Merge tag 'locks-v3.16-2' of git://git.samba.org/jlayton/linux
[linux/fpc-iii.git] / net / ipv4 / ip_sockglue.c
blob64741b9386329ba6a76b00ea660d78983d2efb8a
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.
6 * The IP to API glue.
8 * Authors: see ip.c
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
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 #if IS_ENABLED(CONFIG_IPV6)
41 #include <net/transp_v6.h>
42 #endif
43 #include <net/ip_fib.h>
45 #include <linux/errqueue.h>
46 #include <asm/uaccess.h>
48 #define IP_CMSG_PKTINFO 1
49 #define IP_CMSG_TTL 2
50 #define IP_CMSG_TOS 4
51 #define IP_CMSG_RECVOPTS 8
52 #define IP_CMSG_RETOPTS 16
53 #define IP_CMSG_PASSSEC 32
54 #define IP_CMSG_ORIGDSTADDR 64
57 * SOL_IP control messages.
60 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
62 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
64 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
66 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
69 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
71 int ttl = ip_hdr(skb)->ttl;
72 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
75 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
77 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
80 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
82 if (IPCB(skb)->opt.optlen == 0)
83 return;
85 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
86 ip_hdr(skb) + 1);
90 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
92 unsigned char optbuf[sizeof(struct ip_options) + 40];
93 struct ip_options *opt = (struct ip_options *)optbuf;
95 if (IPCB(skb)->opt.optlen == 0)
96 return;
98 if (ip_options_echo(opt, skb)) {
99 msg->msg_flags |= MSG_CTRUNC;
100 return;
102 ip_options_undo(opt);
104 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
107 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
109 char *secdata;
110 u32 seclen, secid;
111 int err;
113 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
114 if (err)
115 return;
117 err = security_secid_to_secctx(secid, &secdata, &seclen);
118 if (err)
119 return;
121 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
122 security_release_secctx(secdata, seclen);
125 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
127 struct sockaddr_in sin;
128 const struct iphdr *iph = ip_hdr(skb);
129 __be16 *ports = (__be16 *)skb_transport_header(skb);
131 if (skb_transport_offset(skb) + 4 > skb->len)
132 return;
134 /* All current transport protocols have the port numbers in the
135 * first four bytes of the transport header and this function is
136 * written with this assumption in mind.
139 sin.sin_family = AF_INET;
140 sin.sin_addr.s_addr = iph->daddr;
141 sin.sin_port = ports[1];
142 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
144 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
147 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
149 struct inet_sock *inet = inet_sk(skb->sk);
150 unsigned int flags = inet->cmsg_flags;
152 /* Ordered by supposed usage frequency */
153 if (flags & 1)
154 ip_cmsg_recv_pktinfo(msg, skb);
155 if ((flags >>= 1) == 0)
156 return;
158 if (flags & 1)
159 ip_cmsg_recv_ttl(msg, skb);
160 if ((flags >>= 1) == 0)
161 return;
163 if (flags & 1)
164 ip_cmsg_recv_tos(msg, skb);
165 if ((flags >>= 1) == 0)
166 return;
168 if (flags & 1)
169 ip_cmsg_recv_opts(msg, skb);
170 if ((flags >>= 1) == 0)
171 return;
173 if (flags & 1)
174 ip_cmsg_recv_retopts(msg, skb);
175 if ((flags >>= 1) == 0)
176 return;
178 if (flags & 1)
179 ip_cmsg_recv_security(msg, skb);
181 if ((flags >>= 1) == 0)
182 return;
183 if (flags & 1)
184 ip_cmsg_recv_dstaddr(msg, skb);
187 EXPORT_SYMBOL(ip_cmsg_recv);
189 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc,
190 bool allow_ipv6)
192 int err, val;
193 struct cmsghdr *cmsg;
195 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
196 if (!CMSG_OK(msg, cmsg))
197 return -EINVAL;
198 #if defined(CONFIG_IPV6)
199 if (allow_ipv6 &&
200 cmsg->cmsg_level == SOL_IPV6 &&
201 cmsg->cmsg_type == IPV6_PKTINFO) {
202 struct in6_pktinfo *src_info;
204 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
205 return -EINVAL;
206 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
207 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
208 return -EINVAL;
209 ipc->oif = src_info->ipi6_ifindex;
210 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
211 continue;
213 #endif
214 if (cmsg->cmsg_level != SOL_IP)
215 continue;
216 switch (cmsg->cmsg_type) {
217 case IP_RETOPTS:
218 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
219 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
220 err < 40 ? err : 40);
221 if (err)
222 return err;
223 break;
224 case IP_PKTINFO:
226 struct in_pktinfo *info;
227 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
228 return -EINVAL;
229 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
230 ipc->oif = info->ipi_ifindex;
231 ipc->addr = info->ipi_spec_dst.s_addr;
232 break;
234 case IP_TTL:
235 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
236 return -EINVAL;
237 val = *(int *)CMSG_DATA(cmsg);
238 if (val < 1 || val > 255)
239 return -EINVAL;
240 ipc->ttl = val;
241 break;
242 case IP_TOS:
243 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
244 return -EINVAL;
245 val = *(int *)CMSG_DATA(cmsg);
246 if (val < 0 || val > 255)
247 return -EINVAL;
248 ipc->tos = val;
249 ipc->priority = rt_tos2priority(ipc->tos);
250 break;
252 default:
253 return -EINVAL;
256 return 0;
260 /* Special input handler for packets caught by router alert option.
