KVM: PPC: Book3S HV: Prohibit setting illegal transaction state in MSR
[linux/fpc-iii.git] / net / sctp / protocol.c
blob57c2c4c0c97b9b9de8506b980b331c8fa2621e19
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 = t_new(struct sctp_sockaddr_entry, 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 SCTP_DEBUG_PRINTK("%s: DST:%pI4, SRC:%pI4 - ",
455 __func__, &fl4->daddr, &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->fl4_sport = laddr->a.v4.sin_port;
502 flowi4_update_output(fl4,
503 asoc->base.sk->sk_bound_dev_if,
504 RT_CONN_FLAGS(asoc->base.sk),
505 daddr->v4.sin_addr.s_addr,
506 laddr->a.v4.sin_addr.s_addr);
508 rt = ip_route_output_key(sock_net(sk), fl4);
509 if (!IS_ERR(rt)) {
510 dst = &rt->dst;
511 goto out_unlock;
516 out_unlock:
517 rcu_read_unlock();
518 out:
519 t->dst = dst;
520 if (dst)
521 SCTP_DEBUG_PRINTK("rt_dst:%pI4, rt_src:%pI4\n",
522 &fl4->daddr, &fl4->saddr);
523 else
524 SCTP_DEBUG_PRINTK("NO ROUTE\n");
527 /* For v4, the source address is cached in the route entry(dst). So no need
528 * to cache it separately and hence this is an empty routine.
530 static void sctp_v4_get_saddr(struct sctp_sock *sk,
531 struct sctp_transport *t,
532 struct flowi *fl)
534 union sctp_addr *saddr = &t->saddr;
535 struct rtable *rt = (struct rtable *)t->dst;
537 if (rt) {
538 saddr->v4.sin_family = AF_INET;
539 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
543 /* What interface did this skb arrive on? */
544 static int sctp_v4_skb_iif(const struct sk_buff *skb)
546 return inet_iif(skb);
549 /* Was this packet marked by Explicit Congestion Notification? */
550 static int sctp_v4_is_ce(const struct sk_buff *skb)
552 return INET_ECN_is_ce(ip_hdr(skb)->tos);
555 /* Create and initialize a new sk for the socket returned by accept(). */
556 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
557 struct sctp_association *asoc)
559 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
560 sk->sk_prot);
561 struct inet_sock *newinet;
563 if (!newsk)
564 goto out;
566 sock_init_data(NULL, newsk);
568 sctp_copy_sock(newsk, sk, asoc);
569 sock_reset_flag(newsk, SOCK_ZAPPED);
571 newinet = inet_sk(newsk);
573 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
575 sk_refcnt_debug_inc(newsk);
577 if (newsk->sk_prot->init(newsk)) {
578 sk_common_release(newsk);
579 newsk = NULL;
582 out:
583 return newsk;
586 /* Map address, empty for v4 family */
587 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
589 /* Empty */
592 /* Dump the v4 addr to the seq file. */
593 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
595 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
598 static void sctp_v4_ecn_capable(struct sock *sk)
600 INET_ECN_xmit(sk);
603 static void sctp_addr_wq_timeout_handler(unsigned long arg)
605 struct net *net = (struct net *)arg;
606 struct sctp_sockaddr_entry *addrw, *temp;
607 struct sctp_sock *sp;
609 spin_lock_bh(&net->sctp.addr_wq_lock);
611 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
612 SCTP_DEBUG_PRINTK_IPADDR("sctp_addrwq_timo_handler: the first ent in wq %p is ",
613 " for cmd %d at entry %p\n", &net->sctp.addr_waitq, &addrw->a, addrw->state,
614 addrw);
616 #if IS_ENABLED(CONFIG_IPV6)
617 /* Now we send an ASCONF for each association */
618 /* Note. we currently don't handle link local IPv6 addressees */
619 if (addrw->a.sa.sa_family == AF_INET6) {
620 struct in6_addr *in6;
622 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
623 IPV6_ADDR_LINKLOCAL)
624 goto free_next;
626 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
627 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
628 addrw->state == SCTP_ADDR_NEW) {
