Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
[linux-2.6.git] / net / sctp / protocol.c
blob4a17494d736c4b6a50f0e838971204e1e901de43
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 <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>
62 #include <net/ip.h>
63 #include <net/ipv6.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;
73 struct idr sctp_assocs_id;
74 DEFINE_SPINLOCK(sctp_assocs_id_lock);
76 static struct sctp_pf *sctp_pf_inet6_specific;
77 static struct sctp_pf *sctp_pf_inet_specific;
78 static struct sctp_af *sctp_af_v4_specific;
79 static struct sctp_af *sctp_af_v6_specific;
81 struct kmem_cache *sctp_chunk_cachep __read_mostly;
82 struct kmem_cache *sctp_bucket_cachep __read_mostly;
84 long sysctl_sctp_mem[3];
85 int sysctl_sctp_rmem[3];
86 int sysctl_sctp_wmem[3];
88 /* Set up the proc fs entry for the SCTP protocol. */
89 static int __net_init sctp_proc_init(struct net *net)
91 #ifdef CONFIG_PROC_FS
92 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
93 if (!net->sctp.proc_net_sctp)
94 goto out_proc_net_sctp;
95 if (sctp_snmp_proc_init(net))
96 goto out_snmp_proc_init;
97 if (sctp_eps_proc_init(net))
98 goto out_eps_proc_init;
99 if (sctp_assocs_proc_init(net))
100 goto out_assocs_proc_init;
101 if (sctp_remaddr_proc_init(net))
102 goto out_remaddr_proc_init;
104 return 0;
106 out_remaddr_proc_init:
107 sctp_assocs_proc_exit(net);
108 out_assocs_proc_init:
109 sctp_eps_proc_exit(net);
110 out_eps_proc_init:
111 sctp_snmp_proc_exit(net);
112 out_snmp_proc_init:
113 remove_proc_entry("sctp", net->proc_net);
114 net->sctp.proc_net_sctp = NULL;
115 out_proc_net_sctp:
116 return -ENOMEM;
117 #endif /* CONFIG_PROC_FS */
118 return 0;
121 /* Clean up the proc fs entry for the SCTP protocol.
122 * Note: Do not make this __exit as it is used in the init error
123 * path.
125 static void sctp_proc_exit(struct net *net)
127 #ifdef CONFIG_PROC_FS
128 sctp_snmp_proc_exit(net);
129 sctp_eps_proc_exit(net);
130 sctp_assocs_proc_exit(net);
131 sctp_remaddr_proc_exit(net);
133 remove_proc_entry("sctp", net->proc_net);
134 net->sctp.proc_net_sctp = NULL;
135 #endif
138 /* Private helper to extract ipv4 address and stash them in
139 * the protocol structure.
141 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
142 struct net_device *dev)
144 struct in_device *in_dev;
145 struct in_ifaddr *ifa;
146 struct sctp_sockaddr_entry *addr;
148 rcu_read_lock();
149 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
150 rcu_read_unlock();
151 return;
154 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
155 /* Add the address to the local list. */
156 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
157 if (addr) {
158 addr->a.v4.sin_family = AF_INET;
159 addr->a.v4.sin_port = 0;
160 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
161 addr->valid = 1;
162 INIT_LIST_HEAD(&addr->list);
163 list_add_tail(&addr->list, addrlist);
167 rcu_read_unlock();
170 /* Extract our IP addresses from the system and stash them in the
171 * protocol structure.
173 static void sctp_get_local_addr_list(struct net *net)
175 struct net_device *dev;
176 struct list_head *pos;
177 struct sctp_af *af;
179 rcu_read_lock();
180 for_each_netdev_rcu(net, dev) {
181 list_for_each(pos, &sctp_address_families) {
182 af = list_entry(pos, struct sctp_af, list);
183 af->copy_addrlist(&net->sctp.local_addr_list, dev);
186 rcu_read_unlock();
189 /* Free the existing local addresses. */
190 static void sctp_free_local_addr_list(struct net *net)
192 struct sctp_sockaddr_entry *addr;
193 struct list_head *pos, *temp;
195 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
196 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
197 list_del(pos);
198 kfree(addr);
202 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
203 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
204 sctp_scope_t scope, gfp_t gfp, int copy_flags)
206 struct sctp_sockaddr_entry *addr;
207 int error = 0;
209 rcu_read_lock();
210 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
211 if (!addr->valid)
212 continue;
213 if (sctp_in_scope(net, &addr->a, scope)) {
214 /* Now that the address is in scope, check to see if
215 * the address type is really supported by the local
216 * sock as well as the remote peer.
