blob: dc9dff396fad7599252dd3bbff503dbdffbc7a38 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* SCTP kernel reference Implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel reference Implementation
10 *
11 * Initialization/cleanup for SCTP protocol support.
12 *
13 * The SCTP reference implementation is free software;
14 * you can redistribute it and/or modify it under the terms of
15 * the GNU General Public License as published by
16 * the Free Software Foundation; either version 2, or (at your option)
17 * any later version.
18 *
19 * The SCTP reference implementation is distributed in the hope that it
20 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
21 * ************************
22 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
23 * See the GNU General Public License for more details.
24 *
25 * You should have received a copy of the GNU General Public License
26 * along with GNU CC; see the file COPYING. If not, write to
27 * the Free Software Foundation, 59 Temple Place - Suite 330,
28 * Boston, MA 02111-1307, USA.
29 *
30 * Please send any bug reports or fixes you make to the
31 * email address(es):
32 * lksctp developers <lksctp-developers@lists.sourceforge.net>
33 *
34 * Or submit a bug report through the following website:
35 * http://www.sf.net/projects/lksctp
36 *
37 * Written or modified by:
38 * La Monte H.P. Yarroll <piggy@acm.org>
39 * Karl Knutson <karl@athena.chicago.il.us>
40 * Jon Grimm <jgrimm@us.ibm.com>
41 * Sridhar Samudrala <sri@us.ibm.com>
42 * Daisy Chang <daisyc@us.ibm.com>
43 * Ardelle Fan <ardelle.fan@intel.com>
44 *
45 * Any bugs reported given to us we will try to fix... any fixes shared will
46 * be incorporated into the next SCTP release.
47 */
48
49#include <linux/module.h>
50#include <linux/init.h>
51#include <linux/netdevice.h>
52#include <linux/inetdevice.h>
53#include <linux/seq_file.h>
54#include <net/protocol.h>
55#include <net/ip.h>
56#include <net/ipv6.h>
57#include <net/sctp/sctp.h>
58#include <net/addrconf.h>
59#include <net/inet_common.h>
60#include <net/inet_ecn.h>
61
62/* Global data structures. */
63struct sctp_globals sctp_globals;
64struct proc_dir_entry *proc_net_sctp;
Eric Dumazetba899662005-08-26 12:05:31 -070065DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
67struct idr sctp_assocs_id;
68DEFINE_SPINLOCK(sctp_assocs_id_lock);
69
70/* This is the global socket data structure used for responding to
71 * the Out-of-the-blue (OOTB) packets. A control sock will be created
72 * for this socket at the initialization time.
73 */
74static struct socket *sctp_ctl_socket;
75
76static struct sctp_pf *sctp_pf_inet6_specific;
77static struct sctp_pf *sctp_pf_inet_specific;
78static struct sctp_af *sctp_af_v4_specific;
79static struct sctp_af *sctp_af_v6_specific;
80
Eric Dumazetba899662005-08-26 12:05:31 -070081kmem_cache_t *sctp_chunk_cachep __read_mostly;
82kmem_cache_t *sctp_bucket_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070083
84extern int sctp_snmp_proc_init(void);
85extern int sctp_snmp_proc_exit(void);
86extern int sctp_eps_proc_init(void);
87extern int sctp_eps_proc_exit(void);
88extern int sctp_assocs_proc_init(void);
89extern int sctp_assocs_proc_exit(void);
90
91/* Return the address of the control sock. */
92struct sock *sctp_get_ctl_sock(void)
93{
94 return sctp_ctl_socket->sk;
95}
96
97/* Set up the proc fs entry for the SCTP protocol. */
98static __init int sctp_proc_init(void)
99{
100 if (!proc_net_sctp) {
101 struct proc_dir_entry *ent;
102 ent = proc_mkdir("net/sctp", NULL);
103 if (ent) {
104 ent->owner = THIS_MODULE;
105 proc_net_sctp = ent;
106 } else
107 goto out_nomem;
108 }
109
110 if (sctp_snmp_proc_init())
111 goto out_nomem;
112 if (sctp_eps_proc_init())
113 goto out_nomem;
114 if (sctp_assocs_proc_init())
115 goto out_nomem;
116
117 return 0;
118
119out_nomem:
120 return -ENOMEM;
121}
122
123/* Clean up the proc fs entry for the SCTP protocol.
124 * Note: Do not make this __exit as it is used in the init error
125 * path.
126 */
127static void sctp_proc_exit(void)
128{
129 sctp_snmp_proc_exit();
130 sctp_eps_proc_exit();
131 sctp_assocs_proc_exit();
132
133 if (proc_net_sctp) {
134 proc_net_sctp = NULL;
135 remove_proc_entry("net/sctp", NULL);
136 }
137}
138
139/* Private helper to extract ipv4 address and stash them in
140 * the protocol structure.
141 */
142static void sctp_v4_copy_addrlist(struct list_head *addrlist,
143 struct net_device *dev)
144{
145 struct in_device *in_dev;
146 struct in_ifaddr *ifa;
147 struct sctp_sockaddr_entry *addr;
148
149 rcu_read_lock();
Herbert Xue5ed6392005-10-03 14:35:55 -0700150 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 rcu_read_unlock();
152 return;
153 }
154
155 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
156 /* Add the address to the local list. */
157 addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
158 if (addr) {
159 addr->a.v4.sin_family = AF_INET;
160 addr->a.v4.sin_port = 0;
161 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
162 list_add_tail(&addr->list, addrlist);
163 }
164 }
165
166 rcu_read_unlock();
167}
168
169/* Extract our IP addresses from the system and stash them in the
170 * protocol structure.
