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
9 * This file is part of the SCTP kernel reference Implementation
11 * Initialization/cleanup for SCTP protocol support.
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)
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.
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.
30 * Please send any bug reports or fixes you make to the
32 * lksctp developers <lksctp-developers@lists.sourceforge.net>
34 * Or submit a bug report through the following website:
35 * http://www.sf.net/projects/lksctp
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>
45 * Any bugs reported given to us we will try to fix... any fixes shared will
46 * be incorporated into the next SCTP release.
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 <linux/bootmem.h>
55 #include <net/net_namespace.h>
56 #include <net/protocol.h>
59 #include <net/route.h>
60 #include <net/sctp/sctp.h>
61 #include <net/addrconf.h>
62 #include <net/inet_common.h>
63 #include <net/inet_ecn.h>
65 /* Global data structures. */
66 struct sctp_globals sctp_globals __read_mostly;
67 struct proc_dir_entry *proc_net_sctp;
68 DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
70 struct idr sctp_assocs_id;
71 DEFINE_SPINLOCK(sctp_assocs_id_lock);
73 /* This is the global socket data structure used for responding to
74 * the Out-of-the-blue (OOTB) packets. A control sock will be created
75 * for this socket at the initialization time.
77 static struct socket *sctp_ctl_socket;
79 static struct sctp_pf *sctp_pf_inet6_specific;
80 static struct sctp_pf *sctp_pf_inet_specific;
81 static struct sctp_af *sctp_af_v4_specific;
82 static struct sctp_af *sctp_af_v6_specific;
84 struct kmem_cache *sctp_chunk_cachep __read_mostly;
85 struct kmem_cache *sctp_bucket_cachep __read_mostly;
87 extern int sysctl_sctp_mem[3];
88 extern int sysctl_sctp_rmem[3];
89 extern int sysctl_sctp_wmem[3];
91 /* Return the address of the control sock. */
92 struct sock *sctp_get_ctl_sock(void)
94 return sctp_ctl_socket->sk;
97 /* Set up the proc fs entry for the SCTP protocol. */
98 static __init int sctp_proc_init(void)
100 if (!proc_net_sctp) {
101 struct proc_dir_entry *ent;
102 ent = proc_mkdir("sctp", init_net.proc_net);
104 ent->owner = THIS_MODULE;
110 if (sctp_snmp_proc_init())
112 if (sctp_eps_proc_init())
114 if (sctp_assocs_proc_init())
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
127 static void sctp_proc_exit(void)
129 sctp_snmp_proc_exit();
130 sctp_eps_proc_exit();
131 sctp_assocs_proc_exit();
134 proc_net_sctp = NULL;
135 remove_proc_entry("sctp", init_net.proc_net);
139 /* Private helper to extract ipv4 address and stash them in
140 * the protocol structure.
142 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
143 struct net_device *dev)
145 struct in_device *in_dev;
146 struct in_ifaddr *ifa;
147 struct sctp_sockaddr_entry *addr;
150 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
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);
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;
163 INIT_LIST_HEAD(&addr->list);
164 INIT_RCU_HEAD(&addr->rcu);
165 list_add_tail(&addr->list, addrlist);
172 /* Extract our IP addresses from the system and stash them in the
173 * protocol structure.
175 static void sctp_get_local_addr_list(void)
177 struct net_device *dev;
178 struct list_head *pos;
181 read_lock(&dev_base_lock);
182 for_each_netdev(dev) {
183 __list_for_each(pos, &sctp_address_families) {
184 af = list_entry(pos, struct sctp_af, list);
185 af->copy_addrlist(&sctp_local_addr_list, dev);
188 read_unlock(&dev_base_lock);
191 /* Free the existing local addresses. */
192 static void sctp_free_local_addr_list(void)
194 struct sctp_sockaddr_entry *addr;
195 struct list_head *pos, *temp;
197 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
198 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
204 void sctp_local_addr_free(struct rcu_head *head)
206 struct sctp_sockaddr_entry *e = container_of(head,
207 struct sctp_sockaddr_entry, rcu);
211 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
212 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
213 gfp_t gfp, int copy_flags)
215 struct sctp_sockaddr_entry *addr;
219 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
222 if (sctp_in_scope(&addr->a, scope)) {
223 /* Now that the address is in scope, check to see if
224 * the address type is really supported by the local
225 * sock as well as the remote peer.
