Check the arguments are allowable
830 		 */
831 
832 		if (optlen < sizeof(struct in_addr))
833 			goto e_inval;
834 
835 		err = -EFAULT;
836 		if (optlen >= sizeof(struct ip_mreqn)) {
837 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
838 				break;
839 		} else {
840 			memset(&mreq, 0, sizeof(mreq));
841 			if (optlen >= sizeof(struct ip_mreq)) {
842 				if (copy_from_user(&mreq, optval,
843 						   sizeof(struct ip_mreq)))
844 					break;
845 			} else if (optlen >= sizeof(struct in_addr)) {
846 				if (copy_from_user(&mreq.imr_address, optval,
847 						   sizeof(struct in_addr)))
848 					break;
849 			}
850 		}
851 
852 		if (!mreq.imr_ifindex) {
853 			if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
854 				inet->mc_index = 0;
855 				inet->mc_addr  = 0;
856 				err = 0;
857 				break;
858 			}
859 			dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
860 			if (dev)
861 				mreq.imr_ifindex = dev->ifindex;
862 		} else
863 			dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
864 
865 
866 		err = -EADDRNOTAVAIL;
867 		if (!dev)
868 			break;
869 
870 		midx = l3mdev_master_ifindex(dev);
871 
872 		dev_put(dev);
873 
874 		err = -EINVAL;
875 		if (sk->sk_bound_dev_if &&
876 		    mreq.imr_ifindex != sk->sk_bound_dev_if &&
877 		    (!midx || midx != sk->sk_bound_dev_if))
878 			break;
879 
880 		inet->mc_index = mreq.imr_ifindex;
881 		inet->mc_addr  = mreq.imr_address.s_addr;
882 		err = 0;
883 		break;
884 	}
885 
886 	case IP_ADD_MEMBERSHIP:
887 	case IP_DROP_MEMBERSHIP:
888 	{
889 		struct ip_mreqn mreq;
890 
891 		err = -EPROTO;
892 		if (inet_sk(sk)->is_icsk)
893 			break;
894 
895 		if (optlen < sizeof(struct ip_mreq))
896 			goto e_inval;
897 		err = -EFAULT;
898 		if (optlen >= sizeof(struct ip_mreqn)) {
899 			if (copy_from_user(&mreq, optval, sizeof(mreq)))
900 				break;
901 		} else {
902 			memset(&mreq, 0, sizeof(mreq));
903 			if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
904 				break;
905 		}
906 
907 		if (optname == IP_ADD_MEMBERSHIP)
908 			err = ip_mc_join_group(sk, &mreq);
909 		else
910 			err = ip_mc_leave_group(sk, &mreq);
911 		break;
912 	}
913 	case IP_MSFILTER:
914 	{
915 		struct ip_msfilter *msf;
916 
917 		if (optlen < IP_MSFILTER_SIZE(0))
918 			goto e_inval;
919 		if (optlen > sysctl_optmem_max) {
920 			err = -ENOBUFS;
921 			break;
922 		}
923 		msf = kmalloc(optlen, GFP_KERNEL);
924 		if (!msf) {
925 			err = -ENOBUFS;
926 			break;
927 		}
928 		err = -EFAULT;
929 		if (copy_from_user(msf, optval, optlen)) {
930 			kfree(msf);
931 			break;
932 		}
933 		/* numsrc >= (1G-4) overflow in 32 bits */
934 		if (msf->imsf_numsrc >= 0x3ffffffcU ||
935 		    msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
936 			kfree(msf);
937 			err = -ENOBUFS;
938 			break;
939 		}
940 		if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
941 			kfree(msf);
942 			err = -EINVAL;
943 			break;
944 		}
945 		err = ip_mc_msfilter(sk, msf, 0);
946 		kfree(msf);
947 		break;
948 	}
949 	case IP_BLOCK_SOURCE:
950 	case IP_UNBLOCK_SOURCE:
951 	case IP_ADD_SOURCE_MEMBERSHIP:
952 	case IP_DROP_SOURCE_MEMBERSHIP:
953 	{
954 		struct ip_mreq_source mreqs;
955 		int omode, add;
956 
957 		if (optlen != sizeof(struct ip_mreq_source))
958 			goto e_inval;
959 		if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
960 			err = -EFAULT;
961 			break;
962 		}