218 if ((((AF_INET == addr->a.sa.sa_family) &&
219 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
220 (((AF_INET6 == addr->a.sa.sa_family) &&
221 (copy_flags & SCTP_ADDR6_ALLOWED) &&
222 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
223 error = sctp_add_bind_addr(bp, &addr->a,
224 SCTP_ADDR_SRC, GFP_ATOMIC);
225 if (error)
226 goto end_copy;
231 end_copy:
232 rcu_read_unlock();
233 return error;
236 /* Initialize a sctp_addr from in incoming skb. */
237 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
238 int is_saddr)
240 void *from;
241 __be16 *port;
242 struct sctphdr *sh;
244 port = &addr->v4.sin_port;
245 addr->v4.sin_family = AF_INET;
247 sh = sctp_hdr(skb);
248 if (is_saddr) {
249 *port = sh->source;
250 from = &ip_hdr(skb)->saddr;
251 } else {
252 *port = sh->dest;
253 from = &ip_hdr(skb)->daddr;
255 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
258 /* Initialize an sctp_addr from a socket. */
259 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
261 addr->v4.sin_family = AF_INET;
262 addr->v4.sin_port = 0;
263 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
266 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
267 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
269 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
272 /* Initialize sk->sk_daddr from sctp_addr. */
273 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
275 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
278 /* Initialize a sctp_addr from an address parameter. */
279 static void sctp_v4_from_addr_param(union sctp_addr *addr,
280 union sctp_addr_param *param,
281 __be16 port, int iif)
283 addr->v4.sin_family = AF_INET;
284 addr->v4.sin_port = port;
285 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
288 /* Initialize an address parameter from a sctp_addr and return the length
289 * of the address parameter.
291 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
292 union sctp_addr_param *param)
294 int length = sizeof(sctp_ipv4addr_param_t);
296 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
297 param->v4.param_hdr.length = htons(length);
298 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
300 return length;
303 /* Initialize a sctp_addr from a dst_entry. */
304 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
305 __be16 port)
307 saddr->v4.sin_family = AF_INET;
308 saddr->v4.sin_port = port;
309 saddr->v4.sin_addr.s_addr = fl4->saddr;
312 /* Compare two addresses exactly. */
313 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
314 const union sctp_addr *addr2)
316 if (addr1->sa.sa_family != addr2->sa.sa_family)
317 return 0;
318 if (addr1->v4.sin_port != addr2->v4.sin_port)
319 return 0;
320 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
321 return 0;
323 return 1;
326 /* Initialize addr struct to INADDR_ANY. */
327 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
329 addr->v4.sin_family = AF_INET;
330 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
331 addr->v4.sin_port = port;
334 /* Is this a wildcard address? */
335 static int sctp_v4_is_any(const union sctp_addr *addr)
337 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
340 /* This function checks if the address is a valid address to be used for
341 * SCTP binding.
343 * Output:
344 * Return 0 - If the address is a non-unicast or an illegal address.
345 * Return 1 - If the address is a unicast.
347 static int sctp_v4_addr_valid(union sctp_addr *addr,
348 struct sctp_sock *sp,
349 const struct sk_buff *skb)
351 /* IPv4 addresses not allowed */
352 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
353 return 0;
355 /* Is this a non-unicast address or a unusable SCTP address? */
356 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
357 return 0;
359 /* Is this a broadcast address? */
360 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
361 return 0;
363 return 1;
366 /* Should this be available for binding? */
367 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
369 struct net *net = sock_net(&sp->inet.sk);
370 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
373 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
374 ret != RTN_LOCAL &&
375 !sp->inet.freebind &&
376 !sysctl_ip_nonlocal_bind)
377 return 0;
379 if (ipv6_only_sock(sctp_opt2sk(sp)))
380 return 0;
382 return 1;
385 /* Checking the loopback, private and other address scopes as defined in
386 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
387 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
389 * Level 0 - unusable SCTP addresses
390 * Level 1 - loopback address
391 * Level 2 - link-local addresses
392 * Level 3 - private addresses.
393 * Level 4 - global addresses
394 * For INIT and INIT-ACK address list, let L be the level of
395 * of requested destination address, sender and receiver
396 * SHOULD include all of its addresses with level greater
397 * than or equal to L.
399 * IPv4 scoping can be controlled through sysctl option
400 * net.sctp.addr_scope_policy
402 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
404 sctp_scope_t retval;
406 /* Check for unusable SCTP addresses. */
407 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
408 retval = SCTP_SCOPE_UNUSABLE;
409 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
410 retval = SCTP_SCOPE_LOOPBACK;
411 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
412 retval = SCTP_SCOPE_LINK;
413 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
414 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
415 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
416 retval = SCTP_SCOPE_PRIVATE;
417 } else {
418 retval = SCTP_SCOPE_GLOBAL;
421 return retval;
424 /* Returns a valid dst cache entry for the given source and destination ip
425 * addresses. If an association is passed, trys to get a dst entry with a
426 * source address that matches an address in the bind address list.
428 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
429 struct flowi *fl, struct sock *sk)
431 struct sctp_association *asoc = t->asoc;
432 struct rtable *rt;
433 struct flowi4 *fl4 = &fl->u.ip4;
434 struct sctp_bind_addr *bp;
435 struct sctp_sockaddr_entry *laddr;
436 struct dst_entry *dst = NULL;
437 union sctp_addr *daddr = &t->ipaddr;
438 union sctp_addr dst_saddr;
440 memset(fl4, 0x0, sizeof(struct flowi4));
441 fl4->daddr = daddr->v4.sin_addr.s_addr;
442 fl4->fl4_dport = daddr->v4.sin_port;
443 fl4->flowi4_proto = IPPROTO_SCTP;
444 if (asoc) {
445 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
446 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
447 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
449 if (saddr) {
450 fl4->saddr = saddr->v4.sin_addr.s_addr;
451 fl4->fl4_sport = saddr->v4.sin_port;
454 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
455 &fl4->saddr);
457 rt = ip_route_output_key(sock_net(sk), fl4);
458 if (!IS_ERR(rt))
459 dst = &rt->dst;
461 /* If there is no association or if a source address is passed, no
462 * more validation is required.