261 They are selected only by protocol field, and then processed likely
262 local ones; but only if someone wants them! Otherwise, router
263 not running rsvpd will kill RSVP.
265 It is user level problem, what it will make with them.
266 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
267 but receiver should be enough clever f.e. to forward mtrace requests,
268 sent to multicast group to reach destination designated router.
270 struct ip_ra_chain __rcu *ip_ra_chain;
271 static DEFINE_SPINLOCK(ip_ra_lock);
274 static void ip_ra_destroy_rcu(struct rcu_head *head)
276 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
278 sock_put(ra->saved_sk);
279 kfree(ra);
282 int ip_ra_control(struct sock *sk, unsigned char on,
283 void (*destructor)(struct sock *))
285 struct ip_ra_chain *ra, *new_ra;
286 struct ip_ra_chain __rcu **rap;
288 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
289 return -EINVAL;
291 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
293 spin_lock_bh(&ip_ra_lock);
294 for (rap = &ip_ra_chain;
295 (ra = rcu_dereference_protected(*rap,
296 lockdep_is_held(&ip_ra_lock))) != NULL;
297 rap = &ra->next) {
298 if (ra->sk == sk) {
299 if (on) {
300 spin_unlock_bh(&ip_ra_lock);
301 kfree(new_ra);
302 return -EADDRINUSE;
304 /* dont let ip_call_ra_chain() use sk again */
305 ra->sk = NULL;
306 rcu_assign_pointer(*rap, ra->next);
307 spin_unlock_bh(&ip_ra_lock);
309 if (ra->destructor)
310 ra->destructor(sk);
312 * Delay sock_put(sk) and kfree(ra) after one rcu grace
313 * period. This guarantee ip_call_ra_chain() dont need
314 * to mess with socket refcounts.
316 ra->saved_sk = sk;
317 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
318 return 0;
321 if (new_ra == NULL) {
322 spin_unlock_bh(&ip_ra_lock);
323 return -ENOBUFS;
325 new_ra->sk = sk;
326 new_ra->destructor = destructor;
328 new_ra->next = ra;
329 rcu_assign_pointer(*rap, new_ra);
330 sock_hold(sk);
331 spin_unlock_bh(&ip_ra_lock);
333 return 0;
336 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
337 __be16 port, u32 info, u8 *payload)
339 struct sock_exterr_skb *serr;
341 skb = skb_clone(skb, GFP_ATOMIC);
342 if (!skb)
343 return;
345 serr = SKB_EXT_ERR(skb);
346 serr->ee.ee_errno = err;
347 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
348 serr->ee.ee_type = icmp_hdr(skb)->type;
349 serr->ee.ee_code = icmp_hdr(skb)->code;
350 serr->ee.ee_pad = 0;
351 serr->ee.ee_info = info;
352 serr->ee.ee_data = 0;
353 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
354 skb_network_header(skb);
355 serr->port = port;
357 if (skb_pull(skb, payload - skb->data) != NULL) {
358 skb_reset_transport_header(skb);
359 if (sock_queue_err_skb(sk, skb) == 0)
360 return;
362 kfree_skb(skb);
365 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
367 struct inet_sock *inet = inet_sk(sk);
368 struct sock_exterr_skb *serr;
369 struct iphdr *iph;
370 struct sk_buff *skb;
372 if (!inet->recverr)
373 return;
375 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
376 if (!skb)
377 return;
379 skb_put(skb, sizeof(struct iphdr));
380 skb_reset_network_header(skb);
381 iph = ip_hdr(skb);
382 iph->daddr = daddr;
384 serr = SKB_EXT_ERR(skb);
385 serr->ee.ee_errno = err;
386 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
387 serr->ee.ee_type = 0;
388 serr->ee.ee_code = 0;
389 serr->ee.ee_pad = 0;
390 serr->ee.ee_info = info;
391 serr->ee.ee_data = 0;
392 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
393 serr->port = port;
395 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
396 skb_reset_transport_header(skb);
398 if (sock_queue_err_skb(sk, skb))
399 kfree_skb(skb);
403 * Handle MSG_ERRQUEUE
405 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
407 struct sock_exterr_skb *serr;
408 struct sk_buff *skb, *skb2;
409 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
410 struct {
411 struct sock_extended_err ee;
