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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
51 #include <linux/module.h>
52 #include <linux/init.h>
53 #include <linux/netdevice.h>
54 #include <linux/inetdevice.h>
55 #include <linux/seq_file.h>
56 #include <linux/bootmem.h>
57 #include <linux/highmem.h>
58 #include <linux/swap.h>
59 #include <linux/slab.h>
60 #include <net/net_namespace.h>
61 #include <net/protocol.h>
64 #include <net/route.h>
65 #include <net/sctp/sctp.h>
66 #include <net/addrconf.h>
67 #include <net/inet_common.h>
68 #include <net/inet_ecn.h>
70 /* Global data structures. */
71 struct sctp_globals sctp_globals __read_mostly;
72 DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
75 struct proc_dir_entry *proc_net_sctp;
78 struct idr sctp_assocs_id;
79 DEFINE_SPINLOCK(sctp_assocs_id_lock);
81 /* This is the global socket data structure used for responding to
82 * the Out-of-the-blue (OOTB) packets. A control sock will be created
83 * for this socket at the initialization time.
85 static struct sock *sctp_ctl_sock;
87 static struct sctp_pf *sctp_pf_inet6_specific;
88 static struct sctp_pf *sctp_pf_inet_specific;
89 static struct sctp_af *sctp_af_v4_specific;
90 static struct sctp_af *sctp_af_v6_specific;
92 struct kmem_cache *sctp_chunk_cachep __read_mostly;
93 struct kmem_cache *sctp_bucket_cachep __read_mostly;
95 long sysctl_sctp_mem[3];
96 int sysctl_sctp_rmem[3];
97 int sysctl_sctp_wmem[3];
99 /* Return the address of the control sock. */
100 struct sock *sctp_get_ctl_sock(void)
102 return sctp_ctl_sock;
105 /* Set up the proc fs entry for the SCTP protocol. */
106 static __init int sctp_proc_init(void)
108 if (percpu_counter_init(&sctp_sockets_allocated, 0))
110 #ifdef CONFIG_PROC_FS
111 if (!proc_net_sctp) {
112 proc_net_sctp = proc_mkdir("sctp", init_net.proc_net);
114 goto out_free_percpu;
117 if (sctp_snmp_proc_init())
118 goto out_snmp_proc_init;
119 if (sctp_eps_proc_init())
120 goto out_eps_proc_init;
121 if (sctp_assocs_proc_init())
122 goto out_assocs_proc_init;
123 if (sctp_remaddr_proc_init())
124 goto out_remaddr_proc_init;
128 out_remaddr_proc_init:
129 sctp_assocs_proc_exit();
130 out_assocs_proc_init:
131 sctp_eps_proc_exit();
133 sctp_snmp_proc_exit();
136 proc_net_sctp = NULL;
137 remove_proc_entry("sctp", init_net.proc_net);
140 percpu_counter_destroy(&sctp_sockets_allocated);
143 #endif /* CONFIG_PROC_FS */
149 /* Clean up the proc fs entry for the SCTP protocol.
150 * Note: Do not make this __exit as it is used in the init error
153 static void sctp_proc_exit(void)
155 #ifdef CONFIG_PROC_FS
156 sctp_snmp_proc_exit();
157 sctp_eps_proc_exit();
158 sctp_assocs_proc_exit();
159 sctp_remaddr_proc_exit();
162 proc_net_sctp = NULL;
163 remove_proc_entry("sctp", init_net.proc_net);
166 percpu_counter_destroy(&sctp_sockets_allocated);
169 /* Private helper to extract ipv4 address and stash them in
170 * the protocol structure.
172 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
173 struct net_device *dev)
175 struct in_device *in_dev;
176 struct in_ifaddr *ifa;
177 struct sctp_sockaddr_entry *addr;
180 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
185 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
186 /* Add the address to the local list. */
187 addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
189 addr->a.v4.sin_family = AF_INET;
190 addr->a.v4.sin_port = 0;
191 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
193 INIT_LIST_HEAD(&addr->list);
194 list_add_tail(&addr->list, addrlist);
201 /* Extract our IP addresses from the system and stash them in the
202 * protocol structure.
