af_inet.c revision d56f90a7c96da5187f0cdf07ee7434fe6aa78bbc 
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/err.h>
71#include <linux/errno.h>
72#include <linux/types.h>
73#include <linux/socket.h>
74#include <linux/in.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/capability.h>
83#include <linux/fcntl.h>
84#include <linux/mm.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>
90
91#include <asm/uaccess.h>
92#include <asm/system.h>
93
94#include <linux/smp_lock.h>
95#include <linux/inet.h>
96#include <linux/igmp.h>
97#include <linux/inetdevice.h>
98#include <linux/netdevice.h>
99#include <net/ip.h>
100#include <net/protocol.h>
101#include <net/arp.h>
102#include <net/route.h>
103#include <net/ip_fib.h>
104#include <net/inet_connection_sock.h>
105#include <net/tcp.h>
106#include <net/udp.h>
107#include <net/udplite.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, backlog);
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) != 0;
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 __read_mostly;
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		__be32 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 SIOCGSTAMPNS:
759			err = sock_get_timestampns(sk, (struct timespec __user *)arg);
760			break;
761		case SIOCADDRT:
762		case SIOCDELRT:
763		case SIOCRTMSG:
764			err = ip_rt_ioctl(cmd, (void __user *)arg);
765			break;
766		case SIOCDARP:
767		case SIOCGARP:
768		case SIOCSARP:
769			err = arp_ioctl(cmd, (void __user *)arg);
770			break;
771		case SIOCGIFADDR:
772		case SIOCSIFADDR:
773		case SIOCGIFBRDADDR:
774		case SIOCSIFBRDADDR:
775		case SIOCGIFNETMASK:
776		case SIOCSIFNETMASK:
777		case SIOCGIFDSTADDR:
778		case SIOCSIFDSTADDR:
779		case SIOCSIFPFLAGS:
780		case SIOCGIFPFLAGS:
781		case SIOCSIFFLAGS:
782			err = devinet_ioctl(cmd, (void __user *)arg);
783			break;
784		default:
785			if (sk->sk_prot->ioctl)
786				err = sk->sk_prot->ioctl(sk, cmd, arg);
787			else
788				err = -ENOIOCTLCMD;
789			break;
790	}
791	return err;
792}
793
794const struct proto_ops inet_stream_ops = {
795	.family		   = PF_INET,
796	.owner		   = THIS_MODULE,
797	.release	   = inet_release,
798	.bind		   = inet_bind,
799	.connect	   = inet_stream_connect,
800	.socketpair	   = sock_no_socketpair,
801	.accept		   = inet_accept,
802	.getname	   = inet_getname,
803	.poll		   = tcp_poll,
804	.ioctl		   = inet_ioctl,
805	.listen		   = inet_listen,
806	.shutdown	   = inet_shutdown,
807	.setsockopt	   = sock_common_setsockopt,
808	.getsockopt	   = sock_common_getsockopt,
809	.sendmsg	   = inet_sendmsg,
810	.recvmsg	   = sock_common_recvmsg,
811	.mmap		   = sock_no_mmap,
812	.sendpage	   = tcp_sendpage,
813#ifdef CONFIG_COMPAT
814	.compat_setsockopt = compat_sock_common_setsockopt,
815	.compat_getsockopt = compat_sock_common_getsockopt,
816#endif
817};
818
819const struct proto_ops inet_dgram_ops = {
820	.family		   = PF_INET,
821	.owner		   = THIS_MODULE,
822	.release	   = inet_release,
823	.bind		   = inet_bind,
824	.connect	   = inet_dgram_connect,
825	.socketpair	   = sock_no_socketpair,
826	.accept		   = sock_no_accept,
827	.getname	   = inet_getname,
828	.poll		   = udp_poll,
829	.ioctl		   = inet_ioctl,
830	.listen		   = sock_no_listen,
831	.shutdown	   = inet_shutdown,
832	.setsockopt	   = sock_common_setsockopt,
833	.getsockopt	   = sock_common_getsockopt,
834	.sendmsg	   = inet_sendmsg,
835	.recvmsg	   = sock_common_recvmsg,
836	.mmap		   = sock_no_mmap,
