Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
[linux/fpc-iii.git] / net / sctp / protocol.c
blob4e1d0fcb028efa3001a0b1b69eaf80836765b49c
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, see
27 * <http://www.gnu.org/licenses/>.
29 * Please send any bug reports or fixes you make to the
30 * email address(es):
31 * lksctp developers <linux-sctp@vger.kernel.org>
33 * Written or modified by:
34 * La Monte H.P. Yarroll <piggy@acm.org>
35 * Karl Knutson <karl@athena.chicago.il.us>
36 * Jon Grimm <jgrimm@us.ibm.com>
37 * Sridhar Samudrala <sri@us.ibm.com>
38 * Daisy Chang <daisyc@us.ibm.com>
39 * Ardelle Fan <ardelle.fan@intel.com>
42 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
44 #include <linux/module.h>
45 #include <linux/init.h>
46 #include <linux/netdevice.h>
47 #include <linux/inetdevice.h>
48 #include <linux/seq_file.h>
49 #include <linux/bootmem.h>
50 #include <linux/highmem.h>
51 #include <linux/swap.h>
52 #include <linux/slab.h>
53 #include <net/net_namespace.h>
54 #include <net/protocol.h>
55 #include <net/ip.h>
56 #include <net/ipv6.h>
57 #include <net/route.h>
58 #include <net/sctp/sctp.h>
59 #include <net/addrconf.h>
60 #include <net/inet_common.h>
61 #include <net/inet_ecn.h>
63 /* Global data structures. */
64 struct sctp_globals sctp_globals __read_mostly;
66 struct idr sctp_assocs_id;
67 DEFINE_SPINLOCK(sctp_assocs_id_lock);
69 static struct sctp_pf *sctp_pf_inet6_specific;
70 static struct sctp_pf *sctp_pf_inet_specific;
71 static struct sctp_af *sctp_af_v4_specific;
72 static struct sctp_af *sctp_af_v6_specific;
74 struct kmem_cache *sctp_chunk_cachep __read_mostly;
75 struct kmem_cache *sctp_bucket_cachep __read_mostly;
77 long sysctl_sctp_mem[3];
78 int sysctl_sctp_rmem[3];
79 int sysctl_sctp_wmem[3];
81 /* Set up the proc fs entry for the SCTP protocol. */
82 static int __net_init sctp_proc_init(struct net *net)
84 #ifdef CONFIG_PROC_FS
85 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
86 if (!net->sctp.proc_net_sctp)
87 goto out_proc_net_sctp;
88 if (sctp_snmp_proc_init(net))
89 goto out_snmp_proc_init;
90 if (sctp_eps_proc_init(net))
91 goto out_eps_proc_init;
92 if (sctp_assocs_proc_init(net))
93 goto out_assocs_proc_init;
94 if (sctp_remaddr_proc_init(net))
95 goto out_remaddr_proc_init;
97 return 0;
99 out_remaddr_proc_init:
100 sctp_assocs_proc_exit(net);
101 out_assocs_proc_init:
102 sctp_eps_proc_exit(net);
103 out_eps_proc_init:
104 sctp_snmp_proc_exit(net);
105 out_snmp_proc_init:
106 remove_proc_entry("sctp", net->proc_net);
107 net->sctp.proc_net_sctp = NULL;
108 out_proc_net_sctp:
109 return -ENOMEM;
110 #endif /* CONFIG_PROC_FS */
111 return 0;
114 /* Clean up the proc fs entry for the SCTP protocol.
115 * Note: Do not make this __exit as it is used in the init error
116 * path.
118 static void sctp_proc_exit(struct net *net)
120 #ifdef CONFIG_PROC_FS
121 sctp_snmp_proc_exit(net);
122 sctp_eps_proc_exit(net);
123 sctp_assocs_proc_exit(net);
124 sctp_remaddr_proc_exit(net);
126 remove_proc_entry("sctp", net->proc_net);
127 net->sctp.proc_net_sctp = NULL;
128 #endif
131 /* Private helper to extract ipv4 address and stash them in
132 * the protocol structure.
134 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
135 struct net_device *dev)
137 struct in_device *in_dev;
138 struct in_ifaddr *ifa;
139 struct sctp_sockaddr_entry *addr;
141 rcu_read_lock();
142 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
143 rcu_read_unlock();
144 return;
147 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
148 /* Add the address to the local list. */
149 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
150 if (addr) {
151 addr->a.v4.sin_family = AF_INET;
152 addr->a.v4.sin_port = 0;
153 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
154 addr->valid = 1;
155 INIT_LIST_HEAD(&addr->list);
156 list_add_tail(&addr->list, addrlist);
160 rcu_read_unlock();
163 /* Extract our IP addresses from the system and stash them in the
164 * protocol structure.
166 static void sctp_get_local_addr_list(struct net *net)
168 struct net_device *dev;
169 struct list_head *pos;
170 struct sctp_af *af;
172 rcu_read_lock();
173 for_each_netdev_rcu(net, dev) {
174 list_for_each(pos, &sctp_address_families) {
175 af = list_entry(pos, struct sctp_af, list);
176 af->copy_addrlist(&net->sctp.local_addr_list, dev);
179 rcu_read_unlock();
182 /* Free the existing local addresses. */
183 static void sctp_free_local_addr_list(struct net *net)
185 struct sctp_sockaddr_entry *addr;
186 struct list_head *pos, *temp;
188 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
189 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
190 list_del(pos);
191 kfree(addr);
195 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
196 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
197 sctp_scope_t scope, gfp_t gfp, int copy_flags)
199 struct sctp_sockaddr_entry *addr;
200 int error = 0;
202 rcu_read_lock();
203 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
204 if (!addr->valid)
205 continue;
206 if (sctp_in_scope(net, &addr->a, scope)) {
207 /* Now that the address is in scope, check to see if
208 * the address type is really supported by the local
209 * sock as well as the remote peer.
