ia64/linux-2.6.18-xen.hg

view net/sctp/protocol.c @ 871:9cbcc9008446

xen/x86: don't initialize cpu_data[]'s apicid field on generic code

Afaict, this is not only redundant with the intialization done in
drivers/xen/core/smpboot.c, but actually results - at least for
secondary CPUs - in the Xen-specific value written to be later
overwritten with whatever the generic code determines (with no
guarantee that the two values are identical).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
author Keir Fraser <keir.fraser@citrix.com>
date Thu May 14 10:09:15 2009 +0100 (2009-05-14)
parents 831230e53067
children
line source
1 /* 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 */
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/route.h>
58 #include <net/sctp/sctp.h>
59 #include <net/addrconf.h>
60 #include <net/inet_common.h>
61 #include <net/inet_ecn.h>
63 /* Global data structures. */
64 struct sctp_globals sctp_globals;
65 struct proc_dir_entry *proc_net_sctp;
66 DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
68 struct idr sctp_assocs_id;
69 DEFINE_SPINLOCK(sctp_assocs_id_lock);
71 /* This is the global socket data structure used for responding to
72 * the Out-of-the-blue (OOTB) packets. A control sock will be created
73 * for this socket at the initialization time.
74 */
75 static struct socket *sctp_ctl_socket;
77 static struct sctp_pf *sctp_pf_inet6_specific;
78 static struct sctp_pf *sctp_pf_inet_specific;
79 static struct sctp_af *sctp_af_v4_specific;
80 static struct sctp_af *sctp_af_v6_specific;
82 kmem_cache_t *sctp_chunk_cachep __read_mostly;
83 kmem_cache_t *sctp_bucket_cachep __read_mostly;
85 extern int sctp_snmp_proc_init(void);
86 extern int sctp_snmp_proc_exit(void);
87 extern int sctp_eps_proc_init(void);
88 extern int sctp_eps_proc_exit(void);
89 extern int sctp_assocs_proc_init(void);
90 extern int sctp_assocs_proc_exit(void);
92 /* Return the address of the control sock. */
93 struct sock *sctp_get_ctl_sock(void)
94 {
95 return sctp_ctl_socket->sk;
96 }
98 /* Set up the proc fs entry for the SCTP protocol. */
99 static __init int sctp_proc_init(void)
100 {
101 if (!proc_net_sctp) {
102 struct proc_dir_entry *ent;
103 ent = proc_mkdir("net/sctp", NULL);
104 if (ent) {
105 ent->owner = THIS_MODULE;
106 proc_net_sctp = ent;
107 } else
108 goto out_nomem;
109 }
111 if (sctp_snmp_proc_init())
112 goto out_nomem;
113 if (sctp_eps_proc_init())
114 goto out_nomem;
115 if (sctp_assocs_proc_init())
116 goto out_nomem;
118 return 0;
120 out_nomem:
121 return -ENOMEM;
122 }
124 /* Clean up the proc fs entry for the SCTP protocol.
125 * Note: Do not make this __exit as it is used in the init error
126 * path.
127 */
128 static void sctp_proc_exit(void)
129 {
130 sctp_snmp_proc_exit();
131 sctp_eps_proc_exit();
132 sctp_assocs_proc_exit();
134 if (proc_net_sctp) {
135 proc_net_sctp = NULL;
136 remove_proc_entry("net/sctp", NULL);
137 }
138 }
140 /* Private helper to extract ipv4 address and stash them in
141 * the protocol structure.
142 */
143 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
144 struct net_device *dev)
145 {
146 struct in_device *in_dev;
147 struct in_ifaddr *ifa;
148 struct sctp_sockaddr_entry *addr;
150 rcu_read_lock();
151 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
152 rcu_read_unlock();
153 return;
154 }
156 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
157 /* Add the address to the local list. */
158 addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
159 if (addr) {
160 addr->a.v4.sin_family = AF_INET;
161 addr->a.v4.sin_port = 0;
162 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
163 list_add_tail(&addr->list, addrlist);
164 }
165 }
167 rcu_read_unlock();
168 }
170 /* Extract our IP addresses from the system and stash them in the
171 * protocol structure.
172 */
173 static void __sctp_get_local_addr_list(void)
174 {
175 struct net_device *dev;
176 struct list_head *pos;
177 struct sctp_af *af;
179 read_lock(&dev_base_lock);
180 for (dev = dev_base; dev; dev = dev->next) {
181 __list_for_each(pos, &sctp_address_families) {
182 af = list_entry(pos, struct sctp_af, list);
183 af->copy_addrlist(&sctp_local_addr_list, dev);
184 }
185 }
186 read_unlock(&dev_base_lock);
187 }
189 static void sctp_get_local_addr_list(void)
190 {
191 unsigned long flags;
193 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
194 __sctp_get_local_addr_list();
195 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
196 }
198 /* Free the existing local addresses. */
199 static void __sctp_free_local_addr_list(void)
200 {
201 struct sctp_sockaddr_entry *addr;
202 struct list_head *pos, *temp;
204 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
205 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
206 list_del(pos);
207 kfree(addr);
208 }
209 }
211 /* Free the existing local addresses. */
212 static void sctp_free_local_addr_list(void)
213 {
214 unsigned long flags;
216 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
217 __sctp_free_local_addr_list();
218 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
219 }
221 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
222 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
223 gfp_t gfp, int copy_flags)
224 {
225 struct sctp_sockaddr_entry *addr;
226 int error = 0;
227 struct list_head *pos;
228 unsigned long flags;
230 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
231 list_for_each(pos, &sctp_local_addr_list) {
232 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
233 if (sctp_in_scope(&addr->a, scope)) {
234 /* Now that the address is in scope, check to see if
235 * the address type is really supported by the local
236 * sock as well as the remote peer.
