rtlwifi: rtl_pci: Simplify some code be eliminating extraneous variables
[linux/fpc-iii.git] / net / sctp / protocol.c
blobf5172c21349bec865f3812ba63aeb93b6c4fbab9
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 enum sctp_scope scope, gfp_t gfp, int copy_flags)
201 struct sctp_sockaddr_entry *addr;
202 union sctp_addr laddr;
203 int error = 0;
205 rcu_read_lock();
206 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
207 if (!addr->valid)
208 continue;
209 if (!sctp_in_scope(net, &addr->a, scope))
210 continue;
212 /* Now that the address is in scope, check to see if
213 * the address type is really supported by the local
214 * sock as well as the remote peer.
216 if (addr->a.sa.sa_family == AF_INET &&
217 !(copy_flags & SCTP_ADDR4_PEERSUPP))
218 continue;
219 if (addr->a.sa.sa_family == AF_INET6 &&
220 (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
221 !(copy_flags & SCTP_ADDR6_PEERSUPP)))
222 continue;
224 laddr = addr->a;
225 /* also works for setting ipv6 address port */
226 laddr.v4.sin_port = htons(bp->port);
227 if (sctp_bind_addr_state(bp, &laddr) != -1)
228 continue;
230 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
231 SCTP_ADDR_SRC, GFP_ATOMIC);
232 if (error)
233 break;
236 rcu_read_unlock();
237 return error;
240 /* Initialize a sctp_addr from in incoming skb. */
241 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
242 int is_saddr)
244 /* Always called on head skb, so this is safe */
245 struct sctphdr *sh = sctp_hdr(skb);
246 struct sockaddr_in *sa = &addr->v4;
248 addr->v4.sin_family = AF_INET;
250 if (is_saddr) {
251 sa->sin_port = sh->source;
252 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
253 } else {
254 sa->sin_port = sh->dest;
255 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
259 /* Initialize an sctp_addr from a socket. */
260 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
262 addr->v4.sin_family = AF_INET;
263 addr->v4.sin_port = 0;
264 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
267 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
268 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
270 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
273 /* Initialize sk->sk_daddr from sctp_addr. */
274 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
276 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
279 /* Initialize a sctp_addr from an address parameter. */
280 static void sctp_v4_from_addr_param(union sctp_addr *addr,
281 union sctp_addr_param *param,
282 __be16 port, int iif)
284 addr->v4.sin_family = AF_INET;
285 addr->v4.sin_port = port;
286 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
289 /* Initialize an address parameter from a sctp_addr and return the length
290 * of the address parameter.
292 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
293 union sctp_addr_param *param)
295 int length = sizeof(struct sctp_ipv4addr_param);
297 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
298 param->v4.param_hdr.length = htons(length);
299 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
301 return length;
304 /* Initialize a sctp_addr from a dst_entry. */
305 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
306 __be16 port)
308 saddr->v4.sin_family = AF_INET;
309 saddr->v4.sin_port = port;
310 saddr->v4.sin_addr.s_addr = fl4->saddr;
313 /* Compare two addresses exactly. */
314 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
315 const union sctp_addr *addr2)
317 if (addr1->sa.sa_family != addr2->sa.sa_family)
318 return 0;
319 if (addr1->v4.sin_port != addr2->v4.sin_port)
320 return 0;
321 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
322 return 0;
324 return 1;
327 /* Initialize addr struct to INADDR_ANY. */
328 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
330 addr->v4.sin_family = AF_INET;
331 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
332 addr->v4.sin_port = port;
335 /* Is this a wildcard address? */
336 static int sctp_v4_is_any(const union sctp_addr *addr)
338 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
341 /* This function checks if the address is a valid address to be used for
342 * SCTP binding.
344 * Output:
345 * Return 0 - If the address is a non-unicast or an illegal address.
346 * Return 1 - If the address is a unicast.
348 static int sctp_v4_addr_valid(union sctp_addr *addr,
349 struct sctp_sock *sp,
350 const struct sk_buff *skb)
352 /* IPv4 addresses not allowed */
353 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
354 return 0;
356 /* Is this a non-unicast address or a unusable SCTP address? */
357 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
358 return 0;
360 /* Is this a broadcast address? */
361 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
362 return 0;
364 return 1;
367 /* Should this be available for binding? */
368 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
370 struct net *net = sock_net(&sp->inet.sk);
371 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
374 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
375 ret != RTN_LOCAL &&
376 !sp->inet.freebind &&
377 !net->ipv4.sysctl_ip_nonlocal_bind)
378 return 0;
380 if (ipv6_only_sock(sctp_opt2sk(sp)))
381 return 0;
383 return 1;
386 /* Checking the loopback, private and other address scopes as defined in
387 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
388 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
390 * Level 0 - unusable SCTP addresses
391 * Level 1 - loopback address
392 * Level 2 - link-local addresses
393 * Level 3 - private addresses.
394 * Level 4 - global addresses
395 * For INIT and INIT-ACK address list, let L be the level of
396 * of requested destination address, sender and receiver
397 * SHOULD include all of its addresses with level greater
398 * than or equal to L.