227 if ((((AF_INET == addr->a.sa.sa_family) &&
228 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
229 (((AF_INET6 == addr->a.sa.sa_family) &&
230 (copy_flags & SCTP_ADDR6_ALLOWED) &&
231 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
232 error = sctp_add_bind_addr(bp, &addr->a, 1,
245 /* Initialize a sctp_addr from in incoming skb. */
246 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
253 port = &addr->v4.sin_port;
254 addr->v4.sin_family = AF_INET;
259 from = &ip_hdr(skb)->saddr;
262 from = &ip_hdr(skb)->daddr;
264 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
267 /* Initialize an sctp_addr from a socket. */
268 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
270 addr->v4.sin_family = AF_INET;
271 addr->v4.sin_port = 0;
272 addr->v4.sin_addr.s_addr = inet_sk(sk)->rcv_saddr;
275 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
276 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
278 inet_sk(sk)->rcv_saddr = addr->v4.sin_addr.s_addr;
281 /* Initialize sk->sk_daddr from sctp_addr. */
282 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
284 inet_sk(sk)->daddr = addr->v4.sin_addr.s_addr;
287 /* Initialize a sctp_addr from an address parameter. */
288 static void sctp_v4_from_addr_param(union sctp_addr *addr,
289 union sctp_addr_param *param,
290 __be16 port, int iif)
292 addr->v4.sin_family = AF_INET;
293 addr->v4.sin_port = port;
294 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
297 /* Initialize an address parameter from a sctp_addr and return the length
298 * of the address parameter.
300 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
301 union sctp_addr_param *param)
303 int length = sizeof(sctp_ipv4addr_param_t);
305 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
306 param->v4.param_hdr.length = htons(length);
307 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
312 /* Initialize a sctp_addr from a dst_entry. */
313 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct dst_entry *dst,
316 struct rtable *rt = (struct rtable *)dst;
317 saddr->v4.sin_family = AF_INET;
318 saddr->v4.sin_port = port;
319 saddr->v4.sin_addr.s_addr = rt->rt_src;
322 /* Compare two addresses exactly. */
323 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
324 const union sctp_addr *addr2)
326 if (addr1->sa.sa_family != addr2->sa.sa_family)
328 if (addr1->v4.sin_port != addr2->v4.sin_port)
330 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
336 /* Initialize addr struct to INADDR_ANY. */
337 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
339 addr->v4.sin_family = AF_INET;
340 addr->v4.sin_addr.s_addr = INADDR_ANY;
341 addr->v4.sin_port = port;
344 /* Is this a wildcard address? */
345 static int sctp_v4_is_any(const union sctp_addr *addr)
347 return INADDR_ANY == addr->v4.sin_addr.s_addr;
350 /* This function checks if the address is a valid address to be used for
354 * Return 0 - If the address is a non-unicast or an illegal address.
355 * Return 1 - If the address is a unicast.
357 static int sctp_v4_addr_valid(union sctp_addr *addr,
358 struct sctp_sock *sp,
359 const struct sk_buff *skb)
361 /* Is this a non-unicast address or a unusable SCTP address? */
362 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr))
365 /* Is this a broadcast address? */
366 if (skb && ((struct rtable *)skb->dst)->rt_flags & RTCF_BROADCAST)
372 /* Should this be available for binding? */
373 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
375 int ret = inet_addr_type(addr->v4.sin_addr.s_addr);
378 if (addr->v4.sin_addr.s_addr != INADDR_ANY &&
380 !sp->inet.freebind &&
381 !sysctl_ip_nonlocal_bind)
387 /* Checking the loopback, private and other address scopes as defined in
388 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
389 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
391 * Level 0 - unusable SCTP addresses
392 * Level 1 - loopback address
393 * Level 2 - link-local addresses
394 * Level 3 - private addresses.
395 * Level 4 - global addresses
396 * For INIT and INIT-ACK address list, let L be the level of
397 * of requested destination address, sender and receiver
398 * SHOULD include all of its addresses with level greater
399 * than or equal to L.
401 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
405 /* Should IPv4 scoping be a sysctl configurable option
406 * so users can turn it off (default on) for certain
407 * unconventional networking environments?
