Loading...
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
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * Initialization/cleanup for SCTP protocol support.
12 *
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.
18 *
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.
24 *
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/>.
28 *
29 * Please send any bug reports or fixes you make to the
30 * email address(es):
31 * lksctp developers <linux-sctp@vger.kernel.org>
32 *
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>
40 */
41
42#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
43
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>
62
63/* Global data structures. */
64struct sctp_globals sctp_globals __read_mostly;
65
66struct idr sctp_assocs_id;
67DEFINE_SPINLOCK(sctp_assocs_id_lock);
68
69static struct sctp_pf *sctp_pf_inet6_specific;
70static struct sctp_pf *sctp_pf_inet_specific;
71static struct sctp_af *sctp_af_v4_specific;
72static struct sctp_af *sctp_af_v6_specific;
73
74struct kmem_cache *sctp_chunk_cachep __read_mostly;
75struct kmem_cache *sctp_bucket_cachep __read_mostly;
76
77long sysctl_sctp_mem[3];
78int sysctl_sctp_rmem[3];
79int sysctl_sctp_wmem[3];
80
81/* Set up the proc fs entry for the SCTP protocol. */
82static int __net_init sctp_proc_init(struct net *net)
83{
84#ifdef CONFIG_PROC_FS
85 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
86 if (!net->sctp.proc_net_sctp)
87 goto out_proc_net_sctp;
88 if (sctp_snmp_proc_init(net))
89 goto out_snmp_proc_init;
90 if (sctp_eps_proc_init(net))
91 goto out_eps_proc_init;
92 if (sctp_assocs_proc_init(net))
93 goto out_assocs_proc_init;
94 if (sctp_remaddr_proc_init(net))
95 goto out_remaddr_proc_init;
96
97 return 0;
98
99out_remaddr_proc_init:
100 sctp_assocs_proc_exit(net);
101out_assocs_proc_init:
102 sctp_eps_proc_exit(net);
103out_eps_proc_init:
104 sctp_snmp_proc_exit(net);
105out_snmp_proc_init:
106 remove_proc_entry("sctp", net->proc_net);
107 net->sctp.proc_net_sctp = NULL;
108out_proc_net_sctp:
109 return -ENOMEM;
110#endif /* CONFIG_PROC_FS */
111 return 0;
112}
113
114/* Clean up the proc fs entry for the SCTP protocol.
115 * Note: Do not make this __exit as it is used in the init error
116 * path.
117 */
118static void sctp_proc_exit(struct net *net)
119{
120#ifdef CONFIG_PROC_FS
121 sctp_snmp_proc_exit(net);
122 sctp_eps_proc_exit(net);
123 sctp_assocs_proc_exit(net);
124 sctp_remaddr_proc_exit(net);
125
126 remove_proc_entry("sctp", net->proc_net);
127 net->sctp.proc_net_sctp = NULL;
128#endif
129}
130
131/* Private helper to extract ipv4 address and stash them in
132 * the protocol structure.
133 */
134static void sctp_v4_copy_addrlist(struct list_head *addrlist,
135 struct net_device *dev)
136{
137 struct in_device *in_dev;
138 struct in_ifaddr *ifa;
139 struct sctp_sockaddr_entry *addr;
140
141 rcu_read_lock();
142 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
143 rcu_read_unlock();
144 return;
145 }
146
147 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
148 /* Add the address to the local list. */
149 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
150 if (addr) {
151 addr->a.v4.sin_family = AF_INET;
152 addr->a.v4.sin_port = 0;
153 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
154 addr->valid = 1;
155 INIT_LIST_HEAD(&addr->list);
156 list_add_tail(&addr->list, addrlist);
157 }
158 }
159
160 rcu_read_unlock();
161}
162
163/* Extract our IP addresses from the system and stash them in the
164 * protocol structure.