837	.sendpage	   = inet_sendpage,
838#ifdef CONFIG_COMPAT
839	.compat_setsockopt = compat_sock_common_setsockopt,
840	.compat_getsockopt = compat_sock_common_getsockopt,
841#endif
842};
843
844/*
845 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
846 * udp_poll
847 */
848static const struct proto_ops inet_sockraw_ops = {
849	.family		   = PF_INET,
850	.owner		   = THIS_MODULE,
851	.release	   = inet_release,
852	.bind		   = inet_bind,
853	.connect	   = inet_dgram_connect,
854	.socketpair	   = sock_no_socketpair,
855	.accept		   = sock_no_accept,
856	.getname	   = inet_getname,
857	.poll		   = datagram_poll,
858	.ioctl		   = inet_ioctl,
859	.listen		   = sock_no_listen,
860	.shutdown	   = inet_shutdown,
861	.setsockopt	   = sock_common_setsockopt,
862	.getsockopt	   = sock_common_getsockopt,
863	.sendmsg	   = inet_sendmsg,
864	.recvmsg	   = sock_common_recvmsg,
865	.mmap		   = sock_no_mmap,
866	.sendpage	   = inet_sendpage,
867#ifdef CONFIG_COMPAT
868	.compat_setsockopt = compat_sock_common_setsockopt,
869	.compat_getsockopt = compat_sock_common_getsockopt,
870#endif
871};
872
873static struct net_proto_family inet_family_ops = {
874	.family = PF_INET,
875	.create = inet_create,
876	.owner	= THIS_MODULE,
877};
878
879/* Upon startup we insert all the elements in inetsw_array[] into
880 * the linked list inetsw.
881 */
882static struct inet_protosw inetsw_array[] =
883{
884	{
885		.type =       SOCK_STREAM,
886		.protocol =   IPPROTO_TCP,
887		.prot =       &tcp_prot,
888		.ops =        &inet_stream_ops,
889		.capability = -1,
890		.no_check =   0,
891		.flags =      INET_PROTOSW_PERMANENT |
892			      INET_PROTOSW_ICSK,
893	},
894
895	{
896		.type =       SOCK_DGRAM,
897		.protocol =   IPPROTO_UDP,
898		.prot =       &udp_prot,
899		.ops =        &inet_dgram_ops,
900		.capability = -1,
901		.no_check =   UDP_CSUM_DEFAULT,
902		.flags =      INET_PROTOSW_PERMANENT,
903       },
904
905
906       {
907	       .type =       SOCK_RAW,
908	       .protocol =   IPPROTO_IP,	/* wild card */
909	       .prot =       &raw_prot,
910	       .ops =        &inet_sockraw_ops,
911	       .capability = CAP_NET_RAW,
912	       .no_check =   UDP_CSUM_DEFAULT,
913	       .flags =      INET_PROTOSW_REUSE,
914       }
915};
916
917#define INETSW_ARRAY_LEN (sizeof(inetsw_array) / sizeof(struct inet_protosw))
918
919void inet_register_protosw(struct inet_protosw *p)
920{
921	struct list_head *lh;
922	struct inet_protosw *answer;
923	int protocol = p->protocol;
924	struct list_head *last_perm;
925
926	spin_lock_bh(&inetsw_lock);
927
928	if (p->type >= SOCK_MAX)
929		goto out_illegal;
930
931	/* If we are trying to override a permanent protocol, bail. */
932	answer = NULL;
933	last_perm = &inetsw[p->type];
934	list_for_each(lh, &inetsw[p->type]) {
935		answer = list_entry(lh, struct inet_protosw, list);
936
937		/* Check only the non-wild match. */
938		if (INET_PROTOSW_PERMANENT & answer->flags) {
939			if (protocol == answer->protocol)
940				break;
941			last_perm = lh;
942		}
943
944		answer = NULL;
945	}
946	if (answer)
947		goto out_permanent;
948
949	/* Add the new entry after the last permanent entry if any, so that
950	 * the new entry does not override a permanent entry when matched with
951	 * a wild-card protocol. But it is allowed to override any existing
952	 * non-permanent entry.  This means that when we remove this entry, the
953	 * system automatically returns to the old behavior.