410 /* Check for unusable SCTP addresses. */
411 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
412 retval = SCTP_SCOPE_UNUSABLE;
413 } else if (LOOPBACK(addr->v4.sin_addr.s_addr)) {
414 retval = SCTP_SCOPE_LOOPBACK;
415 } else if (IS_IPV4_LINK_ADDRESS(&addr->v4.sin_addr.s_addr)) {
416 retval = SCTP_SCOPE_LINK;
417 } else if (IS_IPV4_PRIVATE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
418 retval = SCTP_SCOPE_PRIVATE;
420 retval = SCTP_SCOPE_GLOBAL;
426 /* Returns a valid dst cache entry for the given source and destination ip
427 * addresses. If an association is passed, trys to get a dst entry with a
428 * source address that matches an address in the bind address list.
430 static struct dst_entry *sctp_v4_get_dst(struct sctp_association *asoc,
431 union sctp_addr *daddr,
432 union sctp_addr *saddr)
436 struct sctp_bind_addr *bp;
437 struct sctp_sockaddr_entry *laddr;
438 struct dst_entry *dst = NULL;
439 union sctp_addr dst_saddr;
441 memset(&fl, 0x0, sizeof(struct flowi));
442 fl.fl4_dst = daddr->v4.sin_addr.s_addr;
443 fl.proto = IPPROTO_SCTP;
445 fl.fl4_tos = RT_CONN_FLAGS(asoc->base.sk);
446 fl.oif = asoc->base.sk->sk_bound_dev_if;
449 fl.fl4_src = saddr->v4.sin_addr.s_addr;
451 SCTP_DEBUG_PRINTK("%s: DST:%u.%u.%u.%u, SRC:%u.%u.%u.%u - ",
452 __FUNCTION__, NIPQUAD(fl.fl4_dst),
453 NIPQUAD(fl.fl4_src));
455 if (!ip_route_output_key(&rt, &fl)) {
459 /* If there is no association or if a source address is passed, no
460 * more validation is required.
465 bp = &asoc->base.bind_addr;
468 /* Walk through the bind address list and look for a bind
469 * address that matches the source address of the returned dst.
472 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
473 if (!laddr->valid || !laddr->use_as_src)
475 sctp_v4_dst_saddr(&dst_saddr, dst, htons(bp->port));
476 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
481 /* None of the bound addresses match the source address of the
482 * dst. So release it.
488 /* Walk through the bind address list and try to get a dst that
489 * matches a bind address as the source address.
492 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
495 if ((laddr->use_as_src) &&
496 (AF_INET == laddr->a.sa.sa_family)) {
497 fl.fl4_src = laddr->a.v4.sin_addr.s_addr;
498 if (!ip_route_output_key(&rt, &fl)) {
509 SCTP_DEBUG_PRINTK("rt_dst:%u.%u.%u.%u, rt_src:%u.%u.%u.%u\n",
510 NIPQUAD(rt->rt_dst), NIPQUAD(rt->rt_src));
512 SCTP_DEBUG_PRINTK("NO ROUTE\n");
517 /* For v4, the source address is cached in the route entry(dst). So no need
518 * to cache it separately and hence this is an empty routine.
520 static void sctp_v4_get_saddr(struct sctp_association *asoc,
521 struct dst_entry *dst,
522 union sctp_addr *daddr,
523 union sctp_addr *saddr)
525 struct rtable *rt = (struct rtable *)dst;
531 saddr->v4.sin_family = AF_INET;
532 saddr->v4.sin_port = htons(asoc->base.bind_addr.port);
533 saddr->v4.sin_addr.s_addr = rt->rt_src;
537 /* What interface did this skb arrive on? */
538 static int sctp_v4_skb_iif(const struct sk_buff *skb)
540 return ((struct rtable *)skb->dst)->rt_iif;
543 /* Was this packet marked by Explicit Congestion Notification? */
544 static int sctp_v4_is_ce(const struct sk_buff *skb)
546 return INET_ECN_is_ce(ip_hdr(skb)->tos);
549 /* Create and initialize a new sk for the socket returned by accept(). */
550 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
551 struct sctp_association *asoc)
553 struct inet_sock *inet = inet_sk(sk);
554 struct inet_sock *newinet;
555 struct sock *newsk = sk_alloc(sk->sk_net, PF_INET, GFP_KERNEL, sk->sk_prot, 1);
560 sock_init_data(NULL, newsk);
562 newsk->sk_type = SOCK_STREAM;
564 newsk->sk_no_check = sk->sk_no_check;
565 newsk->sk_reuse = sk->sk_reuse;
566 newsk->sk_shutdown = sk->sk_shutdown;
568 newsk->sk_destruct = inet_sock_destruct;
569 newsk->sk_family = PF_INET;
570 newsk->sk_protocol = IPPROTO_SCTP;
571 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
572 sock_reset_flag(newsk, SOCK_ZAPPED);
574 newinet = inet_sk(newsk);
576 /* Initialize sk's sport, dport, rcv_saddr and daddr for
577 * getsockname() and getpeername()
579 newinet->sport = inet->sport;
580 newinet->saddr = inet->saddr;
581 newinet->rcv_saddr = inet->rcv_saddr;
582 newinet->dport = htons(asoc->peer.port);
583 newinet->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
584 newinet->pmtudisc = inet->pmtudisc;
585 newinet->id = asoc->next_tsn ^ jiffies;