963 		if (optname == IP_BLOCK_SOURCE) {
964 			omode = MCAST_EXCLUDE;
965 			add = 1;
966 		} else if (optname == IP_UNBLOCK_SOURCE) {
967 			omode = MCAST_EXCLUDE;
968 			add = 0;
969 		} else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
970 			struct ip_mreqn mreq;
971 
972 			mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
973 			mreq.imr_address.s_addr = mreqs.imr_interface;
974 			mreq.imr_ifindex = 0;
975 			err = ip_mc_join_group(sk, &mreq);
976 			if (err && err != -EADDRINUSE)
977 				break;
978 			omode = MCAST_INCLUDE;
979 			add = 1;
980 		} else /* IP_DROP_SOURCE_MEMBERSHIP */ {
981 			omode = MCAST_INCLUDE;
982 			add = 0;
983 		}
984 		err = ip_mc_source(add, omode, sk, &mreqs, 0);
985 		break;
986 	}
987 	case MCAST_JOIN_GROUP:
988 	case MCAST_LEAVE_GROUP:
989 	{
990 		struct group_req greq;
991 		struct sockaddr_in *psin;
992 		struct ip_mreqn mreq;
993 
994 		if (optlen < sizeof(struct group_req))
995 			goto e_inval;
996 		err = -EFAULT;
997 		if (copy_from_user(&greq, optval, sizeof(greq)))
998 			break;
999 		psin = (struct sockaddr_in *)&greq.gr_group;
1000 		if (psin->sin_family != AF_INET)
1001 			goto e_inval;
1002 		memset(&mreq, 0, sizeof(mreq));
1003 		mreq.imr_multiaddr = psin->sin_addr;
1004 		mreq.imr_ifindex = greq.gr_interface;
1005 
1006 		if (optname == MCAST_JOIN_GROUP)
1007 			err = ip_mc_join_group(sk, &mreq);
1008 		else
1009 			err = ip_mc_leave_group(sk, &mreq);
1010 		break;
1011 	}
1012 	case MCAST_JOIN_SOURCE_GROUP:
1013 	case MCAST_LEAVE_SOURCE_GROUP:
1014 	case MCAST_BLOCK_SOURCE:
1015 	case MCAST_UNBLOCK_SOURCE:
1016 	{
1017 		struct group_source_req greqs;
1018 		struct ip_mreq_source mreqs;
1019 		struct sockaddr_in *psin;
1020 		int omode, add;
1021 
1022 		if (optlen != sizeof(struct group_source_req))
1023 			goto e_inval;
1024 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1025 			err = -EFAULT;
1026 			break;
1027 		}
1028 		if (greqs.gsr_group.ss_family != AF_INET ||
1029 		    greqs.gsr_source.ss_family != AF_INET) {
1030 			err = -EADDRNOTAVAIL;
1031 			break;
1032 		}
1033 		psin = (struct sockaddr_in *)&greqs.gsr_group;
1034 		mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1035 		psin = (struct sockaddr_in *)&greqs.gsr_source;
1036 		mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1037 		mreqs.imr_interface = 0; /* use index for mc_source */
1038 
1039 		if (optname == MCAST_BLOCK_SOURCE) {
1040 			omode = MCAST_EXCLUDE;
1041 			add = 1;
1042 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
1043 			omode = MCAST_EXCLUDE;
1044 			add = 0;
1045 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1046 			struct ip_mreqn mreq;
1047 
1048 			psin = (struct sockaddr_in *)&greqs.gsr_group;
1049 			mreq.imr_multiaddr = psin->sin_addr;
1050 			mreq.imr_address.s_addr = 0;
1051 			mreq.imr_ifindex = greqs.gsr_interface;
1052 			err = ip_mc_join_group(sk, &mreq);
1053 			if (err && err != -EADDRINUSE)
1054 				break;
1055 			greqs.gsr_interface = mreq.imr_ifindex;
1056 			omode = MCAST_INCLUDE;
1057 			add = 1;
1058 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
1059 			omode = MCAST_INCLUDE;
1060 			add = 0;
1061 		}
1062 		err = ip_mc_source(add, omode, sk, &mreqs,
1063 				   greqs.gsr_interface);
1064 		break;
1065 	}
1066 	case MCAST_MSFILTER:
1067 	{
1068 		struct sockaddr_in *psin;
1069 		struct ip_msfilter *msf = NULL;
1070 		struct group_filter *gsf = NULL;