237 */
238 if ((((AF_INET == addr->a.sa.sa_family) &&
239 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
240 (((AF_INET6 == addr->a.sa.sa_family) &&
241 (copy_flags & SCTP_ADDR6_ALLOWED) &&
242 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
243 error = sctp_add_bind_addr(bp, &addr->a, 1,
244 GFP_ATOMIC);
245 if (error)
246 goto end_copy;
247 }
248 }
249 }
251 end_copy:
252 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
253 return error;
254 }
256 /* Initialize a sctp_addr from in incoming skb. */
257 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
258 int is_saddr)
259 {
260 void *from;
261 __u16 *port;
262 struct sctphdr *sh;
264 port = &addr->v4.sin_port;
265 addr->v4.sin_family = AF_INET;
267 sh = (struct sctphdr *) skb->h.raw;
268 if (is_saddr) {
269 *port = ntohs(sh->source);
270 from = &skb->nh.iph->saddr;
271 } else {
272 *port = ntohs(sh->dest);
273 from = &skb->nh.iph->daddr;
274 }
275 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
276 }
278 /* Initialize an sctp_addr from a socket. */
279 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
280 {
281 addr->v4.sin_family = AF_INET;
282 addr->v4.sin_port = inet_sk(sk)->num;
283 addr->v4.sin_addr.s_addr = inet_sk(sk)->rcv_saddr;
284 }
286 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
287 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
288 {
289 inet_sk(sk)->rcv_saddr = addr->v4.sin_addr.s_addr;
290 }
292 /* Initialize sk->sk_daddr from sctp_addr. */
293 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
294 {
295 inet_sk(sk)->daddr = addr->v4.sin_addr.s_addr;
296 }
298 /* Initialize a sctp_addr from an address parameter. */
299 static void sctp_v4_from_addr_param(union sctp_addr *addr,
300 union sctp_addr_param *param,
301 __u16 port, int iif)
302 {
303 addr->v4.sin_family = AF_INET;
304 addr->v4.sin_port = port;
305 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
306 }
308 /* Initialize an address parameter from a sctp_addr and return the length
309 * of the address parameter.
310 */
311 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
312 union sctp_addr_param *param)
313 {
314 int length = sizeof(sctp_ipv4addr_param_t);
316 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
317 param->v4.param_hdr.length = ntohs(length);
318 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
320 return length;
321 }
323 /* Initialize a sctp_addr from a dst_entry. */
324 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct dst_entry *dst,
325 unsigned short port)
326 {
327 struct rtable *rt = (struct rtable *)dst;
328 saddr->v4.sin_family = AF_INET;
329 saddr->v4.sin_port = port;
330 saddr->v4.sin_addr.s_addr = rt->rt_src;
331 }
333 /* Compare two addresses exactly. */
334 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
335 const union sctp_addr *addr2)
336 {
337 if (addr1->sa.sa_family != addr2->sa.sa_family)
338 return 0;
339 if (addr1->v4.sin_port != addr2->v4.sin_port)
340 return 0;
341 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
342 return 0;
344 return 1;
345 }
347 /* Initialize addr struct to INADDR_ANY. */
348 static void sctp_v4_inaddr_any(union sctp_addr *addr, unsigned short port)
349 {
350 addr->v4.sin_family = AF_INET;
351 addr->v4.sin_addr.s_addr = INADDR_ANY;
352 addr->v4.sin_port = port;
353 }
355 /* Is this a wildcard address? */
356 static int sctp_v4_is_any(const union sctp_addr *addr)
357 {
358 return INADDR_ANY == addr->v4.sin_addr.s_addr;
359 }
361 /* This function checks if the address is a valid address to be used for
362 * SCTP binding.
363 *
364 * Output:
365 * Return 0 - If the address is a non-unicast or an illegal address.
366 * Return 1 - If the address is a unicast.
367 */
368 static int sctp_v4_addr_valid(union sctp_addr *addr,
369 struct sctp_sock *sp,
370 const struct sk_buff *skb)
371 {
372 /* Is this a non-unicast address or a unusable SCTP address? */
373 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr))
374 return 0;
376 /* Is this a broadcast address? */
377 if (skb && ((struct rtable *)skb->dst)->rt_flags & RTCF_BROADCAST)
378 return 0;
380 return 1;
381 }
383 /* Should this be available for binding? */
384 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
385 {
386 int ret = inet_addr_type(addr->v4.sin_addr.s_addr);
389 if (addr->v4.sin_addr.s_addr != INADDR_ANY &&
390 ret != RTN_LOCAL &&
391 !sp->inet.freebind &&
392 !sysctl_ip_nonlocal_bind)
393 return 0;
395 return 1;
396 }
398 /* Checking the loopback, private and other address scopes as defined in
399 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
400 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
401 *
402 * Level 0 - unusable SCTP addresses
403 * Level 1 - loopback address
404 * Level 2 - link-local addresses
405 * Level 3 - private addresses.