165 */
166static void sctp_get_local_addr_list(struct net *net)
167{
168 struct net_device *dev;
169 struct list_head *pos;
170 struct sctp_af *af;
171
172 rcu_read_lock();
173 for_each_netdev_rcu(net, dev) {
174 list_for_each(pos, &sctp_address_families) {
175 af = list_entry(pos, struct sctp_af, list);
176 af->copy_addrlist(&net->sctp.local_addr_list, dev);
177 }
178 }
179 rcu_read_unlock();
180}
181
182/* Free the existing local addresses. */
183static void sctp_free_local_addr_list(struct net *net)
184{
185 struct sctp_sockaddr_entry *addr;
186 struct list_head *pos, *temp;
187
188 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
189 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
190 list_del(pos);
191 kfree(addr);
192 }
193}
194
195/* Copy the local addresses which are valid for 'scope' into 'bp'. */
196int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
197 sctp_scope_t scope, gfp_t gfp, int copy_flags)
198{
199 struct sctp_sockaddr_entry *addr;
200 int error = 0;
201
202 rcu_read_lock();
203 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
204 if (!addr->valid)
205 continue;
206 if (sctp_in_scope(net, &addr->a, scope)) {
207 /* Now that the address is in scope, check to see if
208 * the address type is really supported by the local
209 * sock as well as the remote peer.
210 */
211 if ((((AF_INET == addr->a.sa.sa_family) &&
212 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
213 (((AF_INET6 == addr->a.sa.sa_family) &&
214 (copy_flags & SCTP_ADDR6_ALLOWED) &&
215 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
216 error = sctp_add_bind_addr(bp, &addr->a,
217 SCTP_ADDR_SRC, GFP_ATOMIC);
218 if (error)
219 goto end_copy;
220 }
221 }
222 }
223
224end_copy:
225 rcu_read_unlock();
226 return error;
227}
228
229/* Initialize a sctp_addr from in incoming skb. */
230static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
231 int is_saddr)
232{
233 void *from;
234 __be16 *port;
235 struct sctphdr *sh;
236
237 port = &addr->v4.sin_port;
238 addr->v4.sin_family = AF_INET;
239
240 sh = sctp_hdr(skb);
241 if (is_saddr) {
242 *port = sh->source;
243 from = &ip_hdr(skb)->saddr;
244 } else {
245 *port = sh->dest;
246 from = &ip_hdr(skb)->daddr;
247 }
248 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
249}
250
251/* Initialize an sctp_addr from a socket. */
252static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
253{
254 addr->v4.sin_family = AF_INET;
255 addr->v4.sin_port = 0;
256 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
257}
258
259/* Initialize sk->sk_rcv_saddr from sctp_addr. */
260static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
261{
262 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
263}
264
265/* Initialize sk->sk_daddr from sctp_addr. */
266static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
267{
268 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
269}
270
271/* Initialize a sctp_addr from an address parameter. */
272static void sctp_v4_from_addr_param(union sctp_addr *addr,
273 union sctp_addr_param *param,
274 __be16 port, int iif)
275{
276 addr->v4.sin_family = AF_INET;
277 addr->v4.sin_port = port;
278 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
279}
280
281/* Initialize an address parameter from a sctp_addr and return the length
282 * of the address parameter.
283 */
284static int sctp_v4_to_addr_param(const union sctp_addr *addr,
285 union sctp_addr_param *param)
286{
287 int length = sizeof(sctp_ipv4addr_param_t);
288
289 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
290 param->v4.param_hdr.length = htons(length);
291 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
292
293 return length;
294}
295
296/* Initialize a sctp_addr from a dst_entry. */
297static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
298 __be16 port)
299{
300 saddr->v4.sin_family = AF_INET;
301 saddr->v4.sin_port = port;
302 saddr->v4.sin_addr.s_addr = fl4->saddr;
303}
304
305/* Compare two addresses exactly. */
306static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
307 const union sctp_addr *addr2)
308{
309 if (addr1->sa.sa_family != addr2->sa.sa_family)
310 return 0;
311 if (addr1->v4.sin_port != addr2->v4.sin_port)
312 return 0;
313 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
314 return 0;
315
316 return 1;
317}
318
319/* Initialize addr struct to INADDR_ANY. */
320static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
321{
322 addr->v4.sin_family = AF_INET;
323 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
324 addr->v4.sin_port = port;
325}
326
327/* Is this a wildcard address? */
328static int sctp_v4_is_any(const union sctp_addr *addr)
329{
330 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
331}
332
333/* This function checks if the address is a valid address to be used for
334 * SCTP binding.
