scsi: use REQ_TYPE_FS for flush request
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv4 / ip_sockglue.c
blobce231780a2b14ef4e15a33baad4b0c887e96df16
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/route.h>
37 #include <net/xfrm.h>
38 #include <net/compat.h>
39 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
40 #include <net/transp_v6.h>
41 #endif
43 #include <linux/errqueue.h>
44 #include <asm/uaccess.h>
46 #define IP_CMSG_PKTINFO 1
47 #define IP_CMSG_TTL 2
48 #define IP_CMSG_TOS 4
49 #define IP_CMSG_RECVOPTS 8
50 #define IP_CMSG_RETOPTS 16
51 #define IP_CMSG_PASSSEC 32
52 #define IP_CMSG_ORIGDSTADDR 64
55 * SOL_IP control messages.
58 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
60 struct in_pktinfo info;
61 struct rtable *rt = skb_rtable(skb);
63 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
64 if (rt) {
65 info.ipi_ifindex = rt->rt_iif;
66 info.ipi_spec_dst.s_addr = rt->rt_spec_dst;
67 } else {
68 info.ipi_ifindex = 0;
69 info.ipi_spec_dst.s_addr = 0;
72 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
75 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
77 int ttl = ip_hdr(skb)->ttl;
78 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
81 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
83 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
86 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
88 if (IPCB(skb)->opt.optlen == 0)
89 return;
91 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
92 ip_hdr(skb) + 1);
96 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
98 unsigned char optbuf[sizeof(struct ip_options) + 40];
99 struct ip_options * opt = (struct ip_options *)optbuf;
101 if (IPCB(skb)->opt.optlen == 0)
102 return;
104 if (ip_options_echo(opt, skb)) {
105 msg->msg_flags |= MSG_CTRUNC;
106 return;
108 ip_options_undo(opt);
110 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
113 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
115 char *secdata;
116 u32 seclen, secid;
117 int err;
119 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
120 if (err)
121 return;
123 err = security_secid_to_secctx(secid, &secdata, &seclen);
124 if (err)
125 return;
127 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
128 security_release_secctx(secdata, seclen);
131 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
133 struct sockaddr_in sin;
134 struct iphdr *iph = ip_hdr(skb);
135 __be16 *ports = (__be16 *)skb_transport_header(skb);
137 if (skb_transport_offset(skb) + 4 > skb->len)
138 return;
140 /* All current transport protocols have the port numbers in the
141 * first four bytes of the transport header and this function is
142 * written with this assumption in mind.
145 sin.sin_family = AF_INET;
146 sin.sin_addr.s_addr = iph->daddr;
147 sin.sin_port = ports[1];
148 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
150 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
153 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
155 struct inet_sock *inet = inet_sk(skb->sk);
156 unsigned flags = inet->cmsg_flags;
158 /* Ordered by supposed usage frequency */
159 if (flags & 1)
160 ip_cmsg_recv_pktinfo(msg, skb);
161 if ((flags >>= 1) == 0)
162 return;
164 if (flags & 1)
165 ip_cmsg_recv_ttl(msg, skb);
166 if ((flags >>= 1) == 0)
167 return;
169 if (flags & 1)
170 ip_cmsg_recv_tos(msg, skb);
171 if ((flags >>= 1) == 0)
172 return;
174 if (flags & 1)
175 ip_cmsg_recv_opts(msg, skb);
176 if ((flags >>= 1) == 0)
177 return;
179 if (flags & 1)
180 ip_cmsg_recv_retopts(msg, skb);
181 if ((flags >>= 1) == 0)
182 return;
184 if (flags & 1)
185 ip_cmsg_recv_security(msg, skb);
187 if ((flags >>= 1) == 0)
188 return;
189 if (flags & 1)
190 ip_cmsg_recv_dstaddr(msg, skb);
193 EXPORT_SYMBOL(ip_cmsg_recv);
195 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
197 int err;
198 struct cmsghdr *cmsg;
200 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
201 if (!CMSG_OK(msg, cmsg))
202 return -EINVAL;
203 if (cmsg->cmsg_level != SOL_IP)
204 continue;
205 switch (cmsg->cmsg_type) {
206 case IP_RETOPTS:
207 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
208 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
209 err < 40 ? err : 40);
210 if (err)
211 return err;
212 break;
213 case IP_PKTINFO:
215 struct in_pktinfo *info;
216 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
217 return -EINVAL;
218 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
219 ipc->oif = info->ipi_ifindex;
220 ipc->addr = info->ipi_spec_dst.s_addr;
221 break;
223 default:
224 return -EINVAL;
227 return 0;
231 /* Special input handler for packets caught by router alert option.
