af_inet.c revision bbcf467dab42ea3c85f368df346c82af2fbba665 
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.
5 *
6 *		PF_INET protocol family socket handler.
7 *
8 * Version:	$Id: af_inet.c,v 1.137 2002/02/01 22:01:03 davem Exp $
9 *
10 * Authors:	Ross Biro
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>
14 *
15 * Changes (see also sock.c)
16 *
17 *		piggy,
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
25 *					structures
26 *					when accept() ed
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
32 *		Alan Cox,
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
42 *					compactness.
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
48 *					dumbly.
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.
63 *
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.
68 */
69
70#include <linux/config.h>
71#include <linux/err.h>
72#include <linux/errno.h>
73#include <linux/types.h>
74#include <linux/socket.h>
75#include <linux/in.h>
76#include <linux/kernel.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/capability.h>
84#include <linux/fcntl.h>
85#include <linux/mm.h>
86#include <linux/interrupt.h>
87#include <linux/stat.h>
88#include <linux/init.h>
89#include <linux/poll.h>
90#include <linux/netfilter_ipv4.h>
91
92#include <asm/uaccess.h>
93#include <asm/system.h>
94
95#include <linux/smp_lock.h>
96#include <linux/inet.h>
97#include <linux/igmp.h>
98#include <linux/inetdevice.h>
99#include <linux/netdevice.h>
100#include <net/ip.h>
101#include <net/protocol.h>
102#include <net/arp.h>
103#include <net/route.h>
104#include <net/ip_fib.h>
105#include <net/inet_connection_sock.h>
106#include <net/tcp.h>
107#include <net/udp.h>
108#include <linux/skbuff.h>
109#include <net/sock.h>
110#include <net/raw.h>
111#include <net/icmp.h>
112#include <net/ipip.h>
113#include <net/inet_common.h>
114#include <net/xfrm.h>
115#ifdef CONFIG_IP_MROUTE
116#include <linux/mroute.h>
117#endif
118
119DEFINE_SNMP_STAT(struct linux_mib, net_statistics) __read_mostly;
120
121extern void ip_mc_drop_socket(struct sock *sk);
122
123/* The inetsw table contains everything that inet_create needs to
124 * build a new socket.
125 */
126static struct list_head inetsw[SOCK_MAX];
127static DEFINE_SPINLOCK(inetsw_lock);
128
129/* New destruction routine */
130
131void inet_sock_destruct(struct sock *sk)
132{
133	struct inet_sock *inet = inet_sk(sk);
134
135	__skb_queue_purge(&sk->sk_receive_queue);
136	__skb_queue_purge(&sk->sk_error_queue);
137
138	if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
139		printk("Attempt to release TCP socket in state %d %p\n",
140		       sk->sk_state, sk);
141		return;
142	}
143	if (!sock_flag(sk, SOCK_DEAD)) {
144		printk("Attempt to release alive inet socket %p\n", sk);
145		return;
146	}
147
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);
152
153	kfree(inet->opt);
154	dst_release(sk->sk_dst_cache);
155	sk_refcnt_debug_dec(sk);
156}
157
158/*
159 *	The routines beyond this point handle the behaviour of an AF_INET
160 *	socket object. Mostly it punts to the subprotocols of IP to do
161 *	the work.
162 */
163
164/*
165 *	Automatically bind an unbound socket.
166 */
167
168static int inet_autobind(struct sock *sk)
169{
170	struct inet_sock *inet;
171	/* We may need to bind the socket. */
172	lock_sock(sk);
173	inet = inet_sk(sk);
174	if (!inet->num) {
175		if (sk->sk_prot->get_port(sk, 0)) {
176			release_sock(sk);
177			return -EAGAIN;
178		}
179		inet->sport = htons(inet->num);
180	}
181	release_sock(sk);
182	return 0;
183}
184
185/*
186 *	Move a socket into listening state.
187 */
188int inet_listen(struct socket *sock, int backlog)
189{
190	struct sock *sk = sock->sk;
191	unsigned char old_state;
192	int err;
193
194	lock_sock(sk);
195
196	err = -EINVAL;
197	if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
198		goto out;
199
200	old_state = sk->sk_state;
201	if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
202		goto out;
203
204	/* Really, if the socket is already in listen state
205	 * we can only allow the backlog to be adjusted.
206	 */
207	if (old_state != TCP_LISTEN) {
208		err = inet_csk_listen_start(sk, TCP_SYNQ_HSIZE);
209		if (err)
210			goto out;
211	}
212	sk->sk_max_ack_backlog = backlog;
213	err = 0;
214
215out:
216	release_sock(sk);
217	return err;
218}
219
220/*
221 *	Create an inet socket.