587 newinet->uc_ttl = -1;
588 newinet->mc_loop = 1;
590 newinet->mc_index = 0;
591 newinet->mc_list = NULL;
593 sk_refcnt_debug_inc(newsk);
595 if (newsk->sk_prot->init(newsk)) {
596 sk_common_release(newsk);
604 /* Map address, empty for v4 family */
605 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
610 /* Dump the v4 addr to the seq file. */
611 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
613 seq_printf(seq, "%d.%d.%d.%d ", NIPQUAD(addr->v4.sin_addr));
616 /* Event handler for inet address addition/deletion events.
617 * The sctp_local_addr_list needs to be protocted by a spin lock since
618 * multiple notifiers (say IPv4 and IPv6) may be running at the same
619 * time and thus corrupt the list.
620 * The reader side is protected with RCU.
622 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
625 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
626 struct sctp_sockaddr_entry *addr = NULL;
627 struct sctp_sockaddr_entry *temp;
631 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
633 addr->a.v4.sin_family = AF_INET;
634 addr->a.v4.sin_port = 0;
635 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
637 spin_lock_bh(&sctp_local_addr_lock);
638 list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
639 spin_unlock_bh(&sctp_local_addr_lock);
643 spin_lock_bh(&sctp_local_addr_lock);
644 list_for_each_entry_safe(addr, temp,
645 &sctp_local_addr_list, list) {
646 if (addr->a.v4.sin_addr.s_addr == ifa->ifa_local) {
648 list_del_rcu(&addr->list);
652 spin_unlock_bh(&sctp_local_addr_lock);
653 if (addr && !addr->valid)
654 call_rcu(&addr->rcu, sctp_local_addr_free);
662 * Initialize the control inode/socket with a control endpoint data
663 * structure. This endpoint is reserved exclusively for the OOTB processing.
665 static int sctp_ctl_sock_init(void)
670 if (sctp_get_pf_specific(PF_INET6))
675 err = sock_create_kern(family, SOCK_SEQPACKET, IPPROTO_SCTP,
679 "SCTP: Failed to create the SCTP control socket.\n");
682 sctp_ctl_socket->sk->sk_allocation = GFP_ATOMIC;
683 inet_sk(sctp_ctl_socket->sk)->uc_ttl = -1;
688 /* Register address family specific functions. */
689 int sctp_register_af(struct sctp_af *af)
691 switch (af->sa_family) {
693 if (sctp_af_v4_specific)
695 sctp_af_v4_specific = af;
698 if (sctp_af_v6_specific)
700 sctp_af_v6_specific = af;
706 INIT_LIST_HEAD(&af->list);
707 list_add_tail(&af->list, &sctp_address_families);
711 /* Get the table of functions for manipulating a particular address
714 struct sctp_af *sctp_get_af_specific(sa_family_t family)
718 return sctp_af_v4_specific;
720 return sctp_af_v6_specific;
726 /* Common code to initialize a AF_INET msg_name. */
727 static void sctp_inet_msgname(char *msgname, int *addr_len)
729 struct sockaddr_in *sin;
731 sin = (struct sockaddr_in *)msgname;
732 *addr_len = sizeof(struct sockaddr_in);
733 sin->sin_family = AF_INET;
734 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
737 /* Copy the primary address of the peer primary address as the msg_name. */
738 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
741 struct sockaddr_in *sin, *sinfrom;
744 struct sctp_association *asoc;
747 sctp_inet_msgname(msgname, addr_len);
748 sin = (struct sockaddr_in *)msgname;
749 sinfrom = &asoc->peer.primary_addr.v4;
750 sin->sin_port = htons(asoc->peer.port);
751 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
755 /* Initialize and copy out a msgname from an inbound skb. */
756 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
759 struct sctphdr *sh = sctp_hdr(skb);
760 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
762 sctp_inet_msgname(msgname, len);
763 sin->sin_port = sh->source;
764 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
768 /* Do we support this AF? */
769 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
771 /* PF_INET only supports AF_INET addresses. */
772 return (AF_INET == family);
775 /* Address matching with wildcards allowed. */
776 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
777 const union sctp_addr *addr2,
778 struct sctp_sock *opt)
780 /* PF_INET only supports AF_INET addresses. */
781 if (addr1->sa.sa_family != addr2->sa.sa_family)
783 if (INADDR_ANY == addr1->v4.sin_addr.s_addr ||
784 INADDR_ANY == addr2->v4.sin_addr.s_addr)
786 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
792 /* Verify that provided sockaddr looks bindable. Common verification has
793 * already been taken care of.