335 *
336 * Output:
337 * Return 0 - If the address is a non-unicast or an illegal address.
338 * Return 1 - If the address is a unicast.
339 */
340static int sctp_v4_addr_valid(union sctp_addr *addr,
341 struct sctp_sock *sp,
342 const struct sk_buff *skb)
343{
344 /* IPv4 addresses not allowed */
345 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
346 return 0;
347
348 /* Is this a non-unicast address or a unusable SCTP address? */
349 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
350 return 0;
351
352 /* Is this a broadcast address? */
353 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
354 return 0;
355
356 return 1;
357}
358
359/* Should this be available for binding? */
360static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
361{
362 struct net *net = sock_net(&sp->inet.sk);
363 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
364
365
366 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
367 ret != RTN_LOCAL &&
368 !sp->inet.freebind &&
369 !sysctl_ip_nonlocal_bind)
370 return 0;
371
372 if (ipv6_only_sock(sctp_opt2sk(sp)))
373 return 0;
374
375 return 1;
376}
377
378/* Checking the loopback, private and other address scopes as defined in
379 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
380 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
381 *
382 * Level 0 - unusable SCTP addresses
383 * Level 1 - loopback address
384 * Level 2 - link-local addresses
385 * Level 3 - private addresses.
386 * Level 4 - global addresses
387 * For INIT and INIT-ACK address list, let L be the level of
388 * of requested destination address, sender and receiver
389 * SHOULD include all of its addresses with level greater
390 * than or equal to L.
391 *
392 * IPv4 scoping can be controlled through sysctl option
393 * net.sctp.addr_scope_policy
394 */
395static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
396{
397 sctp_scope_t retval;
398
399 /* Check for unusable SCTP addresses. */
400 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
401 retval = SCTP_SCOPE_UNUSABLE;
402 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
403 retval = SCTP_SCOPE_LOOPBACK;
404 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
405 retval = SCTP_SCOPE_LINK;
406 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
407 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
408 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
409 retval = SCTP_SCOPE_PRIVATE;
410 } else {
411 retval = SCTP_SCOPE_GLOBAL;
412 }
413
414 return retval;
415}
416
417/* Returns a valid dst cache entry for the given source and destination ip
418 * addresses. If an association is passed, trys to get a dst entry with a
419 * source address that matches an address in the bind address list.
420 */
421static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
422 struct flowi *fl, struct sock *sk)
423{
424 struct sctp_association *asoc = t->asoc;
425 struct rtable *rt;
426 struct flowi4 *fl4 = &fl->u.ip4;
427 struct sctp_bind_addr *bp;
428 struct sctp_sockaddr_entry *laddr;
429 struct dst_entry *dst = NULL;
430 union sctp_addr *daddr = &t->ipaddr;
431 union sctp_addr dst_saddr;
432
433 memset(fl4, 0x0, sizeof(struct flowi4));
434 fl4->daddr = daddr->v4.sin_addr.s_addr;
435 fl4->fl4_dport = daddr->v4.sin_port;
436 fl4->flowi4_proto = IPPROTO_SCTP;
437 if (asoc) {
438 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
439 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
440 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
441 }
442 if (saddr) {
443 fl4->saddr = saddr->v4.sin_addr.s_addr;
444 fl4->fl4_sport = saddr->v4.sin_port;
445 }
446
447 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
448 &fl4->saddr);
449
450 rt = ip_route_output_key(sock_net(sk), fl4);
451 if (!IS_ERR(rt))
452 dst = &rt->dst;
453
454 /* If there is no association or if a source address is passed, no
455 * more validation is required.
456 */
457 if (!asoc || saddr)
458 goto out;
459
460 bp = &asoc->base.bind_addr;
461
462 if (dst) {
463 /* Walk through the bind address list and look for a bind
464 * address that matches the source address of the returned dst.
