1 /* SCTP kernel 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 implementation
11 * Initialization/cleanup for SCTP protocol support.
13 * This SCTP 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 * This SCTP 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 sock *sctp_ctl_sock;
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 int sysctl_sctp_mem[3];
88 int sysctl_sctp_rmem[3];
89 int sysctl_sctp_wmem[3];
91 /* Return the address of the control sock. */
92 struct sock *sctp_get_ctl_sock(void)
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(&init_net, 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,
233 SCTP_ADDR_SRC, GFP_ATOMIC);
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 = htonl(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 htonl(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 && skb->rtable->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(&init_net, addr->v4.sin_addr.s_addr);
378 if (addr->v4.sin_addr.s_addr != htonl(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 (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
414 retval = SCTP_SCOPE_LOOPBACK;
415 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
416 retval = SCTP_SCOPE_LINK;
417 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
418 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
419 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
420 retval = SCTP_SCOPE_PRIVATE;
422 retval = SCTP_SCOPE_GLOBAL;
428 /* Returns a valid dst cache entry for the given source and destination ip
429 * addresses. If an association is passed, trys to get a dst entry with a
430 * source address that matches an address in the bind address list.
432 static struct dst_entry *sctp_v4_get_dst(struct sctp_association *asoc,
433 union sctp_addr *daddr,
434 union sctp_addr *saddr)
438 struct sctp_bind_addr *bp;
439 struct sctp_sockaddr_entry *laddr;
440 struct dst_entry *dst = NULL;
441 union sctp_addr dst_saddr;
443 memset(&fl, 0x0, sizeof(struct flowi));
444 fl.fl4_dst = daddr->v4.sin_addr.s_addr;
445 fl.proto = IPPROTO_SCTP;
447 fl.fl4_tos = RT_CONN_FLAGS(asoc->base.sk);
448 fl.oif = asoc->base.sk->sk_bound_dev_if;
451 fl.fl4_src = saddr->v4.sin_addr.s_addr;
453 SCTP_DEBUG_PRINTK("%s: DST:%u.%u.%u.%u, SRC:%u.%u.%u.%u - ",
454 __func__, NIPQUAD(fl.fl4_dst),
455 NIPQUAD(fl.fl4_src));
457 if (!ip_route_output_key(&init_net, &rt, &fl)) {
461 /* If there is no association or if a source address is passed, no
462 * more validation is required.
467 bp = &asoc->base.bind_addr;
470 /* Walk through the bind address list and look for a bind
471 * address that matches the source address of the returned dst.
474 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
475 if (!laddr->valid || (laddr->state != SCTP_ADDR_SRC))
477 sctp_v4_dst_saddr(&dst_saddr, dst, htons(bp->port));
478 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
483 /* None of the bound addresses match the source address of the
484 * dst. So release it.
490 /* Walk through the bind address list and try to get a dst that
491 * matches a bind address as the source address.
494 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
497 if ((laddr->state == SCTP_ADDR_SRC) &&
498 (AF_INET == laddr->a.sa.sa_family)) {
499 fl.fl4_src = laddr->a.v4.sin_addr.s_addr;
500 if (!ip_route_output_key(&init_net, &rt, &fl)) {
511 SCTP_DEBUG_PRINTK("rt_dst:%u.%u.%u.%u, rt_src:%u.%u.%u.%u\n",
512 NIPQUAD(rt->rt_dst), NIPQUAD(rt->rt_src));
514 SCTP_DEBUG_PRINTK("NO ROUTE\n");
519 /* For v4, the source address is cached in the route entry(dst). So no need
520 * to cache it separately and hence this is an empty routine.
