x86/xen: resume timer irqs early
[linux/fpc-iii.git] / net / sctp / protocol.c
blob2b216f1f6b239f6c105af2e2261235b2f772cd80
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)
17 * any later version.
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
31 * email address(es):
32 * lksctp developers <linux-sctp@vger.kernel.org>
34 * Written or modified by:
35 * La Monte H.P. Yarroll <piggy@acm.org>
36 * Karl Knutson <karl@athena.chicago.il.us>
37 * Jon Grimm <jgrimm@us.ibm.com>
38 * Sridhar Samudrala <sri@us.ibm.com>
39 * Daisy Chang <daisyc@us.ibm.com>
40 * Ardelle Fan <ardelle.fan@intel.com>
43 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45 #include <linux/module.h>
46 #include <linux/init.h>
47 #include <linux/netdevice.h>
48 #include <linux/inetdevice.h>
49 #include <linux/seq_file.h>
50 #include <linux/bootmem.h>
51 #include <linux/highmem.h>
52 #include <linux/swap.h>
53 #include <linux/slab.h>
54 #include <net/net_namespace.h>
55 #include <net/protocol.h>
56 #include <net/ip.h>
57 #include <net/ipv6.h>
58 #include <net/route.h>
59 #include <net/sctp/sctp.h>
60 #include <net/addrconf.h>
61 #include <net/inet_common.h>
62 #include <net/inet_ecn.h>
64 /* Global data structures. */
65 struct sctp_globals sctp_globals __read_mostly;
67 struct idr sctp_assocs_id;
68 DEFINE_SPINLOCK(sctp_assocs_id_lock);
70 static struct sctp_pf *sctp_pf_inet6_specific;
71 static struct sctp_pf *sctp_pf_inet_specific;
72 static struct sctp_af *sctp_af_v4_specific;
73 static struct sctp_af *sctp_af_v6_specific;
75 struct kmem_cache *sctp_chunk_cachep __read_mostly;
76 struct kmem_cache *sctp_bucket_cachep __read_mostly;
78 long sysctl_sctp_mem[3];
79 int sysctl_sctp_rmem[3];
80 int sysctl_sctp_wmem[3];
82 /* Set up the proc fs entry for the SCTP protocol. */
83 static int __net_init sctp_proc_init(struct net *net)
85 #ifdef CONFIG_PROC_FS
86 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
87 if (!net->sctp.proc_net_sctp)
88 goto out_proc_net_sctp;
89 if (sctp_snmp_proc_init(net))
90 goto out_snmp_proc_init;
91 if (sctp_eps_proc_init(net))
92 goto out_eps_proc_init;
93 if (sctp_assocs_proc_init(net))
94 goto out_assocs_proc_init;
95 if (sctp_remaddr_proc_init(net))
96 goto out_remaddr_proc_init;
98 return 0;
100 out_remaddr_proc_init:
101 sctp_assocs_proc_exit(net);
102 out_assocs_proc_init:
103 sctp_eps_proc_exit(net);
104 out_eps_proc_init:
105 sctp_snmp_proc_exit(net);
106 out_snmp_proc_init:
107 remove_proc_entry("sctp", net->proc_net);
108 net->sctp.proc_net_sctp = NULL;
109 out_proc_net_sctp:
110 return -ENOMEM;
111 #endif /* CONFIG_PROC_FS */
112 return 0;
115 /* Clean up the proc fs entry for the SCTP protocol.
116 * Note: Do not make this __exit as it is used in the init error
117 * path.
119 static void sctp_proc_exit(struct net *net)
121 #ifdef CONFIG_PROC_FS
122 sctp_snmp_proc_exit(net);
123 sctp_eps_proc_exit(net);
124 sctp_assocs_proc_exit(net);
125 sctp_remaddr_proc_exit(net);
127 remove_proc_entry("sctp", net->proc_net);
128 net->sctp.proc_net_sctp = NULL;
129 #endif
132 /* Private helper to extract ipv4 address and stash them in
133 * the protocol structure.
135 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
136 struct net_device *dev)
138 struct in_device *in_dev;
139 struct in_ifaddr *ifa;
140 struct sctp_sockaddr_entry *addr;
142 rcu_read_lock();
143 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
144 rcu_read_unlock();
145 return;
148 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
149 /* Add the address to the local list. */
150 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
151 if (addr) {
152 addr->a.v4.sin_family = AF_INET;
153 addr->a.v4.sin_port = 0;
154 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
155 addr->valid = 1;
156 INIT_LIST_HEAD(&addr->list);
157 list_add_tail(&addr->list, addrlist);
161 rcu_read_unlock();
164 /* Extract our IP addresses from the system and stash them in the
165 * protocol structure.
167 static void sctp_get_local_addr_list(struct net *net)
169 struct net_device *dev;
170 struct list_head *pos;
171 struct sctp_af *af;
173 rcu_read_lock();
174 for_each_netdev_rcu(net, dev) {
175 list_for_each(pos, &sctp_address_families) {
176 af = list_entry(pos, struct sctp_af, list);
177 af->copy_addrlist(&net->sctp.local_addr_list, dev);
180 rcu_read_unlock();
183 /* Free the existing local addresses. */
184 static void sctp_free_local_addr_list(struct net *net)
186 struct sctp_sockaddr_entry *addr;
187 struct list_head *pos, *temp;
189 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
190 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
191 list_del(pos);
192 kfree(addr);
196 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
197 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
198 sctp_scope_t scope, gfp_t gfp, int copy_flags)
200 struct sctp_sockaddr_entry *addr;
201 int error = 0;
203 rcu_read_lock();
204 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
205 if (!addr->valid)
206 continue;
207 if (sctp_in_scope(net, &addr->a, scope)) {
208 /* Now that the address is in scope, check to see if
209 * the address type is really supported by the local
210 * sock as well as the remote peer.
