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())
116 if (sctp_remaddr_proc_init())
125 /* Clean up the proc fs entry for the SCTP protocol.
126 * Note: Do not make this __exit as it is used in the init error
129 static void sctp_proc_exit(void)
131 sctp_snmp_proc_exit();
132 sctp_eps_proc_exit();
133 sctp_assocs_proc_exit();
134 sctp_remaddr_proc_exit();
137 proc_net_sctp = NULL;
138 remove_proc_entry("sctp", init_net.proc_net);
142 /* Private helper to extract ipv4 address and stash them in
143 * the protocol structure.
145 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
146 struct net_device *dev)
148 struct in_device *in_dev;
149 struct in_ifaddr *ifa;
150 struct sctp_sockaddr_entry *addr;
153 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
158 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
159 /* Add the address to the local list. */
160 addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
162 addr->a.v4.sin_family = AF_INET;
163 addr->a.v4.sin_port = 0;
164 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
166 INIT_LIST_HEAD(&addr->list);
167 INIT_RCU_HEAD(&addr->rcu);
168 list_add_tail(&addr->list, addrlist);
175 /* Extract our IP addresses from the system and stash them in the
176 * protocol structure.
178 static void sctp_get_local_addr_list(void)
180 struct net_device *dev;
181 struct list_head *pos;
184 read_lock(&dev_base_lock);
185 for_each_netdev(&init_net, dev) {
186 __list_for_each(pos, &sctp_address_families) {
187 af = list_entry(pos, struct sctp_af, list);
188 af->copy_addrlist(&sctp_local_addr_list, dev);
191 read_unlock(&dev_base_lock);
194 /* Free the existing local addresses. */
195 static void sctp_free_local_addr_list(void)
197 struct sctp_sockaddr_entry *addr;
198 struct list_head *pos, *temp;
200 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
201 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
207 void sctp_local_addr_free(struct rcu_head *head)
209 struct sctp_sockaddr_entry *e = container_of(head,
210 struct sctp_sockaddr_entry, rcu);
214 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
215 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
216 gfp_t gfp, int copy_flags)
218 struct sctp_sockaddr_entry *addr;
222 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
225 if (sctp_in_scope(&addr->a, scope)) {
226 /* Now that the address is in scope, check to see if
227 * the address type is really supported by the local
228 * sock as well as the remote peer.
230 if ((((AF_INET == addr->a.sa.sa_family) &&
231 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
232 (((AF_INET6 == addr->a.sa.sa_family) &&
233 (copy_flags & SCTP_ADDR6_ALLOWED) &&
234 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
235 error = sctp_add_bind_addr(bp, &addr->a,
236 SCTP_ADDR_SRC, GFP_ATOMIC);
248 /* Initialize a sctp_addr from in incoming skb. */
249 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
256 port = &addr->v4.sin_port;
257 addr->v4.sin_family = AF_INET;
262 from = &ip_hdr(skb)->saddr;
265 from = &ip_hdr(skb)->daddr;
267 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
270 /* Initialize an sctp_addr from a socket. */
271 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
273 addr->v4.sin_family = AF_INET;
274 addr->v4.sin_port = 0;
275 addr->v4.sin_addr.s_addr = inet_sk(sk)->rcv_saddr;
278 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
279 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
281 inet_sk(sk)->rcv_saddr = addr->v4.sin_addr.s_addr;
284 /* Initialize sk->sk_daddr from sctp_addr. */
285 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
287 inet_sk(sk)->daddr = addr->v4.sin_addr.s_addr;
290 /* Initialize a sctp_addr from an address parameter. */
291 static void sctp_v4_from_addr_param(union sctp_addr *addr,
292 union sctp_addr_param *param,
293 __be16 port, int iif)
295 addr->v4.sin_family = AF_INET;
296 addr->v4.sin_port = port;
297 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
300 /* Initialize an address parameter from a sctp_addr and return the length
301 * of the address parameter.