465 */
466 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
467 rcu_read_lock();
468 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
469 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
470 (laddr->state != SCTP_ADDR_SRC &&
471 !asoc->src_out_of_asoc_ok))
472 continue;
473 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
474 goto out_unlock;
475 }
476 rcu_read_unlock();
477
478 /* None of the bound addresses match the source address of the
479 * dst. So release it.
480 */
481 dst_release(dst);
482 dst = NULL;
483 }
484
485 /* Walk through the bind address list and try to get a dst that
486 * matches a bind address as the source address.
487 */
488 rcu_read_lock();
489 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
490 if (!laddr->valid)
491 continue;
492 if ((laddr->state == SCTP_ADDR_SRC) &&
493 (AF_INET == laddr->a.sa.sa_family)) {
494 fl4->fl4_sport = laddr->a.v4.sin_port;
495 flowi4_update_output(fl4,
496 asoc->base.sk->sk_bound_dev_if,
497 RT_CONN_FLAGS(asoc->base.sk),
498 daddr->v4.sin_addr.s_addr,
499 laddr->a.v4.sin_addr.s_addr);
500
501 rt = ip_route_output_key(sock_net(sk), fl4);
502 if (!IS_ERR(rt)) {
503 dst = &rt->dst;
504 goto out_unlock;
505 }
506 }
507 }
508
509out_unlock:
510 rcu_read_unlock();
511out:
512 t->dst = dst;
513 if (dst)
514 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
515 &fl4->daddr, &fl4->saddr);
516 else
517 pr_debug("no route\n");
518}
519
520/* For v4, the source address is cached in the route entry(dst). So no need
521 * to cache it separately and hence this is an empty routine.
522 */
523static void sctp_v4_get_saddr(struct sctp_sock *sk,
524 struct sctp_transport *t,
525 struct flowi *fl)
526{
527 union sctp_addr *saddr = &t->saddr;
528 struct rtable *rt = (struct rtable *)t->dst;
529
530 if (rt) {
531 saddr->v4.sin_family = AF_INET;
532 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
533 }
534}
535
536/* What interface did this skb arrive on? */
537static int sctp_v4_skb_iif(const struct sk_buff *skb)
538{
539 return inet_iif(skb);
540}
541
542/* Was this packet marked by Explicit Congestion Notification? */
543static int sctp_v4_is_ce(const struct sk_buff *skb)
544{
545 return INET_ECN_is_ce(ip_hdr(skb)->tos);
546}
547
548/* Create and initialize a new sk for the socket returned by accept(). */
549static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
550 struct sctp_association *asoc)
551{
552 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
553 sk->sk_prot);
554 struct inet_sock *newinet;
555
556 if (!newsk)
557 goto out;
558
559 sock_init_data(NULL, newsk);
560
561 sctp_copy_sock(newsk, sk, asoc);
562 sock_reset_flag(newsk, SOCK_ZAPPED);
563
564 newinet = inet_sk(newsk);
565
566 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
567
568 sk_refcnt_debug_inc(newsk);
569
570 if (newsk->sk_prot->init(newsk)) {
571 sk_common_release(newsk);
572 newsk = NULL;
573 }
574
575out:
576 return newsk;
577}
578
579/* Map address, empty for v4 family */
580static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
581{
582 /* Empty */
583}
584
585/* Dump the v4 addr to the seq file. */
586static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
587{
588 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
589}
590
591static void sctp_v4_ecn_capable(struct sock *sk)
592{
593 INET_ECN_xmit(sk);
594}
595
596static void sctp_addr_wq_timeout_handler(unsigned long arg)
597{
598 struct net *net = (struct net *)arg;
599 struct sctp_sockaddr_entry *addrw, *temp;
600 struct sctp_sock *sp;
601
602 spin_lock_bh(&net->sctp.addr_wq_lock);
603
604 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
605 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
606 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
607 addrw->state, addrw);
608
609#if IS_ENABLED(CONFIG_IPV6)
610 /* Now we send an ASCONF for each association */
611 /* Note. we currently don't handle link local IPv6 addressees */
612 if (addrw->a.sa.sa_family == AF_INET6) {
613 struct in6_addr *in6;
