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 * PF_INET protocol family socket handler.
8 * Version: $Id: af_inet.c,v 1.137 2002/02/01 22:01:03 davem Exp $
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
15 * Changes (see also sock.c)
18 * Karl Knutson : Socket protocol table
19 * A.N.Kuznetsov : Socket death error in accept().
20 * John Richardson : Fix non blocking error in connect()
21 * so sockets that fail to connect
22 * don't return -EINPROGRESS.
23 * Alan Cox : Asynchronous I/O support
24 * Alan Cox : Keep correct socket pointer on sock
27 * Alan Cox : Semantics of SO_LINGER aren't state
28 * moved to close when you look carefully.
29 * With this fixed and the accept bug fixed
30 * some RPC stuff seems happier.
31 * Niibe Yutaka : 4.4BSD style write async I/O
33 * Tony Gale : Fixed reuse semantics.
34 * Alan Cox : bind() shouldn't abort existing but dead
35 * sockets. Stops FTP netin:.. I hope.
36 * Alan Cox : bind() works correctly for RAW sockets.
37 * Note that FreeBSD at least was broken
38 * in this respect so be careful with
39 * compatibility tests...
40 * Alan Cox : routing cache support
41 * Alan Cox : memzero the socket structure for
43 * Matt Day : nonblock connect error handler
44 * Alan Cox : Allow large numbers of pending sockets
45 * (eg for big web sites), but only if
46 * specifically application requested.
47 * Alan Cox : New buffering throughout IP. Used
49 * Alan Cox : New buffering now used smartly.
50 * Alan Cox : BSD rather than common sense
51 * interpretation of listen.
52 * Germano Caronni : Assorted small races.
53 * Alan Cox : sendmsg/recvmsg basic support.
54 * Alan Cox : Only sendmsg/recvmsg now supported.
55 * Alan Cox : Locked down bind (see security list).
56 * Alan Cox : Loosened bind a little.
57 * Mike McLagan : ADD/DEL DLCI Ioctls
58 * Willy Konynenberg : Transparent proxying support.
59 * David S. Miller : New socket lookup architecture.
60 * Some other random speedups.
61 * Cyrus Durgin : Cleaned up file for kmod hacks.
62 * Andi Kleen : Fix inet_stream_connect TCP race.
64 * This program is free software; you can redistribute it and/or
65 * modify it under the terms of the GNU General Public License
66 * as published by the Free Software Foundation; either version
67 * 2 of the License, or (at your option) any later version.
70 #include <linux/config.h>
71 #include <linux/errno.h>
72 #include <linux/types.h>
73 #include <linux/socket.h>
75 #include <linux/kernel.h>
76 #include <linux/module.h>
77 #include <linux/sched.h>
78 #include <linux/timer.h>
79 #include <linux/string.h>
80 #include <linux/sockios.h>
81 #include <linux/net.h>
82 #include <linux/fcntl.h>
84 #include <linux/interrupt.h>
85 #include <linux/stat.h>
86 #include <linux/init.h>
87 #include <linux/poll.h>
88 #include <linux/netfilter_ipv4.h>
90 #include <asm/uaccess.h>
91 #include <asm/system.h>
93 #include <linux/smp_lock.h>
94 #include <linux/inet.h>
95 #include <linux/igmp.h>
96 #include <linux/netdevice.h>
98 #include <net/protocol.h>
100 #include <net/route.h>
101 #include <net/ip_fib.h>
104 #include <linux/skbuff.h>
105 #include <net/sock.h>
107 #include <net/icmp.h>
108 #include <net/ipip.h>
109 #include <net/inet_common.h>
110 #include <net/xfrm.h>
111 #ifdef CONFIG_IP_MROUTE
112 #include <linux/mroute.h>
115 DEFINE_SNMP_STAT(struct linux_mib, net_statistics);
117 #ifdef INET_REFCNT_DEBUG
118 atomic_t inet_sock_nr;
121 extern void ip_mc_drop_socket(struct sock *sk);
123 /* The inetsw table contains everything that inet_create needs to
124 * build a new socket.