211 if ((((AF_INET == addr->a.sa.sa_family) &&
212 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
213 (((AF_INET6 == addr->a.sa.sa_family) &&
214 (copy_flags & SCTP_ADDR6_ALLOWED) &&
215 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
216 error = sctp_add_bind_addr(bp, &addr->a,
217 SCTP_ADDR_SRC, GFP_ATOMIC);
218 if (error)
219 goto end_copy;
224 end_copy:
225 rcu_read_unlock();
226 return error;
229 /* Initialize a sctp_addr from in incoming skb. */
230 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
231 int is_saddr)
233 void *from;
234 __be16 *port;
235 struct sctphdr *sh;
237 port = &addr->v4.sin_port;
238 addr->v4.sin_family = AF_INET;
240 sh = sctp_hdr(skb);
241 if (is_saddr) {
242 *port = sh->source;
243 from = &ip_hdr(skb)->saddr;
244 } else {
245 *port = sh->dest;
246 from = &ip_hdr(skb)->daddr;
248 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
251 /* Initialize an sctp_addr from a socket. */
252 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
254 addr->v4.sin_family = AF_INET;
255 addr->v4.sin_port = 0;
256 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
259 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
260 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
262 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
265 /* Initialize sk->sk_daddr from sctp_addr. */
266 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
268 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
271 /* Initialize a sctp_addr from an address parameter. */
272 static void sctp_v4_from_addr_param(union sctp_addr *addr,
273 union sctp_addr_param *param,
274 __be16 port, int iif)
276 addr->v4.sin_family = AF_INET;
277 addr->v4.sin_port = port;
278 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
281 /* Initialize an address parameter from a sctp_addr and return the length
282 * of the address parameter.
284 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
285 union sctp_addr_param *param)
287 int length = sizeof(sctp_ipv4addr_param_t);
289 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
290 param->v4.param_hdr.length = htons(length);
291 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
293 return length;
296 /* Initialize a sctp_addr from a dst_entry. */
297 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
298 __be16 port)
300 saddr->v4.sin_family = AF_INET;
301 saddr->v4.sin_port = port;
302 saddr->v4.sin_addr.s_addr = fl4->saddr;
305 /* Compare two addresses exactly. */
306 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
307 const union sctp_addr *addr2)
309 if (addr1->sa.sa_family != addr2->sa.sa_family)
310 return 0;
311 if (addr1->v4.sin_port != addr2->v4.sin_port)
312 return 0;
313 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
314 return 0;
316 return 1;
319 /* Initialize addr struct to INADDR_ANY. */
320 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
322 addr->v4.sin_family = AF_INET;
323 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
324 addr->v4.sin_port = port;
327 /* Is this a wildcard address? */
328 static int sctp_v4_is_any(const union sctp_addr *addr)
330 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
333 /* This function checks if the address is a valid address to be used for
334 * SCTP binding.
336 * Output:
337 * Return 0 - If the address is a non-unicast or an illegal address.
338 * Return 1 - If the address is a unicast.
340 static int sctp_v4_addr_valid(union sctp_addr *addr,
341 struct sctp_sock *sp,
342 const struct sk_buff *skb)
344 /* IPv4 addresses not allowed */
345 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
346 return 0;
348 /* Is this a non-unicast address or a unusable SCTP address? */
349 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
350 return 0;
352 /* Is this a broadcast address? */
353 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
354 return 0;
356 return 1;
359 /* Should this be available for binding? */
360 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
362 struct net *net = sock_net(&sp->inet.sk);
363 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
366 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
367 ret != RTN_LOCAL &&
368 !sp->inet.freebind &&
369 !sysctl_ip_nonlocal_bind)
370 return 0;
372 if (ipv6_only_sock(sctp_opt2sk(sp)))
373 return 0;
375 return 1;
378 /* Checking the loopback, private and other address scopes as defined in
379 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
380 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
382 * Level 0 - unusable SCTP addresses
383 * Level 1 - loopback address
384 * Level 2 - link-local addresses
385 * Level 3 - private addresses.
386 * Level 4 - global addresses
387 * For INIT and INIT-ACK address list, let L be the level of
388 * of requested destination address, sender and receiver
389 * SHOULD include all of its addresses with level greater
390 * than or equal to L.
392 * IPv4 scoping can be controlled through sysctl option
393 * net.sctp.addr_scope_policy
395 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
397 sctp_scope_t retval;
399 /* Check for unusable SCTP addresses. */
400 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
401 retval = SCTP_SCOPE_UNUSABLE;
402 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
403 retval = SCTP_SCOPE_LOOPBACK;
404 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
405 retval = SCTP_SCOPE_LINK;
406 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
407 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
408 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
409 retval = SCTP_SCOPE_PRIVATE;
410 } else {
411 retval = SCTP_SCOPE_GLOBAL;
414 return retval;
417 /* Returns a valid dst cache entry for the given source and destination ip
418 * addresses. If an association is passed, trys to get a dst entry with a
419 * source address that matches an address in the bind address list.
