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>
116 DEFINE_SNMP_STAT(struct linux_mib
, net_statistics
);
118 #ifdef INET_REFCNT_DEBUG
119 atomic_t inet_sock_nr
;
122 extern void ip_mc_drop_socket(struct sock
*sk
);
124 /* The inetsw table contains everything that inet_create needs to
125 * build a new socket.
127 static struct list_head inetsw
[SOCK_MAX
];
128 static spinlock_t inetsw_lock
= SPIN_LOCK_UNLOCKED
;
130 /* New destruction routine */
132 void inet_sock_destruct(struct sock
*sk
)
134 struct inet_opt
*inet
= inet_sk(sk
);
136 __skb_queue_purge(&sk
->sk_receive_queue
);
137 __skb_queue_purge(&sk
->sk_error_queue
);
139 if (sk
->sk_type
== SOCK_STREAM
&& sk
->sk_state
!= TCP_CLOSE
) {
140 printk("Attempt to release TCP socket in state %d %p\n",
144 if (!sock_flag(sk
, SOCK_DEAD
)) {
145 printk("Attempt to release alive inet socket %p\n", sk
);
149 BUG_TRAP(!atomic_read(&sk
->sk_rmem_alloc
));
150 BUG_TRAP(!atomic_read(&sk
->sk_wmem_alloc
));
151 BUG_TRAP(!sk
->sk_wmem_queued
);
152 BUG_TRAP(!sk
->sk_forward_alloc
);
156 dst_release(sk
->sk_dst_cache
);
157 #ifdef INET_REFCNT_DEBUG
158 atomic_dec(&inet_sock_nr
);
159 printk(KERN_DEBUG
"INET socket %p released, %d are still alive\n",
160 sk
, atomic_read(&inet_sock_nr
));
165 * The routines beyond this point handle the behaviour of an AF_INET
166 * socket object. Mostly it punts to the subprotocols of IP to do
171 * Automatically bind an unbound socket.
174 static int inet_autobind(struct sock
*sk
)
176 struct inet_opt
*inet
;
177 /* We may need to bind the socket. */
181 if (sk
->sk_prot
->get_port(sk
, 0)) {
185 inet
->sport
= htons(inet
->num
);
192 * Move a socket into listening state.
194 int inet_listen(struct socket
*sock
, int backlog
)
196 struct sock
*sk
= sock
->sk
;
197 unsigned char old_state
;
203 if (sock
->state
!= SS_UNCONNECTED
|| sock
->type
!= SOCK_STREAM
)
206 old_state
= sk
->sk_state
;
207 if (!((1 << old_state
) & (TCPF_CLOSE
| TCPF_LISTEN
)))
210 /* Really, if the socket is already in listen state
211 * we can only allow the backlog to be adjusted.
213 if (old_state
!= TCP_LISTEN
) {
214 err
= tcp_listen_start(sk
);
218 sk
->sk_max_ack_backlog
= backlog
;
227 * Create an inet socket.
