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/protocol.h>
58 #include <net/route.h>
59 #include <net/sctp/sctp.h>
60 #include <net/addrconf.h>
61 #include <net/inet_common.h>
62 #include <net/inet_ecn.h>
64 /* Global data structures. */
65 struct sctp_globals sctp_globals __read_mostly;
66 struct proc_dir_entry *proc_net_sctp;
67 DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
69 struct idr sctp_assocs_id;
70 DEFINE_SPINLOCK(sctp_assocs_id_lock);
72 /* This is the global socket data structure used for responding to
73 * the Out-of-the-blue (OOTB) packets. A control sock will be created
74 * for this socket at the initialization time.
76 static struct socket *sctp_ctl_socket;
78 static struct sctp_pf *sctp_pf_inet6_specific;
79 static struct sctp_pf *sctp_pf_inet_specific;
80 static struct sctp_af *sctp_af_v4_specific;
81 static struct sctp_af *sctp_af_v6_specific;
83 struct kmem_cache *sctp_chunk_cachep __read_mostly;
84 struct kmem_cache *sctp_bucket_cachep __read_mostly;
86 extern int sysctl_sctp_mem[3];
87 extern int sysctl_sctp_rmem[3];
88 extern int sysctl_sctp_wmem[3];
90 /* Return the address of the control sock. */
91 struct sock *sctp_get_ctl_sock(void)
93 return sctp_ctl_socket->sk;
96 /* Set up the proc fs entry for the SCTP protocol. */
97 static __init int sctp_proc_init(void)
100 struct proc_dir_entry *ent;
101 ent = proc_mkdir("net/sctp", NULL);
103 ent->owner = THIS_MODULE;
109 if (sctp_snmp_proc_init())
111 if (sctp_eps_proc_init())
113 if (sctp_assocs_proc_init())
122 /* Clean up the proc fs entry for the SCTP protocol.
123 * Note: Do not make this __exit as it is used in the init error
126 static void sctp_proc_exit(void)
128 sctp_snmp_proc_exit();
129 sctp_eps_proc_exit();
130 sctp_assocs_proc_exit();
133 proc_net_sctp = NULL;
134 remove_proc_entry("net/sctp", NULL);
138 /* Private helper to extract ipv4 address and stash them in
139 * the protocol structure.
141 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
142 struct net_device *dev)
144 struct in_device *in_dev;
145 struct in_ifaddr *ifa;
146 struct sctp_sockaddr_entry *addr;
149 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
154 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
155 /* Add the address to the local list. */
156 addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
158 addr->a.v4.sin_family = AF_INET;
159 addr->a.v4.sin_port = 0;
160 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
162 INIT_LIST_HEAD(&addr->list);
163 INIT_RCU_HEAD(&addr->rcu);
164 list_add_tail(&addr->list, addrlist);
171 /* Extract our IP addresses from the system and stash them in the
172 * protocol structure.
174 static void sctp_get_local_addr_list(void)
176 struct net_device *dev;
177 struct list_head *pos;
180 read_lock(&dev_base_lock);
181 for_each_netdev(dev) {
182 __list_for_each(pos, &sctp_address_families) {
183 af = list_entry(pos, struct sctp_af, list);
184 af->copy_addrlist(&sctp_local_addr_list, dev);
187 read_unlock(&dev_base_lock);
190 /* Free the existing local addresses. */
191 static void sctp_free_local_addr_list(void)
193 struct sctp_sockaddr_entry *addr;
194 struct list_head *pos, *temp;
196 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
197 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
203 void sctp_local_addr_free(struct rcu_head *head)
205 struct sctp_sockaddr_entry *e = container_of(head,
206 struct sctp_sockaddr_entry, rcu);
210 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
211 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
212 gfp_t gfp, int copy_flags)
214 struct sctp_sockaddr_entry *addr;
218 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
221 if (sctp_in_scope(&addr->a, scope)) {
222 /* Now that the address is in scope, check to see if
223 * the address type is really supported by the local
224 * sock as well as the remote peer.