412 struct sockaddr_in offender;
413 } errhdr;
414 int err;
415 int copied;
417 err = -EAGAIN;
418 skb = skb_dequeue(&sk->sk_error_queue);
419 if (skb == NULL)
420 goto out;
422 copied = skb->len;
423 if (copied > len) {
424 msg->msg_flags |= MSG_TRUNC;
425 copied = len;
427 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
428 if (err)
429 goto out_free_skb;
431 sock_recv_timestamp(msg, sk, skb);
433 serr = SKB_EXT_ERR(skb);
435 if (sin) {
436 sin->sin_family = AF_INET;
437 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
438 serr->addr_offset);
439 sin->sin_port = serr->port;
440 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
441 *addr_len = sizeof(*sin);
444 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
445 sin = &errhdr.offender;
446 sin->sin_family = AF_UNSPEC;
447 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
448 struct inet_sock *inet = inet_sk(sk);
450 sin->sin_family = AF_INET;
451 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
452 sin->sin_port = 0;
453 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
454 if (inet->cmsg_flags)
455 ip_cmsg_recv(msg, skb);
458 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
460 /* Now we could try to dump offended packet options */
462 msg->msg_flags |= MSG_ERRQUEUE;
463 err = copied;
465 /* Reset and regenerate socket error */
466 spin_lock_bh(&sk->sk_error_queue.lock);
467 sk->sk_err = 0;
468 skb2 = skb_peek(&sk->sk_error_queue);
469 if (skb2 != NULL) {
470 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
471 spin_unlock_bh(&sk->sk_error_queue.lock);
472 sk->sk_error_report(sk);
473 } else
474 spin_unlock_bh(&sk->sk_error_queue.lock);
476 out_free_skb:
477 kfree_skb(skb);
478 out:
479 return err;
484 * Socket option code for IP. This is the end of the line after any
485 * TCP,UDP etc options on an IP socket.
488 static int do_ip_setsockopt(struct sock *sk, int level,
489 int optname, char __user *optval, unsigned int optlen)
491 struct inet_sock *inet = inet_sk(sk);
492 int val = 0, err;
494 switch (optname) {
495 case IP_PKTINFO:
496 case IP_RECVTTL:
497 case IP_RECVOPTS:
498 case IP_RECVTOS:
499 case IP_RETOPTS:
500 case IP_TOS:
501 case IP_TTL:
502 case IP_HDRINCL:
503 case IP_MTU_DISCOVER:
504 case IP_RECVERR:
505 case IP_ROUTER_ALERT:
506 case IP_FREEBIND:
507 case IP_PASSSEC:
508 case IP_TRANSPARENT:
509 case IP_MINTTL:
510 case IP_NODEFRAG:
511 case IP_UNICAST_IF:
512 case IP_MULTICAST_TTL:
513 case IP_MULTICAST_ALL:
514 case IP_MULTICAST_LOOP:
515 case IP_RECVORIGDSTADDR:
516 if (optlen >= sizeof(int)) {
517 if (get_user(val, (int __user *) optval))
518 return -EFAULT;
519 } else if (optlen >= sizeof(char)) {
520 unsigned char ucval;
522 if (get_user(ucval, (unsigned char __user *) optval))
523 return -EFAULT;
524 val = (int) ucval;
528 /* If optlen==0, it is equivalent to val == 0 */
530 if (ip_mroute_opt(optname))
531 return ip_mroute_setsockopt(sk, optname, optval, optlen);
533 err = 0;
534 lock_sock(sk);
536 switch (optname) {
537 case IP_OPTIONS:
539 struct ip_options_rcu *old, *opt = NULL;
541 if (optlen > 40)
542 goto e_inval;
543 err = ip_options_get_from_user(sock_net(sk), &opt,
544 optval, optlen);
545 if (err)
546 break;
547 old = rcu_dereference_protected(inet->inet_opt,
548 sock_owned_by_user(sk));
549 if (inet->is_icsk) {
550 struct inet_connection_sock *icsk = inet_csk(sk);
551 #if IS_ENABLED(CONFIG_IPV6)
552 if (sk->sk_family == PF_INET ||
553 (!((1 << sk->sk_state) &
554 (TCPF_LISTEN | TCPF_CLOSE)) &&
555 inet->inet_daddr != LOOPBACK4_IPV6)) {
556 #endif
557 if (old)
558 icsk->icsk_ext_hdr_len -= old->opt.optlen;
559 if (opt)
560 icsk->icsk_ext_hdr_len += opt->opt.optlen;
561 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
562 #if IS_ENABLED(CONFIG_IPV6)
564 #endif
566 rcu_assign_pointer(inet->inet_opt, opt);
567 if (old)
568 kfree_rcu(old, rcu);
569 break;
571 case IP_PKTINFO:
572 if (val)
573 inet->cmsg_flags |= IP_CMSG_PKTINFO;
574 else
575 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
576 break;
577 case IP_RECVTTL:
578 if (val)
579 inet->cmsg_flags |= IP_CMSG_TTL;
580 else
581 inet->cmsg_flags &= ~IP_CMSG_TTL;
582 break;
583 case IP_RECVTOS:
584 if (val)
585 inet->cmsg_flags |= IP_CMSG_TOS;
586 else
587 inet->cmsg_flags &= ~IP_CMSG_TOS;
588 break;
589 case IP_RECVOPTS:
590 if (val)
591 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
592 else
593 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
594 break;
595 case IP_RETOPTS:
596 if (val)
597 inet->cmsg_flags |= IP_CMSG_RETOPTS;
598 else
599 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
600 break;
601 case IP_PASSSEC:
602 if (val)
603 inet->cmsg_flags |= IP_CMSG_PASSSEC;
604 else
605 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
606 break;
607 case IP_RECVORIGDSTADDR:
608 if (val)
609 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
610 else
611 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
612 break;
613 case IP_TOS: /* This sets both TOS and Precedence */
614 if (sk->sk_type == SOCK_STREAM) {
615 val &= ~INET_ECN_MASK;
616 val |= inet->tos & INET_ECN_MASK;
618 if (inet->tos != val) {
619 inet->tos = val;
620 sk->sk_priority = rt_tos2priority(val);
621 sk_dst_reset(sk);
623 break;
624 case IP_TTL:
625 if (optlen < 1)
626 goto e_inval;
627 if (val != -1 && (val < 1 || val > 255))
628 goto e_inval;
629 inet->uc_ttl = val;
630 break;
631 case IP_HDRINCL:
632 if (sk->sk_type != SOCK_RAW) {
633 err = -ENOPROTOOPT;
634 break;
636 inet->hdrincl = val ? 1 : 0;
637 break;
638 case IP_NODEFRAG:
639 if (sk->sk_type != SOCK_RAW) {
640 err = -ENOPROTOOPT;
641 break;
643 inet->nodefrag = val ? 1 : 0;
644 break;
645 case IP_MTU_DISCOVER:
646 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
647 goto e_inval;
648 inet->pmtudisc = val;
649 break;
650 case IP_RECVERR:
651 inet->recverr = !!val;
652 if (!val)
653 skb_queue_purge(&sk->sk_error_queue);
654 break;
655 case IP_MULTICAST_TTL:
656 if (sk->sk_type == SOCK_STREAM)
657 goto e_inval;
658 if (optlen < 1)
659 goto e_inval;
660 if (val == -1)
661 val = 1;
662 if (val < 0 || val > 255)
663 goto e_inval;
664 inet->mc_ttl = val;
665 break;
666 case IP_MULTICAST_LOOP:
667 if (optlen < 1)
668 goto e_inval;
669 inet->mc_loop = !!val;
670 break;
671 case IP_UNICAST_IF:
673 struct net_device *dev = NULL;
674 int ifindex;
676 if (optlen != sizeof(int))
677 goto e_inval;
679 ifindex = (__force int)ntohl((__force __be32)val);
680 if (ifindex == 0) {
681 inet->uc_index = 0;
682 err = 0;
683 break;
686 dev = dev_get_by_index(sock_net(sk), ifindex);
687 err = -EADDRNOTAVAIL;
688 if (!dev)
689 break;
690 dev_put(dev);
692 err = -EINVAL;
693 if (sk->sk_bound_dev_if)
694 break;
696 inet->uc_index = ifindex;
697 err = 0;
698 break;
700 case IP_MULTICAST_IF:
702 struct ip_mreqn mreq;
703 struct net_device *dev = NULL;
705 if (sk->sk_type == SOCK_STREAM)
706 goto e_inval;
708 * Check the arguments are allowable
711 if (optlen < sizeof(struct in_addr))
712 goto e_inval;
714 err = -EFAULT;
715 if (optlen >= sizeof(struct ip_mreqn)) {
716 if (copy_from_user(&mreq, optval, sizeof(mreq)))
717 break;
718 } else {
719 memset(&mreq, 0, sizeof(mreq));
720 if (optlen >= sizeof(struct ip_mreq)) {
721 if (copy_from_user(&mreq, optval,
722 sizeof(struct ip_mreq)))
723 break;
724 } else if (optlen >= sizeof(struct in_addr)) {
725 if (copy_from_user(&mreq.imr_address, optval,
726 sizeof(struct in_addr)))
727 break;
731 if (!mreq.imr_ifindex) {
732 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
733 inet->mc_index = 0;
734 inet->mc_addr = 0;
735 err = 0;
736 break;
738 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
739 if (dev)
740 mreq.imr_ifindex = dev->ifindex;
741 } else
742 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
745 err = -EADDRNOTAVAIL;