171 */
172static void __sctp_get_local_addr_list(void)
173{
174 struct net_device *dev;
175 struct list_head *pos;
176 struct sctp_af *af;
177
178 read_lock(&dev_base_lock);
179 for (dev = dev_base; dev; dev = dev->next) {
180 __list_for_each(pos, &sctp_address_families) {
181 af = list_entry(pos, struct sctp_af, list);
182 af->copy_addrlist(&sctp_local_addr_list, dev);
183 }
184 }
185 read_unlock(&dev_base_lock);
186}
187
188static void sctp_get_local_addr_list(void)
189{
190 unsigned long flags;
191
192 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
193 __sctp_get_local_addr_list();
194 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
195}
196
197/* Free the existing local addresses. */
198static void __sctp_free_local_addr_list(void)
199{
200 struct sctp_sockaddr_entry *addr;
201 struct list_head *pos, *temp;
202
203 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
204 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
205 list_del(pos);
206 kfree(addr);
207 }
208}
209
210/* Free the existing local addresses. */
211static void sctp_free_local_addr_list(void)
212{
213 unsigned long flags;
214
215 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
216 __sctp_free_local_addr_list();
217 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
218}
219
220/* Copy the local addresses which are valid for 'scope' into 'bp'. */
221int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
Al Virodd0fc662005-10-07 07:46:04 +0100222 gfp_t gfp, int copy_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223{
224 struct sctp_sockaddr_entry *addr;
225 int error = 0;
226 struct list_head *pos;
227 unsigned long flags;
228
229 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
230 list_for_each(pos, &sctp_local_addr_list) {
231 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
232 if (sctp_in_scope(&addr->a, scope)) {
233 /* Now that the address is in scope, check to see if
234 * the address type is really supported by the local
235 * sock as well as the remote peer.
236 */
237 if ((((AF_INET == addr->a.sa.sa_family) &&
238 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
239 (((AF_INET6 == addr->a.sa.sa_family) &&
240 (copy_flags & SCTP_ADDR6_ALLOWED) &&
241 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
242 error = sctp_add_bind_addr(bp, &addr->a,
243 GFP_ATOMIC);
244 if (error)
245 goto end_copy;
246 }
247 }
248 }
249
250end_copy:
251 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
252 return error;
253}
254
255/* Initialize a sctp_addr from in incoming skb. */
256static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
257 int is_saddr)
258{
259 void *from;
260 __u16 *port;
261 struct sctphdr *sh;
262
263 port = &addr->v4.sin_port;
264 addr->v4.sin_family = AF_INET;
265
266 sh = (struct sctphdr *) skb->h.raw;
267 if (is_saddr) {
268 *port = ntohs(sh->source);
269 from = &skb->nh.iph->saddr;
270 } else {
271 *port = ntohs(sh->dest);
272 from = &skb->nh.iph->daddr;
273 }
274 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
275}
276
277/* Initialize an sctp_addr from a socket. */
278static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
279{
280 addr->v4.sin_family = AF_INET;
281 addr->v4.sin_port = inet_sk(sk)->num;
282 addr->v4.sin_addr.s_addr = inet_sk(sk)->rcv_saddr;
283}
284
285/* Initialize sk->sk_rcv_saddr from sctp_addr. */
286static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
287{
288 inet_sk(sk)->rcv_saddr = addr->v4.sin_addr.s_addr;
289}
290
291/* Initialize sk->sk_daddr from sctp_addr. */
292static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
293{
294 inet_sk(sk)->daddr = addr->v4.sin_addr.s_addr;
295}
296
297/* Initialize a sctp_addr from an address parameter. */
298static void sctp_v4_from_addr_param(union sctp_addr *addr,
299 union sctp_addr_param *param,
300 __u16 port, int iif)
301{
302 addr->v4.sin_family = AF_INET;
303 addr->v4.sin_port = port;
304 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
305}
306
307/* Initialize an address parameter from a sctp_addr and return the length
308 * of the address parameter.
309 */
310static int sctp_v4_to_addr_param(const union sctp_addr *addr,
311 union sctp_addr_param *param)
312{
313 int length = sizeof(sctp_ipv4addr_param_t);
314
315 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
316 param->v4.param_hdr.length = ntohs(length);
317 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
318
319 return length;
320}
321
322/* Initialize a sctp_addr from a dst_entry. */
323static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct dst_entry *dst,
324 unsigned short port)
325{
326 struct rtable *rt = (struct rtable *)dst;
327 saddr->v4.sin_family = AF_INET;
328 saddr->v4.sin_port = port;
329 saddr->v4.sin_addr.s_addr = rt->rt_src;
330}
331
332/* Compare two addresses exactly. */
333static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
334 const union sctp_addr *addr2)
335{
336 if (addr1->sa.sa_family != addr2->sa.sa_family)
337 return 0;
338 if (addr1->v4.sin_port != addr2->v4.sin_port)
339 return 0;
340 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
341 return 0;
342
343 return 1;
344}
345
346/* Initialize addr struct to INADDR_ANY. */
347static void sctp_v4_inaddr_any(union sctp_addr *addr, unsigned short port)
348{
349 addr->v4.sin_family = AF_INET;
350 addr->v4.sin_addr.s_addr = INADDR_ANY;
351 addr->v4.sin_port = port;
352}
353
354/* Is this a wildcard address? */
355static int sctp_v4_is_any(const union sctp_addr *addr)
356{
357 return INADDR_ANY == addr->v4.sin_addr.s_addr;
358}
359
360/* This function checks if the address is a valid address to be used for
361 * SCTP binding.