212 if ((((AF_INET == addr->a.sa.sa_family) &&
213 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
214 (((AF_INET6 == addr->a.sa.sa_family) &&
215 (copy_flags & SCTP_ADDR6_ALLOWED) &&
216 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
217 error = sctp_add_bind_addr(bp, &addr->a,
218 SCTP_ADDR_SRC, GFP_ATOMIC);
219 if (error)
220 goto end_copy;
225 end_copy:
226 rcu_read_unlock();
227 return error;
230 /* Initialize a sctp_addr from in incoming skb. */
231 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
232 int is_saddr)
234 void *from;
235 __be16 *port;
236 struct sctphdr *sh;
238 port = &addr->v4.sin_port;
239 addr->v4.sin_family = AF_INET;
241 sh = sctp_hdr(skb);
242 if (is_saddr) {
243 *port = sh->source;
244 from = &ip_hdr(skb)->saddr;
245 } else {
246 *port = sh->dest;
247 from = &ip_hdr(skb)->daddr;
249 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
252 /* Initialize an sctp_addr from a socket. */
253 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
255 addr->v4.sin_family = AF_INET;
256 addr->v4.sin_port = 0;
257 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
260 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
261 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
263 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
266 /* Initialize sk->sk_daddr from sctp_addr. */
267 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
269 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
272 /* Initialize a sctp_addr from an address parameter. */
273 static void sctp_v4_from_addr_param(union sctp_addr *addr,
274 union sctp_addr_param *param,
275 __be16 port, int iif)
277 addr->v4.sin_family = AF_INET;
278 addr->v4.sin_port = port;
279 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
282 /* Initialize an address parameter from a sctp_addr and return the length
283 * of the address parameter.
285 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
286 union sctp_addr_param *param)
288 int length = sizeof(sctp_ipv4addr_param_t);
290 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
291 param->v4.param_hdr.length = htons(length);
292 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
294 return length;
297 /* Initialize a sctp_addr from a dst_entry. */
298 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
299 __be16 port)
301 saddr->v4.sin_family = AF_INET;
302 saddr->v4.sin_port = port;
303 saddr->v4.sin_addr.s_addr = fl4->saddr;
306 /* Compare two addresses exactly. */
307 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
308 const union sctp_addr *addr2)
310 if (addr1->sa.sa_family != addr2->sa.sa_family)
311 return 0;
312 if (addr1->v4.sin_port != addr2->v4.sin_port)
313 return 0;
314 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
315 return 0;
317 return 1;
320 /* Initialize addr struct to INADDR_ANY. */
321 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
323 addr->v4.sin_family = AF_INET;
324 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
325 addr->v4.sin_port = port;
328 /* Is this a wildcard address? */
329 static int sctp_v4_is_any(const union sctp_addr *addr)
331 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
334 /* This function checks if the address is a valid address to be used for
335 * SCTP binding.
337 * Output:
338 * Return 0 - If the address is a non-unicast or an illegal address.
339 * Return 1 - If the address is a unicast.
341 static int sctp_v4_addr_valid(union sctp_addr *addr,
342 struct sctp_sock *sp,
343 const struct sk_buff *skb)
345 /* IPv4 addresses not allowed */
346 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
347 return 0;
349 /* Is this a non-unicast address or a unusable SCTP address? */
350 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
351 return 0;
353 /* Is this a broadcast address? */
354 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
355 return 0;
357 return 1;
360 /* Should this be available for binding? */
361 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
363 struct net *net = sock_net(&sp->inet.sk);
364 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
367 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
368 ret != RTN_LOCAL &&
369 !sp->inet.freebind &&
370 !sysctl_ip_nonlocal_bind)
371 return 0;
373 if (ipv6_only_sock(sctp_opt2sk(sp)))
374 return 0;
376 return 1;
379 /* Checking the loopback, private and other address scopes as defined in
380 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
381 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
383 * Level 0 - unusable SCTP addresses
384 * Level 1 - loopback address
385 * Level 2 - link-local addresses
386 * Level 3 - private addresses.
387 * Level 4 - global addresses
388 * For INIT and INIT-ACK address list, let L be the level of
389 * of requested destination address, sender and receiver
390 * SHOULD include all of its addresses with level greater
391 * than or equal to L.
393 * IPv4 scoping can be controlled through sysctl option
394 * net.sctp.addr_scope_policy
396 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
398 sctp_scope_t retval;
400 /* Check for unusable SCTP addresses. */
401 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
402 retval = SCTP_SCOPE_UNUSABLE;
403 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
404 retval = SCTP_SCOPE_LOOPBACK;
405 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
406 retval = SCTP_SCOPE_LINK;
407 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
408 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
409 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
410 retval = SCTP_SCOPE_PRIVATE;
411 } else {
412 retval = SCTP_SCOPE_GLOBAL;
415 return retval;
418 /* Returns a valid dst cache entry for the given source and destination ip
419 * addresses. If an association is passed, trys to get a dst entry with a
420 * source address that matches an address in the bind address list.