400 * IPv4 scoping can be controlled through sysctl option
401 * net.sctp.addr_scope_policy
403 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
405 enum sctp_scope retval;
407 /* Check for unusable SCTP addresses. */
408 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
409 retval = SCTP_SCOPE_UNUSABLE;
410 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
411 retval = SCTP_SCOPE_LOOPBACK;
412 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
413 retval = SCTP_SCOPE_LINK;
414 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
415 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
416 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
417 retval = SCTP_SCOPE_PRIVATE;
418 } else {
419 retval = SCTP_SCOPE_GLOBAL;
422 return retval;
425 /* Returns a valid dst cache entry for the given source and destination ip
426 * addresses. If an association is passed, trys to get a dst entry with a
427 * source address that matches an address in the bind address list.
429 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
430 struct flowi *fl, struct sock *sk)
432 struct sctp_association *asoc = t->asoc;
433 struct rtable *rt;
434 struct flowi4 *fl4 = &fl->u.ip4;
435 struct sctp_bind_addr *bp;
436 struct sctp_sockaddr_entry *laddr;
437 struct dst_entry *dst = NULL;
438 union sctp_addr *daddr = &t->ipaddr;
439 union sctp_addr dst_saddr;
441 memset(fl4, 0x0, sizeof(struct flowi4));
442 fl4->daddr = daddr->v4.sin_addr.s_addr;
443 fl4->fl4_dport = daddr->v4.sin_port;
444 fl4->flowi4_proto = IPPROTO_SCTP;
445 if (asoc) {
446 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
447 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
448 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
450 if (saddr) {
451 fl4->saddr = saddr->v4.sin_addr.s_addr;
452 fl4->fl4_sport = saddr->v4.sin_port;
455 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
456 &fl4->saddr);
458 rt = ip_route_output_key(sock_net(sk), fl4);
459 if (!IS_ERR(rt))
460 dst = &rt->dst;
462 /* If there is no association or if a source address is passed, no
463 * more validation is required.
465 if (!asoc || saddr)
466 goto out;
468 bp = &asoc->base.bind_addr;
470 if (dst) {
471 /* Walk through the bind address list and look for a bind
472 * address that matches the source address of the returned dst.
474 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
475 rcu_read_lock();
476 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
477 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
478 (laddr->state != SCTP_ADDR_SRC &&
479 !asoc->src_out_of_asoc_ok))
480 continue;
481 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
482 goto out_unlock;
484 rcu_read_unlock();
486 /* None of the bound addresses match the source address of the
487 * dst. So release it.
489 dst_release(dst);
490 dst = NULL;
493 /* Walk through the bind address list and try to get a dst that
494 * matches a bind address as the source address.
496 rcu_read_lock();
497 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
498 struct net_device *odev;
500 if (!laddr->valid)
501 continue;
502 if (laddr->state != SCTP_ADDR_SRC ||
503 AF_INET != laddr->a.sa.sa_family)
504 continue;
506 fl4->fl4_sport = laddr->a.v4.sin_port;
507 flowi4_update_output(fl4,
508 asoc->base.sk->sk_bound_dev_if,
509 RT_CONN_FLAGS(asoc->base.sk),
510 daddr->v4.sin_addr.s_addr,
511 laddr->a.v4.sin_addr.s_addr);
513 rt = ip_route_output_key(sock_net(sk), fl4);
514 if (IS_ERR(rt))
515 continue;
517 if (!dst)
518 dst = &rt->dst;
520 /* Ensure the src address belongs to the output
521 * interface.
523 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
524 false);
525 if (!odev || odev->ifindex != fl4->flowi4_oif) {
526 if (&rt->dst != dst)
527 dst_release(&rt->dst);
528 continue;
531 if (dst != &rt->dst)
532 dst_release(dst);
533 dst = &rt->dst;
534 break;
537 out_unlock:
538 rcu_read_unlock();
539 out:
540 t->dst = dst;
541 if (dst)
542 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
543 &fl4->daddr, &fl4->saddr);
544 else
545 pr_debug("no route\n");
548 /* For v4, the source address is cached in the route entry(dst). So no need
549 * to cache it separately and hence this is an empty routine.