222 */
223
224static int inet_create(struct socket *sock, int protocol)
225{
226	struct sock *sk;
227	struct list_head *p;
228	struct inet_protosw *answer;
229	struct inet_sock *inet;
230	struct proto *answer_prot;
231	unsigned char answer_flags;
232	char answer_no_check;
233	int try_loading_module = 0;
234	int err;
235
236	sock->state = SS_UNCONNECTED;
237
238	/* Look for the requested type/protocol pair. */
239	answer = NULL;
240lookup_protocol:
241	err = -ESOCKTNOSUPPORT;
242	rcu_read_lock();
243	list_for_each_rcu(p, &inetsw[sock->type]) {
244		answer = list_entry(p, struct inet_protosw, list);
245
246		/* Check the non-wild match. */
247		if (protocol == answer->protocol) {
248			if (protocol != IPPROTO_IP)
249				break;
250		} else {
251			/* Check for the two wild cases. */
252			if (IPPROTO_IP == protocol) {
253				protocol = answer->protocol;
254				break;
255			}
256			if (IPPROTO_IP == answer->protocol)
257				break;
258		}
259		err = -EPROTONOSUPPORT;
260		answer = NULL;
261	}
262
263	if (unlikely(answer == NULL)) {
264		if (try_loading_module < 2) {
265			rcu_read_unlock();
266			/*
267			 * Be more specific, e.g. net-pf-2-proto-132-type-1
268			 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
269			 */
270			if (++try_loading_module == 1)
271				request_module("net-pf-%d-proto-%d-type-%d",
272					       PF_INET, protocol, sock->type);
273			/*
274			 * Fall back to generic, e.g. net-pf-2-proto-132
275			 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
276			 */
277			else
278				request_module("net-pf-%d-proto-%d",
279					       PF_INET, protocol);
280			goto lookup_protocol;
281		} else
282			goto out_rcu_unlock;
283	}
284
285	err = -EPERM;
286	if (answer->capability > 0 && !capable(answer->capability))
287		goto out_rcu_unlock;
288
289	sock->ops = answer->ops;
290	answer_prot = answer->prot;
291	answer_no_check = answer->no_check;
292	answer_flags = answer->flags;
293	rcu_read_unlock();
294
295	BUG_TRAP(answer_prot->slab != NULL);
296
297	err = -ENOBUFS;
298	sk = sk_alloc(PF_INET, GFP_KERNEL, answer_prot, 1);
299	if (sk == NULL)
300		goto out;
301
302	err = 0;
303	sk->sk_no_check = answer_no_check;
304	if (INET_PROTOSW_REUSE & answer_flags)
305		sk->sk_reuse = 1;
306
307	inet = inet_sk(sk);
308	inet->is_icsk = INET_PROTOSW_ICSK & answer_flags;
309
310	if (SOCK_RAW == sock->type) {
311		inet->num = protocol;
312		if (IPPROTO_RAW == protocol)
313			inet->hdrincl = 1;
314	}
315
316	if (ipv4_config.no_pmtu_disc)
317		inet->pmtudisc = IP_PMTUDISC_DONT;
318	else
319		inet->pmtudisc = IP_PMTUDISC_WANT;
320
321	inet->id = 0;
322
323	sock_init_data(sock, sk);
324
325	sk->sk_destruct	   = inet_sock_destruct;
326	sk->sk_family	   = PF_INET;
327	sk->sk_protocol	   = protocol;
328	sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
329
330	inet->uc_ttl	= -1;
331	inet->mc_loop	= 1;
332	inet->mc_ttl	= 1;
333	inet->mc_index	= 0;
334	inet->mc_list	= NULL;
335
336	sk_refcnt_debug_inc(sk);
337
338	if (inet->num) {
339		/* It assumes that any protocol which allows
340		 * the user to assign a number at socket
341		 * creation time automatically
342		 * shares.
343		 */
344		inet->sport = htons(inet->num);
345		/* Add to protocol hash chains. */
346		sk->sk_prot->hash(sk);
347	}
348
349	if (sk->sk_prot->init) {
350		err = sk->sk_prot->init(sk);
351		if (err)
352			sk_common_release(sk);
353	}
354out:
355	return err;
356out_rcu_unlock:
357	rcu_read_unlock();
358	goto out;
359}
360
361
362/*
363 *	The peer socket should always be NULL (or else). When we call this
364 *	function we are destroying the object and from then on nobody
365 *	should refer to it.
366 */
367int inet_release(struct socket *sock)
368{
369	struct sock *sk = sock->sk;
370
371	if (sk) {
372		long timeout;
373
374		/* Applications forget to leave groups before exiting */
375		ip_mc_drop_socket(sk);
376
377		/* If linger is set, we don't return until the close
378		 * is complete.  Otherwise we return immediately. The
379		 * actually closing is done the same either way.
380		 *
381		 * If the close is due to the process exiting, we never
382		 * linger..
383		 */
384		timeout = 0;
385		if (sock_flag(sk, SOCK_LINGER) &&
386		    !(current->flags & PF_EXITING))
387			timeout = sk->sk_lingertime;
388		sock->sk = NULL;
389		sk->sk_prot->close(sk, timeout);
390	}
391	return 0;
392}
393
394/* It is off by default, see below. */
395int sysctl_ip_nonlocal_bind;
396
397int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
398{
399	struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
400	struct sock *sk = sock->sk;
401	struct inet_sock *inet = inet_sk(sk);
402	unsigned short snum;
403	int chk_addr_ret;
404	int err;
405
406	/* If the socket has its own bind function then use it. (RAW) */
407	if (sk->sk_prot->bind) {
408		err = sk->sk_prot->bind(sk, uaddr, addr_len);
409		goto out;
410	}
411	err = -EINVAL;
412	if (addr_len < sizeof(struct sockaddr_in))
413		goto out;
414
415	chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
416
417	/* Not specified by any standard per-se, however it breaks too
418	 * many applications when removed.  It is unfortunate since
419	 * allowing applications to make a non-local bind solves
420	 * several problems with systems using dynamic addressing.
