]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/ipv4/ip_sockglue.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'for-4.15/alps' into for-linus
[mirror_ubuntu-bionic-kernel.git] / net / ipv4 / ip_sockglue.c
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 <linux/uaccess.h>
48
49 /*
50 * SOL_IP control messages.
51 */
52
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
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
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
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
83 static void ip_cmsg_recv_retopts(struct net *net, struct msghdr *msg,
84 struct sk_buff *skb)
85 {
86 unsigned char optbuf[sizeof(struct ip_options) + 40];
87 struct ip_options *opt = (struct ip_options *)optbuf;
88
89 if (IPCB(skb)->opt.optlen == 0)
90 return;
91
92 if (ip_options_echo(net, opt, skb)) {
93 msg->msg_flags |= MSG_CTRUNC;
94 return;
95 }
96 ip_options_undo(opt);
97
98 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
99 }
100
101 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb)
102 {
103 int val;
104
105 if (IPCB(skb)->frag_max_size == 0)
106 return;
107
108 val = IPCB(skb)->frag_max_size;
109 put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
110 }
111
112 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
113 int tlen, int offset)
114 {
115 __wsum csum = skb->csum;
116
117 if (skb->ip_summed != CHECKSUM_COMPLETE)
118 return;
119
120 if (offset != 0) {
121 int tend_off = skb_transport_offset(skb) + tlen;
122 csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
123 }
124
125 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
126 }
127
128 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
129 {
130 char *secdata;
131 u32 seclen, secid;
132 int err;
133
134 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
135 if (err)
136 return;
137
138 err = security_secid_to_secctx(secid, &secdata, &seclen);
139 if (err)
140 return;
141
142 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
143 security_release_secctx(secdata, seclen);
144 }
145
146 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
147 {
148 struct sockaddr_in sin;
149 const struct iphdr *iph = ip_hdr(skb);
150 __be16 *ports = (__be16 *)skb_transport_header(skb);
151
152 if (skb_transport_offset(skb) + 4 > (int)skb->len)
153 return;
154
155 /* All current transport protocols have the port numbers in the
156 * first four bytes of the transport header and this function is
157 * written with this assumption in mind.
158 */
159
160 sin.sin_family = AF_INET;
161 sin.sin_addr.s_addr = iph->daddr;
162 sin.sin_port = ports[1];
163 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
164
165 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
166 }
167
168 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
169 struct sk_buff *skb, int tlen, int offset)
170 {
171 struct inet_sock *inet = inet_sk(sk);
172 unsigned int flags = inet->cmsg_flags;
173
174 /* Ordered by supposed usage frequency */
175 if (flags & IP_CMSG_PKTINFO) {
176 ip_cmsg_recv_pktinfo(msg, skb);
177
178 flags &= ~IP_CMSG_PKTINFO;
179 if (!flags)
180 return;
181 }
182
183 if (flags & IP_CMSG_TTL) {
184 ip_cmsg_recv_ttl(msg, skb);
185
186 flags &= ~IP_CMSG_TTL;
187 if (!flags)
188 return;
189 }
190
191 if (flags & IP_CMSG_TOS) {
192 ip_cmsg_recv_tos(msg, skb);
193
194 flags &= ~IP_CMSG_TOS;
195 if (!flags)
196 return;
197 }
198
199 if (flags & IP_CMSG_RECVOPTS) {
200 ip_cmsg_recv_opts(msg, skb);
201
202 flags &= ~IP_CMSG_RECVOPTS;
203 if (!flags)
204 return;
205 }
206
207 if (flags & IP_CMSG_RETOPTS) {
208 ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
209
210 flags &= ~IP_CMSG_RETOPTS;
211 if (!flags)
212 return;
213 }
214
215 if (flags & IP_CMSG_PASSSEC) {
216 ip_cmsg_recv_security(msg, skb);
217
218 flags &= ~IP_CMSG_PASSSEC;
219 if (!flags)
220 return;
221 }
222
223 if (flags & IP_CMSG_ORIGDSTADDR) {
224 ip_cmsg_recv_dstaddr(msg, skb);
225
226 flags &= ~IP_CMSG_ORIGDSTADDR;
227 if (!flags)
228 return;
229 }
230
231 if (flags & IP_CMSG_CHECKSUM)
232 ip_cmsg_recv_checksum(msg, skb, tlen, offset);
233
234 if (flags & IP_CMSG_RECVFRAGSIZE)
235 ip_cmsg_recv_fragsize(msg, skb);
236 }
237 EXPORT_SYMBOL(ip_cmsg_recv_offset);
238
239 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
240 bool allow_ipv6)
241 {
242 int err, val;
243 struct cmsghdr *cmsg;
244 struct net *net = sock_net(sk);
245
246 for_each_cmsghdr(cmsg, msg) {
247 if (!CMSG_OK(msg, cmsg))
248 return -EINVAL;
249 #if IS_ENABLED(CONFIG_IPV6)
250 if (allow_ipv6 &&
251 cmsg->cmsg_level == SOL_IPV6 &&
252 cmsg->cmsg_type == IPV6_PKTINFO) {
253 struct in6_pktinfo *src_info;
254
255 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
256 return -EINVAL;
257 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
258 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
259 return -EINVAL;
260 ipc->oif = src_info->ipi6_ifindex;
261 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
262 continue;
263 }
264 #endif
265 if (cmsg->cmsg_level == SOL_SOCKET) {
266 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
267 if (err)
268 return err;
269 continue;
270 }
271
272 if (cmsg->cmsg_level != SOL_IP)
273 continue;
274 switch (cmsg->cmsg_type) {
275 case IP_RETOPTS:
276 err = cmsg->cmsg_len - sizeof(struct cmsghdr);
277
278 /* Our caller is responsible for freeing ipc->opt */
279 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
280 err < 40 ? err : 40);
281 if (err)
282 return err;
283 break;
284 case IP_PKTINFO:
285 {
286 struct in_pktinfo *info;
287 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
288 return -EINVAL;
289 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
290 ipc->oif = info->ipi_ifindex;
291 ipc->addr = info->ipi_spec_dst.s_addr;
292 break;
293 }
294 case IP_TTL:
295 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
296 return -EINVAL;
297 val = *(int *)CMSG_DATA(cmsg);
298 if (val < 1 || val > 255)
299 return -EINVAL;
300 ipc->ttl = val;
301 break;
302 case IP_TOS:
303 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
304 val = *(int *)CMSG_DATA(cmsg);
305 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
306 val = *(u8 *)CMSG_DATA(cmsg);
307 else
308 return -EINVAL;
309 if (val < 0 || val > 255)
310 return -EINVAL;
311 ipc->tos = val;
312 ipc->priority = rt_tos2priority(ipc->tos);
313 break;
314
315 default:
316 return -EINVAL;
317 }
318 }
319 return 0;
320 }
321
322
323 /* Special input handler for packets caught by router alert option.