551 static void sctp_v4_get_saddr(struct sctp_sock *sk,
552 struct sctp_transport *t,
553 struct flowi *fl)
555 union sctp_addr *saddr = &t->saddr;
556 struct rtable *rt = (struct rtable *)t->dst;
558 if (rt) {
559 saddr->v4.sin_family = AF_INET;
560 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
564 /* What interface did this skb arrive on? */
565 static int sctp_v4_skb_iif(const struct sk_buff *skb)
567 return inet_iif(skb);
570 /* Was this packet marked by Explicit Congestion Notification? */
571 static int sctp_v4_is_ce(const struct sk_buff *skb)
573 return INET_ECN_is_ce(ip_hdr(skb)->tos);
576 /* Create and initialize a new sk for the socket returned by accept(). */
577 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
578 struct sctp_association *asoc,
579 bool kern)
581 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
582 sk->sk_prot, kern);
583 struct inet_sock *newinet;
585 if (!newsk)
586 goto out;
588 sock_init_data(NULL, newsk);
590 sctp_copy_sock(newsk, sk, asoc);
591 sock_reset_flag(newsk, SOCK_ZAPPED);
593 newinet = inet_sk(newsk);
595 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
597 sk_refcnt_debug_inc(newsk);
599 if (newsk->sk_prot->init(newsk)) {
600 sk_common_release(newsk);
601 newsk = NULL;
604 out:
605 return newsk;
608 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
610 /* No address mapping for V4 sockets */
611 return sizeof(struct sockaddr_in);
614 /* Dump the v4 addr to the seq file. */
615 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
617 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
620 static void sctp_v4_ecn_capable(struct sock *sk)
622 INET_ECN_xmit(sk);
625 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
627 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
628 struct sctp_sockaddr_entry *addrw, *temp;
629 struct sctp_sock *sp;
631 spin_lock_bh(&net->sctp.addr_wq_lock);
633 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
634 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
635 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
636 addrw->state, addrw);
638 #if IS_ENABLED(CONFIG_IPV6)
639 /* Now we send an ASCONF for each association */
640 /* Note. we currently don't handle link local IPv6 addressees */
641 if (addrw->a.sa.sa_family == AF_INET6) {
642 struct in6_addr *in6;
644 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
645 IPV6_ADDR_LINKLOCAL)
646 goto free_next;
648 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
649 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
650 addrw->state == SCTP_ADDR_NEW) {
651 unsigned long timeo_val;
653 pr_debug("%s: this is on DAD, trying %d sec "
654 "later\n", __func__,
655 SCTP_ADDRESS_TICK_DELAY);
657 timeo_val = jiffies;
658 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
659 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
660 break;
663 #endif
664 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
665 struct sock *sk;
667 sk = sctp_opt2sk(sp);
668 /* ignore bound-specific endpoints */
669 if (!sctp_is_ep_boundall(sk))
670 continue;
671 bh_lock_sock(sk);
672 if (sctp_asconf_mgmt(sp, addrw) < 0)
673 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
674 bh_unlock_sock(sk);
676 #if IS_ENABLED(CONFIG_IPV6)
677 free_next:
678 #endif
679 list_del(&addrw->list);
680 kfree(addrw);
682 spin_unlock_bh(&net->sctp.addr_wq_lock);
685 static void sctp_free_addr_wq(struct net *net)
687 struct sctp_sockaddr_entry *addrw;
688 struct sctp_sockaddr_entry *temp;
690 spin_lock_bh(&net->sctp.addr_wq_lock);
691 del_timer(&net->sctp.addr_wq_timer);
692 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
693 list_del(&addrw->list);
694 kfree(addrw);
696 spin_unlock_bh(&net->sctp.addr_wq_lock);
699 /* lookup the entry for the same address in the addr_waitq
700 * sctp_addr_wq MUST be locked
702 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
703 struct sctp_sockaddr_entry *addr)
705 struct sctp_sockaddr_entry *addrw;
707 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
708 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
709 continue;
710 if (addrw->a.sa.sa_family == AF_INET) {
711 if (addrw->a.v4.sin_addr.s_addr ==
712 addr->a.v4.sin_addr.s_addr)
713 return addrw;
714 } else if (addrw->a.sa.sa_family == AF_INET6) {
715 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
716 &addr->a.v6.sin6_addr))
717 return addrw;
720 return NULL;
723 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
725 struct sctp_sockaddr_entry *addrw;
726 unsigned long timeo_val;
728 /* first, we check if an opposite message already exist in the queue.
729 * If we found such message, it is removed.
730 * This operation is a bit stupid, but the DHCP client attaches the
731 * new address after a couple of addition and deletion of that address
734 spin_lock_bh(&net->sctp.addr_wq_lock);
735 /* Offsets existing events in addr_wq */
736 addrw = sctp_addr_wq_lookup(net, addr);
737 if (addrw) {
738 if (addrw->state != cmd) {
739 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
740 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
741 &net->sctp.addr_waitq);
743 list_del(&addrw->list);
744 kfree(addrw);
746 spin_unlock_bh(&net->sctp.addr_wq_lock);
747 return;
750 /* OK, we have to add the new address to the wait queue */
751 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
752 if (addrw == NULL) {
753 spin_unlock_bh(&net->sctp.addr_wq_lock);
754 return;
756 addrw->state = cmd;
757 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
759 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
760 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
762 if (!timer_pending(&net->sctp.addr_wq_timer)) {
763 timeo_val = jiffies;
764 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
765 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
767 spin_unlock_bh(&net->sctp.addr_wq_lock);
770 /* Event handler for inet address addition/deletion events.
771 * The sctp_local_addr_list needs to be protocted by a spin lock since
772 * multiple notifiers (say IPv4 and IPv6) may be running at the same
773 * time and thus corrupt the list.