795 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
797 return sctp_v4_available(addr, opt);
800 /* Verify that sockaddr looks sendable. Common verification has already
801 * been taken care of.
803 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
808 /* Fill in Supported Address Type information for INIT and INIT-ACK
809 * chunks. Returns number of addresses supported.
811 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
814 types[0] = SCTP_PARAM_IPV4_ADDRESS;
818 /* Wrapper routine that calls the ip transmit routine. */
819 static inline int sctp_v4_xmit(struct sk_buff *skb,
820 struct sctp_transport *transport, int ipfragok)
822 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, "
823 "src:%u.%u.%u.%u, dst:%u.%u.%u.%u\n",
824 __FUNCTION__, skb, skb->len,
825 NIPQUAD(((struct rtable *)skb->dst)->rt_src),
826 NIPQUAD(((struct rtable *)skb->dst)->rt_dst));
828 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
829 return ip_queue_xmit(skb, ipfragok);
832 static struct sctp_af sctp_ipv4_specific;
834 static struct sctp_pf sctp_pf_inet = {
835 .event_msgname = sctp_inet_event_msgname,
836 .skb_msgname = sctp_inet_skb_msgname,
837 .af_supported = sctp_inet_af_supported,
838 .cmp_addr = sctp_inet_cmp_addr,
839 .bind_verify = sctp_inet_bind_verify,
840 .send_verify = sctp_inet_send_verify,
841 .supported_addrs = sctp_inet_supported_addrs,
842 .create_accept_sk = sctp_v4_create_accept_sk,
843 .addr_v4map = sctp_v4_addr_v4map,
844 .af = &sctp_ipv4_specific,
847 /* Notifier for inetaddr addition/deletion events. */
848 static struct notifier_block sctp_inetaddr_notifier = {
849 .notifier_call = sctp_inetaddr_event,
852 /* Socket operations. */
853 static const struct proto_ops inet_seqpacket_ops = {
855 .owner = THIS_MODULE,
856 .release = inet_release, /* Needs to be wrapped... */
858 .connect = inet_dgram_connect,
859 .socketpair = sock_no_socketpair,
860 .accept = inet_accept,
861 .getname = inet_getname, /* Semantics are different. */
864 .listen = sctp_inet_listen,
865 .shutdown = inet_shutdown, /* Looks harmless. */
866 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
867 .getsockopt = sock_common_getsockopt,
868 .sendmsg = inet_sendmsg,
869 .recvmsg = sock_common_recvmsg,
870 .mmap = sock_no_mmap,
871 .sendpage = sock_no_sendpage,
873 .compat_setsockopt = compat_sock_common_setsockopt,
874 .compat_getsockopt = compat_sock_common_getsockopt,
878 /* Registration with AF_INET family. */
879 static struct inet_protosw sctp_seqpacket_protosw = {
880 .type = SOCK_SEQPACKET,
881 .protocol = IPPROTO_SCTP,
883 .ops = &inet_seqpacket_ops,
886 .flags = SCTP_PROTOSW_FLAG
888 static struct inet_protosw sctp_stream_protosw = {
890 .protocol = IPPROTO_SCTP,
892 .ops = &inet_seqpacket_ops,
895 .flags = SCTP_PROTOSW_FLAG
898 /* Register with IP layer. */
899 static struct net_protocol sctp_protocol = {
901 .err_handler = sctp_v4_err,
905 /* IPv4 address related functions. */
906 static struct sctp_af sctp_ipv4_specific = {
907 .sa_family = AF_INET,
908 .sctp_xmit = sctp_v4_xmit,
909 .setsockopt = ip_setsockopt,
910 .getsockopt = ip_getsockopt,
911 .get_dst = sctp_v4_get_dst,
912 .get_saddr = sctp_v4_get_saddr,
913 .copy_addrlist = sctp_v4_copy_addrlist,
914 .from_skb = sctp_v4_from_skb,
915 .from_sk = sctp_v4_from_sk,
916 .to_sk_saddr = sctp_v4_to_sk_saddr,
917 .to_sk_daddr = sctp_v4_to_sk_daddr,
918 .from_addr_param = sctp_v4_from_addr_param,
919 .to_addr_param = sctp_v4_to_addr_param,
920 .dst_saddr = sctp_v4_dst_saddr,
921 .cmp_addr = sctp_v4_cmp_addr,
922 .addr_valid = sctp_v4_addr_valid,
923 .inaddr_any = sctp_v4_inaddr_any,
924 .is_any = sctp_v4_is_any,
925 .available = sctp_v4_available,
926 .scope = sctp_v4_scope,
