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 $
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
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/major.h>
77 #include <linux/module.h>
78 #include <linux/sched.h>
79 #include <linux/timer.h>
80 #include <linux/string.h>
81 #include <linux/sockios.h>
82 #include <linux/net.h>
83 #include <linux/fcntl.h>
85 #include <linux/interrupt.h>
86 #include <linux/stat.h>
87 #include <linux/init.h>
88 #include <linux/poll.h>
89 #include <linux/netfilter_ipv4.h>
91 #include <asm/uaccess.h>
92 #include <asm/system.h>
94 #include <linux/smp_lock.h>
95 #include <linux/inet.h>
96 #include <linux/igmp.h>
97 #include <linux/netdevice.h>
99 #include <net/protocol.h>
101 #include <net/route.h>
102 #include <net/ip_fib.h>
105 #include <linux/skbuff.h>
106 #include <net/sock.h>
108 #include <net/icmp.h>
109 #include <net/ipip.h>
110 #include <net/inet_common.h>
111 #include <net/xfrm.h>
112 #ifdef CONFIG_IP_MROUTE
113 #include <linux/mroute.h>
115 #include <linux/vs_limit.h>
117 DEFINE_SNMP_STAT(struct linux_mib, net_statistics);
119 #ifdef INET_REFCNT_DEBUG
120 atomic_t inet_sock_nr;
123 extern void ip_mc_drop_socket(struct sock *sk);
125 /* Per protocol sock slabcache */
126 kmem_cache_t *tcp_sk_cachep;
127 static kmem_cache_t *udp_sk_cachep;
128 static kmem_cache_t *raw4_sk_cachep;
130 /* The inetsw table contains everything that inet_create needs to
131 * build a new socket.
133 static struct list_head inetsw[SOCK_MAX];
134 static spinlock_t inetsw_lock = SPIN_LOCK_UNLOCKED;
136 /* New destruction routine */
138 void inet_sock_destruct(struct sock *sk)
140 struct inet_opt *inet = inet_sk(sk);
142 __skb_queue_purge(&sk->sk_receive_queue);
143 __skb_queue_purge(&sk->sk_error_queue);
145 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
146 printk("Attempt to release TCP socket in state %d %p\n",
150 if (!sock_flag(sk, SOCK_DEAD)) {
151 printk("Attempt to release alive inet socket %p\n", sk);
155 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
156 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
157 BUG_TRAP(!sk->sk_wmem_queued);
158 BUG_TRAP(!sk->sk_forward_alloc);
164 clr_vx_info(&sk->sk_vx_info);
166 clr_nx_info(&sk->sk_nx_info);
169 dst_release(sk->sk_dst_cache);
170 #ifdef INET_REFCNT_DEBUG
171 atomic_dec(&inet_sock_nr);
172 printk(KERN_DEBUG "INET socket %p released, %d are still alive\n",
173 sk, atomic_read(&inet_sock_nr));
178 * The routines beyond this point handle the behaviour of an AF_INET
179 * socket object. Mostly it punts to the subprotocols of IP to do
184 * Automatically bind an unbound socket.
187 static int inet_autobind(struct sock *sk)
189 struct inet_opt *inet;
190 /* We may need to bind the socket. */
194 if (sk->sk_prot->get_port(sk, 0)) {
198 inet->sport = htons(inet->num);
205 * Move a socket into listening state.
207 int inet_listen(struct socket *sock, int backlog)
209 struct sock *sk = sock->sk;
210 unsigned char old_state;
216 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
219 old_state = sk->sk_state;
220 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
223 /* Really, if the socket is already in listen state
224 * we can only allow the backlog to be adjusted.
226 if (old_state != TCP_LISTEN) {
227 err = tcp_listen_start(sk);
231 sk->sk_max_ack_backlog = backlog;
239 static __inline__ kmem_cache_t *inet_sk_slab(int protocol)
241 kmem_cache_t* rc = tcp_sk_cachep;
243 if (protocol == IPPROTO_UDP)
245 else if (protocol == IPPROTO_RAW)
250 static __inline__ int inet_sk_size(int protocol)
252 int rc = sizeof(struct tcp_sock);
254 if (protocol == IPPROTO_UDP)
255 rc = sizeof(struct udp_sock);
256 else if (protocol == IPPROTO_RAW)
257 rc = sizeof(struct raw_sock);
262 * Create an inet socket.