774 * The reader side is protected with RCU.
776 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
777 void *ptr)
779 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
780 struct sctp_sockaddr_entry *addr = NULL;
781 struct sctp_sockaddr_entry *temp;
782 struct net *net = dev_net(ifa->ifa_dev->dev);
783 int found = 0;
785 switch (ev) {
786 case NETDEV_UP:
787 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
788 if (addr) {
789 addr->a.v4.sin_family = AF_INET;
790 addr->a.v4.sin_port = 0;
791 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
792 addr->valid = 1;
793 spin_lock_bh(&net->sctp.local_addr_lock);
794 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
795 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
796 spin_unlock_bh(&net->sctp.local_addr_lock);
798 break;
799 case NETDEV_DOWN:
800 spin_lock_bh(&net->sctp.local_addr_lock);
801 list_for_each_entry_safe(addr, temp,
802 &net->sctp.local_addr_list, list) {
803 if (addr->a.sa.sa_family == AF_INET &&
804 addr->a.v4.sin_addr.s_addr ==
805 ifa->ifa_local) {
806 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
807 found = 1;
808 addr->valid = 0;
809 list_del_rcu(&addr->list);
810 break;
813 spin_unlock_bh(&net->sctp.local_addr_lock);
814 if (found)
815 kfree_rcu(addr, rcu);
816 break;
819 return NOTIFY_DONE;
823 * Initialize the control inode/socket with a control endpoint data
824 * structure. This endpoint is reserved exclusively for the OOTB processing.
826 static int sctp_ctl_sock_init(struct net *net)
828 int err;
829 sa_family_t family = PF_INET;
831 if (sctp_get_pf_specific(PF_INET6))
832 family = PF_INET6;
834 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
835 SOCK_SEQPACKET, IPPROTO_SCTP, net);
837 /* If IPv6 socket could not be created, try the IPv4 socket */
838 if (err < 0 && family == PF_INET6)
839 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
840 SOCK_SEQPACKET, IPPROTO_SCTP,
841 net);
843 if (err < 0) {
844 pr_err("Failed to create the SCTP control socket\n");
845 return err;
847 return 0;
850 /* Register address family specific functions. */
851 int sctp_register_af(struct sctp_af *af)
853 switch (af->sa_family) {
854 case AF_INET:
855 if (sctp_af_v4_specific)
856 return 0;
857 sctp_af_v4_specific = af;
858 break;
859 case AF_INET6:
860 if (sctp_af_v6_specific)
861 return 0;
862 sctp_af_v6_specific = af;
863 break;
864 default:
865 return 0;
868 INIT_LIST_HEAD(&af->list);
869 list_add_tail(&af->list, &sctp_address_families);
870 return 1;
873 /* Get the table of functions for manipulating a particular address
874 * family.
876 struct sctp_af *sctp_get_af_specific(sa_family_t family)
878 switch (family) {
879 case AF_INET:
880 return sctp_af_v4_specific;
881 case AF_INET6:
882 return sctp_af_v6_specific;
883 default:
884 return NULL;
888 /* Common code to initialize a AF_INET msg_name. */
889 static void sctp_inet_msgname(char *msgname, int *addr_len)
891 struct sockaddr_in *sin;
893 sin = (struct sockaddr_in *)msgname;
894 *addr_len = sizeof(struct sockaddr_in);
895 sin->sin_family = AF_INET;
896 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
899 /* Copy the primary address of the peer primary address as the msg_name. */
900 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
901 int *addr_len)
903 struct sockaddr_in *sin, *sinfrom;
905 if (msgname) {
906 struct sctp_association *asoc;
908 asoc = event->asoc;
909 sctp_inet_msgname(msgname, addr_len);
910 sin = (struct sockaddr_in *)msgname;
911 sinfrom = &asoc->peer.primary_addr.v4;
912 sin->sin_port = htons(asoc->peer.port);
913 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
917 /* Initialize and copy out a msgname from an inbound skb. */
918 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
920 if (msgname) {
921 struct sctphdr *sh = sctp_hdr(skb);
922 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
924 sctp_inet_msgname(msgname, len);
925 sin->sin_port = sh->source;
926 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
930 /* Do we support this AF? */
931 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
933 /* PF_INET only supports AF_INET addresses. */
934 return AF_INET == family;
937 /* Address matching with wildcards allowed. */
938 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
939 const union sctp_addr *addr2,
940 struct sctp_sock *opt)
942 /* PF_INET only supports AF_INET addresses. */
943 if (addr1->sa.sa_family != addr2->sa.sa_family)
944 return 0;
945 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
946 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
947 return 1;
948 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
949 return 1;
951 return 0;
954 /* Verify that provided sockaddr looks bindable. Common verification has
955 * already been taken care of.
957 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
959 return sctp_v4_available(addr, opt);
962 /* Verify that sockaddr looks sendable. Common verification has already
963 * been taken care of.
965 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
967 return 1;
970 /* Fill in Supported Address Type information for INIT and INIT-ACK
971 * chunks. Returns number of addresses supported.