629 unsigned long timeo_val;
631 SCTP_DEBUG_PRINTK("sctp_timo_handler: this is on DAD, trying %d sec later\n",
632 SCTP_ADDRESS_TICK_DELAY);
633 timeo_val = jiffies;
634 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
635 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
636 break;
639 #endif
640 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
641 struct sock *sk;
643 sk = sctp_opt2sk(sp);
644 /* ignore bound-specific endpoints */
645 if (!sctp_is_ep_boundall(sk))
646 continue;
647 sctp_bh_lock_sock(sk);
648 if (sctp_asconf_mgmt(sp, addrw) < 0)
649 SCTP_DEBUG_PRINTK("sctp_addrwq_timo_handler: sctp_asconf_mgmt failed\n");
650 sctp_bh_unlock_sock(sk);
652 #if IS_ENABLED(CONFIG_IPV6)
653 free_next:
654 #endif
655 list_del(&addrw->list);
656 kfree(addrw);
658 spin_unlock_bh(&net->sctp.addr_wq_lock);
661 static void sctp_free_addr_wq(struct net *net)
663 struct sctp_sockaddr_entry *addrw;
664 struct sctp_sockaddr_entry *temp;
666 spin_lock_bh(&net->sctp.addr_wq_lock);
667 del_timer(&net->sctp.addr_wq_timer);
668 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
669 list_del(&addrw->list);
670 kfree(addrw);
672 spin_unlock_bh(&net->sctp.addr_wq_lock);
675 /* lookup the entry for the same address in the addr_waitq
676 * sctp_addr_wq MUST be locked
678 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
679 struct sctp_sockaddr_entry *addr)
681 struct sctp_sockaddr_entry *addrw;
683 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
684 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
685 continue;
686 if (addrw->a.sa.sa_family == AF_INET) {
687 if (addrw->a.v4.sin_addr.s_addr ==
688 addr->a.v4.sin_addr.s_addr)
689 return addrw;
690 } else if (addrw->a.sa.sa_family == AF_INET6) {
691 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
692 &addr->a.v6.sin6_addr))
693 return addrw;
696 return NULL;
699 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
701 struct sctp_sockaddr_entry *addrw;
702 unsigned long timeo_val;
704 /* first, we check if an opposite message already exist in the queue.
705 * If we found such message, it is removed.
706 * This operation is a bit stupid, but the DHCP client attaches the
707 * new address after a couple of addition and deletion of that address
710 spin_lock_bh(&net->sctp.addr_wq_lock);
711 /* Offsets existing events in addr_wq */
712 addrw = sctp_addr_wq_lookup(net, addr);
713 if (addrw) {
714 if (addrw->state != cmd) {
715 SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt offsets existing entry for %d ",
716 " in wq %p\n", addrw->state, &addrw->a,
717 &net->sctp.addr_waitq);
718 list_del(&addrw->list);
719 kfree(addrw);
721 spin_unlock_bh(&net->sctp.addr_wq_lock);
722 return;
725 /* OK, we have to add the new address to the wait queue */
726 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
727 if (addrw == NULL) {
728 spin_unlock_bh(&net->sctp.addr_wq_lock);
729 return;
731 addrw->state = cmd;
732 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
733 SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt add new entry for cmd:%d ",
734 " in wq %p\n", addrw->state, &addrw->a, &net->sctp.addr_waitq);
736 if (!timer_pending(&net->sctp.addr_wq_timer)) {
737 timeo_val = jiffies;
738 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
739 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
741 spin_unlock_bh(&net->sctp.addr_wq_lock);
744 /* Event handler for inet address addition/deletion events.
745 * The sctp_local_addr_list needs to be protocted by a spin lock since
746 * multiple notifiers (say IPv4 and IPv6) may be running at the same
747 * time and thus corrupt the list.
748 * The reader side is protected with RCU.