422 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
423 struct flowi *fl, struct sock *sk)
425 struct sctp_association *asoc = t->asoc;
426 struct rtable *rt;
427 struct flowi4 *fl4 = &fl->u.ip4;
428 struct sctp_bind_addr *bp;
429 struct sctp_sockaddr_entry *laddr;
430 struct dst_entry *dst = NULL;
431 union sctp_addr *daddr = &t->ipaddr;
432 union sctp_addr dst_saddr;
434 memset(fl4, 0x0, sizeof(struct flowi4));
435 fl4->daddr = daddr->v4.sin_addr.s_addr;
436 fl4->fl4_dport = daddr->v4.sin_port;
437 fl4->flowi4_proto = IPPROTO_SCTP;
438 if (asoc) {
439 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
440 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
441 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
443 if (saddr) {
444 fl4->saddr = saddr->v4.sin_addr.s_addr;
445 fl4->fl4_sport = saddr->v4.sin_port;
448 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
449 &fl4->saddr);
451 rt = ip_route_output_key(sock_net(sk), fl4);
452 if (!IS_ERR(rt))
453 dst = &rt->dst;
455 /* If there is no association or if a source address is passed, no
456 * more validation is required.
458 if (!asoc || saddr)
459 goto out;
461 bp = &asoc->base.bind_addr;
463 if (dst) {
464 /* Walk through the bind address list and look for a bind
465 * address that matches the source address of the returned dst.
467 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
468 rcu_read_lock();
469 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
470 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
471 (laddr->state != SCTP_ADDR_SRC &&
472 !asoc->src_out_of_asoc_ok))
473 continue;
474 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
475 goto out_unlock;
477 rcu_read_unlock();
479 /* None of the bound addresses match the source address of the
480 * dst. So release it.
482 dst_release(dst);
483 dst = NULL;
486 /* Walk through the bind address list and try to get a dst that
487 * matches a bind address as the source address.
489 rcu_read_lock();
490 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
491 if (!laddr->valid)
492 continue;
493 if ((laddr->state == SCTP_ADDR_SRC) &&
494 (AF_INET == laddr->a.sa.sa_family)) {
495 fl4->fl4_sport = laddr->a.v4.sin_port;
496 flowi4_update_output(fl4,
497 asoc->base.sk->sk_bound_dev_if,
498 RT_CONN_FLAGS(asoc->base.sk),
499 daddr->v4.sin_addr.s_addr,
500 laddr->a.v4.sin_addr.s_addr);
502 rt = ip_route_output_key(sock_net(sk), fl4);
503 if (!IS_ERR(rt)) {
504 dst = &rt->dst;
505 goto out_unlock;
510 out_unlock:
511 rcu_read_unlock();
512 out:
513 t->dst = dst;
514 if (dst)
515 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
516 &fl4->daddr, &fl4->saddr);
517 else
518 pr_debug("no route\n");
521 /* For v4, the source address is cached in the route entry(dst). So no need
522 * to cache it separately and hence this is an empty routine.
524 static void sctp_v4_get_saddr(struct sctp_sock *sk,
525 struct sctp_transport *t,
526 struct flowi *fl)
528 union sctp_addr *saddr = &t->saddr;
529 struct rtable *rt = (struct rtable *)t->dst;
531 if (rt) {
532 saddr->v4.sin_family = AF_INET;
533 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
537 /* What interface did this skb arrive on? */
538 static int sctp_v4_skb_iif(const struct sk_buff *skb)
540 return inet_iif(skb);
543 /* Was this packet marked by Explicit Congestion Notification? */
544 static int sctp_v4_is_ce(const struct sk_buff *skb)
546 return INET_ECN_is_ce(ip_hdr(skb)->tos);
549 /* Create and initialize a new sk for the socket returned by accept(). */
550 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
551 struct sctp_association *asoc)
553 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
554 sk->sk_prot);
555 struct inet_sock *newinet;
557 if (!newsk)
558 goto out;
560 sock_init_data(NULL, newsk);
562 sctp_copy_sock(newsk, sk, asoc);
563 sock_reset_flag(newsk, SOCK_ZAPPED);
565 newinet = inet_sk(newsk);
567 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
569 sk_refcnt_debug_inc(newsk);
571 if (newsk->sk_prot->init(newsk)) {
572 sk_common_release(newsk);
573 newsk = NULL;
576 out:
577 return newsk;
580 /* Map address, empty for v4 family */
581 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
583 /* Empty */
586 /* Dump the v4 addr to the seq file. */
587 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
589 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
592 static void sctp_v4_ecn_capable(struct sock *sk)
594 INET_ECN_xmit(sk);
597 static void sctp_addr_wq_timeout_handler(unsigned long arg)
599 struct net *net = (struct net *)arg;
600 struct sctp_sockaddr_entry *addrw, *temp;
601 struct sctp_sock *sp;
603 spin_lock_bh(&net->sctp.addr_wq_lock);
605 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
606 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
607 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
608 addrw->state, addrw);
610 #if IS_ENABLED(CONFIG_IPV6)
611 /* Now we send an ASCONF for each association */
612 /* Note. we currently don't handle link local IPv6 addressees */
613 if (addrw->a.sa.sa_family == AF_INET6) {
614 struct in6_addr *in6;
616 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
617 IPV6_ADDR_LINKLOCAL)
618 goto free_next;
620 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
621 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
622 addrw->state == SCTP_ADDR_NEW) {
623 unsigned long timeo_val;
625 pr_debug("%s: this is on DAD, trying %d sec "
626 "later\n", __func__,
627 SCTP_ADDRESS_TICK_DELAY);
629 timeo_val = jiffies;
630 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
631 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
632 break;
635 #endif
636 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
637 struct sock *sk;