421	 * (ie. your servers still start up even if your ISDN link
422	 *  is temporarily down)
423	 */
424	err = -EADDRNOTAVAIL;
425	if (!sysctl_ip_nonlocal_bind &&
426	    !inet->freebind &&
427	    addr->sin_addr.s_addr != INADDR_ANY &&
428	    chk_addr_ret != RTN_LOCAL &&
429	    chk_addr_ret != RTN_MULTICAST &&
430	    chk_addr_ret != RTN_BROADCAST)
431		goto out;
432
433	snum = ntohs(addr->sin_port);
434	err = -EACCES;
435	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
436		goto out;
437
438	/*      We keep a pair of addresses. rcv_saddr is the one
439	 *      used by hash lookups, and saddr is used for transmit.
440	 *
441	 *      In the BSD API these are the same except where it
442	 *      would be illegal to use them (multicast/broadcast) in
443	 *      which case the sending device address is used.
444	 */
445	lock_sock(sk);
446
447	/* Check these errors (active socket, double bind). */
448	err = -EINVAL;
449	if (sk->sk_state != TCP_CLOSE || inet->num)
450		goto out_release_sock;
451
452	inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
453	if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
454		inet->saddr = 0;  /* Use device */
455
456	/* Make sure we are allowed to bind here. */
457	if (sk->sk_prot->get_port(sk, snum)) {
458		inet->saddr = inet->rcv_saddr = 0;
459		err = -EADDRINUSE;
460		goto out_release_sock;
461	}
462
463	if (inet->rcv_saddr)
464		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
465	if (snum)
466		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
467	inet->sport = htons(inet->num);
468	inet->daddr = 0;
469	inet->dport = 0;
470	sk_dst_reset(sk);
471	err = 0;
472out_release_sock:
473	release_sock(sk);
474out:
475	return err;
476}
477
478int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
479		       int addr_len, int flags)
480{
481	struct sock *sk = sock->sk;
482
483	if (uaddr->sa_family == AF_UNSPEC)
484		return sk->sk_prot->disconnect(sk, flags);
485
486	if (!inet_sk(sk)->num && inet_autobind(sk))
487		return -EAGAIN;
488	return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
489}
490
491static long inet_wait_for_connect(struct sock *sk, long timeo)
492{
493	DEFINE_WAIT(wait);
494
495	prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
496
497	/* Basic assumption: if someone sets sk->sk_err, he _must_
498	 * change state of the socket from TCP_SYN_*.
499	 * Connect() does not allow to get error notifications
500	 * without closing the socket.
501	 */
502	while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
503		release_sock(sk);
504		timeo = schedule_timeout(timeo);
505		lock_sock(sk);
506		if (signal_pending(current) || !timeo)
507			break;
508		prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
509	}
510	finish_wait(sk->sk_sleep, &wait);
511	return timeo;
512}
513
514/*
515 *	Connect to a remote host. There is regrettably still a little
516 *	TCP 'magic' in here.
517 */
518int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
519			int addr_len, int flags)
520{
521	struct sock *sk = sock->sk;
522	int err;
523	long timeo;
524
525	lock_sock(sk);
526
527	if (uaddr->sa_family == AF_UNSPEC) {
528		err = sk->sk_prot->disconnect(sk, flags);
529		sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
530		goto out;
531	}
532
533	switch (sock->state) {
534	default:
535		err = -EINVAL;
536		goto out;
537	case SS_CONNECTED:
538		err = -EISCONN;
539		goto out;
540	case SS_CONNECTING:
541		err = -EALREADY;
542		/* Fall out of switch with err, set for this state */
543		break;
544	case SS_UNCONNECTED:
545		err = -EISCONN;
546		if (sk->sk_state != TCP_CLOSE)
547			goto out;
548
549		err = sk->sk_prot->connect(sk, uaddr, addr_len);
550		if (err < 0)
551			goto out;
552
553  		sock->state = SS_CONNECTING;
554
555		/* Just entered SS_CONNECTING state; the only
556		 * difference is that return value in non-blocking
557		 * case is EINPROGRESS, rather than EALREADY.
558		 */
559		err = -EINPROGRESS;
560		break;
561	}
562
563	timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
564
565	if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
566		/* Error code is set above */
567		if (!timeo || !inet_wait_for_connect(sk, timeo))
568			goto out;
569
570		err = sock_intr_errno(timeo);
571		if (signal_pending(current))
572			goto out;
573	}
574
575	/* Connection was closed by RST, timeout, ICMP error
576	 * or another process disconnected us.
577	 */
578	if (sk->sk_state == TCP_CLOSE)
579		goto sock_error;
580
581	/* sk->sk_err may be not zero now, if RECVERR was ordered by user
582	 * and error was received after socket entered established state.