750 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
751 void *ptr)
753 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
754 struct sctp_sockaddr_entry *addr = NULL;
755 struct sctp_sockaddr_entry *temp;
756 struct net *net = dev_net(ifa->ifa_dev->dev);
757 int found = 0;
759 switch (ev) {
760 case NETDEV_UP:
761 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
762 if (addr) {
763 addr->a.v4.sin_family = AF_INET;
764 addr->a.v4.sin_port = 0;
765 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
766 addr->valid = 1;
767 spin_lock_bh(&net->sctp.local_addr_lock);
768 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
769 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
770 spin_unlock_bh(&net->sctp.local_addr_lock);
772 break;
773 case NETDEV_DOWN:
774 spin_lock_bh(&net->sctp.local_addr_lock);
775 list_for_each_entry_safe(addr, temp,
776 &net->sctp.local_addr_list, list) {
777 if (addr->a.sa.sa_family == AF_INET &&
778 addr->a.v4.sin_addr.s_addr ==
779 ifa->ifa_local) {
780 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
781 found = 1;
782 addr->valid = 0;
783 list_del_rcu(&addr->list);
784 break;
787 spin_unlock_bh(&net->sctp.local_addr_lock);
788 if (found)
789 kfree_rcu(addr, rcu);
790 break;
793 return NOTIFY_DONE;
797 * Initialize the control inode/socket with a control endpoint data
798 * structure. This endpoint is reserved exclusively for the OOTB processing.
800 static int sctp_ctl_sock_init(struct net *net)
802 int err;
803 sa_family_t family = PF_INET;
805 if (sctp_get_pf_specific(PF_INET6))
806 family = PF_INET6;
808 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
809 SOCK_SEQPACKET, IPPROTO_SCTP, net);
811 /* If IPv6 socket could not be created, try the IPv4 socket */
812 if (err < 0 && family == PF_INET6)
813 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
814 SOCK_SEQPACKET, IPPROTO_SCTP,
815 net);
817 if (err < 0) {
818 pr_err("Failed to create the SCTP control socket\n");
819 return err;
821 return 0;
824 /* Register address family specific functions. */
825 int sctp_register_af(struct sctp_af *af)
827 switch (af->sa_family) {
828 case AF_INET:
829 if (sctp_af_v4_specific)
830 return 0;
831 sctp_af_v4_specific = af;
832 break;
833 case AF_INET6:
834 if (sctp_af_v6_specific)
835 return 0;
836 sctp_af_v6_specific = af;
837 break;
838 default:
839 return 0;
842 INIT_LIST_HEAD(&af->list);
843 list_add_tail(&af->list, &sctp_address_families);
844 return 1;
847 /* Get the table of functions for manipulating a particular address
848 * family.
850 struct sctp_af *sctp_get_af_specific(sa_family_t family)
852 switch (family) {
853 case AF_INET:
854 return sctp_af_v4_specific;
855 case AF_INET6:
856 return sctp_af_v6_specific;
857 default:
858 return NULL;
862 /* Common code to initialize a AF_INET msg_name. */
863 static void sctp_inet_msgname(char *msgname, int *addr_len)
865 struct sockaddr_in *sin;
867 sin = (struct sockaddr_in *)msgname;
868 *addr_len = sizeof(struct sockaddr_in);
869 sin->sin_family = AF_INET;
870 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
873 /* Copy the primary address of the peer primary address as the msg_name. */
874 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
875 int *addr_len)
877 struct sockaddr_in *sin, *sinfrom;
879 if (msgname) {
880 struct sctp_association *asoc;
882 asoc = event->asoc;
883 sctp_inet_msgname(msgname, addr_len);
884 sin = (struct sockaddr_in *)msgname;
885 sinfrom = &asoc->peer.primary_addr.v4;
886 sin->sin_port = htons(asoc->peer.port);
887 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
891 /* Initialize and copy out a msgname from an inbound skb. */
892 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
894 if (msgname) {
895 struct sctphdr *sh = sctp_hdr(skb);
896 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
898 sctp_inet_msgname(msgname, len);
899 sin->sin_port = sh->source;
900 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
904 /* Do we support this AF? */
905 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
907 /* PF_INET only supports AF_INET addresses. */
908 return AF_INET == family;
911 /* Address matching with wildcards allowed. */
912 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
913 const union sctp_addr *addr2,
914 struct sctp_sock *opt)
916 /* PF_INET only supports AF_INET addresses. */
917 if (addr1->sa.sa_family != addr2->sa.sa_family)
918 return 0;
919 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
920 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
921 return 1;
922 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
923 return 1;
925 return 0;
928 /* Verify that provided sockaddr looks bindable. Common verification has
929 * already been taken care of.