126 static struct list_head inetsw[SOCK_MAX];
127 static DEFINE_SPINLOCK(inetsw_lock);
129 /* New destruction routine */
131 void inet_sock_destruct(struct sock *sk)
133 struct inet_sock *inet = inet_sk(sk);
135 __skb_queue_purge(&sk->sk_receive_queue);
136 __skb_queue_purge(&sk->sk_error_queue);
138 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
139 printk("Attempt to release TCP socket in state %d %p\n",
143 if (!sock_flag(sk, SOCK_DEAD)) {
144 printk("Attempt to release alive inet socket %p\n", sk);
148 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
149 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
150 BUG_TRAP(!sk->sk_wmem_queued);
151 BUG_TRAP(!sk->sk_forward_alloc);
155 dst_release(sk->sk_dst_cache);
156 #ifdef INET_REFCNT_DEBUG
157 atomic_dec(&inet_sock_nr);
158 printk(KERN_DEBUG "INET socket %p released, %d are still alive\n",
159 sk, atomic_read(&inet_sock_nr));
164 * The routines beyond this point handle the behaviour of an AF_INET
165 * socket object. Mostly it punts to the subprotocols of IP to do
170 * Automatically bind an unbound socket.
173 static int inet_autobind(struct sock *sk)
175 struct inet_sock *inet;
176 /* We may need to bind the socket. */
180 if (sk->sk_prot->get_port(sk, 0)) {
184 inet->sport = htons(inet->num);
191 * Move a socket into listening state.
193 int inet_listen(struct socket *sock, int backlog)
195 struct sock *sk = sock->sk;
196 unsigned char old_state;
202 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
205 old_state = sk->sk_state;
206 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
209 /* Really, if the socket is already in listen state
210 * we can only allow the backlog to be adjusted.
212 if (old_state != TCP_LISTEN) {
213 err = tcp_listen_start(sk);
217 sk->sk_max_ack_backlog = backlog;
226 * Create an inet socket.
229 static int inet_create(struct socket *sock, int protocol)
233 struct inet_protosw *answer;
234 struct inet_sock *inet;
235 struct proto *answer_prot;
236 unsigned char answer_flags;
237 char answer_no_check;
240 sock->state = SS_UNCONNECTED;
242 /* Look for the requested type/protocol pair. */
245 list_for_each_rcu(p, &inetsw[sock->type]) {
246 answer = list_entry(p, struct inet_protosw, list);
248 /* Check the non-wild match. */
249 if (protocol == answer->protocol) {
250 if (protocol != IPPROTO_IP)
253 /* Check for the two wild cases. */
254 if (IPPROTO_IP == protocol) {
255 protocol = answer->protocol;
258 if (IPPROTO_IP == answer->protocol)
264 err = -ESOCKTNOSUPPORT;
268 if (answer->capability > 0 && !capable(answer->capability))
270 err = -EPROTONOSUPPORT;
274 sock->ops = answer->ops;
275 answer_prot = answer->prot;
276 answer_no_check = answer->no_check;
277 answer_flags = answer->flags;
280 BUG_TRAP(answer_prot->slab != NULL);
283 sk = sk_alloc(PF_INET, GFP_KERNEL, answer_prot, 1);
288 sk->sk_no_check = answer_no_check;
289 if (INET_PROTOSW_REUSE & answer_flags)
294 if (SOCK_RAW == sock->type) {
295 inet->num = protocol;
296 if (IPPROTO_RAW == protocol)
300 if (ipv4_config.no_pmtu_disc)
301 inet->pmtudisc = IP_PMTUDISC_DONT;
303 inet->pmtudisc = IP_PMTUDISC_WANT;
307 sock_init_data(sock, sk);
309 sk->sk_destruct = inet_sock_destruct;
310 sk->sk_family = PF_INET;
311 sk->sk_protocol = protocol;
312 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
318 inet->mc_list = NULL;
320 #ifdef INET_REFCNT_DEBUG
321 atomic_inc(&inet_sock_nr);
325 /* It assumes that any protocol which allows
326 * the user to assign a number at socket
327 * creation time automatically
330 inet->sport = htons(inet->num);
331 /* Add to protocol hash chains. */
332 sk->sk_prot->hash(sk);
335 if (sk->sk_prot->init) {
336 err = sk->sk_prot->init(sk);
338 sk_common_release(sk);
349 * The peer socket should always be NULL (or else). When we call this
350 * function we are destroying the object and from then on nobody
351 * should refer to it.
353 int inet_release(struct socket *sock)
355 struct sock *sk = sock->sk;
360 /* Applications forget to leave groups before exiting */
361 ip_mc_drop_socket(sk);
363 /* If linger is set, we don't return until the close
364 * is complete. Otherwise we return immediately. The
365 * actually closing is done the same either way.