421 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
422 struct flowi *fl, struct sock *sk)
424 struct sctp_association *asoc = t->asoc;
425 struct rtable *rt;
426 struct flowi4 *fl4 = &fl->u.ip4;
427 struct sctp_bind_addr *bp;
428 struct sctp_sockaddr_entry *laddr;
429 struct dst_entry *dst = NULL;
430 union sctp_addr *daddr = &t->ipaddr;
431 union sctp_addr dst_saddr;
433 memset(fl4, 0x0, sizeof(struct flowi4));
434 fl4->daddr = daddr->v4.sin_addr.s_addr;
435 fl4->fl4_dport = daddr->v4.sin_port;
436 fl4->flowi4_proto = IPPROTO_SCTP;
437 if (asoc) {
438 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
439 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
440 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
442 if (saddr) {
443 fl4->saddr = saddr->v4.sin_addr.s_addr;
444 fl4->fl4_sport = saddr->v4.sin_port;
447 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
448 &fl4->saddr);
450 rt = ip_route_output_key(sock_net(sk), fl4);
451 if (!IS_ERR(rt))
452 dst = &rt->dst;
454 /* If there is no association or if a source address is passed, no
455 * more validation is required.
457 if (!asoc || saddr)
458 goto out;
460 bp = &asoc->base.bind_addr;
462 if (dst) {
463 /* Walk through the bind address list and look for a bind
464 * address that matches the source address of the returned dst.
466 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
467 rcu_read_lock();
468 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
469 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
470 (laddr->state != SCTP_ADDR_SRC &&
471 !asoc->src_out_of_asoc_ok))
472 continue;
473 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
474 goto out_unlock;
476 rcu_read_unlock();
478 /* None of the bound addresses match the source address of the
479 * dst. So release it.
481 dst_release(dst);
482 dst = NULL;
485 /* Walk through the bind address list and try to get a dst that
486 * matches a bind address as the source address.
488 rcu_read_lock();
489 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
490 if (!laddr->valid)
491 continue;
492 if ((laddr->state == SCTP_ADDR_SRC) &&
493 (AF_INET == laddr->a.sa.sa_family)) {
494 fl4->saddr = laddr->a.v4.sin_addr.s_addr;
495 fl4->fl4_sport = laddr->a.v4.sin_port;
496 rt = ip_route_output_key(sock_net(sk), fl4);
497 if (!IS_ERR(rt)) {
498 dst = &rt->dst;
499 goto out_unlock;
504 out_unlock:
505 rcu_read_unlock();
506 out:
507 t->dst = dst;
508 if (dst)
509 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
510 &fl4->daddr, &fl4->saddr);
511 else
512 pr_debug("no route\n");
515 /* For v4, the source address is cached in the route entry(dst). So no need
516 * to cache it separately and hence this is an empty routine.
518 static void sctp_v4_get_saddr(struct sctp_sock *sk,
519 struct sctp_transport *t,
520 struct flowi *fl)
522 union sctp_addr *saddr = &t->saddr;
523 struct rtable *rt = (struct rtable *)t->dst;
525 if (rt) {
526 saddr->v4.sin_family = AF_INET;
527 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
531 /* What interface did this skb arrive on? */
532 static int sctp_v4_skb_iif(const struct sk_buff *skb)
534 return inet_iif(skb);
537 /* Was this packet marked by Explicit Congestion Notification? */
538 static int sctp_v4_is_ce(const struct sk_buff *skb)
540 return INET_ECN_is_ce(ip_hdr(skb)->tos);
543 /* Create and initialize a new sk for the socket returned by accept(). */
544 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
545 struct sctp_association *asoc)
547 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
548 sk->sk_prot);
549 struct inet_sock *newinet;
551 if (!newsk)
552 goto out;
554 sock_init_data(NULL, newsk);
556 sctp_copy_sock(newsk, sk, asoc);
557 sock_reset_flag(newsk, SOCK_ZAPPED);
559 newinet = inet_sk(newsk);
561 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
563 sk_refcnt_debug_inc(newsk);
565 if (newsk->sk_prot->init(newsk)) {
566 sk_common_release(newsk);
567 newsk = NULL;
570 out:
571 return newsk;
574 /* Map address, empty for v4 family */
575 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
577 /* Empty */
580 /* Dump the v4 addr to the seq file. */
581 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
583 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
586 static void sctp_v4_ecn_capable(struct sock *sk)
588 INET_ECN_xmit(sk);
591 static void sctp_addr_wq_timeout_handler(unsigned long arg)
593 struct net *net = (struct net *)arg;
594 struct sctp_sockaddr_entry *addrw, *temp;
595 struct sctp_sock *sp;
597 spin_lock_bh(&net->sctp.addr_wq_lock);
599 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
600 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
601 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
602 addrw->state, addrw);
604 #if IS_ENABLED(CONFIG_IPV6)
605 /* Now we send an ASCONF for each association */
606 /* Note. we currently don't handle link local IPv6 addressees */
607 if (addrw->a.sa.sa_family == AF_INET6) {
608 struct in6_addr *in6;
610 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
611 IPV6_ADDR_LINKLOCAL)
612 goto free_next;
614 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
615 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
616 addrw->state == SCTP_ADDR_NEW) {
617 unsigned long timeo_val;
619 pr_debug("%s: this is on DAD, trying %d sec "
620 "later\n", __func__,
621 SCTP_ADDRESS_TICK_DELAY);
623 timeo_val = jiffies;
624 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
625 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
626 break;
629 #endif
630 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
631 struct sock *sk;
633 sk = sctp_opt2sk(sp);
634 /* ignore bound-specific endpoints */
635 if (!sctp_is_ep_boundall(sk))
636 continue;
637 bh_lock_sock(sk);
638 if (sctp_asconf_mgmt(sp, addrw) < 0)
639 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
640 bh_unlock_sock(sk);
642 #if IS_ENABLED(CONFIG_IPV6)
643 free_next:
644 #endif
645 list_del(&addrw->list);
646 kfree(addrw);
648 spin_unlock_bh(&net->sctp.addr_wq_lock);
651 static void sctp_free_addr_wq(struct net *net)
653 struct sctp_sockaddr_entry *addrw;
654 struct sctp_sockaddr_entry *temp;
656 spin_lock_bh(&net->sctp.addr_wq_lock);
657 del_timer(&net->sctp.addr_wq_timer);
658 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
659 list_del(&addrw->list);
660 kfree(addrw);
662 spin_unlock_bh(&net->sctp.addr_wq_lock);
665 /* lookup the entry for the same address in the addr_waitq
666 * sctp_addr_wq MUST be locked
668 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
669 struct sctp_sockaddr_entry *addr)
671 struct sctp_sockaddr_entry *addrw;
673 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
674 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
675 continue;
676 if (addrw->a.sa.sa_family == AF_INET) {
677 if (addrw->a.v4.sin_addr.s_addr ==
678 addr->a.v4.sin_addr.s_addr)
679 return addrw;
680 } else if (addrw->a.sa.sa_family == AF_INET6) {
681 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
682 &addr->a.v6.sin6_addr))
683 return addrw;
686 return NULL;
689 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
691 struct sctp_sockaddr_entry *addrw;
692 unsigned long timeo_val;
694 /* first, we check if an opposite message already exist in the queue.