232 They are selected only by protocol field, and then processed likely
233 local ones; but only if someone wants them! Otherwise, router
234 not running rsvpd will kill RSVP.
236 It is user level problem, what it will make with them.
237 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
238 but receiver should be enough clever f.e. to forward mtrace requests,
239 sent to multicast group to reach destination designated router.
241 struct ip_ra_chain *ip_ra_chain;
242 DEFINE_RWLOCK(ip_ra_lock);
244 int ip_ra_control(struct sock *sk, unsigned char on,
245 void (*destructor)(struct sock *))
247 struct ip_ra_chain *ra, *new_ra, **rap;
249 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
250 return -EINVAL;
252 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
254 write_lock_bh(&ip_ra_lock);
255 for (rap = &ip_ra_chain; (ra = *rap) != NULL; rap = &ra->next) {
256 if (ra->sk == sk) {
257 if (on) {
258 write_unlock_bh(&ip_ra_lock);
259 kfree(new_ra);
260 return -EADDRINUSE;
262 *rap = ra->next;
263 write_unlock_bh(&ip_ra_lock);
265 if (ra->destructor)
266 ra->destructor(sk);
267 sock_put(sk);
268 kfree(ra);
269 return 0;
272 if (new_ra == NULL) {
273 write_unlock_bh(&ip_ra_lock);
274 return -ENOBUFS;
276 new_ra->sk = sk;
277 new_ra->destructor = destructor;
279 new_ra->next = ra;
280 *rap = new_ra;
281 sock_hold(sk);
282 write_unlock_bh(&ip_ra_lock);
284 return 0;
287 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
288 __be16 port, u32 info, u8 *payload)
290 struct sock_exterr_skb *serr;
292 skb = skb_clone(skb, GFP_ATOMIC);
293 if (!skb)
294 return;
296 serr = SKB_EXT_ERR(skb);
297 serr->ee.ee_errno = err;
298 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
299 serr->ee.ee_type = icmp_hdr(skb)->type;
300 serr->ee.ee_code = icmp_hdr(skb)->code;
301 serr->ee.ee_pad = 0;
302 serr->ee.ee_info = info;
303 serr->ee.ee_data = 0;
304 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
305 skb_network_header(skb);
306 serr->port = port;
308 if (skb_pull(skb, payload - skb->data) != NULL) {
309 skb_reset_transport_header(skb);
310 if (sock_queue_err_skb(sk, skb) == 0)
311 return;
313 kfree_skb(skb);
316 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
318 struct inet_sock *inet = inet_sk(sk);
319 struct sock_exterr_skb *serr;
320 struct iphdr *iph;
321 struct sk_buff *skb;
323 if (!inet->recverr)
324 return;
326 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
327 if (!skb)
328 return;
330 skb_put(skb, sizeof(struct iphdr));
331 skb_reset_network_header(skb);
332 iph = ip_hdr(skb);
333 iph->daddr = daddr;
335 serr = SKB_EXT_ERR(skb);
336 serr->ee.ee_errno = err;
337 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
338 serr->ee.ee_type = 0;
339 serr->ee.ee_code = 0;
340 serr->ee.ee_pad = 0;
341 serr->ee.ee_info = info;
342 serr->ee.ee_data = 0;
343 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
344 serr->port = port;
346 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
347 skb_reset_transport_header(skb);
349 if (sock_queue_err_skb(sk, skb))
350 kfree_skb(skb);
354 * Handle MSG_ERRQUEUE
356 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
358 struct sock_exterr_skb *serr;
359 struct sk_buff *skb, *skb2;
360 struct sockaddr_in *sin;
361 struct {
362 struct sock_extended_err ee;
363 struct sockaddr_in offender;
364 } errhdr;
365 int err;
366 int copied;
368 err = -EAGAIN;
369 skb = skb_dequeue(&sk->sk_error_queue);
370 if (skb == NULL)
371 goto out;
373 copied = skb->len;
374 if (copied > len) {
375 msg->msg_flags |= MSG_TRUNC;
376 copied = len;
378 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
379 if (err)
380 goto out_free_skb;
382 sock_recv_timestamp(msg, sk, skb);
384 serr = SKB_EXT_ERR(skb);
386 sin = (struct sockaddr_in *)msg->msg_name;
387 if (sin) {
388 sin->sin_family = AF_INET;
389 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
390 serr->addr_offset);
391 sin->sin_port = serr->port;
392 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
395 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
396 sin = &errhdr.offender;
397 sin->sin_family = AF_UNSPEC;
398 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
399 struct inet_sock *inet = inet_sk(sk);
401 sin->sin_family = AF_INET;
402 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
403 sin->sin_port = 0;
404 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
405 if (inet->cmsg_flags)
406 ip_cmsg_recv(msg, skb);
409 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
411 /* Now we could try to dump offended packet options */
413 msg->msg_flags |= MSG_ERRQUEUE;
414 err = copied;
416 /* Reset and regenerate socket error */
417 spin_lock_bh(&sk->sk_error_queue.lock);
418 sk->sk_err = 0;
419 skb2 = skb_peek(&sk->sk_error_queue);
420 if (skb2 != NULL) {
421 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
422 spin_unlock_bh(&sk->sk_error_queue.lock);
423 sk->sk_error_report(sk);
424 } else
425 spin_unlock_bh(&sk->sk_error_queue.lock);
427 out_free_skb:
428 kfree_skb(skb);
429 out:
430 return err;
435 * Socket option code for IP. This is the end of the line after any
436 * TCP,UDP etc options on an IP socket.