954	 */
955	list_add_rcu(&p->list, last_perm);
956out:
957	spin_unlock_bh(&inetsw_lock);
958
959	synchronize_net();
960
961	return;
962
963out_permanent:
964	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
965	       protocol);
966	goto out;
967
968out_illegal:
969	printk(KERN_ERR
970	       "Ignoring attempt to register invalid socket type %d.\n",
971	       p->type);
972	goto out;
973}
974
975void inet_unregister_protosw(struct inet_protosw *p)
976{
977	if (INET_PROTOSW_PERMANENT & p->flags) {
978		printk(KERN_ERR
979		       "Attempt to unregister permanent protocol %d.\n",
980		       p->protocol);
981	} else {
982		spin_lock_bh(&inetsw_lock);
983		list_del_rcu(&p->list);
984		spin_unlock_bh(&inetsw_lock);
985
986		synchronize_net();
987	}
988}
989
990/*
991 *      Shall we try to damage output packets if routing dev changes?
992 */
993
994int sysctl_ip_dynaddr __read_mostly;
995
996static int inet_sk_reselect_saddr(struct sock *sk)
997{
998	struct inet_sock *inet = inet_sk(sk);
999	int err;
1000	struct rtable *rt;
1001	__be32 old_saddr = inet->saddr;
1002	__be32 new_saddr;
1003	__be32 daddr = inet->daddr;
1004
1005	if (inet->opt && inet->opt->srr)
1006		daddr = inet->opt->faddr;
1007
1008	/* Query new route. */
1009	err = ip_route_connect(&rt, daddr, 0,
1010			       RT_CONN_FLAGS(sk),
1011			       sk->sk_bound_dev_if,
1012			       sk->sk_protocol,
1013			       inet->sport, inet->dport, sk, 0);
1014	if (err)
1015		return err;
1016
1017	sk_setup_caps(sk, &rt->u.dst);
1018
1019	new_saddr = rt->rt_src;
1020
1021	if (new_saddr == old_saddr)
1022		return 0;
1023
1024	if (sysctl_ip_dynaddr > 1) {
1025		printk(KERN_INFO "%s(): shifting inet->"
1026				 "saddr from %d.%d.%d.%d to %d.%d.%d.%d\n",
1027		       __FUNCTION__,
1028		       NIPQUAD(old_saddr),
1029		       NIPQUAD(new_saddr));
1030	}
1031
1032	inet->saddr = inet->rcv_saddr = new_saddr;
1033
1034	/*
1035	 * XXX The only one ugly spot where we need to
1036	 * XXX really change the sockets identity after
1037	 * XXX it has entered the hashes. -DaveM
1038	 *
1039	 * Besides that, it does not check for connection
1040	 * uniqueness. Wait for troubles.