265 static int inet_create(struct socket *sock, int protocol)
269 struct inet_protosw *answer;
270 struct inet_opt *inet;
273 sock->state = SS_UNCONNECTED;
274 sk = sk_alloc(PF_INET, GFP_KERNEL, inet_sk_size(protocol),
275 inet_sk_slab(protocol));
279 /* Look for the requested type/protocol pair. */
282 list_for_each_rcu(p, &inetsw[sock->type]) {
283 answer = list_entry(p, struct inet_protosw, list);
285 /* Check the non-wild match. */
286 if (protocol == answer->protocol) {
287 if (protocol != IPPROTO_IP)
290 /* Check for the two wild cases. */
291 if (IPPROTO_IP == protocol) {
292 protocol = answer->protocol;
295 if (IPPROTO_IP == answer->protocol)
301 err = -ESOCKTNOSUPPORT;
305 if ((protocol == IPPROTO_ICMP) && vx_ccaps(VXC_RAW_ICMP))
307 if (answer->capability > 0 && !capable(answer->capability))
310 err = -EPROTONOSUPPORT;
314 sock->ops = answer->ops;
315 sk->sk_prot = answer->prot;
316 sk->sk_no_check = answer->no_check;
317 if (INET_PROTOSW_REUSE & answer->flags)
323 if (SOCK_RAW == sock->type) {
324 inet->num = protocol;
325 if (IPPROTO_RAW == protocol)
329 if (ipv4_config.no_pmtu_disc)
330 inet->pmtudisc = IP_PMTUDISC_DONT;
332 inet->pmtudisc = IP_PMTUDISC_WANT;
336 sock_init_data(sock, sk);
337 sk_set_owner(sk, THIS_MODULE);
339 sk->sk_destruct = inet_sock_destruct;
341 sk->sk_family = PF_INET;
342 sk->sk_protocol = protocol;
343 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
345 set_vx_info(&sk->sk_vx_info, current->vx_info);
346 sk->sk_xid = vx_current_xid();
348 set_nx_info(&sk->sk_nx_info, current->nx_info);
349 sk->sk_nid = nx_current_nid();
355 inet->mc_list = NULL;
357 #ifdef INET_REFCNT_DEBUG
358 atomic_inc(&inet_sock_nr);
362 /* It assumes that any protocol which allows
363 * the user to assign a number at socket
364 * creation time automatically
367 inet->sport = htons(inet->num);
368 /* Add to protocol hash chains. */
369 sk->sk_prot->hash(sk);
372 if (sk->sk_prot->init) {
373 err = sk->sk_prot->init(sk);
375 sk_common_release(sk);
387 * The peer socket should always be NULL (or else). When we call this
388 * function we are destroying the object and from then on nobody
389 * should refer to it.
391 int inet_release(struct socket *sock)
393 struct sock *sk = sock->sk;
398 /* Applications forget to leave groups before exiting */
399 ip_mc_drop_socket(sk);
401 /* If linger is set, we don't return until the close
402 * is complete. Otherwise we return immediately. The
403 * actually closing is done the same either way.
405 * If the close is due to the process exiting, we never
409 if (sock_flag(sk, SOCK_LINGER) &&
410 !(current->flags & PF_EXITING))
411 timeout = sk->sk_lingertime;
414 clr_vx_info(&sk->sk_vx_info);
416 clr_nx_info(&sk->sk_nx_info);
418 sk->sk_prot->close(sk, timeout);
423 /* It is off by default, see below. */
424 int sysctl_ip_nonlocal_bind;
426 int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
428 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
429 struct sock *sk = sock->sk;
430 struct inet_opt *inet = inet_sk(sk);
435 /* If the socket has its own bind function then use it. (RAW) */
436 if (sk->sk_prot->bind) {
437 err = sk->sk_prot->bind(sk, uaddr, addr_len);
441 if (addr_len < sizeof(struct sockaddr_in))
444 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
446 /* Not specified by any standard per-se, however it breaks too
447 * many applications when removed. It is unfortunate since
448 * allowing applications to make a non-local bind solves
449 * several problems with systems using dynamic addressing.