439 static int do_ip_setsockopt(struct sock *sk, int level,
440 int optname, char __user *optval, unsigned int optlen)
442 struct inet_sock *inet = inet_sk(sk);
443 int val = 0, err;
445 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
446 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
447 (1<<IP_RETOPTS) | (1<<IP_TOS) |
448 (1<<IP_TTL) | (1<<IP_HDRINCL) |
449 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
450 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) |
451 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) |
452 (1<<IP_MINTTL))) ||
453 optname == IP_MULTICAST_TTL ||
454 optname == IP_MULTICAST_ALL ||
455 optname == IP_MULTICAST_LOOP ||
456 optname == IP_RECVORIGDSTADDR) {
457 if (optlen >= sizeof(int)) {
458 if (get_user(val, (int __user *) optval))
459 return -EFAULT;
460 } else if (optlen >= sizeof(char)) {
461 unsigned char ucval;
463 if (get_user(ucval, (unsigned char __user *) optval))
464 return -EFAULT;
465 val = (int) ucval;
469 /* If optlen==0, it is equivalent to val == 0 */
471 if (ip_mroute_opt(optname))
472 return ip_mroute_setsockopt(sk, optname, optval, optlen);
474 err = 0;
475 lock_sock(sk);
477 switch (optname) {
478 case IP_OPTIONS:
480 struct ip_options *opt = NULL;
481 if (optlen > 40)
482 goto e_inval;
483 err = ip_options_get_from_user(sock_net(sk), &opt,
484 optval, optlen);
485 if (err)
486 break;
487 if (inet->is_icsk) {
488 struct inet_connection_sock *icsk = inet_csk(sk);
489 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
490 if (sk->sk_family == PF_INET ||
491 (!((1 << sk->sk_state) &
492 (TCPF_LISTEN | TCPF_CLOSE)) &&
493 inet->inet_daddr != LOOPBACK4_IPV6)) {
494 #endif
495 if (inet->opt)
496 icsk->icsk_ext_hdr_len -= inet->opt->optlen;
497 if (opt)
498 icsk->icsk_ext_hdr_len += opt->optlen;
499 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
500 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
502 #endif
504 opt = xchg(&inet->opt, opt);
505 kfree(opt);
506 break;
508 case IP_PKTINFO:
509 if (val)
510 inet->cmsg_flags |= IP_CMSG_PKTINFO;
511 else
512 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
513 break;
514 case IP_RECVTTL:
515 if (val)
516 inet->cmsg_flags |= IP_CMSG_TTL;
517 else
518 inet->cmsg_flags &= ~IP_CMSG_TTL;
519 break;
520 case IP_RECVTOS:
521 if (val)
522 inet->cmsg_flags |= IP_CMSG_TOS;
523 else
524 inet->cmsg_flags &= ~IP_CMSG_TOS;
525 break;
526 case IP_RECVOPTS:
527 if (val)
528 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
529 else
530 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
531 break;
532 case IP_RETOPTS:
533 if (val)
534 inet->cmsg_flags |= IP_CMSG_RETOPTS;
535 else
536 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
537 break;
538 case IP_PASSSEC:
539 if (val)
540 inet->cmsg_flags |= IP_CMSG_PASSSEC;
541 else
542 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
543 break;
544 case IP_RECVORIGDSTADDR:
545 if (val)
546 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
547 else
548 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
549 break;
550 case IP_TOS: /* This sets both TOS and Precedence */
551 if (sk->sk_type == SOCK_STREAM) {
552 val &= ~3;