367 * If the close is due to the process exiting, we never
371 if (sock_flag(sk, SOCK_LINGER) &&
372 !(current->flags & PF_EXITING))
373 timeout = sk->sk_lingertime;
375 sk->sk_prot->close(sk, timeout);
380 /* It is off by default, see below. */
381 int sysctl_ip_nonlocal_bind;
383 int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
385 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
386 struct sock *sk = sock->sk;
387 struct inet_sock *inet = inet_sk(sk);
392 /* If the socket has its own bind function then use it. (RAW) */
393 if (sk->sk_prot->bind) {
394 err = sk->sk_prot->bind(sk, uaddr, addr_len);
398 if (addr_len < sizeof(struct sockaddr_in))
401 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
403 /* Not specified by any standard per-se, however it breaks too
404 * many applications when removed. It is unfortunate since
405 * allowing applications to make a non-local bind solves
406 * several problems with systems using dynamic addressing.
407 * (ie. your servers still start up even if your ISDN link
408 * is temporarily down)
410 err = -EADDRNOTAVAIL;
411 if (!sysctl_ip_nonlocal_bind &&
413 addr->sin_addr.s_addr != INADDR_ANY &&
414 chk_addr_ret != RTN_LOCAL &&
415 chk_addr_ret != RTN_MULTICAST &&
416 chk_addr_ret != RTN_BROADCAST)
419 snum = ntohs(addr->sin_port);
421 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
424 /* We keep a pair of addresses. rcv_saddr is the one
425 * used by hash lookups, and saddr is used for transmit.
427 * In the BSD API these are the same except where it
428 * would be illegal to use them (multicast/broadcast) in
429 * which case the sending device address is used.
433 /* Check these errors (active socket, double bind). */
435 if (sk->sk_state != TCP_CLOSE || inet->num)
436 goto out_release_sock;
438 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
439 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
440 inet->saddr = 0; /* Use device */
442 /* Make sure we are allowed to bind here. */
443 if (sk->sk_prot->get_port(sk, snum)) {
444 inet->saddr = inet->rcv_saddr = 0;
446 goto out_release_sock;
450 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
452 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
453 inet->sport = htons(inet->num);
464 int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
465 int addr_len, int flags)
467 struct sock *sk = sock->sk;
469 if (uaddr->sa_family == AF_UNSPEC)
470 return sk->sk_prot->disconnect(sk, flags);
472 if (!inet_sk(sk)->num && inet_autobind(sk))
474 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
477 static long inet_wait_for_connect(struct sock *sk, long timeo)
481 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
483 /* Basic assumption: if someone sets sk->sk_err, he _must_
484 * change state of the socket from TCP_SYN_*.
485 * Connect() does not allow to get error notifications
486 * without closing the socket.
488 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
490 timeo = schedule_timeout(timeo);
492 if (signal_pending(current) || !timeo)
494 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
496 finish_wait(sk->sk_sleep, &wait);
501 * Connect to a remote host. There is regrettably still a little
502 * TCP 'magic' in here.
504 int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
505 int addr_len, int flags)
507 struct sock *sk = sock->sk;
513 if (uaddr->sa_family == AF_UNSPEC) {
514 err = sk->sk_prot->disconnect(sk, flags);
515 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
519 switch (sock->state) {
528 /* Fall out of switch with err, set for this state */
532 if (sk->sk_state != TCP_CLOSE)
535 err = sk->sk_prot->connect(sk, uaddr, addr_len);
539 sock->state = SS_CONNECTING;
541 /* Just entered SS_CONNECTING state; the only
542 * difference is that return value in non-blocking
543 * case is EINPROGRESS, rather than EALREADY.
549 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
551 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
552 /* Error code is set above */
553 if (!timeo || !inet_wait_for_connect(sk, timeo))
556 err = sock_intr_errno(timeo);
557 if (signal_pending(current))
561 /* Connection was closed by RST, timeout, ICMP error
562 * or another process disconnected us.
564 if (sk->sk_state == TCP_CLOSE)
567 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
568 * and error was received after socket entered established state.
569 * Hence, it is handled normally after connect() return successfully.
572 sock->state = SS_CONNECTED;
579 err = sock_error(sk) ? : -ECONNABORTED;
580 sock->state = SS_UNCONNECTED;
581 if (sk->sk_prot->disconnect(sk, flags))
582 sock->state = SS_DISCONNECTING;
587 * Accept a pending connection. The TCP layer now gives BSD semantics.
590 int inet_accept(struct socket *sock, struct socket *newsock, int flags)
592 struct sock *sk1 = sock->sk;
594 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
601 BUG_TRAP((1 << sk2->sk_state) &
602 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE));
604 sock_graft(sk2, newsock);
606 newsock->state = SS_CONNECTED;
615 * This does both peername and sockname.