450 * (ie. your servers still start up even if your ISDN link
451 * is temporarily down)
453 err = -EADDRNOTAVAIL;
454 if (!sysctl_ip_nonlocal_bind &&
456 addr->sin_addr.s_addr != INADDR_ANY &&
457 chk_addr_ret != RTN_LOCAL &&
458 chk_addr_ret != RTN_MULTICAST &&
459 chk_addr_ret != RTN_BROADCAST)
462 snum = ntohs(addr->sin_port);
464 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
467 /* We keep a pair of addresses. rcv_saddr is the one
468 * used by hash lookups, and saddr is used for transmit.
470 * In the BSD API these are the same except where it
471 * would be illegal to use them (multicast/broadcast) in
472 * which case the sending device address is used.
476 /* Check these errors (active socket, double bind). */
478 if (sk->sk_state != TCP_CLOSE || inet->num)
479 goto out_release_sock;
481 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
482 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
483 inet->saddr = 0; /* Use device */
485 /* Make sure we are allowed to bind here. */
486 if (sk->sk_prot->get_port(sk, snum)) {
487 inet->saddr = inet->rcv_saddr = 0;
489 goto out_release_sock;
493 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
495 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
496 inet->sport = htons(inet->num);
507 int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
508 int addr_len, int flags)
510 struct sock *sk = sock->sk;
512 if (uaddr->sa_family == AF_UNSPEC)
513 return sk->sk_prot->disconnect(sk, flags);
515 if (!inet_sk(sk)->num && inet_autobind(sk))
517 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
520 static long inet_wait_for_connect(struct sock *sk, long timeo)
524 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
526 /* Basic assumption: if someone sets sk->sk_err, he _must_
527 * change state of the socket from TCP_SYN_*.
528 * Connect() does not allow to get error notifications
529 * without closing the socket.
531 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
533 timeo = schedule_timeout(timeo);
535 if (signal_pending(current) || !timeo)
537 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
539 finish_wait(sk->sk_sleep, &wait);
544 * Connect to a remote host. There is regrettably still a little
545 * TCP 'magic' in here.
547 int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
548 int addr_len, int flags)
550 struct sock *sk = sock->sk;
556 if (uaddr->sa_family == AF_UNSPEC) {
557 err = sk->sk_prot->disconnect(sk, flags);
558 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
562 switch (sock->state) {
571 /* Fall out of switch with err, set for this state */
575 if (sk->sk_state != TCP_CLOSE)
578 err = sk->sk_prot->connect(sk, uaddr, addr_len);
582 sock->state = SS_CONNECTING;
584 /* Just entered SS_CONNECTING state; the only
585 * difference is that return value in non-blocking
586 * case is EINPROGRESS, rather than EALREADY.
592 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
594 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
595 /* Error code is set above */
596 if (!timeo || !inet_wait_for_connect(sk, timeo))
599 err = sock_intr_errno(timeo);
600 if (signal_pending(current))
604 /* Connection was closed by RST, timeout, ICMP error
605 * or another process disconnected us.
607 if (sk->sk_state == TCP_CLOSE)
610 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
611 * and error was received after socket entered established state.
612 * Hence, it is handled normally after connect() return successfully.
615 sock->state = SS_CONNECTED;
622 err = sock_error(sk) ? : -ECONNABORTED;
623 sock->state = SS_UNCONNECTED;
624 if (sk->sk_prot->disconnect(sk, flags))
625 sock->state = SS_DISCONNECTING;
630 * Accept a pending connection. The TCP layer now gives BSD semantics.
633 int inet_accept(struct socket *sock, struct socket *newsock, int flags)
635 struct sock *sk1 = sock->sk;
637 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
644 BUG_TRAP((1 << sk2->sk_state) &
645 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE));
647 sock_graft(sk2, newsock);
649 newsock->state = SS_CONNECTED;
658 * This does both peername and sockname.