362 *
363 * Output:
364 * Return 0 - If the address is a non-unicast or an illegal address.
365 * Return 1 - If the address is a unicast.
366 */
367static int sctp_v4_addr_valid(union sctp_addr *addr, struct sctp_sock *sp)
368{
369 /* Is this a non-unicast address or a unusable SCTP address? */
370 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr))
371 return 0;
372
373 return 1;
374}
375
376/* Should this be available for binding? */
377static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
378{
379 int ret = inet_addr_type(addr->v4.sin_addr.s_addr);
380
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381
Neil Hormancdac4e02005-06-13 15:12:33 -0700382 if (addr->v4.sin_addr.s_addr != INADDR_ANY &&
383 ret != RTN_LOCAL &&
384 !sp->inet.freebind &&
385 !sysctl_ip_nonlocal_bind)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 return 0;
Neil Hormancdac4e02005-06-13 15:12:33 -0700387
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 return 1;
389}
390
391/* Checking the loopback, private and other address scopes as defined in
392 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
393 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
394 *
395 * Level 0 - unusable SCTP addresses
396 * Level 1 - loopback address
397 * Level 2 - link-local addresses
398 * Level 3 - private addresses.
399 * Level 4 - global addresses
400 * For INIT and INIT-ACK address list, let L be the level of
401 * of requested destination address, sender and receiver
402 * SHOULD include all of its addresses with level greater
403 * than or equal to L.
404 */
405static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
406{
407 sctp_scope_t retval;
408
409 /* Should IPv4 scoping be a sysctl configurable option
410 * so users can turn it off (default on) for certain
411 * unconventional networking environments?
412 */
413
414 /* Check for unusable SCTP addresses. */
415 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
416 retval = SCTP_SCOPE_UNUSABLE;
417 } else if (LOOPBACK(addr->v4.sin_addr.s_addr)) {
418 retval = SCTP_SCOPE_LOOPBACK;
419 } else if (IS_IPV4_LINK_ADDRESS(&addr->v4.sin_addr.s_addr)) {
420 retval = SCTP_SCOPE_LINK;
421 } else if (IS_IPV4_PRIVATE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
422 retval = SCTP_SCOPE_PRIVATE;
423 } else {
424 retval = SCTP_SCOPE_GLOBAL;
425 }
426
427 return retval;
428}
429
430/* Returns a valid dst cache entry for the given source and destination ip
431 * addresses. If an association is passed, trys to get a dst entry with a
432 * source address that matches an address in the bind address list.
433 */
434static struct dst_entry *sctp_v4_get_dst(struct sctp_association *asoc,
435 union sctp_addr *daddr,
436 union sctp_addr *saddr)
437{
438 struct rtable *rt;
439 struct flowi fl;
440 struct sctp_bind_addr *bp;
441 rwlock_t *addr_lock;
442 struct sctp_sockaddr_entry *laddr;
443 struct list_head *pos;
444 struct dst_entry *dst = NULL;
445 union sctp_addr dst_saddr;
446
447 memset(&fl, 0x0, sizeof(struct flowi));
448 fl.fl4_dst = daddr->v4.sin_addr.s_addr;
449 fl.proto = IPPROTO_SCTP;
450 if (asoc) {
451 fl.fl4_tos = RT_CONN_FLAGS(asoc->base.sk);
452 fl.oif = asoc->base.sk->sk_bound_dev_if;
453 }
454 if (saddr)
455 fl.fl4_src = saddr->v4.sin_addr.s_addr;
456
457 SCTP_DEBUG_PRINTK("%s: DST:%u.%u.%u.%u, SRC:%u.%u.%u.%u - ",
458 __FUNCTION__, NIPQUAD(fl.fl4_dst),
459 NIPQUAD(fl.fl4_src));
460
461 if (!ip_route_output_key(&rt, &fl)) {
462 dst = &rt->u.dst;
463 }
464
465 /* If there is no association or if a source address is passed, no
466 * more validation is required.
467 */
468 if (!asoc || saddr)
469 goto out;
470
471 bp = &asoc->base.bind_addr;
472 addr_lock = &asoc->base.addr_lock;
473
474 if (dst) {
475 /* Walk through the bind address list and look for a bind
476 * address that matches the source address of the returned dst.
477 */
478 sctp_read_lock(addr_lock);
479 list_for_each(pos, &bp->address_list) {
480 laddr = list_entry(pos, struct sctp_sockaddr_entry,
481 list);
482 sctp_v4_dst_saddr(&dst_saddr, dst, bp->port);
483 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
484 goto out_unlock;
485 }
486 sctp_read_unlock(addr_lock);
487
488 /* None of the bound addresses match the source address of the
489 * dst. So release it.