230 static int inet_create(struct socket
*sock
, int protocol
)
234 struct inet_protosw
*answer
;
235 struct inet_opt
*inet
;
236 struct proto
*answer_prot
;
237 unsigned char answer_flags
;
238 char answer_no_check
;
241 sock
->state
= SS_UNCONNECTED
;
243 /* Look for the requested type/protocol pair. */
246 list_for_each_rcu(p
, &inetsw
[sock
->type
]) {
247 answer
= list_entry(p
, struct inet_protosw
, list
);
249 /* Check the non-wild match. */
250 if (protocol
== answer
->protocol
) {
251 if (protocol
!= IPPROTO_IP
)
254 /* Check for the two wild cases. */
255 if (IPPROTO_IP
== protocol
) {
256 protocol
= answer
->protocol
;
259 if (IPPROTO_IP
== answer
->protocol
)
265 err
= -ESOCKTNOSUPPORT
;
269 if (answer
->capability
> 0 && !capable(answer
->capability
))
271 err
= -EPROTONOSUPPORT
;
275 sock
->ops
= answer
->ops
;
276 answer_prot
= answer
->prot
;
277 answer_no_check
= answer
->no_check
;
278 answer_flags
= answer
->flags
;
281 BUG_TRAP(answer_prot
->slab
!= NULL
);
284 sk
= sk_alloc(PF_INET
, GFP_KERNEL
,
285 answer_prot
->slab_obj_size
,
291 sk
->sk_prot
= answer_prot
;
292 sk
->sk_no_check
= answer_no_check
;
293 if (INET_PROTOSW_REUSE
& answer_flags
)
298 if (SOCK_RAW
== sock
->type
) {
299 inet
->num
= protocol
;
300 if (IPPROTO_RAW
== protocol
)
304 if (ipv4_config
.no_pmtu_disc
)
305 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
307 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
311 sock_init_data(sock
, sk
);
312 sk_set_owner(sk
, THIS_MODULE
);
314 sk
->sk_destruct
= inet_sock_destruct
;
315 sk
->sk_family
= PF_INET
;
316 sk
->sk_protocol
= protocol
;
317 sk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
323 inet
->mc_list
= NULL
;
325 #ifdef INET_REFCNT_DEBUG
326 atomic_inc(&inet_sock_nr
);
330 /* It assumes that any protocol which allows
331 * the user to assign a number at socket
332 * creation time automatically
335 inet
->sport
= htons(inet
->num
);
336 /* Add to protocol hash chains. */
337 sk
->sk_prot
->hash(sk
);
340 if (sk
->sk_prot
->init
) {
341 err
= sk
->sk_prot
->init(sk
);
343 sk_common_release(sk
);
354 * The peer socket should always be NULL (or else). When we call this
355 * function we are destroying the object and from then on nobody
356 * should refer to it.
358 int inet_release(struct socket
*sock
)
360 struct sock
*sk
= sock
->sk
;
365 /* Applications forget to leave groups before exiting */
366 ip_mc_drop_socket(sk
);
368 /* If linger is set, we don't return until the close
369 * is complete. Otherwise we return immediately. The
370 * actually closing is done the same either way.
372 * If the close is due to the process exiting, we never
376 if (sock_flag(sk
, SOCK_LINGER
) &&
377 !(current
->flags
& PF_EXITING
))
378 timeout
= sk
->sk_lingertime
;
380 sk
->sk_prot
->close(sk
, timeout
);
385 /* It is off by default, see below. */
386 int sysctl_ip_nonlocal_bind
;
388 int inet_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
390 struct sockaddr_in
*addr
= (struct sockaddr_in
*)uaddr
;
391 struct sock
*sk
= sock
->sk
;
392 struct inet_opt
*inet
= inet_sk(sk
);
397 /* If the socket has its own bind function then use it. (RAW) */
398 if (sk
->sk_prot
->bind
) {
399 err
= sk
->sk_prot
->bind(sk
, uaddr
, addr_len
);
403 if (addr_len
< sizeof(struct sockaddr_in
))
406 chk_addr_ret
= inet_addr_type(addr
->sin_addr
.s_addr
);
408 /* Not specified by any standard per-se, however it breaks too
409 * many applications when removed. It is unfortunate since
410 * allowing applications to make a non-local bind solves
411 * several problems with systems using dynamic addressing.
412 * (ie. your servers still start up even if your ISDN link
413 * is temporarily down)
415 err
= -EADDRNOTAVAIL
;
416 if (!sysctl_ip_nonlocal_bind
&&
418 addr
->sin_addr
.s_addr
!= INADDR_ANY
&&
419 chk_addr_ret
!= RTN_LOCAL
&&
420 chk_addr_ret
!= RTN_MULTICAST
&&
421 chk_addr_ret
!= RTN_BROADCAST
)
424 snum
= ntohs(addr
->sin_port
);
426 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
429 /* We keep a pair of addresses. rcv_saddr is the one
430 * used by hash lookups, and saddr is used for transmit.