639 sk = sctp_opt2sk(sp);
640 /* ignore bound-specific endpoints */
641 if (!sctp_is_ep_boundall(sk))
642 continue;
643 sctp_bh_lock_sock(sk);
644 if (sctp_asconf_mgmt(sp, addrw) < 0)
645 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
646 sctp_bh_unlock_sock(sk);
648 #if IS_ENABLED(CONFIG_IPV6)
649 free_next:
650 #endif
651 list_del(&addrw->list);
652 kfree(addrw);
654 spin_unlock_bh(&net->sctp.addr_wq_lock);
657 static void sctp_free_addr_wq(struct net *net)
659 struct sctp_sockaddr_entry *addrw;
660 struct sctp_sockaddr_entry *temp;
662 spin_lock_bh(&net->sctp.addr_wq_lock);
663 del_timer(&net->sctp.addr_wq_timer);
664 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
665 list_del(&addrw->list);
666 kfree(addrw);
668 spin_unlock_bh(&net->sctp.addr_wq_lock);
671 /* lookup the entry for the same address in the addr_waitq
672 * sctp_addr_wq MUST be locked
674 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
675 struct sctp_sockaddr_entry *addr)
677 struct sctp_sockaddr_entry *addrw;
679 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
680 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
681 continue;
682 if (addrw->a.sa.sa_family == AF_INET) {
683 if (addrw->a.v4.sin_addr.s_addr ==
684 addr->a.v4.sin_addr.s_addr)
685 return addrw;
686 } else if (addrw->a.sa.sa_family == AF_INET6) {
687 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
688 &addr->a.v6.sin6_addr))
689 return addrw;
692 return NULL;
695 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
697 struct sctp_sockaddr_entry *addrw;
698 unsigned long timeo_val;
700 /* first, we check if an opposite message already exist in the queue.
701 * If we found such message, it is removed.
702 * This operation is a bit stupid, but the DHCP client attaches the
703 * new address after a couple of addition and deletion of that address
706 spin_lock_bh(&net->sctp.addr_wq_lock);
707 /* Offsets existing events in addr_wq */
708 addrw = sctp_addr_wq_lookup(net, addr);
709 if (addrw) {
710 if (addrw->state != cmd) {
711 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
712 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
713 &net->sctp.addr_waitq);
715 list_del(&addrw->list);
716 kfree(addrw);
718 spin_unlock_bh(&net->sctp.addr_wq_lock);
719 return;
722 /* OK, we have to add the new address to the wait queue */
723 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
724 if (addrw == NULL) {
725 spin_unlock_bh(&net->sctp.addr_wq_lock);
726 return;
728 addrw->state = cmd;
729 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
731 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
732 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
734 if (!timer_pending(&net->sctp.addr_wq_timer)) {
735 timeo_val = jiffies;
736 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
737 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
739 spin_unlock_bh(&net->sctp.addr_wq_lock);
742 /* Event handler for inet address addition/deletion events.
743 * The sctp_local_addr_list needs to be protocted by a spin lock since
744 * multiple notifiers (say IPv4 and IPv6) may be running at the same
745 * time and thus corrupt the list.
746 * The reader side is protected with RCU.
748 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
749 void *ptr)
751 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
752 struct sctp_sockaddr_entry *addr = NULL;
753 struct sctp_sockaddr_entry *temp;
754 struct net *net = dev_net(ifa->ifa_dev->dev);
755 int found = 0;
757 switch (ev) {
758 case NETDEV_UP:
759 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
760 if (addr) {
761 addr->a.v4.sin_family = AF_INET;
762 addr->a.v4.sin_port = 0;
763 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
764 addr->valid = 1;
765 spin_lock_bh(&net->sctp.local_addr_lock);
766 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
767 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
768 spin_unlock_bh(&net->sctp.local_addr_lock);
770 break;
771 case NETDEV_DOWN:
772 spin_lock_bh(&net->sctp.local_addr_lock);
773 list_for_each_entry_safe(addr, temp,
774 &net->sctp.local_addr_list, list) {
775 if (addr->a.sa.sa_family == AF_INET &&
776 addr->a.v4.sin_addr.s_addr ==
777 ifa->ifa_local) {
778 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
779 found = 1;
780 addr->valid = 0;
781 list_del_rcu(&addr->list);
782 break;
785 spin_unlock_bh(&net->sctp.local_addr_lock);
786 if (found)
787 kfree_rcu(addr, rcu);
788 break;
791 return NOTIFY_DONE;
795 * Initialize the control inode/socket with a control endpoint data
796 * structure. This endpoint is reserved exclusively for the OOTB processing.
798 static int sctp_ctl_sock_init(struct net *net)
800 int err;
801 sa_family_t family = PF_INET;
803 if (sctp_get_pf_specific(PF_INET6))
804 family = PF_INET6;
806 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
807 SOCK_SEQPACKET, IPPROTO_SCTP, net);
809 /* If IPv6 socket could not be created, try the IPv4 socket */
810 if (err < 0 && family == PF_INET6)
811 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
812 SOCK_SEQPACKET, IPPROTO_SCTP,
813 net);
815 if (err < 0) {
816 pr_err("Failed to create the SCTP control socket\n");
817 return err;
819 return 0;
822 /* Register address family specific functions. */
823 int sctp_register_af(struct sctp_af *af)
825 switch (af->sa_family) {
826 case AF_INET:
827 if (sctp_af_v4_specific)
828 return 0;
829 sctp_af_v4_specific = af;
830 break;
831 case AF_INET6:
832 if (sctp_af_v6_specific)
833 return 0;
834 sctp_af_v6_specific = af;
835 break;
836 default:
837 return 0;
840 INIT_LIST_HEAD(&af->list);
841 list_add_tail(&af->list, &sctp_address_families);
842 return 1;
845 /* Get the table of functions for manipulating a particular address
846 * family.