583	 * Hence, it is handled normally after connect() return successfully.
584	 */
585
586	sock->state = SS_CONNECTED;
587	err = 0;
588out:
589	release_sock(sk);
590	return err;
591
592sock_error:
593	err = sock_error(sk) ? : -ECONNABORTED;
594	sock->state = SS_UNCONNECTED;
595	if (sk->sk_prot->disconnect(sk, flags))
596		sock->state = SS_DISCONNECTING;
597	goto out;
598}
599
600/*
601 *	Accept a pending connection. The TCP layer now gives BSD semantics.
602 */
603
604int inet_accept(struct socket *sock, struct socket *newsock, int flags)
605{
606	struct sock *sk1 = sock->sk;
607	int err = -EINVAL;
608	struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
609
610	if (!sk2)
611		goto do_err;
612
613	lock_sock(sk2);
614
615	BUG_TRAP((1 << sk2->sk_state) &
616		 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE));
617
618	sock_graft(sk2, newsock);
619
620	newsock->state = SS_CONNECTED;
621	err = 0;
622	release_sock(sk2);
623do_err:
624	return err;
625}
626
627
628/*
629 *	This does both peername and sockname.
630 */
631int inet_getname(struct socket *sock, struct sockaddr *uaddr,
632			int *uaddr_len, int peer)
633{
634	struct sock *sk		= sock->sk;
635	struct inet_sock *inet	= inet_sk(sk);
636	struct sockaddr_in *sin	= (struct sockaddr_in *)uaddr;
637
638	sin->sin_family = AF_INET;
639	if (peer) {
640		if (!inet->dport ||
641		    (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
642		     peer == 1))
643			return -ENOTCONN;
644		sin->sin_port = inet->dport;
645		sin->sin_addr.s_addr = inet->daddr;
646	} else {
647		__u32 addr = inet->rcv_saddr;
648		if (!addr)
649			addr = inet->saddr;
650		sin->sin_port = inet->sport;
651		sin->sin_addr.s_addr = addr;
652	}
653	memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
654	*uaddr_len = sizeof(*sin);
655	return 0;
656}
657
658int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
659		 size_t size)
660{
661	struct sock *sk = sock->sk;
662
663	/* We may need to bind the socket. */
664	if (!inet_sk(sk)->num && inet_autobind(sk))
665		return -EAGAIN;
666
667	return sk->sk_prot->sendmsg(iocb, sk, msg, size);
668}
669
670
671static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
672{
673	struct sock *sk = sock->sk;
674
675	/* We may need to bind the socket. */
676	if (!inet_sk(sk)->num && inet_autobind(sk))
677		return -EAGAIN;
678
679	if (sk->sk_prot->sendpage)
680		return sk->sk_prot->sendpage(sk, page, offset, size, flags);
681	return sock_no_sendpage(sock, page, offset, size, flags);
682}
683
684
685int inet_shutdown(struct socket *sock, int how)
686{
687	struct sock *sk = sock->sk;
688	int err = 0;
689
690	/* This should really check to make sure
691	 * the socket is a TCP socket. (WHY AC...)
692	 */
693	how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
694		       1->2 bit 2 snds.
695		       2->3 */
696	if ((how & ~SHUTDOWN_MASK) || !how)	/* MAXINT->0 */
697		return -EINVAL;
698
699	lock_sock(sk);
700	if (sock->state == SS_CONNECTING) {
701		if ((1 << sk->sk_state) &
702		    (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
703			sock->state = SS_DISCONNECTING;
704		else
705			sock->state = SS_CONNECTED;
706	}
707
708	switch (sk->sk_state) {
709	case TCP_CLOSE:
710		err = -ENOTCONN;
711		/* Hack to wake up other listeners, who can poll for
712		   POLLHUP, even on eg. unconnected UDP sockets -- RR */
713	default:
714		sk->sk_shutdown |= how;
715		if (sk->sk_prot->shutdown)
716			sk->sk_prot->shutdown(sk, how);
717		break;
718
719	/* Remaining two branches are temporary solution for missing
720	 * close() in multithreaded environment. It is _not_ a good idea,
721	 * but we have no choice until close() is repaired at VFS level.
722	 */
723	case TCP_LISTEN:
724		if (!(how & RCV_SHUTDOWN))
725			break;
726		/* Fall through */
727	case TCP_SYN_SENT:
728		err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
729		sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
730		break;
731	}
732
733	/* Wake up anyone sleeping in poll. */
734	sk->sk_state_change(sk);
735	release_sock(sk);
736	return err;
737}
738
739/*
740 *	ioctl() calls you can issue on an INET socket. Most of these are
741 *	device configuration and stuff and very rarely used. Some ioctls
742 *	pass on to the socket itself.
743 *
744 *	NOTE: I like the idea of a module for the config stuff. ie ifconfig
745 *	loads the devconfigure module does its configuring and unloads it.
746 *	There's a good 20K of config code hanging around the kernel.