204 static void sctp_get_local_addr_list(void)
206 struct net_device *dev;
207 struct list_head *pos;
211 for_each_netdev_rcu(&init_net, dev) {
212 __list_for_each(pos, &sctp_address_families) {
213 af = list_entry(pos, struct sctp_af, list);
214 af->copy_addrlist(&sctp_local_addr_list, dev);
220 /* Free the existing local addresses. */
221 static void sctp_free_local_addr_list(void)
223 struct sctp_sockaddr_entry *addr;
224 struct list_head *pos, *temp;
226 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
227 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
233 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
234 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
235 gfp_t gfp, int copy_flags)
237 struct sctp_sockaddr_entry *addr;
241 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
244 if (sctp_in_scope(&addr->a, scope)) {
245 /* Now that the address is in scope, check to see if
246 * the address type is really supported by the local
247 * sock as well as the remote peer.
249 if ((((AF_INET == addr->a.sa.sa_family) &&
250 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
251 (((AF_INET6 == addr->a.sa.sa_family) &&
252 (copy_flags & SCTP_ADDR6_ALLOWED) &&
253 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
254 error = sctp_add_bind_addr(bp, &addr->a,
255 SCTP_ADDR_SRC, GFP_ATOMIC);
267 /* Initialize a sctp_addr from in incoming skb. */
268 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
275 port = &addr->v4.sin_port;
276 addr->v4.sin_family = AF_INET;
281 from = &ip_hdr(skb)->saddr;
284 from = &ip_hdr(skb)->daddr;
286 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
289 /* Initialize an sctp_addr from a socket. */
290 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
292 addr->v4.sin_family = AF_INET;
293 addr->v4.sin_port = 0;
294 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
297 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
298 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
300 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
303 /* Initialize sk->sk_daddr from sctp_addr. */
304 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
306 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
309 /* Initialize a sctp_addr from an address parameter. */
310 static void sctp_v4_from_addr_param(union sctp_addr *addr,
311 union sctp_addr_param *param,
312 __be16 port, int iif)
314 addr->v4.sin_family = AF_INET;
315 addr->v4.sin_port = port;
316 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
319 /* Initialize an address parameter from a sctp_addr and return the length
320 * of the address parameter.
322 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
323 union sctp_addr_param *param)
325 int length = sizeof(sctp_ipv4addr_param_t);
327 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
328 param->v4.param_hdr.length = htons(length);
329 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
334 /* Initialize a sctp_addr from a dst_entry. */
335 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct dst_entry *dst,
338 struct rtable *rt = (struct rtable *)dst;
339 saddr->v4.sin_family = AF_INET;
340 saddr->v4.sin_port = port;
341 saddr->v4.sin_addr.s_addr = rt->rt_src;
344 /* Compare two addresses exactly. */
345 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
346 const union sctp_addr *addr2)
348 if (addr1->sa.sa_family != addr2->sa.sa_family)
350 if (addr1->v4.sin_port != addr2->v4.sin_port)
352 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
358 /* Initialize addr struct to INADDR_ANY. */
359 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
361 addr->v4.sin_family = AF_INET;
362 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
363 addr->v4.sin_port = port;
366 /* Is this a wildcard address? */
367 static int sctp_v4_is_any(const union sctp_addr *addr)
369 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
372 /* This function checks if the address is a valid address to be used for
376 * Return 0 - If the address is a non-unicast or an illegal address.
377 * Return 1 - If the address is a unicast.
379 static int sctp_v4_addr_valid(union sctp_addr *addr,
380 struct sctp_sock *sp,
381 const struct sk_buff *skb)
383 /* IPv4 addresses not allowed */
384 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
387 /* Is this a non-unicast address or a unusable SCTP address? */
388 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
391 /* Is this a broadcast address? */
392 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
398 /* Should this be available for binding? */
399 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
401 int ret = inet_addr_type(&init_net, addr->v4.sin_addr.s_addr);
404 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
406 !sp->inet.freebind &&
407 !sysctl_ip_nonlocal_bind)
410 if (ipv6_only_sock(sctp_opt2sk(sp)))
416 /* Checking the loopback, private and other address scopes as defined in
417 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
418 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
420 * Level 0 - unusable SCTP addresses
421 * Level 1 - loopback address
422 * Level 2 - link-local addresses
423 * Level 3 - private addresses.