931 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
933 return sctp_v4_available(addr, opt);
936 /* Verify that sockaddr looks sendable. Common verification has already
937 * been taken care of.
939 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
941 return 1;
944 /* Fill in Supported Address Type information for INIT and INIT-ACK
945 * chunks. Returns number of addresses supported.
947 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
948 __be16 *types)
950 types[0] = SCTP_PARAM_IPV4_ADDRESS;
951 return 1;
954 /* Wrapper routine that calls the ip transmit routine. */
955 static inline int sctp_v4_xmit(struct sk_buff *skb,
956 struct sctp_transport *transport)
958 struct inet_sock *inet = inet_sk(skb->sk);
960 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n",
961 __func__, skb, skb->len,
962 &transport->fl.u.ip4.saddr,
963 &transport->fl.u.ip4.daddr);
965 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
966 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
968 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
969 return ip_queue_xmit(skb, &transport->fl);
972 static struct sctp_af sctp_af_inet;
974 static struct sctp_pf sctp_pf_inet = {
975 .event_msgname = sctp_inet_event_msgname,
976 .skb_msgname = sctp_inet_skb_msgname,
977 .af_supported = sctp_inet_af_supported,
978 .cmp_addr = sctp_inet_cmp_addr,
979 .bind_verify = sctp_inet_bind_verify,
980 .send_verify = sctp_inet_send_verify,
981 .supported_addrs = sctp_inet_supported_addrs,
982 .create_accept_sk = sctp_v4_create_accept_sk,
983 .addr_v4map = sctp_v4_addr_v4map,
984 .af = &sctp_af_inet
987 /* Notifier for inetaddr addition/deletion events. */
988 static struct notifier_block sctp_inetaddr_notifier = {
989 .notifier_call = sctp_inetaddr_event,
992 /* Socket operations. */
993 static const struct proto_ops inet_seqpacket_ops = {
994 .family = PF_INET,
995 .owner = THIS_MODULE,
996 .release = inet_release, /* Needs to be wrapped... */
997 .bind = inet_bind,
998 .connect = inet_dgram_connect,
999 .socketpair = sock_no_socketpair,
1000 .accept = inet_accept,
1001 .getname = inet_getname, /* Semantics are different. */
1002 .poll = sctp_poll,
1003 .ioctl = inet_ioctl,
1004 .listen = sctp_inet_listen,
1005 .shutdown = inet_shutdown, /* Looks harmless. */
1006 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1007 .getsockopt = sock_common_getsockopt,
1008 .sendmsg = inet_sendmsg,
1009 .recvmsg = sock_common_recvmsg,
1010 .mmap = sock_no_mmap,
1011 .sendpage = sock_no_sendpage,
1012 #ifdef CONFIG_COMPAT
1013 .compat_setsockopt = compat_sock_common_setsockopt,
1014 .compat_getsockopt = compat_sock_common_getsockopt,
1015 #endif
1018 /* Registration with AF_INET family. */
1019 static struct inet_protosw sctp_seqpacket_protosw = {
1020 .type = SOCK_SEQPACKET,
1021 .protocol = IPPROTO_SCTP,
1022 .prot = &sctp_prot,
1023 .ops = &inet_seqpacket_ops,
1024 .no_check = 0,
1025 .flags = SCTP_PROTOSW_FLAG
1027 static struct inet_protosw sctp_stream_protosw = {
1028 .type = SOCK_STREAM,
1029 .protocol = IPPROTO_SCTP,
1030 .prot = &sctp_prot,
1031 .ops = &inet_seqpacket_ops,
1032 .no_check = 0,
1033 .flags = SCTP_PROTOSW_FLAG
1036 /* Register with IP layer. */
1037 static const struct net_protocol sctp_protocol = {
1038 .handler = sctp_rcv,
1039 .err_handler = sctp_v4_err,
1040 .no_policy = 1,
1041 .netns_ok = 1,
1044 /* IPv4 address related functions. */
1045 static struct sctp_af sctp_af_inet = {
1046 .sa_family = AF_INET,
1047 .sctp_xmit = sctp_v4_xmit,
1048 .setsockopt = ip_setsockopt,
1049 .getsockopt = ip_getsockopt,
1050 .get_dst = sctp_v4_get_dst,
1051 .get_saddr = sctp_v4_get_saddr,
1052 .copy_addrlist = sctp_v4_copy_addrlist,
1053 .from_skb = sctp_v4_from_skb,