747 */
748
749int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
750{
751	struct sock *sk = sock->sk;
752	int err = 0;
753
754	switch (cmd) {
755		case SIOCGSTAMP:
756			err = sock_get_timestamp(sk, (struct timeval __user *)arg);
757			break;
758		case SIOCADDRT:
759		case SIOCDELRT:
760		case SIOCRTMSG:
761			err = ip_rt_ioctl(cmd, (void __user *)arg);
762			break;
763		case SIOCDARP:
764		case SIOCGARP:
765		case SIOCSARP:
766			err = arp_ioctl(cmd, (void __user *)arg);
767			break;
768		case SIOCGIFADDR:
769		case SIOCSIFADDR:
770		case SIOCGIFBRDADDR:
771		case SIOCSIFBRDADDR:
772		case SIOCGIFNETMASK:
773		case SIOCSIFNETMASK:
774		case SIOCGIFDSTADDR:
775		case SIOCSIFDSTADDR:
776		case SIOCSIFPFLAGS:
777		case SIOCGIFPFLAGS:
778		case SIOCSIFFLAGS:
779			err = devinet_ioctl(cmd, (void __user *)arg);
780			break;
781		default:
782			if (sk->sk_prot->ioctl)
783				err = sk->sk_prot->ioctl(sk, cmd, arg);
784			else
785				err = -ENOIOCTLCMD;
786			break;
787	}
788	return err;
789}
790
791const struct proto_ops inet_stream_ops = {
792	.family		   = PF_INET,
793	.owner		   = THIS_MODULE,
794	.release	   = inet_release,
795	.bind		   = inet_bind,
796	.connect	   = inet_stream_connect,
797	.socketpair	   = sock_no_socketpair,
798	.accept		   = inet_accept,
799	.getname	   = inet_getname,
800	.poll		   = tcp_poll,
801	.ioctl		   = inet_ioctl,
802	.listen		   = inet_listen,
803	.shutdown	   = inet_shutdown,
804	.setsockopt	   = sock_common_setsockopt,
805	.getsockopt	   = sock_common_getsockopt,
806	.sendmsg	   = inet_sendmsg,
807	.recvmsg	   = sock_common_recvmsg,
808	.mmap		   = sock_no_mmap,
809	.sendpage	   = tcp_sendpage,
810#ifdef CONFIG_COMPAT
811	.compat_setsockopt = compat_sock_common_setsockopt,
812	.compat_getsockopt = compat_sock_common_getsockopt,
813#endif
814};
815
816const struct proto_ops inet_dgram_ops = {
817	.family		   = PF_INET,
818	.owner		   = THIS_MODULE,
819	.release	   = inet_release,
820	.bind		   = inet_bind,
821	.connect	   = inet_dgram_connect,
822	.socketpair	   = sock_no_socketpair,
823	.accept		   = sock_no_accept,
824	.getname	   = inet_getname,
825	.poll		   = udp_poll,
826	.ioctl		   = inet_ioctl,
827	.listen		   = sock_no_listen,
828	.shutdown	   = inet_shutdown,
829	.setsockopt	   = sock_common_setsockopt,
830	.getsockopt	   = sock_common_getsockopt,
831	.sendmsg	   = inet_sendmsg,
832	.recvmsg	   = sock_common_recvmsg,
833	.mmap		   = sock_no_mmap,
834	.sendpage	   = inet_sendpage,
835#ifdef CONFIG_COMPAT
836	.compat_setsockopt = compat_sock_common_setsockopt,
837	.compat_getsockopt = compat_sock_common_getsockopt,
838#endif
839};
840
841/*
842 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
843 * udp_poll
844 */
845static const struct proto_ops inet_sockraw_ops = {
846	.family		   = PF_INET,
847	.owner		   = THIS_MODULE,
848	.release	   = inet_release,
849	.bind		   = inet_bind,
850	.connect	   = inet_dgram_connect,
851	.socketpair	   = sock_no_socketpair,
852	.accept		   = sock_no_accept,
853	.getname	   = inet_getname,
854	.poll		   = datagram_poll,
855	.ioctl		   = inet_ioctl,
856	.listen		   = sock_no_listen,
857	.shutdown	   = inet_shutdown,
858	.setsockopt	   = sock_common_setsockopt,
859	.getsockopt	   = sock_common_getsockopt,
860	.sendmsg	   = inet_sendmsg,
861	.recvmsg	   = sock_common_recvmsg,
862	.mmap		   = sock_no_mmap,
863	.sendpage	   = inet_sendpage,
864#ifdef CONFIG_COMPAT
865	.compat_setsockopt = compat_sock_common_setsockopt,
866	.compat_getsockopt = compat_sock_common_getsockopt,
867#endif
868};
869
870static struct net_proto_family inet_family_ops = {
871	.family = PF_INET,
872	.create = inet_create,
873	.owner	= THIS_MODULE,
874};
875
876/* Upon startup we insert all the elements in inetsw_array[] into
877 * the linked list inetsw.