927 .skb_iif = sctp_v4_skb_iif,
928 .is_ce = sctp_v4_is_ce,
929 .seq_dump_addr = sctp_v4_seq_dump_addr,
930 .net_header_len = sizeof(struct iphdr),
931 .sockaddr_len = sizeof(struct sockaddr_in),
933 .compat_setsockopt = compat_ip_setsockopt,
934 .compat_getsockopt = compat_ip_getsockopt,
938 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
942 return sctp_pf_inet_specific;
944 return sctp_pf_inet6_specific;
950 /* Register the PF specific function table. */
951 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
955 if (sctp_pf_inet_specific)
957 sctp_pf_inet_specific = pf;
960 if (sctp_pf_inet6_specific)
962 sctp_pf_inet6_specific = pf;
970 static int __init init_sctp_mibs(void)
972 sctp_statistics[0] = alloc_percpu(struct sctp_mib);
973 if (!sctp_statistics[0])
975 sctp_statistics[1] = alloc_percpu(struct sctp_mib);
976 if (!sctp_statistics[1]) {
977 free_percpu(sctp_statistics[0]);
984 static void cleanup_sctp_mibs(void)
986 free_percpu(sctp_statistics[0]);
987 free_percpu(sctp_statistics[1]);
990 /* Initialize the universe into something sensible. */
991 SCTP_STATIC __init int sctp_init(void)
994 int status = -EINVAL;
1000 /* SCTP_DEBUG sanity check. */
1001 if (!sctp_sanity_check())
1004 /* Allocate bind_bucket and chunk caches. */
1006 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1007 sizeof(struct sctp_bind_bucket),
1008 0, SLAB_HWCACHE_ALIGN,
1010 if (!sctp_bucket_cachep)
1013 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1014 sizeof(struct sctp_chunk),
1015 0, SLAB_HWCACHE_ALIGN,
1017 if (!sctp_chunk_cachep)
1018 goto err_chunk_cachep;
1020 /* Allocate and initialise sctp mibs. */
1021 status = init_sctp_mibs();
1025 /* Initialize proc fs directory. */
1026 status = sctp_proc_init();
1030 /* Initialize object count debugging. */
1031 sctp_dbg_objcnt_init();
1033 /* Initialize the SCTP specific PF functions. */
1034 sctp_register_pf(&sctp_pf_inet, PF_INET);
1036 * 14. Suggested SCTP Protocol Parameter Values
1038 /* The following protocol parameters are RECOMMENDED: */
1039 /* RTO.Initial - 3 seconds */
1040 sctp_rto_initial = SCTP_RTO_INITIAL;
1041 /* RTO.Min - 1 second */
1042 sctp_rto_min = SCTP_RTO_MIN;
1043 /* RTO.Max - 60 seconds */
1044 sctp_rto_max = SCTP_RTO_MAX;
1045 /* RTO.Alpha - 1/8 */
1046 sctp_rto_alpha = SCTP_RTO_ALPHA;
1047 /* RTO.Beta - 1/4 */
1048 sctp_rto_beta = SCTP_RTO_BETA;
1050 /* Valid.Cookie.Life - 60 seconds */
1051 sctp_valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1053 /* Whether Cookie Preservative is enabled(1) or not(0) */
1054 sctp_cookie_preserve_enable = 1;
1057 sctp_max_burst = SCTP_DEFAULT_MAX_BURST;
1059 /* Association.Max.Retrans - 10 attempts
1060 * Path.Max.Retrans - 5 attempts (per destination address)
1061 * Max.Init.Retransmits - 8 attempts
1063 sctp_max_retrans_association = 10;
1064 sctp_max_retrans_path = 5;
1065 sctp_max_retrans_init = 8;
1067 /* Sendbuffer growth - do per-socket accounting */
1068 sctp_sndbuf_policy = 0;
1070 /* Rcvbuffer growth - do per-socket accounting */
1071 sctp_rcvbuf_policy = 0;
1073 /* HB.interval - 30 seconds */
1074 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1076 /* delayed SACK timeout */
1077 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1079 /* Implementation specific variables. */
1081 /* Initialize default stream count setup information. */
1082 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1083 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1085 /* Initialize handle used for association ids. */
1086 idr_init(&sctp_assocs_id);
1088 /* Set the pressure threshold to be a fraction of global memory that
1089 * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
1090 * memory, with a floor of 128 pages.