553 val |= inet->tos & 3;
555 if (inet->tos != val) {
556 inet->tos = val;
557 sk->sk_priority = rt_tos2priority(val);
558 sk_dst_reset(sk);
560 break;
561 case IP_TTL:
562 if (optlen < 1)
563 goto e_inval;
564 if (val != -1 && (val < 0 || val > 255))
565 goto e_inval;
566 inet->uc_ttl = val;
567 break;
568 case IP_HDRINCL:
569 if (sk->sk_type != SOCK_RAW) {
570 err = -ENOPROTOOPT;
571 break;
573 inet->hdrincl = val ? 1 : 0;
574 break;
575 case IP_MTU_DISCOVER:
576 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
577 goto e_inval;
578 inet->pmtudisc = val;
579 break;
580 case IP_RECVERR:
581 inet->recverr = !!val;
582 if (!val)
583 skb_queue_purge(&sk->sk_error_queue);
584 break;
585 case IP_MULTICAST_TTL:
586 if (sk->sk_type == SOCK_STREAM)
587 goto e_inval;
588 if (optlen < 1)
589 goto e_inval;
590 if (val == -1)
591 val = 1;
592 if (val < 0 || val > 255)
593 goto e_inval;
594 inet->mc_ttl = val;
595 break;
596 case IP_MULTICAST_LOOP:
597 if (optlen < 1)
598 goto e_inval;
599 inet->mc_loop = !!val;
600 break;
601 case IP_MULTICAST_IF:
603 struct ip_mreqn mreq;
604 struct net_device *dev = NULL;
606 if (sk->sk_type == SOCK_STREAM)
607 goto e_inval;
609 * Check the arguments are allowable
612 if (optlen < sizeof(struct in_addr))
613 goto e_inval;
615 err = -EFAULT;
616 if (optlen >= sizeof(struct ip_mreqn)) {
617 if (copy_from_user(&mreq, optval, sizeof(mreq)))
618 break;
619 } else {
620 memset(&mreq, 0, sizeof(mreq));
621 if (optlen >= sizeof(struct in_addr) &&
622 copy_from_user(&mreq.imr_address, optval,
623 sizeof(struct in_addr)))
624 break;
627 if (!mreq.imr_ifindex) {
628 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
629 inet->mc_index = 0;
630 inet->mc_addr = 0;
631 err = 0;
632 break;
634 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
635 if (dev)
636 mreq.imr_ifindex = dev->ifindex;
637 } else
638 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
641 err = -EADDRNOTAVAIL;
642 if (!dev)
643 break;
644 dev_put(dev);
646 err = -EINVAL;
647 if (sk->sk_bound_dev_if &&
648 mreq.imr_ifindex != sk->sk_bound_dev_if)
649 break;
651 inet->mc_index = mreq.imr_ifindex;
652 inet->mc_addr = mreq.imr_address.s_addr;
653 err = 0;
654 break;
657 case IP_ADD_MEMBERSHIP:
658 case IP_DROP_MEMBERSHIP:
660 struct ip_mreqn mreq;
662 err = -EPROTO;
663 if (inet_sk(sk)->is_icsk)
664 break;
666 if (optlen < sizeof(struct ip_mreq))
667 goto e_inval;
668 err = -EFAULT;
669 if (optlen >= sizeof(struct ip_mreqn)) {
670 if (copy_from_user(&mreq, optval, sizeof(mreq)))
671 break;
672 } else {
673 memset(&mreq, 0, sizeof(mreq));
674 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
675 break;
678 if (optname == IP_ADD_MEMBERSHIP)
679 err = ip_mc_join_group(sk, &mreq);
680 else
681 err = ip_mc_leave_group(sk, &mreq);
682 break;
684 case IP_MSFILTER:
686 struct ip_msfilter *msf;
688 if (optlen < IP_MSFILTER_SIZE(0))
689 goto e_inval;
690 if (optlen > sysctl_optmem_max) {
691 err = -ENOBUFS;
692 break;
694 msf = kmalloc(optlen, GFP_KERNEL);
695 if (!msf) {
696 err = -ENOBUFS;
697 break;
699 err = -EFAULT;
700 if (copy_from_user(msf, optval, optlen)) {
701 kfree(msf);
702 break;
704 /* numsrc >= (1G-4) overflow in 32 bits */