1071 		int msize, i, ifindex;
1072 
1073 		if (optlen < GROUP_FILTER_SIZE(0))
1074 			goto e_inval;
1075 		if (optlen > sysctl_optmem_max) {
1076 			err = -ENOBUFS;
1077 			break;
1078 		}
1079 		gsf = kmalloc(optlen, GFP_KERNEL);
1080 		if (!gsf) {
1081 			err = -ENOBUFS;
1082 			break;
1083 		}
1084 		err = -EFAULT;
1085 		if (copy_from_user(gsf, optval, optlen))
1086 			goto mc_msf_out;
1087 
1088 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
1089 		if (gsf->gf_numsrc >= 0x1ffffff ||
1090 		    gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1091 			err = -ENOBUFS;
1092 			goto mc_msf_out;
1093 		}
1094 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1095 			err = -EINVAL;
1096 			goto mc_msf_out;
1097 		}
1098 		msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1099 		msf = kmalloc(msize, GFP_KERNEL);
1100 		if (!msf) {
1101 			err = -ENOBUFS;
1102 			goto mc_msf_out;
1103 		}
1104 		ifindex = gsf->gf_interface;
1105 		psin = (struct sockaddr_in *)&gsf->gf_group;
1106 		if (psin->sin_family != AF_INET) {
1107 			err = -EADDRNOTAVAIL;
1108 			goto mc_msf_out;
1109 		}
1110 		msf->imsf_multiaddr = psin->sin_addr.s_addr;
1111 		msf->imsf_interface = 0;
1112 		msf->imsf_fmode = gsf->gf_fmode;
1113 		msf->imsf_numsrc = gsf->gf_numsrc;
1114 		err = -EADDRNOTAVAIL;
1115 		for (i = 0; i < gsf->gf_numsrc; ++i) {
1116 			psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1117 
1118 			if (psin->sin_family != AF_INET)
1119 				goto mc_msf_out;
1120 			msf->imsf_slist[i] = psin->sin_addr.s_addr;
1121 		}
1122 		kfree(gsf);
1123 		gsf = NULL;
1124 
1125 		err = ip_mc_msfilter(sk, msf, ifindex);
1126 mc_msf_out:
1127 		kfree(msf);
1128 		kfree(gsf);
1129 		break;
1130 	}
1131 	case IP_MULTICAST_ALL:
1132 		if (optlen < 1)
1133 			goto e_inval;
1134 		if (val != 0 && val != 1)
1135 			goto e_inval;
1136 		inet->mc_all = val;
1137 		break;
1138 	case IP_ROUTER_ALERT:
1139 		err = ip_ra_control(sk, val ? 1 : 0, NULL);
1140 		break;
1141 
1142 	case IP_FREEBIND:
1143 		if (optlen < 1)
1144 			goto e_inval;
1145 		inet->freebind = !!val;
1146 		break;
1147 
1148 	case IP_IPSEC_POLICY:
1149 	case IP_XFRM_POLICY:
1150 		err = -EPERM;
1151 		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1152 			break;
1153 		err = xfrm_user_policy(sk, optname, optval, optlen);
1154 		break;
1155 
1156 	case IP_TRANSPARENT:
1157 		if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1158 		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1159 			err = -EPERM;
1160 			break;
1161 		}
1162 		if (optlen < 1)
1163 			goto e_inval;
1164 		inet->transparent = !!val;
1165 		break;
1166 
1167 	case IP_MINTTL:
1168 		if (optlen < 1)
1169 			goto e_inval;
1170 		if (val < 0 || val > 255)
1171 			goto e_inval;
1172 		inet->min_ttl = val;
1173 		break;
1174 
1175 	default:
1176 		err = -ENOPROTOOPT;
1177 		break;
1178 	}
1179 	release_sock(sk);
1180 	if (needs_rtnl)
1181 		rtnl_unlock();
1182 	return err;
1183 
1184 e_inval:
1185 	release_sock(sk);
1186 	if (needs_rtnl)
1187 		rtnl_unlock();
1188 	return -EINVAL;
1189 }
1190 
1191 /**
1192  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1193  * @sk: socket
1194  * @skb: buffer
1195  *
1196  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1197  * destination in skb->cb[] before dst drop.
1198  * This way, receiver doesn't make cache line misses to read rtable.