424 * Level 4 - global addresses
425 * For INIT and INIT-ACK address list, let L be the level of
426 * of requested destination address, sender and receiver
427 * SHOULD include all of its addresses with level greater
428 * than or equal to L.
430 * IPv4 scoping can be controlled through sysctl option
431 * net.sctp.addr_scope_policy
433 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
437 /* Check for unusable SCTP addresses. */
438 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
439 retval = SCTP_SCOPE_UNUSABLE;
440 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
441 retval = SCTP_SCOPE_LOOPBACK;
442 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
443 retval = SCTP_SCOPE_LINK;
444 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
445 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
446 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
447 retval = SCTP_SCOPE_PRIVATE;
449 retval = SCTP_SCOPE_GLOBAL;
455 /* Returns a valid dst cache entry for the given source and destination ip
456 * addresses. If an association is passed, trys to get a dst entry with a
457 * source address that matches an address in the bind address list.
459 static struct dst_entry *sctp_v4_get_dst(struct sctp_association *asoc,
460 union sctp_addr *daddr,
461 union sctp_addr *saddr)
465 struct sctp_bind_addr *bp;
466 struct sctp_sockaddr_entry *laddr;
467 struct dst_entry *dst = NULL;
468 union sctp_addr dst_saddr;
470 memset(&fl4, 0x0, sizeof(struct flowi4));
471 fl4.daddr = daddr->v4.sin_addr.s_addr;
472 fl4.fl4_dport = daddr->v4.sin_port;
473 fl4.flowi4_proto = IPPROTO_SCTP;
475 fl4.flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
476 fl4.flowi4_oif = asoc->base.sk->sk_bound_dev_if;
477 fl4.fl4_sport = htons(asoc->base.bind_addr.port);
480 fl4.saddr = saddr->v4.sin_addr.s_addr;
481 fl4.fl4_sport = saddr->v4.sin_port;
484 SCTP_DEBUG_PRINTK("%s: DST:%pI4, SRC:%pI4 - ",
485 __func__, &fl4.daddr, &fl4.saddr);
487 rt = ip_route_output_key(&init_net, &fl4);
491 /* If there is no association or if a source address is passed, no
492 * more validation is required.
497 bp = &asoc->base.bind_addr;
500 /* Walk through the bind address list and look for a bind
501 * address that matches the source address of the returned dst.
503 sctp_v4_dst_saddr(&dst_saddr, dst, htons(bp->port));
505 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
506 if (!laddr->valid || (laddr->state != SCTP_ADDR_SRC))
508 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
513 /* None of the bound addresses match the source address of the
514 * dst. So release it.
520 /* Walk through the bind address list and try to get a dst that
521 * matches a bind address as the source address.
524 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
527 if ((laddr->state == SCTP_ADDR_SRC) &&
528 (AF_INET == laddr->a.sa.sa_family)) {
529 fl4.saddr = laddr->a.v4.sin_addr.s_addr;
530 fl4.fl4_sport = laddr->a.v4.sin_port;
531 rt = ip_route_output_key(&init_net, &fl4);
543 SCTP_DEBUG_PRINTK("rt_dst:%pI4, rt_src:%pI4\n",
544 &rt->rt_dst, &rt->rt_src);
546 SCTP_DEBUG_PRINTK("NO ROUTE\n");
551 /* For v4, the source address is cached in the route entry(dst). So no need
552 * to cache it separately and hence this is an empty routine.
554 static void sctp_v4_get_saddr(struct sctp_sock *sk,
555 struct sctp_association *asoc,
556 struct dst_entry *dst,
557 union sctp_addr *daddr,
558 union sctp_addr *saddr)
560 struct rtable *rt = (struct rtable *)dst;
566 saddr->v4.sin_family = AF_INET;
567 saddr->v4.sin_port = htons(asoc->base.bind_addr.port);
568 saddr->v4.sin_addr.s_addr = rt->rt_src;
572 /* What interface did this skb arrive on? */
573 static int sctp_v4_skb_iif(const struct sk_buff *skb)
575 return skb_rtable(skb)->rt_iif;
578 /* Was this packet marked by Explicit Congestion Notification? */
579 static int sctp_v4_is_ce(const struct sk_buff *skb)
581 return INET_ECN_is_ce(ip_hdr(skb)->tos);
584 /* Create and initialize a new sk for the socket returned by accept(). */
585 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
586 struct sctp_association *asoc)
588 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
590 struct inet_sock *newinet;
595 sock_init_data(NULL, newsk);
597 sctp_copy_sock(newsk, sk, asoc);
598 sock_reset_flag(newsk, SOCK_ZAPPED);
600 newinet = inet_sk(newsk);
602 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
604 sk_refcnt_debug_inc(newsk);
606 if (newsk->sk_prot->init(newsk)) {
607 sk_common_release(newsk);
615 /* Map address, empty for v4 family */
616 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
621 /* Dump the v4 addr to the seq file. */
622 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
624 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
627 static void sctp_v4_ecn_capable(struct sock *sk)
632 /* Event handler for inet address addition/deletion events.