226 if ((((AF_INET == addr->a.sa.sa_family) &&
227 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
228 (((AF_INET6 == addr->a.sa.sa_family) &&
229 (copy_flags & SCTP_ADDR6_ALLOWED) &&
230 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
231 error = sctp_add_bind_addr(bp, &addr->a, 1,
244 /* Initialize a sctp_addr from in incoming skb. */
245 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
252 port = &addr->v4.sin_port;
253 addr->v4.sin_family = AF_INET;
258 from = &ip_hdr(skb)->saddr;
261 from = &ip_hdr(skb)->daddr;
263 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
266 /* Initialize an sctp_addr from a socket. */
267 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
269 addr->v4.sin_family = AF_INET;
270 addr->v4.sin_port = 0;
271 addr->v4.sin_addr.s_addr = inet_sk(sk)->rcv_saddr;
274 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
275 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
277 inet_sk(sk)->rcv_saddr = addr->v4.sin_addr.s_addr;
280 /* Initialize sk->sk_daddr from sctp_addr. */
281 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
283 inet_sk(sk)->daddr = addr->v4.sin_addr.s_addr;
286 /* Initialize a sctp_addr from an address parameter. */
287 static void sctp_v4_from_addr_param(union sctp_addr *addr,
288 union sctp_addr_param *param,
289 __be16 port, int iif)
291 addr->v4.sin_family = AF_INET;
292 addr->v4.sin_port = port;
293 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
296 /* Initialize an address parameter from a sctp_addr and return the length
297 * of the address parameter.
299 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
300 union sctp_addr_param *param)
302 int length = sizeof(sctp_ipv4addr_param_t);
304 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
305 param->v4.param_hdr.length = htons(length);
306 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
311 /* Initialize a sctp_addr from a dst_entry. */
312 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct dst_entry *dst,
315 struct rtable *rt = (struct rtable *)dst;
316 saddr->v4.sin_family = AF_INET;
317 saddr->v4.sin_port = port;
318 saddr->v4.sin_addr.s_addr = rt->rt_src;
321 /* Compare two addresses exactly. */
322 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
323 const union sctp_addr *addr2)
325 if (addr1->sa.sa_family != addr2->sa.sa_family)
327 if (addr1->v4.sin_port != addr2->v4.sin_port)
329 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
335 /* Initialize addr struct to INADDR_ANY. */
336 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
338 addr->v4.sin_family = AF_INET;
339 addr->v4.sin_addr.s_addr = INADDR_ANY;
340 addr->v4.sin_port = port;
343 /* Is this a wildcard address? */
344 static int sctp_v4_is_any(const union sctp_addr *addr)
346 return INADDR_ANY == addr->v4.sin_addr.s_addr;
349 /* This function checks if the address is a valid address to be used for
353 * Return 0 - If the address is a non-unicast or an illegal address.
354 * Return 1 - If the address is a unicast.
356 static int sctp_v4_addr_valid(union sctp_addr *addr,
357 struct sctp_sock *sp,
358 const struct sk_buff *skb)
360 /* Is this a non-unicast address or a unusable SCTP address? */
361 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr))
364 /* Is this a broadcast address? */
365 if (skb && ((struct rtable *)skb->dst)->rt_flags & RTCF_BROADCAST)
371 /* Should this be available for binding? */
372 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
374 int ret = inet_addr_type(addr->v4.sin_addr.s_addr);
377 if (addr->v4.sin_addr.s_addr != INADDR_ANY &&
379 !sp->inet.freebind &&
380 !sysctl_ip_nonlocal_bind)
386 /* Checking the loopback, private and other address scopes as defined in
387 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
388 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
390 * Level 0 - unusable SCTP addresses
391 * Level 1 - loopback address
392 * Level 2 - link-local addresses
393 * Level 3 - private addresses.
394 * Level 4 - global addresses
395 * For INIT and INIT-ACK address list, let L be the level of
396 * of requested destination address, sender and receiver
397 * SHOULD include all of its addresses with level greater
398 * than or equal to L.
400 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
404 /* Should IPv4 scoping be a sysctl configurable option
405 * so users can turn it off (default on) for certain
406 * unconventional networking environments?
409 /* Check for unusable SCTP addresses. */
410 if (IS_IPV4_UNUSABLE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
411 retval = SCTP_SCOPE_UNUSABLE;
412 } else if (LOOPBACK(addr->v4.sin_addr.s_addr)) {
413 retval = SCTP_SCOPE_LOOPBACK;
414 } else if (IS_IPV4_LINK_ADDRESS(&addr->v4.sin_addr.s_addr)) {
415 retval = SCTP_SCOPE_LINK;
416 } else if (IS_IPV4_PRIVATE_ADDRESS(&addr->v4.sin_addr.s_addr)) {
417 retval = SCTP_SCOPE_PRIVATE;
419 retval = SCTP_SCOPE_GLOBAL;
425 /* Returns a valid dst cache entry for the given source and destination ip
426 * addresses. If an association is passed, trys to get a dst entry with a
427 * source address that matches an address in the bind address list.