490 */
491 dst_release(dst);
492 dst = NULL;
493 }
494
495 /* Walk through the bind address list and try to get a dst that
496 * matches a bind address as the source address.
497 */
498 sctp_read_lock(addr_lock);
499 list_for_each(pos, &bp->address_list) {
500 laddr = list_entry(pos, struct sctp_sockaddr_entry, list);
501
502 if (AF_INET == laddr->a.sa.sa_family) {
503 fl.fl4_src = laddr->a.v4.sin_addr.s_addr;
504 if (!ip_route_output_key(&rt, &fl)) {
505 dst = &rt->u.dst;
506 goto out_unlock;
507 }
508 }
509 }
510
511out_unlock:
512 sctp_read_unlock(addr_lock);
513out:
514 if (dst)
515 SCTP_DEBUG_PRINTK("rt_dst:%u.%u.%u.%u, rt_src:%u.%u.%u.%u\n",
516 NIPQUAD(rt->rt_dst), NIPQUAD(rt->rt_src));
517 else
518 SCTP_DEBUG_PRINTK("NO ROUTE\n");
519
520 return dst;
521}
522
523/* For v4, the source address is cached in the route entry(dst). So no need
524 * to cache it separately and hence this is an empty routine.
525 */
526static void sctp_v4_get_saddr(struct sctp_association *asoc,
527 struct dst_entry *dst,
528 union sctp_addr *daddr,
529 union sctp_addr *saddr)
530{
531 struct rtable *rt = (struct rtable *)dst;
532
Vladislav Yasevich23ec47a2005-11-11 16:05:55 -0800533 if (!asoc)
534 return;
535
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 if (rt) {
537 saddr->v4.sin_family = AF_INET;
538 saddr->v4.sin_port = asoc->base.bind_addr.port;
539 saddr->v4.sin_addr.s_addr = rt->rt_src;
540 }
541}
542
543/* What interface did this skb arrive on? */
544static int sctp_v4_skb_iif(const struct sk_buff *skb)
545{
546 return ((struct rtable *)skb->dst)->rt_iif;
547}
548
549/* Was this packet marked by Explicit Congestion Notification? */
550static int sctp_v4_is_ce(const struct sk_buff *skb)
551{
552 return INET_ECN_is_ce(skb->nh.iph->tos);
553}
554
555/* Create and initialize a new sk for the socket returned by accept(). */
556static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
557 struct sctp_association *asoc)
558{
559 struct inet_sock *inet = inet_sk(sk);
560 struct inet_sock *newinet;
561 struct sock *newsk = sk_alloc(PF_INET, GFP_KERNEL, sk->sk_prot, 1);
562
563 if (!newsk)
564 goto out;
565
566 sock_init_data(NULL, newsk);
567
568 newsk->sk_type = SOCK_STREAM;
569
570 newsk->sk_no_check = sk->sk_no_check;
571 newsk->sk_reuse = sk->sk_reuse;
572 newsk->sk_shutdown = sk->sk_shutdown;
573
574 newsk->sk_destruct = inet_sock_destruct;
575 newsk->sk_family = PF_INET;
576 newsk->sk_protocol = IPPROTO_SCTP;
577 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
578 sock_reset_flag(newsk, SOCK_ZAPPED);
579
580 newinet = inet_sk(newsk);
581
582 /* Initialize sk's sport, dport, rcv_saddr and daddr for
583 * getsockname() and getpeername()
584 */
585 newinet->sport = inet->sport;
586 newinet->saddr = inet->saddr;
587 newinet->rcv_saddr = inet->rcv_saddr;
588 newinet->dport = htons(asoc->peer.port);
589 newinet->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
590 newinet->pmtudisc = inet->pmtudisc;
591 newinet->id = 0;
592
593 newinet->uc_ttl = -1;
594 newinet->mc_loop = 1;
595 newinet->mc_ttl = 1;
596 newinet->mc_index = 0;
597 newinet->mc_list = NULL;
598
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700599 sk_refcnt_debug_inc(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601 if (newsk->sk_prot->init(newsk)) {
602 sk_common_release(newsk);
603 newsk = NULL;
604 }
605
606out:
607 return newsk;
608}
609
610/* Map address, empty for v4 family */
611static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
612{
613 /* Empty */
614}
615
616/* Dump the v4 addr to the seq file. */
617static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
618{
619 seq_printf(seq, "%d.%d.%d.%d ", NIPQUAD(addr->v4.sin_addr));
620}
621
622/* Event handler for inet address addition/deletion events.
623 * Basically, whenever there is an event, we re-build our local address list.
624 */
625int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
626 void *ptr)
627{
628 unsigned long flags;
629
630 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
631 __sctp_free_local_addr_list();
632 __sctp_get_local_addr_list();
633 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
634
635 return NOTIFY_DONE;
636}
637
638/*
639 * Initialize the control inode/socket with a control endpoint data
640 * structure. This endpoint is reserved exclusively for the OOTB processing.