406 * Level 4 - global addresses
407 * For INIT and INIT-ACK address list, let L be the level of
408 * of requested destination address, sender and receiver
409 * SHOULD include all of its addresses with level greater
410 * than or equal to L.
411 */
412 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
413 {
414 sctp_scope_t retval;
416 /* Should IPv4 scoping be a sysctl configurable option
417 * so users can turn it off (default on) for certain
418 * unconventional networking environments?
419 */
421 /* Check for unusable SCTP addresses. */
422 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
423 retval = SCTP_SCOPE_UNUSABLE;
424 } else if (LOOPBACK(addr->v4.sin_addr.s_addr)) {
425 retval = SCTP_SCOPE_LOOPBACK;
426 } else if (IS_IPV4_LINK_ADDRESS(&addr->v4.sin_addr.s_addr)) {
427 retval = SCTP_SCOPE_LINK;
428 } else if (IS_IPV4_PRIVATE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
429 retval = SCTP_SCOPE_PRIVATE;
430 } else {
431 retval = SCTP_SCOPE_GLOBAL;
432 }
434 return retval;
435 }
437 /* Returns a valid dst cache entry for the given source and destination ip
438 * addresses. If an association is passed, trys to get a dst entry with a
439 * source address that matches an address in the bind address list.
440 */
441 static struct dst_entry *sctp_v4_get_dst(struct sctp_association *asoc,
442 union sctp_addr *daddr,
443 union sctp_addr *saddr)
444 {
445 struct rtable *rt;
446 struct flowi fl;
447 struct sctp_bind_addr *bp;
448 rwlock_t *addr_lock;
449 struct sctp_sockaddr_entry *laddr;
450 struct list_head *pos;
451 struct dst_entry *dst = NULL;
452 union sctp_addr dst_saddr;
454 memset(&fl, 0x0, sizeof(struct flowi));
455 fl.fl4_dst = daddr->v4.sin_addr.s_addr;
456 fl.proto = IPPROTO_SCTP;
457 if (asoc) {
458 fl.fl4_tos = RT_CONN_FLAGS(asoc->base.sk);
459 fl.oif = asoc->base.sk->sk_bound_dev_if;
460 }
461 if (saddr)
462 fl.fl4_src = saddr->v4.sin_addr.s_addr;
464 SCTP_DEBUG_PRINTK("%s: DST:%u.%u.%u.%u, SRC:%u.%u.%u.%u - ",
465 __FUNCTION__, NIPQUAD(fl.fl4_dst),
466 NIPQUAD(fl.fl4_src));
468 if (!ip_route_output_key(&rt, &fl)) {
469 dst = &rt->u.dst;
470 }
472 /* If there is no association or if a source address is passed, no
473 * more validation is required.
474 */
475 if (!asoc || saddr)
476 goto out;
478 bp = &asoc->base.bind_addr;
479 addr_lock = &asoc->base.addr_lock;
481 if (dst) {
482 /* Walk through the bind address list and look for a bind
483 * address that matches the source address of the returned dst.
484 */
485 sctp_read_lock(addr_lock);
486 list_for_each(pos, &bp->address_list) {
487 laddr = list_entry(pos, struct sctp_sockaddr_entry,
488 list);
489 if (!laddr->use_as_src)
490 continue;
491 sctp_v4_dst_saddr(&dst_saddr, dst, bp->port);
492 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
493 goto out_unlock;
494 }
495 sctp_read_unlock(addr_lock);
497 /* None of the bound addresses match the source address of the
498 * dst. So release it.
499 */
500 dst_release(dst);
501 dst = NULL;
502 }
504 /* Walk through the bind address list and try to get a dst that
505 * matches a bind address as the source address.
506 */
507 sctp_read_lock(addr_lock);
508 list_for_each(pos, &bp->address_list) {
509 laddr = list_entry(pos, struct sctp_sockaddr_entry, list);
511 if ((laddr->use_as_src) &&
512 (AF_INET == laddr->a.sa.sa_family)) {
513 fl.fl4_src = laddr->a.v4.sin_addr.s_addr;
514 if (!ip_route_output_key(&rt, &fl)) {
515 dst = &rt->u.dst;
516 goto out_unlock;
517 }
518 }
519 }
521 out_unlock:
522 sctp_read_unlock(addr_lock);
523 out:
524 if (dst)
525 SCTP_DEBUG_PRINTK("rt_dst:%u.%u.%u.%u, rt_src:%u.%u.%u.%u\n",
526 NIPQUAD(rt->rt_dst), NIPQUAD(rt->rt_src));
527 else
528 SCTP_DEBUG_PRINTK("NO ROUTE\n");
530 return dst;
531 }
533 /* For v4, the source address is cached in the route entry(dst). So no need
534 * to cache it separately and hence this is an empty routine.