324 They are selected only by protocol field, and then processed likely
325 local ones; but only if someone wants them! Otherwise, router
326 not running rsvpd will kill RSVP.
327
328 It is user level problem, what it will make with them.
329 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
330 but receiver should be enough clever f.e. to forward mtrace requests,
331 sent to multicast group to reach destination designated router.
332 */
333 struct ip_ra_chain __rcu *ip_ra_chain;
334
335
336 static void ip_ra_destroy_rcu(struct rcu_head *head)
337 {
338 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
339
340 sock_put(ra->saved_sk);
341 kfree(ra);
342 }
343
344 int ip_ra_control(struct sock *sk, unsigned char on,
345 void (*destructor)(struct sock *))
346 {
347 struct ip_ra_chain *ra, *new_ra;
348 struct ip_ra_chain __rcu **rap;
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
355 for (rap = &ip_ra_chain;
356 (ra = rtnl_dereference(*rap)) != NULL;
357 rap = &ra->next) {
358 if (ra->sk == sk) {
359 if (on) {
360 kfree(new_ra);
361 return -EADDRINUSE;
362 }
363 /* dont let ip_call_ra_chain() use sk again */
364 ra->sk = NULL;
365 RCU_INIT_POINTER(*rap, ra->next);
366
367 if (ra->destructor)
368 ra->destructor(sk);
369 /*
370 * Delay sock_put(sk) and kfree(ra) after one rcu grace
371 * period. This guarantee ip_call_ra_chain() dont need
372 * to mess with socket refcounts.
373 */
374 ra->saved_sk = sk;
375 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
376 return 0;
377 }
378 }
379 if (!new_ra)
380 return -ENOBUFS;
381 new_ra->sk = sk;
382 new_ra->destructor = destructor;
383
384 RCU_INIT_POINTER(new_ra->next, ra);
385 rcu_assign_pointer(*rap, new_ra);
386 sock_hold(sk);
387
388 return 0;
389 }
390
391 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
392 __be16 port, u32 info, u8 *payload)
393 {
394 struct sock_exterr_skb *serr;
395
396 skb = skb_clone(skb, GFP_ATOMIC);
397 if (!skb)
398 return;
399
400 serr = SKB_EXT_ERR(skb);
401 serr->ee.ee_errno = err;
402 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
403 serr->ee.ee_type = icmp_hdr(skb)->type;
404 serr->ee.ee_code = icmp_hdr(skb)->code;
405 serr->ee.ee_pad = 0;
406 serr->ee.ee_info = info;
407 serr->ee.ee_data = 0;
408 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
409 skb_network_header(skb);
410 serr->port = port;
411
412 if (skb_pull(skb, payload - skb->data)) {
413 skb_reset_transport_header(skb);
414 if (sock_queue_err_skb(sk, skb) == 0)
415 return;
416 }
417 kfree_skb(skb);
418 }
419
420 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
421 {
422 struct inet_sock *inet = inet_sk(sk);
423 struct sock_exterr_skb *serr;
424 struct iphdr *iph;
425 struct sk_buff *skb;
426
427 if (!inet->recverr)
428 return;
429
430 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
431 if (!skb)
432 return;
433
434 skb_put(skb, sizeof(struct iphdr));
435 skb_reset_network_header(skb);
436 iph = ip_hdr(skb);
437 iph->daddr = daddr;
438
439 serr = SKB_EXT_ERR(skb);
440 serr->ee.ee_errno = err;
441 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
442 serr->ee.ee_type = 0;
443 serr->ee.ee_code = 0;
444 serr->ee.ee_pad = 0;
445 serr->ee.ee_info = info;
446 serr->ee.ee_data = 0;
447 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
448 serr->port = port;
449
450 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
451 skb_reset_transport_header(skb);
452
453 if (sock_queue_err_skb(sk, skb))
454 kfree_skb(skb);
455 }
456
457 /* For some errors we have valid addr_offset even with zero payload and
458 * zero port. Also, addr_offset should be supported if port is set.
459 */
460 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
461 {
462 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
463 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
464 }
465
466 /* IPv4 supports cmsg on all imcp errors and some timestamps
467 *
468 * Timestamp code paths do not initialize the fields expected by cmsg:
469 * the PKTINFO fields in skb->cb[]. Fill those in here.