848 struct sctp_af *sctp_get_af_specific(sa_family_t family)
850 switch (family) {
851 case AF_INET:
852 return sctp_af_v4_specific;
853 case AF_INET6:
854 return sctp_af_v6_specific;
855 default:
856 return NULL;
860 /* Common code to initialize a AF_INET msg_name. */
861 static void sctp_inet_msgname(char *msgname, int *addr_len)
863 struct sockaddr_in *sin;
865 sin = (struct sockaddr_in *)msgname;
866 *addr_len = sizeof(struct sockaddr_in);
867 sin->sin_family = AF_INET;
868 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
871 /* Copy the primary address of the peer primary address as the msg_name. */
872 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
873 int *addr_len)
875 struct sockaddr_in *sin, *sinfrom;
877 if (msgname) {
878 struct sctp_association *asoc;
880 asoc = event->asoc;
881 sctp_inet_msgname(msgname, addr_len);
882 sin = (struct sockaddr_in *)msgname;
883 sinfrom = &asoc->peer.primary_addr.v4;
884 sin->sin_port = htons(asoc->peer.port);
885 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
889 /* Initialize and copy out a msgname from an inbound skb. */
890 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
892 if (msgname) {
893 struct sctphdr *sh = sctp_hdr(skb);
894 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
896 sctp_inet_msgname(msgname, len);
897 sin->sin_port = sh->source;
898 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
902 /* Do we support this AF? */
903 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
905 /* PF_INET only supports AF_INET addresses. */
906 return AF_INET == family;
909 /* Address matching with wildcards allowed. */
910 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
911 const union sctp_addr *addr2,
912 struct sctp_sock *opt)
914 /* PF_INET only supports AF_INET addresses. */
915 if (addr1->sa.sa_family != addr2->sa.sa_family)
916 return 0;
917 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
918 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
919 return 1;
920 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
921 return 1;
923 return 0;
926 /* Verify that provided sockaddr looks bindable. Common verification has
927 * already been taken care of.
929 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
931 return sctp_v4_available(addr, opt);
934 /* Verify that sockaddr looks sendable. Common verification has already
935 * been taken care of.
937 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
939 return 1;
942 /* Fill in Supported Address Type information for INIT and INIT-ACK
943 * chunks. Returns number of addresses supported.
945 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
946 __be16 *types)
948 types[0] = SCTP_PARAM_IPV4_ADDRESS;
949 return 1;
952 /* Wrapper routine that calls the ip transmit routine. */
953 static inline int sctp_v4_xmit(struct sk_buff *skb,
954 struct sctp_transport *transport)
956 struct inet_sock *inet = inet_sk(skb->sk);
958 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
959 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
961 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
962 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
964 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
966 return ip_queue_xmit(skb, &transport->fl);
969 static struct sctp_af sctp_af_inet;
971 static struct sctp_pf sctp_pf_inet = {
972 .event_msgname = sctp_inet_event_msgname,
973 .skb_msgname = sctp_inet_skb_msgname,
974 .af_supported = sctp_inet_af_supported,
975 .cmp_addr = sctp_inet_cmp_addr,
976 .bind_verify = sctp_inet_bind_verify,
977 .send_verify = sctp_inet_send_verify,
978 .supported_addrs = sctp_inet_supported_addrs,
979 .create_accept_sk = sctp_v4_create_accept_sk,
980 .addr_v4map = sctp_v4_addr_v4map,
981 .af = &sctp_af_inet
984 /* Notifier for inetaddr addition/deletion events. */
985 static struct notifier_block sctp_inetaddr_notifier = {
986 .notifier_call = sctp_inetaddr_event,
989 /* Socket operations. */
990 static const struct proto_ops inet_seqpacket_ops = {
991 .family = PF_INET,
992 .owner = THIS_MODULE,
993 .release = inet_release, /* Needs to be wrapped... */
994 .bind = inet_bind,
995 .connect = inet_dgram_connect,
996 .socketpair = sock_no_socketpair,
997 .accept = inet_accept,
998 .getname = inet_getname, /* Semantics are different. */
999 .poll = sctp_poll,
1000 .ioctl = inet_ioctl,
1001 .listen = sctp_inet_listen,
1002 .shutdown = inet_shutdown, /* Looks harmless. */
1003 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1004 .getsockopt = sock_common_getsockopt,
1005 .sendmsg = inet_sendmsg,
1006 .recvmsg = sock_common_recvmsg,
1007 .mmap = sock_no_mmap,
1008 .sendpage = sock_no_sendpage,
1009 #ifdef CONFIG_COMPAT
1010 .compat_setsockopt = compat_sock_common_setsockopt,
1011 .compat_getsockopt = compat_sock_common_getsockopt,
1012 #endif
1015 /* Registration with AF_INET family. */
1016 static struct inet_protosw sctp_seqpacket_protosw = {
1017 .type = SOCK_SEQPACKET,
1018 .protocol = IPPROTO_SCTP,
1019 .prot = &sctp_prot,