1199  */
ipv4_pktinfo_prepare(const struct sock * sk,struct sk_buff * skb)1200 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1201 {
1202 	struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1203 	bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1204 		       ipv6_sk_rxinfo(sk);
1205 
1206 	if (prepare && skb_rtable(skb)) {
1207 		/* skb->cb is overloaded: prior to this point it is IP{6}CB
1208 		 * which has interface index (iif) as the first member of the
1209 		 * underlying inet{6}_skb_parm struct. This code then overlays
1210 		 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1211 		 * element so the iif is picked up from the prior IPCB. If iif
1212 		 * is the loopback interface, then return the sending interface
1213 		 * (e.g., process binds socket to eth0 for Tx which is
1214 		 * redirected to loopback in the rtable/dst).
1215 		 */
1216 		if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1217 			pktinfo->ipi_ifindex = inet_iif(skb);
1218 
1219 		pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1220 	} else {
1221 		pktinfo->ipi_ifindex = 0;
1222 		pktinfo->ipi_spec_dst.s_addr = 0;
1223 	}
1224 	/* We need to keep the dst for __ip_options_echo()
1225 	 * We could restrict the test to opt.ts_needtime || opt.srr,
1226 	 * but the following is good enough as IP options are not often used.
1227 	 */
1228 	if (unlikely(IPCB(skb)->opt.optlen))
1229 		skb_dst_force(skb);
1230 	else
1231 		skb_dst_drop(skb);
1232 }
1233 
ip_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1234 int ip_setsockopt(struct sock *sk, int level,
1235 		int optname, char __user *optval, unsigned int optlen)
1236 {
1237 	int err;
1238 
1239 	if (level != SOL_IP)
1240 		return -ENOPROTOOPT;
1241 
1242 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1243 #ifdef CONFIG_NETFILTER
1244 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1245 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1246 			optname != IP_IPSEC_POLICY &&
1247 			optname != IP_XFRM_POLICY &&
1248 			!ip_mroute_opt(optname))
1249 		err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1250 #endif
1251 	return err;
1252 }
1253 EXPORT_SYMBOL(ip_setsockopt);
1254 
1255 #ifdef CONFIG_COMPAT
compat_ip_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1256 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1257 			 char __user *optval, unsigned int optlen)
1258 {
1259 	int err;
1260 
1261 	if (level != SOL_IP)
1262 		return -ENOPROTOOPT;
1263 
1264 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1265 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1266 			ip_setsockopt);
1267 
1268 	err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1269 #ifdef CONFIG_NETFILTER
1270 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1271 	if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1272 			optname != IP_IPSEC_POLICY &&
1273 			optname != IP_XFRM_POLICY &&
1274 			!ip_mroute_opt(optname))
1275 		err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1276 					   optlen);
1277 #endif
1278 	return err;
1279 }
1280 EXPORT_SYMBOL(compat_ip_setsockopt);
1281 #endif
1282 
1283 /*
1284  *	Get the options. Note for future reference. The GET of IP options gets
1285  *	the _received_ ones. The set sets the _sent_ ones.