470 */
471 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
472 struct sk_buff *skb,
473 int ee_origin)
474 {
475 struct in_pktinfo *info;
476
477 if (ee_origin == SO_EE_ORIGIN_ICMP)
478 return true;
479
480 if (ee_origin == SO_EE_ORIGIN_LOCAL)
481 return false;
482
483 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
484 * timestamp with egress dev. Not possible for packets without iif
485 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
486 */
487 info = PKTINFO_SKB_CB(skb);
488 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
489 !info->ipi_ifindex)
490 return false;
491
492 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
493 return true;
494 }
495
496 /*
497 * Handle MSG_ERRQUEUE
498 */
499 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
500 {
501 struct sock_exterr_skb *serr;
502 struct sk_buff *skb;
503 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
504 struct {
505 struct sock_extended_err ee;
506 struct sockaddr_in offender;
507 } errhdr;
508 int err;
509 int copied;
510
511 WARN_ON_ONCE(sk->sk_family == AF_INET6);
512
513 err = -EAGAIN;
514 skb = sock_dequeue_err_skb(sk);
515 if (!skb)
516 goto out;
517
518 copied = skb->len;
519 if (copied > len) {
520 msg->msg_flags |= MSG_TRUNC;
521 copied = len;
522 }
523 err = skb_copy_datagram_msg(skb, 0, msg, copied);
524 if (unlikely(err)) {
525 kfree_skb(skb);
526 return err;
527 }
528 sock_recv_timestamp(msg, sk, skb);
529
530 serr = SKB_EXT_ERR(skb);
531
532 if (sin && ipv4_datagram_support_addr(serr)) {
533 sin->sin_family = AF_INET;
534 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
535 serr->addr_offset);
536 sin->sin_port = serr->port;
537 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
538 *addr_len = sizeof(*sin);
539 }
540
541 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
542 sin = &errhdr.offender;
543 memset(sin, 0, sizeof(*sin));
544
545 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
546 sin->sin_family = AF_INET;
547 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
548 if (inet_sk(sk)->cmsg_flags)
549 ip_cmsg_recv(msg, skb);
550 }
551
552 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
553
554 /* Now we could try to dump offended packet options */
555
556 msg->msg_flags |= MSG_ERRQUEUE;
557 err = copied;
558
559 consume_skb(skb);
560 out:
561 return err;
562 }
563
564
565 /*
566 * Socket option code for IP. This is the end of the line after any
567 * TCP,UDP etc options on an IP socket.
568 */
569 static bool setsockopt_needs_rtnl(int optname)
570 {
571 switch (optname) {
572 case IP_ADD_MEMBERSHIP:
573 case IP_ADD_SOURCE_MEMBERSHIP:
574 case IP_BLOCK_SOURCE:
575 case IP_DROP_MEMBERSHIP:
576 case IP_DROP_SOURCE_MEMBERSHIP:
577 case IP_MSFILTER:
578 case IP_UNBLOCK_SOURCE:
579 case MCAST_BLOCK_SOURCE:
580 case MCAST_MSFILTER:
581 case MCAST_JOIN_GROUP:
582 case MCAST_JOIN_SOURCE_GROUP:
583 case MCAST_LEAVE_GROUP:
584 case MCAST_LEAVE_SOURCE_GROUP:
585 case MCAST_UNBLOCK_SOURCE:
586 case IP_ROUTER_ALERT:
587 return true;
588 }
589 return false;
590 }
591
592 static int do_ip_setsockopt(struct sock *sk, int level,
593 int optname, char __user *optval, unsigned int optlen)
594 {
595 struct inet_sock *inet = inet_sk(sk);
596 struct net *net = sock_net(sk);
597 int val = 0, err;
598 bool needs_rtnl = setsockopt_needs_rtnl(optname);
599
600 switch (optname) {
601 case IP_PKTINFO:
602 case IP_RECVTTL:
603 case IP_RECVOPTS:
604 case IP_RECVTOS:
605 case IP_RETOPTS:
606 case IP_TOS:
607 case IP_TTL:
608 case IP_HDRINCL:
609 case IP_MTU_DISCOVER:
610 case IP_RECVERR:
611 case IP_ROUTER_ALERT:
612 case IP_FREEBIND:
613 case IP_PASSSEC:
614 case IP_TRANSPARENT:
615 case IP_MINTTL:
616 case IP_NODEFRAG:
617 case IP_BIND_ADDRESS_NO_PORT:
618 case IP_UNICAST_IF:
619 case IP_MULTICAST_TTL:
620 case IP_MULTICAST_ALL:
621 case IP_MULTICAST_LOOP:
622 case IP_RECVORIGDSTADDR:
623 case IP_CHECKSUM:
624 case IP_RECVFRAGSIZE:
625 if (optlen >= sizeof(int)) {
626 if (get_user(val, (int __user *) optval))
627 return -EFAULT;
628 } else if (optlen >= sizeof(char)) {
629 unsigned char ucval;
630
631 if (get_user(ucval, (unsigned char __user *) optval))
632 return -EFAULT;
633 val = (int) ucval;
634 }
635 }
636
637 /* If optlen==0, it is equivalent to val == 0 */
638
639 if (ip_mroute_opt(optname))
640 return ip_mroute_setsockopt(sk, optname, optval, optlen);
641
642 err = 0;
643 if (needs_rtnl)
644 rtnl_lock();
645 lock_sock(sk);
646
647 switch (optname) {
648 case IP_OPTIONS:
649 {
650 struct ip_options_rcu *old, *opt = NULL;
651
652 if (optlen > 40)
653 goto e_inval;
654 err = ip_options_get_from_user(sock_net(sk), &opt,
655 optval, optlen);
656 if (err)
657 break;
658 old = rcu_dereference_protected(inet->inet_opt,
659 lockdep_sock_is_held(sk));
660 if (inet->is_icsk) {