429 static struct dst_entry *sctp_v4_get_dst(struct sctp_association *asoc,
430 union sctp_addr *daddr,
431 union sctp_addr *saddr)
435 struct sctp_bind_addr *bp;
436 struct sctp_sockaddr_entry *laddr;
437 struct dst_entry *dst = NULL;
438 union sctp_addr dst_saddr;
440 memset(&fl, 0x0, sizeof(struct flowi));
441 fl.fl4_dst = daddr->v4.sin_addr.s_addr;
442 fl.proto = IPPROTO_SCTP;
444 fl.fl4_tos = RT_CONN_FLAGS(asoc->base.sk);
445 fl.oif = asoc->base.sk->sk_bound_dev_if;
448 fl.fl4_src = saddr->v4.sin_addr.s_addr;
450 SCTP_DEBUG_PRINTK("%s: DST:%u.%u.%u.%u, SRC:%u.%u.%u.%u - ",
451 __FUNCTION__, NIPQUAD(fl.fl4_dst),
452 NIPQUAD(fl.fl4_src));
454 if (!ip_route_output_key(&rt, &fl)) {
458 /* If there is no association or if a source address is passed, no
459 * more validation is required.
464 bp = &asoc->base.bind_addr;
467 /* Walk through the bind address list and look for a bind
468 * address that matches the source address of the returned dst.
471 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
472 if (!laddr->valid || !laddr->use_as_src)
474 sctp_v4_dst_saddr(&dst_saddr, dst, htons(bp->port));
475 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
480 /* None of the bound addresses match the source address of the
481 * dst. So release it.
487 /* Walk through the bind address list and try to get a dst that
488 * matches a bind address as the source address.
491 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
494 if ((laddr->use_as_src) &&
495 (AF_INET == laddr->a.sa.sa_family)) {
496 fl.fl4_src = laddr->a.v4.sin_addr.s_addr;
497 if (!ip_route_output_key(&rt, &fl)) {
508 SCTP_DEBUG_PRINTK("rt_dst:%u.%u.%u.%u, rt_src:%u.%u.%u.%u\n",
509 NIPQUAD(rt->rt_dst), NIPQUAD(rt->rt_src));
511 SCTP_DEBUG_PRINTK("NO ROUTE\n");
516 /* For v4, the source address is cached in the route entry(dst). So no need
517 * to cache it separately and hence this is an empty routine.
519 static void sctp_v4_get_saddr(struct sctp_association *asoc,
520 struct dst_entry *dst,
521 union sctp_addr *daddr,
522 union sctp_addr *saddr)
524 struct rtable *rt = (struct rtable *)dst;
530 saddr->v4.sin_family = AF_INET;
531 saddr->v4.sin_port = htons(asoc->base.bind_addr.port);
532 saddr->v4.sin_addr.s_addr = rt->rt_src;
536 /* What interface did this skb arrive on? */
537 static int sctp_v4_skb_iif(const struct sk_buff *skb)
539 return ((struct rtable *)skb->dst)->rt_iif;
542 /* Was this packet marked by Explicit Congestion Notification? */
543 static int sctp_v4_is_ce(const struct sk_buff *skb)
545 return INET_ECN_is_ce(ip_hdr(skb)->tos);
548 /* Create and initialize a new sk for the socket returned by accept(). */
549 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
550 struct sctp_association *asoc)
552 struct inet_sock *inet = inet_sk(sk);
553 struct inet_sock *newinet;
554 struct sock *newsk = sk_alloc(PF_INET, GFP_KERNEL, sk->sk_prot, 1);
559 sock_init_data(NULL, newsk);
561 newsk->sk_type = SOCK_STREAM;
563 newsk->sk_no_check = sk->sk_no_check;
564 newsk->sk_reuse = sk->sk_reuse;
565 newsk->sk_shutdown = sk->sk_shutdown;
567 newsk->sk_destruct = inet_sock_destruct;
568 newsk->sk_family = PF_INET;
569 newsk->sk_protocol = IPPROTO_SCTP;
570 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
571 sock_reset_flag(newsk, SOCK_ZAPPED);
573 newinet = inet_sk(newsk);
575 /* Initialize sk's sport, dport, rcv_saddr and daddr for
576 * getsockname() and getpeername()
578 newinet->sport = inet->sport;
579 newinet->saddr = inet->saddr;
580 newinet->rcv_saddr = inet->rcv_saddr;
581 newinet->dport = htons(asoc->peer.port);
582 newinet->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
583 newinet->pmtudisc = inet->pmtudisc;
584 newinet->id = asoc->next_tsn ^ jiffies;
586 newinet->uc_ttl = -1;
587 newinet->mc_loop = 1;
589 newinet->mc_index = 0;
590 newinet->mc_list = NULL;
592 sk_refcnt_debug_inc(newsk);
594 if (newsk->sk_prot->init(newsk)) {
595 sk_common_release(newsk);
603 /* Map address, empty for v4 family */
604 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
609 /* Dump the v4 addr to the seq file. */
610 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
612 seq_printf(seq, "%d.%d.%d.%d ", NIPQUAD(addr->v4.sin_addr));
615 /* Event handler for inet address addition/deletion events.