695 * If we found such message, it is removed.
696 * This operation is a bit stupid, but the DHCP client attaches the
697 * new address after a couple of addition and deletion of that address
700 spin_lock_bh(&net->sctp.addr_wq_lock);
701 /* Offsets existing events in addr_wq */
702 addrw = sctp_addr_wq_lookup(net, addr);
703 if (addrw) {
704 if (addrw->state != cmd) {
705 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
706 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
707 &net->sctp.addr_waitq);
709 list_del(&addrw->list);
710 kfree(addrw);
712 spin_unlock_bh(&net->sctp.addr_wq_lock);
713 return;
716 /* OK, we have to add the new address to the wait queue */
717 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
718 if (addrw == NULL) {
719 spin_unlock_bh(&net->sctp.addr_wq_lock);
720 return;
722 addrw->state = cmd;
723 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
725 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
726 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
728 if (!timer_pending(&net->sctp.addr_wq_timer)) {
729 timeo_val = jiffies;
730 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
731 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
733 spin_unlock_bh(&net->sctp.addr_wq_lock);
736 /* Event handler for inet address addition/deletion events.
737 * The sctp_local_addr_list needs to be protocted by a spin lock since
738 * multiple notifiers (say IPv4 and IPv6) may be running at the same
739 * time and thus corrupt the list.
740 * The reader side is protected with RCU.
742 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
743 void *ptr)
745 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
746 struct sctp_sockaddr_entry *addr = NULL;
747 struct sctp_sockaddr_entry *temp;
748 struct net *net = dev_net(ifa->ifa_dev->dev);
749 int found = 0;
751 switch (ev) {
752 case NETDEV_UP:
753 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
754 if (addr) {
755 addr->a.v4.sin_family = AF_INET;
756 addr->a.v4.sin_port = 0;
757 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
758 addr->valid = 1;
759 spin_lock_bh(&net->sctp.local_addr_lock);
760 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
761 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
762 spin_unlock_bh(&net->sctp.local_addr_lock);
764 break;
765 case NETDEV_DOWN:
766 spin_lock_bh(&net->sctp.local_addr_lock);
767 list_for_each_entry_safe(addr, temp,
768 &net->sctp.local_addr_list, list) {
769 if (addr->a.sa.sa_family == AF_INET &&
770 addr->a.v4.sin_addr.s_addr ==
771 ifa->ifa_local) {
772 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
773 found = 1;
774 addr->valid = 0;
775 list_del_rcu(&addr->list);
776 break;
779 spin_unlock_bh(&net->sctp.local_addr_lock);
780 if (found)
781 kfree_rcu(addr, rcu);
782 break;
785 return NOTIFY_DONE;
789 * Initialize the control inode/socket with a control endpoint data
790 * structure. This endpoint is reserved exclusively for the OOTB processing.
792 static int sctp_ctl_sock_init(struct net *net)
794 int err;
795 sa_family_t family = PF_INET;
797 if (sctp_get_pf_specific(PF_INET6))
798 family = PF_INET6;
800 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
801 SOCK_SEQPACKET, IPPROTO_SCTP, net);
803 /* If IPv6 socket could not be created, try the IPv4 socket */
804 if (err < 0 && family == PF_INET6)
805 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
806 SOCK_SEQPACKET, IPPROTO_SCTP,
807 net);
809 if (err < 0) {
810 pr_err("Failed to create the SCTP control socket\n");
811 return err;
813 return 0;
816 /* Register address family specific functions. */
817 int sctp_register_af(struct sctp_af *af)
819 switch (af->sa_family) {
820 case AF_INET:
821 if (sctp_af_v4_specific)
822 return 0;
823 sctp_af_v4_specific = af;
824 break;
825 case AF_INET6:
826 if (sctp_af_v6_specific)
827 return 0;
828 sctp_af_v6_specific = af;
829 break;
830 default:
831 return 0;
834 INIT_LIST_HEAD(&af->list);
835 list_add_tail(&af->list, &sctp_address_families);
836 return 1;
839 /* Get the table of functions for manipulating a particular address
840 * family.