973 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
974 __be16 *types)
976 types[0] = SCTP_PARAM_IPV4_ADDRESS;
977 return 1;
980 /* Wrapper routine that calls the ip transmit routine. */
981 static inline int sctp_v4_xmit(struct sk_buff *skb,
982 struct sctp_transport *transport)
984 struct inet_sock *inet = inet_sk(skb->sk);
986 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
987 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
989 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
990 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
992 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
994 return ip_queue_xmit(&inet->sk, skb, &transport->fl);
997 static struct sctp_af sctp_af_inet;
999 static struct sctp_pf sctp_pf_inet = {
1000 .event_msgname = sctp_inet_event_msgname,
1001 .skb_msgname = sctp_inet_skb_msgname,
1002 .af_supported = sctp_inet_af_supported,
1003 .cmp_addr = sctp_inet_cmp_addr,
1004 .bind_verify = sctp_inet_bind_verify,
1005 .send_verify = sctp_inet_send_verify,
1006 .supported_addrs = sctp_inet_supported_addrs,
1007 .create_accept_sk = sctp_v4_create_accept_sk,
1008 .addr_to_user = sctp_v4_addr_to_user,
1009 .to_sk_saddr = sctp_v4_to_sk_saddr,
1010 .to_sk_daddr = sctp_v4_to_sk_daddr,
1011 .af = &sctp_af_inet
1014 /* Notifier for inetaddr addition/deletion events. */
1015 static struct notifier_block sctp_inetaddr_notifier = {
1016 .notifier_call = sctp_inetaddr_event,
1019 /* Socket operations. */
1020 static const struct proto_ops inet_seqpacket_ops = {
1021 .family = PF_INET,
1022 .owner = THIS_MODULE,
1023 .release = inet_release, /* Needs to be wrapped... */
1024 .bind = inet_bind,
1025 .connect = inet_dgram_connect,
1026 .socketpair = sock_no_socketpair,
1027 .accept = inet_accept,
1028 .getname = inet_getname, /* Semantics are different. */
1029 .poll = sctp_poll,
1030 .ioctl = inet_ioctl,
1031 .listen = sctp_inet_listen,
1032 .shutdown = inet_shutdown, /* Looks harmless. */
1033 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1034 .getsockopt = sock_common_getsockopt,
1035 .sendmsg = inet_sendmsg,
1036 .recvmsg = inet_recvmsg,
1037 .mmap = sock_no_mmap,
1038 .sendpage = sock_no_sendpage,
1039 #ifdef CONFIG_COMPAT
1040 .compat_setsockopt = compat_sock_common_setsockopt,
1041 .compat_getsockopt = compat_sock_common_getsockopt,
1042 #endif
1045 /* Registration with AF_INET family. */
1046 static struct inet_protosw sctp_seqpacket_protosw = {
1047 .type = SOCK_SEQPACKET,
1048 .protocol = IPPROTO_SCTP,
1049 .prot = &sctp_prot,
1050 .ops = &inet_seqpacket_ops,
1051 .flags = SCTP_PROTOSW_FLAG
1053 static struct inet_protosw sctp_stream_protosw = {
1054 .type = SOCK_STREAM,
1055 .protocol = IPPROTO_SCTP,
1056 .prot = &sctp_prot,
1057 .ops = &inet_seqpacket_ops,
1058 .flags = SCTP_PROTOSW_FLAG
1061 /* Register with IP layer. */
1062 static const struct net_protocol sctp_protocol = {
1063 .handler = sctp_rcv,
1064 .err_handler = sctp_v4_err,
1065 .no_policy = 1,
1066 .netns_ok = 1,
1067 .icmp_strict_tag_validation = 1,
1070 /* IPv4 address related functions. */
1071 static struct sctp_af sctp_af_inet = {
1072 .sa_family = AF_INET,
1073 .sctp_xmit = sctp_v4_xmit,
1074 .setsockopt = ip_setsockopt,
1075 .getsockopt = ip_getsockopt,
1076 .get_dst = sctp_v4_get_dst,
1077 .get_saddr = sctp_v4_get_saddr,
1078 .copy_addrlist = sctp_v4_copy_addrlist,
1079 .from_skb = sctp_v4_from_skb,
1080 .from_sk = sctp_v4_from_sk,
1081 .from_addr_param = sctp_v4_from_addr_param,
1082 .to_addr_param = sctp_v4_to_addr_param,
1083 .cmp_addr = sctp_v4_cmp_addr,
1084 .addr_valid = sctp_v4_addr_valid,
1085 .inaddr_any = sctp_v4_inaddr_any,
1086 .is_any = sctp_v4_is_any,
1087 .available = sctp_v4_available,
1088 .scope = sctp_v4_scope,
1089 .skb_iif = sctp_v4_skb_iif,
1090 .is_ce = sctp_v4_is_ce,
1091 .seq_dump_addr = sctp_v4_seq_dump_addr,
1092 .ecn_capable = sctp_v4_ecn_capable,
1093 .net_header_len = sizeof(struct iphdr),
1094 .sockaddr_len = sizeof(struct sockaddr_in),
1095 #ifdef CONFIG_COMPAT
1096 .compat_setsockopt = compat_ip_setsockopt,
1097 .compat_getsockopt = compat_ip_getsockopt,
1098 #endif
1101 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1103 switch (family) {
1104 case PF_INET:
1105 return sctp_pf_inet_specific;
1106 case PF_INET6:
1107 return sctp_pf_inet6_specific;
1108 default:
1109 return NULL;
1113 /* Register the PF specific function table. */
1114 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1116 switch (family) {
1117 case PF_INET:
1118 if (sctp_pf_inet_specific)
1119 return 0;
1120 sctp_pf_inet_specific = pf;
1121 break;
1122 case PF_INET6:
1123 if (sctp_pf_inet6_specific)
1124 return 0;