616 * The sctp_local_addr_list needs to be protocted by a spin lock since
617 * multiple notifiers (say IPv4 and IPv6) may be running at the same
618 * time and thus corrupt the list.
619 * The reader side is protected with RCU.
621 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
624 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
625 struct sctp_sockaddr_entry *addr = NULL;
626 struct sctp_sockaddr_entry *temp;
630 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
632 addr->a.v4.sin_family = AF_INET;
633 addr->a.v4.sin_port = 0;
634 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
636 spin_lock_bh(&sctp_local_addr_lock);
637 list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
638 spin_unlock_bh(&sctp_local_addr_lock);
642 spin_lock_bh(&sctp_local_addr_lock);
643 list_for_each_entry_safe(addr, temp,
644 &sctp_local_addr_list, list) {
645 if (addr->a.v4.sin_addr.s_addr == ifa->ifa_local) {
647 list_del_rcu(&addr->list);
651 spin_unlock_bh(&sctp_local_addr_lock);
652 if (addr && !addr->valid)
653 call_rcu(&addr->rcu, sctp_local_addr_free);
661 * Initialize the control inode/socket with a control endpoint data
662 * structure. This endpoint is reserved exclusively for the OOTB processing.
664 static int sctp_ctl_sock_init(void)
669 if (sctp_get_pf_specific(PF_INET6))
674 err = sock_create_kern(family, SOCK_SEQPACKET, IPPROTO_SCTP,
678 "SCTP: Failed to create the SCTP control socket.\n");
681 sctp_ctl_socket->sk->sk_allocation = GFP_ATOMIC;
682 inet_sk(sctp_ctl_socket->sk)->uc_ttl = -1;
687 /* Register address family specific functions. */
688 int sctp_register_af(struct sctp_af *af)
690 switch (af->sa_family) {
692 if (sctp_af_v4_specific)
694 sctp_af_v4_specific = af;
697 if (sctp_af_v6_specific)
699 sctp_af_v6_specific = af;
705 INIT_LIST_HEAD(&af->list);
706 list_add_tail(&af->list, &sctp_address_families);
710 /* Get the table of functions for manipulating a particular address
713 struct sctp_af *sctp_get_af_specific(sa_family_t family)
717 return sctp_af_v4_specific;
719 return sctp_af_v6_specific;
725 /* Common code to initialize a AF_INET msg_name. */
726 static void sctp_inet_msgname(char *msgname, int *addr_len)
728 struct sockaddr_in *sin;
730 sin = (struct sockaddr_in *)msgname;
731 *addr_len = sizeof(struct sockaddr_in);
732 sin->sin_family = AF_INET;
733 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
736 /* Copy the primary address of the peer primary address as the msg_name. */
737 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
740 struct sockaddr_in *sin, *sinfrom;
743 struct sctp_association *asoc;
746 sctp_inet_msgname(msgname, addr_len);
747 sin = (struct sockaddr_in *)msgname;
748 sinfrom = &asoc->peer.primary_addr.v4;
749 sin->sin_port = htons(asoc->peer.port);
750 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
754 /* Initialize and copy out a msgname from an inbound skb. */
755 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
758 struct sctphdr *sh = sctp_hdr(skb);
759 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
761 sctp_inet_msgname(msgname, len);
762 sin->sin_port = sh->source;
763 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
767 /* Do we support this AF? */
768 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
770 /* PF_INET only supports AF_INET addresses. */
771 return (AF_INET == family);
774 /* Address matching with wildcards allowed. */
775 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
776 const union sctp_addr *addr2,
777 struct sctp_sock *opt)
779 /* PF_INET only supports AF_INET addresses. */
780 if (addr1->sa.sa_family != addr2->sa.sa_family)
782 if (INADDR_ANY == addr1->v4.sin_addr.s_addr ||
783 INADDR_ANY == addr2->v4.sin_addr.s_addr)
785 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
791 /* Verify that provided sockaddr looks bindable. Common verification has
792 * already been taken care of.