705 if (msf->imsf_numsrc >= 0x3ffffffcU ||
706 msf->imsf_numsrc > sysctl_igmp_max_msf) {
707 kfree(msf);
708 err = -ENOBUFS;
709 break;
711 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
712 kfree(msf);
713 err = -EINVAL;
714 break;
716 err = ip_mc_msfilter(sk, msf, 0);
717 kfree(msf);
718 break;
720 case IP_BLOCK_SOURCE:
721 case IP_UNBLOCK_SOURCE:
722 case IP_ADD_SOURCE_MEMBERSHIP:
723 case IP_DROP_SOURCE_MEMBERSHIP:
725 struct ip_mreq_source mreqs;
726 int omode, add;
728 if (optlen != sizeof(struct ip_mreq_source))
729 goto e_inval;
730 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
731 err = -EFAULT;
732 break;
734 if (optname == IP_BLOCK_SOURCE) {
735 omode = MCAST_EXCLUDE;
736 add = 1;
737 } else if (optname == IP_UNBLOCK_SOURCE) {
738 omode = MCAST_EXCLUDE;
739 add = 0;
740 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
741 struct ip_mreqn mreq;
743 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
744 mreq.imr_address.s_addr = mreqs.imr_interface;
745 mreq.imr_ifindex = 0;
746 err = ip_mc_join_group(sk, &mreq);
747 if (err && err != -EADDRINUSE)
748 break;
749 omode = MCAST_INCLUDE;
750 add = 1;
751 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
752 omode = MCAST_INCLUDE;
753 add = 0;
755 err = ip_mc_source(add, omode, sk, &mreqs, 0);
756 break;
758 case MCAST_JOIN_GROUP:
759 case MCAST_LEAVE_GROUP:
761 struct group_req greq;
762 struct sockaddr_in *psin;
763 struct ip_mreqn mreq;
765 if (optlen < sizeof(struct group_req))
766 goto e_inval;
767 err = -EFAULT;
768 if (copy_from_user(&greq, optval, sizeof(greq)))
769 break;
770 psin = (struct sockaddr_in *)&greq.gr_group;
771 if (psin->sin_family != AF_INET)
772 goto e_inval;
773 memset(&mreq, 0, sizeof(mreq));
774 mreq.imr_multiaddr = psin->sin_addr;
775 mreq.imr_ifindex = greq.gr_interface;
777 if (optname == MCAST_JOIN_GROUP)
778 err = ip_mc_join_group(sk, &mreq);
779 else
780 err = ip_mc_leave_group(sk, &mreq);
781 break;
783 case MCAST_JOIN_SOURCE_GROUP:
784 case MCAST_LEAVE_SOURCE_GROUP:
785 case MCAST_BLOCK_SOURCE:
786 case MCAST_UNBLOCK_SOURCE:
788 struct group_source_req greqs;
789 struct ip_mreq_source mreqs;
790 struct sockaddr_in *psin;
791 int omode, add;
793 if (optlen != sizeof(struct group_source_req))
794 goto e_inval;
795 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
796 err = -EFAULT;
797 break;
799 if (greqs.gsr_group.ss_family != AF_INET ||
800 greqs.gsr_source.ss_family != AF_INET) {
801 err = -EADDRNOTAVAIL;
802 break;
804 psin = (struct sockaddr_in *)&greqs.gsr_group;
805 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
806 psin = (struct sockaddr_in *)&greqs.gsr_source;
807 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
808 mreqs.imr_interface = 0; /* use index for mc_source */
810 if (optname == MCAST_BLOCK_SOURCE) {
811 omode = MCAST_EXCLUDE;
812 add = 1;
813 } else if (optname == MCAST_UNBLOCK_SOURCE) {
814 omode = MCAST_EXCLUDE;
815 add = 0;
816 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
817 struct ip_mreqn mreq;
819 psin = (struct sockaddr_in *)&greqs.gsr_group;
820 mreq.imr_multiaddr = psin->sin_addr;
821 mreq.imr_address.s_addr = 0;