633 * The sctp_local_addr_list needs to be protocted by a spin lock since
634 * multiple notifiers (say IPv4 and IPv6) may be running at the same
635 * time and thus corrupt the list.
636 * The reader side is protected with RCU.
638 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
641 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
642 struct sctp_sockaddr_entry *addr = NULL;
643 struct sctp_sockaddr_entry *temp;
646 if (!net_eq(dev_net(ifa->ifa_dev->dev), &init_net))
651 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
653 addr->a.v4.sin_family = AF_INET;
654 addr->a.v4.sin_port = 0;
655 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
657 spin_lock_bh(&sctp_local_addr_lock);
658 list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
659 spin_unlock_bh(&sctp_local_addr_lock);
663 spin_lock_bh(&sctp_local_addr_lock);
664 list_for_each_entry_safe(addr, temp,
665 &sctp_local_addr_list, list) {
666 if (addr->a.sa.sa_family == AF_INET &&
667 addr->a.v4.sin_addr.s_addr ==
671 list_del_rcu(&addr->list);
675 spin_unlock_bh(&sctp_local_addr_lock);
677 kfree_rcu(addr, rcu);
685 * Initialize the control inode/socket with a control endpoint data
686 * structure. This endpoint is reserved exclusively for the OOTB processing.
688 static int sctp_ctl_sock_init(void)
691 sa_family_t family = PF_INET;
693 if (sctp_get_pf_specific(PF_INET6))
696 err = inet_ctl_sock_create(&sctp_ctl_sock, family,
697 SOCK_SEQPACKET, IPPROTO_SCTP, &init_net);
699 /* If IPv6 socket could not be created, try the IPv4 socket */
700 if (err < 0 && family == PF_INET6)
701 err = inet_ctl_sock_create(&sctp_ctl_sock, AF_INET,
702 SOCK_SEQPACKET, IPPROTO_SCTP,
706 pr_err("Failed to create the SCTP control socket\n");
712 /* Register address family specific functions. */
713 int sctp_register_af(struct sctp_af *af)
715 switch (af->sa_family) {
717 if (sctp_af_v4_specific)
719 sctp_af_v4_specific = af;
722 if (sctp_af_v6_specific)
724 sctp_af_v6_specific = af;
730 INIT_LIST_HEAD(&af->list);
731 list_add_tail(&af->list, &sctp_address_families);
735 /* Get the table of functions for manipulating a particular address
738 struct sctp_af *sctp_get_af_specific(sa_family_t family)
742 return sctp_af_v4_specific;
744 return sctp_af_v6_specific;
750 /* Common code to initialize a AF_INET msg_name. */
751 static void sctp_inet_msgname(char *msgname, int *addr_len)
753 struct sockaddr_in *sin;
755 sin = (struct sockaddr_in *)msgname;
756 *addr_len = sizeof(struct sockaddr_in);
757 sin->sin_family = AF_INET;
758 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
761 /* Copy the primary address of the peer primary address as the msg_name. */
762 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
765 struct sockaddr_in *sin, *sinfrom;
768 struct sctp_association *asoc;
771 sctp_inet_msgname(msgname, addr_len);
772 sin = (struct sockaddr_in *)msgname;
773 sinfrom = &asoc->peer.primary_addr.v4;
774 sin->sin_port = htons(asoc->peer.port);
775 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
779 /* Initialize and copy out a msgname from an inbound skb. */
780 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
783 struct sctphdr *sh = sctp_hdr(skb);
784 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
786 sctp_inet_msgname(msgname, len);
787 sin->sin_port = sh->source;
788 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
792 /* Do we support this AF? */
793 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
795 /* PF_INET only supports AF_INET addresses. */
796 return AF_INET == family;
799 /* Address matching with wildcards allowed. */
800 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
801 const union sctp_addr *addr2,
802 struct sctp_sock *opt)
804 /* PF_INET only supports AF_INET addresses. */
805 if (addr1->sa.sa_family != addr2->sa.sa_family)
807 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
808 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
810 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
816 /* Verify that provided sockaddr looks bindable. Common verification has
817 * already been taken care of.