303 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
304 union sctp_addr_param *param)
306 int length = sizeof(sctp_ipv4addr_param_t);
308 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
309 param->v4.param_hdr.length = htons(length);
310 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
315 /* Initialize a sctp_addr from a dst_entry. */
316 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct dst_entry *dst,
319 struct rtable *rt = (struct rtable *)dst;
320 saddr->v4.sin_family = AF_INET;
321 saddr->v4.sin_port = port;
322 saddr->v4.sin_addr.s_addr = rt->rt_src;
325 /* Compare two addresses exactly. */
326 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
327 const union sctp_addr *addr2)
329 if (addr1->sa.sa_family != addr2->sa.sa_family)
331 if (addr1->v4.sin_port != addr2->v4.sin_port)
333 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
339 /* Initialize addr struct to INADDR_ANY. */
340 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
342 addr->v4.sin_family = AF_INET;
343 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
344 addr->v4.sin_port = port;
347 /* Is this a wildcard address? */
348 static int sctp_v4_is_any(const union sctp_addr *addr)
350 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
353 /* This function checks if the address is a valid address to be used for
357 * Return 0 - If the address is a non-unicast or an illegal address.
358 * Return 1 - If the address is a unicast.
360 static int sctp_v4_addr_valid(union sctp_addr *addr,
361 struct sctp_sock *sp,
362 const struct sk_buff *skb)
364 /* Is this a non-unicast address or a unusable SCTP address? */
365 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
368 /* Is this a broadcast address? */
369 if (skb && skb->rtable->rt_flags & RTCF_BROADCAST)
375 /* Should this be available for binding? */
376 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
378 int ret = inet_addr_type(&init_net, addr->v4.sin_addr.s_addr);
381 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
383 !sp->inet.freebind &&
384 !sysctl_ip_nonlocal_bind)
390 /* Checking the loopback, private and other address scopes as defined in
391 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
392 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
394 * Level 0 - unusable SCTP addresses
395 * Level 1 - loopback address
396 * Level 2 - link-local addresses
397 * Level 3 - private addresses.
398 * Level 4 - global addresses
399 * For INIT and INIT-ACK address list, let L be the level of
400 * of requested destination address, sender and receiver
401 * SHOULD include all of its addresses with level greater
402 * than or equal to L.
404 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
408 /* Should IPv4 scoping be a sysctl configurable option
409 * so users can turn it off (default on) for certain
410 * unconventional networking environments?
413 /* Check for unusable SCTP addresses. */
414 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
415 retval = SCTP_SCOPE_UNUSABLE;
416 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
417 retval = SCTP_SCOPE_LOOPBACK;
418 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
419 retval = SCTP_SCOPE_LINK;
420 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
421 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
422 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
423 retval = SCTP_SCOPE_PRIVATE;
425 retval = SCTP_SCOPE_GLOBAL;
431 /* Returns a valid dst cache entry for the given source and destination ip
432 * addresses. If an association is passed, trys to get a dst entry with a
433 * source address that matches an address in the bind address list.
435 static struct dst_entry *sctp_v4_get_dst(struct sctp_association *asoc,
436 union sctp_addr *daddr,
437 union sctp_addr *saddr)
441 struct sctp_bind_addr *bp;
442 struct sctp_sockaddr_entry *laddr;
443 struct dst_entry *dst = NULL;
444 union sctp_addr dst_saddr;
446 memset(&fl, 0x0, sizeof(struct flowi));
447 fl.fl4_dst = daddr->v4.sin_addr.s_addr;
448 fl.proto = IPPROTO_SCTP;
450 fl.fl4_tos = RT_CONN_FLAGS(asoc->base.sk);
451 fl.oif = asoc->base.sk->sk_bound_dev_if;
454 fl.fl4_src = saddr->v4.sin_addr.s_addr;
456 SCTP_DEBUG_PRINTK("%s: DST:%u.%u.%u.%u, SRC:%u.%u.%u.%u - ",
457 __func__, NIPQUAD(fl.fl4_dst),
458 NIPQUAD(fl.fl4_src));
460 if (!ip_route_output_key(&init_net, &rt, &fl)) {
464 /* If there is no association or if a source address is passed, no
465 * more validation is required.