878 */
879static struct inet_protosw inetsw_array[] =
880{
881        {
882                .type =       SOCK_STREAM,
883                .protocol =   IPPROTO_TCP,
884                .prot =       &tcp_prot,
885                .ops =        &inet_stream_ops,
886                .capability = -1,
887                .no_check =   0,
888                .flags =      INET_PROTOSW_PERMANENT |
889			      INET_PROTOSW_ICSK,
890        },
891
892        {
893                .type =       SOCK_DGRAM,
894                .protocol =   IPPROTO_UDP,
895                .prot =       &udp_prot,
896                .ops =        &inet_dgram_ops,
897                .capability = -1,
898                .no_check =   UDP_CSUM_DEFAULT,
899                .flags =      INET_PROTOSW_PERMANENT,
900       },
901
902
903       {
904               .type =       SOCK_RAW,
905               .protocol =   IPPROTO_IP,	/* wild card */
906               .prot =       &raw_prot,
907               .ops =        &inet_sockraw_ops,
908               .capability = CAP_NET_RAW,
909               .no_check =   UDP_CSUM_DEFAULT,
910               .flags =      INET_PROTOSW_REUSE,
911       }
912};
913
914#define INETSW_ARRAY_LEN (sizeof(inetsw_array) / sizeof(struct inet_protosw))
915
916void inet_register_protosw(struct inet_protosw *p)
917{
918	struct list_head *lh;
919	struct inet_protosw *answer;
920	int protocol = p->protocol;
921	struct list_head *last_perm;
922
923	spin_lock_bh(&inetsw_lock);
924
925	if (p->type >= SOCK_MAX)
926		goto out_illegal;
927
928	/* If we are trying to override a permanent protocol, bail. */
929	answer = NULL;
930	last_perm = &inetsw[p->type];
931	list_for_each(lh, &inetsw[p->type]) {
932		answer = list_entry(lh, struct inet_protosw, list);
933
934		/* Check only the non-wild match. */
935		if (INET_PROTOSW_PERMANENT & answer->flags) {
936			if (protocol == answer->protocol)
937				break;
938			last_perm = lh;
939		}
940
941		answer = NULL;
942	}
943	if (answer)
944		goto out_permanent;
945
946	/* Add the new entry after the last permanent entry if any, so that
947	 * the new entry does not override a permanent entry when matched with
948	 * a wild-card protocol. But it is allowed to override any existing
949	 * non-permanent entry.  This means that when we remove this entry, the
950	 * system automatically returns to the old behavior.
951	 */
952	list_add_rcu(&p->list, last_perm);
953out:
954	spin_unlock_bh(&inetsw_lock);
955
956	synchronize_net();
957
958	return;
959
960out_permanent:
961	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
962	       protocol);
963	goto out;
964
965out_illegal:
966	printk(KERN_ERR
967	       "Ignoring attempt to register invalid socket type %d.\n",
968	       p->type);
969	goto out;
970}
971
972void inet_unregister_protosw(struct inet_protosw *p)
973{
974	if (INET_PROTOSW_PERMANENT & p->flags) {
975		printk(KERN_ERR
976		       "Attempt to unregister permanent protocol %d.\n",
977		       p->protocol);
978	} else {
979		spin_lock_bh(&inetsw_lock);
980		list_del_rcu(&p->list);
981		spin_unlock_bh(&inetsw_lock);
982
983		synchronize_net();
984	}
985}
986
987/*
988 *      Shall we try to damage output packets if routing dev changes?
989 */
990
991int sysctl_ip_dynaddr;
992
993static int inet_sk_reselect_saddr(struct sock *sk)
994{
995	struct inet_sock *inet = inet_sk(sk);
996	int err;
997	struct rtable *rt;
998	__u32 old_saddr = inet->saddr;
999	__u32 new_saddr;
1000	__u32 daddr = inet->daddr;
1001
1002	if (inet->opt && inet->opt->srr)
1003		daddr = inet->opt->faddr;
1004
1005	/* Query new route. */
1006	err = ip_route_connect(&rt, daddr, 0,
1007			       RT_CONN_FLAGS(sk),
1008			       sk->sk_bound_dev_if,
1009			       sk->sk_protocol,
1010			       inet->sport, inet->dport, sk);
1011	if (err)
1012		return err;
1013
1014	sk_setup_caps(sk, &rt->u.dst);
1015
1016	new_saddr = rt->rt_src;
1017
1018	if (new_saddr == old_saddr)
1019		return 0;
1020
1021	if (sysctl_ip_dynaddr > 1) {
1022		printk(KERN_INFO "%s(): shifting inet->"
1023				 "saddr from %d.%d.%d.%d to %d.%d.%d.%d\n",
1024		       __FUNCTION__,
1025		       NIPQUAD(old_saddr),
1026		       NIPQUAD(new_saddr));
1027	}
1028
1029	inet->saddr = inet->rcv_saddr = new_saddr;
1030
1031	/*
1032	 * XXX The only one ugly spot where we need to
1033	 * XXX really change the sockets identity after
1034	 * XXX it has entered the hashes. -DaveM
1035	 *
1036	 * Besides that, it does not check for connection
1037	 * uniqueness. Wait for troubles.