822 mreq.imr_ifindex = greqs.gsr_interface;
823 err = ip_mc_join_group(sk, &mreq);
824 if (err && err != -EADDRINUSE)
825 break;
826 greqs.gsr_interface = mreq.imr_ifindex;
827 omode = MCAST_INCLUDE;
828 add = 1;
829 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
830 omode = MCAST_INCLUDE;
831 add = 0;
833 err = ip_mc_source(add, omode, sk, &mreqs,
834 greqs.gsr_interface);
835 break;
837 case MCAST_MSFILTER:
839 struct sockaddr_in *psin;
840 struct ip_msfilter *msf = NULL;
841 struct group_filter *gsf = NULL;
842 int msize, i, ifindex;
844 if (optlen < GROUP_FILTER_SIZE(0))
845 goto e_inval;
846 if (optlen > sysctl_optmem_max) {
847 err = -ENOBUFS;
848 break;
850 gsf = kmalloc(optlen, GFP_KERNEL);
851 if (!gsf) {
852 err = -ENOBUFS;
853 break;
855 err = -EFAULT;
856 if (copy_from_user(gsf, optval, optlen))
857 goto mc_msf_out;
859 /* numsrc >= (4G-140)/128 overflow in 32 bits */
860 if (gsf->gf_numsrc >= 0x1ffffff ||
861 gsf->gf_numsrc > sysctl_igmp_max_msf) {
862 err = -ENOBUFS;
863 goto mc_msf_out;
865 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
866 err = -EINVAL;
867 goto mc_msf_out;
869 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
870 msf = kmalloc(msize, GFP_KERNEL);
871 if (!msf) {
872 err = -ENOBUFS;
873 goto mc_msf_out;
875 ifindex = gsf->gf_interface;
876 psin = (struct sockaddr_in *)&gsf->gf_group;
877 if (psin->sin_family != AF_INET) {
878 err = -EADDRNOTAVAIL;
879 goto mc_msf_out;
881 msf->imsf_multiaddr = psin->sin_addr.s_addr;
882 msf->imsf_interface = 0;
883 msf->imsf_fmode = gsf->gf_fmode;
884 msf->imsf_numsrc = gsf->gf_numsrc;
885 err = -EADDRNOTAVAIL;
886 for (i = 0; i < gsf->gf_numsrc; ++i) {
887 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
889 if (psin->sin_family != AF_INET)
890 goto mc_msf_out;
891 msf->imsf_slist[i] = psin->sin_addr.s_addr;
893 kfree(gsf);
894 gsf = NULL;
896 err = ip_mc_msfilter(sk, msf, ifindex);
897 mc_msf_out:
898 kfree(msf);
899 kfree(gsf);
900 break;
902 case IP_MULTICAST_ALL:
903 if (optlen < 1)
904 goto e_inval;
905 if (val != 0 && val != 1)
906 goto e_inval;
907 inet->mc_all = val;
908 break;
909 case IP_ROUTER_ALERT:
910 err = ip_ra_control(sk, val ? 1 : 0, NULL);
911 break;
913 case IP_FREEBIND:
914 if (optlen < 1)
915 goto e_inval;
916 inet->freebind = !!val;
917 break;
919 case IP_IPSEC_POLICY:
920 case IP_XFRM_POLICY:
921 err = -EPERM;
922 if (!capable(CAP_NET_ADMIN))
923 break;
924 err = xfrm_user_policy(sk, optname, optval, optlen);
925 break;
927 case IP_TRANSPARENT:
928 if (!capable(CAP_NET_ADMIN)) {
929 err = -EPERM;
930 break;
932 if (optlen < 1)
933 goto e_inval;
934 inet->transparent = !!val;
935 break;
937 case IP_MINTTL:
938 if (optlen < 1)
939 goto e_inval;
940 if (val < 0 || val > 255)
941 goto e_inval;
942 inet->min_ttl = val;
943 break;
945 default:
946 err = -ENOPROTOOPT;
947 break;
949 release_sock(sk);
950 return err;
952 e_inval:
953 release_sock(sk);
954 return -EINVAL;
958 * ip_queue_rcv_skb - Queue an skb into sock receive queue
959 * @sk: socket
960 * @skb: buffer
962 * Queues an skb into socket receive queue. If IP_CMSG_PKTINFO option
963 * is not set, we drop skb dst entry now, while dst cache line is hot.