470 bp = &asoc->base.bind_addr;
473 /* Walk through the bind address list and look for a bind
474 * address that matches the source address of the returned dst.
476 sctp_v4_dst_saddr(&dst_saddr, dst, htons(bp->port));
478 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
479 if (!laddr->valid || (laddr->state != SCTP_ADDR_SRC))
481 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
486 /* None of the bound addresses match the source address of the
487 * dst. So release it.
493 /* Walk through the bind address list and try to get a dst that
494 * matches a bind address as the source address.
497 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
500 if ((laddr->state == SCTP_ADDR_SRC) &&
501 (AF_INET == laddr->a.sa.sa_family)) {
502 fl.fl4_src = laddr->a.v4.sin_addr.s_addr;
503 if (!ip_route_output_key(&init_net, &rt, &fl)) {
514 SCTP_DEBUG_PRINTK("rt_dst:%u.%u.%u.%u, rt_src:%u.%u.%u.%u\n",
515 NIPQUAD(rt->rt_dst), NIPQUAD(rt->rt_src));
517 SCTP_DEBUG_PRINTK("NO ROUTE\n");
522 /* For v4, the source address is cached in the route entry(dst). So no need
523 * to cache it separately and hence this is an empty routine.
525 static void sctp_v4_get_saddr(struct sctp_sock *sk,
526 struct sctp_association *asoc,
527 struct dst_entry *dst,
528 union sctp_addr *daddr,
529 union sctp_addr *saddr)
531 struct rtable *rt = (struct rtable *)dst;
537 saddr->v4.sin_family = AF_INET;
538 saddr->v4.sin_port = htons(asoc->base.bind_addr.port);
539 saddr->v4.sin_addr.s_addr = rt->rt_src;
543 /* What interface did this skb arrive on? */
544 static int sctp_v4_skb_iif(const struct sk_buff *skb)
546 return skb->rtable->rt_iif;
549 /* Was this packet marked by Explicit Congestion Notification? */
550 static int sctp_v4_is_ce(const struct sk_buff *skb)
552 return INET_ECN_is_ce(ip_hdr(skb)->tos);
555 /* Create and initialize a new sk for the socket returned by accept(). */
556 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
557 struct sctp_association *asoc)
559 struct inet_sock *inet = inet_sk(sk);
560 struct inet_sock *newinet;
561 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
567 sock_init_data(NULL, newsk);
569 newsk->sk_type = SOCK_STREAM;
571 newsk->sk_no_check = sk->sk_no_check;
572 newsk->sk_reuse = sk->sk_reuse;
573 newsk->sk_shutdown = sk->sk_shutdown;
575 newsk->sk_destruct = inet_sock_destruct;
576 newsk->sk_family = PF_INET;
577 newsk->sk_protocol = IPPROTO_SCTP;
578 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
579 sock_reset_flag(newsk, SOCK_ZAPPED);
581 newinet = inet_sk(newsk);
583 /* Initialize sk's sport, dport, rcv_saddr and daddr for
584 * getsockname() and getpeername()
586 newinet->sport = inet->sport;
587 newinet->saddr = inet->saddr;
588 newinet->rcv_saddr = inet->rcv_saddr;
589 newinet->dport = htons(asoc->peer.port);
590 newinet->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
591 newinet->pmtudisc = inet->pmtudisc;
592 newinet->id = asoc->next_tsn ^ jiffies;
594 newinet->uc_ttl = -1;
595 newinet->mc_loop = 1;
597 newinet->mc_index = 0;
598 newinet->mc_list = NULL;
600 sk_refcnt_debug_inc(newsk);
602 if (newsk->sk_prot->init(newsk)) {
603 sk_common_release(newsk);
611 /* Map address, empty for v4 family */
612 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
617 /* Dump the v4 addr to the seq file. */
618 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
620 seq_printf(seq, "%d.%d.%d.%d ", NIPQUAD(addr->v4.sin_addr));
623 static void sctp_v4_ecn_capable(struct sock *sk)
628 /* Event handler for inet address addition/deletion events.