1125 sctp_pf_inet6_specific = pf;
1126 break;
1127 default:
1128 return 0;
1130 return 1;
1133 static inline int init_sctp_mibs(struct net *net)
1135 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1136 if (!net->sctp.sctp_statistics)
1137 return -ENOMEM;
1138 return 0;
1141 static inline void cleanup_sctp_mibs(struct net *net)
1143 free_percpu(net->sctp.sctp_statistics);
1146 static void sctp_v4_pf_init(void)
1148 /* Initialize the SCTP specific PF functions. */
1149 sctp_register_pf(&sctp_pf_inet, PF_INET);
1150 sctp_register_af(&sctp_af_inet);
1153 static void sctp_v4_pf_exit(void)
1155 list_del(&sctp_af_inet.list);
1158 static int sctp_v4_protosw_init(void)
1160 int rc;
1162 rc = proto_register(&sctp_prot, 1);
1163 if (rc)
1164 return rc;
1166 /* Register SCTP(UDP and TCP style) with socket layer. */
1167 inet_register_protosw(&sctp_seqpacket_protosw);
1168 inet_register_protosw(&sctp_stream_protosw);
1170 return 0;
1173 static void sctp_v4_protosw_exit(void)
1175 inet_unregister_protosw(&sctp_stream_protosw);
1176 inet_unregister_protosw(&sctp_seqpacket_protosw);
1177 proto_unregister(&sctp_prot);
1180 static int sctp_v4_add_protocol(void)
1182 /* Register notifier for inet address additions/deletions. */
1183 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1185 /* Register SCTP with inet layer. */
1186 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1187 return -EAGAIN;
1189 return 0;
1192 static void sctp_v4_del_protocol(void)
1194 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1195 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1198 static int __net_init sctp_defaults_init(struct net *net)
1200 int status;
1203 * 14. Suggested SCTP Protocol Parameter Values
1205 /* The following protocol parameters are RECOMMENDED: */
1206 /* RTO.Initial - 3 seconds */
1207 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1208 /* RTO.Min - 1 second */
1209 net->sctp.rto_min = SCTP_RTO_MIN;
1210 /* RTO.Max - 60 seconds */
1211 net->sctp.rto_max = SCTP_RTO_MAX;
1212 /* RTO.Alpha - 1/8 */
1213 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1214 /* RTO.Beta - 1/4 */
1215 net->sctp.rto_beta = SCTP_RTO_BETA;
1217 /* Valid.Cookie.Life - 60 seconds */
1218 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1220 /* Whether Cookie Preservative is enabled(1) or not(0) */
1221 net->sctp.cookie_preserve_enable = 1;
1223 /* Default sctp sockets to use md5 as their hmac alg */
1224 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1225 net->sctp.sctp_hmac_alg = "md5";
1226 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1227 net->sctp.sctp_hmac_alg = "sha1";
1228 #else
1229 net->sctp.sctp_hmac_alg = NULL;
1230 #endif
1232 /* Max.Burst - 4 */
1233 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1235 /* Enable pf state by default */
1236 net->sctp.pf_enable = 1;
1238 /* Association.Max.Retrans - 10 attempts
1239 * Path.Max.Retrans - 5 attempts (per destination address)
1240 * Max.Init.Retransmits - 8 attempts
1242 net->sctp.max_retrans_association = 10;
1243 net->sctp.max_retrans_path = 5;
1244 net->sctp.max_retrans_init = 8;
1246 /* Sendbuffer growth - do per-socket accounting */
1247 net->sctp.sndbuf_policy = 0;
1249 /* Rcvbuffer growth - do per-socket accounting */
1250 net->sctp.rcvbuf_policy = 0;
1252 /* HB.interval - 30 seconds */
1253 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1255 /* delayed SACK timeout */
1256 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1258 /* Disable ADDIP by default. */
1259 net->sctp.addip_enable = 0;
1260 net->sctp.addip_noauth = 0;
1261 net->sctp.default_auto_asconf = 0;
1263 /* Enable PR-SCTP by default. */
1264 net->sctp.prsctp_enable = 1;
1266 /* Disable RECONF by default. */
1267 net->sctp.reconf_enable = 0;
1269 /* Disable AUTH by default. */
1270 net->sctp.auth_enable = 0;
1272 /* Set SCOPE policy to enabled */
1273 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1275 /* Set the default rwnd update threshold */
1276 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1278 /* Initialize maximum autoclose timeout. */
1279 net->sctp.max_autoclose = INT_MAX / HZ;
1281 status = sctp_sysctl_net_register(net);
1282 if (status)
1283 goto err_sysctl_register;
1285 /* Allocate and initialise sctp mibs. */
1286 status = init_sctp_mibs(net);
1287 if (status)
1288 goto err_init_mibs;
1290 /* Initialize proc fs directory. */
1291 status = sctp_proc_init(net);
1292 if (status)
1293 goto err_init_proc;
1295 sctp_dbg_objcnt_init(net);
1297 /* Initialize the local address list. */
1298 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1299 spin_lock_init(&net->sctp.local_addr_lock);