522 static void sctp_v4_get_saddr(struct sctp_sock *sk,
523 struct sctp_association *asoc,
524 struct dst_entry *dst,
525 union sctp_addr *daddr,
526 union sctp_addr *saddr)
528 struct rtable *rt = (struct rtable *)dst;
534 saddr->v4.sin_family = AF_INET;
535 saddr->v4.sin_port = htons(asoc->base.bind_addr.port);
536 saddr->v4.sin_addr.s_addr = rt->rt_src;
540 /* What interface did this skb arrive on? */
541 static int sctp_v4_skb_iif(const struct sk_buff *skb)
543 return skb->rtable->rt_iif;
546 /* Was this packet marked by Explicit Congestion Notification? */
547 static int sctp_v4_is_ce(const struct sk_buff *skb)
549 return INET_ECN_is_ce(ip_hdr(skb)->tos);
552 /* Create and initialize a new sk for the socket returned by accept(). */
553 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
554 struct sctp_association *asoc)
556 struct inet_sock *inet = inet_sk(sk);
557 struct inet_sock *newinet;
558 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
564 sock_init_data(NULL, newsk);
566 newsk->sk_type = SOCK_STREAM;
568 newsk->sk_no_check = sk->sk_no_check;
569 newsk->sk_reuse = sk->sk_reuse;
570 newsk->sk_shutdown = sk->sk_shutdown;
572 newsk->sk_destruct = inet_sock_destruct;
573 newsk->sk_family = PF_INET;
574 newsk->sk_protocol = IPPROTO_SCTP;
575 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
576 sock_reset_flag(newsk, SOCK_ZAPPED);
578 newinet = inet_sk(newsk);
580 /* Initialize sk's sport, dport, rcv_saddr and daddr for
581 * getsockname() and getpeername()
583 newinet->sport = inet->sport;
584 newinet->saddr = inet->saddr;
585 newinet->rcv_saddr = inet->rcv_saddr;
586 newinet->dport = htons(asoc->peer.port);
587 newinet->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
588 newinet->pmtudisc = inet->pmtudisc;
589 newinet->id = asoc->next_tsn ^ jiffies;
591 newinet->uc_ttl = -1;
592 newinet->mc_loop = 1;
594 newinet->mc_index = 0;
595 newinet->mc_list = NULL;
597 sk_refcnt_debug_inc(newsk);
599 if (newsk->sk_prot->init(newsk)) {
600 sk_common_release(newsk);
608 /* Map address, empty for v4 family */
609 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
614 /* Dump the v4 addr to the seq file. */
615 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
617 seq_printf(seq, "%d.%d.%d.%d ", NIPQUAD(addr->v4.sin_addr));
620 /* Event handler for inet address addition/deletion events.
621 * The sctp_local_addr_list needs to be protocted by a spin lock since
622 * multiple notifiers (say IPv4 and IPv6) may be running at the same
623 * time and thus corrupt the list.
624 * The reader side is protected with RCU.
626 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
629 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
630 struct sctp_sockaddr_entry *addr = NULL;
631 struct sctp_sockaddr_entry *temp;
634 if (dev_net(ifa->ifa_dev->dev) != &init_net)
639 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
641 addr->a.v4.sin_family = AF_INET;
642 addr->a.v4.sin_port = 0;
643 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
645 spin_lock_bh(&sctp_local_addr_lock);
646 list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
647 spin_unlock_bh(&sctp_local_addr_lock);
651 spin_lock_bh(&sctp_local_addr_lock);
652 list_for_each_entry_safe(addr, temp,
653 &sctp_local_addr_list, list) {
654 if (addr->a.sa.sa_family == AF_INET &&
655 addr->a.v4.sin_addr.s_addr ==
659 list_del_rcu(&addr->list);
663 spin_unlock_bh(&sctp_local_addr_lock);
665 call_rcu(&addr->rcu, sctp_local_addr_free);
673 * Initialize the control inode/socket with a control endpoint data
674 * structure. This endpoint is reserved exclusively for the OOTB processing.