614
615 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
616 IPV6_ADDR_LINKLOCAL)
617 goto free_next;
618
619 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
620 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
621 addrw->state == SCTP_ADDR_NEW) {
622 unsigned long timeo_val;
623
624 pr_debug("%s: this is on DAD, trying %d sec "
625 "later\n", __func__,
626 SCTP_ADDRESS_TICK_DELAY);
627
628 timeo_val = jiffies;
629 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
630 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
631 break;
632 }
633 }
634#endif
635 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
636 struct sock *sk;
637
638 sk = sctp_opt2sk(sp);
639 /* ignore bound-specific endpoints */
640 if (!sctp_is_ep_boundall(sk))
641 continue;
642 bh_lock_sock(sk);
643 if (sctp_asconf_mgmt(sp, addrw) < 0)
644 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
645 bh_unlock_sock(sk);
646 }
647#if IS_ENABLED(CONFIG_IPV6)
648free_next:
649#endif
650 list_del(&addrw->list);
651 kfree(addrw);
652 }
653 spin_unlock_bh(&net->sctp.addr_wq_lock);
654}
655
656static void sctp_free_addr_wq(struct net *net)
657{
658 struct sctp_sockaddr_entry *addrw;
659 struct sctp_sockaddr_entry *temp;
660
661 spin_lock_bh(&net->sctp.addr_wq_lock);
662 del_timer(&net->sctp.addr_wq_timer);
663 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
664 list_del(&addrw->list);
665 kfree(addrw);
666 }
667 spin_unlock_bh(&net->sctp.addr_wq_lock);
668}
669
670/* lookup the entry for the same address in the addr_waitq
671 * sctp_addr_wq MUST be locked
672 */
673static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
674 struct sctp_sockaddr_entry *addr)
675{
676 struct sctp_sockaddr_entry *addrw;
677
678 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
679 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
680 continue;
681 if (addrw->a.sa.sa_family == AF_INET) {
682 if (addrw->a.v4.sin_addr.s_addr ==
683 addr->a.v4.sin_addr.s_addr)
684 return addrw;
685 } else if (addrw->a.sa.sa_family == AF_INET6) {
686 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
687 &addr->a.v6.sin6_addr))
688 return addrw;
689 }
690 }
691 return NULL;
692}
693
694void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
695{
696 struct sctp_sockaddr_entry *addrw;
697 unsigned long timeo_val;
698
699 /* first, we check if an opposite message already exist in the queue.
700 * If we found such message, it is removed.
701 * This operation is a bit stupid, but the DHCP client attaches the
702 * new address after a couple of addition and deletion of that address
703 */
704
705 spin_lock_bh(&net->sctp.addr_wq_lock);
706 /* Offsets existing events in addr_wq */
707 addrw = sctp_addr_wq_lookup(net, addr);
708 if (addrw) {
709 if (addrw->state != cmd) {
710 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
711 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
712 &net->sctp.addr_waitq);
713
714 list_del(&addrw->list);
715 kfree(addrw);
716 }
717 spin_unlock_bh(&net->sctp.addr_wq_lock);
718 return;
719 }
720
721 /* OK, we have to add the new address to the wait queue */
722 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
723 if (addrw == NULL) {
724 spin_unlock_bh(&net->sctp.addr_wq_lock);
725 return;
726 }
727 addrw->state = cmd;
728 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
729
730 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
731 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
732
733 if (!timer_pending(&net->sctp.addr_wq_timer)) {
734 timeo_val = jiffies;
735 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
736 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
737 }
738 spin_unlock_bh(&net->sctp.addr_wq_lock);
739}
740
741/* Event handler for inet address addition/deletion events.
742 * The sctp_local_addr_list needs to be protocted by a spin lock since
743 * multiple notifiers (say IPv4 and IPv6) may be running at the same
744 * time and thus corrupt the list.
745 * The reader side is protected with RCU.
