Merge tag 'gpio-v4.9-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux...
[linux/fpc-iii.git] / net / sctp / protocol.c
blob7b523e3f551f1761d891b606f293b2b721a87ca8
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 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
65 /* Global data structures. */
66 struct sctp_globals sctp_globals __read_mostly;
68 struct idr sctp_assocs_id;
69 DEFINE_SPINLOCK(sctp_assocs_id_lock);
71 static struct sctp_pf *sctp_pf_inet6_specific;
72 static struct sctp_pf *sctp_pf_inet_specific;
73 static struct sctp_af *sctp_af_v4_specific;
74 static struct sctp_af *sctp_af_v6_specific;
76 struct kmem_cache *sctp_chunk_cachep __read_mostly;
77 struct kmem_cache *sctp_bucket_cachep __read_mostly;
79 long sysctl_sctp_mem[3];
80 int sysctl_sctp_rmem[3];
81 int sysctl_sctp_wmem[3];
83 /* Set up the proc fs entry for the SCTP protocol. */
84 static int __net_init sctp_proc_init(struct net *net)
86 #ifdef CONFIG_PROC_FS
87 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
88 if (!net->sctp.proc_net_sctp)
89 goto out_proc_net_sctp;
90 if (sctp_snmp_proc_init(net))
91 goto out_snmp_proc_init;
92 if (sctp_eps_proc_init(net))
93 goto out_eps_proc_init;
94 if (sctp_assocs_proc_init(net))
95 goto out_assocs_proc_init;
96 if (sctp_remaddr_proc_init(net))
97 goto out_remaddr_proc_init;
99 return 0;
101 out_remaddr_proc_init:
102 sctp_assocs_proc_exit(net);
103 out_assocs_proc_init:
104 sctp_eps_proc_exit(net);
105 out_eps_proc_init:
106 sctp_snmp_proc_exit(net);
107 out_snmp_proc_init:
108 remove_proc_entry("sctp", net->proc_net);
109 net->sctp.proc_net_sctp = NULL;
110 out_proc_net_sctp:
111 return -ENOMEM;
112 #endif /* CONFIG_PROC_FS */
113 return 0;
116 /* Clean up the proc fs entry for the SCTP protocol.
117 * Note: Do not make this __exit as it is used in the init error
118 * path.
120 static void sctp_proc_exit(struct net *net)
122 #ifdef CONFIG_PROC_FS
123 sctp_snmp_proc_exit(net);
124 sctp_eps_proc_exit(net);
125 sctp_assocs_proc_exit(net);
126 sctp_remaddr_proc_exit(net);
128 remove_proc_entry("sctp", net->proc_net);
129 net->sctp.proc_net_sctp = NULL;
130 #endif
133 /* Private helper to extract ipv4 address and stash them in
134 * the protocol structure.
136 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
137 struct net_device *dev)
139 struct in_device *in_dev;
140 struct in_ifaddr *ifa;
141 struct sctp_sockaddr_entry *addr;
143 rcu_read_lock();
144 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
145 rcu_read_unlock();
146 return;
149 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
150 /* Add the address to the local list. */
151 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
152 if (addr) {
153 addr->a.v4.sin_family = AF_INET;
154 addr->a.v4.sin_port = 0;
155 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
156 addr->valid = 1;
157 INIT_LIST_HEAD(&addr->list);
158 list_add_tail(&addr->list, addrlist);
162 rcu_read_unlock();
165 /* Extract our IP addresses from the system and stash them in the
166 * protocol structure.
168 static void sctp_get_local_addr_list(struct net *net)
170 struct net_device *dev;
171 struct list_head *pos;
172 struct sctp_af *af;
174 rcu_read_lock();
175 for_each_netdev_rcu(net, dev) {
176 list_for_each(pos, &sctp_address_families) {
177 af = list_entry(pos, struct sctp_af, list);
178 af->copy_addrlist(&net->sctp.local_addr_list, dev);
181 rcu_read_unlock();
184 /* Free the existing local addresses. */
185 static void sctp_free_local_addr_list(struct net *net)
187 struct sctp_sockaddr_entry *addr;
188 struct list_head *pos, *temp;
190 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
191 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
192 list_del(pos);
193 kfree(addr);
197 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
198 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
199 sctp_scope_t scope, gfp_t gfp, int copy_flags)
201 struct sctp_sockaddr_entry *addr;
202 int error = 0;
204 rcu_read_lock();
205 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
206 if (!addr->valid)
207 continue;
208 if (sctp_in_scope(net, &addr->a, scope)) {
209 /* Now that the address is in scope, check to see if
210 * the address type is really supported by the local
211 * sock as well as the remote peer.