660 int inet_getname(struct socket *sock, struct sockaddr *uaddr,
661 int *uaddr_len, int peer)
663 struct sock *sk = sock->sk;
664 struct inet_opt *inet = inet_sk(sk);
665 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
667 sin->sin_family = AF_INET;
670 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
673 sin->sin_port = inet->dport;
674 sin->sin_addr.s_addr = inet->daddr;
676 __u32 addr = inet->rcv_saddr;
679 sin->sin_port = inet->sport;
680 sin->sin_addr.s_addr = addr;
682 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
683 *uaddr_len = sizeof(*sin);
687 int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
690 struct sock *sk = sock->sk;
692 /* We may need to bind the socket. */
693 if (!inet_sk(sk)->num && inet_autobind(sk))
696 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
700 ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
702 struct sock *sk = sock->sk;
704 /* We may need to bind the socket. */
705 if (!inet_sk(sk)->num && inet_autobind(sk))
708 if (sk->sk_prot->sendpage)
709 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
710 return sock_no_sendpage(sock, page, offset, size, flags);
714 int inet_shutdown(struct socket *sock, int how)
716 struct sock *sk = sock->sk;
719 /* This should really check to make sure
720 * the socket is a TCP socket. (WHY AC...)
722 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
725 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
729 if (sock->state == SS_CONNECTING) {
730 if ((1 << sk->sk_state) &
731 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
732 sock->state = SS_DISCONNECTING;
734 sock->state = SS_CONNECTED;
737 switch (sk->sk_state) {
740 /* Hack to wake up other listeners, who can poll for
741 POLLHUP, even on eg. unconnected UDP sockets -- RR */
743 sk->sk_shutdown |= how;
744 if (sk->sk_prot->shutdown)
745 sk->sk_prot->shutdown(sk, how);
748 /* Remaining two branches are temporary solution for missing
749 * close() in multithreaded environment. It is _not_ a good idea,
750 * but we have no choice until close() is repaired at VFS level.
753 if (!(how & RCV_SHUTDOWN))
757 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
758 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
762 /* Wake up anyone sleeping in poll. */
763 sk->sk_state_change(sk);
769 * ioctl() calls you can issue on an INET socket. Most of these are
770 * device configuration and stuff and very rarely used. Some ioctls
771 * pass on to the socket itself.
773 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
774 * loads the devconfigure module does its configuring and unloads it.
775 * There's a good 20K of config code hanging around the kernel.
778 int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
780 struct sock *sk = sock->sk;
785 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
790 err = ip_rt_ioctl(cmd, (void __user *)arg);
795 err = arp_ioctl(cmd, (void __user *)arg);
808 err = devinet_ioctl(cmd, (void __user *)arg);
811 if (!sk->sk_prot->ioctl ||
812 (err = sk->sk_prot->ioctl(sk, cmd, arg)) ==
814 err = dev_ioctl(cmd, (void __user *)arg);
820 struct proto_ops inet_stream_ops = {
822 .owner = THIS_MODULE,
823 .release = inet_release,
825 .connect = inet_stream_connect,
826 .socketpair = sock_no_socketpair,
827 .accept = inet_accept,
828 .getname = inet_getname,
831 .listen = inet_listen,
832 .shutdown = inet_shutdown,
833 .setsockopt = sock_common_setsockopt,
834 .getsockopt = sock_common_getsockopt,
835 .sendmsg = inet_sendmsg,
836 .recvmsg = sock_common_recvmsg,
837 .mmap = sock_no_mmap,
838 .sendpage = tcp_sendpage
841 struct proto_ops inet_dgram_ops = {
843 .owner = THIS_MODULE,
844 .release = inet_release,
846 .connect = inet_dgram_connect,
847 .socketpair = sock_no_socketpair,
848 .accept = sock_no_accept,
849 .getname = inet_getname,
850 .poll = datagram_poll,
852 .listen = sock_no_listen,
853 .shutdown = inet_shutdown,
854 .setsockopt = sock_common_setsockopt,
855 .getsockopt = sock_common_getsockopt,
856 .sendmsg = inet_sendmsg,
857 .recvmsg = sock_common_recvmsg,
858 .mmap = sock_no_mmap,
859 .sendpage = inet_sendpage,
862 struct net_proto_family inet_family_ops = {
864 .create = inet_create,
865 .owner = THIS_MODULE,
869 extern void tcp_init(void);
870 extern void tcp_v4_init(struct net_proto_family *);
872 /* Upon startup we insert all the elements in inetsw_array[] into
873 * the linked list inetsw.