432 * In the BSD API these are the same except where it
433 * would be illegal to use them (multicast/broadcast) in
434 * which case the sending device address is used.
438 /* Check these errors (active socket, double bind). */
440 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->num
)
441 goto out_release_sock
;
443 inet
->rcv_saddr
= inet
->saddr
= addr
->sin_addr
.s_addr
;
444 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
445 inet
->saddr
= 0; /* Use device */
447 /* Make sure we are allowed to bind here. */
448 if (sk
->sk_prot
->get_port(sk
, snum
)) {
449 inet
->saddr
= inet
->rcv_saddr
= 0;
451 goto out_release_sock
;
455 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
457 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
458 inet
->sport
= htons(inet
->num
);
469 int inet_dgram_connect(struct socket
*sock
, struct sockaddr
* uaddr
,
470 int addr_len
, int flags
)
472 struct sock
*sk
= sock
->sk
;
474 if (uaddr
->sa_family
== AF_UNSPEC
)
475 return sk
->sk_prot
->disconnect(sk
, flags
);
477 if (!inet_sk(sk
)->num
&& inet_autobind(sk
))
479 return sk
->sk_prot
->connect(sk
, (struct sockaddr
*)uaddr
, addr_len
);
482 static long inet_wait_for_connect(struct sock
*sk
, long timeo
)
486 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
488 /* Basic assumption: if someone sets sk->sk_err, he _must_
489 * change state of the socket from TCP_SYN_*.
490 * Connect() does not allow to get error notifications
491 * without closing the socket.
493 while ((1 << sk
->sk_state
) & (TCPF_SYN_SENT
| TCPF_SYN_RECV
)) {
495 timeo
= schedule_timeout(timeo
);
497 if (signal_pending(current
) || !timeo
)
499 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
501 finish_wait(sk
->sk_sleep
, &wait
);
506 * Connect to a remote host. There is regrettably still a little
507 * TCP 'magic' in here.
509 int inet_stream_connect(struct socket
*sock
, struct sockaddr
*uaddr
,
510 int addr_len
, int flags
)
512 struct sock
*sk
= sock
->sk
;
518 if (uaddr
->sa_family
== AF_UNSPEC
) {
519 err
= sk
->sk_prot
->disconnect(sk
, flags
);
520 sock
->state
= err
? SS_DISCONNECTING
: SS_UNCONNECTED
;
524 switch (sock
->state
) {
533 /* Fall out of switch with err, set for this state */
537 if (sk
->sk_state
!= TCP_CLOSE
)
540 err
= sk
->sk_prot
->connect(sk
, uaddr
, addr_len
);
544 sock
->state
= SS_CONNECTING
;
546 /* Just entered SS_CONNECTING state; the only
547 * difference is that return value in non-blocking
548 * case is EINPROGRESS, rather than EALREADY.
554 timeo
= sock_sndtimeo(sk
, flags
& O_NONBLOCK
);
556 if ((1 << sk
->sk_state
) & (TCPF_SYN_SENT
| TCPF_SYN_RECV
)) {
557 /* Error code is set above */
558 if (!timeo
|| !inet_wait_for_connect(sk
, timeo
))
561 err
= sock_intr_errno(timeo
);
562 if (signal_pending(current
))
566 /* Connection was closed by RST, timeout, ICMP error
567 * or another process disconnected us.
569 if (sk
->sk_state
== TCP_CLOSE
)
572 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
573 * and error was received after socket entered established state.
574 * Hence, it is handled normally after connect() return successfully.
577 sock
->state
= SS_CONNECTED
;
584 err
= sock_error(sk
) ? : -ECONNABORTED
;
585 sock
->state
= SS_UNCONNECTED
;
586 if (sk
->sk_prot
->disconnect(sk
, flags
))
587 sock
->state
= SS_DISCONNECTING
;
592 * Accept a pending connection. The TCP layer now gives BSD semantics.