794 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
796 return sctp_v4_available(addr, opt);
799 /* Verify that sockaddr looks sendable. Common verification has already
800 * been taken care of.
802 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
807 /* Fill in Supported Address Type information for INIT and INIT-ACK
808 * chunks. Returns number of addresses supported.
810 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
813 types[0] = SCTP_PARAM_IPV4_ADDRESS;
817 /* Wrapper routine that calls the ip transmit routine. */
818 static inline int sctp_v4_xmit(struct sk_buff *skb,
819 struct sctp_transport *transport, int ipfragok)
821 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, "
822 "src:%u.%u.%u.%u, dst:%u.%u.%u.%u\n",
823 __FUNCTION__, skb, skb->len,
824 NIPQUAD(((struct rtable *)skb->dst)->rt_src),
825 NIPQUAD(((struct rtable *)skb->dst)->rt_dst));
827 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
828 return ip_queue_xmit(skb, ipfragok);
831 static struct sctp_af sctp_ipv4_specific;
833 static struct sctp_pf sctp_pf_inet = {
834 .event_msgname = sctp_inet_event_msgname,
835 .skb_msgname = sctp_inet_skb_msgname,
836 .af_supported = sctp_inet_af_supported,
837 .cmp_addr = sctp_inet_cmp_addr,
838 .bind_verify = sctp_inet_bind_verify,
839 .send_verify = sctp_inet_send_verify,
840 .supported_addrs = sctp_inet_supported_addrs,
841 .create_accept_sk = sctp_v4_create_accept_sk,
842 .addr_v4map = sctp_v4_addr_v4map,
843 .af = &sctp_ipv4_specific,
846 /* Notifier for inetaddr addition/deletion events. */
847 static struct notifier_block sctp_inetaddr_notifier = {
848 .notifier_call = sctp_inetaddr_event,
851 /* Socket operations. */
852 static const struct proto_ops inet_seqpacket_ops = {
854 .owner = THIS_MODULE,
855 .release = inet_release, /* Needs to be wrapped... */
857 .connect = inet_dgram_connect,
858 .socketpair = sock_no_socketpair,
859 .accept = inet_accept,
860 .getname = inet_getname, /* Semantics are different. */
863 .listen = sctp_inet_listen,
864 .shutdown = inet_shutdown, /* Looks harmless. */
865 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
866 .getsockopt = sock_common_getsockopt,
867 .sendmsg = inet_sendmsg,
868 .recvmsg = sock_common_recvmsg,
869 .mmap = sock_no_mmap,
870 .sendpage = sock_no_sendpage,
872 .compat_setsockopt = compat_sock_common_setsockopt,
873 .compat_getsockopt = compat_sock_common_getsockopt,
877 /* Registration with AF_INET family. */
878 static struct inet_protosw sctp_seqpacket_protosw = {
879 .type = SOCK_SEQPACKET,
880 .protocol = IPPROTO_SCTP,
882 .ops = &inet_seqpacket_ops,
885 .flags = SCTP_PROTOSW_FLAG
887 static struct inet_protosw sctp_stream_protosw = {
889 .protocol = IPPROTO_SCTP,
891 .ops = &inet_seqpacket_ops,
894 .flags = SCTP_PROTOSW_FLAG
897 /* Register with IP layer. */
898 static struct net_protocol sctp_protocol = {
900 .err_handler = sctp_v4_err,
904 /* IPv4 address related functions. */
905 static struct sctp_af sctp_ipv4_specific = {
906 .sa_family = AF_INET,
907 .sctp_xmit = sctp_v4_xmit,
908 .setsockopt = ip_setsockopt,
909 .getsockopt = ip_getsockopt,
910 .get_dst = sctp_v4_get_dst,
911 .get_saddr = sctp_v4_get_saddr,
912 .copy_addrlist = sctp_v4_copy_addrlist,
913 .from_skb = sctp_v4_from_skb,
914 .from_sk = sctp_v4_from_sk,
915 .to_sk_saddr = sctp_v4_to_sk_saddr,
916 .to_sk_daddr = sctp_v4_to_sk_daddr,
917 .from_addr_param = sctp_v4_from_addr_param,
918 .to_addr_param = sctp_v4_to_addr_param,
919 .dst_saddr = sctp_v4_dst_saddr,
920 .cmp_addr = sctp_v4_cmp_addr,
921 .addr_valid = sctp_v4_addr_valid,
922 .inaddr_any = sctp_v4_inaddr_any,
923 .is_any = sctp_v4_is_any,
924 .available = sctp_v4_available,
925 .scope = sctp_v4_scope,