617 int inet_getname(struct socket *sock, struct sockaddr *uaddr,
618 int *uaddr_len, int peer)
620 struct sock *sk = sock->sk;
621 struct inet_sock *inet = inet_sk(sk);
622 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
624 sin->sin_family = AF_INET;
627 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
630 sin->sin_port = inet->dport;
631 sin->sin_addr.s_addr = inet->daddr;
633 __u32 addr = inet->rcv_saddr;
636 sin->sin_port = inet->sport;
637 sin->sin_addr.s_addr = addr;
639 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
640 *uaddr_len = sizeof(*sin);
644 int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
647 struct sock *sk = sock->sk;
649 /* We may need to bind the socket. */
650 if (!inet_sk(sk)->num && inet_autobind(sk))
653 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
657 static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
659 struct sock *sk = sock->sk;
661 /* We may need to bind the socket. */
662 if (!inet_sk(sk)->num && inet_autobind(sk))
665 if (sk->sk_prot->sendpage)
666 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
667 return sock_no_sendpage(sock, page, offset, size, flags);
671 int inet_shutdown(struct socket *sock, int how)
673 struct sock *sk = sock->sk;
676 /* This should really check to make sure
677 * the socket is a TCP socket. (WHY AC...)
679 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
682 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
686 if (sock->state == SS_CONNECTING) {
687 if ((1 << sk->sk_state) &
688 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
689 sock->state = SS_DISCONNECTING;
691 sock->state = SS_CONNECTED;
694 switch (sk->sk_state) {
697 /* Hack to wake up other listeners, who can poll for
698 POLLHUP, even on eg. unconnected UDP sockets -- RR */
700 sk->sk_shutdown |= how;
701 if (sk->sk_prot->shutdown)
702 sk->sk_prot->shutdown(sk, how);
705 /* Remaining two branches are temporary solution for missing
706 * close() in multithreaded environment. It is _not_ a good idea,
707 * but we have no choice until close() is repaired at VFS level.
710 if (!(how & RCV_SHUTDOWN))
714 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
715 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
719 /* Wake up anyone sleeping in poll. */
720 sk->sk_state_change(sk);
726 * ioctl() calls you can issue on an INET socket. Most of these are
727 * device configuration and stuff and very rarely used. Some ioctls
728 * pass on to the socket itself.
730 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
731 * loads the devconfigure module does its configuring and unloads it.
732 * There's a good 20K of config code hanging around the kernel.
735 int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
737 struct sock *sk = sock->sk;
742 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
747 err = ip_rt_ioctl(cmd, (void __user *)arg);
752 err = arp_ioctl(cmd, (void __user *)arg);
765 err = devinet_ioctl(cmd, (void __user *)arg);
768 if (!sk->sk_prot->ioctl ||
769 (err = sk->sk_prot->ioctl(sk, cmd, arg)) ==
771 err = dev_ioctl(cmd, (void __user *)arg);
777 struct proto_ops inet_stream_ops = {
779 .owner = THIS_MODULE,
780 .release = inet_release,
782 .connect = inet_stream_connect,
783 .socketpair = sock_no_socketpair,
784 .accept = inet_accept,
785 .getname = inet_getname,
788 .listen = inet_listen,
789 .shutdown = inet_shutdown,
790 .setsockopt = sock_common_setsockopt,
791 .getsockopt = sock_common_getsockopt,
792 .sendmsg = inet_sendmsg,
793 .recvmsg = sock_common_recvmsg,
794 .mmap = sock_no_mmap,
795 .sendpage = tcp_sendpage
798 struct proto_ops inet_dgram_ops = {
800 .owner = THIS_MODULE,
801 .release = inet_release,
803 .connect = inet_dgram_connect,
804 .socketpair = sock_no_socketpair,
805 .accept = sock_no_accept,
806 .getname = inet_getname,
809 .listen = sock_no_listen,
810 .shutdown = inet_shutdown,
811 .setsockopt = sock_common_setsockopt,
812 .getsockopt = sock_common_getsockopt,
813 .sendmsg = inet_sendmsg,
814 .recvmsg = sock_common_recvmsg,
815 .mmap = sock_no_mmap,
816 .sendpage = inet_sendpage,
820 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
823 static struct proto_ops inet_sockraw_ops = {
825 .owner = THIS_MODULE,
826 .release = inet_release,
828 .connect = inet_dgram_connect,
829 .socketpair = sock_no_socketpair,
830 .accept = sock_no_accept,
831 .getname = inet_getname,
832 .poll = datagram_poll,
834 .listen = sock_no_listen,
835 .shutdown = inet_shutdown,
836 .setsockopt = sock_common_setsockopt,
837 .getsockopt = sock_common_getsockopt,
838 .sendmsg = inet_sendmsg,
839 .recvmsg = sock_common_recvmsg,
840 .mmap = sock_no_mmap,
841 .sendpage = inet_sendpage,
844 static struct net_proto_family inet_family_ops = {
846 .create = inet_create,
847 .owner = THIS_MODULE,
851 extern void tcp_init(void);
852 extern void tcp_v4_init(struct net_proto_family *);
854 /* Upon startup we insert all the elements in inetsw_array[] into
855 * the linked list inetsw.