641 */
642static int sctp_ctl_sock_init(void)
643{
644 int err;
645 sa_family_t family;
646
647 if (sctp_get_pf_specific(PF_INET6))
648 family = PF_INET6;
649 else
650 family = PF_INET;
651
652 err = sock_create_kern(family, SOCK_SEQPACKET, IPPROTO_SCTP,
653 &sctp_ctl_socket);
654 if (err < 0) {
655 printk(KERN_ERR
656 "SCTP: Failed to create the SCTP control socket.\n");
657 return err;
658 }
659 sctp_ctl_socket->sk->sk_allocation = GFP_ATOMIC;
660 inet_sk(sctp_ctl_socket->sk)->uc_ttl = -1;
661
662 return 0;
663}
664
665/* Register address family specific functions. */
666int sctp_register_af(struct sctp_af *af)
667{
668 switch (af->sa_family) {
669 case AF_INET:
670 if (sctp_af_v4_specific)
671 return 0;
672 sctp_af_v4_specific = af;
673 break;
674 case AF_INET6:
675 if (sctp_af_v6_specific)
676 return 0;
677 sctp_af_v6_specific = af;
678 break;
679 default:
680 return 0;
681 }
682
683 INIT_LIST_HEAD(&af->list);
684 list_add_tail(&af->list, &sctp_address_families);
685 return 1;
686}
687
688/* Get the table of functions for manipulating a particular address
689 * family.
690 */
691struct sctp_af *sctp_get_af_specific(sa_family_t family)
692{
693 switch (family) {
694 case AF_INET:
695 return sctp_af_v4_specific;
696 case AF_INET6:
697 return sctp_af_v6_specific;
698 default:
699 return NULL;
700 }
701}
702
703/* Common code to initialize a AF_INET msg_name. */
704static void sctp_inet_msgname(char *msgname, int *addr_len)
705{
706 struct sockaddr_in *sin;
707
708 sin = (struct sockaddr_in *)msgname;
709 *addr_len = sizeof(struct sockaddr_in);
710 sin->sin_family = AF_INET;
711 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
712}
713
714/* Copy the primary address of the peer primary address as the msg_name. */
715static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
716 int *addr_len)
717{
718 struct sockaddr_in *sin, *sinfrom;
719
720 if (msgname) {
721 struct sctp_association *asoc;
722
723 asoc = event->asoc;
724 sctp_inet_msgname(msgname, addr_len);
725 sin = (struct sockaddr_in *)msgname;
726 sinfrom = &asoc->peer.primary_addr.v4;
727 sin->sin_port = htons(asoc->peer.port);
728 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
729 }
730}
731
732/* Initialize and copy out a msgname from an inbound skb. */
733static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
734{
735 struct sctphdr *sh;
736 struct sockaddr_in *sin;
737
738 if (msgname) {
739 sctp_inet_msgname(msgname, len);
740 sin = (struct sockaddr_in *)msgname;
741 sh = (struct sctphdr *)skb->h.raw;
742 sin->sin_port = sh->source;
743 sin->sin_addr.s_addr = skb->nh.iph->saddr;
744 }
745}
746
747/* Do we support this AF? */
748static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
749{
750 /* PF_INET only supports AF_INET addresses. */
751 return (AF_INET == family);
752}
753
754/* Address matching with wildcards allowed. */
755static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
756 const union sctp_addr *addr2,
757 struct sctp_sock *opt)
758{
759 /* PF_INET only supports AF_INET addresses. */
760 if (addr1->sa.sa_family != addr2->sa.sa_family)
761 return 0;
762 if (INADDR_ANY == addr1->v4.sin_addr.s_addr ||
763 INADDR_ANY == addr2->v4.sin_addr.s_addr)
764 return 1;
765 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
766 return 1;
767
768 return 0;
769}
770
771/* Verify that provided sockaddr looks bindable. Common verification has
772 * already been taken care of.
773 */
774static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
775{
776 return sctp_v4_available(addr, opt);
777}
778
779/* Verify that sockaddr looks sendable. Common verification has already
780 * been taken care of.
781 */
782static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
783{
784 return 1;
785}
786
787/* Fill in Supported Address Type information for INIT and INIT-ACK
788 * chunks. Returns number of addresses supported.