213 if ((((AF_INET == addr->a.sa.sa_family) &&
214 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
215 (((AF_INET6 == addr->a.sa.sa_family) &&
216 (copy_flags & SCTP_ADDR6_ALLOWED) &&
217 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
218 error = sctp_add_bind_addr(bp, &addr->a,
219 sizeof(addr->a),
220 SCTP_ADDR_SRC, GFP_ATOMIC);
221 if (error)
222 goto end_copy;
227 end_copy:
228 rcu_read_unlock();
229 return error;
232 /* Initialize a sctp_addr from in incoming skb. */
233 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
234 int is_saddr)
236 void *from;
237 __be16 *port;
238 struct sctphdr *sh;
240 port = &addr->v4.sin_port;
241 addr->v4.sin_family = AF_INET;
243 /* Always called on head skb, so this is safe */
244 sh = sctp_hdr(skb);
245 if (is_saddr) {
246 *port = sh->source;
247 from = &ip_hdr(skb)->saddr;
248 } else {
249 *port = sh->dest;
250 from = &ip_hdr(skb)->daddr;
252 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
255 /* Initialize an sctp_addr from a socket. */
256 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
258 addr->v4.sin_family = AF_INET;
259 addr->v4.sin_port = 0;
260 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
263 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
264 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
266 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
269 /* Initialize sk->sk_daddr from sctp_addr. */
270 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
272 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
275 /* Initialize a sctp_addr from an address parameter. */
276 static void sctp_v4_from_addr_param(union sctp_addr *addr,
277 union sctp_addr_param *param,
278 __be16 port, int iif)
280 addr->v4.sin_family = AF_INET;
281 addr->v4.sin_port = port;
282 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
285 /* Initialize an address parameter from a sctp_addr and return the length
286 * of the address parameter.
288 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
289 union sctp_addr_param *param)
291 int length = sizeof(sctp_ipv4addr_param_t);
293 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
294 param->v4.param_hdr.length = htons(length);
295 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
297 return length;
300 /* Initialize a sctp_addr from a dst_entry. */
301 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
302 __be16 port)
304 saddr->v4.sin_family = AF_INET;
305 saddr->v4.sin_port = port;
306 saddr->v4.sin_addr.s_addr = fl4->saddr;
309 /* Compare two addresses exactly. */
310 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
311 const union sctp_addr *addr2)
313 if (addr1->sa.sa_family != addr2->sa.sa_family)
314 return 0;
315 if (addr1->v4.sin_port != addr2->v4.sin_port)
316 return 0;
317 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
318 return 0;
320 return 1;
323 /* Initialize addr struct to INADDR_ANY. */
324 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
326 addr->v4.sin_family = AF_INET;
327 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
328 addr->v4.sin_port = port;
331 /* Is this a wildcard address? */
332 static int sctp_v4_is_any(const union sctp_addr *addr)
334 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
337 /* This function checks if the address is a valid address to be used for
338 * SCTP binding.
340 * Output:
341 * Return 0 - If the address is a non-unicast or an illegal address.
342 * Return 1 - If the address is a unicast.
344 static int sctp_v4_addr_valid(union sctp_addr *addr,
345 struct sctp_sock *sp,
346 const struct sk_buff *skb)
348 /* IPv4 addresses not allowed */
349 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
350 return 0;
352 /* Is this a non-unicast address or a unusable SCTP address? */
353 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
354 return 0;
356 /* Is this a broadcast address? */
357 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
358 return 0;
360 return 1;
363 /* Should this be available for binding? */
364 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
366 struct net *net = sock_net(&sp->inet.sk);
367 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
370 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
371 ret != RTN_LOCAL &&
372 !sp->inet.freebind &&
373 !net->ipv4.sysctl_ip_nonlocal_bind)
374 return 0;
376 if (ipv6_only_sock(sctp_opt2sk(sp)))
377 return 0;
379 return 1;
382 /* Checking the loopback, private and other address scopes as defined in
383 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
384 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
386 * Level 0 - unusable SCTP addresses
387 * Level 1 - loopback address
388 * Level 2 - link-local addresses
389 * Level 3 - private addresses.
390 * Level 4 - global addresses
391 * For INIT and INIT-ACK address list, let L be the level of
392 * of requested destination address, sender and receiver
393 * SHOULD include all of its addresses with level greater
394 * than or equal to L.
396 * IPv4 scoping can be controlled through sysctl option
397 * net.sctp.addr_scope_policy
399 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
401 sctp_scope_t retval;
403 /* Check for unusable SCTP addresses. */
404 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
405 retval = SCTP_SCOPE_UNUSABLE;
406 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
407 retval = SCTP_SCOPE_LOOPBACK;
408 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
409 retval = SCTP_SCOPE_LINK;
410 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
411 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
412 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
413 retval = SCTP_SCOPE_PRIVATE;
414 } else {
415 retval = SCTP_SCOPE_GLOBAL;
418 return retval;
421 /* Returns a valid dst cache entry for the given source and destination ip
422 * addresses. If an association is passed, trys to get a dst entry with a
423 * source address that matches an address in the bind address list.
425 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
426 struct flowi *fl, struct sock *sk)
428 struct sctp_association *asoc = t->asoc;
429 struct rtable *rt;
430 struct flowi4 *fl4 = &fl->u.ip4;
431 struct sctp_bind_addr *bp;
432 struct sctp_sockaddr_entry *laddr;
433 struct dst_entry *dst = NULL;
434 union sctp_addr *daddr = &t->ipaddr;
435 union sctp_addr dst_saddr;
437 memset(fl4, 0x0, sizeof(struct flowi4));
438 fl4->daddr = daddr->v4.sin_addr.s_addr;
439 fl4->fl4_dport = daddr->v4.sin_port;
440 fl4->flowi4_proto = IPPROTO_SCTP;
441 if (asoc) {
442 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
443 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
444 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
446 if (saddr) {
447 fl4->saddr = saddr->v4.sin_addr.s_addr;
448 fl4->fl4_sport = saddr->v4.sin_port;
451 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
452 &fl4->saddr);
454 rt = ip_route_output_key(sock_net(sk), fl4);
455 if (!IS_ERR(rt))
456 dst = &rt->dst;
458 /* If there is no association or if a source address is passed, no
459 * more validation is required.