819 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
821 return sctp_v4_available(addr, opt);
824 /* Verify that sockaddr looks sendable. Common verification has already
825 * been taken care of.
827 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
832 /* Fill in Supported Address Type information for INIT and INIT-ACK
833 * chunks. Returns number of addresses supported.
835 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
838 types[0] = SCTP_PARAM_IPV4_ADDRESS;
842 /* Wrapper routine that calls the ip transmit routine. */
843 static inline int sctp_v4_xmit(struct sk_buff *skb,
844 struct sctp_transport *transport)
846 struct inet_sock *inet = inet_sk(skb->sk);
848 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n",
849 __func__, skb, skb->len,
850 &skb_rtable(skb)->rt_src,
851 &skb_rtable(skb)->rt_dst);
853 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
854 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
856 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
857 return ip_queue_xmit(skb);
860 static struct sctp_af sctp_af_inet;
862 static struct sctp_pf sctp_pf_inet = {
863 .event_msgname = sctp_inet_event_msgname,
864 .skb_msgname = sctp_inet_skb_msgname,
865 .af_supported = sctp_inet_af_supported,
866 .cmp_addr = sctp_inet_cmp_addr,
867 .bind_verify = sctp_inet_bind_verify,
868 .send_verify = sctp_inet_send_verify,
869 .supported_addrs = sctp_inet_supported_addrs,
870 .create_accept_sk = sctp_v4_create_accept_sk,
871 .addr_v4map = sctp_v4_addr_v4map,
875 /* Notifier for inetaddr addition/deletion events. */
876 static struct notifier_block sctp_inetaddr_notifier = {
877 .notifier_call = sctp_inetaddr_event,
880 /* Socket operations. */
881 static const struct proto_ops inet_seqpacket_ops = {
883 .owner = THIS_MODULE,
884 .release = inet_release, /* Needs to be wrapped... */
886 .connect = inet_dgram_connect,
887 .socketpair = sock_no_socketpair,
888 .accept = inet_accept,
889 .getname = inet_getname, /* Semantics are different. */
892 .listen = sctp_inet_listen,
893 .shutdown = inet_shutdown, /* Looks harmless. */
894 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
895 .getsockopt = sock_common_getsockopt,
896 .sendmsg = inet_sendmsg,
897 .recvmsg = sock_common_recvmsg,
898 .mmap = sock_no_mmap,
899 .sendpage = sock_no_sendpage,
901 .compat_setsockopt = compat_sock_common_setsockopt,
902 .compat_getsockopt = compat_sock_common_getsockopt,
906 /* Registration with AF_INET family. */
907 static struct inet_protosw sctp_seqpacket_protosw = {
908 .type = SOCK_SEQPACKET,
909 .protocol = IPPROTO_SCTP,
911 .ops = &inet_seqpacket_ops,
913 .flags = SCTP_PROTOSW_FLAG
915 static struct inet_protosw sctp_stream_protosw = {
917 .protocol = IPPROTO_SCTP,
919 .ops = &inet_seqpacket_ops,
921 .flags = SCTP_PROTOSW_FLAG
924 /* Register with IP layer. */
925 static const struct net_protocol sctp_protocol = {
927 .err_handler = sctp_v4_err,
931 /* IPv4 address related functions. */
932 static struct sctp_af sctp_af_inet = {
933 .sa_family = AF_INET,
934 .sctp_xmit = sctp_v4_xmit,
935 .setsockopt = ip_setsockopt,
936 .getsockopt = ip_getsockopt,
937 .get_dst = sctp_v4_get_dst,
938 .get_saddr = sctp_v4_get_saddr,
939 .copy_addrlist = sctp_v4_copy_addrlist,
940 .from_skb = sctp_v4_from_skb,
941 .from_sk = sctp_v4_from_sk,
942 .to_sk_saddr = sctp_v4_to_sk_saddr,
943 .to_sk_daddr = sctp_v4_to_sk_daddr,
944 .from_addr_param = sctp_v4_from_addr_param,
945 .to_addr_param = sctp_v4_to_addr_param,
946 .dst_saddr = sctp_v4_dst_saddr,
947 .cmp_addr = sctp_v4_cmp_addr,
948 .addr_valid = sctp_v4_addr_valid,
949 .inaddr_any = sctp_v4_inaddr_any,
950 .is_any = sctp_v4_is_any,
951 .available = sctp_v4_available,
952 .scope = sctp_v4_scope,
953 .skb_iif = sctp_v4_skb_iif,