857 static struct inet_protosw inetsw_array[] =
861 .protocol = IPPROTO_TCP,
863 .ops = &inet_stream_ops,
866 .flags = INET_PROTOSW_PERMANENT,
871 .protocol = IPPROTO_UDP,
873 .ops = &inet_dgram_ops,
875 .no_check = UDP_CSUM_DEFAULT,
876 .flags = INET_PROTOSW_PERMANENT,
882 .protocol = IPPROTO_IP, /* wild card */
884 .ops = &inet_sockraw_ops,
885 .capability = CAP_NET_RAW,
886 .no_check = UDP_CSUM_DEFAULT,
887 .flags = INET_PROTOSW_REUSE,
891 #define INETSW_ARRAY_LEN (sizeof(inetsw_array) / sizeof(struct inet_protosw))
893 void inet_register_protosw(struct inet_protosw *p)
895 struct list_head *lh;
896 struct inet_protosw *answer;
897 int protocol = p->protocol;
898 struct list_head *last_perm;
900 spin_lock_bh(&inetsw_lock);
902 if (p->type >= SOCK_MAX)
905 /* If we are trying to override a permanent protocol, bail. */
907 last_perm = &inetsw[p->type];
908 list_for_each(lh, &inetsw[p->type]) {
909 answer = list_entry(lh, struct inet_protosw, list);
911 /* Check only the non-wild match. */
912 if (INET_PROTOSW_PERMANENT & answer->flags) {
913 if (protocol == answer->protocol)
923 /* Add the new entry after the last permanent entry if any, so that
924 * the new entry does not override a permanent entry when matched with
925 * a wild-card protocol. But it is allowed to override any existing
926 * non-permanent entry. This means that when we remove this entry, the
927 * system automatically returns to the old behavior.
929 list_add_rcu(&p->list, last_perm);
931 spin_unlock_bh(&inetsw_lock);
938 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
944 "Ignoring attempt to register invalid socket type %d.\n",
949 void inet_unregister_protosw(struct inet_protosw *p)
951 if (INET_PROTOSW_PERMANENT & p->flags) {
953 "Attempt to unregister permanent protocol %d.\n",
956 spin_lock_bh(&inetsw_lock);
957 list_del_rcu(&p->list);
958 spin_unlock_bh(&inetsw_lock);
964 #ifdef CONFIG_IP_MULTICAST
965 static struct net_protocol igmp_protocol = {
970 static struct net_protocol tcp_protocol = {
971 .handler = tcp_v4_rcv,
972 .err_handler = tcp_v4_err,
976 static struct net_protocol udp_protocol = {
978 .err_handler = udp_err,
982 static struct net_protocol icmp_protocol = {
986 static int __init init_ipv4_mibs(void)
988 net_statistics[0] = alloc_percpu(struct linux_mib);
989 net_statistics[1] = alloc_percpu(struct linux_mib);
990 ip_statistics[0] = alloc_percpu(struct ipstats_mib);
991 ip_statistics[1] = alloc_percpu(struct ipstats_mib);
992 icmp_statistics[0] = alloc_percpu(struct icmp_mib);
993 icmp_statistics[1] = alloc_percpu(struct icmp_mib);
994 tcp_statistics[0] = alloc_percpu(struct tcp_mib);
995 tcp_statistics[1] = alloc_percpu(struct tcp_mib);
996 udp_statistics[0] = alloc_percpu(struct udp_mib);
997 udp_statistics[1] = alloc_percpu(struct udp_mib);
999 (net_statistics[0] && net_statistics[1] && ip_statistics[0]
1000 && ip_statistics[1] && tcp_statistics[0] && tcp_statistics[1]
1001 && udp_statistics[0] && udp_statistics[1]))
1004 (void) tcp_mib_init();
1009 static int ipv4_proc_init(void);
1010 extern void ipfrag_init(void);
1013 * IP protocol layer initialiser
1016 static struct packet_type ip_packet_type = {
1017 .type = __constant_htons(ETH_P_IP),
1021 static int __init inet_init(void)
1023 struct sk_buff *dummy_skb;
1024 struct inet_protosw *q;
1025 struct list_head *r;
1028 if (sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb)) {
1029 printk(KERN_CRIT "%s: panic\n", __FUNCTION__);
1033 rc = proto_register(&tcp_prot, 1);
1037 rc = proto_register(&udp_prot, 1);
1039 goto out_unregister_tcp_proto;
1041 rc = proto_register(&raw_prot, 1);
1043 goto out_unregister_udp_proto;
1046 * Tell SOCKET that we are alive...