629 * The sctp_local_addr_list needs to be protocted by a spin lock since
630 * multiple notifiers (say IPv4 and IPv6) may be running at the same
631 * time and thus corrupt the list.
632 * The reader side is protected with RCU.
634 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
637 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
638 struct sctp_sockaddr_entry *addr = NULL;
639 struct sctp_sockaddr_entry *temp;
642 if (dev_net(ifa->ifa_dev->dev) != &init_net)
647 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
649 addr->a.v4.sin_family = AF_INET;
650 addr->a.v4.sin_port = 0;
651 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
653 spin_lock_bh(&sctp_local_addr_lock);
654 list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
655 spin_unlock_bh(&sctp_local_addr_lock);
659 spin_lock_bh(&sctp_local_addr_lock);
660 list_for_each_entry_safe(addr, temp,
661 &sctp_local_addr_list, list) {
662 if (addr->a.sa.sa_family == AF_INET &&
663 addr->a.v4.sin_addr.s_addr ==
667 list_del_rcu(&addr->list);
671 spin_unlock_bh(&sctp_local_addr_lock);
673 call_rcu(&addr->rcu, sctp_local_addr_free);
681 * Initialize the control inode/socket with a control endpoint data
682 * structure. This endpoint is reserved exclusively for the OOTB processing.
684 static int sctp_ctl_sock_init(void)
689 if (sctp_get_pf_specific(PF_INET6))
694 err = inet_ctl_sock_create(&sctp_ctl_sock, family,
695 SOCK_SEQPACKET, IPPROTO_SCTP, &init_net);
698 "SCTP: Failed to create the SCTP control socket.\n");
704 /* Register address family specific functions. */
705 int sctp_register_af(struct sctp_af *af)
707 switch (af->sa_family) {
709 if (sctp_af_v4_specific)
711 sctp_af_v4_specific = af;
714 if (sctp_af_v6_specific)
716 sctp_af_v6_specific = af;
722 INIT_LIST_HEAD(&af->list);
723 list_add_tail(&af->list, &sctp_address_families);
727 /* Get the table of functions for manipulating a particular address
730 struct sctp_af *sctp_get_af_specific(sa_family_t family)
734 return sctp_af_v4_specific;
736 return sctp_af_v6_specific;
742 /* Common code to initialize a AF_INET msg_name. */
743 static void sctp_inet_msgname(char *msgname, int *addr_len)
745 struct sockaddr_in *sin;
747 sin = (struct sockaddr_in *)msgname;
748 *addr_len = sizeof(struct sockaddr_in);
749 sin->sin_family = AF_INET;
750 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
753 /* Copy the primary address of the peer primary address as the msg_name. */
754 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
757 struct sockaddr_in *sin, *sinfrom;
760 struct sctp_association *asoc;
763 sctp_inet_msgname(msgname, addr_len);
764 sin = (struct sockaddr_in *)msgname;
765 sinfrom = &asoc->peer.primary_addr.v4;
766 sin->sin_port = htons(asoc->peer.port);
767 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
771 /* Initialize and copy out a msgname from an inbound skb. */
772 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
775 struct sctphdr *sh = sctp_hdr(skb);
776 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
778 sctp_inet_msgname(msgname, len);
779 sin->sin_port = sh->source;
780 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
784 /* Do we support this AF? */
785 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
787 /* PF_INET only supports AF_INET addresses. */
788 return (AF_INET == family);
791 /* Address matching with wildcards allowed. */
792 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
793 const union sctp_addr *addr2,
794 struct sctp_sock *opt)
796 /* PF_INET only supports AF_INET addresses. */
797 if (addr1->sa.sa_family != addr2->sa.sa_family)
799 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
800 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
802 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
808 /* Verify that provided sockaddr looks bindable. Common verification has
809 * already been taken care of.
811 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
813 return sctp_v4_available(addr, opt);
816 /* Verify that sockaddr looks sendable. Common verification has already
817 * been taken care of.
819 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
824 /* Fill in Supported Address Type information for INIT and INIT-ACK
825 * chunks. Returns number of addresses supported.