789 */
790static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
791 __u16 *types)
792{
793 types[0] = SCTP_PARAM_IPV4_ADDRESS;
794 return 1;
795}
796
797/* Wrapper routine that calls the ip transmit routine. */
798static inline int sctp_v4_xmit(struct sk_buff *skb,
799 struct sctp_transport *transport, int ipfragok)
800{
801 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, "
802 "src:%u.%u.%u.%u, dst:%u.%u.%u.%u\n",
803 __FUNCTION__, skb, skb->len,
804 NIPQUAD(((struct rtable *)skb->dst)->rt_src),
805 NIPQUAD(((struct rtable *)skb->dst)->rt_dst));
806
807 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
808 return ip_queue_xmit(skb, ipfragok);
809}
810
811static struct sctp_af sctp_ipv4_specific;
812
813static struct sctp_pf sctp_pf_inet = {
814 .event_msgname = sctp_inet_event_msgname,
815 .skb_msgname = sctp_inet_skb_msgname,
816 .af_supported = sctp_inet_af_supported,
817 .cmp_addr = sctp_inet_cmp_addr,
818 .bind_verify = sctp_inet_bind_verify,
819 .send_verify = sctp_inet_send_verify,
820 .supported_addrs = sctp_inet_supported_addrs,
821 .create_accept_sk = sctp_v4_create_accept_sk,
822 .addr_v4map = sctp_v4_addr_v4map,
823 .af = &sctp_ipv4_specific,
824};
825
826/* Notifier for inetaddr addition/deletion events. */
827static struct notifier_block sctp_inetaddr_notifier = {
828 .notifier_call = sctp_inetaddr_event,
829};
830
831/* Socket operations. */
832static struct proto_ops inet_seqpacket_ops = {
833 .family = PF_INET,
834 .owner = THIS_MODULE,
835 .release = inet_release, /* Needs to be wrapped... */
836 .bind = inet_bind,
837 .connect = inet_dgram_connect,
838 .socketpair = sock_no_socketpair,
839 .accept = inet_accept,
840 .getname = inet_getname, /* Semantics are different. */
841 .poll = sctp_poll,
842 .ioctl = inet_ioctl,
843 .listen = sctp_inet_listen,
844 .shutdown = inet_shutdown, /* Looks harmless. */
845 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem. */
846 .getsockopt = sock_common_getsockopt,
847 .sendmsg = inet_sendmsg,
848 .recvmsg = sock_common_recvmsg,
849 .mmap = sock_no_mmap,
850 .sendpage = sock_no_sendpage,
851};
852
853/* Registration with AF_INET family. */
854static struct inet_protosw sctp_seqpacket_protosw = {
855 .type = SOCK_SEQPACKET,
856 .protocol = IPPROTO_SCTP,
857 .prot = &sctp_prot,
858 .ops = &inet_seqpacket_ops,
859 .capability = -1,
860 .no_check = 0,
861 .flags = SCTP_PROTOSW_FLAG
862};
863static struct inet_protosw sctp_stream_protosw = {
864 .type = SOCK_STREAM,
865 .protocol = IPPROTO_SCTP,
866 .prot = &sctp_prot,
867 .ops = &inet_seqpacket_ops,
868 .capability = -1,
869 .no_check = 0,
870 .flags = SCTP_PROTOSW_FLAG
871};
872
873/* Register with IP layer. */
874static struct net_protocol sctp_protocol = {
875 .handler = sctp_rcv,
876 .err_handler = sctp_v4_err,
877 .no_policy = 1,
878};
879
880/* IPv4 address related functions. */
881static struct sctp_af sctp_ipv4_specific = {
882 .sctp_xmit = sctp_v4_xmit,
883 .setsockopt = ip_setsockopt,
884 .getsockopt = ip_getsockopt,
885 .get_dst = sctp_v4_get_dst,
886 .get_saddr = sctp_v4_get_saddr,
887 .copy_addrlist = sctp_v4_copy_addrlist,
888 .from_skb = sctp_v4_from_skb,
889 .from_sk = sctp_v4_from_sk,
890 .to_sk_saddr = sctp_v4_to_sk_saddr,
891 .to_sk_daddr = sctp_v4_to_sk_daddr,
892 .from_addr_param= sctp_v4_from_addr_param,
893 .to_addr_param = sctp_v4_to_addr_param,
894 .dst_saddr = sctp_v4_dst_saddr,
895 .cmp_addr = sctp_v4_cmp_addr,
896 .addr_valid = sctp_v4_addr_valid,
897 .inaddr_any = sctp_v4_inaddr_any,
898 .is_any = sctp_v4_is_any,
899 .available = sctp_v4_available,
900 .scope = sctp_v4_scope,
901 .skb_iif = sctp_v4_skb_iif,
902 .is_ce = sctp_v4_is_ce,
903 .seq_dump_addr = sctp_v4_seq_dump_addr,
904 .net_header_len = sizeof(struct iphdr),
905 .sockaddr_len = sizeof(struct sockaddr_in),
906 .sa_family = AF_INET,
907};
908
909struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
910
911 switch (family) {
912 case PF_INET:
913 return sctp_pf_inet_specific;
914 case PF_INET6:
915 return sctp_pf_inet6_specific;
916 default:
917 return NULL;
918 }
919}
920
921/* Register the PF specific function table. */
922int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
923{
924 switch (family) {
925 case PF_INET:
926 if (sctp_pf_inet_specific)
927 return 0;
928 sctp_pf_inet_specific = pf;
929 break;
930 case PF_INET6:
931 if (sctp_pf_inet6_specific)
932 return 0;
933 sctp_pf_inet6_specific = pf;
934 break;
935 default:
936 return 0;
937 }
938 return 1;
939}
940
941static int __init init_sctp_mibs(void)
942{
943 sctp_statistics[0] = alloc_percpu(struct sctp_mib);
944 if (!sctp_statistics[0])
945 return -ENOMEM;
946 sctp_statistics[1] = alloc_percpu(struct sctp_mib);
947 if (!sctp_statistics[1]) {
948 free_percpu(sctp_statistics[0]);
949 return -ENOMEM;
950 }
951 return 0;