464 if (!asoc || saddr)
465 goto out;
467 bp = &asoc->base.bind_addr;
469 if (dst) {
470 /* Walk through the bind address list and look for a bind
471 * address that matches the source address of the returned dst.
473 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
474 rcu_read_lock();
475 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
476 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
477 (laddr->state != SCTP_ADDR_SRC &&
478 !asoc->src_out_of_asoc_ok))
479 continue;
480 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
481 goto out_unlock;
483 rcu_read_unlock();
485 /* None of the bound addresses match the source address of the
486 * dst. So release it.
488 dst_release(dst);
489 dst = NULL;
492 /* Walk through the bind address list and try to get a dst that
493 * matches a bind address as the source address.
495 rcu_read_lock();
496 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
497 if (!laddr->valid)
498 continue;
499 if ((laddr->state == SCTP_ADDR_SRC) &&
500 (AF_INET == laddr->a.sa.sa_family)) {
501 fl4->saddr = laddr->a.v4.sin_addr.s_addr;
502 fl4->fl4_sport = laddr->a.v4.sin_port;
503 rt = ip_route_output_key(sock_net(sk), fl4);
504 if (!IS_ERR(rt)) {
505 dst = &rt->dst;
506 goto out_unlock;
511 out_unlock:
512 rcu_read_unlock();
513 out:
514 t->dst = dst;
515 if (dst)
516 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
517 &fl4->daddr, &fl4->saddr);
518 else
519 pr_debug("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_transport *t,
527 struct flowi *fl)
529 union sctp_addr *saddr = &t->saddr;
530 struct rtable *rt = (struct rtable *)t->dst;
532 if (rt) {
533 saddr->v4.sin_family = AF_INET;
534 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
538 /* What interface did this skb arrive on? */
539 static int sctp_v4_skb_iif(const struct sk_buff *skb)
541 return inet_iif(skb);
544 /* Was this packet marked by Explicit Congestion Notification? */
545 static int sctp_v4_is_ce(const struct sk_buff *skb)
547 return INET_ECN_is_ce(ip_hdr(skb)->tos);
550 /* Create and initialize a new sk for the socket returned by accept(). */
551 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
552 struct sctp_association *asoc)
554 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
555 sk->sk_prot);
556 struct inet_sock *newinet;
558 if (!newsk)
559 goto out;
561 sock_init_data(NULL, newsk);
563 sctp_copy_sock(newsk, sk, asoc);
564 sock_reset_flag(newsk, SOCK_ZAPPED);
566 newinet = inet_sk(newsk);
568 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
570 sk_refcnt_debug_inc(newsk);
572 if (newsk->sk_prot->init(newsk)) {
573 sk_common_release(newsk);
574 newsk = NULL;
577 out:
578 return newsk;
581 /* Map address, empty for v4 family */
582 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
584 /* Empty */
587 /* Dump the v4 addr to the seq file. */
588 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
590 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
593 static void sctp_v4_ecn_capable(struct sock *sk)
595 INET_ECN_xmit(sk);
598 static void sctp_addr_wq_timeout_handler(unsigned long arg)
600 struct net *net = (struct net *)arg;
601 struct sctp_sockaddr_entry *addrw, *temp;
602 struct sctp_sock *sp;
604 spin_lock_bh(&net->sctp.addr_wq_lock);
606 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
607 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
608 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
609 addrw->state, addrw);
611 #if IS_ENABLED(CONFIG_IPV6)
612 /* Now we send an ASCONF for each association */
613 /* Note. we currently don't handle link local IPv6 addressees */
614 if (addrw->a.sa.sa_family == AF_INET6) {
615 struct in6_addr *in6;
617 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
618 IPV6_ADDR_LINKLOCAL)
619 goto free_next;
621 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
622 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
623 addrw->state == SCTP_ADDR_NEW) {
624 unsigned long timeo_val;
626 pr_debug("%s: this is on DAD, trying %d sec "
627 "later\n", __func__,
628 SCTP_ADDRESS_TICK_DELAY);
630 timeo_val = jiffies;
631 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
632 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
633 break;
636 #endif
637 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
638 struct sock *sk;
640 sk = sctp_opt2sk(sp);
641 /* ignore bound-specific endpoints */
642 if (!sctp_is_ep_boundall(sk))
643 continue;
644 sctp_bh_lock_sock(sk);
645 if (sctp_asconf_mgmt(sp, addrw) < 0)
646 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
647 sctp_bh_unlock_sock(sk);
649 #if IS_ENABLED(CONFIG_IPV6)
650 free_next:
651 #endif
652 list_del(&addrw->list);
653 kfree(addrw);
655 spin_unlock_bh(&net->sctp.addr_wq_lock);
658 static void sctp_free_addr_wq(struct net *net)
660 struct sctp_sockaddr_entry *addrw;
661 struct sctp_sockaddr_entry *temp;
663 spin_lock_bh(&net->sctp.addr_wq_lock);
664 del_timer(&net->sctp.addr_wq_timer);
665 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
666 list_del(&addrw->list);
667 kfree(addrw);
669 spin_unlock_bh(&net->sctp.addr_wq_lock);
672 /* lookup the entry for the same address in the addr_waitq
673 * sctp_addr_wq MUST be locked
675 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
676 struct sctp_sockaddr_entry *addr)
678 struct sctp_sockaddr_entry *addrw;
680 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
681 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
682 continue;
683 if (addrw->a.sa.sa_family == AF_INET) {
684 if (addrw->a.v4.sin_addr.s_addr ==
685 addr->a.v4.sin_addr.s_addr)
686 return addrw;
687 } else if (addrw->a.sa.sa_family == AF_INET6) {
688 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
689 &addr->a.v6.sin6_addr))
690 return addrw;
693 return NULL;
696 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
698 struct sctp_sockaddr_entry *addrw;
699 unsigned long timeo_val;
701 /* first, we check if an opposite message already exist in the queue.