1300 sctp_get_local_addr_list(net);
1302 /* Initialize the address event list */
1303 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1304 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1305 spin_lock_init(&net->sctp.addr_wq_lock);
1306 net->sctp.addr_wq_timer.expires = 0;
1307 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1309 return 0;
1311 err_init_proc:
1312 cleanup_sctp_mibs(net);
1313 err_init_mibs:
1314 sctp_sysctl_net_unregister(net);
1315 err_sysctl_register:
1316 return status;
1319 static void __net_exit sctp_defaults_exit(struct net *net)
1321 /* Free the local address list */
1322 sctp_free_addr_wq(net);
1323 sctp_free_local_addr_list(net);
1325 sctp_dbg_objcnt_exit(net);
1327 sctp_proc_exit(net);
1328 cleanup_sctp_mibs(net);
1329 sctp_sysctl_net_unregister(net);
1332 static struct pernet_operations sctp_defaults_ops = {
1333 .init = sctp_defaults_init,
1334 .exit = sctp_defaults_exit,
1337 static int __net_init sctp_ctrlsock_init(struct net *net)
1339 int status;
1341 /* Initialize the control inode/socket for handling OOTB packets. */
1342 status = sctp_ctl_sock_init(net);
1343 if (status)
1344 pr_err("Failed to initialize the SCTP control sock\n");
1346 return status;
1349 static void __net_init sctp_ctrlsock_exit(struct net *net)
1351 /* Free the control endpoint. */
1352 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1355 static struct pernet_operations sctp_ctrlsock_ops = {
1356 .init = sctp_ctrlsock_init,
1357 .exit = sctp_ctrlsock_exit,
1360 /* Initialize the universe into something sensible. */
1361 static __init int sctp_init(void)
1363 int i;
1364 int status = -EINVAL;
1365 unsigned long goal;
1366 unsigned long limit;
1367 int max_share;
1368 int order;
1369 int num_entries;
1370 int max_entry_order;
1372 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1374 /* Allocate bind_bucket and chunk caches. */
1375 status = -ENOBUFS;
1376 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1377 sizeof(struct sctp_bind_bucket),
1378 0, SLAB_HWCACHE_ALIGN,
1379 NULL);
1380 if (!sctp_bucket_cachep)
1381 goto out;
1383 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1384 sizeof(struct sctp_chunk),
1385 0, SLAB_HWCACHE_ALIGN,
1386 NULL);
1387 if (!sctp_chunk_cachep)
1388 goto err_chunk_cachep;
1390 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1391 if (status)
1392 goto err_percpu_counter_init;
1394 /* Implementation specific variables. */
1396 /* Initialize default stream count setup information. */
1397 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1398 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1400 /* Initialize handle used for association ids. */
1401 idr_init(&sctp_assocs_id);
1403 limit = nr_free_buffer_pages() / 8;
1404 limit = max(limit, 128UL);
1405 sysctl_sctp_mem[0] = limit / 4 * 3;
1406 sysctl_sctp_mem[1] = limit;
1407 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1409 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1410 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1411 max_share = min(4UL*1024*1024, limit);
1413 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1414 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1415 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1417 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1418 sysctl_sctp_wmem[1] = 16*1024;
1419 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1421 /* Size and allocate the association hash table.
1422 * The methodology is similar to that of the tcp hash tables.
1423 * Though not identical. Start by getting a goal size
1425 if (totalram_pages >= (128 * 1024))
1426 goal = totalram_pages >> (22 - PAGE_SHIFT);
1427 else
1428 goal = totalram_pages >> (24 - PAGE_SHIFT);
1430 /* Then compute the page order for said goal */
1431 order = get_order(goal);
1433 /* Now compute the required page order for the maximum sized table we
1434 * want to create
1436 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1437 sizeof(struct sctp_bind_hashbucket));
1439 /* Limit the page order by that maximum hash table size */
1440 order = min(order, max_entry_order);
1442 /* Allocate and initialize the endpoint hash table. */
1443 sctp_ep_hashsize = 64;
1444 sctp_ep_hashtable =
1445 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1446 if (!sctp_ep_hashtable) {
1447 pr_err("Failed endpoint_hash alloc\n");
1448 status = -ENOMEM;
1449 goto err_ehash_alloc;
1451 for (i = 0; i < sctp_ep_hashsize; i++) {
1452 rwlock_init(&sctp_ep_hashtable[i].lock);
1453 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1456 /* Allocate and initialize the SCTP port hash table.