595 int inet_accept(struct socket
*sock
, struct socket
*newsock
, int flags
)
597 struct sock
*sk1
= sock
->sk
;
599 struct sock
*sk2
= sk1
->sk_prot
->accept(sk1
, flags
, &err
);
606 BUG_TRAP((1 << sk2
->sk_state
) &
607 (TCPF_ESTABLISHED
| TCPF_CLOSE_WAIT
| TCPF_CLOSE
));
609 sock_graft(sk2
, newsock
);
611 newsock
->state
= SS_CONNECTED
;
620 * This does both peername and sockname.
622 int inet_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
623 int *uaddr_len
, int peer
)
625 struct sock
*sk
= sock
->sk
;
626 struct inet_opt
*inet
= inet_sk(sk
);
627 struct sockaddr_in
*sin
= (struct sockaddr_in
*)uaddr
;
629 sin
->sin_family
= AF_INET
;
632 (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
635 sin
->sin_port
= inet
->dport
;
636 sin
->sin_addr
.s_addr
= inet
->daddr
;
638 __u32 addr
= inet
->rcv_saddr
;
641 sin
->sin_port
= inet
->sport
;
642 sin
->sin_addr
.s_addr
= addr
;
644 memset(sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
645 *uaddr_len
= sizeof(*sin
);
649 int inet_sendmsg(struct kiocb
*iocb
, struct socket
*sock
, struct msghdr
*msg
,
652 struct sock
*sk
= sock
->sk
;
654 /* We may need to bind the socket. */
655 if (!inet_sk(sk
)->num
&& inet_autobind(sk
))
658 return sk
->sk_prot
->sendmsg(iocb
, sk
, msg
, size
);
662 ssize_t
inet_sendpage(struct socket
*sock
, struct page
*page
, int offset
, size_t size
, int flags
)
664 struct sock
*sk
= sock
->sk
;
666 /* We may need to bind the socket. */
667 if (!inet_sk(sk
)->num
&& inet_autobind(sk
))
670 if (sk
->sk_prot
->sendpage
)
671 return sk
->sk_prot
->sendpage(sk
, page
, offset
, size
, flags
);
672 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
676 int inet_shutdown(struct socket
*sock
, int how
)
678 struct sock
*sk
= sock
->sk
;
681 /* This should really check to make sure
682 * the socket is a TCP socket. (WHY AC...)
684 how
++; /* maps 0->1 has the advantage of making bit 1 rcvs and
687 if ((how
& ~SHUTDOWN_MASK
) || !how
) /* MAXINT->0 */
691 if (sock
->state
== SS_CONNECTING
) {
692 if ((1 << sk
->sk_state
) &
693 (TCPF_SYN_SENT
| TCPF_SYN_RECV
| TCPF_CLOSE
))
694 sock
->state
= SS_DISCONNECTING
;
696 sock
->state
= SS_CONNECTED
;
699 switch (sk
->sk_state
) {
702 /* Hack to wake up other listeners, who can poll for
703 POLLHUP, even on eg. unconnected UDP sockets -- RR */
705 sk
->sk_shutdown
|= how
;
706 if (sk
->sk_prot
->shutdown
)
707 sk
->sk_prot
->shutdown(sk
, how
);
710 /* Remaining two branches are temporary solution for missing
711 * close() in multithreaded environment. It is _not_ a good idea,
712 * but we have no choice until close() is repaired at VFS level.
715 if (!(how
& RCV_SHUTDOWN
))
719 err
= sk
->sk_prot
->disconnect(sk
, O_NONBLOCK
);
720 sock
->state
= err
? SS_DISCONNECTING
: SS_UNCONNECTED
;
724 /* Wake up anyone sleeping in poll. */
725 sk
->sk_state_change(sk
);
731 * ioctl() calls you can issue on an INET socket. Most of these are
732 * device configuration and stuff and very rarely used. Some ioctls
733 * pass on to the socket itself.