842 struct sctp_af *sctp_get_af_specific(sa_family_t family)
844 switch (family) {
845 case AF_INET:
846 return sctp_af_v4_specific;
847 case AF_INET6:
848 return sctp_af_v6_specific;
849 default:
850 return NULL;
854 /* Common code to initialize a AF_INET msg_name. */
855 static void sctp_inet_msgname(char *msgname, int *addr_len)
857 struct sockaddr_in *sin;
859 sin = (struct sockaddr_in *)msgname;
860 *addr_len = sizeof(struct sockaddr_in);
861 sin->sin_family = AF_INET;
862 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
865 /* Copy the primary address of the peer primary address as the msg_name. */
866 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
867 int *addr_len)
869 struct sockaddr_in *sin, *sinfrom;
871 if (msgname) {
872 struct sctp_association *asoc;
874 asoc = event->asoc;
875 sctp_inet_msgname(msgname, addr_len);
876 sin = (struct sockaddr_in *)msgname;
877 sinfrom = &asoc->peer.primary_addr.v4;
878 sin->sin_port = htons(asoc->peer.port);
879 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
883 /* Initialize and copy out a msgname from an inbound skb. */
884 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
886 if (msgname) {
887 struct sctphdr *sh = sctp_hdr(skb);
888 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
890 sctp_inet_msgname(msgname, len);
891 sin->sin_port = sh->source;
892 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
896 /* Do we support this AF? */
897 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
899 /* PF_INET only supports AF_INET addresses. */
900 return AF_INET == family;
903 /* Address matching with wildcards allowed. */
904 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
905 const union sctp_addr *addr2,
906 struct sctp_sock *opt)
908 /* PF_INET only supports AF_INET addresses. */
909 if (addr1->sa.sa_family != addr2->sa.sa_family)
910 return 0;
911 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
912 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
913 return 1;
914 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
915 return 1;
917 return 0;
920 /* Verify that provided sockaddr looks bindable. Common verification has
921 * already been taken care of.
923 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
925 return sctp_v4_available(addr, opt);
928 /* Verify that sockaddr looks sendable. Common verification has already
929 * been taken care of.
931 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
933 return 1;
936 /* Fill in Supported Address Type information for INIT and INIT-ACK
937 * chunks. Returns number of addresses supported.
939 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
940 __be16 *types)
942 types[0] = SCTP_PARAM_IPV4_ADDRESS;
943 return 1;
946 /* Wrapper routine that calls the ip transmit routine. */
947 static inline int sctp_v4_xmit(struct sk_buff *skb,
948 struct sctp_transport *transport)
950 struct inet_sock *inet = inet_sk(skb->sk);
952 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
953 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
955 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
956 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
958 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
960 return ip_queue_xmit(skb, &transport->fl);
963 static struct sctp_af sctp_af_inet;
965 static struct sctp_pf sctp_pf_inet = {
966 .event_msgname = sctp_inet_event_msgname,
967 .skb_msgname = sctp_inet_skb_msgname,
968 .af_supported = sctp_inet_af_supported,
969 .cmp_addr = sctp_inet_cmp_addr,
970 .bind_verify = sctp_inet_bind_verify,
971 .send_verify = sctp_inet_send_verify,
972 .supported_addrs = sctp_inet_supported_addrs,
973 .create_accept_sk = sctp_v4_create_accept_sk,
974 .addr_v4map = sctp_v4_addr_v4map,
975 .af = &sctp_af_inet
978 /* Notifier for inetaddr addition/deletion events. */
979 static struct notifier_block sctp_inetaddr_notifier = {
980 .notifier_call = sctp_inetaddr_event,
983 /* Socket operations. */
984 static const struct proto_ops inet_seqpacket_ops = {
985 .family = PF_INET,
986 .owner = THIS_MODULE,
987 .release = inet_release, /* Needs to be wrapped... */
988 .bind = inet_bind,
989 .connect = inet_dgram_connect,
990 .socketpair = sock_no_socketpair,
991 .accept = inet_accept,
992 .getname = inet_getname, /* Semantics are different. */
993 .poll = sctp_poll,
994 .ioctl = inet_ioctl,
995 .listen = sctp_inet_listen,
996 .shutdown = inet_shutdown, /* Looks harmless. */
997 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
998 .getsockopt = sock_common_getsockopt,
999 .sendmsg = inet_sendmsg,
1000 .recvmsg = sock_common_recvmsg,
1001 .mmap = sock_no_mmap,
1002 .sendpage = sock_no_sendpage,
1003 #ifdef CONFIG_COMPAT
1004 .compat_setsockopt = compat_sock_common_setsockopt,
1005 .compat_getsockopt = compat_sock_common_getsockopt,
1006 #endif
1009 /* Registration with AF_INET family. */
1010 static struct inet_protosw sctp_seqpacket_protosw = {
1011 .type = SOCK_SEQPACKET,
1012 .protocol = IPPROTO_SCTP,
1013 .prot = &sctp_prot,
1014 .ops = &inet_seqpacket_ops,