1020 .ops = &inet_seqpacket_ops,
1021 .no_check = 0,
1022 .flags = SCTP_PROTOSW_FLAG
1024 static struct inet_protosw sctp_stream_protosw = {
1025 .type = SOCK_STREAM,
1026 .protocol = IPPROTO_SCTP,
1027 .prot = &sctp_prot,
1028 .ops = &inet_seqpacket_ops,
1029 .no_check = 0,
1030 .flags = SCTP_PROTOSW_FLAG
1033 /* Register with IP layer. */
1034 static const struct net_protocol sctp_protocol = {
1035 .handler = sctp_rcv,
1036 .err_handler = sctp_v4_err,
1037 .no_policy = 1,
1038 .netns_ok = 1,
1041 /* IPv4 address related functions. */
1042 static struct sctp_af sctp_af_inet = {
1043 .sa_family = AF_INET,
1044 .sctp_xmit = sctp_v4_xmit,
1045 .setsockopt = ip_setsockopt,
1046 .getsockopt = ip_getsockopt,
1047 .get_dst = sctp_v4_get_dst,
1048 .get_saddr = sctp_v4_get_saddr,
1049 .copy_addrlist = sctp_v4_copy_addrlist,
1050 .from_skb = sctp_v4_from_skb,
1051 .from_sk = sctp_v4_from_sk,
1052 .to_sk_saddr = sctp_v4_to_sk_saddr,
1053 .to_sk_daddr = sctp_v4_to_sk_daddr,
1054 .from_addr_param = sctp_v4_from_addr_param,
1055 .to_addr_param = sctp_v4_to_addr_param,
1056 .cmp_addr = sctp_v4_cmp_addr,
1057 .addr_valid = sctp_v4_addr_valid,
1058 .inaddr_any = sctp_v4_inaddr_any,
1059 .is_any = sctp_v4_is_any,
1060 .available = sctp_v4_available,
1061 .scope = sctp_v4_scope,
1062 .skb_iif = sctp_v4_skb_iif,
1063 .is_ce = sctp_v4_is_ce,
1064 .seq_dump_addr = sctp_v4_seq_dump_addr,
1065 .ecn_capable = sctp_v4_ecn_capable,
1066 .net_header_len = sizeof(struct iphdr),
1067 .sockaddr_len = sizeof(struct sockaddr_in),
1068 #ifdef CONFIG_COMPAT
1069 .compat_setsockopt = compat_ip_setsockopt,
1070 .compat_getsockopt = compat_ip_getsockopt,
1071 #endif
1074 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
1076 switch (family) {
1077 case PF_INET:
1078 return sctp_pf_inet_specific;
1079 case PF_INET6:
1080 return sctp_pf_inet6_specific;
1081 default:
1082 return NULL;
1086 /* Register the PF specific function table. */
1087 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1089 switch (family) {
1090 case PF_INET:
1091 if (sctp_pf_inet_specific)
1092 return 0;
1093 sctp_pf_inet_specific = pf;
1094 break;
1095 case PF_INET6:
1096 if (sctp_pf_inet6_specific)
1097 return 0;
1098 sctp_pf_inet6_specific = pf;
1099 break;
1100 default:
1101 return 0;
1103 return 1;
1106 static inline int init_sctp_mibs(struct net *net)
1108 return snmp_mib_init((void __percpu **)net->sctp.sctp_statistics,
1109 sizeof(struct sctp_mib),
1110 __alignof__(struct sctp_mib));
1113 static inline void cleanup_sctp_mibs(struct net *net)
1115 snmp_mib_free((void __percpu **)net->sctp.sctp_statistics);
1118 static void sctp_v4_pf_init(void)
1120 /* Initialize the SCTP specific PF functions. */
1121 sctp_register_pf(&sctp_pf_inet, PF_INET);
1122 sctp_register_af(&sctp_af_inet);
1125 static void sctp_v4_pf_exit(void)
1127 list_del(&sctp_af_inet.list);
1130 static int sctp_v4_protosw_init(void)
1132 int rc;
1134 rc = proto_register(&sctp_prot, 1);
1135 if (rc)
1136 return rc;
1138 /* Register SCTP(UDP and TCP style) with socket layer. */
1139 inet_register_protosw(&sctp_seqpacket_protosw);
1140 inet_register_protosw(&sctp_stream_protosw);
1142 return 0;
1145 static void sctp_v4_protosw_exit(void)
1147 inet_unregister_protosw(&sctp_stream_protosw);
1148 inet_unregister_protosw(&sctp_seqpacket_protosw);
1149 proto_unregister(&sctp_prot);
1152 static int sctp_v4_add_protocol(void)
1154 /* Register notifier for inet address additions/deletions. */
1155 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1157 /* Register SCTP with inet layer. */
1158 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1159 return -EAGAIN;
1161 return 0;
1164 static void sctp_v4_del_protocol(void)
1166 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1167 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1170 static int __net_init sctp_net_init(struct net *net)
1172 int status;
1175 * 14. Suggested SCTP Protocol Parameter Values
1177 /* The following protocol parameters are RECOMMENDED: */
1178 /* RTO.Initial - 3 seconds */
1179 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1180 /* RTO.Min - 1 second */
1181 net->sctp.rto_min = SCTP_RTO_MIN;
1182 /* RTO.Max - 60 seconds */
1183 net->sctp.rto_max = SCTP_RTO_MAX;
1184 /* RTO.Alpha - 1/8 */
1185 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1186 /* RTO.Beta - 1/4 */
1187 net->sctp.rto_beta = SCTP_RTO_BETA;
1189 /* Valid.Cookie.Life - 60 seconds */
1190 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1192 /* Whether Cookie Preservative is enabled(1) or not(0) */
1193 net->sctp.cookie_preserve_enable = 1;
1195 /* Default sctp sockets to use md5 as their hmac alg */
1196 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1197 net->sctp.sctp_hmac_alg = "md5";
1198 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1199 net->sctp.sctp_hmac_alg = "sha1";
1200 #else
1201 net->sctp.sctp_hmac_alg = NULL;
1202 #endif
1204 /* Max.Burst - 4 */
1205 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1207 /* Association.Max.Retrans - 10 attempts