735 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
736 * loads the devconfigure module does its configuring and unloads it.
737 * There's a good 20K of config code hanging around the kernel.
740 int inet_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
742 struct sock
*sk
= sock
->sk
;
747 err
= sock_get_timestamp(sk
, (struct timeval __user
*)arg
);
752 err
= ip_rt_ioctl(cmd
, (void __user
*)arg
);
757 err
= arp_ioctl(cmd
, (void __user
*)arg
);
770 err
= devinet_ioctl(cmd
, (void __user
*)arg
);
773 if (!sk
->sk_prot
->ioctl
||
774 (err
= sk
->sk_prot
->ioctl(sk
, cmd
, arg
)) ==
776 err
= dev_ioctl(cmd
, (void __user
*)arg
);
782 struct proto_ops inet_stream_ops
= {
784 .owner
= THIS_MODULE
,
785 .release
= inet_release
,
787 .connect
= inet_stream_connect
,
788 .socketpair
= sock_no_socketpair
,
789 .accept
= inet_accept
,
790 .getname
= inet_getname
,
793 .listen
= inet_listen
,
794 .shutdown
= inet_shutdown
,
795 .setsockopt
= sock_common_setsockopt
,
796 .getsockopt
= sock_common_getsockopt
,
797 .sendmsg
= inet_sendmsg
,
798 .recvmsg
= sock_common_recvmsg
,
799 .mmap
= sock_no_mmap
,
800 .sendpage
= tcp_sendpage
803 struct proto_ops inet_dgram_ops
= {
805 .owner
= THIS_MODULE
,
806 .release
= inet_release
,
808 .connect
= inet_dgram_connect
,
809 .socketpair
= sock_no_socketpair
,
810 .accept
= sock_no_accept
,
811 .getname
= inet_getname
,
812 .poll
= datagram_poll
,
814 .listen
= sock_no_listen
,
815 .shutdown
= inet_shutdown
,
816 .setsockopt
= sock_common_setsockopt
,
817 .getsockopt
= sock_common_getsockopt
,
818 .sendmsg
= inet_sendmsg
,
819 .recvmsg
= sock_common_recvmsg
,
820 .mmap
= sock_no_mmap
,
821 .sendpage
= inet_sendpage
,
824 static struct net_proto_family inet_family_ops
= {
826 .create
= inet_create
,
827 .owner
= THIS_MODULE
,
831 extern void tcp_init(void);
832 extern void tcp_v4_init(struct net_proto_family
*);
834 /* Upon startup we insert all the elements in inetsw_array[] into
835 * the linked list inetsw.
837 static struct inet_protosw inetsw_array
[] =
841 .protocol
= IPPROTO_TCP
,
843 .ops
= &inet_stream_ops
,
846 .flags
= INET_PROTOSW_PERMANENT
,
851 .protocol
= IPPROTO_UDP
,
853 .ops
= &inet_dgram_ops
,
855 .no_check
= UDP_CSUM_DEFAULT
,
856 .flags
= INET_PROTOSW_PERMANENT
,
862 .protocol
= IPPROTO_IP
, /* wild card */
864 .ops
= &inet_dgram_ops
,
865 .capability
= CAP_NET_RAW
,
866 .no_check
= UDP_CSUM_DEFAULT
,
867 .flags
= INET_PROTOSW_REUSE
,
871 #define INETSW_ARRAY_LEN (sizeof(inetsw_array) / sizeof(struct inet_protosw))
873 void inet_register_protosw(struct inet_protosw
*p
)
875 struct list_head
*lh
;
876 struct inet_protosw
*answer
;
877 int protocol
= p
->protocol
;
878 struct list_head
*last_perm
;
880 spin_lock_bh(&inetsw_lock
);
882 if (p
->type
>= SOCK_MAX
)
885 /* If we are trying to override a permanent protocol, bail. */
887 last_perm
= &inetsw
[p
->type
];
888 list_for_each(lh
, &inetsw
[p
->type
]) {
889 answer
= list_entry(lh
, struct inet_protosw
, list
);
891 /* Check only the non-wild match. */
892 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
893 if (protocol
== answer
->protocol
)
903 /* Add the new entry after the last permanent entry if any, so that
904 * the new entry does not override a permanent entry when matched with
905 * a wild-card protocol. But it is allowed to override any existing
906 * non-permanent entry. This means that when we remove this entry, the
907 * system automatically returns to the old behavior.