461 if (!asoc || saddr)
462 goto out;
464 bp = &asoc->base.bind_addr;
466 if (dst) {
467 /* Walk through the bind address list and look for a bind
468 * address that matches the source address of the returned dst.
470 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
471 rcu_read_lock();
472 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
473 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
474 (laddr->state != SCTP_ADDR_SRC &&
475 !asoc->src_out_of_asoc_ok))
476 continue;
477 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
478 goto out_unlock;
480 rcu_read_unlock();
482 /* None of the bound addresses match the source address of the
483 * dst. So release it.
485 dst_release(dst);
486 dst = NULL;
489 /* Walk through the bind address list and try to get a dst that
490 * matches a bind address as the source address.
492 rcu_read_lock();
493 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
494 struct net_device *odev;
496 if (!laddr->valid)
497 continue;
498 if (laddr->state != SCTP_ADDR_SRC ||
499 AF_INET != laddr->a.sa.sa_family)
500 continue;
502 fl4->fl4_sport = laddr->a.v4.sin_port;
503 flowi4_update_output(fl4,
504 asoc->base.sk->sk_bound_dev_if,
505 RT_CONN_FLAGS(asoc->base.sk),
506 daddr->v4.sin_addr.s_addr,
507 laddr->a.v4.sin_addr.s_addr);
509 rt = ip_route_output_key(sock_net(sk), fl4);
510 if (IS_ERR(rt))
511 continue;
513 if (!dst)
514 dst = &rt->dst;
516 /* Ensure the src address belongs to the output
517 * interface.
519 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
520 false);
521 if (!odev || odev->ifindex != fl4->flowi4_oif) {
522 if (&rt->dst != dst)
523 dst_release(&rt->dst);
524 continue;
527 if (dst != &rt->dst)
528 dst_release(dst);
529 dst = &rt->dst;
530 break;
533 out_unlock:
534 rcu_read_unlock();
535 out:
536 t->dst = dst;
537 if (dst)
538 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
539 &fl4->daddr, &fl4->saddr);
540 else
541 pr_debug("no route\n");
544 /* For v4, the source address is cached in the route entry(dst). So no need
545 * to cache it separately and hence this is an empty routine.
547 static void sctp_v4_get_saddr(struct sctp_sock *sk,
548 struct sctp_transport *t,
549 struct flowi *fl)
551 union sctp_addr *saddr = &t->saddr;
552 struct rtable *rt = (struct rtable *)t->dst;
554 if (rt) {
555 saddr->v4.sin_family = AF_INET;
556 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
560 /* What interface did this skb arrive on? */
561 static int sctp_v4_skb_iif(const struct sk_buff *skb)
563 return inet_iif(skb);
566 /* Was this packet marked by Explicit Congestion Notification? */
567 static int sctp_v4_is_ce(const struct sk_buff *skb)
569 return INET_ECN_is_ce(ip_hdr(skb)->tos);
572 /* Create and initialize a new sk for the socket returned by accept(). */
573 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
574 struct sctp_association *asoc)
576 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
577 sk->sk_prot, 0);
578 struct inet_sock *newinet;
580 if (!newsk)
581 goto out;
583 sock_init_data(NULL, newsk);
585 sctp_copy_sock(newsk, sk, asoc);
586 sock_reset_flag(newsk, SOCK_ZAPPED);
588 newinet = inet_sk(newsk);
590 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
592 sk_refcnt_debug_inc(newsk);
594 if (newsk->sk_prot->init(newsk)) {
595 sk_common_release(newsk);
596 newsk = NULL;
599 out:
600 return newsk;
603 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
605 /* No address mapping for V4 sockets */
606 return sizeof(struct sockaddr_in);
609 /* Dump the v4 addr to the seq file. */
610 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
612 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
615 static void sctp_v4_ecn_capable(struct sock *sk)
617 INET_ECN_xmit(sk);
620 static void sctp_addr_wq_timeout_handler(unsigned long arg)
622 struct net *net = (struct net *)arg;
623 struct sctp_sockaddr_entry *addrw, *temp;
624 struct sctp_sock *sp;
626 spin_lock_bh(&net->sctp.addr_wq_lock);
628 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
629 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
630 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
631 addrw->state, addrw);
633 #if IS_ENABLED(CONFIG_IPV6)
634 /* Now we send an ASCONF for each association */
635 /* Note. we currently don't handle link local IPv6 addressees */
636 if (addrw->a.sa.sa_family == AF_INET6) {
637 struct in6_addr *in6;
639 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
640 IPV6_ADDR_LINKLOCAL)
641 goto free_next;
643 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
644 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
645 addrw->state == SCTP_ADDR_NEW) {
646 unsigned long timeo_val;
648 pr_debug("%s: this is on DAD, trying %d sec "
649 "later\n", __func__,
650 SCTP_ADDRESS_TICK_DELAY);
652 timeo_val = jiffies;
653 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
654 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
655 break;
658 #endif
659 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
660 struct sock *sk;
662 sk = sctp_opt2sk(sp);
663 /* ignore bound-specific endpoints */
664 if (!sctp_is_ep_boundall(sk))
665 continue;
666 bh_lock_sock(sk);
667 if (sctp_asconf_mgmt(sp, addrw) < 0)
668 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
669 bh_unlock_sock(sk);
671 #if IS_ENABLED(CONFIG_IPV6)
672 free_next:
673 #endif
674 list_del(&addrw->list);
675 kfree(addrw);
677 spin_unlock_bh(&net->sctp.addr_wq_lock);
680 static void sctp_free_addr_wq(struct net *net)
682 struct sctp_sockaddr_entry *addrw;
683 struct sctp_sockaddr_entry *temp;
685 spin_lock_bh(&net->sctp.addr_wq_lock);
686 del_timer(&net->sctp.addr_wq_timer);
687 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
688 list_del(&addrw->list);
689 kfree(addrw);
691 spin_unlock_bh(&net->sctp.addr_wq_lock);
694 /* lookup the entry for the same address in the addr_waitq
695 * sctp_addr_wq MUST be locked
697 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
698 struct sctp_sockaddr_entry *addr)
700 struct sctp_sockaddr_entry *addrw;
702 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
703 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
704 continue;
705 if (addrw->a.sa.sa_family == AF_INET) {
706 if (addrw->a.v4.sin_addr.s_addr ==
707 addr->a.v4.sin_addr.s_addr)
708 return addrw;
709 } else if (addrw->a.sa.sa_family == AF_INET6) {
710 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
711 &addr->a.v6.sin6_addr))
712 return addrw;
715 return NULL;
718 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
720 struct sctp_sockaddr_entry *addrw;
721 unsigned long timeo_val;
723 /* first, we check if an opposite message already exist in the queue.