1208 * Path.Max.Retrans - 5 attempts (per destination address)
1209 * Max.Init.Retransmits - 8 attempts
1211 net->sctp.max_retrans_association = 10;
1212 net->sctp.max_retrans_path = 5;
1213 net->sctp.max_retrans_init = 8;
1215 /* Sendbuffer growth - do per-socket accounting */
1216 net->sctp.sndbuf_policy = 0;
1218 /* Rcvbuffer growth - do per-socket accounting */
1219 net->sctp.rcvbuf_policy = 0;
1221 /* HB.interval - 30 seconds */
1222 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1224 /* delayed SACK timeout */
1225 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1227 /* Disable ADDIP by default. */
1228 net->sctp.addip_enable = 0;
1229 net->sctp.addip_noauth = 0;
1230 net->sctp.default_auto_asconf = 0;
1232 /* Enable PR-SCTP by default. */
1233 net->sctp.prsctp_enable = 1;
1235 /* Disable AUTH by default. */
1236 net->sctp.auth_enable = 0;
1238 /* Set SCOPE policy to enabled */
1239 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1241 /* Set the default rwnd update threshold */
1242 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1244 /* Initialize maximum autoclose timeout. */
1245 net->sctp.max_autoclose = INT_MAX / HZ;
1247 status = sctp_sysctl_net_register(net);
1248 if (status)
1249 goto err_sysctl_register;
1251 /* Allocate and initialise sctp mibs. */
1252 status = init_sctp_mibs(net);
1253 if (status)
1254 goto err_init_mibs;
1256 /* Initialize proc fs directory. */
1257 status = sctp_proc_init(net);
1258 if (status)
1259 goto err_init_proc;
1261 sctp_dbg_objcnt_init(net);
1263 /* Initialize the control inode/socket for handling OOTB packets. */
1264 if ((status = sctp_ctl_sock_init(net))) {
1265 pr_err("Failed to initialize the SCTP control sock\n");
1266 goto err_ctl_sock_init;
1269 /* Initialize the local address list. */
1270 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1271 spin_lock_init(&net->sctp.local_addr_lock);
1272 sctp_get_local_addr_list(net);
1274 /* Initialize the address event list */
1275 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1276 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1277 spin_lock_init(&net->sctp.addr_wq_lock);
1278 net->sctp.addr_wq_timer.expires = 0;
1279 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1280 (unsigned long)net);
1282 return 0;
1284 err_ctl_sock_init:
1285 sctp_dbg_objcnt_exit(net);
1286 sctp_proc_exit(net);
1287 err_init_proc:
1288 cleanup_sctp_mibs(net);
1289 err_init_mibs:
1290 sctp_sysctl_net_unregister(net);
1291 err_sysctl_register:
1292 return status;
1295 static void __net_exit sctp_net_exit(struct net *net)
1297 /* Free the local address list */
1298 sctp_free_addr_wq(net);
1299 sctp_free_local_addr_list(net);
1301 /* Free the control endpoint. */
1302 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1304 sctp_dbg_objcnt_exit(net);
1306 sctp_proc_exit(net);
1307 cleanup_sctp_mibs(net);
1308 sctp_sysctl_net_unregister(net);
1311 static struct pernet_operations sctp_net_ops = {
1312 .init = sctp_net_init,
1313 .exit = sctp_net_exit,
1316 /* Initialize the universe into something sensible. */
1317 static __init int sctp_init(void)
1319 int i;
1320 int status = -EINVAL;
1321 unsigned long goal;
1322 unsigned long limit;
1323 int max_share;
1324 int order;
1326 BUILD_BUG_ON(sizeof(struct sctp_ulpevent) >
1327 sizeof(((struct sk_buff *) 0)->cb));
1329 /* Allocate bind_bucket and chunk caches. */
1330 status = -ENOBUFS;
1331 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1332 sizeof(struct sctp_bind_bucket),
1333 0, SLAB_HWCACHE_ALIGN,
1334 NULL);
1335 if (!sctp_bucket_cachep)
1336 goto out;
1338 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1339 sizeof(struct sctp_chunk),
1340 0, SLAB_HWCACHE_ALIGN,
1341 NULL);
1342 if (!sctp_chunk_cachep)
1343 goto err_chunk_cachep;
1345 status = percpu_counter_init(&sctp_sockets_allocated, 0);
1346 if (status)
1347 goto err_percpu_counter_init;
1349 /* Implementation specific variables. */
1351 /* Initialize default stream count setup information. */
1352 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1353 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1355 /* Initialize handle used for association ids. */
1356 idr_init(&sctp_assocs_id);
1358 limit = nr_free_buffer_pages() / 8;
1359 limit = max(limit, 128UL);
1360 sysctl_sctp_mem[0] = limit / 4 * 3;
1361 sysctl_sctp_mem[1] = limit;
1362 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1364 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1365 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1366 max_share = min(4UL*1024*1024, limit);
1368 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1369 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1370 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1372 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1373 sysctl_sctp_wmem[1] = 16*1024;
1374 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1376 /* Size and allocate the association hash table.
1377 * The methodology is similar to that of the tcp hash tables.