676 static int sctp_ctl_sock_init(void)
681 if (sctp_get_pf_specific(PF_INET6))
686 err = inet_ctl_sock_create(&sctp_ctl_sock, family,
687 SOCK_SEQPACKET, IPPROTO_SCTP, &init_net);
690 "SCTP: Failed to create the SCTP control socket.\n");
696 /* Register address family specific functions. */
697 int sctp_register_af(struct sctp_af *af)
699 switch (af->sa_family) {
701 if (sctp_af_v4_specific)
703 sctp_af_v4_specific = af;
706 if (sctp_af_v6_specific)
708 sctp_af_v6_specific = af;
714 INIT_LIST_HEAD(&af->list);
715 list_add_tail(&af->list, &sctp_address_families);
719 /* Get the table of functions for manipulating a particular address
722 struct sctp_af *sctp_get_af_specific(sa_family_t family)
726 return sctp_af_v4_specific;
728 return sctp_af_v6_specific;
734 /* Common code to initialize a AF_INET msg_name. */
735 static void sctp_inet_msgname(char *msgname, int *addr_len)
737 struct sockaddr_in *sin;
739 sin = (struct sockaddr_in *)msgname;
740 *addr_len = sizeof(struct sockaddr_in);
741 sin->sin_family = AF_INET;
742 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
745 /* Copy the primary address of the peer primary address as the msg_name. */
746 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
749 struct sockaddr_in *sin, *sinfrom;
752 struct sctp_association *asoc;
755 sctp_inet_msgname(msgname, addr_len);
756 sin = (struct sockaddr_in *)msgname;
757 sinfrom = &asoc->peer.primary_addr.v4;
758 sin->sin_port = htons(asoc->peer.port);
759 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
763 /* Initialize and copy out a msgname from an inbound skb. */
764 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
767 struct sctphdr *sh = sctp_hdr(skb);
768 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
770 sctp_inet_msgname(msgname, len);
771 sin->sin_port = sh->source;
772 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
776 /* Do we support this AF? */
777 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
779 /* PF_INET only supports AF_INET addresses. */
780 return (AF_INET == family);
783 /* Address matching with wildcards allowed. */
784 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
785 const union sctp_addr *addr2,
786 struct sctp_sock *opt)
788 /* PF_INET only supports AF_INET addresses. */
789 if (addr1->sa.sa_family != addr2->sa.sa_family)
791 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
792 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
794 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
800 /* Verify that provided sockaddr looks bindable. Common verification has
801 * already been taken care of.
803 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
805 return sctp_v4_available(addr, opt);
808 /* Verify that sockaddr looks sendable. Common verification has already
809 * been taken care of.
811 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
816 /* Fill in Supported Address Type information for INIT and INIT-ACK
817 * chunks. Returns number of addresses supported.
819 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
822 types[0] = SCTP_PARAM_IPV4_ADDRESS;
826 /* Wrapper routine that calls the ip transmit routine. */
827 static inline int sctp_v4_xmit(struct sk_buff *skb,
828 struct sctp_transport *transport, int ipfragok)
830 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, "
831 "src:%u.%u.%u.%u, dst:%u.%u.%u.%u\n",
832 __func__, skb, skb->len,
833 NIPQUAD(skb->rtable->rt_src),
834 NIPQUAD(skb->rtable->rt_dst));
836 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
837 return ip_queue_xmit(skb, ipfragok);
840 static struct sctp_af sctp_af_inet;
842 static struct sctp_pf sctp_pf_inet = {
843 .event_msgname = sctp_inet_event_msgname,
844 .skb_msgname = sctp_inet_skb_msgname,
845 .af_supported = sctp_inet_af_supported,
846 .cmp_addr = sctp_inet_cmp_addr,
847 .bind_verify = sctp_inet_bind_verify,
848 .send_verify = sctp_inet_send_verify,
849 .supported_addrs = sctp_inet_supported_addrs,
850 .create_accept_sk = sctp_v4_create_accept_sk,
851 .addr_v4map = sctp_v4_addr_v4map,
855 /* Notifier for inetaddr addition/deletion events. */
856 static struct notifier_block sctp_inetaddr_notifier = {
857 .notifier_call = sctp_inetaddr_event,
860 /* Socket operations. */
861 static const struct proto_ops inet_seqpacket_ops = {