702 * If we found such message, it is removed.
703 * This operation is a bit stupid, but the DHCP client attaches the
704 * new address after a couple of addition and deletion of that address
707 spin_lock_bh(&net->sctp.addr_wq_lock);
708 /* Offsets existing events in addr_wq */
709 addrw = sctp_addr_wq_lookup(net, addr);
710 if (addrw) {
711 if (addrw->state != cmd) {
712 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
713 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
714 &net->sctp.addr_waitq);
716 list_del(&addrw->list);
717 kfree(addrw);
719 spin_unlock_bh(&net->sctp.addr_wq_lock);
720 return;
723 /* OK, we have to add the new address to the wait queue */
724 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
725 if (addrw == NULL) {
726 spin_unlock_bh(&net->sctp.addr_wq_lock);
727 return;
729 addrw->state = cmd;
730 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
732 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
733 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
735 if (!timer_pending(&net->sctp.addr_wq_timer)) {
736 timeo_val = jiffies;
737 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
738 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
740 spin_unlock_bh(&net->sctp.addr_wq_lock);
743 /* Event handler for inet address addition/deletion events.
744 * The sctp_local_addr_list needs to be protocted by a spin lock since
745 * multiple notifiers (say IPv4 and IPv6) may be running at the same
746 * time and thus corrupt the list.
747 * The reader side is protected with RCU.
749 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
750 void *ptr)
752 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
753 struct sctp_sockaddr_entry *addr = NULL;
754 struct sctp_sockaddr_entry *temp;
755 struct net *net = dev_net(ifa->ifa_dev->dev);
756 int found = 0;
758 switch (ev) {
759 case NETDEV_UP:
760 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
761 if (addr) {
762 addr->a.v4.sin_family = AF_INET;
763 addr->a.v4.sin_port = 0;
764 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
765 addr->valid = 1;
766 spin_lock_bh(&net->sctp.local_addr_lock);
767 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
768 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
769 spin_unlock_bh(&net->sctp.local_addr_lock);
771 break;
772 case NETDEV_DOWN:
773 spin_lock_bh(&net->sctp.local_addr_lock);
774 list_for_each_entry_safe(addr, temp,
775 &net->sctp.local_addr_list, list) {
776 if (addr->a.sa.sa_family == AF_INET &&
777 addr->a.v4.sin_addr.s_addr ==
778 ifa->ifa_local) {
779 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
780 found = 1;
781 addr->valid = 0;
782 list_del_rcu(&addr->list);
783 break;
786 spin_unlock_bh(&net->sctp.local_addr_lock);
787 if (found)
788 kfree_rcu(addr, rcu);
789 break;
792 return NOTIFY_DONE;
796 * Initialize the control inode/socket with a control endpoint data
797 * structure. This endpoint is reserved exclusively for the OOTB processing.
799 static int sctp_ctl_sock_init(struct net *net)
801 int err;
802 sa_family_t family = PF_INET;
804 if (sctp_get_pf_specific(PF_INET6))
805 family = PF_INET6;
807 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
808 SOCK_SEQPACKET, IPPROTO_SCTP, net);
810 /* If IPv6 socket could not be created, try the IPv4 socket */
811 if (err < 0 && family == PF_INET6)
812 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
813 SOCK_SEQPACKET, IPPROTO_SCTP,
814 net);
816 if (err < 0) {
817 pr_err("Failed to create the SCTP control socket\n");
818 return err;
820 return 0;
823 /* Register address family specific functions. */
824 int sctp_register_af(struct sctp_af *af)
826 switch (af->sa_family) {
827 case AF_INET:
828 if (sctp_af_v4_specific)
829 return 0;
830 sctp_af_v4_specific = af;
831 break;
832 case AF_INET6:
833 if (sctp_af_v6_specific)
834 return 0;
835 sctp_af_v6_specific = af;
836 break;
837 default:
838 return 0;
841 INIT_LIST_HEAD(&af->list);
842 list_add_tail(&af->list, &sctp_address_families);
843 return 1;
846 /* Get the table of functions for manipulating a particular address
847 * family.