746 if (!dev)
747 break;
748 dev_put(dev);
750 err = -EINVAL;
751 if (sk->sk_bound_dev_if &&
752 mreq.imr_ifindex != sk->sk_bound_dev_if)
753 break;
755 inet->mc_index = mreq.imr_ifindex;
756 inet->mc_addr = mreq.imr_address.s_addr;
757 err = 0;
758 break;
761 case IP_ADD_MEMBERSHIP:
762 case IP_DROP_MEMBERSHIP:
764 struct ip_mreqn mreq;
766 err = -EPROTO;
767 if (inet_sk(sk)->is_icsk)
768 break;
770 if (optlen < sizeof(struct ip_mreq))
771 goto e_inval;
772 err = -EFAULT;
773 if (optlen >= sizeof(struct ip_mreqn)) {
774 if (copy_from_user(&mreq, optval, sizeof(mreq)))
775 break;
776 } else {
777 memset(&mreq, 0, sizeof(mreq));
778 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
779 break;
782 if (optname == IP_ADD_MEMBERSHIP)
783 err = ip_mc_join_group(sk, &mreq);
784 else
785 err = ip_mc_leave_group(sk, &mreq);
786 break;
788 case IP_MSFILTER:
790 struct ip_msfilter *msf;
792 if (optlen < IP_MSFILTER_SIZE(0))
793 goto e_inval;
794 if (optlen > sysctl_optmem_max) {
795 err = -ENOBUFS;
796 break;
798 msf = kmalloc(optlen, GFP_KERNEL);
799 if (!msf) {
800 err = -ENOBUFS;
801 break;
803 err = -EFAULT;
804 if (copy_from_user(msf, optval, optlen)) {
805 kfree(msf);
806 break;
808 /* numsrc >= (1G-4) overflow in 32 bits */
809 if (msf->imsf_numsrc >= 0x3ffffffcU ||
810 msf->imsf_numsrc > sysctl_igmp_max_msf) {
811 kfree(msf);
812 err = -ENOBUFS;
813 break;
815 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
816 kfree(msf);
817 err = -EINVAL;
818 break;
820 err = ip_mc_msfilter(sk, msf, 0);
821 kfree(msf);
822 break;
824 case IP_BLOCK_SOURCE:
825 case IP_UNBLOCK_SOURCE:
826 case IP_ADD_SOURCE_MEMBERSHIP:
827 case IP_DROP_SOURCE_MEMBERSHIP:
829 struct ip_mreq_source mreqs;
830 int omode, add;
832 if (optlen != sizeof(struct ip_mreq_source))
833 goto e_inval;
834 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
835 err = -EFAULT;
836 break;
838 if (optname == IP_BLOCK_SOURCE) {
839 omode = MCAST_EXCLUDE;
840 add = 1;
841 } else if (optname == IP_UNBLOCK_SOURCE) {
842 omode = MCAST_EXCLUDE;
843 add = 0;
844 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
845 struct ip_mreqn mreq;
847 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
848 mreq.imr_address.s_addr = mreqs.imr_interface;
849 mreq.imr_ifindex = 0;
850 err = ip_mc_join_group(sk, &mreq);
851 if (err && err != -EADDRINUSE)
852 break;
853 omode = MCAST_INCLUDE;
854 add = 1;
855 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
856 omode = MCAST_INCLUDE;
857 add = 0;
859 err = ip_mc_source(add, omode, sk, &mreqs, 0);
860 break;
862 case MCAST_JOIN_GROUP:
863 case MCAST_LEAVE_GROUP:
865 struct group_req greq;
866 struct sockaddr_in *psin;
867 struct ip_mreqn mreq;
869 if (optlen < sizeof(struct group_req))
870 goto e_inval;
871 err = -EFAULT;
872 if (copy_from_user(&greq, optval, sizeof(greq)))
873 break;
874 psin = (struct sockaddr_in *)&greq.gr_group;
875 if (psin->sin_family != AF_INET)
876 goto e_inval;
877 memset(&mreq, 0, sizeof(mreq));
878 mreq.imr_multiaddr = psin->sin_addr;
879 mreq.imr_ifindex = greq.gr_interface;
881 if (optname == MCAST_JOIN_GROUP)
882 err = ip_mc_join_group(sk, &mreq);
883 else
884 err = ip_mc_leave_group(sk, &mreq);
885 break;
887 case MCAST_JOIN_SOURCE_GROUP:
888 case MCAST_LEAVE_SOURCE_GROUP:
889 case MCAST_BLOCK_SOURCE:
890 case MCAST_UNBLOCK_SOURCE:
892 struct group_source_req greqs;
893 struct ip_mreq_source mreqs;
894 struct sockaddr_in *psin;
895 int omode, add;
897 if (optlen != sizeof(struct group_source_req))
898 goto e_inval;
899 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
900 err = -EFAULT;
901 break;
903 if (greqs.gsr_group.ss_family != AF_INET ||
904 greqs.gsr_source.ss_family != AF_INET) {
905 err = -EADDRNOTAVAIL;
906 break;
908 psin = (struct sockaddr_in *)&greqs.gsr_group;
909 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
910 psin = (struct sockaddr_in *)&greqs.gsr_source;