1054 .from_sk = sctp_v4_from_sk,
1055 .to_sk_saddr = sctp_v4_to_sk_saddr,
1056 .to_sk_daddr = sctp_v4_to_sk_daddr,
1057 .from_addr_param = sctp_v4_from_addr_param,
1058 .to_addr_param = sctp_v4_to_addr_param,
1059 .cmp_addr = sctp_v4_cmp_addr,
1060 .addr_valid = sctp_v4_addr_valid,
1061 .inaddr_any = sctp_v4_inaddr_any,
1062 .is_any = sctp_v4_is_any,
1063 .available = sctp_v4_available,
1064 .scope = sctp_v4_scope,
1065 .skb_iif = sctp_v4_skb_iif,
1066 .is_ce = sctp_v4_is_ce,
1067 .seq_dump_addr = sctp_v4_seq_dump_addr,
1068 .ecn_capable = sctp_v4_ecn_capable,
1069 .net_header_len = sizeof(struct iphdr),
1070 .sockaddr_len = sizeof(struct sockaddr_in),
1071 #ifdef CONFIG_COMPAT
1072 .compat_setsockopt = compat_ip_setsockopt,
1073 .compat_getsockopt = compat_ip_getsockopt,
1074 #endif
1077 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
1079 switch (family) {
1080 case PF_INET:
1081 return sctp_pf_inet_specific;
1082 case PF_INET6:
1083 return sctp_pf_inet6_specific;
1084 default:
1085 return NULL;
1089 /* Register the PF specific function table. */
1090 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1092 switch (family) {
1093 case PF_INET:
1094 if (sctp_pf_inet_specific)
1095 return 0;
1096 sctp_pf_inet_specific = pf;
1097 break;
1098 case PF_INET6:
1099 if (sctp_pf_inet6_specific)
1100 return 0;
1101 sctp_pf_inet6_specific = pf;
1102 break;
1103 default:
1104 return 0;
1106 return 1;
1109 static inline int init_sctp_mibs(struct net *net)
1111 return snmp_mib_init((void __percpu **)net->sctp.sctp_statistics,
1112 sizeof(struct sctp_mib),
1113 __alignof__(struct sctp_mib));
1116 static inline void cleanup_sctp_mibs(struct net *net)
1118 snmp_mib_free((void __percpu **)net->sctp.sctp_statistics);
1121 static void sctp_v4_pf_init(void)
1123 /* Initialize the SCTP specific PF functions. */
1124 sctp_register_pf(&sctp_pf_inet, PF_INET);
1125 sctp_register_af(&sctp_af_inet);
1128 static void sctp_v4_pf_exit(void)
1130 list_del(&sctp_af_inet.list);
1133 static int sctp_v4_protosw_init(void)
1135 int rc;
1137 rc = proto_register(&sctp_prot, 1);
1138 if (rc)
1139 return rc;
1141 /* Register SCTP(UDP and TCP style) with socket layer. */
1142 inet_register_protosw(&sctp_seqpacket_protosw);
1143 inet_register_protosw(&sctp_stream_protosw);
1145 return 0;
1148 static void sctp_v4_protosw_exit(void)
1150 inet_unregister_protosw(&sctp_stream_protosw);
1151 inet_unregister_protosw(&sctp_seqpacket_protosw);
1152 proto_unregister(&sctp_prot);
1155 static int sctp_v4_add_protocol(void)
1157 /* Register notifier for inet address additions/deletions. */
1158 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1160 /* Register SCTP with inet layer. */
1161 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1162 return -EAGAIN;
1164 return 0;
1167 static void sctp_v4_del_protocol(void)
1169 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1170 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1173 static int __net_init sctp_defaults_init(struct net *net)
1175 int status;
1178 * 14. Suggested SCTP Protocol Parameter Values
1180 /* The following protocol parameters are RECOMMENDED: */
1181 /* RTO.Initial - 3 seconds */
1182 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1183 /* RTO.Min - 1 second */
1184 net->sctp.rto_min = SCTP_RTO_MIN;
1185 /* RTO.Max - 60 seconds */
1186 net->sctp.rto_max = SCTP_RTO_MAX;
1187 /* RTO.Alpha - 1/8 */
1188 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1189 /* RTO.Beta - 1/4 */
1190 net->sctp.rto_beta = SCTP_RTO_BETA;
1192 /* Valid.Cookie.Life - 60 seconds */
1193 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1195 /* Whether Cookie Preservative is enabled(1) or not(0) */
1196 net->sctp.cookie_preserve_enable = 1;