849 struct sctp_af *sctp_get_af_specific(sa_family_t family)
851 switch (family) {
852 case AF_INET:
853 return sctp_af_v4_specific;
854 case AF_INET6:
855 return sctp_af_v6_specific;
856 default:
857 return NULL;
861 /* Common code to initialize a AF_INET msg_name. */
862 static void sctp_inet_msgname(char *msgname, int *addr_len)
864 struct sockaddr_in *sin;
866 sin = (struct sockaddr_in *)msgname;
867 *addr_len = sizeof(struct sockaddr_in);
868 sin->sin_family = AF_INET;
869 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
872 /* Copy the primary address of the peer primary address as the msg_name. */
873 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
874 int *addr_len)
876 struct sockaddr_in *sin, *sinfrom;
878 if (msgname) {
879 struct sctp_association *asoc;
881 asoc = event->asoc;
882 sctp_inet_msgname(msgname, addr_len);
883 sin = (struct sockaddr_in *)msgname;
884 sinfrom = &asoc->peer.primary_addr.v4;
885 sin->sin_port = htons(asoc->peer.port);
886 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
890 /* Initialize and copy out a msgname from an inbound skb. */
891 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
893 if (msgname) {
894 struct sctphdr *sh = sctp_hdr(skb);
895 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
897 sctp_inet_msgname(msgname, len);
898 sin->sin_port = sh->source;
899 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
903 /* Do we support this AF? */
904 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
906 /* PF_INET only supports AF_INET addresses. */
907 return AF_INET == family;
910 /* Address matching with wildcards allowed. */
911 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
912 const union sctp_addr *addr2,
913 struct sctp_sock *opt)
915 /* PF_INET only supports AF_INET addresses. */
916 if (addr1->sa.sa_family != addr2->sa.sa_family)
917 return 0;
918 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
919 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
920 return 1;
921 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
922 return 1;
924 return 0;
927 /* Verify that provided sockaddr looks bindable. Common verification has
928 * already been taken care of.
930 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
932 return sctp_v4_available(addr, opt);
935 /* Verify that sockaddr looks sendable. Common verification has already
936 * been taken care of.
938 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
940 return 1;
943 /* Fill in Supported Address Type information for INIT and INIT-ACK
944 * chunks. Returns number of addresses supported.
946 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
947 __be16 *types)
949 types[0] = SCTP_PARAM_IPV4_ADDRESS;
950 return 1;
953 /* Wrapper routine that calls the ip transmit routine. */
954 static inline int sctp_v4_xmit(struct sk_buff *skb,
955 struct sctp_transport *transport)
957 struct inet_sock *inet = inet_sk(skb->sk);
959 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
960 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
962 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
963 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
965 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
967 return ip_queue_xmit(skb, &transport->fl);
970 static struct sctp_af sctp_af_inet;
972 static struct sctp_pf sctp_pf_inet = {
973 .event_msgname = sctp_inet_event_msgname,
974 .skb_msgname = sctp_inet_skb_msgname,
975 .af_supported = sctp_inet_af_supported,
976 .cmp_addr = sctp_inet_cmp_addr,
977 .bind_verify = sctp_inet_bind_verify,
978 .send_verify = sctp_inet_send_verify,
979 .supported_addrs = sctp_inet_supported_addrs,
980 .create_accept_sk = sctp_v4_create_accept_sk,
981 .addr_v4map = sctp_v4_addr_v4map,
982 .af = &sctp_af_inet
985 /* Notifier for inetaddr addition/deletion events. */
986 static struct notifier_block sctp_inetaddr_notifier = {
987 .notifier_call = sctp_inetaddr_event,
990 /* Socket operations. */
991 static const struct proto_ops inet_seqpacket_ops = {
992 .family = PF_INET,
993 .owner = THIS_MODULE,
994 .release = inet_release, /* Needs to be wrapped... */
995 .bind = inet_bind,
996 .connect = inet_dgram_connect,
997 .socketpair = sock_no_socketpair,
998 .accept = inet_accept,
999 .getname = inet_getname, /* Semantics are different. */
1000 .poll = sctp_poll,
1001 .ioctl = inet_ioctl,
1002 .listen = sctp_inet_listen,
1003 .shutdown = inet_shutdown, /* Looks harmless. */
1004 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1005 .getsockopt = sock_common_getsockopt,
1006 .sendmsg = inet_sendmsg,
1007 .recvmsg = sock_common_recvmsg,
1008 .mmap = sock_no_mmap,
1009 .sendpage = sock_no_sendpage,
1010 #ifdef CONFIG_COMPAT
1011 .compat_setsockopt = compat_sock_common_setsockopt,
1012 .compat_getsockopt = compat_sock_common_getsockopt,
1013 #endif
1016 /* Registration with AF_INET family. */
1017 static struct inet_protosw sctp_seqpacket_protosw = {
1018 .type = SOCK_SEQPACKET,
1019 .protocol = IPPROTO_SCTP,
1020 .prot = &sctp_prot,