965 int ip_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
967 if (!(inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO))
968 skb_dst_drop(skb);
969 return sock_queue_rcv_skb(sk, skb);
971 EXPORT_SYMBOL(ip_queue_rcv_skb);
973 int ip_setsockopt(struct sock *sk, int level,
974 int optname, char __user *optval, unsigned int optlen)
976 int err;
978 if (level != SOL_IP)
979 return -ENOPROTOOPT;
981 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
982 #ifdef CONFIG_NETFILTER
983 /* we need to exclude all possible ENOPROTOOPTs except default case */
984 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
985 optname != IP_IPSEC_POLICY &&
986 optname != IP_XFRM_POLICY &&
987 !ip_mroute_opt(optname)) {
988 lock_sock(sk);
989 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
990 release_sock(sk);
992 #endif
993 return err;
995 EXPORT_SYMBOL(ip_setsockopt);
997 #ifdef CONFIG_COMPAT
998 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
999 char __user *optval, unsigned int optlen)
1001 int err;
1003 if (level != SOL_IP)
1004 return -ENOPROTOOPT;
1006 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1007 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1008 ip_setsockopt);
1010 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1011 #ifdef CONFIG_NETFILTER
1012 /* we need to exclude all possible ENOPROTOOPTs except default case */
1013 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1014 optname != IP_IPSEC_POLICY &&
1015 optname != IP_XFRM_POLICY &&
1016 !ip_mroute_opt(optname)) {
1017 lock_sock(sk);
1018 err = compat_nf_setsockopt(sk, PF_INET, optname,
1019 optval, optlen);
1020 release_sock(sk);
1022 #endif
1023 return err;
1025 EXPORT_SYMBOL(compat_ip_setsockopt);
1026 #endif
1029 * Get the options. Note for future reference. The GET of IP options gets
1030 * the _received_ ones. The set sets the _sent_ ones.
1033 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1034 char __user *optval, int __user *optlen)
1036 struct inet_sock *inet = inet_sk(sk);
1037 int val;
1038 int len;
1040 if (level != SOL_IP)
1041 return -EOPNOTSUPP;
1043 if (ip_mroute_opt(optname))
1044 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1046 if (get_user(len, optlen))
1047 return -EFAULT;
1048 if (len < 0)
1049 return -EINVAL;
1051 lock_sock(sk);
1053 switch (optname) {
1054 case IP_OPTIONS:
1056 unsigned char optbuf[sizeof(struct ip_options)+40];
1057 struct ip_options * opt = (struct ip_options *)optbuf;
1058 opt->optlen = 0;
1059 if (inet->opt)
1060 memcpy(optbuf, inet->opt,
1061 sizeof(struct ip_options)+
1062 inet->opt->optlen);
1063 release_sock(sk);
1065 if (opt->optlen == 0)
1066 return put_user(0, optlen);
1068 ip_options_undo(opt);
1070 len = min_t(unsigned int, len, opt->optlen);
1071 if (put_user(len, optlen))
1072 return -EFAULT;
1073 if (copy_to_user(optval, opt->__data, len))
1074 return -EFAULT;
1075 return 0;
1077 case IP_PKTINFO:
1078 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1079 break;
1080 case IP_RECVTTL:
1081 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1082 break;
1083 case IP_RECVTOS:
1084 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1085 break;
1086 case IP_RECVOPTS:
1087 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1088 break;
1089 case IP_RETOPTS:
1090 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1091 break;
1092 case IP_PASSSEC:
1093 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1094 break;
1095 case IP_RECVORIGDSTADDR:
1096 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1097 break;
1098 case IP_TOS:
1099 val = inet->tos;
1100 break;
1101 case IP_TTL:
1102 val = (inet->uc_ttl == -1 ?