1091 * Note this initalizes the data in sctpv6_prot too
1092 * Unabashedly stolen from tcp_init
1094 limit = min(num_physpages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1095 limit = (limit * (num_physpages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
1096 limit = max(limit, 128UL);
1097 sysctl_sctp_mem[0] = limit / 4 * 3;
1098 sysctl_sctp_mem[1] = limit;
1099 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1101 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1102 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1103 max_share = min(4UL*1024*1024, limit);
1105 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
1106 sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
1107 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1109 sysctl_sctp_wmem[0] = SK_STREAM_MEM_QUANTUM;
1110 sysctl_sctp_wmem[1] = 16*1024;
1111 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1113 /* Size and allocate the association hash table.
1114 * The methodology is similar to that of the tcp hash tables.
1116 if (num_physpages >= (128 * 1024))
1117 goal = num_physpages >> (22 - PAGE_SHIFT);
1119 goal = num_physpages >> (24 - PAGE_SHIFT);
1121 for (order = 0; (1UL << order) < goal; order++)
1125 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1126 sizeof(struct sctp_hashbucket);
1127 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1129 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1130 __get_free_pages(GFP_ATOMIC, order);
1131 } while (!sctp_assoc_hashtable && --order > 0);
1132 if (!sctp_assoc_hashtable) {
1133 printk(KERN_ERR "SCTP: Failed association hash alloc.\n");
1135 goto err_ahash_alloc;
1137 for (i = 0; i < sctp_assoc_hashsize; i++) {
1138 rwlock_init(&sctp_assoc_hashtable[i].lock);
1139 sctp_assoc_hashtable[i].chain = NULL;
1142 /* Allocate and initialize the endpoint hash table. */
1143 sctp_ep_hashsize = 64;
1144 sctp_ep_hashtable = (struct sctp_hashbucket *)
1145 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1146 if (!sctp_ep_hashtable) {
1147 printk(KERN_ERR "SCTP: Failed endpoint_hash alloc.\n");
1149 goto err_ehash_alloc;
1151 for (i = 0; i < sctp_ep_hashsize; i++) {
1152 rwlock_init(&sctp_ep_hashtable[i].lock);
1153 sctp_ep_hashtable[i].chain = NULL;
1156 /* Allocate and initialize the SCTP port hash table. */
1158 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1159 sizeof(struct sctp_bind_hashbucket);
1160 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1162 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1163 __get_free_pages(GFP_ATOMIC, order);
1164 } while (!sctp_port_hashtable && --order > 0);
1165 if (!sctp_port_hashtable) {
1166 printk(KERN_ERR "SCTP: Failed bind hash alloc.");
1168 goto err_bhash_alloc;
1170 for (i = 0; i < sctp_port_hashsize; i++) {
1171 spin_lock_init(&sctp_port_hashtable[i].lock);
1172 sctp_port_hashtable[i].chain = NULL;
1175 spin_lock_init(&sctp_port_alloc_lock);
1176 sctp_port_rover = sysctl_local_port_range[0] - 1;
1178 printk(KERN_INFO "SCTP: Hash tables configured "
1179 "(established %d bind %d)\n",
1180 sctp_assoc_hashsize, sctp_port_hashsize);
1182 /* Disable ADDIP by default. */
1183 sctp_addip_enable = 0;
1185 /* Enable PR-SCTP by default. */
1186 sctp_prsctp_enable = 1;
1188 sctp_sysctl_register();
1190 INIT_LIST_HEAD(&sctp_address_families);
1191 sctp_register_af(&sctp_ipv4_specific);
1193 status = proto_register(&sctp_prot, 1);
1195 goto err_proto_register;
1197 /* Register SCTP(UDP and TCP style) with socket layer. */