1021 .ops = &inet_seqpacket_ops,
1022 .no_check = 0,
1023 .flags = SCTP_PROTOSW_FLAG
1025 static struct inet_protosw sctp_stream_protosw = {
1026 .type = SOCK_STREAM,
1027 .protocol = IPPROTO_SCTP,
1028 .prot = &sctp_prot,
1029 .ops = &inet_seqpacket_ops,
1030 .no_check = 0,
1031 .flags = SCTP_PROTOSW_FLAG
1034 /* Register with IP layer. */
1035 static const struct net_protocol sctp_protocol = {
1036 .handler = sctp_rcv,
1037 .err_handler = sctp_v4_err,
1038 .no_policy = 1,
1039 .netns_ok = 1,
1042 /* IPv4 address related functions. */
1043 static struct sctp_af sctp_af_inet = {
1044 .sa_family = AF_INET,
1045 .sctp_xmit = sctp_v4_xmit,
1046 .setsockopt = ip_setsockopt,
1047 .getsockopt = ip_getsockopt,
1048 .get_dst = sctp_v4_get_dst,
1049 .get_saddr = sctp_v4_get_saddr,
1050 .copy_addrlist = sctp_v4_copy_addrlist,
1051 .from_skb = sctp_v4_from_skb,
1052 .from_sk = sctp_v4_from_sk,
1053 .to_sk_saddr = sctp_v4_to_sk_saddr,
1054 .to_sk_daddr = sctp_v4_to_sk_daddr,
1055 .from_addr_param = sctp_v4_from_addr_param,
1056 .to_addr_param = sctp_v4_to_addr_param,
1057 .cmp_addr = sctp_v4_cmp_addr,
1058 .addr_valid = sctp_v4_addr_valid,
1059 .inaddr_any = sctp_v4_inaddr_any,
1060 .is_any = sctp_v4_is_any,
1061 .available = sctp_v4_available,
1062 .scope = sctp_v4_scope,
1063 .skb_iif = sctp_v4_skb_iif,
1064 .is_ce = sctp_v4_is_ce,
1065 .seq_dump_addr = sctp_v4_seq_dump_addr,
1066 .ecn_capable = sctp_v4_ecn_capable,
1067 .net_header_len = sizeof(struct iphdr),
1068 .sockaddr_len = sizeof(struct sockaddr_in),
1069 #ifdef CONFIG_COMPAT
1070 .compat_setsockopt = compat_ip_setsockopt,
1071 .compat_getsockopt = compat_ip_getsockopt,
1072 #endif
1075 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
1077 switch (family) {
1078 case PF_INET:
1079 return sctp_pf_inet_specific;
1080 case PF_INET6:
1081 return sctp_pf_inet6_specific;
1082 default:
1083 return NULL;
1087 /* Register the PF specific function table. */
1088 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1090 switch (family) {
1091 case PF_INET:
1092 if (sctp_pf_inet_specific)
1093 return 0;
1094 sctp_pf_inet_specific = pf;
1095 break;
1096 case PF_INET6:
1097 if (sctp_pf_inet6_specific)
1098 return 0;
1099 sctp_pf_inet6_specific = pf;
1100 break;
1101 default:
1102 return 0;
1104 return 1;
1107 static inline int init_sctp_mibs(struct net *net)
1109 return snmp_mib_init((void __percpu **)net->sctp.sctp_statistics,
1110 sizeof(struct sctp_mib),
1111 __alignof__(struct sctp_mib));
1114 static inline void cleanup_sctp_mibs(struct net *net)
1116 snmp_mib_free((void __percpu **)net->sctp.sctp_statistics);
1119 static void sctp_v4_pf_init(void)
1121 /* Initialize the SCTP specific PF functions. */
1122 sctp_register_pf(&sctp_pf_inet, PF_INET);
1123 sctp_register_af(&sctp_af_inet);
1126 static void sctp_v4_pf_exit(void)
1128 list_del(&sctp_af_inet.list);
1131 static int sctp_v4_protosw_init(void)
1133 int rc;
1135 rc = proto_register(&sctp_prot, 1);
1136 if (rc)
1137 return rc;
1139 /* Register SCTP(UDP and TCP style) with socket layer. */
1140 inet_register_protosw(&sctp_seqpacket_protosw);
1141 inet_register_protosw(&sctp_stream_protosw);
1143 return 0;
1146 static void sctp_v4_protosw_exit(void)
1148 inet_unregister_protosw(&sctp_stream_protosw);
1149 inet_unregister_protosw(&sctp_seqpacket_protosw);
1150 proto_unregister(&sctp_prot);
1153 static int sctp_v4_add_protocol(void)
1155 /* Register notifier for inet address additions/deletions. */
1156 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1158 /* Register SCTP with inet layer. */
1159 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1160 return -EAGAIN;
1162 return 0;
1165 static void sctp_v4_del_protocol(void)
1167 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1168 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1171 static int __net_init sctp_net_init(struct net *net)
1173 int status;
1176 * 14. Suggested SCTP Protocol Parameter Values
1178 /* The following protocol parameters are RECOMMENDED: */
1179 /* RTO.Initial - 3 seconds */
1180 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1181 /* RTO.Min - 1 second */
1182 net->sctp.rto_min = SCTP_RTO_MIN;
1183 /* RTO.Max - 60 seconds */
1184 net->sctp.rto_max = SCTP_RTO_MAX;
1185 /* RTO.Alpha - 1/8 */
1186 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1187 /* RTO.Beta - 1/4 */
1188 net->sctp.rto_beta = SCTP_RTO_BETA;
1190 /* Valid.Cookie.Life - 60 seconds */
1191 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1193 /* Whether Cookie Preservative is enabled(1) or not(0) */
1194 net->sctp.cookie_preserve_enable = 1;
1196 /* Default sctp sockets to use md5 as their hmac alg */
1197 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1198 net->sctp.sctp_hmac_alg = "md5";
1199 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1200 net->sctp.sctp_hmac_alg = "sha1";
1201 #else
1202 net->sctp.sctp_hmac_alg = NULL;
1203 #endif
1205 /* Max.Burst - 4 */
1206 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1208 /* Association.Max.Retrans - 10 attempts