535 */
536 static void sctp_v4_get_saddr(struct sctp_association *asoc,
537 struct dst_entry *dst,
538 union sctp_addr *daddr,
539 union sctp_addr *saddr)
540 {
541 struct rtable *rt = (struct rtable *)dst;
543 if (!asoc)
544 return;
546 if (rt) {
547 saddr->v4.sin_family = AF_INET;
548 saddr->v4.sin_port = asoc->base.bind_addr.port;
549 saddr->v4.sin_addr.s_addr = rt->rt_src;
550 }
551 }
553 /* What interface did this skb arrive on? */
554 static int sctp_v4_skb_iif(const struct sk_buff *skb)
555 {
556 return ((struct rtable *)skb->dst)->rt_iif;
557 }
559 /* Was this packet marked by Explicit Congestion Notification? */
560 static int sctp_v4_is_ce(const struct sk_buff *skb)
561 {
562 return INET_ECN_is_ce(skb->nh.iph->tos);
563 }
565 /* Create and initialize a new sk for the socket returned by accept(). */
566 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
567 struct sctp_association *asoc)
568 {
569 struct inet_sock *inet = inet_sk(sk);
570 struct inet_sock *newinet;
571 struct sock *newsk = sk_alloc(PF_INET, GFP_KERNEL, sk->sk_prot, 1);
573 if (!newsk)
574 goto out;
576 sock_init_data(NULL, newsk);
578 newsk->sk_type = SOCK_STREAM;
580 newsk->sk_no_check = sk->sk_no_check;
581 newsk->sk_reuse = sk->sk_reuse;
582 newsk->sk_shutdown = sk->sk_shutdown;
584 newsk->sk_destruct = inet_sock_destruct;
585 newsk->sk_family = PF_INET;
586 newsk->sk_protocol = IPPROTO_SCTP;
587 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
588 sock_reset_flag(newsk, SOCK_ZAPPED);
590 newinet = inet_sk(newsk);
592 /* Initialize sk's sport, dport, rcv_saddr and daddr for
593 * getsockname() and getpeername()
594 */
595 newinet->sport = inet->sport;
596 newinet->saddr = inet->saddr;
597 newinet->rcv_saddr = inet->rcv_saddr;
598 newinet->dport = htons(asoc->peer.port);
599 newinet->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
600 newinet->pmtudisc = inet->pmtudisc;
601 newinet->id = 0;
603 newinet->uc_ttl = -1;
604 newinet->mc_loop = 1;
605 newinet->mc_ttl = 1;
606 newinet->mc_index = 0;
607 newinet->mc_list = NULL;
609 sk_refcnt_debug_inc(newsk);
611 if (newsk->sk_prot->init(newsk)) {
612 sk_common_release(newsk);
613 newsk = NULL;
614 }
616 out:
617 return newsk;
618 }
620 /* Map address, empty for v4 family */
621 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
622 {
623 /* Empty */
624 }
626 /* Dump the v4 addr to the seq file. */
627 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
628 {
629 seq_printf(seq, "%d.%d.%d.%d ", NIPQUAD(addr->v4.sin_addr));
630 }
632 /* Event handler for inet address addition/deletion events.
633 * Basically, whenever there is an event, we re-build our local address list.
634 */
635 int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
636 void *ptr)
637 {
638 unsigned long flags;
640 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
641 __sctp_free_local_addr_list();
642 __sctp_get_local_addr_list();
643 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
645 return NOTIFY_DONE;
646 }
648 /*
649 * Initialize the control inode/socket with a control endpoint data
650 * structure. This endpoint is reserved exclusively for the OOTB processing.
651 */
652 static int sctp_ctl_sock_init(void)
653 {
654 int err;
655 sa_family_t family;
657 if (sctp_get_pf_specific(PF_INET6))
658 family = PF_INET6;
659 else
660 family = PF_INET;
662 err = sock_create_kern(family, SOCK_SEQPACKET, IPPROTO_SCTP,
663 &sctp_ctl_socket);
664 if (err < 0) {
665 printk(KERN_ERR
666 "SCTP: Failed to create the SCTP control socket.\n");
667 return err;
668 }
669 sctp_ctl_socket->sk->sk_allocation = GFP_ATOMIC;
670 inet_sk(sctp_ctl_socket->sk)->uc_ttl = -1;
672 return 0;
673 }
675 /* Register address family specific functions. */
676 int sctp_register_af(struct sctp_af *af)
677 {
678 switch (af->sa_family) {
679 case AF_INET:
680 if (sctp_af_v4_specific)
681 return 0;
682 sctp_af_v4_specific = af;
683 break;
684 case AF_INET6:
685 if (sctp_af_v6_specific)
686 return 0;
687 sctp_af_v6_specific = af;
688 break;
689 default:
690 return 0;
691 }
693 INIT_LIST_HEAD(&af->list);
694 list_add_tail(&af->list, &sctp_address_families);
695 return 1;
696 }
698 /* Get the table of functions for manipulating a particular address
699 * family.