911 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
912 mreqs.imr_interface = 0; /* use index for mc_source */
914 if (optname == MCAST_BLOCK_SOURCE) {
915 omode = MCAST_EXCLUDE;
916 add = 1;
917 } else if (optname == MCAST_UNBLOCK_SOURCE) {
918 omode = MCAST_EXCLUDE;
919 add = 0;
920 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
921 struct ip_mreqn mreq;
923 psin = (struct sockaddr_in *)&greqs.gsr_group;
924 mreq.imr_multiaddr = psin->sin_addr;
925 mreq.imr_address.s_addr = 0;
926 mreq.imr_ifindex = greqs.gsr_interface;
927 err = ip_mc_join_group(sk, &mreq);
928 if (err && err != -EADDRINUSE)
929 break;
930 greqs.gsr_interface = mreq.imr_ifindex;
931 omode = MCAST_INCLUDE;
932 add = 1;
933 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
934 omode = MCAST_INCLUDE;
935 add = 0;
937 err = ip_mc_source(add, omode, sk, &mreqs,
938 greqs.gsr_interface);
939 break;
941 case MCAST_MSFILTER:
943 struct sockaddr_in *psin;
944 struct ip_msfilter *msf = NULL;
945 struct group_filter *gsf = NULL;
946 int msize, i, ifindex;
948 if (optlen < GROUP_FILTER_SIZE(0))
949 goto e_inval;
950 if (optlen > sysctl_optmem_max) {
951 err = -ENOBUFS;
952 break;
954 gsf = kmalloc(optlen, GFP_KERNEL);
955 if (!gsf) {
956 err = -ENOBUFS;
957 break;
959 err = -EFAULT;
960 if (copy_from_user(gsf, optval, optlen))
961 goto mc_msf_out;
963 /* numsrc >= (4G-140)/128 overflow in 32 bits */
964 if (gsf->gf_numsrc >= 0x1ffffff ||
965 gsf->gf_numsrc > sysctl_igmp_max_msf) {
966 err = -ENOBUFS;
967 goto mc_msf_out;
969 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
970 err = -EINVAL;
971 goto mc_msf_out;
973 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
974 msf = kmalloc(msize, GFP_KERNEL);
975 if (!msf) {
976 err = -ENOBUFS;
977 goto mc_msf_out;
979 ifindex = gsf->gf_interface;
980 psin = (struct sockaddr_in *)&gsf->gf_group;
981 if (psin->sin_family != AF_INET) {
982 err = -EADDRNOTAVAIL;
983 goto mc_msf_out;
985 msf->imsf_multiaddr = psin->sin_addr.s_addr;
986 msf->imsf_interface = 0;
987 msf->imsf_fmode = gsf->gf_fmode;
988 msf->imsf_numsrc = gsf->gf_numsrc;
989 err = -EADDRNOTAVAIL;
990 for (i = 0; i < gsf->gf_numsrc; ++i) {
991 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
993 if (psin->sin_family != AF_INET)
994 goto mc_msf_out;
995 msf->imsf_slist[i] = psin->sin_addr.s_addr;
997 kfree(gsf);
998 gsf = NULL;
1000 err = ip_mc_msfilter(sk, msf, ifindex);
1001 mc_msf_out:
1002 kfree(msf);
1003 kfree(gsf);
1004 break;
1006 case IP_MULTICAST_ALL:
1007 if (optlen < 1)
1008 goto e_inval;
1009 if (val != 0 && val != 1)
1010 goto e_inval;
1011 inet->mc_all = val;
1012 break;
1013 case IP_ROUTER_ALERT:
1014 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1015 break;
1017 case IP_FREEBIND:
1018 if (optlen < 1)
1019 goto e_inval;
1020 inet->freebind = !!val;
1021 break;
1023 case IP_IPSEC_POLICY:
1024 case IP_XFRM_POLICY:
1025 err = -EPERM;
1026 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1027 break;
1028 err = xfrm_user_policy(sk, optname, optval, optlen);
1029 break;
1031 case IP_TRANSPARENT:
1032 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1033 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1034 err = -EPERM;
1035 break;
1037 if (optlen < 1)
1038 goto e_inval;
1039 inet->transparent = !!val;
1040 break;
1042 case IP_MINTTL:
1043 if (optlen < 1)
1044 goto e_inval;
1045 if (val < 0 || val > 255)
1046 goto e_inval;
1047 inet->min_ttl = val;
1048 break;
1050 default:
1051 err = -ENOPROTOOPT;
1052 break;
1054 release_sock(sk);
1055 return err;
1057 e_inval:
1058 release_sock(sk);
1059 return -EINVAL;
1063 * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1064 * @sk: socket
1065 * @skb: buffer
1067 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1068 * destination in skb->cb[] before dst drop.
1069 * This way, receiver doesn't make cache line misses to read rtable.