746 */
747static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
748 void *ptr)
749{
750 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
751 struct sctp_sockaddr_entry *addr = NULL;
752 struct sctp_sockaddr_entry *temp;
753 struct net *net = dev_net(ifa->ifa_dev->dev);
754 int found = 0;
755
756 switch (ev) {
757 case NETDEV_UP:
758 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
759 if (addr) {
760 addr->a.v4.sin_family = AF_INET;
761 addr->a.v4.sin_port = 0;
762 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
763 addr->valid = 1;
764 spin_lock_bh(&net->sctp.local_addr_lock);
765 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
766 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
767 spin_unlock_bh(&net->sctp.local_addr_lock);
768 }
769 break;
770 case NETDEV_DOWN:
771 spin_lock_bh(&net->sctp.local_addr_lock);
772 list_for_each_entry_safe(addr, temp,
773 &net->sctp.local_addr_list, list) {
774 if (addr->a.sa.sa_family == AF_INET &&
775 addr->a.v4.sin_addr.s_addr ==
776 ifa->ifa_local) {
777 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
778 found = 1;
779 addr->valid = 0;
780 list_del_rcu(&addr->list);
781 break;
782 }
783 }
784 spin_unlock_bh(&net->sctp.local_addr_lock);
785 if (found)
786 kfree_rcu(addr, rcu);
787 break;
788 }
789
790 return NOTIFY_DONE;
791}
792
793/*
794 * Initialize the control inode/socket with a control endpoint data
795 * structure. This endpoint is reserved exclusively for the OOTB processing.
796 */
797static int sctp_ctl_sock_init(struct net *net)
798{
799 int err;
800 sa_family_t family = PF_INET;
801
802 if (sctp_get_pf_specific(PF_INET6))
803 family = PF_INET6;
804
805 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
806 SOCK_SEQPACKET, IPPROTO_SCTP, net);
807
808 /* If IPv6 socket could not be created, try the IPv4 socket */
809 if (err < 0 && family == PF_INET6)
810 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
811 SOCK_SEQPACKET, IPPROTO_SCTP,
812 net);
813
814 if (err < 0) {
815 pr_err("Failed to create the SCTP control socket\n");
816 return err;
817 }
818 return 0;
819}
820
821/* Register address family specific functions. */
822int sctp_register_af(struct sctp_af *af)
823{
824 switch (af->sa_family) {
825 case AF_INET:
826 if (sctp_af_v4_specific)
827 return 0;
828 sctp_af_v4_specific = af;
829 break;
830 case AF_INET6:
831 if (sctp_af_v6_specific)
832 return 0;
833 sctp_af_v6_specific = af;
834 break;
835 default:
836 return 0;
837 }
838
839 INIT_LIST_HEAD(&af->list);
840 list_add_tail(&af->list, &sctp_address_families);
841 return 1;
842}
843
844/* Get the table of functions for manipulating a particular address
845 * family.
846 */
847struct sctp_af *sctp_get_af_specific(sa_family_t family)
848{
849 switch (family) {
850 case AF_INET:
851 return sctp_af_v4_specific;
852 case AF_INET6:
853 return sctp_af_v6_specific;
854 default:
855 return NULL;
856 }
857}
858
859/* Common code to initialize a AF_INET msg_name. */
860static void sctp_inet_msgname(char *msgname, int *addr_len)
861{
862 struct sockaddr_in *sin;
863
864 sin = (struct sockaddr_in *)msgname;
865 *addr_len = sizeof(struct sockaddr_in);
866 sin->sin_family = AF_INET;
867 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
868}
869
870/* Copy the primary address of the peer primary address as the msg_name. */
871static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
872 int *addr_len)
873{
874 struct sockaddr_in *sin, *sinfrom;
875
876 if (msgname) {
877 struct sctp_association *asoc;
878
879 asoc = event->asoc;
880 sctp_inet_msgname(msgname, addr_len);
881 sin = (struct sockaddr_in *)msgname;
882 sinfrom = &asoc->peer.primary_addr.v4;
883 sin->sin_port = htons(asoc->peer.port);
884 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
885 }
886}
887
888/* Initialize and copy out a msgname from an inbound skb. */
889static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
890{
891 if (msgname) {
892 struct sctphdr *sh = sctp_hdr(skb);
893 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
894
895 sctp_inet_msgname(msgname, len);
896 sin->sin_port = sh->source;
897 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
898 }
899}
900
901/* Do we support this AF? */
902static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
903{
904 /* PF_INET only supports AF_INET addresses. */
905 return AF_INET == family;
906}
907
908/* Address matching with wildcards allowed. */
909static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
910 const union sctp_addr *addr2,
911 struct sctp_sock *opt)
912{
913 /* PF_INET only supports AF_INET addresses. */
914 if (addr1->sa.sa_family != addr2->sa.sa_family)
915 return 0;
916 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
917 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
918 return 1;
919 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
920 return 1;
921
922 return 0;
923}
924
925/* Verify that provided sockaddr looks bindable. Common verification has
926 * already been taken care of.