700 */
701 struct sctp_af *sctp_get_af_specific(sa_family_t family)
702 {
703 switch (family) {
704 case AF_INET:
705 return sctp_af_v4_specific;
706 case AF_INET6:
707 return sctp_af_v6_specific;
708 default:
709 return NULL;
710 }
711 }
713 /* Common code to initialize a AF_INET msg_name. */
714 static void sctp_inet_msgname(char *msgname, int *addr_len)
715 {
716 struct sockaddr_in *sin;
718 sin = (struct sockaddr_in *)msgname;
719 *addr_len = sizeof(struct sockaddr_in);
720 sin->sin_family = AF_INET;
721 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
722 }
724 /* Copy the primary address of the peer primary address as the msg_name. */
725 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
726 int *addr_len)
727 {
728 struct sockaddr_in *sin, *sinfrom;
730 if (msgname) {
731 struct sctp_association *asoc;
733 asoc = event->asoc;
734 sctp_inet_msgname(msgname, addr_len);
735 sin = (struct sockaddr_in *)msgname;
736 sinfrom = &asoc->peer.primary_addr.v4;
737 sin->sin_port = htons(asoc->peer.port);
738 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
739 }
740 }
742 /* Initialize and copy out a msgname from an inbound skb. */
743 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
744 {
745 struct sctphdr *sh;
746 struct sockaddr_in *sin;
748 if (msgname) {
749 sctp_inet_msgname(msgname, len);
750 sin = (struct sockaddr_in *)msgname;
751 sh = (struct sctphdr *)skb->h.raw;
752 sin->sin_port = sh->source;
753 sin->sin_addr.s_addr = skb->nh.iph->saddr;
754 }
755 }
757 /* Do we support this AF? */
758 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
759 {
760 /* PF_INET only supports AF_INET addresses. */
761 return (AF_INET == family);
762 }
764 /* Address matching with wildcards allowed. */
765 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
766 const union sctp_addr *addr2,
767 struct sctp_sock *opt)
768 {
769 /* PF_INET only supports AF_INET addresses. */
770 if (addr1->sa.sa_family != addr2->sa.sa_family)
771 return 0;
772 if (INADDR_ANY == addr1->v4.sin_addr.s_addr ||
773 INADDR_ANY == addr2->v4.sin_addr.s_addr)
774 return 1;
775 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
776 return 1;
778 return 0;
779 }
781 /* Verify that provided sockaddr looks bindable. Common verification has
782 * already been taken care of.
783 */
784 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
785 {
786 return sctp_v4_available(addr, opt);
787 }
789 /* Verify that sockaddr looks sendable. Common verification has already
790 * been taken care of.
791 */
792 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
793 {
794 return 1;
795 }
797 /* Fill in Supported Address Type information for INIT and INIT-ACK
798 * chunks. Returns number of addresses supported.
799 */
800 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
801 __u16 *types)
802 {
803 types[0] = SCTP_PARAM_IPV4_ADDRESS;
804 return 1;
805 }
807 /* Wrapper routine that calls the ip transmit routine. */
808 static inline int sctp_v4_xmit(struct sk_buff *skb,
809 struct sctp_transport *transport, int ipfragok)
810 {
811 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, "
812 "src:%u.%u.%u.%u, dst:%u.%u.%u.%u\n",
813 __FUNCTION__, skb, skb->len,
814 NIPQUAD(((struct rtable *)skb->dst)->rt_src),
815 NIPQUAD(((struct rtable *)skb->dst)->rt_dst));
817 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
818 return ip_queue_xmit(skb, ipfragok);
819 }
821 static struct sctp_af sctp_ipv4_specific;
823 static struct sctp_pf sctp_pf_inet = {
824 .event_msgname = sctp_inet_event_msgname,
825 .skb_msgname = sctp_inet_skb_msgname,
826 .af_supported = sctp_inet_af_supported,
827 .cmp_addr = sctp_inet_cmp_addr,
828 .bind_verify = sctp_inet_bind_verify,
829 .send_verify = sctp_inet_send_verify,
830 .supported_addrs = sctp_inet_supported_addrs,
831 .create_accept_sk = sctp_v4_create_accept_sk,
832 .addr_v4map = sctp_v4_addr_v4map,
833 .af = &sctp_ipv4_specific,
834 };
836 /* Notifier for inetaddr addition/deletion events. */
837 static struct notifier_block sctp_inetaddr_notifier = {
838 .notifier_call = sctp_inetaddr_event,
839 };
841 /* Socket operations. */
842 static const struct proto_ops inet_seqpacket_ops = {
843 .family = PF_INET,
844 .owner = THIS_MODULE,