1457 * Note that order is initalized to start at the max sized
1458 * table we want to support. If we can't get that many pages
1459 * reduce the order and try again
1461 do {
1462 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1463 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1464 } while (!sctp_port_hashtable && --order > 0);
1466 if (!sctp_port_hashtable) {
1467 pr_err("Failed bind hash alloc\n");
1468 status = -ENOMEM;
1469 goto err_bhash_alloc;
1472 /* Now compute the number of entries that will fit in the
1473 * port hash space we allocated
1475 num_entries = (1UL << order) * PAGE_SIZE /
1476 sizeof(struct sctp_bind_hashbucket);
1478 /* And finish by rounding it down to the nearest power of two
1479 * this wastes some memory of course, but its needed because
1480 * the hash function operates based on the assumption that
1481 * that the number of entries is a power of two
1483 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1485 for (i = 0; i < sctp_port_hashsize; i++) {
1486 spin_lock_init(&sctp_port_hashtable[i].lock);
1487 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1490 status = sctp_transport_hashtable_init();
1491 if (status)
1492 goto err_thash_alloc;
1494 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1495 num_entries);
1497 sctp_sysctl_register();
1499 INIT_LIST_HEAD(&sctp_address_families);
1500 sctp_v4_pf_init();
1501 sctp_v6_pf_init();
1503 status = register_pernet_subsys(&sctp_defaults_ops);
1504 if (status)
1505 goto err_register_defaults;
1507 status = sctp_v4_protosw_init();
1508 if (status)
1509 goto err_protosw_init;
1511 status = sctp_v6_protosw_init();
1512 if (status)
1513 goto err_v6_protosw_init;
1515 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1516 if (status)
1517 goto err_register_ctrlsock;
1519 status = sctp_v4_add_protocol();
1520 if (status)
1521 goto err_add_protocol;
1523 /* Register SCTP with inet6 layer. */
1524 status = sctp_v6_add_protocol();
1525 if (status)
1526 goto err_v6_add_protocol;
1528 if (sctp_offload_init() < 0)
1529 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1531 out:
1532 return status;
1533 err_v6_add_protocol:
1534 sctp_v4_del_protocol();
1535 err_add_protocol:
1536 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1537 err_register_ctrlsock:
1538 sctp_v6_protosw_exit();
1539 err_v6_protosw_init:
1540 sctp_v4_protosw_exit();
1541 err_protosw_init:
1542 unregister_pernet_subsys(&sctp_defaults_ops);
1543 err_register_defaults:
1544 sctp_v4_pf_exit();
1545 sctp_v6_pf_exit();
1546 sctp_sysctl_unregister();
1547 free_pages((unsigned long)sctp_port_hashtable,
1548 get_order(sctp_port_hashsize *
1549 sizeof(struct sctp_bind_hashbucket)));
1550 err_bhash_alloc:
1551 sctp_transport_hashtable_destroy();
1552 err_thash_alloc:
1553 kfree(sctp_ep_hashtable);
1554 err_ehash_alloc:
1555 percpu_counter_destroy(&sctp_sockets_allocated);
1556 err_percpu_counter_init:
1557 kmem_cache_destroy(sctp_chunk_cachep);
1558 err_chunk_cachep:
1559 kmem_cache_destroy(sctp_bucket_cachep);
1560 goto out;
1563 /* Exit handler for the SCTP protocol. */
1564 static __exit void sctp_exit(void)
1566 /* BUG. This should probably do something useful like clean
1567 * up all the remaining associations and all that memory.
1570 /* Unregister with inet6/inet layers. */
1571 sctp_v6_del_protocol();
1572 sctp_v4_del_protocol();
1574 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1576 /* Free protosw registrations */
1577 sctp_v6_protosw_exit();
1578 sctp_v4_protosw_exit();
1580 unregister_pernet_subsys(&sctp_defaults_ops);
1582 /* Unregister with socket layer. */
1583 sctp_v6_pf_exit();
1584 sctp_v4_pf_exit();
1586 sctp_sysctl_unregister();
1588 free_pages((unsigned long)sctp_port_hashtable,
1589 get_order(sctp_port_hashsize *
1590 sizeof(struct sctp_bind_hashbucket)));
1591 kfree(sctp_ep_hashtable);
1592 sctp_transport_hashtable_destroy();
1594 percpu_counter_destroy(&sctp_sockets_allocated);
1596 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1598 kmem_cache_destroy(sctp_chunk_cachep);
1599 kmem_cache_destroy(sctp_bucket_cachep);
1602 module_init(sctp_init);
1603 module_exit(sctp_exit);
1606 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1608 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1609 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1610 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1611 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1612 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1613 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1614 MODULE_LICENSE("GPL");