1049 (void)sock_register(&inet_family_ops);
1052 * Add all the base protocols.
1055 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1056 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1057 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1058 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1059 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1060 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1061 #ifdef CONFIG_IP_MULTICAST
1062 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1063 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1066 /* Register the socket-side information for inet_create. */
1067 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1070 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1071 inet_register_protosw(q);
1074 * Set the ARP module up
1080 * Set the IP module up
1085 tcp_v4_init(&inet_family_ops);
1087 /* Setup TCP slab cache for open requests. */
1092 * Set the ICMP layer up
1095 icmp_init(&inet_family_ops);
1098 * Initialise the multicast router
1100 #if defined(CONFIG_IP_MROUTE)
1104 * Initialise per-cpu ipv4 mibs
1107 if(init_ipv4_mibs())
1108 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
1114 dev_add_pack(&ip_packet_type);
1119 out_unregister_tcp_proto:
1120 proto_unregister(&tcp_prot);
1121 out_unregister_udp_proto:
1122 proto_unregister(&udp_prot);
1126 module_init(inet_init);
1128 /* ------------------------------------------------------------------------ */
1130 #ifdef CONFIG_PROC_FS
1131 extern int fib_proc_init(void);
1132 extern void fib_proc_exit(void);
1133 #ifdef CONFIG_IP_FIB_TRIE
1134 extern int fib_stat_proc_init(void);
1135 extern void fib_stat_proc_exit(void);
1137 extern int ip_misc_proc_init(void);
1138 extern int raw_proc_init(void);
1139 extern void raw_proc_exit(void);
1140 extern int tcp4_proc_init(void);
1141 extern void tcp4_proc_exit(void);
1142 extern int udp4_proc_init(void);
1143 extern void udp4_proc_exit(void);
1145 static int __init ipv4_proc_init(void)
1149 if (raw_proc_init())
1151 if (tcp4_proc_init())
1153 if (udp4_proc_init())
1155 if (fib_proc_init())
1157 #ifdef CONFIG_IP_FIB_TRIE
1158 if (fib_stat_proc_init())
1161 if (ip_misc_proc_init())
1166 #ifdef CONFIG_IP_FIB_TRIE
1167 fib_stat_proc_exit();
1182 #else /* CONFIG_PROC_FS */
1183 static int __init ipv4_proc_init(void)
1187 #endif /* CONFIG_PROC_FS */
1189 MODULE_ALIAS_NETPROTO(PF_INET);
1191 EXPORT_SYMBOL(inet_accept);
1192 EXPORT_SYMBOL(inet_bind);
1193 EXPORT_SYMBOL(inet_dgram_connect);
1194 EXPORT_SYMBOL(inet_dgram_ops);
1195 EXPORT_SYMBOL(inet_getname);
1196 EXPORT_SYMBOL(inet_ioctl);
1197 EXPORT_SYMBOL(inet_listen);
1198 EXPORT_SYMBOL(inet_register_protosw);
1199 EXPORT_SYMBOL(inet_release);
1200 EXPORT_SYMBOL(inet_sendmsg);
1201 EXPORT_SYMBOL(inet_shutdown);
1202 EXPORT_SYMBOL(inet_sock_destruct);
1203 EXPORT_SYMBOL(inet_stream_connect);
1204 EXPORT_SYMBOL(inet_stream_ops);
1205 EXPORT_SYMBOL(inet_unregister_protosw);
1206 EXPORT_SYMBOL(net_statistics);
1207 EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
1209 #ifdef INET_REFCNT_DEBUG
1210 EXPORT_SYMBOL(inet_sock_nr);