661 struct inet_connection_sock *icsk = inet_csk(sk);
662 #if IS_ENABLED(CONFIG_IPV6)
663 if (sk->sk_family == PF_INET ||
664 (!((1 << sk->sk_state) &
665 (TCPF_LISTEN | TCPF_CLOSE)) &&
666 inet->inet_daddr != LOOPBACK4_IPV6)) {
667 #endif
668 if (old)
669 icsk->icsk_ext_hdr_len -= old->opt.optlen;
670 if (opt)
671 icsk->icsk_ext_hdr_len += opt->opt.optlen;
672 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
673 #if IS_ENABLED(CONFIG_IPV6)
674 }
675 #endif
676 }
677 rcu_assign_pointer(inet->inet_opt, opt);
678 if (old)
679 kfree_rcu(old, rcu);
680 break;
681 }
682 case IP_PKTINFO:
683 if (val)
684 inet->cmsg_flags |= IP_CMSG_PKTINFO;
685 else
686 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
687 break;
688 case IP_RECVTTL:
689 if (val)
690 inet->cmsg_flags |= IP_CMSG_TTL;
691 else
692 inet->cmsg_flags &= ~IP_CMSG_TTL;
693 break;
694 case IP_RECVTOS:
695 if (val)
696 inet->cmsg_flags |= IP_CMSG_TOS;
697 else
698 inet->cmsg_flags &= ~IP_CMSG_TOS;
699 break;
700 case IP_RECVOPTS:
701 if (val)
702 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
703 else
704 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
705 break;
706 case IP_RETOPTS:
707 if (val)
708 inet->cmsg_flags |= IP_CMSG_RETOPTS;
709 else
710 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
711 break;
712 case IP_PASSSEC:
713 if (val)
714 inet->cmsg_flags |= IP_CMSG_PASSSEC;
715 else
716 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
717 break;
718 case IP_RECVORIGDSTADDR:
719 if (val)
720 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
721 else
722 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
723 break;
724 case IP_CHECKSUM:
725 if (val) {
726 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
727 inet_inc_convert_csum(sk);
728 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
729 }
730 } else {
731 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
732 inet_dec_convert_csum(sk);
733 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
734 }
735 }
736 break;
737 case IP_RECVFRAGSIZE:
738 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
739 goto e_inval;
740 if (val)
741 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
742 else
743 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
744 break;
745 case IP_TOS: /* This sets both TOS and Precedence */
746 if (sk->sk_type == SOCK_STREAM) {
747 val &= ~INET_ECN_MASK;
748 val |= inet->tos & INET_ECN_MASK;
749 }
750 if (inet->tos != val) {
751 inet->tos = val;
752 sk->sk_priority = rt_tos2priority(val);
753 sk_dst_reset(sk);
754 }
755 break;
756 case IP_TTL:
757 if (optlen < 1)
758 goto e_inval;
759 if (val != -1 && (val < 1 || val > 255))
760 goto e_inval;
761 inet->uc_ttl = val;
762 break;
763 case IP_HDRINCL:
764 if (sk->sk_type != SOCK_RAW) {
765 err = -ENOPROTOOPT;
766 break;
767 }
768 inet->hdrincl = val ? 1 : 0;
769 break;
770 case IP_NODEFRAG:
771 if (sk->sk_type != SOCK_RAW) {
772 err = -ENOPROTOOPT;
773 break;
774 }
775 inet->nodefrag = val ? 1 : 0;
776 break;
777 case IP_BIND_ADDRESS_NO_PORT:
778 inet->bind_address_no_port = val ? 1 : 0;
779 break;
780 case IP_MTU_DISCOVER:
781 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
782 goto e_inval;
783 inet->pmtudisc = val;
784 break;
785 case IP_RECVERR:
786 inet->recverr = !!val;
787 if (!val)
788 skb_queue_purge(&sk->sk_error_queue);
789 break;
790 case IP_MULTICAST_TTL:
791 if (sk->sk_type == SOCK_STREAM)
792 goto e_inval;
793 if (optlen < 1)
794 goto e_inval;
795 if (val == -1)
796 val = 1;
797 if (val < 0 || val > 255)
798 goto e_inval;
799 inet->mc_ttl = val;
800 break;
801 case IP_MULTICAST_LOOP:
802 if (optlen < 1)
803 goto e_inval;
804 inet->mc_loop = !!val;
805 break;
806 case IP_UNICAST_IF:
807 {
808 struct net_device *dev = NULL;
809 int ifindex;
810
811 if (optlen != sizeof(int))
812 goto e_inval;
813
814 ifindex = (__force int)ntohl((__force __be32)val);
815 if (ifindex == 0) {
816 inet->uc_index = 0;
817 err = 0;
818 break;
819 }
820
821 dev = dev_get_by_index(sock_net(sk), ifindex);
822 err = -EADDRNOTAVAIL;
823 if (!dev)
824 break;
825 dev_put(dev);
826
827 err = -EINVAL;
828 if (sk->sk_bound_dev_if)
829 break;
830
831 inet->uc_index = ifindex;
832 err = 0;
833 break;
834 }
835 case IP_MULTICAST_IF:
836 {
837 struct ip_mreqn mreq;
838 struct net_device *dev = NULL;
839 int midx;
840
841 if (sk->sk_type == SOCK_STREAM)
842 goto e_inval;
843 /*
844 * Check the arguments are allowable
845 */
846
847 if (optlen < sizeof(struct in_addr))
848 goto e_inval;
849
850 err = -EFAULT;
851 if (optlen >= sizeof(struct ip_mreqn)) {
852 if (copy_from_user(&mreq, optval, sizeof(mreq)))
853 break;
854 } else {
855 memset(&mreq, 0, sizeof(mreq));
856 if (optlen >= sizeof(struct ip_mreq)) {
857 if (copy_from_user(&mreq, optval,
858 sizeof(struct ip_mreq)))
859 break;