827 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
830 types[0] = SCTP_PARAM_IPV4_ADDRESS;
834 /* Wrapper routine that calls the ip transmit routine. */
835 static inline int sctp_v4_xmit(struct sk_buff *skb,
836 struct sctp_transport *transport, int ipfragok)
838 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, "
839 "src:%u.%u.%u.%u, dst:%u.%u.%u.%u\n",
840 __func__, skb, skb->len,
841 NIPQUAD(skb->rtable->rt_src),
842 NIPQUAD(skb->rtable->rt_dst));
844 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
845 return ip_queue_xmit(skb, ipfragok);
848 static struct sctp_af sctp_af_inet;
850 static struct sctp_pf sctp_pf_inet = {
851 .event_msgname = sctp_inet_event_msgname,
852 .skb_msgname = sctp_inet_skb_msgname,
853 .af_supported = sctp_inet_af_supported,
854 .cmp_addr = sctp_inet_cmp_addr,
855 .bind_verify = sctp_inet_bind_verify,
856 .send_verify = sctp_inet_send_verify,
857 .supported_addrs = sctp_inet_supported_addrs,
858 .create_accept_sk = sctp_v4_create_accept_sk,
859 .addr_v4map = sctp_v4_addr_v4map,
863 /* Notifier for inetaddr addition/deletion events. */
864 static struct notifier_block sctp_inetaddr_notifier = {
865 .notifier_call = sctp_inetaddr_event,
868 /* Socket operations. */
869 static const struct proto_ops inet_seqpacket_ops = {
871 .owner = THIS_MODULE,
872 .release = inet_release, /* Needs to be wrapped... */
874 .connect = inet_dgram_connect,
875 .socketpair = sock_no_socketpair,
876 .accept = inet_accept,
877 .getname = inet_getname, /* Semantics are different. */
880 .listen = sctp_inet_listen,
881 .shutdown = inet_shutdown, /* Looks harmless. */
882 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
883 .getsockopt = sock_common_getsockopt,
884 .sendmsg = inet_sendmsg,
885 .recvmsg = sock_common_recvmsg,
886 .mmap = sock_no_mmap,
887 .sendpage = sock_no_sendpage,
889 .compat_setsockopt = compat_sock_common_setsockopt,
890 .compat_getsockopt = compat_sock_common_getsockopt,
894 /* Registration with AF_INET family. */
895 static struct inet_protosw sctp_seqpacket_protosw = {
896 .type = SOCK_SEQPACKET,
897 .protocol = IPPROTO_SCTP,
899 .ops = &inet_seqpacket_ops,
902 .flags = SCTP_PROTOSW_FLAG
904 static struct inet_protosw sctp_stream_protosw = {
906 .protocol = IPPROTO_SCTP,
908 .ops = &inet_seqpacket_ops,
911 .flags = SCTP_PROTOSW_FLAG
914 /* Register with IP layer. */
915 static struct net_protocol sctp_protocol = {
917 .err_handler = sctp_v4_err,
921 /* IPv4 address related functions. */
922 static struct sctp_af sctp_af_inet = {
923 .sa_family = AF_INET,
924 .sctp_xmit = sctp_v4_xmit,
925 .setsockopt = ip_setsockopt,
926 .getsockopt = ip_getsockopt,
927 .get_dst = sctp_v4_get_dst,
928 .get_saddr = sctp_v4_get_saddr,
929 .copy_addrlist = sctp_v4_copy_addrlist,
930 .from_skb = sctp_v4_from_skb,
931 .from_sk = sctp_v4_from_sk,
932 .to_sk_saddr = sctp_v4_to_sk_saddr,
933 .to_sk_daddr = sctp_v4_to_sk_daddr,
934 .from_addr_param = sctp_v4_from_addr_param,
935 .to_addr_param = sctp_v4_to_addr_param,
936 .dst_saddr = sctp_v4_dst_saddr,
937 .cmp_addr = sctp_v4_cmp_addr,
938 .addr_valid = sctp_v4_addr_valid,
939 .inaddr_any = sctp_v4_inaddr_any,
940 .is_any = sctp_v4_is_any,
941 .available = sctp_v4_available,
942 .scope = sctp_v4_scope,
943 .skb_iif = sctp_v4_skb_iif,
944 .is_ce = sctp_v4_is_ce,
945 .seq_dump_addr = sctp_v4_seq_dump_addr,
946 .ecn_capable = sctp_v4_ecn_capable,
947 .net_header_len = sizeof(struct iphdr),
948 .sockaddr_len = sizeof(struct sockaddr_in),
950 .compat_setsockopt = compat_ip_setsockopt,