1041	 */
1042	__sk_prot_rehash(sk);
1043	return 0;
1044}
1045
1046int inet_sk_rebuild_header(struct sock *sk)
1047{
1048	struct inet_sock *inet = inet_sk(sk);
1049	struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
1050	__be32 daddr;
1051	int err;
1052
1053	/* Route is OK, nothing to do. */
1054	if (rt)
1055		return 0;
1056
1057	/* Reroute. */
1058	daddr = inet->daddr;
1059	if (inet->opt && inet->opt->srr)
1060		daddr = inet->opt->faddr;
1061{
1062	struct flowi fl = {
1063		.oif = sk->sk_bound_dev_if,
1064		.nl_u = {
1065			.ip4_u = {
1066				.daddr	= daddr,
1067				.saddr	= inet->saddr,
1068				.tos	= RT_CONN_FLAGS(sk),
1069			},
1070		},
1071		.proto = sk->sk_protocol,
1072		.uli_u = {
1073			.ports = {
1074				.sport = inet->sport,
1075				.dport = inet->dport,
1076			},
1077		},
1078	};
1079
1080	security_sk_classify_flow(sk, &fl);
1081	err = ip_route_output_flow(&rt, &fl, sk, 0);
1082}
1083	if (!err)
1084		sk_setup_caps(sk, &rt->u.dst);
1085	else {
1086		/* Routing failed... */
1087		sk->sk_route_caps = 0;
1088		/*
1089		 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1090		 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1091		 */
1092		if (!sysctl_ip_dynaddr ||
1093		    sk->sk_state != TCP_SYN_SENT ||
1094		    (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1095		    (err = inet_sk_reselect_saddr(sk)) != 0)
1096			sk->sk_err_soft = -err;
1097	}
1098
1099	return err;
1100}
1101
1102EXPORT_SYMBOL(inet_sk_rebuild_header);
1103
1104static int inet_gso_send_check(struct sk_buff *skb)
1105{
1106	struct iphdr *iph;
1107	struct net_protocol *ops;
1108	int proto;
1109	int ihl;
1110	int err = -EINVAL;
1111
1112	if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1113		goto out;
1114
1115	iph = skb->nh.iph;
1116	ihl = iph->ihl * 4;
1117	if (ihl < sizeof(*iph))
1118		goto out;
1119
1120	if (unlikely(!pskb_may_pull(skb, ihl)))
1121		goto out;
1122
1123	skb->h.raw = __skb_pull(skb, ihl);
1124	iph = skb->nh.iph;
1125	proto = iph->protocol & (MAX_INET_PROTOS - 1);
1126	err = -EPROTONOSUPPORT;
1127
1128	rcu_read_lock();
1129	ops = rcu_dereference(inet_protos[proto]);
1130	if (likely(ops && ops->gso_send_check))
1131		err = ops->gso_send_check(skb);
1132	rcu_read_unlock();
1133
1134out:
1135	return err;
1136}
1137
1138static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
1139{
1140	struct sk_buff *segs = ERR_PTR(-EINVAL);
1141	struct iphdr *iph;
1142	struct net_protocol *ops;
1143	int proto;
1144	int ihl;
1145	int id;
1146
1147	if (unlikely(skb_shinfo(skb)->gso_type &
1148		     ~(SKB_GSO_TCPV4 |
1149		       SKB_GSO_UDP |
1150		       SKB_GSO_DODGY |
1151		       SKB_GSO_TCP_ECN |
1152		       0)))
1153		goto out;
1154
1155	if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1156		goto out;
1157
1158	iph = skb->nh.iph;
1159	ihl = iph->ihl * 4;
1160	if (ihl < sizeof(*iph))
1161		goto out;
1162
1163	if (unlikely(!pskb_may_pull(skb, ihl)))
1164		goto out;
1165
1166	skb->h.raw = __skb_pull(skb, ihl);
1167	iph = skb->nh.iph;
1168	id = ntohs(iph->id);
1169	proto = iph->protocol & (MAX_INET_PROTOS - 1);
1170	segs = ERR_PTR(-EPROTONOSUPPORT);
1171
1172	rcu_read_lock();