1198 /* Default sctp sockets to use md5 as their hmac alg */
1199 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1200 net->sctp.sctp_hmac_alg = "md5";
1201 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1202 net->sctp.sctp_hmac_alg = "sha1";
1203 #else
1204 net->sctp.sctp_hmac_alg = NULL;
1205 #endif
1207 /* Max.Burst - 4 */
1208 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1210 /* Association.Max.Retrans - 10 attempts
1211 * Path.Max.Retrans - 5 attempts (per destination address)
1212 * Max.Init.Retransmits - 8 attempts
1214 net->sctp.max_retrans_association = 10;
1215 net->sctp.max_retrans_path = 5;
1216 net->sctp.max_retrans_init = 8;
1218 /* Sendbuffer growth - do per-socket accounting */
1219 net->sctp.sndbuf_policy = 0;
1221 /* Rcvbuffer growth - do per-socket accounting */
1222 net->sctp.rcvbuf_policy = 0;
1224 /* HB.interval - 30 seconds */
1225 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1227 /* delayed SACK timeout */
1228 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1230 /* Disable ADDIP by default. */
1231 net->sctp.addip_enable = 0;
1232 net->sctp.addip_noauth = 0;
1233 net->sctp.default_auto_asconf = 0;
1235 /* Enable PR-SCTP by default. */
1236 net->sctp.prsctp_enable = 1;
1238 /* Disable AUTH by default. */
1239 net->sctp.auth_enable = 0;
1241 /* Set SCOPE policy to enabled */
1242 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1244 /* Set the default rwnd update threshold */
1245 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1247 /* Initialize maximum autoclose timeout. */
1248 net->sctp.max_autoclose = INT_MAX / HZ;
1250 status = sctp_sysctl_net_register(net);
1251 if (status)
1252 goto err_sysctl_register;
1254 /* Allocate and initialise sctp mibs. */
1255 status = init_sctp_mibs(net);
1256 if (status)
1257 goto err_init_mibs;
1259 /* Initialize proc fs directory. */
1260 status = sctp_proc_init(net);
1261 if (status)
1262 goto err_init_proc;
1264 sctp_dbg_objcnt_init(net);
1266 /* Initialize the local address list. */
1267 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1268 spin_lock_init(&net->sctp.local_addr_lock);
1269 sctp_get_local_addr_list(net);
1271 /* Initialize the address event list */
1272 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1273 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1274 spin_lock_init(&net->sctp.addr_wq_lock);
1275 net->sctp.addr_wq_timer.expires = 0;
1276 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1277 (unsigned long)net);
1279 return 0;
1281 err_init_proc:
1282 cleanup_sctp_mibs(net);
1283 err_init_mibs:
1284 sctp_sysctl_net_unregister(net);
1285 err_sysctl_register:
1286 return status;
1289 static void __net_exit sctp_defaults_exit(struct net *net)
1291 /* Free the local address list */
1292 sctp_free_addr_wq(net);
1293 sctp_free_local_addr_list(net);
1295 sctp_dbg_objcnt_exit(net);
1297 sctp_proc_exit(net);
1298 cleanup_sctp_mibs(net);
1299 sctp_sysctl_net_unregister(net);
1302 static struct pernet_operations sctp_defaults_ops = {
1303 .init = sctp_defaults_init,
1304 .exit = sctp_defaults_exit,
1307 static int __net_init sctp_ctrlsock_init(struct net *net)
1309 int status;
1311 /* Initialize the control inode/socket for handling OOTB packets. */
1312 status = sctp_ctl_sock_init(net);
1313 if (status)
1314 pr_err("Failed to initialize the SCTP control sock\n");
1316 return status;
1319 static void __net_init sctp_ctrlsock_exit(struct net *net)
1321 /* Free the control endpoint. */
1322 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1325 static struct pernet_operations sctp_ctrlsock_ops = {
1326 .init = sctp_ctrlsock_init,
1327 .exit = sctp_ctrlsock_exit,
1330 /* Initialize the universe into something sensible. */