927 */
928static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
929{
930 return sctp_v4_available(addr, opt);
931}
932
933/* Verify that sockaddr looks sendable. Common verification has already
934 * been taken care of.
935 */
936static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
937{
938 return 1;
939}
940
941/* Fill in Supported Address Type information for INIT and INIT-ACK
942 * chunks. Returns number of addresses supported.
943 */
944static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
945 __be16 *types)
946{
947 types[0] = SCTP_PARAM_IPV4_ADDRESS;
948 return 1;
949}
950
951/* Wrapper routine that calls the ip transmit routine. */
952static inline int sctp_v4_xmit(struct sk_buff *skb,
953 struct sctp_transport *transport)
954{
955 struct inet_sock *inet = inet_sk(skb->sk);
956
957 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
958 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
959
960 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
961 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
962
963 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
964
965 return ip_queue_xmit(&inet->sk, skb, &transport->fl);
966}
967
968static struct sctp_af sctp_af_inet;
969
970static struct sctp_pf sctp_pf_inet = {
971 .event_msgname = sctp_inet_event_msgname,
972 .skb_msgname = sctp_inet_skb_msgname,
973 .af_supported = sctp_inet_af_supported,
974 .cmp_addr = sctp_inet_cmp_addr,
975 .bind_verify = sctp_inet_bind_verify,
976 .send_verify = sctp_inet_send_verify,
977 .supported_addrs = sctp_inet_supported_addrs,
978 .create_accept_sk = sctp_v4_create_accept_sk,
979 .addr_v4map = sctp_v4_addr_v4map,
980 .af = &sctp_af_inet
981};
982
983/* Notifier for inetaddr addition/deletion events. */
984static struct notifier_block sctp_inetaddr_notifier = {
985 .notifier_call = sctp_inetaddr_event,
986};
987
988/* Socket operations. */
989static const struct proto_ops inet_seqpacket_ops = {
990 .family = PF_INET,
991 .owner = THIS_MODULE,
992 .release = inet_release, /* Needs to be wrapped... */
993 .bind = inet_bind,
994 .connect = inet_dgram_connect,
995 .socketpair = sock_no_socketpair,
996 .accept = inet_accept,
997 .getname = inet_getname, /* Semantics are different. */
998 .poll = sctp_poll,
999 .ioctl = inet_ioctl,
1000 .listen = sctp_inet_listen,
1001 .shutdown = inet_shutdown, /* Looks harmless. */
1002 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1003 .getsockopt = sock_common_getsockopt,
1004 .sendmsg = inet_sendmsg,
1005 .recvmsg = sock_common_recvmsg,
1006 .mmap = sock_no_mmap,
1007 .sendpage = sock_no_sendpage,
1008#ifdef CONFIG_COMPAT
1009 .compat_setsockopt = compat_sock_common_setsockopt,
1010 .compat_getsockopt = compat_sock_common_getsockopt,
1011#endif
1012};
1013
1014/* Registration with AF_INET family. */
1015static struct inet_protosw sctp_seqpacket_protosw = {
1016 .type = SOCK_SEQPACKET,
1017 .protocol = IPPROTO_SCTP,
1018 .prot = &sctp_prot,
1019 .ops = &inet_seqpacket_ops,
1020 .no_check = 0,
1021 .flags = SCTP_PROTOSW_FLAG
1022};