875 static struct inet_protosw inetsw_array[] =
879 .protocol = IPPROTO_TCP,
881 .ops = &inet_stream_ops,
884 .flags = INET_PROTOSW_PERMANENT,
889 .protocol = IPPROTO_UDP,
891 .ops = &inet_dgram_ops,
893 .no_check = UDP_CSUM_DEFAULT,
894 .flags = INET_PROTOSW_PERMANENT,
900 .protocol = IPPROTO_IP, /* wild card */
902 .ops = &inet_dgram_ops,
903 .capability = CAP_NET_RAW,
904 .no_check = UDP_CSUM_DEFAULT,
905 .flags = INET_PROTOSW_REUSE,
909 #define INETSW_ARRAY_LEN (sizeof(inetsw_array) / sizeof(struct inet_protosw))
911 void inet_register_protosw(struct inet_protosw *p)
913 struct list_head *lh;
914 struct inet_protosw *answer;
915 int protocol = p->protocol;
916 struct list_head *last_perm;
918 spin_lock_bh(&inetsw_lock);
920 if (p->type >= SOCK_MAX)
923 /* If we are trying to override a permanent protocol, bail. */
925 last_perm = &inetsw[p->type];
926 list_for_each(lh, &inetsw[p->type]) {
927 answer = list_entry(lh, struct inet_protosw, list);
929 /* Check only the non-wild match. */
930 if (INET_PROTOSW_PERMANENT & answer->flags) {
931 if (protocol == answer->protocol)
941 /* Add the new entry after the last permanent entry if any, so that
942 * the new entry does not override a permanent entry when matched with
943 * a wild-card protocol. But it is allowed to override any existing
944 * non-permanent entry. This means that when we remove this entry, the
945 * system automatically returns to the old behavior.
947 list_add_rcu(&p->list, last_perm);
949 spin_unlock_bh(&inetsw_lock);
956 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
962 "Ignoring attempt to register invalid socket type %d.\n",
967 void inet_unregister_protosw(struct inet_protosw *p)
969 if (INET_PROTOSW_PERMANENT & p->flags) {
971 "Attempt to unregister permanent protocol %d.\n",
974 spin_lock_bh(&inetsw_lock);
975 list_del_rcu(&p->list);
976 spin_unlock_bh(&inetsw_lock);
982 #ifdef CONFIG_IP_MULTICAST
983 static struct net_protocol igmp_protocol = {
988 static struct net_protocol tcp_protocol = {
989 .handler = tcp_v4_rcv,
990 .err_handler = tcp_v4_err,
994 static struct net_protocol udp_protocol = {
996 .err_handler = udp_err,
1000 static struct net_protocol icmp_protocol = {
1001 .handler = icmp_rcv,
1004 static int __init init_ipv4_mibs(void)
1006 net_statistics[0] = alloc_percpu(struct linux_mib);
1007 net_statistics[1] = alloc_percpu(struct linux_mib);
1008 ip_statistics[0] = alloc_percpu(struct ipstats_mib);
1009 ip_statistics[1] = alloc_percpu(struct ipstats_mib);
1010 icmp_statistics[0] = alloc_percpu(struct icmp_mib);
1011 icmp_statistics[1] = alloc_percpu(struct icmp_mib);
1012 tcp_statistics[0] = alloc_percpu(struct tcp_mib);
1013 tcp_statistics[1] = alloc_percpu(struct tcp_mib);
1014 udp_statistics[0] = alloc_percpu(struct udp_mib);
1015 udp_statistics[1] = alloc_percpu(struct udp_mib);
1017 (net_statistics[0] && net_statistics[1] && ip_statistics[0]
1018 && ip_statistics[1] && tcp_statistics[0] && tcp_statistics[1]
1019 && udp_statistics[0] && udp_statistics[1]))
1022 (void) tcp_mib_init();
1027 int ipv4_proc_init(void);
1028 extern void ipfrag_init(void);
1030 static int __init inet_init(void)
1032 struct sk_buff *dummy_skb;
1033 struct inet_protosw *q;
1034 struct list_head *r;
1036 if (sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb)) {
1037 printk(KERN_CRIT "%s: panic\n", __FUNCTION__);
1041 tcp_sk_cachep = kmem_cache_create("tcp_sock",
1042 sizeof(struct tcp_sock), 0,
1043 SLAB_HWCACHE_ALIGN, NULL, NULL);
1044 udp_sk_cachep = kmem_cache_create("udp_sock",
1045 sizeof(struct udp_sock), 0,
1046 SLAB_HWCACHE_ALIGN, NULL, NULL);
1047 raw4_sk_cachep = kmem_cache_create("raw4_sock",
1048 sizeof(struct raw_sock), 0,
1049 SLAB_HWCACHE_ALIGN, NULL, NULL);
1050 if (!tcp_sk_cachep || !udp_sk_cachep || !raw4_sk_cachep)
1052 "inet_init: Can't create protocol sock SLAB caches!\n");
1054 * Tell SOCKET that we are alive...