860 } else if (optlen >= sizeof(struct in_addr)) {
861 if (copy_from_user(&mreq.imr_address, optval,
862 sizeof(struct in_addr)))
863 break;
864 }
865 }
866
867 if (!mreq.imr_ifindex) {
868 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
869 inet->mc_index = 0;
870 inet->mc_addr = 0;
871 err = 0;
872 break;
873 }
874 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
875 if (dev)
876 mreq.imr_ifindex = dev->ifindex;
877 } else
878 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
879
880
881 err = -EADDRNOTAVAIL;
882 if (!dev)
883 break;
884
885 midx = l3mdev_master_ifindex(dev);
886
887 dev_put(dev);
888
889 err = -EINVAL;
890 if (sk->sk_bound_dev_if &&
891 mreq.imr_ifindex != sk->sk_bound_dev_if &&
892 (!midx || midx != sk->sk_bound_dev_if))
893 break;
894
895 inet->mc_index = mreq.imr_ifindex;
896 inet->mc_addr = mreq.imr_address.s_addr;
897 err = 0;
898 break;
899 }
900
901 case IP_ADD_MEMBERSHIP:
902 case IP_DROP_MEMBERSHIP:
903 {
904 struct ip_mreqn mreq;
905
906 err = -EPROTO;
907 if (inet_sk(sk)->is_icsk)
908 break;
909
910 if (optlen < sizeof(struct ip_mreq))
911 goto e_inval;
912 err = -EFAULT;
913 if (optlen >= sizeof(struct ip_mreqn)) {
914 if (copy_from_user(&mreq, optval, sizeof(mreq)))
915 break;
916 } else {
917 memset(&mreq, 0, sizeof(mreq));
918 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
919 break;
920 }
921
922 if (optname == IP_ADD_MEMBERSHIP)
923 err = ip_mc_join_group(sk, &mreq);
924 else
925 err = ip_mc_leave_group(sk, &mreq);
926 break;
927 }
928 case IP_MSFILTER:
929 {
930 struct ip_msfilter *msf;
931
932 if (optlen < IP_MSFILTER_SIZE(0))
933 goto e_inval;
934 if (optlen > sysctl_optmem_max) {
935 err = -ENOBUFS;
936 break;
937 }
938 msf = memdup_user(optval, optlen);
939 if (IS_ERR(msf)) {
940 err = PTR_ERR(msf);
941 break;
942 }
943 /* numsrc >= (1G-4) overflow in 32 bits */
944 if (msf->imsf_numsrc >= 0x3ffffffcU ||
945 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
946 kfree(msf);
947 err = -ENOBUFS;
948 break;
949 }
950 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
951 kfree(msf);
952 err = -EINVAL;
953 break;
954 }
955 err = ip_mc_msfilter(sk, msf, 0);
956 kfree(msf);
957 break;
958 }
959 case IP_BLOCK_SOURCE:
960 case IP_UNBLOCK_SOURCE:
961 case IP_ADD_SOURCE_MEMBERSHIP:
962 case IP_DROP_SOURCE_MEMBERSHIP:
963 {
964 struct ip_mreq_source mreqs;
965 int omode, add;
966
967 if (optlen != sizeof(struct ip_mreq_source))
968 goto e_inval;
969 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
970 err = -EFAULT;
971 break;
972 }
973 if (optname == IP_BLOCK_SOURCE) {
974 omode = MCAST_EXCLUDE;
975 add = 1;
976 } else if (optname == IP_UNBLOCK_SOURCE) {
977 omode = MCAST_EXCLUDE;
978 add = 0;
979 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
980 struct ip_mreqn mreq;
981
982 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
983 mreq.imr_address.s_addr = mreqs.imr_interface;
984 mreq.imr_ifindex = 0;
985 err = ip_mc_join_group(sk, &mreq);
986 if (err && err != -EADDRINUSE)
987 break;
988 omode = MCAST_INCLUDE;
989 add = 1;
990 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
991 omode = MCAST_INCLUDE;
992 add = 0;
993 }
994 err = ip_mc_source(add, omode, sk, &mreqs, 0);
995 break;
996 }
997 case MCAST_JOIN_GROUP:
998 case MCAST_LEAVE_GROUP:
999 {
1000 struct group_req greq;
1001 struct sockaddr_in *psin;
1002 struct ip_mreqn mreq;
1003
1004 if (optlen < sizeof(struct group_req))
1005 goto e_inval;
1006 err = -EFAULT;
1007 if (copy_from_user(&greq, optval, sizeof(greq)))
1008 break;
1009 psin = (struct sockaddr_in *)&greq.gr_group;
1010 if (psin->sin_family != AF_INET)
1011 goto e_inval;
1012 memset(&mreq, 0, sizeof(mreq));
1013 mreq.imr_multiaddr = psin->sin_addr;
1014 mreq.imr_ifindex = greq.gr_interface;
1015
1016 if (optname == MCAST_JOIN_GROUP)
1017 err = ip_mc_join_group(sk, &mreq);
1018 else
1019 err = ip_mc_leave_group(sk, &mreq);
1020 break;
1021 }
1022 case MCAST_JOIN_SOURCE_GROUP:
1023 case MCAST_LEAVE_SOURCE_GROUP:
1024 case MCAST_BLOCK_SOURCE:
1025 case MCAST_UNBLOCK_SOURCE:
1026 {
1027 struct group_source_req greqs;
1028 struct ip_mreq_source mreqs;
1029 struct sockaddr_in *psin;
1030 int omode, add;
1031
1032 if (optlen != sizeof(struct group_source_req))
1033 goto e_inval;
1034 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1035 err = -EFAULT;
1036 break;
1037 }
1038 if (greqs.gsr_group.ss_family != AF_INET ||
1039 greqs.gsr_source.ss_family != AF_INET) {
1040 err = -EADDRNOTAVAIL;
1041 break;
1042 }
1043 psin = (struct sockaddr_in *)&greqs.gsr_group;
1044 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1045 psin = (struct sockaddr_in *)&greqs.gsr_source;
1046 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1047 mreqs.imr_interface = 0; /* use index for mc_source */
1048