952
953}
954
955static void cleanup_sctp_mibs(void)
956{
957 free_percpu(sctp_statistics[0]);
958 free_percpu(sctp_statistics[1]);
959}
960
961/* Initialize the universe into something sensible. */
962SCTP_STATIC __init int sctp_init(void)
963{
964 int i;
965 int status = -EINVAL;
966 unsigned long goal;
967 int order;
968
969 /* SCTP_DEBUG sanity check. */
970 if (!sctp_sanity_check())
971 goto out;
972
973 status = proto_register(&sctp_prot, 1);
974 if (status)
975 goto out;
976
977 /* Add SCTP to inet_protos hash table. */
978 status = -EAGAIN;
979 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
980 goto err_add_protocol;
981
982 /* Add SCTP(TCP and UDP style) to inetsw linked list. */
983 inet_register_protosw(&sctp_seqpacket_protosw);
984 inet_register_protosw(&sctp_stream_protosw);
985
986 /* Allocate a cache pools. */
987 status = -ENOBUFS;
988 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
989 sizeof(struct sctp_bind_bucket),
990 0, SLAB_HWCACHE_ALIGN,
991 NULL, NULL);
992
993 if (!sctp_bucket_cachep)
994 goto err_bucket_cachep;
995
996 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
997 sizeof(struct sctp_chunk),
998 0, SLAB_HWCACHE_ALIGN,
999 NULL, NULL);
1000 if (!sctp_chunk_cachep)
1001 goto err_chunk_cachep;
1002
1003 /* Allocate and initialise sctp mibs. */
1004 status = init_sctp_mibs();
1005 if (status)
1006 goto err_init_mibs;
1007
1008 /* Initialize proc fs directory. */
1009 status = sctp_proc_init();
1010 if (status)
1011 goto err_init_proc;
1012
1013 /* Initialize object count debugging. */
1014 sctp_dbg_objcnt_init();
1015
1016 /* Initialize the SCTP specific PF functions. */
1017 sctp_register_pf(&sctp_pf_inet, PF_INET);
1018 /*
1019 * 14. Suggested SCTP Protocol Parameter Values
1020 */
1021 /* The following protocol parameters are RECOMMENDED: */
1022 /* RTO.Initial - 3 seconds */
1023 sctp_rto_initial = SCTP_RTO_INITIAL;
1024 /* RTO.Min - 1 second */
1025 sctp_rto_min = SCTP_RTO_MIN;
1026 /* RTO.Max - 60 seconds */
1027 sctp_rto_max = SCTP_RTO_MAX;
1028 /* RTO.Alpha - 1/8 */
1029 sctp_rto_alpha = SCTP_RTO_ALPHA;
1030 /* RTO.Beta - 1/4 */
1031 sctp_rto_beta = SCTP_RTO_BETA;
1032
1033 /* Valid.Cookie.Life - 60 seconds */
1034 sctp_valid_cookie_life = 60 * HZ;
1035
1036 /* Whether Cookie Preservative is enabled(1) or not(0) */
1037 sctp_cookie_preserve_enable = 1;
1038
1039 /* Max.Burst - 4 */
1040 sctp_max_burst = SCTP_MAX_BURST;
1041
1042 /* Association.Max.Retrans - 10 attempts
1043 * Path.Max.Retrans - 5 attempts (per destination address)
1044 * Max.Init.Retransmits - 8 attempts
1045 */
1046 sctp_max_retrans_association = 10;
1047 sctp_max_retrans_path = 5;
1048 sctp_max_retrans_init = 8;
1049
Neil Horman4eb701d2005-04-28 12:02:04 -07001050 /* Sendbuffer growth - do per-socket accounting */
1051 sctp_sndbuf_policy = 0;
1052
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 /* HB.interval - 30 seconds */
Vlad Yasevich2f85a422005-06-28 13:24:23 -07001054 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1055
1056 /* delayed SACK timeout */
1057 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
1059 /* Implementation specific variables. */
1060
1061 /* Initialize default stream count setup information. */
1062 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1063 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1064
1065 /* Initialize handle used for association ids. */
1066 idr_init(&sctp_assocs_id);
1067
1068 /* Size and allocate the association hash table.
1069 * The methodology is similar to that of the tcp hash tables.
1070 */
1071 if (num_physpages >= (128 * 1024))
1072 goal = num_physpages >> (22 - PAGE_SHIFT);
1073 else
1074 goal = num_physpages >> (24 - PAGE_SHIFT);
1075
1076 for (order = 0; (1UL << order) < goal; order++)
1077 ;
1078
1079 do {
1080 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1081 sizeof(struct sctp_hashbucket);
1082 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1083 continue;
1084 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1085 __get_free_pages(GFP_ATOMIC, order);
1086 } while (!sctp_assoc_hashtable && --order > 0);
1087 if (!sctp_assoc_hashtable) {
1088 printk(KERN_ERR "SCTP: Failed association hash alloc.\n");
1089 status = -ENOMEM;
1090 goto err_ahash_alloc;
1091 }
1092 for (i = 0; i < sctp_assoc_hashsize; i++) {
1093 rwlock_init(&sctp_assoc_hashtable[i].lock);
1094 sctp_assoc_hashtable[i].chain = NULL;
1095 }
1096
1097 /* Allocate and initialize the endpoint hash table. */
1098 sctp_ep_hashsize = 64;
1099 sctp_ep_hashtable = (struct sctp_hashbucket *)
1100 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1101 if (!sctp_ep_hashtable) {
1102 printk(KERN_ERR "SCTP: Failed endpoint_hash alloc.\n");
1103 status = -ENOMEM;
1104 goto err_ehash_alloc;
1105 }