863 .owner = THIS_MODULE,
864 .release = inet_release, /* Needs to be wrapped... */
866 .connect = inet_dgram_connect,
867 .socketpair = sock_no_socketpair,
868 .accept = inet_accept,
869 .getname = inet_getname, /* Semantics are different. */
872 .listen = sctp_inet_listen,
873 .shutdown = inet_shutdown, /* Looks harmless. */
874 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
875 .getsockopt = sock_common_getsockopt,
876 .sendmsg = inet_sendmsg,
877 .recvmsg = sock_common_recvmsg,
878 .mmap = sock_no_mmap,
879 .sendpage = sock_no_sendpage,
881 .compat_setsockopt = compat_sock_common_setsockopt,
882 .compat_getsockopt = compat_sock_common_getsockopt,
886 /* Registration with AF_INET family. */
887 static struct inet_protosw sctp_seqpacket_protosw = {
888 .type = SOCK_SEQPACKET,
889 .protocol = IPPROTO_SCTP,
891 .ops = &inet_seqpacket_ops,
894 .flags = SCTP_PROTOSW_FLAG
896 static struct inet_protosw sctp_stream_protosw = {
898 .protocol = IPPROTO_SCTP,
900 .ops = &inet_seqpacket_ops,
903 .flags = SCTP_PROTOSW_FLAG
906 /* Register with IP layer. */
907 static struct net_protocol sctp_protocol = {
909 .err_handler = sctp_v4_err,
913 /* IPv4 address related functions. */
914 static struct sctp_af sctp_af_inet = {
915 .sa_family = AF_INET,
916 .sctp_xmit = sctp_v4_xmit,
917 .setsockopt = ip_setsockopt,
918 .getsockopt = ip_getsockopt,
919 .get_dst = sctp_v4_get_dst,
920 .get_saddr = sctp_v4_get_saddr,
921 .copy_addrlist = sctp_v4_copy_addrlist,
922 .from_skb = sctp_v4_from_skb,
923 .from_sk = sctp_v4_from_sk,
924 .to_sk_saddr = sctp_v4_to_sk_saddr,
925 .to_sk_daddr = sctp_v4_to_sk_daddr,
926 .from_addr_param = sctp_v4_from_addr_param,
927 .to_addr_param = sctp_v4_to_addr_param,
928 .dst_saddr = sctp_v4_dst_saddr,
929 .cmp_addr = sctp_v4_cmp_addr,
930 .addr_valid = sctp_v4_addr_valid,
931 .inaddr_any = sctp_v4_inaddr_any,
932 .is_any = sctp_v4_is_any,
933 .available = sctp_v4_available,
934 .scope = sctp_v4_scope,
935 .skb_iif = sctp_v4_skb_iif,
936 .is_ce = sctp_v4_is_ce,
937 .seq_dump_addr = sctp_v4_seq_dump_addr,
938 .net_header_len = sizeof(struct iphdr),
939 .sockaddr_len = sizeof(struct sockaddr_in),
941 .compat_setsockopt = compat_ip_setsockopt,
942 .compat_getsockopt = compat_ip_getsockopt,
946 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
950 return sctp_pf_inet_specific;
952 return sctp_pf_inet6_specific;
958 /* Register the PF specific function table. */
959 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
963 if (sctp_pf_inet_specific)
965 sctp_pf_inet_specific = pf;
968 if (sctp_pf_inet6_specific)
970 sctp_pf_inet6_specific = pf;
978 static inline int init_sctp_mibs(void)
980 return snmp_mib_init((void**)sctp_statistics, sizeof(struct sctp_mib));
983 static inline void cleanup_sctp_mibs(void)
985 snmp_mib_free((void**)sctp_statistics);
988 static void sctp_v4_pf_init(void)
990 /* Initialize the SCTP specific PF functions. */
991 sctp_register_pf(&sctp_pf_inet, PF_INET);
992 sctp_register_af(&sctp_af_inet);
995 static void sctp_v4_pf_exit(void)
997 list_del(&sctp_af_inet.list);
1000 static int sctp_v4_protosw_init(void)
1004 rc = proto_register(&sctp_prot, 1);
1008 /* Register SCTP(UDP and TCP style) with socket layer. */
1009 inet_register_protosw(&sctp_seqpacket_protosw);
1010 inet_register_protosw(&sctp_stream_protosw);
1015 static void sctp_v4_protosw_exit(void)
1017 inet_unregister_protosw(&sctp_stream_protosw);
1018 inet_unregister_protosw(&sctp_seqpacket_protosw);
1019 proto_unregister(&sctp_prot);
1022 static int sctp_v4_add_protocol(void)
1024 /* Register notifier for inet address additions/deletions. */
1025 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1027 /* Register SCTP with inet layer. */
1028 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1034 static void sctp_v4_del_protocol(void)
1036 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1037 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1040 /* Initialize the universe into something sensible. */
1041 SCTP_STATIC __init int sctp_init(void)
1044 int status = -EINVAL;
1046 unsigned long limit;
1050 /* SCTP_DEBUG sanity check. */
1051 if (!sctp_sanity_check())
1054 /* Allocate bind_bucket and chunk caches. */
1056 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1057 sizeof(struct sctp_bind_bucket),