1049 if (optname == MCAST_BLOCK_SOURCE) {
1050 omode = MCAST_EXCLUDE;
1051 add = 1;
1052 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1053 omode = MCAST_EXCLUDE;
1054 add = 0;
1055 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1056 struct ip_mreqn mreq;
1057
1058 psin = (struct sockaddr_in *)&greqs.gsr_group;
1059 mreq.imr_multiaddr = psin->sin_addr;
1060 mreq.imr_address.s_addr = 0;
1061 mreq.imr_ifindex = greqs.gsr_interface;
1062 err = ip_mc_join_group(sk, &mreq);
1063 if (err && err != -EADDRINUSE)
1064 break;
1065 greqs.gsr_interface = mreq.imr_ifindex;
1066 omode = MCAST_INCLUDE;
1067 add = 1;
1068 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1069 omode = MCAST_INCLUDE;
1070 add = 0;
1071 }
1072 err = ip_mc_source(add, omode, sk, &mreqs,
1073 greqs.gsr_interface);
1074 break;
1075 }
1076 case MCAST_MSFILTER:
1077 {
1078 struct sockaddr_in *psin;
1079 struct ip_msfilter *msf = NULL;
1080 struct group_filter *gsf = NULL;
1081 int msize, i, ifindex;
1082
1083 if (optlen < GROUP_FILTER_SIZE(0))
1084 goto e_inval;
1085 if (optlen > sysctl_optmem_max) {
1086 err = -ENOBUFS;
1087 break;
1088 }
1089 gsf = memdup_user(optval, optlen);
1090 if (IS_ERR(gsf)) {
1091 err = PTR_ERR(gsf);
1092 break;
1093 }
1094
1095 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1096 if (gsf->gf_numsrc >= 0x1ffffff ||
1097 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1098 err = -ENOBUFS;
1099 goto mc_msf_out;
1100 }
1101 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1102 err = -EINVAL;
1103 goto mc_msf_out;
1104 }
1105 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1106 msf = kmalloc(msize, GFP_KERNEL);
1107 if (!msf) {
1108 err = -ENOBUFS;
1109 goto mc_msf_out;
1110 }
1111 ifindex = gsf->gf_interface;
1112 psin = (struct sockaddr_in *)&gsf->gf_group;
1113 if (psin->sin_family != AF_INET) {
1114 err = -EADDRNOTAVAIL;
1115 goto mc_msf_out;
1116 }
1117 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1118 msf->imsf_interface = 0;
1119 msf->imsf_fmode = gsf->gf_fmode;
1120 msf->imsf_numsrc = gsf->gf_numsrc;
1121 err = -EADDRNOTAVAIL;
1122 for (i = 0; i < gsf->gf_numsrc; ++i) {
1123 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1124
1125 if (psin->sin_family != AF_INET)
1126 goto mc_msf_out;
1127 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1128 }
1129 kfree(gsf);
1130 gsf = NULL;
1131
1132 err = ip_mc_msfilter(sk, msf, ifindex);
1133 mc_msf_out:
1134 kfree(msf);
1135 kfree(gsf);
1136 break;
1137 }
1138 case IP_MULTICAST_ALL:
1139 if (optlen < 1)
1140 goto e_inval;
1141 if (val != 0 && val != 1)
1142 goto e_inval;
1143 inet->mc_all = val;
1144 break;
1145 case IP_ROUTER_ALERT:
1146 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1147 break;
1148
1149 case IP_FREEBIND:
1150 if (optlen < 1)
1151 goto e_inval;
1152 inet->freebind = !!val;
1153 break;
1154
1155 case IP_IPSEC_POLICY:
1156 case IP_XFRM_POLICY:
1157 err = -EPERM;
1158 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1159 break;
1160 err = xfrm_user_policy(sk, optname, optval, optlen);
1161 break;
1162
1163 case IP_TRANSPARENT:
1164 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1165 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1166 err = -EPERM;
1167 break;
1168 }
1169 if (optlen < 1)
1170 goto e_inval;
1171 inet->transparent = !!val;
1172 break;
1173
1174 case IP_MINTTL:
1175 if (optlen < 1)
1176 goto e_inval;
1177 if (val < 0 || val > 255)
1178 goto e_inval;
1179 inet->min_ttl = val;
1180 break;
1181
1182 default:
1183 err = -ENOPROTOOPT;
1184 break;
1185 }
1186 release_sock(sk);
1187 if (needs_rtnl)
1188 rtnl_unlock();
1189 return err;
1190
1191 e_inval:
1192 release_sock(sk);
1193 if (needs_rtnl)
1194 rtnl_unlock();
1195 return -EINVAL;
1196 }
1197
1198 /**
1199 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1200 * @sk: socket
1201 * @skb: buffer
1202 *
1203 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1204 * destination in skb->cb[] before dst drop.
1205 * This way, receiver doesn't make cache line misses to read rtable.
1206 */
1207 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1208 {
1209 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1210 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1211 ipv6_sk_rxinfo(sk);
1212
1213 if (prepare && skb_rtable(skb)) {
1214 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1215 * which has interface index (iif) as the first member of the
1216 * underlying inet{6}_skb_parm struct. This code then overlays
1217 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1218 * element so the iif is picked up from the prior IPCB. If iif
1219 * is the loopback interface, then return the sending interface
1220 * (e.g., process binds socket to eth0 for Tx which is
1221 * redirected to loopback in the rtable/dst).