1331 SCTP_STATIC __init int sctp_init(void)
1333 int i;
1334 int status = -EINVAL;
1335 unsigned long goal;
1336 unsigned long limit;
1337 int max_share;
1338 int order;
1340 /* SCTP_DEBUG sanity check. */
1341 if (!sctp_sanity_check())
1342 goto out;
1344 /* Allocate bind_bucket and chunk caches. */
1345 status = -ENOBUFS;
1346 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1347 sizeof(struct sctp_bind_bucket),
1348 0, SLAB_HWCACHE_ALIGN,
1349 NULL);
1350 if (!sctp_bucket_cachep)
1351 goto out;
1353 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1354 sizeof(struct sctp_chunk),
1355 0, SLAB_HWCACHE_ALIGN,
1356 NULL);
1357 if (!sctp_chunk_cachep)
1358 goto err_chunk_cachep;
1360 status = percpu_counter_init(&sctp_sockets_allocated, 0);
1361 if (status)
1362 goto err_percpu_counter_init;
1364 /* Implementation specific variables. */
1366 /* Initialize default stream count setup information. */
1367 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1368 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1370 /* Initialize handle used for association ids. */
1371 idr_init(&sctp_assocs_id);
1373 limit = nr_free_buffer_pages() / 8;
1374 limit = max(limit, 128UL);
1375 sysctl_sctp_mem[0] = limit / 4 * 3;
1376 sysctl_sctp_mem[1] = limit;
1377 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1379 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1380 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1381 max_share = min(4UL*1024*1024, limit);
1383 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1384 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1385 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1387 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1388 sysctl_sctp_wmem[1] = 16*1024;
1389 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1391 /* Size and allocate the association hash table.
1392 * The methodology is similar to that of the tcp hash tables.
1394 if (totalram_pages >= (128 * 1024))
1395 goal = totalram_pages >> (22 - PAGE_SHIFT);
1396 else
1397 goal = totalram_pages >> (24 - PAGE_SHIFT);
1399 for (order = 0; (1UL << order) < goal; order++)
1402 do {
1403 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1404 sizeof(struct sctp_hashbucket);
1405 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1406 continue;
1407 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1408 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1409 } while (!sctp_assoc_hashtable && --order > 0);
1410 if (!sctp_assoc_hashtable) {
1411 pr_err("Failed association hash alloc\n");
1412 status = -ENOMEM;
1413 goto err_ahash_alloc;
1415 for (i = 0; i < sctp_assoc_hashsize; i++) {
1416 rwlock_init(&sctp_assoc_hashtable[i].lock);
1417 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1420 /* Allocate and initialize the endpoint hash table. */
1421 sctp_ep_hashsize = 64;
1422 sctp_ep_hashtable =
1423 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1424 if (!sctp_ep_hashtable) {
1425 pr_err("Failed endpoint_hash alloc\n");
1426 status = -ENOMEM;
1427 goto err_ehash_alloc;
1429 for (i = 0; i < sctp_ep_hashsize; i++) {
1430 rwlock_init(&sctp_ep_hashtable[i].lock);
1431 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1434 /* Allocate and initialize the SCTP port hash table. */
1435 do {
1436 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1437 sizeof(struct sctp_bind_hashbucket);
1438 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1439 continue;
1440 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1441 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1442 } while (!sctp_port_hashtable && --order > 0);
1443 if (!sctp_port_hashtable) {
1444 pr_err("Failed bind hash alloc\n");