1071 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1073 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1074 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1075 ipv6_sk_rxinfo(sk);
1077 if (prepare && skb_rtable(skb)) {
1078 pktinfo->ipi_ifindex = inet_iif(skb);
1079 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1080 } else {
1081 pktinfo->ipi_ifindex = 0;
1082 pktinfo->ipi_spec_dst.s_addr = 0;
1084 skb_dst_drop(skb);
1087 int ip_setsockopt(struct sock *sk, int level,
1088 int optname, char __user *optval, unsigned int optlen)
1090 int err;
1092 if (level != SOL_IP)
1093 return -ENOPROTOOPT;
1095 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1096 #ifdef CONFIG_NETFILTER
1097 /* we need to exclude all possible ENOPROTOOPTs except default case */
1098 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1099 optname != IP_IPSEC_POLICY &&
1100 optname != IP_XFRM_POLICY &&
1101 !ip_mroute_opt(optname)) {
1102 lock_sock(sk);
1103 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1104 release_sock(sk);
1106 #endif
1107 return err;
1109 EXPORT_SYMBOL(ip_setsockopt);
1111 #ifdef CONFIG_COMPAT
1112 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1113 char __user *optval, unsigned int optlen)
1115 int err;
1117 if (level != SOL_IP)
1118 return -ENOPROTOOPT;
1120 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1121 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1122 ip_setsockopt);
1124 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1125 #ifdef CONFIG_NETFILTER
1126 /* we need to exclude all possible ENOPROTOOPTs except default case */
1127 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1128 optname != IP_IPSEC_POLICY &&
1129 optname != IP_XFRM_POLICY &&
1130 !ip_mroute_opt(optname)) {
1131 lock_sock(sk);
1132 err = compat_nf_setsockopt(sk, PF_INET, optname,
1133 optval, optlen);
1134 release_sock(sk);
1136 #endif
1137 return err;
1139 EXPORT_SYMBOL(compat_ip_setsockopt);
1140 #endif
1143 * Get the options. Note for future reference. The GET of IP options gets
1144 * the _received_ ones. The set sets the _sent_ ones.
1147 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1148 char __user *optval, int __user *optlen, unsigned int flags)
1150 struct inet_sock *inet = inet_sk(sk);
1151 int val;
1152 int len;
1154 if (level != SOL_IP)
1155 return -EOPNOTSUPP;
1157 if (ip_mroute_opt(optname))
1158 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1160 if (get_user(len, optlen))
1161 return -EFAULT;
1162 if (len < 0)
1163 return -EINVAL;
1165 lock_sock(sk);
1167 switch (optname) {
1168 case IP_OPTIONS:
1170 unsigned char optbuf[sizeof(struct ip_options)+40];
1171 struct ip_options *opt = (struct ip_options *)optbuf;
1172 struct ip_options_rcu *inet_opt;
1174 inet_opt = rcu_dereference_protected(inet->inet_opt,
1175 sock_owned_by_user(sk));
1176 opt->optlen = 0;
1177 if (inet_opt)
1178 memcpy(optbuf, &inet_opt->opt,
1179 sizeof(struct ip_options) +
1180 inet_opt->opt.optlen);
1181 release_sock(sk);
1183 if (opt->optlen == 0)
1184 return put_user(0, optlen);
1186 ip_options_undo(opt);
1188 len = min_t(unsigned int, len, opt->optlen);
1189 if (put_user(len, optlen))
1190 return -EFAULT;
1191 if (copy_to_user(optval, opt->__data, len))
1192 return -EFAULT;
1193 return 0;
1195 case IP_PKTINFO:
1196 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1197 break;
1198 case IP_RECVTTL:
1199 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1200 break;
1201 case IP_RECVTOS:
1202 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1203 break;
1204 case IP_RECVOPTS:
1205 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1206 break;
1207 case IP_RETOPTS:
1208 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1209 break;
1210 case IP_PASSSEC:
1211 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1212 break;
1213 case IP_RECVORIGDSTADDR:
1214 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1215 break;
1216 case IP_TOS:
1217 val = inet->tos;
1218 break;
1219 case IP_TTL:
1220 val = (inet->uc_ttl == -1 ?