1209 * Path.Max.Retrans - 5 attempts (per destination address)
1210 * Max.Init.Retransmits - 8 attempts
1212 net->sctp.max_retrans_association = 10;
1213 net->sctp.max_retrans_path = 5;
1214 net->sctp.max_retrans_init = 8;
1216 /* Sendbuffer growth - do per-socket accounting */
1217 net->sctp.sndbuf_policy = 0;
1219 /* Rcvbuffer growth - do per-socket accounting */
1220 net->sctp.rcvbuf_policy = 0;
1222 /* HB.interval - 30 seconds */
1223 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1225 /* delayed SACK timeout */
1226 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1228 /* Disable ADDIP by default. */
1229 net->sctp.addip_enable = 0;
1230 net->sctp.addip_noauth = 0;
1231 net->sctp.default_auto_asconf = 0;
1233 /* Enable PR-SCTP by default. */
1234 net->sctp.prsctp_enable = 1;
1236 /* Disable AUTH by default. */
1237 net->sctp.auth_enable = 0;
1239 /* Set SCOPE policy to enabled */
1240 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1242 /* Set the default rwnd update threshold */
1243 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1245 /* Initialize maximum autoclose timeout. */
1246 net->sctp.max_autoclose = INT_MAX / HZ;
1248 status = sctp_sysctl_net_register(net);
1249 if (status)
1250 goto err_sysctl_register;
1252 /* Allocate and initialise sctp mibs. */
1253 status = init_sctp_mibs(net);
1254 if (status)
1255 goto err_init_mibs;
1257 /* Initialize proc fs directory. */
1258 status = sctp_proc_init(net);
1259 if (status)
1260 goto err_init_proc;
1262 sctp_dbg_objcnt_init(net);
1264 /* Initialize the control inode/socket for handling OOTB packets. */
1265 if ((status = sctp_ctl_sock_init(net))) {
1266 pr_err("Failed to initialize the SCTP control sock\n");
1267 goto err_ctl_sock_init;
1270 /* Initialize the local address list. */
1271 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1272 spin_lock_init(&net->sctp.local_addr_lock);
1273 sctp_get_local_addr_list(net);
1275 /* Initialize the address event list */
1276 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1277 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1278 spin_lock_init(&net->sctp.addr_wq_lock);
1279 net->sctp.addr_wq_timer.expires = 0;
1280 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1281 (unsigned long)net);
1283 return 0;
1285 err_ctl_sock_init:
1286 sctp_dbg_objcnt_exit(net);
1287 sctp_proc_exit(net);
1288 err_init_proc:
1289 cleanup_sctp_mibs(net);
1290 err_init_mibs:
1291 sctp_sysctl_net_unregister(net);
1292 err_sysctl_register:
1293 return status;
1296 static void __net_exit sctp_net_exit(struct net *net)
1298 /* Free the local address list */
1299 sctp_free_addr_wq(net);
1300 sctp_free_local_addr_list(net);
1302 /* Free the control endpoint. */
1303 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1305 sctp_dbg_objcnt_exit(net);
1307 sctp_proc_exit(net);
1308 cleanup_sctp_mibs(net);
1309 sctp_sysctl_net_unregister(net);
1312 static struct pernet_operations sctp_net_ops = {
1313 .init = sctp_net_init,
1314 .exit = sctp_net_exit,
1317 /* Initialize the universe into something sensible. */
1318 static __init int sctp_init(void)
1320 int i;
1321 int status = -EINVAL;
1322 unsigned long goal;
1323 unsigned long limit;
1324 int max_share;
1325 int order;
1327 BUILD_BUG_ON(sizeof(struct sctp_ulpevent) >
1328 sizeof(((struct sk_buff *) 0)->cb));
1330 /* Allocate bind_bucket and chunk caches. */
1331 status = -ENOBUFS;
1332 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1333 sizeof(struct sctp_bind_bucket),
1334 0, SLAB_HWCACHE_ALIGN,
1335 NULL);
1336 if (!sctp_bucket_cachep)
1337 goto out;
1339 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1340 sizeof(struct sctp_chunk),
1341 0, SLAB_HWCACHE_ALIGN,
1342 NULL);
1343 if (!sctp_chunk_cachep)
1344 goto err_chunk_cachep;
1346 status = percpu_counter_init(&sctp_sockets_allocated, 0);
1347 if (status)
1348 goto err_percpu_counter_init;
1350 /* Implementation specific variables. */
1352 /* Initialize default stream count setup information. */
1353 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1354 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1356 /* Initialize handle used for association ids. */
1357 idr_init(&sctp_assocs_id);
1359 limit = nr_free_buffer_pages() / 8;
1360 limit = max(limit, 128UL);
1361 sysctl_sctp_mem[0] = limit / 4 * 3;
1362 sysctl_sctp_mem[1] = limit;
1363 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1365 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1366 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1367 max_share = min(4UL*1024*1024, limit);
1369 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1370 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1371 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1373 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1374 sysctl_sctp_wmem[1] = 16*1024;
1375 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1377 /* Size and allocate the association hash table.
1378 * The methodology is similar to that of the tcp hash tables.