954 .is_ce = sctp_v4_is_ce,
955 .seq_dump_addr = sctp_v4_seq_dump_addr,
956 .ecn_capable = sctp_v4_ecn_capable,
957 .net_header_len = sizeof(struct iphdr),
958 .sockaddr_len = sizeof(struct sockaddr_in),
960 .compat_setsockopt = compat_ip_setsockopt,
961 .compat_getsockopt = compat_ip_getsockopt,
965 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
969 return sctp_pf_inet_specific;
971 return sctp_pf_inet6_specific;
977 /* Register the PF specific function table. */
978 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
982 if (sctp_pf_inet_specific)
984 sctp_pf_inet_specific = pf;
987 if (sctp_pf_inet6_specific)
989 sctp_pf_inet6_specific = pf;
997 static inline int init_sctp_mibs(void)
999 return snmp_mib_init((void __percpu **)sctp_statistics,
1000 sizeof(struct sctp_mib),
1001 __alignof__(struct sctp_mib));
1004 static inline void cleanup_sctp_mibs(void)
1006 snmp_mib_free((void __percpu **)sctp_statistics);
1009 static void sctp_v4_pf_init(void)
1011 /* Initialize the SCTP specific PF functions. */
1012 sctp_register_pf(&sctp_pf_inet, PF_INET);
1013 sctp_register_af(&sctp_af_inet);
1016 static void sctp_v4_pf_exit(void)
1018 list_del(&sctp_af_inet.list);
1021 static int sctp_v4_protosw_init(void)
1025 rc = proto_register(&sctp_prot, 1);
1029 /* Register SCTP(UDP and TCP style) with socket layer. */
1030 inet_register_protosw(&sctp_seqpacket_protosw);
1031 inet_register_protosw(&sctp_stream_protosw);
1036 static void sctp_v4_protosw_exit(void)
1038 inet_unregister_protosw(&sctp_stream_protosw);
1039 inet_unregister_protosw(&sctp_seqpacket_protosw);
1040 proto_unregister(&sctp_prot);
1043 static int sctp_v4_add_protocol(void)
1045 /* Register notifier for inet address additions/deletions. */
1046 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1048 /* Register SCTP with inet layer. */
1049 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1055 static void sctp_v4_del_protocol(void)
1057 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1058 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1061 /* Initialize the universe into something sensible. */
1062 SCTP_STATIC __init int sctp_init(void)
1065 int status = -EINVAL;
1067 unsigned long limit;
1068 unsigned long nr_pages;
1072 /* SCTP_DEBUG sanity check. */
1073 if (!sctp_sanity_check())
1076 /* Allocate bind_bucket and chunk caches. */
1078 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1079 sizeof(struct sctp_bind_bucket),
1080 0, SLAB_HWCACHE_ALIGN,
1082 if (!sctp_bucket_cachep)
1085 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1086 sizeof(struct sctp_chunk),
1087 0, SLAB_HWCACHE_ALIGN,
1089 if (!sctp_chunk_cachep)
1090 goto err_chunk_cachep;
1092 /* Allocate and initialise sctp mibs. */
1093 status = init_sctp_mibs();
1097 /* Initialize proc fs directory. */
1098 status = sctp_proc_init();
1102 /* Initialize object count debugging. */
1103 sctp_dbg_objcnt_init();
1106 * 14. Suggested SCTP Protocol Parameter Values
1108 /* The following protocol parameters are RECOMMENDED: */
1109 /* RTO.Initial - 3 seconds */
1110 sctp_rto_initial = SCTP_RTO_INITIAL;
1111 /* RTO.Min - 1 second */
1112 sctp_rto_min = SCTP_RTO_MIN;
1113 /* RTO.Max - 60 seconds */
1114 sctp_rto_max = SCTP_RTO_MAX;
1115 /* RTO.Alpha - 1/8 */
1116 sctp_rto_alpha = SCTP_RTO_ALPHA;
1117 /* RTO.Beta - 1/4 */
1118 sctp_rto_beta = SCTP_RTO_BETA;
1120 /* Valid.Cookie.Life - 60 seconds */
1121 sctp_valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1123 /* Whether Cookie Preservative is enabled(1) or not(0) */
1124 sctp_cookie_preserve_enable = 1;
1127 sctp_max_burst = SCTP_DEFAULT_MAX_BURST;
1129 /* Association.Max.Retrans - 10 attempts
1130 * Path.Max.Retrans - 5 attempts (per destination address)
1131 * Max.Init.Retransmits - 8 attempts