845 .release = inet_release, /* Needs to be wrapped... */
846 .bind = inet_bind,
847 .connect = inet_dgram_connect,
848 .socketpair = sock_no_socketpair,
849 .accept = inet_accept,
850 .getname = inet_getname, /* Semantics are different. */
851 .poll = sctp_poll,
852 .ioctl = inet_ioctl,
853 .listen = sctp_inet_listen,
854 .shutdown = inet_shutdown, /* Looks harmless. */
855 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
856 .getsockopt = sock_common_getsockopt,
857 .sendmsg = inet_sendmsg,
858 .recvmsg = sock_common_recvmsg,
859 .mmap = sock_no_mmap,
860 .sendpage = sock_no_sendpage,
861 #ifdef CONFIG_COMPAT
862 .compat_setsockopt = compat_sock_common_setsockopt,
863 .compat_getsockopt = compat_sock_common_getsockopt,
864 #endif
865 };
867 /* Registration with AF_INET family. */
868 static struct inet_protosw sctp_seqpacket_protosw = {
869 .type = SOCK_SEQPACKET,
870 .protocol = IPPROTO_SCTP,
871 .prot = &sctp_prot,
872 .ops = &inet_seqpacket_ops,
873 .capability = -1,
874 .no_check = 0,
875 .flags = SCTP_PROTOSW_FLAG
876 };
877 static struct inet_protosw sctp_stream_protosw = {
878 .type = SOCK_STREAM,
879 .protocol = IPPROTO_SCTP,
880 .prot = &sctp_prot,
881 .ops = &inet_seqpacket_ops,
882 .capability = -1,
883 .no_check = 0,
884 .flags = SCTP_PROTOSW_FLAG
885 };
887 /* Register with IP layer. */
888 static struct net_protocol sctp_protocol = {
889 .handler = sctp_rcv,
890 .err_handler = sctp_v4_err,
891 .no_policy = 1,
892 };
894 /* IPv4 address related functions. */
895 static struct sctp_af sctp_ipv4_specific = {
896 .sa_family = AF_INET,
897 .sctp_xmit = sctp_v4_xmit,
898 .setsockopt = ip_setsockopt,
899 .getsockopt = ip_getsockopt,
900 .get_dst = sctp_v4_get_dst,
901 .get_saddr = sctp_v4_get_saddr,
902 .copy_addrlist = sctp_v4_copy_addrlist,
903 .from_skb = sctp_v4_from_skb,
904 .from_sk = sctp_v4_from_sk,
905 .to_sk_saddr = sctp_v4_to_sk_saddr,
906 .to_sk_daddr = sctp_v4_to_sk_daddr,
907 .from_addr_param = sctp_v4_from_addr_param,
908 .to_addr_param = sctp_v4_to_addr_param,
909 .dst_saddr = sctp_v4_dst_saddr,
910 .cmp_addr = sctp_v4_cmp_addr,
911 .addr_valid = sctp_v4_addr_valid,
912 .inaddr_any = sctp_v4_inaddr_any,
913 .is_any = sctp_v4_is_any,
914 .available = sctp_v4_available,
915 .scope = sctp_v4_scope,
916 .skb_iif = sctp_v4_skb_iif,
917 .is_ce = sctp_v4_is_ce,
918 .seq_dump_addr = sctp_v4_seq_dump_addr,
919 .net_header_len = sizeof(struct iphdr),
920 .sockaddr_len = sizeof(struct sockaddr_in),
921 #ifdef CONFIG_COMPAT
922 .compat_setsockopt = compat_ip_setsockopt,
923 .compat_getsockopt = compat_ip_getsockopt,
924 #endif
925 };
927 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
929 switch (family) {
930 case PF_INET:
931 return sctp_pf_inet_specific;
932 case PF_INET6:
933 return sctp_pf_inet6_specific;
934 default:
935 return NULL;
936 }
937 }
939 /* Register the PF specific function table. */
940 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
941 {
942 switch (family) {
943 case PF_INET:
944 if (sctp_pf_inet_specific)
945 return 0;
946 sctp_pf_inet_specific = pf;
947 break;
948 case PF_INET6:
949 if (sctp_pf_inet6_specific)
950 return 0;
951 sctp_pf_inet6_specific = pf;
952 break;
953 default:
954 return 0;
955 }
956 return 1;
957 }
959 static int __init init_sctp_mibs(void)
960 {
961 sctp_statistics[0] = alloc_percpu(struct sctp_mib);
962 if (!sctp_statistics[0])
963 return -ENOMEM;
964 sctp_statistics[1] = alloc_percpu(struct sctp_mib);
965 if (!sctp_statistics[1]) {
966 free_percpu(sctp_statistics[0]);
967 return -ENOMEM;
968 }
969 return 0;
971 }
973 static void cleanup_sctp_mibs(void)
974 {
975 free_percpu(sctp_statistics[0]);
976 free_percpu(sctp_statistics[1]);
977 }
979 /* Initialize the universe into something sensible. */
980 SCTP_STATIC __init int sctp_init(void)
981 {
982 int i;
983 int status = -EINVAL;
984 unsigned long goal;
985 int order;
987 /* SCTP_DEBUG sanity check. */
988 if (!sctp_sanity_check())
989 goto out;
991 status = proto_register(&sctp_prot, 1);
992 if (status)
993 goto out;
995 /* Add SCTP to inet_protos hash table. */
996 status = -EAGAIN;
997 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
998 goto err_add_protocol;
1000 /* Add SCTP(TCP and UDP style) to inetsw linked list. */