926 .skb_iif = sctp_v4_skb_iif,
927 .is_ce = sctp_v4_is_ce,
928 .seq_dump_addr = sctp_v4_seq_dump_addr,
929 .net_header_len = sizeof(struct iphdr),
930 .sockaddr_len = sizeof(struct sockaddr_in),
932 .compat_setsockopt = compat_ip_setsockopt,
933 .compat_getsockopt = compat_ip_getsockopt,
937 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
941 return sctp_pf_inet_specific;
943 return sctp_pf_inet6_specific;
949 /* Register the PF specific function table. */
950 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
954 if (sctp_pf_inet_specific)
956 sctp_pf_inet_specific = pf;
959 if (sctp_pf_inet6_specific)
961 sctp_pf_inet6_specific = pf;
969 static int __init init_sctp_mibs(void)
971 sctp_statistics[0] = alloc_percpu(struct sctp_mib);
972 if (!sctp_statistics[0])
974 sctp_statistics[1] = alloc_percpu(struct sctp_mib);
975 if (!sctp_statistics[1]) {
976 free_percpu(sctp_statistics[0]);
983 static void cleanup_sctp_mibs(void)
985 free_percpu(sctp_statistics[0]);
986 free_percpu(sctp_statistics[1]);
989 /* Initialize the universe into something sensible. */
990 SCTP_STATIC __init int sctp_init(void)
993 int status = -EINVAL;
999 /* SCTP_DEBUG sanity check. */
1000 if (!sctp_sanity_check())
1003 /* Allocate bind_bucket and chunk caches. */
1005 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1006 sizeof(struct sctp_bind_bucket),
1007 0, SLAB_HWCACHE_ALIGN,
1009 if (!sctp_bucket_cachep)
1012 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1013 sizeof(struct sctp_chunk),
1014 0, SLAB_HWCACHE_ALIGN,
1016 if (!sctp_chunk_cachep)
1017 goto err_chunk_cachep;
1019 /* Allocate and initialise sctp mibs. */
1020 status = init_sctp_mibs();
1024 /* Initialize proc fs directory. */
1025 status = sctp_proc_init();
1029 /* Initialize object count debugging. */
1030 sctp_dbg_objcnt_init();
1032 /* Initialize the SCTP specific PF functions. */
1033 sctp_register_pf(&sctp_pf_inet, PF_INET);
1035 * 14. Suggested SCTP Protocol Parameter Values
1037 /* The following protocol parameters are RECOMMENDED: */
1038 /* RTO.Initial - 3 seconds */
1039 sctp_rto_initial = SCTP_RTO_INITIAL;
1040 /* RTO.Min - 1 second */
1041 sctp_rto_min = SCTP_RTO_MIN;
1042 /* RTO.Max - 60 seconds */
1043 sctp_rto_max = SCTP_RTO_MAX;
1044 /* RTO.Alpha - 1/8 */
1045 sctp_rto_alpha = SCTP_RTO_ALPHA;
1046 /* RTO.Beta - 1/4 */
1047 sctp_rto_beta = SCTP_RTO_BETA;
1049 /* Valid.Cookie.Life - 60 seconds */
1050 sctp_valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1052 /* Whether Cookie Preservative is enabled(1) or not(0) */
1053 sctp_cookie_preserve_enable = 1;
1056 sctp_max_burst = SCTP_DEFAULT_MAX_BURST;
1058 /* Association.Max.Retrans - 10 attempts
1059 * Path.Max.Retrans - 5 attempts (per destination address)
1060 * Max.Init.Retransmits - 8 attempts
1062 sctp_max_retrans_association = 10;
1063 sctp_max_retrans_path = 5;
1064 sctp_max_retrans_init = 8;
1066 /* Sendbuffer growth - do per-socket accounting */
1067 sctp_sndbuf_policy = 0;
1069 /* Rcvbuffer growth - do per-socket accounting */
1070 sctp_rcvbuf_policy = 0;
1072 /* HB.interval - 30 seconds */
1073 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1075 /* delayed SACK timeout */
1076 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1078 /* Implementation specific variables. */
1080 /* Initialize default stream count setup information. */
1081 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1082 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1084 /* Initialize handle used for association ids. */
1085 idr_init(&sctp_assocs_id);
1087 /* Set the pressure threshold to be a fraction of global memory that
1088 * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
1089 * memory, with a floor of 128 pages.