1015 .no_check = 0,
1016 .flags = SCTP_PROTOSW_FLAG
1018 static struct inet_protosw sctp_stream_protosw = {
1019 .type = SOCK_STREAM,
1020 .protocol = IPPROTO_SCTP,
1021 .prot = &sctp_prot,
1022 .ops = &inet_seqpacket_ops,
1023 .no_check = 0,
1024 .flags = SCTP_PROTOSW_FLAG
1027 /* Register with IP layer. */
1028 static const struct net_protocol sctp_protocol = {
1029 .handler = sctp_rcv,
1030 .err_handler = sctp_v4_err,
1031 .no_policy = 1,
1032 .netns_ok = 1,
1033 .icmp_strict_tag_validation = 1,
1036 /* IPv4 address related functions. */
1037 static struct sctp_af sctp_af_inet = {
1038 .sa_family = AF_INET,
1039 .sctp_xmit = sctp_v4_xmit,
1040 .setsockopt = ip_setsockopt,
1041 .getsockopt = ip_getsockopt,
1042 .get_dst = sctp_v4_get_dst,
1043 .get_saddr = sctp_v4_get_saddr,
1044 .copy_addrlist = sctp_v4_copy_addrlist,
1045 .from_skb = sctp_v4_from_skb,
1046 .from_sk = sctp_v4_from_sk,
1047 .to_sk_saddr = sctp_v4_to_sk_saddr,
1048 .to_sk_daddr = sctp_v4_to_sk_daddr,
1049 .from_addr_param = sctp_v4_from_addr_param,
1050 .to_addr_param = sctp_v4_to_addr_param,
1051 .cmp_addr = sctp_v4_cmp_addr,
1052 .addr_valid = sctp_v4_addr_valid,
1053 .inaddr_any = sctp_v4_inaddr_any,
1054 .is_any = sctp_v4_is_any,
1055 .available = sctp_v4_available,
1056 .scope = sctp_v4_scope,
1057 .skb_iif = sctp_v4_skb_iif,
1058 .is_ce = sctp_v4_is_ce,
1059 .seq_dump_addr = sctp_v4_seq_dump_addr,
1060 .ecn_capable = sctp_v4_ecn_capable,
1061 .net_header_len = sizeof(struct iphdr),
1062 .sockaddr_len = sizeof(struct sockaddr_in),
1063 #ifdef CONFIG_COMPAT
1064 .compat_setsockopt = compat_ip_setsockopt,
1065 .compat_getsockopt = compat_ip_getsockopt,
1066 #endif
1069 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1071 switch (family) {
1072 case PF_INET:
1073 return sctp_pf_inet_specific;
1074 case PF_INET6:
1075 return sctp_pf_inet6_specific;
1076 default:
1077 return NULL;
1081 /* Register the PF specific function table. */
1082 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1084 switch (family) {
1085 case PF_INET:
1086 if (sctp_pf_inet_specific)
1087 return 0;
1088 sctp_pf_inet_specific = pf;
1089 break;
1090 case PF_INET6:
1091 if (sctp_pf_inet6_specific)
1092 return 0;
1093 sctp_pf_inet6_specific = pf;
1094 break;
1095 default:
1096 return 0;
1098 return 1;
1101 static inline int init_sctp_mibs(struct net *net)
1103 return snmp_mib_init((void __percpu **)net->sctp.sctp_statistics,
1104 sizeof(struct sctp_mib),
1105 __alignof__(struct sctp_mib));
1108 static inline void cleanup_sctp_mibs(struct net *net)
1110 snmp_mib_free((void __percpu **)net->sctp.sctp_statistics);
1113 static void sctp_v4_pf_init(void)
1115 /* Initialize the SCTP specific PF functions. */
1116 sctp_register_pf(&sctp_pf_inet, PF_INET);
1117 sctp_register_af(&sctp_af_inet);
1120 static void sctp_v4_pf_exit(void)
1122 list_del(&sctp_af_inet.list);
1125 static int sctp_v4_protosw_init(void)
1127 int rc;
1129 rc = proto_register(&sctp_prot, 1);
1130 if (rc)
1131 return rc;
1133 /* Register SCTP(UDP and TCP style) with socket layer. */
1134 inet_register_protosw(&sctp_seqpacket_protosw);
1135 inet_register_protosw(&sctp_stream_protosw);
1137 return 0;
1140 static void sctp_v4_protosw_exit(void)
1142 inet_unregister_protosw(&sctp_stream_protosw);
1143 inet_unregister_protosw(&sctp_seqpacket_protosw);
1144 proto_unregister(&sctp_prot);
1147 static int sctp_v4_add_protocol(void)
1149 /* Register notifier for inet address additions/deletions. */
1150 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1152 /* Register SCTP with inet layer. */
1153 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1154 return -EAGAIN;
1156 return 0;
1159 static void sctp_v4_del_protocol(void)
1161 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1162 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1165 static int __net_init sctp_net_init(struct net *net)
1167 int status;
1170 * 14. Suggested SCTP Protocol Parameter Values
1172 /* The following protocol parameters are RECOMMENDED: */
1173 /* RTO.Initial - 3 seconds */
1174 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1175 /* RTO.Min - 1 second */
1176 net->sctp.rto_min = SCTP_RTO_MIN;
1177 /* RTO.Max - 60 seconds */
1178 net->sctp.rto_max = SCTP_RTO_MAX;
1179 /* RTO.Alpha - 1/8 */
1180 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1181 /* RTO.Beta - 1/4 */
1182 net->sctp.rto_beta = SCTP_RTO_BETA;
1184 /* Valid.Cookie.Life - 60 seconds */
1185 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1187 /* Whether Cookie Preservative is enabled(1) or not(0) */
1188 net->sctp.cookie_preserve_enable = 1;
1190 /* Default sctp sockets to use md5 as their hmac alg */
1191 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1192 net->sctp.sctp_hmac_alg = "md5";
1193 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1194 net->sctp.sctp_hmac_alg = "sha1";
1195 #else
1196 net->sctp.sctp_hmac_alg = NULL;
1197 #endif
1199 /* Max.Burst - 4 */
1200 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1202 /* Association.Max.Retrans - 10 attempts