724 * If we found such message, it is removed.
725 * This operation is a bit stupid, but the DHCP client attaches the
726 * new address after a couple of addition and deletion of that address
729 spin_lock_bh(&net->sctp.addr_wq_lock);
730 /* Offsets existing events in addr_wq */
731 addrw = sctp_addr_wq_lookup(net, addr);
732 if (addrw) {
733 if (addrw->state != cmd) {
734 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
735 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
736 &net->sctp.addr_waitq);
738 list_del(&addrw->list);
739 kfree(addrw);
741 spin_unlock_bh(&net->sctp.addr_wq_lock);
742 return;
745 /* OK, we have to add the new address to the wait queue */
746 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
747 if (addrw == NULL) {
748 spin_unlock_bh(&net->sctp.addr_wq_lock);
749 return;
751 addrw->state = cmd;
752 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
754 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
755 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
757 if (!timer_pending(&net->sctp.addr_wq_timer)) {
758 timeo_val = jiffies;
759 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
760 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
762 spin_unlock_bh(&net->sctp.addr_wq_lock);
765 /* Event handler for inet address addition/deletion events.
766 * The sctp_local_addr_list needs to be protocted by a spin lock since
767 * multiple notifiers (say IPv4 and IPv6) may be running at the same
768 * time and thus corrupt the list.
769 * The reader side is protected with RCU.
771 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
772 void *ptr)
774 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
775 struct sctp_sockaddr_entry *addr = NULL;
776 struct sctp_sockaddr_entry *temp;
777 struct net *net = dev_net(ifa->ifa_dev->dev);
778 int found = 0;
780 switch (ev) {
781 case NETDEV_UP:
782 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
783 if (addr) {
784 addr->a.v4.sin_family = AF_INET;
785 addr->a.v4.sin_port = 0;
786 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
787 addr->valid = 1;
788 spin_lock_bh(&net->sctp.local_addr_lock);
789 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
790 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
791 spin_unlock_bh(&net->sctp.local_addr_lock);
793 break;
794 case NETDEV_DOWN:
795 spin_lock_bh(&net->sctp.local_addr_lock);
796 list_for_each_entry_safe(addr, temp,
797 &net->sctp.local_addr_list, list) {
798 if (addr->a.sa.sa_family == AF_INET &&
799 addr->a.v4.sin_addr.s_addr ==
800 ifa->ifa_local) {
801 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
802 found = 1;
803 addr->valid = 0;
804 list_del_rcu(&addr->list);
805 break;
808 spin_unlock_bh(&net->sctp.local_addr_lock);
809 if (found)
810 kfree_rcu(addr, rcu);
811 break;
814 return NOTIFY_DONE;
818 * Initialize the control inode/socket with a control endpoint data
819 * structure. This endpoint is reserved exclusively for the OOTB processing.
821 static int sctp_ctl_sock_init(struct net *net)
823 int err;
824 sa_family_t family = PF_INET;
826 if (sctp_get_pf_specific(PF_INET6))
827 family = PF_INET6;
829 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
830 SOCK_SEQPACKET, IPPROTO_SCTP, net);
832 /* If IPv6 socket could not be created, try the IPv4 socket */
833 if (err < 0 && family == PF_INET6)
834 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
835 SOCK_SEQPACKET, IPPROTO_SCTP,
836 net);
838 if (err < 0) {
839 pr_err("Failed to create the SCTP control socket\n");
840 return err;
842 return 0;
845 /* Register address family specific functions. */
846 int sctp_register_af(struct sctp_af *af)
848 switch (af->sa_family) {
849 case AF_INET:
850 if (sctp_af_v4_specific)
851 return 0;
852 sctp_af_v4_specific = af;
853 break;
854 case AF_INET6:
855 if (sctp_af_v6_specific)
856 return 0;
857 sctp_af_v6_specific = af;
858 break;
859 default:
860 return 0;
863 INIT_LIST_HEAD(&af->list);
864 list_add_tail(&af->list, &sctp_address_families);
865 return 1;
868 /* Get the table of functions for manipulating a particular address
869 * family.