1001 inet_register_protosw(&sctp_seqpacket_protosw);
1002 inet_register_protosw(&sctp_stream_protosw);
1004 /* Allocate a cache pools. */
1005 status = -ENOBUFS;
1006 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1007 sizeof(struct sctp_bind_bucket),
1008 0, SLAB_HWCACHE_ALIGN,
1009 NULL, NULL);
1011 if (!sctp_bucket_cachep)
1012 goto err_bucket_cachep;
1014 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1015 sizeof(struct sctp_chunk),
1016 0, SLAB_HWCACHE_ALIGN,
1017 NULL, NULL);
1018 if (!sctp_chunk_cachep)
1019 goto err_chunk_cachep;
1021 /* Allocate and initialise sctp mibs. */
1022 status = init_sctp_mibs();
1023 if (status)
1024 goto err_init_mibs;
1026 /* Initialize proc fs directory. */
1027 status = sctp_proc_init();
1028 if (status)
1029 goto err_init_proc;
1031 /* Initialize object count debugging. */
1032 sctp_dbg_objcnt_init();
1034 /* Initialize the SCTP specific PF functions. */
1035 sctp_register_pf(&sctp_pf_inet, PF_INET);
1036 /*
1037 * 14. Suggested SCTP Protocol Parameter Values
1038 */
1039 /* The following protocol parameters are RECOMMENDED: */
1040 /* RTO.Initial - 3 seconds */
1041 sctp_rto_initial = SCTP_RTO_INITIAL;
1042 /* RTO.Min - 1 second */
1043 sctp_rto_min = SCTP_RTO_MIN;
1044 /* RTO.Max - 60 seconds */
1045 sctp_rto_max = SCTP_RTO_MAX;
1046 /* RTO.Alpha - 1/8 */
1047 sctp_rto_alpha = SCTP_RTO_ALPHA;
1048 /* RTO.Beta - 1/4 */
1049 sctp_rto_beta = SCTP_RTO_BETA;
1051 /* Valid.Cookie.Life - 60 seconds */
1052 sctp_valid_cookie_life = 60 * HZ;
1054 /* Whether Cookie Preservative is enabled(1) or not(0) */
1055 sctp_cookie_preserve_enable = 1;
1057 /* Max.Burst - 4 */
1058 sctp_max_burst = SCTP_MAX_BURST;
1060 /* Association.Max.Retrans - 10 attempts
1061 * Path.Max.Retrans - 5 attempts (per destination address)
1062 * Max.Init.Retransmits - 8 attempts
1063 */
1064 sctp_max_retrans_association = 10;
1065 sctp_max_retrans_path = 5;
1066 sctp_max_retrans_init = 8;
1068 /* Sendbuffer growth - do per-socket accounting */
1069 sctp_sndbuf_policy = 0;
1071 /* Rcvbuffer growth - do per-socket accounting */
1072 sctp_rcvbuf_policy = 0;
1074 /* HB.interval - 30 seconds */
1075 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1077 /* delayed SACK timeout */
1078 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1080 /* Implementation specific variables. */
1082 /* Initialize default stream count setup information. */
1083 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1084 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1086 /* Initialize handle used for association ids. */
1087 idr_init(&sctp_assocs_id);
1089 /* Size and allocate the association hash table.
1090 * The methodology is similar to that of the tcp hash tables.
1091 */
1092 if (num_physpages >= (128 * 1024))
1093 goal = num_physpages >> (22 - PAGE_SHIFT);
1094 else
1095 goal = num_physpages >> (24 - PAGE_SHIFT);
1097 for (order = 0; (1UL << order) < goal; order++)
1100 do {
1101 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1102 sizeof(struct sctp_hashbucket);
1103 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1104 continue;
1105 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1106 __get_free_pages(GFP_ATOMIC, order);
1107 } while (!sctp_assoc_hashtable && --order > 0);
1108 if (!sctp_assoc_hashtable) {
1109 printk(KERN_ERR "SCTP: Failed association hash alloc.\n");
1110 status = -ENOMEM;
1111 goto err_ahash_alloc;
1113 for (i = 0; i < sctp_assoc_hashsize; i++) {
1114 rwlock_init(&sctp_assoc_hashtable[i].lock);
1115 sctp_assoc_hashtable[i].chain = NULL;
1118 /* Allocate and initialize the endpoint hash table. */
1119 sctp_ep_hashsize = 64;
1120 sctp_ep_hashtable = (struct sctp_hashbucket *)
1121 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1122 if (!sctp_ep_hashtable) {
1123 printk(KERN_ERR "SCTP: Failed endpoint_hash alloc.\n");
1124 status = -ENOMEM;
1125 goto err_ehash_alloc;
1127 for (i = 0; i < sctp_ep_hashsize; i++) {
1128 rwlock_init(&sctp_ep_hashtable[i].lock);
1129 sctp_ep_hashtable[i].chain = NULL;
1132 /* Allocate and initialize the SCTP port hash table. */
1133 do {
1134 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1135 sizeof(struct sctp_bind_hashbucket);