1203 * Path.Max.Retrans - 5 attempts (per destination address)
1204 * Max.Init.Retransmits - 8 attempts
1206 net->sctp.max_retrans_association = 10;
1207 net->sctp.max_retrans_path = 5;
1208 net->sctp.max_retrans_init = 8;
1210 /* Sendbuffer growth - do per-socket accounting */
1211 net->sctp.sndbuf_policy = 0;
1213 /* Rcvbuffer growth - do per-socket accounting */
1214 net->sctp.rcvbuf_policy = 0;
1216 /* HB.interval - 30 seconds */
1217 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1219 /* delayed SACK timeout */
1220 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1222 /* Disable ADDIP by default. */
1223 net->sctp.addip_enable = 0;
1224 net->sctp.addip_noauth = 0;
1225 net->sctp.default_auto_asconf = 0;
1227 /* Enable PR-SCTP by default. */
1228 net->sctp.prsctp_enable = 1;
1230 /* Disable AUTH by default. */
1231 net->sctp.auth_enable = 0;
1233 /* Set SCOPE policy to enabled */
1234 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1236 /* Set the default rwnd update threshold */
1237 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1239 /* Initialize maximum autoclose timeout. */
1240 net->sctp.max_autoclose = INT_MAX / HZ;
1242 status = sctp_sysctl_net_register(net);
1243 if (status)
1244 goto err_sysctl_register;
1246 /* Allocate and initialise sctp mibs. */
1247 status = init_sctp_mibs(net);
1248 if (status)
1249 goto err_init_mibs;
1251 /* Initialize proc fs directory. */
1252 status = sctp_proc_init(net);
1253 if (status)
1254 goto err_init_proc;
1256 sctp_dbg_objcnt_init(net);
1258 /* Initialize the control inode/socket for handling OOTB packets. */
1259 if ((status = sctp_ctl_sock_init(net))) {
1260 pr_err("Failed to initialize the SCTP control sock\n");
1261 goto err_ctl_sock_init;
1264 /* Initialize the local address list. */
1265 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1266 spin_lock_init(&net->sctp.local_addr_lock);
1267 sctp_get_local_addr_list(net);
1269 /* Initialize the address event list */
1270 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1271 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1272 spin_lock_init(&net->sctp.addr_wq_lock);
1273 net->sctp.addr_wq_timer.expires = 0;
1274 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1275 (unsigned long)net);
1277 return 0;
1279 err_ctl_sock_init:
1280 sctp_dbg_objcnt_exit(net);
1281 sctp_proc_exit(net);
1282 err_init_proc:
1283 cleanup_sctp_mibs(net);
1284 err_init_mibs:
1285 sctp_sysctl_net_unregister(net);
1286 err_sysctl_register:
1287 return status;
1290 static void __net_exit sctp_net_exit(struct net *net)
1292 /* Free the local address list */
1293 sctp_free_addr_wq(net);
1294 sctp_free_local_addr_list(net);
1296 /* Free the control endpoint. */
1297 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1299 sctp_dbg_objcnt_exit(net);
1301 sctp_proc_exit(net);
1302 cleanup_sctp_mibs(net);
1303 sctp_sysctl_net_unregister(net);
1306 static struct pernet_operations sctp_net_ops = {
1307 .init = sctp_net_init,
1308 .exit = sctp_net_exit,
1311 /* Initialize the universe into something sensible. */
1312 static __init int sctp_init(void)
1314 int i;
1315 int status = -EINVAL;
1316 unsigned long goal;
1317 unsigned long limit;
1318 int max_share;
1319 int order;
1321 BUILD_BUG_ON(sizeof(struct sctp_ulpevent) >
1322 sizeof(((struct sk_buff *) 0)->cb));
1324 /* Allocate bind_bucket and chunk caches. */
1325 status = -ENOBUFS;
1326 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1327 sizeof(struct sctp_bind_bucket),
1328 0, SLAB_HWCACHE_ALIGN,
1329 NULL);
1330 if (!sctp_bucket_cachep)
1331 goto out;
1333 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1334 sizeof(struct sctp_chunk),
1335 0, SLAB_HWCACHE_ALIGN,
1336 NULL);
1337 if (!sctp_chunk_cachep)
1338 goto err_chunk_cachep;
1340 status = percpu_counter_init(&sctp_sockets_allocated, 0);
1341 if (status)
1342 goto err_percpu_counter_init;
1344 /* Implementation specific variables. */
1346 /* Initialize default stream count setup information. */
1347 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1348 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1350 /* Initialize handle used for association ids. */
1351 idr_init(&sctp_assocs_id);
1353 limit = nr_free_buffer_pages() / 8;
1354 limit = max(limit, 128UL);
1355 sysctl_sctp_mem[0] = limit / 4 * 3;
1356 sysctl_sctp_mem[1] = limit;
1357 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1359 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1360 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1361 max_share = min(4UL*1024*1024, limit);
1363 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1364 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1365 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1367 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1368 sysctl_sctp_wmem[1] = 16*1024;
1369 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1371 /* Size and allocate the association hash table.
1372 * The methodology is similar to that of the tcp hash tables.