909 list_add_rcu(&p
->list
, last_perm
);
911 spin_unlock_bh(&inetsw_lock
);
913 #if 1 // mask by Victor Yu. 05-26-2005, I don't know what happen on it.
920 printk(KERN_ERR
"Attempt to override permanent protocol %d.\n",
926 "Ignoring attempt to register invalid socket type %d.\n",
931 void inet_unregister_protosw(struct inet_protosw
*p
)
933 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
935 "Attempt to unregister permanent protocol %d.\n",
938 spin_lock_bh(&inetsw_lock
);
939 list_del_rcu(&p
->list
);
940 spin_unlock_bh(&inetsw_lock
);
946 #ifdef CONFIG_IP_MULTICAST
947 static struct net_protocol igmp_protocol
= {
952 static struct net_protocol tcp_protocol
= {
953 .handler
= tcp_v4_rcv
,
954 .err_handler
= tcp_v4_err
,
958 static struct net_protocol udp_protocol
= {
960 .err_handler
= udp_err
,
964 static struct net_protocol icmp_protocol
= {
968 static int __init
init_ipv4_mibs(void)
970 net_statistics
[0] = alloc_percpu(struct linux_mib
);
971 net_statistics
[1] = alloc_percpu(struct linux_mib
);
972 ip_statistics
[0] = alloc_percpu(struct ipstats_mib
);
973 ip_statistics
[1] = alloc_percpu(struct ipstats_mib
);
974 icmp_statistics
[0] = alloc_percpu(struct icmp_mib
);
975 icmp_statistics
[1] = alloc_percpu(struct icmp_mib
);
976 tcp_statistics
[0] = alloc_percpu(struct tcp_mib
);
977 tcp_statistics
[1] = alloc_percpu(struct tcp_mib
);
978 udp_statistics
[0] = alloc_percpu(struct udp_mib
);
979 udp_statistics
[1] = alloc_percpu(struct udp_mib
);
981 (net_statistics
[0] && net_statistics
[1] && ip_statistics
[0]
982 && ip_statistics
[1] && tcp_statistics
[0] && tcp_statistics
[1]
983 && udp_statistics
[0] && udp_statistics
[1]))
986 (void) tcp_mib_init();
991 int ipv4_proc_init(void);
992 extern void ipfrag_init(void);
994 static int __init
inet_init(void)
996 struct sk_buff
*dummy_skb
;
997 struct inet_protosw
*q
;
1001 if (sizeof(struct inet_skb_parm
) > sizeof(dummy_skb
->cb
)) {
1002 printk(KERN_CRIT
"%s: panic\n", __FUNCTION__
);
1006 rc
= sk_alloc_slab(&tcp_prot
, "tcp_sock");
1008 sk_alloc_slab_error(&tcp_prot
);
1011 rc
= sk_alloc_slab(&udp_prot
, "udp_sock");
1013 sk_alloc_slab_error(&udp_prot
);
1014 goto out_tcp_free_slab
;
1016 rc
= sk_alloc_slab(&raw_prot
, "raw_sock");
1018 sk_alloc_slab_error(&raw_prot
);
1019 goto out_udp_free_slab
;
1023 * Tell SOCKET that we are alive...
1027 (void)sock_register(&inet_family_ops
);
1030 * Add all the base protocols.