951 .compat_getsockopt = compat_ip_getsockopt,
955 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
959 return sctp_pf_inet_specific;
961 return sctp_pf_inet6_specific;
967 /* Register the PF specific function table. */
968 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
972 if (sctp_pf_inet_specific)
974 sctp_pf_inet_specific = pf;
977 if (sctp_pf_inet6_specific)
979 sctp_pf_inet6_specific = pf;
987 static inline int init_sctp_mibs(void)
989 return snmp_mib_init((void**)sctp_statistics, sizeof(struct sctp_mib));
992 static inline void cleanup_sctp_mibs(void)
994 snmp_mib_free((void**)sctp_statistics);
997 static void sctp_v4_pf_init(void)
999 /* Initialize the SCTP specific PF functions. */
1000 sctp_register_pf(&sctp_pf_inet, PF_INET);
1001 sctp_register_af(&sctp_af_inet);
1004 static void sctp_v4_pf_exit(void)
1006 list_del(&sctp_af_inet.list);
1009 static int sctp_v4_protosw_init(void)
1013 rc = proto_register(&sctp_prot, 1);
1017 /* Register SCTP(UDP and TCP style) with socket layer. */
1018 inet_register_protosw(&sctp_seqpacket_protosw);
1019 inet_register_protosw(&sctp_stream_protosw);
1024 static void sctp_v4_protosw_exit(void)
1026 inet_unregister_protosw(&sctp_stream_protosw);
1027 inet_unregister_protosw(&sctp_seqpacket_protosw);
1028 proto_unregister(&sctp_prot);
1031 static int sctp_v4_add_protocol(void)
1033 /* Register notifier for inet address additions/deletions. */
1034 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1036 /* Register SCTP with inet layer. */
1037 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1043 static void sctp_v4_del_protocol(void)
1045 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1046 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1049 /* Initialize the universe into something sensible. */
1050 SCTP_STATIC __init int sctp_init(void)
1053 int status = -EINVAL;
1055 unsigned long limit;
1059 /* SCTP_DEBUG sanity check. */
1060 if (!sctp_sanity_check())
1063 /* Allocate bind_bucket and chunk caches. */
1065 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1066 sizeof(struct sctp_bind_bucket),
1067 0, SLAB_HWCACHE_ALIGN,
1069 if (!sctp_bucket_cachep)
1072 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1073 sizeof(struct sctp_chunk),
1074 0, SLAB_HWCACHE_ALIGN,
1076 if (!sctp_chunk_cachep)
1077 goto err_chunk_cachep;
1079 /* Allocate and initialise sctp mibs. */
1080 status = init_sctp_mibs();
1084 /* Initialize proc fs directory. */
1085 status = sctp_proc_init();
1089 /* Initialize object count debugging. */
1090 sctp_dbg_objcnt_init();
1093 * 14. Suggested SCTP Protocol Parameter Values
1095 /* The following protocol parameters are RECOMMENDED: */
1096 /* RTO.Initial - 3 seconds */
1097 sctp_rto_initial = SCTP_RTO_INITIAL;
1098 /* RTO.Min - 1 second */
1099 sctp_rto_min = SCTP_RTO_MIN;
1100 /* RTO.Max - 60 seconds */
1101 sctp_rto_max = SCTP_RTO_MAX;
1102 /* RTO.Alpha - 1/8 */
1103 sctp_rto_alpha = SCTP_RTO_ALPHA;
1104 /* RTO.Beta - 1/4 */
1105 sctp_rto_beta = SCTP_RTO_BETA;
1107 /* Valid.Cookie.Life - 60 seconds */
1108 sctp_valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1110 /* Whether Cookie Preservative is enabled(1) or not(0) */
1111 sctp_cookie_preserve_enable = 1;
1114 sctp_max_burst = SCTP_DEFAULT_MAX_BURST;
1116 /* Association.Max.Retrans - 10 attempts
1117 * Path.Max.Retrans - 5 attempts (per destination address)
1118 * Max.Init.Retransmits - 8 attempts
1120 sctp_max_retrans_association = 10;
1121 sctp_max_retrans_path = 5;
1122 sctp_max_retrans_init = 8;
1124 /* Sendbuffer growth - do per-socket accounting */