1286  */
1287 
getsockopt_needs_rtnl(int optname)1288 static bool getsockopt_needs_rtnl(int optname)
1289 {
1290 	switch (optname) {
1291 	case IP_MSFILTER:
1292 	case MCAST_MSFILTER:
1293 		return true;
1294 	}
1295 	return false;
1296 }
1297 
do_ip_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen,unsigned int flags)1298 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1299 			    char __user *optval, int __user *optlen, unsigned int flags)
1300 {
1301 	struct inet_sock *inet = inet_sk(sk);
1302 	bool needs_rtnl = getsockopt_needs_rtnl(optname);
1303 	int val, err = 0;
1304 	int len;
1305 
1306 	if (level != SOL_IP)
1307 		return -EOPNOTSUPP;
1308 
1309 	if (ip_mroute_opt(optname))
1310 		return ip_mroute_getsockopt(sk, optname, optval, optlen);
1311 
1312 	if (get_user(len, optlen))
1313 		return -EFAULT;
1314 	if (len < 0)
1315 		return -EINVAL;
1316 
1317 	if (needs_rtnl)
1318 		rtnl_lock();
1319 	lock_sock(sk);
1320 
1321 	switch (optname) {
1322 	case IP_OPTIONS:
1323 	{
1324 		unsigned char optbuf[sizeof(struct ip_options)+40];
1325 		struct ip_options *opt = (struct ip_options *)optbuf;
1326 		struct ip_options_rcu *inet_opt;
1327 
1328 		inet_opt = rcu_dereference_protected(inet->inet_opt,
1329 						     lockdep_sock_is_held(sk));
1330 		opt->optlen = 0;
1331 		if (inet_opt)
1332 			memcpy(optbuf, &inet_opt->opt,
1333 			       sizeof(struct ip_options) +
1334 			       inet_opt->opt.optlen);
1335 		release_sock(sk);
1336 
1337 		if (opt->optlen == 0)
1338 			return put_user(0, optlen);
1339 
1340 		ip_options_undo(opt);
1341 
1342 		len = min_t(unsigned int, len, opt->optlen);
1343 		if (put_user(len, optlen))
1344 			return -EFAULT;
1345 		if (copy_to_user(optval, opt->__data, len))
1346 			return -EFAULT;
1347 		return 0;
1348 	}
1349 	case IP_PKTINFO:
1350 		val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1351 		break;
1352 	case IP_RECVTTL:
1353 		val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1354 		break;
1355 	case IP_RECVTOS:
1356 		val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1357 		break;
1358 	case IP_RECVOPTS:
1359 		val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1360 		break;
1361 	case IP_RETOPTS:
1362 		val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1363 		break;
1364 	case IP_PASSSEC:
1365 		val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1366 		break;
1367 	case IP_RECVORIGDSTADDR:
1368 		val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1369 		break;
1370 	case IP_CHECKSUM:
1371 		val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1372 		break;
1373 	case IP_TOS:
1374 		val = inet->tos;
1375 		break;
1376 	case IP_TTL:
1377 	{
1378 		struct net *net = sock_net(sk);
1379 		val = (inet->uc_ttl == -1 ?
1380 		       net->ipv4.sysctl_ip_default_ttl :
1381 		       inet->uc_ttl);
1382 		break;
1383 	}
1384 	case IP_HDRINCL:
1385 		val = inet->hdrincl;
1386 		break;
1387 	case IP_NODEFRAG:
1388 		val = inet->nodefrag;
1389 		break;
1390 	case IP_BIND_ADDRESS_NO_PORT:
1391 		val = inet->bind_address_no_port;
1392 		break;
1393 	case IP_MTU_DISCOVER:
1394 		val = inet->pmtudisc;
1395 		break;
1396 	case IP_MTU:
1397 	{
1398 		struct dst_entry *dst;
1399 		val = 0;
1400 		dst = sk_dst_get(sk);
1401 		if (dst) {
1402 			val = dst_mtu(dst);
1403 			dst_release(dst);
1404 		}
1405 		if (!val) {
1406 			release_sock(sk);
1407 			return -ENOTCONN;
1408 		}
1409 		break;
1410 	}
1411 	case IP_RECVERR:
1412 		val = inet->recverr;
1413 		break;
1414 	case IP_MULTICAST_TTL:
1415 		val = inet->mc_ttl;
1416 		break;
1417 	case IP_MULTICAST_LOOP:
1418 		val = inet->mc_loop;
1419 		break;
1420 	case IP_UNICAST_IF:
1421 		val = (__force int)htonl((__u32) inet->uc_index);
1422 		break;
1423 	case IP_MULTICAST_IF:
1424 	{
1425 		struct in_addr addr;
1426 		len = min_t(unsigned int, len, sizeof(struct in_addr));
1427 		addr.s_addr = inet->mc_addr;
1428 		release_sock(sk);
1429 
1430 		if (put_user(len, optlen))
1431 			return -EFAULT;
1432 		if (copy_to_user(optval, &addr, len))
1433 			return -EFAULT;
1434 		return 0;
1435 	}
1436 	case IP_MSFILTER:
1437 	{
1438 		struct ip_msfilter msf;
1439 
1440 		if (len < IP_MSFILTER_SIZE(0)) {
1441 			err = -EINVAL;
1442 			goto out;
1443 		}