1034 if (inet_add_protocol(&icmp_protocol
, IPPROTO_ICMP
) < 0)
1035 printk(KERN_CRIT
"inet_init: Cannot add ICMP protocol\n");
1036 if (inet_add_protocol(&udp_protocol
, IPPROTO_UDP
) < 0)
1037 printk(KERN_CRIT
"inet_init: Cannot add UDP protocol\n");
1038 if (inet_add_protocol(&tcp_protocol
, IPPROTO_TCP
) < 0)
1039 printk(KERN_CRIT
"inet_init: Cannot add TCP protocol\n");
1040 #ifdef CONFIG_IP_MULTICAST
1041 if (inet_add_protocol(&igmp_protocol
, IPPROTO_IGMP
) < 0)
1042 printk(KERN_CRIT
"inet_init: Cannot add IGMP protocol\n");
1045 /* Register the socket-side information for inet_create. */
1046 for (r
= &inetsw
[0]; r
< &inetsw
[SOCK_MAX
]; ++r
)
1049 for (q
= inetsw_array
; q
< &inetsw_array
[INETSW_ARRAY_LEN
]; ++q
) {
1050 inet_register_protosw(q
);
1054 * Set the ARP module up
1060 * Set the IP module up
1068 tcp_v4_init(&inet_family_ops
);
1070 /* Setup TCP slab cache for open requests. */
1075 * Set the ICMP layer up
1078 icmp_init(&inet_family_ops
);
1081 * Initialise the multicast router
1083 #if defined(CONFIG_IP_MROUTE)
1087 * Initialise per-cpu ipv4 mibs
1090 if(init_ipv4_mibs())
1091 printk(KERN_CRIT
"inet_init: Cannot init ipv4 mibs\n"); ;
1104 sk_free_slab(&tcp_prot
);
1106 sk_free_slab(&udp_prot
);
1110 module_init(inet_init
);
1112 /* ------------------------------------------------------------------------ */
1114 #ifdef CONFIG_PROC_FS
1115 extern int fib_proc_init(void);
1116 extern void fib_proc_exit(void);
1117 extern int ip_misc_proc_init(void);
1118 extern int raw_proc_init(void);
1119 extern void raw_proc_exit(void);
1120 extern int tcp4_proc_init(void);
1121 extern void tcp4_proc_exit(void);
1122 extern int udp4_proc_init(void);
1123 extern void udp4_proc_exit(void);
1125 int __init
ipv4_proc_init(void)
1129 if (raw_proc_init())
1131 if (tcp4_proc_init())
1133 if (udp4_proc_init())
1135 if (fib_proc_init())
1137 if (ip_misc_proc_init())
1154 #else /* CONFIG_PROC_FS */
1155 int __init
ipv4_proc_init(void)
1159 #endif /* CONFIG_PROC_FS */
1161 MODULE_ALIAS_NETPROTO(PF_INET
);
1163 EXPORT_SYMBOL(inet_accept
);
1164 EXPORT_SYMBOL(inet_bind
);
1165 EXPORT_SYMBOL(inet_dgram_connect
);
1166 EXPORT_SYMBOL(inet_dgram_ops
);
1167 EXPORT_SYMBOL(inet_getname
);
1168 EXPORT_SYMBOL(inet_ioctl
);
1169 EXPORT_SYMBOL(inet_listen
);
1170 EXPORT_SYMBOL(inet_register_protosw
);
1171 EXPORT_SYMBOL(inet_release
);
1172 EXPORT_SYMBOL(inet_sendmsg
);
1173 EXPORT_SYMBOL(inet_shutdown
);
1174 EXPORT_SYMBOL(inet_sock_destruct
);
1175 EXPORT_SYMBOL(inet_stream_connect
);
1176 EXPORT_SYMBOL(inet_stream_ops
);
1177 EXPORT_SYMBOL(inet_unregister_protosw
);
1178 EXPORT_SYMBOL(net_statistics
);
1179 EXPORT_SYMBOL(tcp_protocol
);
1180 EXPORT_SYMBOL(udp_protocol
);
1182 #ifdef INET_REFCNT_DEBUG
1183 EXPORT_SYMBOL(inet_sock_nr
);