1198 inet_register_protosw(&sctp_seqpacket_protosw);
1199 inet_register_protosw(&sctp_stream_protosw);
1201 status = sctp_v6_init();
1205 /* Initialize the control inode/socket for handling OOTB packets. */
1206 if ((status = sctp_ctl_sock_init())) {
1208 "SCTP: Failed to initialize the SCTP control sock.\n");
1209 goto err_ctl_sock_init;
1212 /* Initialize the local address list. */
1213 INIT_LIST_HEAD(&sctp_local_addr_list);
1214 spin_lock_init(&sctp_local_addr_lock);
1215 sctp_get_local_addr_list();
1217 /* Register notifier for inet address additions/deletions. */
1218 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1220 /* Register SCTP with inet layer. */
1221 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) {
1223 goto err_add_protocol;
1226 /* Register SCTP with inet6 layer. */
1227 status = sctp_v6_add_protocol();
1229 goto err_v6_add_protocol;
1231 __unsafe(THIS_MODULE);
1235 err_v6_add_protocol:
1236 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1237 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1239 sctp_free_local_addr_list();
1240 sock_release(sctp_ctl_socket);
1244 inet_unregister_protosw(&sctp_stream_protosw);
1245 inet_unregister_protosw(&sctp_seqpacket_protosw);
1246 proto_unregister(&sctp_prot);
1248 sctp_sysctl_unregister();
1249 list_del(&sctp_ipv4_specific.list);
1250 free_pages((unsigned long)sctp_port_hashtable,
1251 get_order(sctp_port_hashsize *
1252 sizeof(struct sctp_bind_hashbucket)));
1254 kfree(sctp_ep_hashtable);
1256 free_pages((unsigned long)sctp_assoc_hashtable,
1257 get_order(sctp_assoc_hashsize *
1258 sizeof(struct sctp_hashbucket)));
1260 sctp_dbg_objcnt_exit();
1263 cleanup_sctp_mibs();
1265 kmem_cache_destroy(sctp_chunk_cachep);
1267 kmem_cache_destroy(sctp_bucket_cachep);
1271 /* Exit handler for the SCTP protocol. */
1272 SCTP_STATIC __exit void sctp_exit(void)
1274 /* BUG. This should probably do something useful like clean
1275 * up all the remaining associations and all that memory.
1278 /* Unregister with inet6/inet layers. */
1279 sctp_v6_del_protocol();
1280 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1282 /* Unregister notifier for inet address additions/deletions. */
1283 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1285 /* Free the local address list. */
1286 sctp_free_local_addr_list();
1288 /* Free the control endpoint. */
1289 sock_release(sctp_ctl_socket);
1291 /* Cleanup v6 initializations. */
1294 /* Unregister with socket layer. */
1295 inet_unregister_protosw(&sctp_stream_protosw);
1296 inet_unregister_protosw(&sctp_seqpacket_protosw);
1298 sctp_sysctl_unregister();
1299 list_del(&sctp_ipv4_specific.list);
1301 free_pages((unsigned long)sctp_assoc_hashtable,
1302 get_order(sctp_assoc_hashsize *
1303 sizeof(struct sctp_hashbucket)));
1304 kfree(sctp_ep_hashtable);
1305 free_pages((unsigned long)sctp_port_hashtable,
1306 get_order(sctp_port_hashsize *
1307 sizeof(struct sctp_bind_hashbucket)));
1309 sctp_dbg_objcnt_exit();
1311 cleanup_sctp_mibs();
1313 kmem_cache_destroy(sctp_chunk_cachep);
1314 kmem_cache_destroy(sctp_bucket_cachep);
1316 proto_unregister(&sctp_prot);
1319 module_init(sctp_init);
1320 module_exit(sctp_exit);
1323 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1325 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1326 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1327 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1328 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1329 MODULE_LICENSE("GPL");