1058 0, SLAB_HWCACHE_ALIGN,
1060 if (!sctp_bucket_cachep)
1063 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1064 sizeof(struct sctp_chunk),
1065 0, SLAB_HWCACHE_ALIGN,
1067 if (!sctp_chunk_cachep)
1068 goto err_chunk_cachep;
1070 /* Allocate and initialise sctp mibs. */
1071 status = init_sctp_mibs();
1075 /* Initialize proc fs directory. */
1076 status = sctp_proc_init();
1080 /* Initialize object count debugging. */
1081 sctp_dbg_objcnt_init();
1084 * 14. Suggested SCTP Protocol Parameter Values
1086 /* The following protocol parameters are RECOMMENDED: */
1087 /* RTO.Initial - 3 seconds */
1088 sctp_rto_initial = SCTP_RTO_INITIAL;
1089 /* RTO.Min - 1 second */
1090 sctp_rto_min = SCTP_RTO_MIN;
1091 /* RTO.Max - 60 seconds */
1092 sctp_rto_max = SCTP_RTO_MAX;
1093 /* RTO.Alpha - 1/8 */
1094 sctp_rto_alpha = SCTP_RTO_ALPHA;
1095 /* RTO.Beta - 1/4 */
1096 sctp_rto_beta = SCTP_RTO_BETA;
1098 /* Valid.Cookie.Life - 60 seconds */
1099 sctp_valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1101 /* Whether Cookie Preservative is enabled(1) or not(0) */
1102 sctp_cookie_preserve_enable = 1;
1105 sctp_max_burst = SCTP_DEFAULT_MAX_BURST;
1107 /* Association.Max.Retrans - 10 attempts
1108 * Path.Max.Retrans - 5 attempts (per destination address)
1109 * Max.Init.Retransmits - 8 attempts
1111 sctp_max_retrans_association = 10;
1112 sctp_max_retrans_path = 5;
1113 sctp_max_retrans_init = 8;
1115 /* Sendbuffer growth - do per-socket accounting */
1116 sctp_sndbuf_policy = 0;
1118 /* Rcvbuffer growth - do per-socket accounting */
1119 sctp_rcvbuf_policy = 0;
1121 /* HB.interval - 30 seconds */
1122 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1124 /* delayed SACK timeout */
1125 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1127 /* Implementation specific variables. */
1129 /* Initialize default stream count setup information. */
1130 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1131 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1133 /* Initialize handle used for association ids. */
1134 idr_init(&sctp_assocs_id);
1136 /* Set the pressure threshold to be a fraction of global memory that
1137 * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
1138 * memory, with a floor of 128 pages.
1139 * Note this initalizes the data in sctpv6_prot too
1140 * Unabashedly stolen from tcp_init
1142 limit = min(num_physpages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1143 limit = (limit * (num_physpages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
1144 limit = max(limit, 128UL);
1145 sysctl_sctp_mem[0] = limit / 4 * 3;
1146 sysctl_sctp_mem[1] = limit;
1147 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1149 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1150 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1151 max_share = min(4UL*1024*1024, limit);
1153 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
1154 sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
1155 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1157 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1158 sysctl_sctp_wmem[1] = 16*1024;
1159 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1161 /* Size and allocate the association hash table.
1162 * The methodology is similar to that of the tcp hash tables.
1164 if (num_physpages >= (128 * 1024))
1165 goal = num_physpages >> (22 - PAGE_SHIFT);
1167 goal = num_physpages >> (24 - PAGE_SHIFT);
1169 for (order = 0; (1UL << order) < goal; order++)
1173 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1174 sizeof(struct sctp_hashbucket);
1175 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1177 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1178 __get_free_pages(GFP_ATOMIC, order);
1179 } while (!sctp_assoc_hashtable && --order > 0);
1180 if (!sctp_assoc_hashtable) {
1181 printk(KERN_ERR "SCTP: Failed association hash alloc.\n");
1183 goto err_ahash_alloc;
1185 for (i = 0; i < sctp_assoc_hashsize; i++) {
1186 rwlock_init(&sctp_assoc_hashtable[i].lock);
1187 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1190 /* Allocate and initialize the endpoint hash table. */
1191 sctp_ep_hashsize = 64;
1192 sctp_ep_hashtable = (struct sctp_hashbucket *)
1193 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1194 if (!sctp_ep_hashtable) {