1444 		if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1445 			err = -EFAULT;
1446 			goto out;
1447 		}
1448 		err = ip_mc_msfget(sk, &msf,
1449 				   (struct ip_msfilter __user *)optval, optlen);
1450 		goto out;
1451 	}
1452 	case MCAST_MSFILTER:
1453 	{
1454 		struct group_filter gsf;
1455 
1456 		if (len < GROUP_FILTER_SIZE(0)) {
1457 			err = -EINVAL;
1458 			goto out;
1459 		}
1460 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1461 			err = -EFAULT;
1462 			goto out;
1463 		}
1464 		err = ip_mc_gsfget(sk, &gsf,
1465 				   (struct group_filter __user *)optval,
1466 				   optlen);
1467 		goto out;
1468 	}
1469 	case IP_MULTICAST_ALL:
1470 		val = inet->mc_all;
1471 		break;
1472 	case IP_PKTOPTIONS:
1473 	{
1474 		struct msghdr msg;
1475 
1476 		release_sock(sk);
1477 
1478 		if (sk->sk_type != SOCK_STREAM)
1479 			return -ENOPROTOOPT;
1480 
1481 		msg.msg_control = (__force void *) optval;
1482 		msg.msg_controllen = len;
1483 		msg.msg_flags = flags;
1484 
1485 		if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1486 			struct in_pktinfo info;
1487 
1488 			info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1489 			info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1490 			info.ipi_ifindex = inet->mc_index;
1491 			put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1492 		}
1493 		if (inet->cmsg_flags & IP_CMSG_TTL) {
1494 			int hlim = inet->mc_ttl;
1495 			put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1496 		}
1497 		if (inet->cmsg_flags & IP_CMSG_TOS) {
1498 			int tos = inet->rcv_tos;
1499 			put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1500 		}
1501 		len -= msg.msg_controllen;
1502 		return put_user(len, optlen);
1503 	}
1504 	case IP_FREEBIND:
1505 		val = inet->freebind;
1506 		break;
1507 	case IP_TRANSPARENT:
1508 		val = inet->transparent;
1509 		break;
1510 	case IP_MINTTL:
1511 		val = inet->min_ttl;
1512 		break;
1513 	default:
1514 		release_sock(sk);
1515 		return -ENOPROTOOPT;
1516 	}
1517 	release_sock(sk);
1518 
1519 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1520 		unsigned char ucval = (unsigned char)val;
1521 		len = 1;
1522 		if (put_user(len, optlen))
1523 			return -EFAULT;
1524 		if (copy_to_user(optval, &ucval, 1))
1525 			return -EFAULT;
1526 	} else {
1527 		len = min_t(unsigned int, sizeof(int), len);
1528 		if (put_user(len, optlen))
1529 			return -EFAULT;
1530 		if (copy_to_user(optval, &val, len))
1531 			return -EFAULT;
1532 	}
1533 	return 0;
1534 
1535 out:
1536 	release_sock(sk);
1537 	if (needs_rtnl)
1538 		rtnl_unlock();
1539 	return err;
1540 }
1541 
ip_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1542 int ip_getsockopt(struct sock *sk, int level,
1543 		  int optname, char __user *optval, int __user *optlen)
1544 {
1545 	int err;
1546 
1547 	err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1548 #ifdef CONFIG_NETFILTER
1549 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1550 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1551 			!ip_mroute_opt(optname)) {
1552 		int len;
1553 
1554 		if (get_user(len, optlen))
1555 			return -EFAULT;
1556 
1557 		err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1558 		if (err >= 0)
1559 			err = put_user(len, optlen);
1560 		return err;
1561 	}
1562 #endif
1563 	return err;
1564 }
1565 EXPORT_SYMBOL(ip_getsockopt);
1566 
1567 #ifdef CONFIG_COMPAT
compat_ip_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1568 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1569 			 char __user *optval, int __user *optlen)
1570 {
1571 	int err;
1572 
1573 	if (optname == MCAST_MSFILTER)
1574 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1575 			ip_getsockopt);
1576 
1577 	err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1578 		MSG_CMSG_COMPAT);
1579 
1580 #ifdef CONFIG_NETFILTER
1581 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1582 	if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1583 			!ip_mroute_opt(optname)) {
1584 		int len;
1585 
1586 		if (get_user(len, optlen))
1587 			return -EFAULT;
1588 
1589 		err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1590 		if (err >= 0)
1591 			err = put_user(len, optlen);
1592 		return err;
1593 	}
1594 #endif
1595 	return err;
1596 }
1597 EXPORT_SYMBOL(compat_ip_getsockopt);
1598 #endif
1599