1221 sysctl_ip_default_ttl :
1222 inet->uc_ttl);
1223 break;
1224 case IP_HDRINCL:
1225 val = inet->hdrincl;
1226 break;
1227 case IP_NODEFRAG:
1228 val = inet->nodefrag;
1229 break;
1230 case IP_MTU_DISCOVER:
1231 val = inet->pmtudisc;
1232 break;
1233 case IP_MTU:
1235 struct dst_entry *dst;
1236 val = 0;
1237 dst = sk_dst_get(sk);
1238 if (dst) {
1239 val = dst_mtu(dst);
1240 dst_release(dst);
1242 if (!val) {
1243 release_sock(sk);
1244 return -ENOTCONN;
1246 break;
1248 case IP_RECVERR:
1249 val = inet->recverr;
1250 break;
1251 case IP_MULTICAST_TTL:
1252 val = inet->mc_ttl;
1253 break;
1254 case IP_MULTICAST_LOOP:
1255 val = inet->mc_loop;
1256 break;
1257 case IP_UNICAST_IF:
1258 val = (__force int)htonl((__u32) inet->uc_index);
1259 break;
1260 case IP_MULTICAST_IF:
1262 struct in_addr addr;
1263 len = min_t(unsigned int, len, sizeof(struct in_addr));
1264 addr.s_addr = inet->mc_addr;
1265 release_sock(sk);
1267 if (put_user(len, optlen))
1268 return -EFAULT;
1269 if (copy_to_user(optval, &addr, len))
1270 return -EFAULT;
1271 return 0;
1273 case IP_MSFILTER:
1275 struct ip_msfilter msf;
1276 int err;
1278 if (len < IP_MSFILTER_SIZE(0)) {
1279 release_sock(sk);
1280 return -EINVAL;
1282 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1283 release_sock(sk);
1284 return -EFAULT;
1286 err = ip_mc_msfget(sk, &msf,
1287 (struct ip_msfilter __user *)optval, optlen);
1288 release_sock(sk);
1289 return err;
1291 case MCAST_MSFILTER:
1293 struct group_filter gsf;
1294 int err;
1296 if (len < GROUP_FILTER_SIZE(0)) {
1297 release_sock(sk);
1298 return -EINVAL;
1300 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1301 release_sock(sk);
1302 return -EFAULT;
1304 err = ip_mc_gsfget(sk, &gsf,
1305 (struct group_filter __user *)optval,
1306 optlen);
1307 release_sock(sk);
1308 return err;
1310 case IP_MULTICAST_ALL:
1311 val = inet->mc_all;
1312 break;
1313 case IP_PKTOPTIONS:
1315 struct msghdr msg;
1317 release_sock(sk);
1319 if (sk->sk_type != SOCK_STREAM)
1320 return -ENOPROTOOPT;
1322 msg.msg_control = optval;
1323 msg.msg_controllen = len;
1324 msg.msg_flags = flags;
1326 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1327 struct in_pktinfo info;
1329 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1330 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1331 info.ipi_ifindex = inet->mc_index;
1332 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1334 if (inet->cmsg_flags & IP_CMSG_TTL) {
1335 int hlim = inet->mc_ttl;
1336 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1338 if (inet->cmsg_flags & IP_CMSG_TOS) {
1339 int tos = inet->rcv_tos;
1340 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1342 len -= msg.msg_controllen;
1343 return put_user(len, optlen);
1345 case IP_FREEBIND:
1346 val = inet->freebind;
1347 break;
1348 case IP_TRANSPARENT:
1349 val = inet->transparent;
1350 break;
1351 case IP_MINTTL:
1352 val = inet->min_ttl;
1353 break;
1354 default:
1355 release_sock(sk);
1356 return -ENOPROTOOPT;
1358 release_sock(sk);
1360 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1361 unsigned char ucval = (unsigned char)val;
1362 len = 1;
1363 if (put_user(len, optlen))
1364 return -EFAULT;
1365 if (copy_to_user(optval, &ucval, 1))
1366 return -EFAULT;
1367 } else {
1368 len = min_t(unsigned int, sizeof(int), len);
1369 if (put_user(len, optlen))
1370 return -EFAULT;
1371 if (copy_to_user(optval, &val, len))
1372 return -EFAULT;
1374 return 0;
1377 int ip_getsockopt(struct sock *sk, int level,
1378 int optname, char __user *optval, int __user *optlen)
1380 int err;
1382 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1383 #ifdef CONFIG_NETFILTER
1384 /* we need to exclude all possible ENOPROTOOPTs except default case */
1385 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1386 !ip_mroute_opt(optname)) {
1387 int len;
1389 if (get_user(len, optlen))
1390 return -EFAULT;
1392 lock_sock(sk);
1393 err = nf_getsockopt(sk, PF_INET, optname, optval,
1394 &len);
1395 release_sock(sk);
1396 if (err >= 0)
1397 err = put_user(len, optlen);
1398 return err;
1400 #endif
1401 return err;
1403 EXPORT_SYMBOL(ip_getsockopt);
1405 #ifdef CONFIG_COMPAT
1406 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1407 char __user *optval, int __user *optlen)
1409 int err;
1411 if (optname == MCAST_MSFILTER)
1412 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1413 ip_getsockopt);
1415 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1416 MSG_CMSG_COMPAT);
1418 #ifdef CONFIG_NETFILTER
1419 /* we need to exclude all possible ENOPROTOOPTs except default case */
1420 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1421 !ip_mroute_opt(optname)) {
1422 int len;
1424 if (get_user(len, optlen))
1425 return -EFAULT;
1427 lock_sock(sk);
1428 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1429 release_sock(sk);
1430 if (err >= 0)
1431 err = put_user(len, optlen);
1432 return err;
1434 #endif
1435 return err;
1437 EXPORT_SYMBOL(compat_ip_getsockopt);
1438 #endif