1445 status = -ENOMEM;
1446 goto err_bhash_alloc;
1448 for (i = 0; i < sctp_port_hashsize; i++) {
1449 spin_lock_init(&sctp_port_hashtable[i].lock);
1450 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1453 pr_info("Hash tables configured (established %d bind %d)\n",
1454 sctp_assoc_hashsize, sctp_port_hashsize);
1456 sctp_sysctl_register();
1458 INIT_LIST_HEAD(&sctp_address_families);
1459 sctp_v4_pf_init();
1460 sctp_v6_pf_init();
1462 status = register_pernet_subsys(&sctp_defaults_ops);
1463 if (status)
1464 goto err_register_defaults;
1466 status = sctp_v4_protosw_init();
1467 if (status)
1468 goto err_protosw_init;
1470 status = sctp_v6_protosw_init();
1471 if (status)
1472 goto err_v6_protosw_init;
1474 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1475 if (status)
1476 goto err_register_ctrlsock;
1478 status = sctp_v4_add_protocol();
1479 if (status)
1480 goto err_add_protocol;
1482 /* Register SCTP with inet6 layer. */
1483 status = sctp_v6_add_protocol();
1484 if (status)
1485 goto err_v6_add_protocol;
1487 status = 0;
1488 out:
1489 return status;
1490 err_v6_add_protocol:
1491 sctp_v4_del_protocol();
1492 err_add_protocol:
1493 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1494 err_register_ctrlsock:
1495 sctp_v6_protosw_exit();
1496 err_v6_protosw_init:
1497 sctp_v4_protosw_exit();
1498 err_protosw_init:
1499 unregister_pernet_subsys(&sctp_defaults_ops);
1500 err_register_defaults:
1501 sctp_v4_pf_exit();
1502 sctp_v6_pf_exit();
1503 sctp_sysctl_unregister();
1504 free_pages((unsigned long)sctp_port_hashtable,
1505 get_order(sctp_port_hashsize *
1506 sizeof(struct sctp_bind_hashbucket)));
1507 err_bhash_alloc:
1508 kfree(sctp_ep_hashtable);
1509 err_ehash_alloc:
1510 free_pages((unsigned long)sctp_assoc_hashtable,
1511 get_order(sctp_assoc_hashsize *
1512 sizeof(struct sctp_hashbucket)));
1513 err_ahash_alloc:
1514 percpu_counter_destroy(&sctp_sockets_allocated);
1515 err_percpu_counter_init:
1516 kmem_cache_destroy(sctp_chunk_cachep);
1517 err_chunk_cachep:
1518 kmem_cache_destroy(sctp_bucket_cachep);
1519 goto out;
1522 /* Exit handler for the SCTP protocol. */
1523 SCTP_STATIC __exit void sctp_exit(void)
1525 /* BUG. This should probably do something useful like clean
1526 * up all the remaining associations and all that memory.
1529 /* Unregister with inet6/inet layers. */
1530 sctp_v6_del_protocol();
1531 sctp_v4_del_protocol();
1533 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1535 /* Free protosw registrations */
1536 sctp_v6_protosw_exit();
1537 sctp_v4_protosw_exit();
1539 unregister_pernet_subsys(&sctp_defaults_ops);
1541 /* Unregister with socket layer. */
1542 sctp_v6_pf_exit();
1543 sctp_v4_pf_exit();
1545 sctp_sysctl_unregister();
1547 free_pages((unsigned long)sctp_assoc_hashtable,
1548 get_order(sctp_assoc_hashsize *
1549 sizeof(struct sctp_hashbucket)));
1550 kfree(sctp_ep_hashtable);
1551 free_pages((unsigned long)sctp_port_hashtable,
1552 get_order(sctp_port_hashsize *
1553 sizeof(struct sctp_bind_hashbucket)));
1555 percpu_counter_destroy(&sctp_sockets_allocated);
1557 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1559 kmem_cache_destroy(sctp_chunk_cachep);
1560 kmem_cache_destroy(sctp_bucket_cachep);
1563 module_init(sctp_init);
1564 module_exit(sctp_exit);
1567 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1569 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1570 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1571 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1572 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1573 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1574 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1575 MODULE_LICENSE("GPL");