871 struct sctp_af *sctp_get_af_specific(sa_family_t family)
873 switch (family) {
874 case AF_INET:
875 return sctp_af_v4_specific;
876 case AF_INET6:
877 return sctp_af_v6_specific;
878 default:
879 return NULL;
883 /* Common code to initialize a AF_INET msg_name. */
884 static void sctp_inet_msgname(char *msgname, int *addr_len)
886 struct sockaddr_in *sin;
888 sin = (struct sockaddr_in *)msgname;
889 *addr_len = sizeof(struct sockaddr_in);
890 sin->sin_family = AF_INET;
891 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
894 /* Copy the primary address of the peer primary address as the msg_name. */
895 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
896 int *addr_len)
898 struct sockaddr_in *sin, *sinfrom;
900 if (msgname) {
901 struct sctp_association *asoc;
903 asoc = event->asoc;
904 sctp_inet_msgname(msgname, addr_len);
905 sin = (struct sockaddr_in *)msgname;
906 sinfrom = &asoc->peer.primary_addr.v4;
907 sin->sin_port = htons(asoc->peer.port);
908 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
912 /* Initialize and copy out a msgname from an inbound skb. */
913 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
915 if (msgname) {
916 struct sctphdr *sh = sctp_hdr(skb);
917 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
919 sctp_inet_msgname(msgname, len);
920 sin->sin_port = sh->source;
921 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
925 /* Do we support this AF? */
926 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
928 /* PF_INET only supports AF_INET addresses. */
929 return AF_INET == family;
932 /* Address matching with wildcards allowed. */
933 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
934 const union sctp_addr *addr2,
935 struct sctp_sock *opt)
937 /* PF_INET only supports AF_INET addresses. */
938 if (addr1->sa.sa_family != addr2->sa.sa_family)
939 return 0;
940 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
941 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
942 return 1;
943 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
944 return 1;
946 return 0;
949 /* Verify that provided sockaddr looks bindable. Common verification has
950 * already been taken care of.
952 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
954 return sctp_v4_available(addr, opt);
957 /* Verify that sockaddr looks sendable. Common verification has already
958 * been taken care of.
960 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
962 return 1;
965 /* Fill in Supported Address Type information for INIT and INIT-ACK
966 * chunks. Returns number of addresses supported.
968 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
969 __be16 *types)
971 types[0] = SCTP_PARAM_IPV4_ADDRESS;
972 return 1;
975 /* Wrapper routine that calls the ip transmit routine. */
976 static inline int sctp_v4_xmit(struct sk_buff *skb,
977 struct sctp_transport *transport)
979 struct inet_sock *inet = inet_sk(skb->sk);
981 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
982 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
984 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
985 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
987 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
989 return ip_queue_xmit(&inet->sk, skb, &transport->fl);
992 static struct sctp_af sctp_af_inet;
994 static struct sctp_pf sctp_pf_inet = {
995 .event_msgname = sctp_inet_event_msgname,
996 .skb_msgname = sctp_inet_skb_msgname,
997 .af_supported = sctp_inet_af_supported,
998 .cmp_addr = sctp_inet_cmp_addr,
999 .bind_verify = sctp_inet_bind_verify,
1000 .send_verify = sctp_inet_send_verify,
1001 .supported_addrs = sctp_inet_supported_addrs,
1002 .create_accept_sk = sctp_v4_create_accept_sk,
1003 .addr_to_user = sctp_v4_addr_to_user,
1004 .to_sk_saddr = sctp_v4_to_sk_saddr,
1005 .to_sk_daddr = sctp_v4_to_sk_daddr,
1006 .af = &sctp_af_inet
1009 /* Notifier for inetaddr addition/deletion events. */
1010 static struct notifier_block sctp_inetaddr_notifier = {
1011 .notifier_call = sctp_inetaddr_event,
1014 /* Socket operations. */
1015 static const struct proto_ops inet_seqpacket_ops = {
1016 .family = PF_INET,
1017 .owner = THIS_MODULE,
1018 .release = inet_release, /* Needs to be wrapped... */
1019 .bind = inet_bind,
1020 .connect = inet_dgram_connect,
1021 .socketpair = sock_no_socketpair,
1022 .accept = inet_accept,
1023 .getname = inet_getname, /* Semantics are different. */
1024 .poll = sctp_poll,
1025 .ioctl = inet_ioctl,
1026 .listen = sctp_inet_listen,
1027 .shutdown = inet_shutdown, /* Looks harmless. */
1028 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1029 .getsockopt = sock_common_getsockopt,
1030 .sendmsg = inet_sendmsg,
1031 .recvmsg = inet_recvmsg,
1032 .mmap = sock_no_mmap,
1033 .sendpage = sock_no_sendpage,
1034 #ifdef CONFIG_COMPAT
1035 .compat_setsockopt = compat_sock_common_setsockopt,
1036 .compat_getsockopt = compat_sock_common_getsockopt,
1037 #endif
1040 /* Registration with AF_INET family. */
1041 static struct inet_protosw sctp_seqpacket_protosw = {
1042 .type = SOCK_SEQPACKET,
1043 .protocol = IPPROTO_SCTP,
1044 .prot = &sctp_prot,