1133 sctp_max_retrans_association = 10;
1134 sctp_max_retrans_path = 5;
1135 sctp_max_retrans_init = 8;
1137 /* Sendbuffer growth - do per-socket accounting */
1138 sctp_sndbuf_policy = 0;
1140 /* Rcvbuffer growth - do per-socket accounting */
1141 sctp_rcvbuf_policy = 0;
1143 /* HB.interval - 30 seconds */
1144 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1146 /* delayed SACK timeout */
1147 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1149 /* Implementation specific variables. */
1151 /* Initialize default stream count setup information. */
1152 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1153 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1155 /* Initialize handle used for association ids. */
1156 idr_init(&sctp_assocs_id);
1158 /* Set the pressure threshold to be a fraction of global memory that
1159 * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
1160 * memory, with a floor of 128 pages.
1161 * Note this initializes the data in sctpv6_prot too
1162 * Unabashedly stolen from tcp_init
1164 nr_pages = totalram_pages - totalhigh_pages;
1165 limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1166 limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
1167 limit = max(limit, 128UL);
1168 sysctl_sctp_mem[0] = limit / 4 * 3;
1169 sysctl_sctp_mem[1] = limit;
1170 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1172 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1173 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1174 max_share = min(4UL*1024*1024, limit);
1176 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1177 sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
1178 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1180 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1181 sysctl_sctp_wmem[1] = 16*1024;
1182 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1184 /* Size and allocate the association hash table.
1185 * The methodology is similar to that of the tcp hash tables.
1187 if (totalram_pages >= (128 * 1024))
1188 goal = totalram_pages >> (22 - PAGE_SHIFT);
1190 goal = totalram_pages >> (24 - PAGE_SHIFT);
1192 for (order = 0; (1UL << order) < goal; order++)
1196 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1197 sizeof(struct sctp_hashbucket);
1198 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1200 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1201 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1202 } while (!sctp_assoc_hashtable && --order > 0);
1203 if (!sctp_assoc_hashtable) {
1204 pr_err("Failed association hash alloc\n");
1206 goto err_ahash_alloc;
1208 for (i = 0; i < sctp_assoc_hashsize; i++) {
1209 rwlock_init(&sctp_assoc_hashtable[i].lock);
1210 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1213 /* Allocate and initialize the endpoint hash table. */
1214 sctp_ep_hashsize = 64;
1215 sctp_ep_hashtable = (struct sctp_hashbucket *)
1216 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1217 if (!sctp_ep_hashtable) {
1218 pr_err("Failed endpoint_hash alloc\n");
1220 goto err_ehash_alloc;
1222 for (i = 0; i < sctp_ep_hashsize; i++) {
1223 rwlock_init(&sctp_ep_hashtable[i].lock);
1224 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1227 /* Allocate and initialize the SCTP port hash table. */
1229 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1230 sizeof(struct sctp_bind_hashbucket);
1231 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1233 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1234 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1235 } while (!sctp_port_hashtable && --order > 0);
1236 if (!sctp_port_hashtable) {
1237 pr_err("Failed bind hash alloc\n");
1239 goto err_bhash_alloc;
1241 for (i = 0; i < sctp_port_hashsize; i++) {