1374 if (totalram_pages >= (128 * 1024))
1375 goal = totalram_pages >> (22 - PAGE_SHIFT);
1376 else
1377 goal = totalram_pages >> (24 - PAGE_SHIFT);
1379 for (order = 0; (1UL << order) < goal; order++)
1382 do {
1383 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1384 sizeof(struct sctp_hashbucket);
1385 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1386 continue;
1387 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1388 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1389 } while (!sctp_assoc_hashtable && --order > 0);
1390 if (!sctp_assoc_hashtable) {
1391 pr_err("Failed association hash alloc\n");
1392 status = -ENOMEM;
1393 goto err_ahash_alloc;
1395 for (i = 0; i < sctp_assoc_hashsize; i++) {
1396 rwlock_init(&sctp_assoc_hashtable[i].lock);
1397 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1400 /* Allocate and initialize the endpoint hash table. */
1401 sctp_ep_hashsize = 64;
1402 sctp_ep_hashtable =
1403 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1404 if (!sctp_ep_hashtable) {
1405 pr_err("Failed endpoint_hash alloc\n");
1406 status = -ENOMEM;
1407 goto err_ehash_alloc;
1409 for (i = 0; i < sctp_ep_hashsize; i++) {
1410 rwlock_init(&sctp_ep_hashtable[i].lock);
1411 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1414 /* Allocate and initialize the SCTP port hash table. */
1415 do {
1416 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1417 sizeof(struct sctp_bind_hashbucket);
1418 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1419 continue;
1420 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1421 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1422 } while (!sctp_port_hashtable && --order > 0);
1423 if (!sctp_port_hashtable) {
1424 pr_err("Failed bind hash alloc\n");
1425 status = -ENOMEM;
1426 goto err_bhash_alloc;
1428 for (i = 0; i < sctp_port_hashsize; i++) {
1429 spin_lock_init(&sctp_port_hashtable[i].lock);
1430 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1433 pr_info("Hash tables configured (established %d bind %d)\n",
1434 sctp_assoc_hashsize, sctp_port_hashsize);
1436 sctp_sysctl_register();
1438 INIT_LIST_HEAD(&sctp_address_families);
1439 sctp_v4_pf_init();
1440 sctp_v6_pf_init();
1442 status = sctp_v4_protosw_init();
1444 if (status)
1445 goto err_protosw_init;
1447 status = sctp_v6_protosw_init();
1448 if (status)
1449 goto err_v6_protosw_init;
1451 status = register_pernet_subsys(&sctp_net_ops);
1452 if (status)
1453 goto err_register_pernet_subsys;
1455 status = sctp_v4_add_protocol();
1456 if (status)
1457 goto err_add_protocol;
1459 /* Register SCTP with inet6 layer. */
1460 status = sctp_v6_add_protocol();
1461 if (status)
1462 goto err_v6_add_protocol;
1464 out:
1465 return status;
1466 err_v6_add_protocol:
1467 sctp_v4_del_protocol();
1468 err_add_protocol:
1469 unregister_pernet_subsys(&sctp_net_ops);
1470 err_register_pernet_subsys:
1471 sctp_v6_protosw_exit();
1472 err_v6_protosw_init:
1473 sctp_v4_protosw_exit();
1474 err_protosw_init:
1475 sctp_v4_pf_exit();
1476 sctp_v6_pf_exit();
1477 sctp_sysctl_unregister();
1478 free_pages((unsigned long)sctp_port_hashtable,
1479 get_order(sctp_port_hashsize *
1480 sizeof(struct sctp_bind_hashbucket)));
1481 err_bhash_alloc:
1482 kfree(sctp_ep_hashtable);
1483 err_ehash_alloc:
1484 free_pages((unsigned long)sctp_assoc_hashtable,
1485 get_order(sctp_assoc_hashsize *
1486 sizeof(struct sctp_hashbucket)));
1487 err_ahash_alloc:
1488 percpu_counter_destroy(&sctp_sockets_allocated);
1489 err_percpu_counter_init:
1490 kmem_cache_destroy(sctp_chunk_cachep);
1491 err_chunk_cachep:
1492 kmem_cache_destroy(sctp_bucket_cachep);
1493 goto out;
1496 /* Exit handler for the SCTP protocol. */
1497 static __exit void sctp_exit(void)
1499 /* BUG. This should probably do something useful like clean
1500 * up all the remaining associations and all that memory.
1503 /* Unregister with inet6/inet layers. */
1504 sctp_v6_del_protocol();
1505 sctp_v4_del_protocol();
1507 unregister_pernet_subsys(&sctp_net_ops);
1509 /* Free protosw registrations */
1510 sctp_v6_protosw_exit();
1511 sctp_v4_protosw_exit();
1513 /* Unregister with socket layer. */
1514 sctp_v6_pf_exit();
1515 sctp_v4_pf_exit();
1517 sctp_sysctl_unregister();
1519 free_pages((unsigned long)sctp_assoc_hashtable,
1520 get_order(sctp_assoc_hashsize *
1521 sizeof(struct sctp_hashbucket)));
1522 kfree(sctp_ep_hashtable);
1523 free_pages((unsigned long)sctp_port_hashtable,
1524 get_order(sctp_port_hashsize *
1525 sizeof(struct sctp_bind_hashbucket)));
1527 percpu_counter_destroy(&sctp_sockets_allocated);
1529 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1531 kmem_cache_destroy(sctp_chunk_cachep);
1532 kmem_cache_destroy(sctp_bucket_cachep);
1535 module_init(sctp_init);
1536 module_exit(sctp_exit);
1539 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1541 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1542 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1543 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1544 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1545 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1546 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1547 MODULE_LICENSE("GPL");