1195 printk(KERN_ERR "SCTP: Failed endpoint_hash alloc.\n");
1197 goto err_ehash_alloc;
1199 for (i = 0; i < sctp_ep_hashsize; i++) {
1200 rwlock_init(&sctp_ep_hashtable[i].lock);
1201 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1204 /* Allocate and initialize the SCTP port hash table. */
1206 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1207 sizeof(struct sctp_bind_hashbucket);
1208 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1210 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1211 __get_free_pages(GFP_ATOMIC, order);
1212 } while (!sctp_port_hashtable && --order > 0);
1213 if (!sctp_port_hashtable) {
1214 printk(KERN_ERR "SCTP: Failed bind hash alloc.");
1216 goto err_bhash_alloc;
1218 for (i = 0; i < sctp_port_hashsize; i++) {
1219 spin_lock_init(&sctp_port_hashtable[i].lock);
1220 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1223 printk(KERN_INFO "SCTP: Hash tables configured "
1224 "(established %d bind %d)\n",
1225 sctp_assoc_hashsize, sctp_port_hashsize);
1227 /* Disable ADDIP by default. */
1228 sctp_addip_enable = 0;
1229 sctp_addip_noauth = 0;
1231 /* Enable PR-SCTP by default. */
1232 sctp_prsctp_enable = 1;
1234 /* Disable AUTH by default. */
1235 sctp_auth_enable = 0;
1237 sctp_sysctl_register();
1239 INIT_LIST_HEAD(&sctp_address_families);
1243 /* Initialize the local address list. */
1244 INIT_LIST_HEAD(&sctp_local_addr_list);
1245 spin_lock_init(&sctp_local_addr_lock);
1246 sctp_get_local_addr_list();
1248 status = sctp_v4_protosw_init();
1251 goto err_protosw_init;
1253 status = sctp_v6_protosw_init();
1255 goto err_v6_protosw_init;
1257 /* Initialize the control inode/socket for handling OOTB packets. */
1258 if ((status = sctp_ctl_sock_init())) {
1260 "SCTP: Failed to initialize the SCTP control sock.\n");
1261 goto err_ctl_sock_init;
1264 status = sctp_v4_add_protocol();
1266 goto err_add_protocol;
1268 /* Register SCTP with inet6 layer. */
1269 status = sctp_v6_add_protocol();
1271 goto err_v6_add_protocol;
1276 err_v6_add_protocol:
1277 sctp_v6_del_protocol();
1279 sctp_v4_del_protocol();
1280 inet_ctl_sock_destroy(sctp_ctl_sock);
1282 sctp_v6_protosw_exit();
1283 err_v6_protosw_init:
1284 sctp_v4_protosw_exit();
1286 sctp_free_local_addr_list();
1289 sctp_sysctl_unregister();
1290 list_del(&sctp_af_inet.list);
1291 free_pages((unsigned long)sctp_port_hashtable,
1292 get_order(sctp_port_hashsize *
1293 sizeof(struct sctp_bind_hashbucket)));
1295 kfree(sctp_ep_hashtable);
1297 free_pages((unsigned long)sctp_assoc_hashtable,
1298 get_order(sctp_assoc_hashsize *
1299 sizeof(struct sctp_hashbucket)));
1301 sctp_dbg_objcnt_exit();
1304 cleanup_sctp_mibs();
1306 kmem_cache_destroy(sctp_chunk_cachep);
1308 kmem_cache_destroy(sctp_bucket_cachep);
1312 /* Exit handler for the SCTP protocol. */
1313 SCTP_STATIC __exit void sctp_exit(void)
1315 /* BUG. This should probably do something useful like clean
1316 * up all the remaining associations and all that memory.
1319 /* Unregister with inet6/inet layers. */
1320 sctp_v6_del_protocol();
1321 sctp_v4_del_protocol();
1323 /* Free the control endpoint. */
1324 inet_ctl_sock_destroy(sctp_ctl_sock);
1326 /* Free protosw registrations */
1327 sctp_v6_protosw_exit();
1328 sctp_v4_protosw_exit();
1330 /* Free the local address list. */
1331 sctp_free_local_addr_list();
1333 /* Unregister with socket layer. */
1337 sctp_sysctl_unregister();
1338 list_del(&sctp_af_inet.list);
1340 free_pages((unsigned long)sctp_assoc_hashtable,
1341 get_order(sctp_assoc_hashsize *
1342 sizeof(struct sctp_hashbucket)));
1343 kfree(sctp_ep_hashtable);
1344 free_pages((unsigned long)sctp_port_hashtable,
1345 get_order(sctp_port_hashsize *
1346 sizeof(struct sctp_bind_hashbucket)));
1348 sctp_dbg_objcnt_exit();
1350 cleanup_sctp_mibs();
1352 kmem_cache_destroy(sctp_chunk_cachep);
1353 kmem_cache_destroy(sctp_bucket_cachep);
1356 module_init(sctp_init);
1357 module_exit(sctp_exit);
1360 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1362 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1363 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1364 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1365 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1366 MODULE_LICENSE("GPL");