1045 .ops = &inet_seqpacket_ops,
1046 .flags = SCTP_PROTOSW_FLAG
1048 static struct inet_protosw sctp_stream_protosw = {
1049 .type = SOCK_STREAM,
1050 .protocol = IPPROTO_SCTP,
1051 .prot = &sctp_prot,
1052 .ops = &inet_seqpacket_ops,
1053 .flags = SCTP_PROTOSW_FLAG
1056 /* Register with IP layer. */
1057 static const struct net_protocol sctp_protocol = {
1058 .handler = sctp_rcv,
1059 .err_handler = sctp_v4_err,
1060 .no_policy = 1,
1061 .netns_ok = 1,
1062 .icmp_strict_tag_validation = 1,
1065 /* IPv4 address related functions. */
1066 static struct sctp_af sctp_af_inet = {
1067 .sa_family = AF_INET,
1068 .sctp_xmit = sctp_v4_xmit,
1069 .setsockopt = ip_setsockopt,
1070 .getsockopt = ip_getsockopt,
1071 .get_dst = sctp_v4_get_dst,
1072 .get_saddr = sctp_v4_get_saddr,
1073 .copy_addrlist = sctp_v4_copy_addrlist,
1074 .from_skb = sctp_v4_from_skb,
1075 .from_sk = sctp_v4_from_sk,
1076 .from_addr_param = sctp_v4_from_addr_param,
1077 .to_addr_param = sctp_v4_to_addr_param,
1078 .cmp_addr = sctp_v4_cmp_addr,
1079 .addr_valid = sctp_v4_addr_valid,
1080 .inaddr_any = sctp_v4_inaddr_any,
1081 .is_any = sctp_v4_is_any,
1082 .available = sctp_v4_available,
1083 .scope = sctp_v4_scope,
1084 .skb_iif = sctp_v4_skb_iif,
1085 .is_ce = sctp_v4_is_ce,
1086 .seq_dump_addr = sctp_v4_seq_dump_addr,
1087 .ecn_capable = sctp_v4_ecn_capable,
1088 .net_header_len = sizeof(struct iphdr),
1089 .sockaddr_len = sizeof(struct sockaddr_in),
1090 #ifdef CONFIG_COMPAT
1091 .compat_setsockopt = compat_ip_setsockopt,
1092 .compat_getsockopt = compat_ip_getsockopt,
1093 #endif
1096 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1098 switch (family) {
1099 case PF_INET:
1100 return sctp_pf_inet_specific;
1101 case PF_INET6:
1102 return sctp_pf_inet6_specific;
1103 default:
1104 return NULL;
1108 /* Register the PF specific function table. */
1109 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1111 switch (family) {
1112 case PF_INET:
1113 if (sctp_pf_inet_specific)
1114 return 0;
1115 sctp_pf_inet_specific = pf;
1116 break;
1117 case PF_INET6:
1118 if (sctp_pf_inet6_specific)
1119 return 0;
1120 sctp_pf_inet6_specific = pf;
1121 break;
1122 default:
1123 return 0;
1125 return 1;
1128 static inline int init_sctp_mibs(struct net *net)
1130 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1131 if (!net->sctp.sctp_statistics)
1132 return -ENOMEM;
1133 return 0;
1136 static inline void cleanup_sctp_mibs(struct net *net)
1138 free_percpu(net->sctp.sctp_statistics);
1141 static void sctp_v4_pf_init(void)
1143 /* Initialize the SCTP specific PF functions. */
1144 sctp_register_pf(&sctp_pf_inet, PF_INET);
1145 sctp_register_af(&sctp_af_inet);
1148 static void sctp_v4_pf_exit(void)
1150 list_del(&sctp_af_inet.list);
1153 static int sctp_v4_protosw_init(void)
1155 int rc;
1157 rc = proto_register(&sctp_prot, 1);
1158 if (rc)
1159 return rc;
1161 /* Register SCTP(UDP and TCP style) with socket layer. */
1162 inet_register_protosw(&sctp_seqpacket_protosw);
1163 inet_register_protosw(&sctp_stream_protosw);
1165 return 0;
1168 static void sctp_v4_protosw_exit(void)
1170 inet_unregister_protosw(&sctp_stream_protosw);
1171 inet_unregister_protosw(&sctp_seqpacket_protosw);
1172 proto_unregister(&sctp_prot);
1175 static int sctp_v4_add_protocol(void)
1177 /* Register notifier for inet address additions/deletions. */
1178 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1180 /* Register SCTP with inet layer. */
1181 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1182 return -EAGAIN;
1184 return 0;
1187 static void sctp_v4_del_protocol(void)
1189 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1190 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1193 static int __net_init sctp_defaults_init(struct net *net)
1195 int status;
1198 * 14. Suggested SCTP Protocol Parameter Values
1200 /* The following protocol parameters are RECOMMENDED: */
1201 /* RTO.Initial - 3 seconds */
1202 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1203 /* RTO.Min - 1 second */
1204 net->sctp.rto_min = SCTP_RTO_MIN;
1205 /* RTO.Max - 60 seconds */
1206 net->sctp.rto_max = SCTP_RTO_MAX;
1207 /* RTO.Alpha - 1/8 */
1208 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1209 /* RTO.Beta - 1/4 */
1210 net->sctp.rto_beta = SCTP_RTO_BETA;
1212 /* Valid.Cookie.Life - 60 seconds */
1213 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1215 /* Whether Cookie Preservative is enabled(1) or not(0) */
1216 net->sctp.cookie_preserve_enable = 1;
1218 /* Default sctp sockets to use md5 as their hmac alg */
1219 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1220 net->sctp.sctp_hmac_alg = "md5";
1221 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1222 net->sctp.sctp_hmac_alg = "sha1";
1223 #else
1224 net->sctp.sctp_hmac_alg = NULL;
1225 #endif
1227 /* Max.Burst - 4 */
1228 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1230 /* Enable pf state by default */
1231 net->sctp.pf_enable = 1;
1233 /* Association.Max.Retrans - 10 attempts
1234 * Path.Max.Retrans - 5 attempts (per destination address)
1235 * Max.Init.Retransmits - 8 attempts