1380 if (totalram_pages >= (128 * 1024))
1381 goal = totalram_pages >> (22 - PAGE_SHIFT);
1382 else
1383 goal = totalram_pages >> (24 - PAGE_SHIFT);
1385 for (order = 0; (1UL << order) < goal; order++)
1388 do {
1389 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1390 sizeof(struct sctp_hashbucket);
1391 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1392 continue;
1393 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1394 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1395 } while (!sctp_assoc_hashtable && --order > 0);
1396 if (!sctp_assoc_hashtable) {
1397 pr_err("Failed association hash alloc\n");
1398 status = -ENOMEM;
1399 goto err_ahash_alloc;
1401 for (i = 0; i < sctp_assoc_hashsize; i++) {
1402 rwlock_init(&sctp_assoc_hashtable[i].lock);
1403 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1406 /* Allocate and initialize the endpoint hash table. */
1407 sctp_ep_hashsize = 64;
1408 sctp_ep_hashtable =
1409 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1410 if (!sctp_ep_hashtable) {
1411 pr_err("Failed endpoint_hash alloc\n");
1412 status = -ENOMEM;
1413 goto err_ehash_alloc;
1415 for (i = 0; i < sctp_ep_hashsize; i++) {
1416 rwlock_init(&sctp_ep_hashtable[i].lock);
1417 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1420 /* Allocate and initialize the SCTP port hash table. */
1421 do {
1422 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1423 sizeof(struct sctp_bind_hashbucket);
1424 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1425 continue;
1426 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1427 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1428 } while (!sctp_port_hashtable && --order > 0);
1429 if (!sctp_port_hashtable) {
1430 pr_err("Failed bind hash alloc\n");
1431 status = -ENOMEM;
1432 goto err_bhash_alloc;
1434 for (i = 0; i < sctp_port_hashsize; i++) {
1435 spin_lock_init(&sctp_port_hashtable[i].lock);
1436 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1439 pr_info("Hash tables configured (established %d bind %d)\n",
1440 sctp_assoc_hashsize, sctp_port_hashsize);
1442 sctp_sysctl_register();
1444 INIT_LIST_HEAD(&sctp_address_families);
1445 sctp_v4_pf_init();
1446 sctp_v6_pf_init();
1448 status = sctp_v4_protosw_init();
1450 if (status)
1451 goto err_protosw_init;
1453 status = sctp_v6_protosw_init();
1454 if (status)
1455 goto err_v6_protosw_init;
1457 status = register_pernet_subsys(&sctp_net_ops);
1458 if (status)
1459 goto err_register_pernet_subsys;
1461 status = sctp_v4_add_protocol();
1462 if (status)
1463 goto err_add_protocol;
1465 /* Register SCTP with inet6 layer. */
1466 status = sctp_v6_add_protocol();
1467 if (status)
1468 goto err_v6_add_protocol;
1470 status = 0;
1471 out:
1472 return status;
1473 err_v6_add_protocol:
1474 sctp_v4_del_protocol();
1475 err_add_protocol:
1476 unregister_pernet_subsys(&sctp_net_ops);
1477 err_register_pernet_subsys:
1478 sctp_v6_protosw_exit();
1479 err_v6_protosw_init:
1480 sctp_v4_protosw_exit();
1481 err_protosw_init:
1482 sctp_v4_pf_exit();
1483 sctp_v6_pf_exit();
1484 sctp_sysctl_unregister();
1485 free_pages((unsigned long)sctp_port_hashtable,
1486 get_order(sctp_port_hashsize *
1487 sizeof(struct sctp_bind_hashbucket)));
1488 err_bhash_alloc:
1489 kfree(sctp_ep_hashtable);
1490 err_ehash_alloc:
1491 free_pages((unsigned long)sctp_assoc_hashtable,
1492 get_order(sctp_assoc_hashsize *
1493 sizeof(struct sctp_hashbucket)));
1494 err_ahash_alloc:
1495 percpu_counter_destroy(&sctp_sockets_allocated);
1496 err_percpu_counter_init:
1497 kmem_cache_destroy(sctp_chunk_cachep);
1498 err_chunk_cachep:
1499 kmem_cache_destroy(sctp_bucket_cachep);
1500 goto out;
1503 /* Exit handler for the SCTP protocol. */
1504 static __exit void sctp_exit(void)
1506 /* BUG. This should probably do something useful like clean
1507 * up all the remaining associations and all that memory.
1510 /* Unregister with inet6/inet layers. */
1511 sctp_v6_del_protocol();
1512 sctp_v4_del_protocol();
1514 unregister_pernet_subsys(&sctp_net_ops);
1516 /* Free protosw registrations */
1517 sctp_v6_protosw_exit();
1518 sctp_v4_protosw_exit();
1520 /* Unregister with socket layer. */
1521 sctp_v6_pf_exit();
1522 sctp_v4_pf_exit();
1524 sctp_sysctl_unregister();
1526 free_pages((unsigned long)sctp_assoc_hashtable,
1527 get_order(sctp_assoc_hashsize *
1528 sizeof(struct sctp_hashbucket)));
1529 kfree(sctp_ep_hashtable);
1530 free_pages((unsigned long)sctp_port_hashtable,
1531 get_order(sctp_port_hashsize *
1532 sizeof(struct sctp_bind_hashbucket)));
1534 percpu_counter_destroy(&sctp_sockets_allocated);
1536 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1538 kmem_cache_destroy(sctp_chunk_cachep);
1539 kmem_cache_destroy(sctp_bucket_cachep);
1542 module_init(sctp_init);
1543 module_exit(sctp_exit);
1546 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1548 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1549 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1550 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1551 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1552 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1553 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1554 MODULE_LICENSE("GPL");