1379 if (totalram_pages >= (128 * 1024))
1380 goal = totalram_pages >> (22 - PAGE_SHIFT);
1381 else
1382 goal = totalram_pages >> (24 - PAGE_SHIFT);
1384 for (order = 0; (1UL << order) < goal; order++)
1387 do {
1388 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1389 sizeof(struct sctp_hashbucket);
1390 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1391 continue;
1392 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1393 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1394 } while (!sctp_assoc_hashtable && --order > 0);
1395 if (!sctp_assoc_hashtable) {
1396 pr_err("Failed association hash alloc\n");
1397 status = -ENOMEM;
1398 goto err_ahash_alloc;
1400 for (i = 0; i < sctp_assoc_hashsize; i++) {
1401 rwlock_init(&sctp_assoc_hashtable[i].lock);
1402 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1405 /* Allocate and initialize the endpoint hash table. */
1406 sctp_ep_hashsize = 64;
1407 sctp_ep_hashtable =
1408 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1409 if (!sctp_ep_hashtable) {
1410 pr_err("Failed endpoint_hash alloc\n");
1411 status = -ENOMEM;
1412 goto err_ehash_alloc;
1414 for (i = 0; i < sctp_ep_hashsize; i++) {
1415 rwlock_init(&sctp_ep_hashtable[i].lock);
1416 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1419 /* Allocate and initialize the SCTP port hash table. */
1420 do {
1421 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1422 sizeof(struct sctp_bind_hashbucket);
1423 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1424 continue;
1425 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1426 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1427 } while (!sctp_port_hashtable && --order > 0);
1428 if (!sctp_port_hashtable) {
1429 pr_err("Failed bind hash alloc\n");
1430 status = -ENOMEM;
1431 goto err_bhash_alloc;
1433 for (i = 0; i < sctp_port_hashsize; i++) {
1434 spin_lock_init(&sctp_port_hashtable[i].lock);
1435 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1438 pr_info("Hash tables configured (established %d bind %d)\n",
1439 sctp_assoc_hashsize, sctp_port_hashsize);
1441 sctp_sysctl_register();
1443 INIT_LIST_HEAD(&sctp_address_families);
1444 sctp_v4_pf_init();
1445 sctp_v6_pf_init();
1447 status = sctp_v4_protosw_init();
1449 if (status)
1450 goto err_protosw_init;
1452 status = sctp_v6_protosw_init();
1453 if (status)
1454 goto err_v6_protosw_init;
1456 status = register_pernet_subsys(&sctp_net_ops);
1457 if (status)
1458 goto err_register_pernet_subsys;
1460 status = sctp_v4_add_protocol();
1461 if (status)
1462 goto err_add_protocol;
1464 /* Register SCTP with inet6 layer. */
1465 status = sctp_v6_add_protocol();
1466 if (status)
1467 goto err_v6_add_protocol;
1469 status = 0;
1470 out:
1471 return status;
1472 err_v6_add_protocol:
1473 sctp_v4_del_protocol();
1474 err_add_protocol:
1475 unregister_pernet_subsys(&sctp_net_ops);
1476 err_register_pernet_subsys:
1477 sctp_v6_protosw_exit();
1478 err_v6_protosw_init:
1479 sctp_v4_protosw_exit();
1480 err_protosw_init:
1481 sctp_v4_pf_exit();
1482 sctp_v6_pf_exit();
1483 sctp_sysctl_unregister();
1484 free_pages((unsigned long)sctp_port_hashtable,
1485 get_order(sctp_port_hashsize *
1486 sizeof(struct sctp_bind_hashbucket)));
1487 err_bhash_alloc:
1488 kfree(sctp_ep_hashtable);
1489 err_ehash_alloc:
1490 free_pages((unsigned long)sctp_assoc_hashtable,
1491 get_order(sctp_assoc_hashsize *
1492 sizeof(struct sctp_hashbucket)));
1493 err_ahash_alloc:
1494 percpu_counter_destroy(&sctp_sockets_allocated);
1495 err_percpu_counter_init:
1496 kmem_cache_destroy(sctp_chunk_cachep);
1497 err_chunk_cachep:
1498 kmem_cache_destroy(sctp_bucket_cachep);
1499 goto out;
1502 /* Exit handler for the SCTP protocol. */
1503 static __exit void sctp_exit(void)
1505 /* BUG. This should probably do something useful like clean
1506 * up all the remaining associations and all that memory.
1509 /* Unregister with inet6/inet layers. */
1510 sctp_v6_del_protocol();
1511 sctp_v4_del_protocol();
1513 unregister_pernet_subsys(&sctp_net_ops);
1515 /* Free protosw registrations */
1516 sctp_v6_protosw_exit();
1517 sctp_v4_protosw_exit();
1519 /* Unregister with socket layer. */
1520 sctp_v6_pf_exit();
1521 sctp_v4_pf_exit();
1523 sctp_sysctl_unregister();
1525 free_pages((unsigned long)sctp_assoc_hashtable,
1526 get_order(sctp_assoc_hashsize *
1527 sizeof(struct sctp_hashbucket)));
1528 kfree(sctp_ep_hashtable);
1529 free_pages((unsigned long)sctp_port_hashtable,
1530 get_order(sctp_port_hashsize *
1531 sizeof(struct sctp_bind_hashbucket)));
1533 percpu_counter_destroy(&sctp_sockets_allocated);
1535 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1537 kmem_cache_destroy(sctp_chunk_cachep);
1538 kmem_cache_destroy(sctp_bucket_cachep);
1541 module_init(sctp_init);
1542 module_exit(sctp_exit);
1545 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1547 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1548 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1549 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1550 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1551 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1552 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1553 MODULE_LICENSE("GPL");