1136 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1137 continue;
1138 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1139 __get_free_pages(GFP_ATOMIC, order);
1140 } while (!sctp_port_hashtable && --order > 0);
1141 if (!sctp_port_hashtable) {
1142 printk(KERN_ERR "SCTP: Failed bind hash alloc.");
1143 status = -ENOMEM;
1144 goto err_bhash_alloc;
1146 for (i = 0; i < sctp_port_hashsize; i++) {
1147 spin_lock_init(&sctp_port_hashtable[i].lock);
1148 sctp_port_hashtable[i].chain = NULL;
1151 spin_lock_init(&sctp_port_alloc_lock);
1152 sctp_port_rover = sysctl_local_port_range[0] - 1;
1154 printk(KERN_INFO "SCTP: Hash tables configured "
1155 "(established %d bind %d)\n",
1156 sctp_assoc_hashsize, sctp_port_hashsize);
1158 /* Disable ADDIP by default. */
1159 sctp_addip_enable = 0;
1161 /* Enable PR-SCTP by default. */
1162 sctp_prsctp_enable = 1;
1164 sctp_sysctl_register();
1166 INIT_LIST_HEAD(&sctp_address_families);
1167 sctp_register_af(&sctp_ipv4_specific);
1169 status = sctp_v6_init();
1170 if (status)
1171 goto err_v6_init;
1173 /* Initialize the control inode/socket for handling OOTB packets. */
1174 if ((status = sctp_ctl_sock_init())) {
1175 printk (KERN_ERR
1176 "SCTP: Failed to initialize the SCTP control sock.\n");
1177 goto err_ctl_sock_init;
1180 /* Initialize the local address list. */
1181 INIT_LIST_HEAD(&sctp_local_addr_list);
1182 spin_lock_init(&sctp_local_addr_lock);
1184 /* Register notifier for inet address additions/deletions. */
1185 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1187 sctp_get_local_addr_list();
1189 __unsafe(THIS_MODULE);
1190 status = 0;
1191 out:
1192 return status;
1193 err_ctl_sock_init:
1194 sctp_v6_exit();
1195 err_v6_init:
1196 sctp_sysctl_unregister();
1197 list_del(&sctp_ipv4_specific.list);
1198 free_pages((unsigned long)sctp_port_hashtable,
1199 get_order(sctp_port_hashsize *
1200 sizeof(struct sctp_bind_hashbucket)));
1201 err_bhash_alloc:
1202 kfree(sctp_ep_hashtable);
1203 err_ehash_alloc:
1204 free_pages((unsigned long)sctp_assoc_hashtable,
1205 get_order(sctp_assoc_hashsize *
1206 sizeof(struct sctp_hashbucket)));
1207 err_ahash_alloc:
1208 sctp_dbg_objcnt_exit();
1209 err_init_proc:
1210 sctp_proc_exit();
1211 cleanup_sctp_mibs();
1212 err_init_mibs:
1213 kmem_cache_destroy(sctp_chunk_cachep);
1214 err_chunk_cachep:
1215 kmem_cache_destroy(sctp_bucket_cachep);
1216 err_bucket_cachep:
1217 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1218 inet_unregister_protosw(&sctp_seqpacket_protosw);
1219 inet_unregister_protosw(&sctp_stream_protosw);
1220 err_add_protocol:
1221 proto_unregister(&sctp_prot);
1222 goto out;
1225 /* Exit handler for the SCTP protocol. */
1226 SCTP_STATIC __exit void sctp_exit(void)
1228 /* BUG. This should probably do something useful like clean
1229 * up all the remaining associations and all that memory.
1230 */
1232 /* Unregister notifier for inet address additions/deletions. */
1233 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1235 /* Free the local address list. */
1236 sctp_free_local_addr_list();
1238 /* Free the control endpoint. */
1239 sock_release(sctp_ctl_socket);
1241 sctp_v6_exit();
1242 sctp_sysctl_unregister();
1243 list_del(&sctp_ipv4_specific.list);
1245 free_pages((unsigned long)sctp_assoc_hashtable,
1246 get_order(sctp_assoc_hashsize *
1247 sizeof(struct sctp_hashbucket)));
1248 kfree(sctp_ep_hashtable);
1249 free_pages((unsigned long)sctp_port_hashtable,
1250 get_order(sctp_port_hashsize *
1251 sizeof(struct sctp_bind_hashbucket)));
1253 kmem_cache_destroy(sctp_chunk_cachep);
1254 kmem_cache_destroy(sctp_bucket_cachep);
1256 sctp_dbg_objcnt_exit();
1257 sctp_proc_exit();
1258 cleanup_sctp_mibs();
1260 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1261 inet_unregister_protosw(&sctp_seqpacket_protosw);
1262 inet_unregister_protosw(&sctp_stream_protosw);
1263 proto_unregister(&sctp_prot);
1266 module_init(sctp_init);
1267 module_exit(sctp_exit);
1269 /*
1270 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1271 */
1272 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1273 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1274 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1275 MODULE_LICENSE("GPL");