1090 * Note this initalizes the data in sctpv6_prot too
1091 * Unabashedly stolen from tcp_init
1093 limit = min(num_physpages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1094 limit = (limit * (num_physpages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
1095 limit = max(limit, 128UL);
1096 sysctl_sctp_mem[0] = limit / 4 * 3;
1097 sysctl_sctp_mem[1] = limit;
1098 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1100 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1101 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1102 max_share = min(4UL*1024*1024, limit);
1104 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
1105 sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
1106 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1108 sysctl_sctp_wmem[0] = SK_STREAM_MEM_QUANTUM;
1109 sysctl_sctp_wmem[1] = 16*1024;
1110 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1112 /* Size and allocate the association hash table.
1113 * The methodology is similar to that of the tcp hash tables.
1115 if (num_physpages >= (128 * 1024))
1116 goal = num_physpages >> (22 - PAGE_SHIFT);
1118 goal = num_physpages >> (24 - PAGE_SHIFT);
1120 for (order = 0; (1UL << order) < goal; order++)
1124 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1125 sizeof(struct sctp_hashbucket);
1126 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1128 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1129 __get_free_pages(GFP_ATOMIC, order);
1130 } while (!sctp_assoc_hashtable && --order > 0);
1131 if (!sctp_assoc_hashtable) {
1132 printk(KERN_ERR "SCTP: Failed association hash alloc.\n");
1134 goto err_ahash_alloc;
1136 for (i = 0; i < sctp_assoc_hashsize; i++) {
1137 rwlock_init(&sctp_assoc_hashtable[i].lock);
1138 sctp_assoc_hashtable[i].chain = NULL;
1141 /* Allocate and initialize the endpoint hash table. */
1142 sctp_ep_hashsize = 64;
1143 sctp_ep_hashtable = (struct sctp_hashbucket *)
1144 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1145 if (!sctp_ep_hashtable) {
1146 printk(KERN_ERR "SCTP: Failed endpoint_hash alloc.\n");
1148 goto err_ehash_alloc;
1150 for (i = 0; i < sctp_ep_hashsize; i++) {
1151 rwlock_init(&sctp_ep_hashtable[i].lock);
1152 sctp_ep_hashtable[i].chain = NULL;
1155 /* Allocate and initialize the SCTP port hash table. */
1157 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1158 sizeof(struct sctp_bind_hashbucket);
1159 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1161 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1162 __get_free_pages(GFP_ATOMIC, order);
1163 } while (!sctp_port_hashtable && --order > 0);
1164 if (!sctp_port_hashtable) {
1165 printk(KERN_ERR "SCTP: Failed bind hash alloc.");
1167 goto err_bhash_alloc;
1169 for (i = 0; i < sctp_port_hashsize; i++) {
1170 spin_lock_init(&sctp_port_hashtable[i].lock);
1171 sctp_port_hashtable[i].chain = NULL;
1174 spin_lock_init(&sctp_port_alloc_lock);
1175 sctp_port_rover = sysctl_local_port_range[0] - 1;
1177 printk(KERN_INFO "SCTP: Hash tables configured "
1178 "(established %d bind %d)\n",
1179 sctp_assoc_hashsize, sctp_port_hashsize);
1181 /* Disable ADDIP by default. */
1182 sctp_addip_enable = 0;
1184 /* Enable PR-SCTP by default. */
1185 sctp_prsctp_enable = 1;
1187 sctp_sysctl_register();
1189 INIT_LIST_HEAD(&sctp_address_families);
1190 sctp_register_af(&sctp_ipv4_specific);
1192 status = proto_register(&sctp_prot, 1);
1194 goto err_proto_register;
1196 /* Register SCTP(UDP and TCP style) with socket layer. */