1173	ops = rcu_dereference(inet_protos[proto]);
1174	if (likely(ops && ops->gso_segment))
1175		segs = ops->gso_segment(skb, features);
1176	rcu_read_unlock();
1177
1178	if (!segs || unlikely(IS_ERR(segs)))
1179		goto out;
1180
1181	skb = segs;
1182	do {
1183		iph = skb->nh.iph;
1184		iph->id = htons(id++);
1185		iph->tot_len = htons(skb->len - skb->mac_len);
1186		iph->check = 0;
1187		iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
1188	} while ((skb = skb->next));
1189
1190out:
1191	return segs;
1192}
1193
1194#ifdef CONFIG_IP_MULTICAST
1195static struct net_protocol igmp_protocol = {
1196	.handler =	igmp_rcv,
1197};
1198#endif
1199
1200static struct net_protocol tcp_protocol = {
1201	.handler =	tcp_v4_rcv,
1202	.err_handler =	tcp_v4_err,
1203	.gso_send_check = tcp_v4_gso_send_check,
1204	.gso_segment =	tcp_tso_segment,
1205	.no_policy =	1,
1206};
1207
1208static struct net_protocol udp_protocol = {
1209	.handler =	udp_rcv,
1210	.err_handler =	udp_err,
1211	.no_policy =	1,
1212};
1213
1214static struct net_protocol icmp_protocol = {
1215	.handler =	icmp_rcv,
1216};
1217
1218static int __init init_ipv4_mibs(void)
1219{
1220	net_statistics[0] = alloc_percpu(struct linux_mib);
1221	net_statistics[1] = alloc_percpu(struct linux_mib);
1222	ip_statistics[0] = alloc_percpu(struct ipstats_mib);
1223	ip_statistics[1] = alloc_percpu(struct ipstats_mib);
1224	icmp_statistics[0] = alloc_percpu(struct icmp_mib);
1225	icmp_statistics[1] = alloc_percpu(struct icmp_mib);
1226	tcp_statistics[0] = alloc_percpu(struct tcp_mib);
1227	tcp_statistics[1] = alloc_percpu(struct tcp_mib);
1228	udp_statistics[0] = alloc_percpu(struct udp_mib);
1229	udp_statistics[1] = alloc_percpu(struct udp_mib);
1230	udplite_statistics[0] = alloc_percpu(struct udp_mib);
1231	udplite_statistics[1] = alloc_percpu(struct udp_mib);
1232	if (!
1233	    (net_statistics[0] && net_statistics[1] && ip_statistics[0]
1234	     && ip_statistics[1] && tcp_statistics[0] && tcp_statistics[1]
1235	     && udp_statistics[0] && udp_statistics[1]
1236	     && udplite_statistics[0] && udplite_statistics[1]             ) )
1237		return -ENOMEM;
1238
1239	(void) tcp_mib_init();
1240
1241	return 0;
1242}
1243
1244static int ipv4_proc_init(void);
1245
1246/*
1247 *	IP protocol layer initialiser
1248 */
1249
1250static struct packet_type ip_packet_type = {
1251	.type = __constant_htons(ETH_P_IP),
1252	.func = ip_rcv,
1253	.gso_send_check = inet_gso_send_check,
1254	.gso_segment = inet_gso_segment,
1255};
1256
1257static int __init inet_init(void)
1258{
1259	struct sk_buff *dummy_skb;
1260	struct inet_protosw *q;
1261	struct list_head *r;
1262	int rc = -EINVAL;
1263
1264	BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
1265
1266	rc = proto_register(&tcp_prot, 1);
1267	if (rc)
1268		goto out;
1269
1270	rc = proto_register(&udp_prot, 1);
1271	if (rc)
1272		goto out_unregister_tcp_proto;
1273
1274	rc = proto_register(&raw_prot, 1);
1275	if (rc)
1276		goto out_unregister_udp_proto;
1277
1278	/*
1279	 *	Tell SOCKET that we are alive...
1280	 */
1281
1282	(void)sock_register(&inet_family_ops);
1283
1284	/*
1285	 *	Add all the base protocols.