1222 */
1223 struct rtable *rt = skb_rtable(skb);
1224 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1225
1226 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1227 pktinfo->ipi_ifindex = inet_iif(skb);
1228 else if (l3slave && rt && rt->rt_iif)
1229 pktinfo->ipi_ifindex = rt->rt_iif;
1230
1231 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1232 } else {
1233 pktinfo->ipi_ifindex = 0;
1234 pktinfo->ipi_spec_dst.s_addr = 0;
1235 }
1236 skb_dst_drop(skb);
1237 }
1238
1239 int ip_setsockopt(struct sock *sk, int level,
1240 int optname, char __user *optval, unsigned int optlen)
1241 {
1242 int err;
1243
1244 if (level != SOL_IP)
1245 return -ENOPROTOOPT;
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 lock_sock(sk);
1255 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1256 release_sock(sk);
1257 }
1258 #endif
1259 return err;
1260 }
1261 EXPORT_SYMBOL(ip_setsockopt);
1262
1263 #ifdef CONFIG_COMPAT
1264 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1265 char __user *optval, unsigned int optlen)
1266 {
1267 int err;
1268
1269 if (level != SOL_IP)
1270 return -ENOPROTOOPT;
1271
1272 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1273 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1274 ip_setsockopt);
1275
1276 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1277 #ifdef CONFIG_NETFILTER
1278 /* we need to exclude all possible ENOPROTOOPTs except default case */
1279 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1280 optname != IP_IPSEC_POLICY &&
1281 optname != IP_XFRM_POLICY &&
1282 !ip_mroute_opt(optname)) {
1283 lock_sock(sk);
1284 err = compat_nf_setsockopt(sk, PF_INET, optname,
1285 optval, optlen);
1286 release_sock(sk);
1287 }
1288 #endif
1289 return err;
1290 }
1291 EXPORT_SYMBOL(compat_ip_setsockopt);
1292 #endif
1293
1294 /*
1295 * Get the options. Note for future reference. The GET of IP options gets
1296 * the _received_ ones. The set sets the _sent_ ones.
1297 */
1298
1299 static bool getsockopt_needs_rtnl(int optname)
1300 {
1301 switch (optname) {
1302 case IP_MSFILTER:
1303 case MCAST_MSFILTER:
1304 return true;
1305 }
1306 return false;
1307 }
1308
1309 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1310 char __user *optval, int __user *optlen, unsigned int flags)
1311 {
1312 struct inet_sock *inet = inet_sk(sk);
1313 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1314 int val, err = 0;
1315 int len;
1316
1317 if (level != SOL_IP)
1318 return -EOPNOTSUPP;
1319
1320 if (ip_mroute_opt(optname))
1321 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1322
1323 if (get_user(len, optlen))
1324 return -EFAULT;
1325 if (len < 0)
1326 return -EINVAL;
1327
1328 if (needs_rtnl)
1329 rtnl_lock();
1330 lock_sock(sk);
1331
1332 switch (optname) {
1333 case IP_OPTIONS:
1334 {
1335 unsigned char optbuf[sizeof(struct ip_options)+40];
1336 struct ip_options *opt = (struct ip_options *)optbuf;
1337 struct ip_options_rcu *inet_opt;
1338
1339 inet_opt = rcu_dereference_protected(inet->inet_opt,
1340 lockdep_sock_is_held(sk));
1341 opt->optlen = 0;
1342 if (inet_opt)
1343 memcpy(optbuf, &inet_opt->opt,
1344 sizeof(struct ip_options) +
1345 inet_opt->opt.optlen);
1346 release_sock(sk);
1347
1348 if (opt->optlen == 0)
1349 return put_user(0, optlen);
1350
1351 ip_options_undo(opt);
1352
1353 len = min_t(unsigned int, len, opt->optlen);
1354 if (put_user(len, optlen))
1355 return -EFAULT;
1356 if (copy_to_user(optval, opt->__data, len))
1357 return -EFAULT;
1358 return 0;
1359 }
1360 case IP_PKTINFO:
1361 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1362 break;
1363 case IP_RECVTTL:
1364 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1365 break;
1366 case IP_RECVTOS:
1367 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1368 break;
1369 case IP_RECVOPTS:
1370 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1371 break;
1372 case IP_RETOPTS:
1373 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1374 break;
1375 case IP_PASSSEC:
1376 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1377 break;
1378 case IP_RECVORIGDSTADDR:
1379 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1380 break;
1381 case IP_CHECKSUM:
1382 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1383 break;
1384 case IP_RECVFRAGSIZE:
1385 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1386 break;
1387 case IP_TOS:
1388 val = inet->tos;
1389 break;
1390 case IP_TTL:
1391 {
1392 struct net *net = sock_net(sk);
1393 val = (inet->uc_ttl == -1 ?