1038	 */
1039	__sk_prot_rehash(sk);
1040	return 0;
1041}
1042
1043int inet_sk_rebuild_header(struct sock *sk)
1044{
1045	struct inet_sock *inet = inet_sk(sk);
1046	struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
1047	u32 daddr;
1048	int err;
1049
1050	/* Route is OK, nothing to do. */
1051	if (rt)
1052		return 0;
1053
1054	/* Reroute. */
1055	daddr = inet->daddr;
1056	if (inet->opt && inet->opt->srr)
1057		daddr = inet->opt->faddr;
1058{
1059	struct flowi fl = {
1060		.oif = sk->sk_bound_dev_if,
1061		.nl_u = {
1062			.ip4_u = {
1063				.daddr	= daddr,
1064				.saddr	= inet->saddr,
1065				.tos	= RT_CONN_FLAGS(sk),
1066			},
1067		},
1068		.proto = sk->sk_protocol,
1069		.uli_u = {
1070			.ports = {
1071				.sport = inet->sport,
1072				.dport = inet->dport,
1073			},
1074		},
1075	};
1076
1077	err = ip_route_output_flow(&rt, &fl, sk, 0);
1078}
1079	if (!err)
1080		sk_setup_caps(sk, &rt->u.dst);
1081	else {
1082		/* Routing failed... */
1083		sk->sk_route_caps = 0;
1084		/*
1085		 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1086		 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1087		 */
1088		if (!sysctl_ip_dynaddr ||
1089		    sk->sk_state != TCP_SYN_SENT ||
1090		    (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1091		    (err = inet_sk_reselect_saddr(sk)) != 0)
1092			sk->sk_err_soft = -err;
1093	}
1094
1095	return err;
1096}
1097
1098EXPORT_SYMBOL(inet_sk_rebuild_header);
1099
1100static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
1101{
1102	struct sk_buff *segs = ERR_PTR(-EINVAL);
1103	struct iphdr *iph;
1104	struct net_protocol *ops;
1105	int proto;
1106	int ihl;
1107	int id;
1108
1109	if (unlikely(skb_shinfo(skb)->gso_type &
1110		     ~(SKB_GSO_TCPV4 |
1111		       SKB_GSO_UDP |
1112		       SKB_GSO_DODGY |
1113		       SKB_GSO_TCP_ECN |
1114		       0)))
1115		goto out;
1116
1117	if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1118		goto out;
1119
1120	iph = skb->nh.iph;
1121	ihl = iph->ihl * 4;
1122	if (ihl < sizeof(*iph))
1123		goto out;
1124
1125	if (unlikely(!pskb_may_pull(skb, ihl)))
1126		goto out;
1127
1128	skb->h.raw = __skb_pull(skb, ihl);
1129	iph = skb->nh.iph;
1130	id = ntohs(iph->id);
1131	proto = iph->protocol & (MAX_INET_PROTOS - 1);
1132	segs = ERR_PTR(-EPROTONOSUPPORT);
1133
1134	rcu_read_lock();
1135	ops = rcu_dereference(inet_protos[proto]);
1136	if (likely(ops && ops->gso_segment))
1137		segs = ops->gso_segment(skb, features);
1138	rcu_read_unlock();
1139
1140	if (!segs || unlikely(IS_ERR(segs)))
1141		goto out;
1142
1143	skb = segs;
1144	do {
1145		iph = skb->nh.iph;
1146		iph->id = htons(id++);
1147		iph->tot_len = htons(skb->len - skb->mac_len);
1148		iph->check = 0;
1149		iph->check = ip_fast_csum(skb->nh.raw, iph->ihl);
1150	} while ((skb = skb->next));
1151
1152out:
1153	return segs;
1154}
1155
1156#ifdef CONFIG_IP_MULTICAST
1157static struct net_protocol igmp_protocol = {
1158	.handler =	igmp_rcv,
1159};
1160#endif
1161
1162static struct net_protocol tcp_protocol = {
1163	.handler =	tcp_v4_rcv,
1164	.err_handler =	tcp_v4_err,
1165	.gso_segment =	tcp_tso_segment,
1166	.no_policy =	1,
1167};
1168
1169static struct net_protocol udp_protocol = {
1170	.handler =	udp_rcv,
1171	.err_handler =	udp_err,
1172	.no_policy =	1,
1173};
1174
1175static struct net_protocol icmp_protocol = {
1176	.handler =	icmp_rcv,
1177};
1178
1179static int __init init_ipv4_mibs(void)
1180{
1181	net_statistics[0] = alloc_percpu(struct linux_mib);
1182	net_statistics[1] = alloc_percpu(struct linux_mib);
1183	ip_statistics[0] = alloc_percpu(struct ipstats_mib);
1184	ip_statistics[1] = alloc_percpu(struct ipstats_mib);
1185	icmp_statistics[0] = alloc_percpu(struct icmp_mib);
1186	icmp_statistics[1] = alloc_percpu(struct icmp_mib);
1187	tcp_statistics[0] = alloc_percpu(struct tcp_mib);
1188	tcp_statistics[1] = alloc_percpu(struct tcp_mib);
1189	udp_statistics[0] = alloc_percpu(struct udp_mib);
1190	udp_statistics[1] = alloc_percpu(struct udp_mib);
1191	if (!