1286	 */
1287
1288	if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1289		printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1290	if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1291		printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1292	if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1293		printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1294#ifdef CONFIG_IP_MULTICAST
1295	if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1296		printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1297#endif
1298
1299	/* Register the socket-side information for inet_create. */
1300	for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1301		INIT_LIST_HEAD(r);
1302
1303	for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1304		inet_register_protosw(q);
1305
1306	/*
1307	 *	Set the ARP module up
1308	 */
1309
1310	arp_init();
1311
1312	/*
1313	 *	Set the IP module up
1314	 */
1315
1316	ip_init();
1317
1318	tcp_v4_init(&inet_family_ops);
1319
1320	/* Setup TCP slab cache for open requests. */
1321	tcp_init();
1322
1323	/* Add UDP-Lite (RFC 3828) */
1324	udplite4_register();
1325
1326	/*
1327	 *	Set the ICMP layer up
1328	 */
1329
1330	icmp_init(&inet_family_ops);
1331
1332	/*
1333	 *	Initialise the multicast router
1334	 */
1335#if defined(CONFIG_IP_MROUTE)
1336	ip_mr_init();
1337#endif
1338	/*
1339	 *	Initialise per-cpu ipv4 mibs
1340	 */
1341
1342	if (init_ipv4_mibs())
1343		printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
1344
1345	ipv4_proc_init();
1346
1347	ipfrag_init();
1348
1349	dev_add_pack(&ip_packet_type);
1350
1351	rc = 0;
1352out:
1353	return rc;
1354out_unregister_udp_proto:
1355	proto_unregister(&udp_prot);
1356out_unregister_tcp_proto:
1357	proto_unregister(&tcp_prot);
1358	goto out;
1359}
1360
1361fs_initcall(inet_init);
1362
1363/* ------------------------------------------------------------------------ */
1364
1365#ifdef CONFIG_PROC_FS
1366static int __init ipv4_proc_init(void)
1367{
1368	int rc = 0;
1369
1370	if (raw_proc_init())
1371		goto out_raw;
1372	if (tcp4_proc_init())
1373		goto out_tcp;
1374	if (udp4_proc_init())
1375		goto out_udp;
1376	if (fib_proc_init())
1377		goto out_fib;
1378	if (ip_misc_proc_init())
1379		goto out_misc;
1380out:
1381	return rc;
1382out_misc:
1383	fib_proc_exit();
1384out_fib:
1385	udp4_proc_exit();
1386out_udp:
1387	tcp4_proc_exit();
1388out_tcp:
1389	raw_proc_exit();
1390out_raw:
1391	rc = -ENOMEM;
1392	goto out;
1393}
1394
1395#else /* CONFIG_PROC_FS */
1396static int __init ipv4_proc_init(void)
1397{
1398	return 0;
1399}
1400#endif /* CONFIG_PROC_FS */
1401
1402MODULE_ALIAS_NETPROTO(PF_INET);
1403
1404EXPORT_SYMBOL(inet_accept);
1405EXPORT_SYMBOL(inet_bind);
1406EXPORT_SYMBOL(inet_dgram_connect);
1407EXPORT_SYMBOL(inet_dgram_ops);
1408EXPORT_SYMBOL(inet_getname);
1409EXPORT_SYMBOL(inet_ioctl);
1410EXPORT_SYMBOL(inet_listen);
1411EXPORT_SYMBOL(inet_register_protosw);
1412EXPORT_SYMBOL(inet_release);
1413EXPORT_SYMBOL(inet_sendmsg);
1414EXPORT_SYMBOL(inet_shutdown);
1415EXPORT_SYMBOL(inet_sock_destruct);
1416EXPORT_SYMBOL(inet_stream_connect);
1417EXPORT_SYMBOL(inet_stream_ops);
1418EXPORT_SYMBOL(inet_unregister_protosw);
1419EXPORT_SYMBOL(net_statistics);
1420EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
1421