1106 for (i = 0; i < sctp_ep_hashsize; i++) {
1107 rwlock_init(&sctp_ep_hashtable[i].lock);
1108 sctp_ep_hashtable[i].chain = NULL;
1109 }
1110
1111 /* Allocate and initialize the SCTP port hash table. */
1112 do {
1113 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1114 sizeof(struct sctp_bind_hashbucket);
1115 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1116 continue;
1117 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1118 __get_free_pages(GFP_ATOMIC, order);
1119 } while (!sctp_port_hashtable && --order > 0);
1120 if (!sctp_port_hashtable) {
1121 printk(KERN_ERR "SCTP: Failed bind hash alloc.");
1122 status = -ENOMEM;
1123 goto err_bhash_alloc;
1124 }
1125 for (i = 0; i < sctp_port_hashsize; i++) {
1126 spin_lock_init(&sctp_port_hashtable[i].lock);
1127 sctp_port_hashtable[i].chain = NULL;
1128 }
1129
1130 spin_lock_init(&sctp_port_alloc_lock);
1131 sctp_port_rover = sysctl_local_port_range[0] - 1;
1132
1133 printk(KERN_INFO "SCTP: Hash tables configured "
1134 "(established %d bind %d)\n",
1135 sctp_assoc_hashsize, sctp_port_hashsize);
1136
1137 /* Disable ADDIP by default. */
1138 sctp_addip_enable = 0;
1139
1140 /* Enable PR-SCTP by default. */
1141 sctp_prsctp_enable = 1;
1142
1143 sctp_sysctl_register();
1144
1145 INIT_LIST_HEAD(&sctp_address_families);
1146 sctp_register_af(&sctp_ipv4_specific);
1147
1148 status = sctp_v6_init();
1149 if (status)
1150 goto err_v6_init;
1151
1152 /* Initialize the control inode/socket for handling OOTB packets. */
1153 if ((status = sctp_ctl_sock_init())) {
1154 printk (KERN_ERR
1155 "SCTP: Failed to initialize the SCTP control sock.\n");
1156 goto err_ctl_sock_init;
1157 }
1158
1159 /* Initialize the local address list. */
1160 INIT_LIST_HEAD(&sctp_local_addr_list);
1161 spin_lock_init(&sctp_local_addr_lock);
1162
1163 /* Register notifier for inet address additions/deletions. */
1164 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1165
1166 sctp_get_local_addr_list();
1167
1168 __unsafe(THIS_MODULE);
1169 status = 0;
1170out:
1171 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172err_ctl_sock_init:
1173 sctp_v6_exit();
1174err_v6_init:
1175 sctp_sysctl_unregister();
1176 list_del(&sctp_ipv4_specific.list);
1177 free_pages((unsigned long)sctp_port_hashtable,
1178 get_order(sctp_port_hashsize *
1179 sizeof(struct sctp_bind_hashbucket)));
1180err_bhash_alloc:
1181 kfree(sctp_ep_hashtable);
1182err_ehash_alloc:
1183 free_pages((unsigned long)sctp_assoc_hashtable,
1184 get_order(sctp_assoc_hashsize *
1185 sizeof(struct sctp_hashbucket)));
1186err_ahash_alloc:
1187 sctp_dbg_objcnt_exit();
1188err_init_proc:
1189 sctp_proc_exit();
1190 cleanup_sctp_mibs();
1191err_init_mibs:
1192 kmem_cache_destroy(sctp_chunk_cachep);
1193err_chunk_cachep:
1194 kmem_cache_destroy(sctp_bucket_cachep);
1195err_bucket_cachep:
1196 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1197 inet_unregister_protosw(&sctp_seqpacket_protosw);
1198 inet_unregister_protosw(&sctp_stream_protosw);
Neil Horman5e6bc342005-04-28 11:59:49 -07001199err_add_protocol:
1200 proto_unregister(&sctp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 goto out;
1202}
1203
1204/* Exit handler for the SCTP protocol. */
1205SCTP_STATIC __exit void sctp_exit(void)
1206{
1207 /* BUG. This should probably do something useful like clean
1208 * up all the remaining associations and all that memory.
1209 */
1210
1211 /* Unregister notifier for inet address additions/deletions. */
1212 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1213
1214 /* Free the local address list. */
1215 sctp_free_local_addr_list();
1216
1217 /* Free the control endpoint. */
1218 sock_release(sctp_ctl_socket);
1219
1220 sctp_v6_exit();
1221 sctp_sysctl_unregister();
1222 list_del(&sctp_ipv4_specific.list);
1223
1224 free_pages((unsigned long)sctp_assoc_hashtable,
1225 get_order(sctp_assoc_hashsize *
1226 sizeof(struct sctp_hashbucket)));
1227 kfree(sctp_ep_hashtable);
1228 free_pages((unsigned long)sctp_port_hashtable,
1229 get_order(sctp_port_hashsize *
1230 sizeof(struct sctp_bind_hashbucket)));
1231
1232 kmem_cache_destroy(sctp_chunk_cachep);
1233 kmem_cache_destroy(sctp_bucket_cachep);
1234
1235 sctp_dbg_objcnt_exit();
1236 sctp_proc_exit();
1237 cleanup_sctp_mibs();
1238
1239 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1240 inet_unregister_protosw(&sctp_seqpacket_protosw);
1241 inet_unregister_protosw(&sctp_stream_protosw);
1242 proto_unregister(&sctp_prot);
1243}
1244
1245module_init(sctp_init);
1246module_exit(sctp_exit);
1247
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -07001248/*
1249 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1250 */
1251MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1253MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1254MODULE_LICENSE("GPL");