1394 net->ipv4.sysctl_ip_default_ttl :
1395 inet->uc_ttl);
1396 break;
1397 }
1398 case IP_HDRINCL:
1399 val = inet->hdrincl;
1400 break;
1401 case IP_NODEFRAG:
1402 val = inet->nodefrag;
1403 break;
1404 case IP_BIND_ADDRESS_NO_PORT:
1405 val = inet->bind_address_no_port;
1406 break;
1407 case IP_MTU_DISCOVER:
1408 val = inet->pmtudisc;
1409 break;
1410 case IP_MTU:
1411 {
1412 struct dst_entry *dst;
1413 val = 0;
1414 dst = sk_dst_get(sk);
1415 if (dst) {
1416 val = dst_mtu(dst);
1417 dst_release(dst);
1418 }
1419 if (!val) {
1420 release_sock(sk);
1421 return -ENOTCONN;
1422 }
1423 break;
1424 }
1425 case IP_RECVERR:
1426 val = inet->recverr;
1427 break;
1428 case IP_MULTICAST_TTL:
1429 val = inet->mc_ttl;
1430 break;
1431 case IP_MULTICAST_LOOP:
1432 val = inet->mc_loop;
1433 break;
1434 case IP_UNICAST_IF:
1435 val = (__force int)htonl((__u32) inet->uc_index);
1436 break;
1437 case IP_MULTICAST_IF:
1438 {
1439 struct in_addr addr;
1440 len = min_t(unsigned int, len, sizeof(struct in_addr));
1441 addr.s_addr = inet->mc_addr;
1442 release_sock(sk);
1443
1444 if (put_user(len, optlen))
1445 return -EFAULT;
1446 if (copy_to_user(optval, &addr, len))
1447 return -EFAULT;
1448 return 0;
1449 }
1450 case IP_MSFILTER:
1451 {
1452 struct ip_msfilter msf;
1453
1454 if (len < IP_MSFILTER_SIZE(0)) {
1455 err = -EINVAL;
1456 goto out;
1457 }
1458 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1459 err = -EFAULT;
1460 goto out;
1461 }
1462 err = ip_mc_msfget(sk, &msf,
1463 (struct ip_msfilter __user *)optval, optlen);
1464 goto out;
1465 }
1466 case MCAST_MSFILTER:
1467 {
1468 struct group_filter gsf;
1469
1470 if (len < GROUP_FILTER_SIZE(0)) {
1471 err = -EINVAL;
1472 goto out;
1473 }
1474 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1475 err = -EFAULT;
1476 goto out;
1477 }
1478 err = ip_mc_gsfget(sk, &gsf,
1479 (struct group_filter __user *)optval,
1480 optlen);
1481 goto out;
1482 }
1483 case IP_MULTICAST_ALL:
1484 val = inet->mc_all;
1485 break;
1486 case IP_PKTOPTIONS:
1487 {
1488 struct msghdr msg;
1489
1490 release_sock(sk);
1491
1492 if (sk->sk_type != SOCK_STREAM)
1493 return -ENOPROTOOPT;
1494
1495 msg.msg_control = (__force void *) optval;
1496 msg.msg_controllen = len;
1497 msg.msg_flags = flags;
1498
1499 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1500 struct in_pktinfo info;
1501
1502 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1503 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1504 info.ipi_ifindex = inet->mc_index;
1505 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1506 }
1507 if (inet->cmsg_flags & IP_CMSG_TTL) {
1508 int hlim = inet->mc_ttl;
1509 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1510 }
1511 if (inet->cmsg_flags & IP_CMSG_TOS) {
1512 int tos = inet->rcv_tos;
1513 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1514 }
1515 len -= msg.msg_controllen;
1516 return put_user(len, optlen);
1517 }
1518 case IP_FREEBIND:
1519 val = inet->freebind;
1520 break;
1521 case IP_TRANSPARENT:
1522 val = inet->transparent;
1523 break;
1524 case IP_MINTTL:
1525 val = inet->min_ttl;
1526 break;
1527 default:
1528 release_sock(sk);
1529 return -ENOPROTOOPT;
1530 }
1531 release_sock(sk);
1532
1533 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1534 unsigned char ucval = (unsigned char)val;
1535 len = 1;
1536 if (put_user(len, optlen))
1537 return -EFAULT;
1538 if (copy_to_user(optval, &ucval, 1))
1539 return -EFAULT;
1540 } else {
1541 len = min_t(unsigned int, sizeof(int), len);
1542 if (put_user(len, optlen))
1543 return -EFAULT;
1544 if (copy_to_user(optval, &val, len))
1545 return -EFAULT;
1546 }
1547 return 0;
1548
1549 out:
1550 release_sock(sk);
1551 if (needs_rtnl)
1552 rtnl_unlock();
1553 return err;
1554 }
1555
1556 int ip_getsockopt(struct sock *sk, int level,
1557 int optname, char __user *optval, int __user *optlen)
1558 {
1559 int err;
1560
1561 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1562 #ifdef CONFIG_NETFILTER
1563 /* we need to exclude all possible ENOPROTOOPTs except default case */
1564 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1565 !ip_mroute_opt(optname)) {
1566 int len;
1567
1568 if (get_user(len, optlen))
1569 return -EFAULT;
1570
1571 lock_sock(sk);
1572 err = nf_getsockopt(sk, PF_INET, optname, optval,
1573 &len);
1574 release_sock(sk);
1575 if (err >= 0)
1576 err = put_user(len, optlen);
1577 return err;
1578 }
1579 #endif
1580 return err;
1581 }
1582 EXPORT_SYMBOL(ip_getsockopt);
1583
1584 #ifdef CONFIG_COMPAT
1585 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1586 char __user *optval, int __user *optlen)
1587 {
1588 int err;
1589
1590 if (optname == MCAST_MSFILTER)
1591 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1592 ip_getsockopt);
1593
1594 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1595 MSG_CMSG_COMPAT);
1596
1597 #ifdef CONFIG_NETFILTER
1598 /* we need to exclude all possible ENOPROTOOPTs except default case */
1599 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1600 !ip_mroute_opt(optname)) {
1601 int len;
1602
1603 if (get_user(len, optlen))
1604 return -EFAULT;
1605
1606 lock_sock(sk);
1607 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1608 release_sock(sk);
1609 if (err >= 0)
1610 err = put_user(len, optlen);
1611 return err;
1612 }
1613 #endif
1614 return err;
1615 }
1616 EXPORT_SYMBOL(compat_ip_getsockopt);
1617 #endif
ubuntu-bionic-kernel mirror