1237 net->sctp.max_retrans_association = 10;
1238 net->sctp.max_retrans_path = 5;
1239 net->sctp.max_retrans_init = 8;
1241 /* Sendbuffer growth - do per-socket accounting */
1242 net->sctp.sndbuf_policy = 0;
1244 /* Rcvbuffer growth - do per-socket accounting */
1245 net->sctp.rcvbuf_policy = 0;
1247 /* HB.interval - 30 seconds */
1248 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1250 /* delayed SACK timeout */
1251 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1253 /* Disable ADDIP by default. */
1254 net->sctp.addip_enable = 0;
1255 net->sctp.addip_noauth = 0;
1256 net->sctp.default_auto_asconf = 0;
1258 /* Enable PR-SCTP by default. */
1259 net->sctp.prsctp_enable = 1;
1261 /* Disable AUTH by default. */
1262 net->sctp.auth_enable = 0;
1264 /* Set SCOPE policy to enabled */
1265 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1267 /* Set the default rwnd update threshold */
1268 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1270 /* Initialize maximum autoclose timeout. */
1271 net->sctp.max_autoclose = INT_MAX / HZ;
1273 status = sctp_sysctl_net_register(net);
1274 if (status)
1275 goto err_sysctl_register;
1277 /* Allocate and initialise sctp mibs. */
1278 status = init_sctp_mibs(net);
1279 if (status)
1280 goto err_init_mibs;
1282 /* Initialize proc fs directory. */
1283 status = sctp_proc_init(net);
1284 if (status)
1285 goto err_init_proc;
1287 sctp_dbg_objcnt_init(net);
1289 /* Initialize the local address list. */
1290 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1291 spin_lock_init(&net->sctp.local_addr_lock);
1292 sctp_get_local_addr_list(net);
1294 /* Initialize the address event list */
1295 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1296 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1297 spin_lock_init(&net->sctp.addr_wq_lock);
1298 net->sctp.addr_wq_timer.expires = 0;
1299 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1300 (unsigned long)net);
1302 return 0;
1304 err_init_proc:
1305 cleanup_sctp_mibs(net);
1306 err_init_mibs:
1307 sctp_sysctl_net_unregister(net);
1308 err_sysctl_register:
1309 return status;
1312 static void __net_exit sctp_defaults_exit(struct net *net)
1314 /* Free the local address list */
1315 sctp_free_addr_wq(net);
1316 sctp_free_local_addr_list(net);
1318 sctp_dbg_objcnt_exit(net);
1320 sctp_proc_exit(net);
1321 cleanup_sctp_mibs(net);
1322 sctp_sysctl_net_unregister(net);
1325 static struct pernet_operations sctp_defaults_ops = {
1326 .init = sctp_defaults_init,
1327 .exit = sctp_defaults_exit,
1330 static int __net_init sctp_ctrlsock_init(struct net *net)
1332 int status;
1334 /* Initialize the control inode/socket for handling OOTB packets. */
1335 status = sctp_ctl_sock_init(net);
1336 if (status)
1337 pr_err("Failed to initialize the SCTP control sock\n");
1339 return status;
1342 static void __net_init sctp_ctrlsock_exit(struct net *net)
1344 /* Free the control endpoint. */
1345 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1348 static struct pernet_operations sctp_ctrlsock_ops = {
1349 .init = sctp_ctrlsock_init,
1350 .exit = sctp_ctrlsock_exit,
1353 /* Initialize the universe into something sensible. */
1354 static __init int sctp_init(void)
1356 int i;
1357 int status = -EINVAL;
1358 unsigned long goal;
1359 unsigned long limit;
1360 int max_share;
1361 int order;
1362 int num_entries;
1363 int max_entry_order;
1365 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1367 /* Allocate bind_bucket and chunk caches. */
1368 status = -ENOBUFS;
1369 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1370 sizeof(struct sctp_bind_bucket),
1371 0, SLAB_HWCACHE_ALIGN,
1372 NULL);
1373 if (!sctp_bucket_cachep)
1374 goto out;
1376 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1377 sizeof(struct sctp_chunk),
1378 0, SLAB_HWCACHE_ALIGN,
1379 NULL);
1380 if (!sctp_chunk_cachep)
1381 goto err_chunk_cachep;
1383 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1384 if (status)
1385 goto err_percpu_counter_init;
1387 /* Implementation specific variables. */
1389 /* Initialize default stream count setup information. */
1390 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1391 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1393 /* Initialize handle used for association ids. */
1394 idr_init(&sctp_assocs_id);
1396 limit = nr_free_buffer_pages() / 8;
1397 limit = max(limit, 128UL);
1398 sysctl_sctp_mem[0] = limit / 4 * 3;
1399 sysctl_sctp_mem[1] = limit;
1400 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1402 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1403 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1404 max_share = min(4UL*1024*1024, limit);
1406 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1407 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1408 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1410 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1411 sysctl_sctp_wmem[1] = 16*1024;
1412 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1414 /* Size and allocate the association hash table.
1415 * The methodology is similar to that of the tcp hash tables.