1242 spin_lock_init(&sctp_port_hashtable[i].lock);
1243 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1246 pr_info("Hash tables configured (established %d bind %d)\n",
1247 sctp_assoc_hashsize, sctp_port_hashsize);
1249 /* Disable ADDIP by default. */
1250 sctp_addip_enable = 0;
1251 sctp_addip_noauth = 0;
1253 /* Enable PR-SCTP by default. */
1254 sctp_prsctp_enable = 1;
1256 /* Disable AUTH by default. */
1257 sctp_auth_enable = 0;
1259 /* Set SCOPE policy to enabled */
1260 sctp_scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1262 /* Set the default rwnd update threshold */
1263 sctp_rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1265 sctp_sysctl_register();
1267 INIT_LIST_HEAD(&sctp_address_families);
1271 /* Initialize the local address list. */
1272 INIT_LIST_HEAD(&sctp_local_addr_list);
1273 spin_lock_init(&sctp_local_addr_lock);
1274 sctp_get_local_addr_list();
1276 status = sctp_v4_protosw_init();
1279 goto err_protosw_init;
1281 status = sctp_v6_protosw_init();
1283 goto err_v6_protosw_init;
1285 /* Initialize the control inode/socket for handling OOTB packets. */
1286 if ((status = sctp_ctl_sock_init())) {
1287 pr_err("Failed to initialize the SCTP control sock\n");
1288 goto err_ctl_sock_init;
1291 status = sctp_v4_add_protocol();
1293 goto err_add_protocol;
1295 /* Register SCTP with inet6 layer. */
1296 status = sctp_v6_add_protocol();
1298 goto err_v6_add_protocol;
1303 err_v6_add_protocol:
1304 sctp_v4_del_protocol();
1306 inet_ctl_sock_destroy(sctp_ctl_sock);
1308 sctp_v6_protosw_exit();
1309 err_v6_protosw_init:
1310 sctp_v4_protosw_exit();
1312 sctp_free_local_addr_list();
1315 sctp_sysctl_unregister();
1316 free_pages((unsigned long)sctp_port_hashtable,
1317 get_order(sctp_port_hashsize *
1318 sizeof(struct sctp_bind_hashbucket)));
1320 kfree(sctp_ep_hashtable);
1322 free_pages((unsigned long)sctp_assoc_hashtable,
1323 get_order(sctp_assoc_hashsize *
1324 sizeof(struct sctp_hashbucket)));
1326 sctp_dbg_objcnt_exit();
1329 cleanup_sctp_mibs();
1331 kmem_cache_destroy(sctp_chunk_cachep);
1333 kmem_cache_destroy(sctp_bucket_cachep);
1337 /* Exit handler for the SCTP protocol. */
1338 SCTP_STATIC __exit void sctp_exit(void)
1340 /* BUG. This should probably do something useful like clean
1341 * up all the remaining associations and all that memory.
1344 /* Unregister with inet6/inet layers. */
1345 sctp_v6_del_protocol();
1346 sctp_v4_del_protocol();
1348 /* Free the control endpoint. */
1349 inet_ctl_sock_destroy(sctp_ctl_sock);
1351 /* Free protosw registrations */
1352 sctp_v6_protosw_exit();
1353 sctp_v4_protosw_exit();
1355 /* Free the local address list. */
1356 sctp_free_local_addr_list();
1358 /* Unregister with socket layer. */
1362 sctp_sysctl_unregister();
1364 free_pages((unsigned long)sctp_assoc_hashtable,
1365 get_order(sctp_assoc_hashsize *
1366 sizeof(struct sctp_hashbucket)));
1367 kfree(sctp_ep_hashtable);
1368 free_pages((unsigned long)sctp_port_hashtable,
1369 get_order(sctp_port_hashsize *
1370 sizeof(struct sctp_bind_hashbucket)));
1372 sctp_dbg_objcnt_exit();
1374 cleanup_sctp_mibs();
1376 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1378 kmem_cache_destroy(sctp_chunk_cachep);
1379 kmem_cache_destroy(sctp_bucket_cachep);
1382 module_init(sctp_init);
1383 module_exit(sctp_exit);
1386 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1388 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1389 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1390 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1391 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1392 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1393 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1394 MODULE_LICENSE("GPL");