1125 sctp_sndbuf_policy = 0;
1127 /* Rcvbuffer growth - do per-socket accounting */
1128 sctp_rcvbuf_policy = 0;
1130 /* HB.interval - 30 seconds */
1131 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1133 /* delayed SACK timeout */
1134 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1136 /* Implementation specific variables. */
1138 /* Initialize default stream count setup information. */
1139 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1140 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1142 /* Initialize handle used for association ids. */
1143 idr_init(&sctp_assocs_id);
1145 /* Set the pressure threshold to be a fraction of global memory that
1146 * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
1147 * memory, with a floor of 128 pages.
1148 * Note this initalizes the data in sctpv6_prot too
1149 * Unabashedly stolen from tcp_init
1151 limit = min(num_physpages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1152 limit = (limit * (num_physpages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
1153 limit = max(limit, 128UL);
1154 sysctl_sctp_mem[0] = limit / 4 * 3;
1155 sysctl_sctp_mem[1] = limit;
1156 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1158 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1159 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1160 max_share = min(4UL*1024*1024, limit);
1162 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
1163 sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
1164 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1166 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1167 sysctl_sctp_wmem[1] = 16*1024;
1168 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1170 /* Size and allocate the association hash table.
1171 * The methodology is similar to that of the tcp hash tables.
1173 if (num_physpages >= (128 * 1024))
1174 goal = num_physpages >> (22 - PAGE_SHIFT);
1176 goal = num_physpages >> (24 - PAGE_SHIFT);
1178 for (order = 0; (1UL << order) < goal; order++)
1182 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1183 sizeof(struct sctp_hashbucket);
1184 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1186 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1187 __get_free_pages(GFP_ATOMIC, order);
1188 } while (!sctp_assoc_hashtable && --order > 0);
1189 if (!sctp_assoc_hashtable) {
1190 printk(KERN_ERR "SCTP: Failed association hash alloc.\n");
1192 goto err_ahash_alloc;
1194 for (i = 0; i < sctp_assoc_hashsize; i++) {
1195 rwlock_init(&sctp_assoc_hashtable[i].lock);
1196 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1199 /* Allocate and initialize the endpoint hash table. */
1200 sctp_ep_hashsize = 64;
1201 sctp_ep_hashtable = (struct sctp_hashbucket *)
1202 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1203 if (!sctp_ep_hashtable) {
1204 printk(KERN_ERR "SCTP: Failed endpoint_hash alloc.\n");
1206 goto err_ehash_alloc;
1208 for (i = 0; i < sctp_ep_hashsize; i++) {
1209 rwlock_init(&sctp_ep_hashtable[i].lock);
1210 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1213 /* Allocate and initialize the SCTP port hash table. */
1215 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1216 sizeof(struct sctp_bind_hashbucket);
1217 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1219 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1220 __get_free_pages(GFP_ATOMIC, order);
1221 } while (!sctp_port_hashtable && --order > 0);
1222 if (!sctp_port_hashtable) {
1223 printk(KERN_ERR "SCTP: Failed bind hash alloc.");