1416 * Though not identical. Start by getting a goal size
1418 if (totalram_pages >= (128 * 1024))
1419 goal = totalram_pages >> (22 - PAGE_SHIFT);
1420 else
1421 goal = totalram_pages >> (24 - PAGE_SHIFT);
1423 /* Then compute the page order for said goal */
1424 order = get_order(goal);
1426 /* Now compute the required page order for the maximum sized table we
1427 * want to create
1429 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1430 sizeof(struct sctp_bind_hashbucket));
1432 /* Limit the page order by that maximum hash table size */
1433 order = min(order, max_entry_order);
1435 /* Allocate and initialize the endpoint hash table. */
1436 sctp_ep_hashsize = 64;
1437 sctp_ep_hashtable =
1438 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1439 if (!sctp_ep_hashtable) {
1440 pr_err("Failed endpoint_hash alloc\n");
1441 status = -ENOMEM;
1442 goto err_ehash_alloc;
1444 for (i = 0; i < sctp_ep_hashsize; i++) {
1445 rwlock_init(&sctp_ep_hashtable[i].lock);
1446 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1449 /* Allocate and initialize the SCTP port hash table.
1450 * Note that order is initalized to start at the max sized
1451 * table we want to support. If we can't get that many pages
1452 * reduce the order and try again
1454 do {
1455 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1456 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1457 } while (!sctp_port_hashtable && --order > 0);
1459 if (!sctp_port_hashtable) {
1460 pr_err("Failed bind hash alloc\n");
1461 status = -ENOMEM;
1462 goto err_bhash_alloc;
1465 /* Now compute the number of entries that will fit in the
1466 * port hash space we allocated
1468 num_entries = (1UL << order) * PAGE_SIZE /
1469 sizeof(struct sctp_bind_hashbucket);
1471 /* And finish by rounding it down to the nearest power of two
1472 * this wastes some memory of course, but its needed because
1473 * the hash function operates based on the assumption that
1474 * that the number of entries is a power of two
1476 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1478 for (i = 0; i < sctp_port_hashsize; i++) {
1479 spin_lock_init(&sctp_port_hashtable[i].lock);
1480 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1483 status = sctp_transport_hashtable_init();
1484 if (status)
1485 goto err_thash_alloc;
1487 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1488 num_entries);
1490 sctp_sysctl_register();
1492 INIT_LIST_HEAD(&sctp_address_families);
1493 sctp_v4_pf_init();
1494 sctp_v6_pf_init();
1496 status = register_pernet_subsys(&sctp_defaults_ops);
1497 if (status)
1498 goto err_register_defaults;
1500 status = sctp_v4_protosw_init();
1501 if (status)
1502 goto err_protosw_init;
1504 status = sctp_v6_protosw_init();
1505 if (status)
1506 goto err_v6_protosw_init;
1508 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1509 if (status)
1510 goto err_register_ctrlsock;
1512 status = sctp_v4_add_protocol();
1513 if (status)
1514 goto err_add_protocol;
1516 /* Register SCTP with inet6 layer. */
1517 status = sctp_v6_add_protocol();
1518 if (status)
1519 goto err_v6_add_protocol;
1521 if (sctp_offload_init() < 0)
1522 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1524 out:
1525 return status;
1526 err_v6_add_protocol:
1527 sctp_v4_del_protocol();
1528 err_add_protocol:
1529 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1530 err_register_ctrlsock:
1531 sctp_v6_protosw_exit();
1532 err_v6_protosw_init:
1533 sctp_v4_protosw_exit();
1534 err_protosw_init:
1535 unregister_pernet_subsys(&sctp_defaults_ops);
1536 err_register_defaults:
1537 sctp_v4_pf_exit();
1538 sctp_v6_pf_exit();
1539 sctp_sysctl_unregister();
1540 free_pages((unsigned long)sctp_port_hashtable,
1541 get_order(sctp_port_hashsize *
1542 sizeof(struct sctp_bind_hashbucket)));
1543 err_bhash_alloc:
1544 sctp_transport_hashtable_destroy();
1545 err_thash_alloc:
1546 kfree(sctp_ep_hashtable);
1547 err_ehash_alloc:
1548 percpu_counter_destroy(&sctp_sockets_allocated);
1549 err_percpu_counter_init:
1550 kmem_cache_destroy(sctp_chunk_cachep);
1551 err_chunk_cachep:
1552 kmem_cache_destroy(sctp_bucket_cachep);
1553 goto out;
1556 /* Exit handler for the SCTP protocol. */
1557 static __exit void sctp_exit(void)
1559 /* BUG. This should probably do something useful like clean
1560 * up all the remaining associations and all that memory.
1563 /* Unregister with inet6/inet layers. */
1564 sctp_v6_del_protocol();
1565 sctp_v4_del_protocol();
1567 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1569 /* Free protosw registrations */
1570 sctp_v6_protosw_exit();
1571 sctp_v4_protosw_exit();
1573 unregister_pernet_subsys(&sctp_defaults_ops);
1575 /* Unregister with socket layer. */
1576 sctp_v6_pf_exit();
1577 sctp_v4_pf_exit();
1579 sctp_sysctl_unregister();
1581 free_pages((unsigned long)sctp_port_hashtable,
1582 get_order(sctp_port_hashsize *
1583 sizeof(struct sctp_bind_hashbucket)));
1584 kfree(sctp_ep_hashtable);
1585 sctp_transport_hashtable_destroy();
1587 percpu_counter_destroy(&sctp_sockets_allocated);
1589 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1591 kmem_cache_destroy(sctp_chunk_cachep);
1592 kmem_cache_destroy(sctp_bucket_cachep);
1595 module_init(sctp_init);
1596 module_exit(sctp_exit);
1599 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1601 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1602 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1603 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1604 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1605 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1606 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1607 MODULE_LICENSE("GPL");