1103 sysctl_ip_default_ttl :
1104 inet->uc_ttl);
1105 break;
1106 case IP_HDRINCL:
1107 val = inet->hdrincl;
1108 break;
1109 case IP_MTU_DISCOVER:
1110 val = inet->pmtudisc;
1111 break;
1112 case IP_MTU:
1114 struct dst_entry *dst;
1115 val = 0;
1116 dst = sk_dst_get(sk);
1117 if (dst) {
1118 val = dst_mtu(dst);
1119 dst_release(dst);
1121 if (!val) {
1122 release_sock(sk);
1123 return -ENOTCONN;
1125 break;
1127 case IP_RECVERR:
1128 val = inet->recverr;
1129 break;
1130 case IP_MULTICAST_TTL:
1131 val = inet->mc_ttl;
1132 break;
1133 case IP_MULTICAST_LOOP:
1134 val = inet->mc_loop;
1135 break;
1136 case IP_MULTICAST_IF:
1138 struct in_addr addr;
1139 len = min_t(unsigned int, len, sizeof(struct in_addr));
1140 addr.s_addr = inet->mc_addr;
1141 release_sock(sk);
1143 if (put_user(len, optlen))
1144 return -EFAULT;
1145 if (copy_to_user(optval, &addr, len))
1146 return -EFAULT;
1147 return 0;
1149 case IP_MSFILTER:
1151 struct ip_msfilter msf;
1152 int err;
1154 if (len < IP_MSFILTER_SIZE(0)) {
1155 release_sock(sk);
1156 return -EINVAL;
1158 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1159 release_sock(sk);
1160 return -EFAULT;
1162 err = ip_mc_msfget(sk, &msf,
1163 (struct ip_msfilter __user *)optval, optlen);
1164 release_sock(sk);
1165 return err;
1167 case MCAST_MSFILTER:
1169 struct group_filter gsf;
1170 int err;
1172 if (len < GROUP_FILTER_SIZE(0)) {
1173 release_sock(sk);
1174 return -EINVAL;
1176 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1177 release_sock(sk);
1178 return -EFAULT;
1180 err = ip_mc_gsfget(sk, &gsf,
1181 (struct group_filter __user *)optval,
1182 optlen);
1183 release_sock(sk);
1184 return err;
1186 case IP_MULTICAST_ALL:
1187 val = inet->mc_all;
1188 break;
1189 case IP_PKTOPTIONS:
1191 struct msghdr msg;
1193 release_sock(sk);
1195 if (sk->sk_type != SOCK_STREAM)
1196 return -ENOPROTOOPT;
1198 msg.msg_control = optval;
1199 msg.msg_controllen = len;
1200 msg.msg_flags = 0;
1202 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1203 struct in_pktinfo info;
1205 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1206 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1207 info.ipi_ifindex = inet->mc_index;
1208 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1210 if (inet->cmsg_flags & IP_CMSG_TTL) {
1211 int hlim = inet->mc_ttl;
1212 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1214 len -= msg.msg_controllen;
1215 return put_user(len, optlen);
1217 case IP_FREEBIND:
1218 val = inet->freebind;
1219 break;
1220 case IP_TRANSPARENT:
1221 val = inet->transparent;
1222 break;
1223 case IP_MINTTL:
1224 val = inet->min_ttl;
1225 break;
1226 default:
1227 release_sock(sk);
1228 return -ENOPROTOOPT;
1230 release_sock(sk);
1232 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1233 unsigned char ucval = (unsigned char)val;
1234 len = 1;
1235 if (put_user(len, optlen))
1236 return -EFAULT;
1237 if (copy_to_user(optval, &ucval, 1))
1238 return -EFAULT;
1239 } else {
1240 len = min_t(unsigned int, sizeof(int), len);
1241 if (put_user(len, optlen))
1242 return -EFAULT;
1243 if (copy_to_user(optval, &val, len))
1244 return -EFAULT;
1246 return 0;
1249 int ip_getsockopt(struct sock *sk, int level,
1250 int optname, char __user *optval, int __user *optlen)
1252 int err;
1254 err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1255 #ifdef CONFIG_NETFILTER
1256 /* we need to exclude all possible ENOPROTOOPTs except default case */
1257 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1258 !ip_mroute_opt(optname)) {
1259 int len;
1261 if (get_user(len, optlen))
1262 return -EFAULT;
1264 lock_sock(sk);
1265 err = nf_getsockopt(sk, PF_INET, optname, optval,
1266 &len);
1267 release_sock(sk);
1268 if (err >= 0)
1269 err = put_user(len, optlen);
1270 return err;
1272 #endif
1273 return err;
1275 EXPORT_SYMBOL(ip_getsockopt);
1277 #ifdef CONFIG_COMPAT
1278 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1279 char __user *optval, int __user *optlen)
1281 int err;
1283 if (optname == MCAST_MSFILTER)
1284 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1285 ip_getsockopt);
1287 err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1289 #ifdef CONFIG_NETFILTER
1290 /* we need to exclude all possible ENOPROTOOPTs except default case */
1291 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1292 !ip_mroute_opt(optname)) {
1293 int len;
1295 if (get_user(len, optlen))
1296 return -EFAULT;
1298 lock_sock(sk);
1299 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1300 release_sock(sk);
1301 if (err >= 0)
1302 err = put_user(len, optlen);
1303 return err;
1305 #endif
1306 return err;
1308 EXPORT_SYMBOL(compat_ip_getsockopt);
1309 #endif