1225 goto err_bhash_alloc;
1227 for (i = 0; i < sctp_port_hashsize; i++) {
1228 spin_lock_init(&sctp_port_hashtable[i].lock);
1229 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1232 printk(KERN_INFO "SCTP: Hash tables configured "
1233 "(established %d bind %d)\n",
1234 sctp_assoc_hashsize, sctp_port_hashsize);
1236 /* Disable ADDIP by default. */
1237 sctp_addip_enable = 0;
1238 sctp_addip_noauth = 0;
1240 /* Enable PR-SCTP by default. */
1241 sctp_prsctp_enable = 1;
1243 /* Disable AUTH by default. */
1244 sctp_auth_enable = 0;
1246 sctp_sysctl_register();
1248 INIT_LIST_HEAD(&sctp_address_families);
1252 /* Initialize the local address list. */
1253 INIT_LIST_HEAD(&sctp_local_addr_list);
1254 spin_lock_init(&sctp_local_addr_lock);
1255 sctp_get_local_addr_list();
1257 status = sctp_v4_protosw_init();
1260 goto err_protosw_init;
1262 status = sctp_v6_protosw_init();
1264 goto err_v6_protosw_init;
1266 /* Initialize the control inode/socket for handling OOTB packets. */
1267 if ((status = sctp_ctl_sock_init())) {
1269 "SCTP: Failed to initialize the SCTP control sock.\n");
1270 goto err_ctl_sock_init;
1273 status = sctp_v4_add_protocol();
1275 goto err_add_protocol;
1277 /* Register SCTP with inet6 layer. */
1278 status = sctp_v6_add_protocol();
1280 goto err_v6_add_protocol;
1285 err_v6_add_protocol:
1286 sctp_v6_del_protocol();
1288 sctp_v4_del_protocol();
1289 inet_ctl_sock_destroy(sctp_ctl_sock);
1291 sctp_v6_protosw_exit();
1292 err_v6_protosw_init:
1293 sctp_v4_protosw_exit();
1295 sctp_free_local_addr_list();
1298 sctp_sysctl_unregister();
1299 list_del(&sctp_af_inet.list);
1300 free_pages((unsigned long)sctp_port_hashtable,
1301 get_order(sctp_port_hashsize *
1302 sizeof(struct sctp_bind_hashbucket)));
1304 kfree(sctp_ep_hashtable);
1306 free_pages((unsigned long)sctp_assoc_hashtable,
1307 get_order(sctp_assoc_hashsize *
1308 sizeof(struct sctp_hashbucket)));
1310 sctp_dbg_objcnt_exit();
1313 cleanup_sctp_mibs();
1315 kmem_cache_destroy(sctp_chunk_cachep);
1317 kmem_cache_destroy(sctp_bucket_cachep);
1321 /* Exit handler for the SCTP protocol. */
1322 SCTP_STATIC __exit void sctp_exit(void)
1324 /* BUG. This should probably do something useful like clean
1325 * up all the remaining associations and all that memory.
1328 /* Unregister with inet6/inet layers. */
1329 sctp_v6_del_protocol();
1330 sctp_v4_del_protocol();
1332 /* Free the control endpoint. */
1333 inet_ctl_sock_destroy(sctp_ctl_sock);
1335 /* Free protosw registrations */
1336 sctp_v6_protosw_exit();
1337 sctp_v4_protosw_exit();
1339 /* Free the local address list. */
1340 sctp_free_local_addr_list();
1342 /* Unregister with socket layer. */
1346 sctp_sysctl_unregister();
1347 list_del(&sctp_af_inet.list);
1349 free_pages((unsigned long)sctp_assoc_hashtable,
1350 get_order(sctp_assoc_hashsize *
1351 sizeof(struct sctp_hashbucket)));
1352 kfree(sctp_ep_hashtable);
1353 free_pages((unsigned long)sctp_port_hashtable,
1354 get_order(sctp_port_hashsize *
1355 sizeof(struct sctp_bind_hashbucket)));
1357 sctp_dbg_objcnt_exit();
1359 cleanup_sctp_mibs();
1361 kmem_cache_destroy(sctp_chunk_cachep);
1362 kmem_cache_destroy(sctp_bucket_cachep);
1365 module_init(sctp_init);
1366 module_exit(sctp_exit);
1369 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1371 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1372 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1373 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1374 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1375 MODULE_LICENSE("GPL");