1057 (void)sock_register(&inet_family_ops);
1060 * Add all the base protocols.
1063 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1064 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1065 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1066 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1067 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1068 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1069 #ifdef CONFIG_IP_MULTICAST
1070 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1071 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1074 /* Register the socket-side information for inet_create. */
1075 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1078 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1079 inet_register_protosw(q);
1082 * Set the ARP module up
1088 * Set the IP module up
1093 tcp_v4_init(&inet_family_ops);
1095 /* Setup TCP slab cache for open requests. */
1100 * Set the ICMP layer up
1103 icmp_init(&inet_family_ops);
1106 * Initialise the multicast router
1108 #if defined(CONFIG_IP_MROUTE)
1112 * Initialise per-cpu ipv4 mibs
1115 if(init_ipv4_mibs())
1116 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
1125 module_init(inet_init);
1127 /* ------------------------------------------------------------------------ */
1129 #ifdef CONFIG_PROC_FS
1130 extern int fib_proc_init(void);
1131 extern void fib_proc_exit(void);
1132 extern int ip_misc_proc_init(void);
1133 extern int raw_proc_init(void);
1134 extern void raw_proc_exit(void);
1135 extern int tcp4_proc_init(void);
1136 extern void tcp4_proc_exit(void);
1137 extern int udp4_proc_init(void);
1138 extern void udp4_proc_exit(void);
1140 int __init ipv4_proc_init(void)
1144 if (raw_proc_init())
1146 if (tcp4_proc_init())
1148 if (udp4_proc_init())
1150 if (fib_proc_init())
1152 if (ip_misc_proc_init())
1169 #else /* CONFIG_PROC_FS */
1170 int __init ipv4_proc_init(void)
1174 #endif /* CONFIG_PROC_FS */
1176 MODULE_ALIAS_NETPROTO(PF_INET);
1178 EXPORT_SYMBOL(inet_accept);
1179 EXPORT_SYMBOL(inet_bind);
1180 EXPORT_SYMBOL(inet_dgram_connect);
1181 EXPORT_SYMBOL(inet_dgram_ops);
1182 EXPORT_SYMBOL(inet_family_ops);
1183 EXPORT_SYMBOL(inet_getname);
1184 EXPORT_SYMBOL(inet_ioctl);
1185 EXPORT_SYMBOL(inet_listen);
1186 EXPORT_SYMBOL(inet_register_protosw);
1187 EXPORT_SYMBOL(inet_release);
1188 EXPORT_SYMBOL(inet_sendmsg);
1189 EXPORT_SYMBOL(inet_shutdown);
1190 EXPORT_SYMBOL(inet_sock_destruct);
1191 EXPORT_SYMBOL(inet_stream_connect);
1192 EXPORT_SYMBOL(inet_stream_ops);
1193 EXPORT_SYMBOL(inet_unregister_protosw);
1194 EXPORT_SYMBOL(net_statistics);
1195 EXPORT_SYMBOL(tcp_protocol);
1196 EXPORT_SYMBOL(udp_protocol);
1198 #ifdef INET_REFCNT_DEBUG
1199 EXPORT_SYMBOL(inet_sock_nr);