1197 inet_register_protosw(&sctp_seqpacket_protosw);
1198 inet_register_protosw(&sctp_stream_protosw);
1200 status = sctp_v6_init();
1204 /* Initialize the control inode/socket for handling OOTB packets. */
1205 if ((status = sctp_ctl_sock_init())) {
1207 "SCTP: Failed to initialize the SCTP control sock.\n");
1208 goto err_ctl_sock_init;
1211 /* Initialize the local address list. */
1212 INIT_LIST_HEAD(&sctp_local_addr_list);
1213 spin_lock_init(&sctp_local_addr_lock);
1214 sctp_get_local_addr_list();
1216 /* Register notifier for inet address additions/deletions. */
1217 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1219 /* Register SCTP with inet layer. */
1220 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) {
1222 goto err_add_protocol;
1225 /* Register SCTP with inet6 layer. */
1226 status = sctp_v6_add_protocol();
1228 goto err_v6_add_protocol;
1230 __unsafe(THIS_MODULE);
1234 err_v6_add_protocol:
1235 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1236 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1238 sctp_free_local_addr_list();
1239 sock_release(sctp_ctl_socket);
1243 inet_unregister_protosw(&sctp_stream_protosw);
1244 inet_unregister_protosw(&sctp_seqpacket_protosw);
1245 proto_unregister(&sctp_prot);
1247 sctp_sysctl_unregister();
1248 list_del(&sctp_ipv4_specific.list);
1249 free_pages((unsigned long)sctp_port_hashtable,
1250 get_order(sctp_port_hashsize *
1251 sizeof(struct sctp_bind_hashbucket)));
1253 kfree(sctp_ep_hashtable);
1255 free_pages((unsigned long)sctp_assoc_hashtable,
1256 get_order(sctp_assoc_hashsize *
1257 sizeof(struct sctp_hashbucket)));
1259 sctp_dbg_objcnt_exit();
1262 cleanup_sctp_mibs();
1264 kmem_cache_destroy(sctp_chunk_cachep);
1266 kmem_cache_destroy(sctp_bucket_cachep);
1270 /* Exit handler for the SCTP protocol. */
1271 SCTP_STATIC __exit void sctp_exit(void)
1273 /* BUG. This should probably do something useful like clean
1274 * up all the remaining associations and all that memory.
1277 /* Unregister with inet6/inet layers. */
1278 sctp_v6_del_protocol();
1279 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1281 /* Unregister notifier for inet address additions/deletions. */
1282 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1284 /* Free the local address list. */
1285 sctp_free_local_addr_list();
1287 /* Free the control endpoint. */
1288 sock_release(sctp_ctl_socket);
1290 /* Cleanup v6 initializations. */
1293 /* Unregister with socket layer. */
1294 inet_unregister_protosw(&sctp_stream_protosw);
1295 inet_unregister_protosw(&sctp_seqpacket_protosw);
1297 sctp_sysctl_unregister();
1298 list_del(&sctp_ipv4_specific.list);
1300 free_pages((unsigned long)sctp_assoc_hashtable,
1301 get_order(sctp_assoc_hashsize *
1302 sizeof(struct sctp_hashbucket)));
1303 kfree(sctp_ep_hashtable);
1304 free_pages((unsigned long)sctp_port_hashtable,
1305 get_order(sctp_port_hashsize *
1306 sizeof(struct sctp_bind_hashbucket)));
1308 sctp_dbg_objcnt_exit();
1310 cleanup_sctp_mibs();
1312 kmem_cache_destroy(sctp_chunk_cachep);
1313 kmem_cache_destroy(sctp_bucket_cachep);
1315 proto_unregister(&sctp_prot);
1318 module_init(sctp_init);
1319 module_exit(sctp_exit);
1322 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1324 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1325 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1326 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1327 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1328 MODULE_LICENSE("GPL");