1192	    (net_statistics[0] && net_statistics[1] && ip_statistics[0]
1193	     && ip_statistics[1] && tcp_statistics[0] && tcp_statistics[1]
1194	     && udp_statistics[0] && udp_statistics[1]))
1195		return -ENOMEM;
1196
1197	(void) tcp_mib_init();
1198
1199	return 0;
1200}
1201
1202static int ipv4_proc_init(void);
1203
1204/*
1205 *	IP protocol layer initialiser
1206 */
1207
1208static struct packet_type ip_packet_type = {
1209	.type = __constant_htons(ETH_P_IP),
1210	.func = ip_rcv,
1211	.gso_segment = inet_gso_segment,
1212};
1213
1214static int __init inet_init(void)
1215{
1216	struct sk_buff *dummy_skb;
1217	struct inet_protosw *q;
1218	struct list_head *r;
1219	int rc = -EINVAL;
1220
1221	if (sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb)) {
1222		printk(KERN_CRIT "%s: panic\n", __FUNCTION__);
1223		goto out;
1224	}
1225
1226	rc = proto_register(&tcp_prot, 1);
1227	if (rc)
1228		goto out;
1229
1230	rc = proto_register(&udp_prot, 1);
1231	if (rc)
1232		goto out_unregister_tcp_proto;
1233
1234	rc = proto_register(&raw_prot, 1);
1235	if (rc)
1236		goto out_unregister_udp_proto;
1237
1238	/*
1239	 *	Tell SOCKET that we are alive...
1240	 */
1241
1242  	(void)sock_register(&inet_family_ops);
1243
1244	/*
1245	 *	Add all the base protocols.
1246	 */
1247
1248	if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1249		printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1250	if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1251		printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1252	if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1253		printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1254#ifdef CONFIG_IP_MULTICAST
1255	if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1256		printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1257#endif
1258
1259	/* Register the socket-side information for inet_create. */
1260	for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1261		INIT_LIST_HEAD(r);
1262
1263	for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1264		inet_register_protosw(q);
1265
1266	/*
1267	 *	Set the ARP module up
1268	 */
1269
1270	arp_init();
1271
1272  	/*
1273  	 *	Set the IP module up
1274  	 */
1275
1276	ip_init();
1277
1278	tcp_v4_init(&inet_family_ops);
1279
1280	/* Setup TCP slab cache for open requests. */
1281	tcp_init();
1282
1283
1284	/*
1285	 *	Set the ICMP layer up
1286	 */
1287
1288	icmp_init(&inet_family_ops);
1289
1290	/*
1291	 *	Initialise the multicast router
1292	 */
1293#if defined(CONFIG_IP_MROUTE)
1294	ip_mr_init();
1295#endif
1296	/*
1297	 *	Initialise per-cpu ipv4 mibs
1298	 */
1299
1300	if(init_ipv4_mibs())
1301		printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
1302
1303	ipv4_proc_init();
1304
1305	ipfrag_init();
1306
1307	dev_add_pack(&ip_packet_type);
1308
1309	rc = 0;
1310out:
1311	return rc;
1312out_unregister_tcp_proto:
1313	proto_unregister(&tcp_prot);
1314out_unregister_udp_proto:
1315	proto_unregister(&udp_prot);
1316	goto out;
1317}
1318
1319fs_initcall(inet_init);
1320
1321/* ------------------------------------------------------------------------ */
1322
1323#ifdef CONFIG_PROC_FS
1324static int __init ipv4_proc_init(void)
1325{
1326	int rc = 0;
1327
1328	if (raw_proc_init())
1329		goto out_raw;
1330	if (tcp4_proc_init())
1331		goto out_tcp;
1332	if (udp4_proc_init())
1333		goto out_udp;
1334	if (fib_proc_init())
1335		goto out_fib;
1336	if (ip_misc_proc_init())
1337		goto out_misc;
1338out:
1339	return rc;
1340out_misc:
1341	fib_proc_exit();
1342out_fib:
1343	udp4_proc_exit();
1344out_udp:
1345	tcp4_proc_exit();
1346out_tcp:
1347	raw_proc_exit();
1348out_raw:
1349	rc = -ENOMEM;
1350	goto out;
1351}
1352
1353#else /* CONFIG_PROC_FS */
1354static int __init ipv4_proc_init(void)
1355{
1356	return 0;
1357}
1358#endif /* CONFIG_PROC_FS */
1359
1360MODULE_ALIAS_NETPROTO(PF_INET);
1361
1362EXPORT_SYMBOL(inet_accept);
1363EXPORT_SYMBOL(inet_bind);
1364EXPORT_SYMBOL(inet_dgram_connect);
1365EXPORT_SYMBOL(inet_dgram_ops);
1366EXPORT_SYMBOL(inet_getname);
1367EXPORT_SYMBOL(inet_ioctl);
1368EXPORT_SYMBOL(inet_listen);
1369EXPORT_SYMBOL(inet_register_protosw);
1370EXPORT_SYMBOL(inet_release);
1371EXPORT_SYMBOL(inet_sendmsg);
1372EXPORT_SYMBOL(inet_shutdown);
1373EXPORT_SYMBOL(inet_sock_destruct);
1374EXPORT_SYMBOL(inet_stream_connect);
1375EXPORT_SYMBOL(inet_stream_ops);
1376EXPORT_SYMBOL(inet_unregister_protosw);
1377EXPORT_SYMBOL(net_statistics);
1378EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
1379