1023static struct inet_protosw sctp_stream_protosw = {
1024 .type = SOCK_STREAM,
1025 .protocol = IPPROTO_SCTP,
1026 .prot = &sctp_prot,
1027 .ops = &inet_seqpacket_ops,
1028 .no_check = 0,
1029 .flags = SCTP_PROTOSW_FLAG
1030};
1031
1032/* Register with IP layer. */
1033static const struct net_protocol sctp_protocol = {
1034 .handler = sctp_rcv,
1035 .err_handler = sctp_v4_err,
1036 .no_policy = 1,
1037 .netns_ok = 1,
1038 .icmp_strict_tag_validation = 1,
1039};
1040
1041/* IPv4 address related functions. */
1042static struct sctp_af sctp_af_inet = {
1043 .sa_family = AF_INET,
1044 .sctp_xmit = sctp_v4_xmit,
1045 .setsockopt = ip_setsockopt,
1046 .getsockopt = ip_getsockopt,
1047 .get_dst = sctp_v4_get_dst,
1048 .get_saddr = sctp_v4_get_saddr,
1049 .copy_addrlist = sctp_v4_copy_addrlist,
1050 .from_skb = sctp_v4_from_skb,
1051 .from_sk = sctp_v4_from_sk,
1052 .to_sk_saddr = sctp_v4_to_sk_saddr,
1053 .to_sk_daddr = sctp_v4_to_sk_daddr,
1054 .from_addr_param = sctp_v4_from_addr_param,
1055 .to_addr_param = sctp_v4_to_addr_param,
1056 .cmp_addr = sctp_v4_cmp_addr,
1057 .addr_valid = sctp_v4_addr_valid,
1058 .inaddr_any = sctp_v4_inaddr_any,
1059 .is_any = sctp_v4_is_any,
1060 .available = sctp_v4_available,
1061 .scope = sctp_v4_scope,
1062 .skb_iif = sctp_v4_skb_iif,
1063 .is_ce = sctp_v4_is_ce,
1064 .seq_dump_addr = sctp_v4_seq_dump_addr,
1065 .ecn_capable = sctp_v4_ecn_capable,
1066 .net_header_len = sizeof(struct iphdr),
1067 .sockaddr_len = sizeof(struct sockaddr_in),
1068#ifdef CONFIG_COMPAT
1069 .compat_setsockopt = compat_ip_setsockopt,
1070 .compat_getsockopt = compat_ip_getsockopt,
1071#endif
1072};
1073
1074struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1075{
1076 switch (family) {
1077 case PF_INET:
1078 return sctp_pf_inet_specific;
1079 case PF_INET6:
1080 return sctp_pf_inet6_specific;
1081 default:
1082 return NULL;
1083 }
1084}
1085
1086/* Register the PF specific function table. */
1087int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1088{
1089 switch (family) {
1090 case PF_INET:
1091 if (sctp_pf_inet_specific)
1092 return 0;
1093 sctp_pf_inet_specific = pf;
1094 break;
1095 case PF_INET6:
1096 if (sctp_pf_inet6_specific)
1097 return 0;
1098 sctp_pf_inet6_specific = pf;
1099 break;
1100 default:
1101 return 0;
1102 }
1103 return 1;
1104}
1105
1106static inline int init_sctp_mibs(struct net *net)
1107{
1108 return snmp_mib_init((void __percpu **)net->sctp.sctp_statistics,
1109 sizeof(struct sctp_mib),
1110 __alignof__(struct sctp_mib));
1111}
1112
1113static inline void cleanup_sctp_mibs(struct net *net)
1114{
1115 snmp_mib_free((void __percpu **)net->sctp.sctp_statistics);
1116}
1117
1118static void sctp_v4_pf_init(void)
1119{
1120 /* Initialize the SCTP specific PF functions. */
1121 sctp_register_pf(&sctp_pf_inet, PF_INET);
1122 sctp_register_af(&sctp_af_inet);
1123}
1124
1125static void sctp_v4_pf_exit(void)
1126{
1127 list_del(&sctp_af_inet.list);
1128}
1129
1130static int sctp_v4_protosw_init(void)
1131{
1132 int rc;
1133
1134 rc = proto_register(&sctp_prot, 1);
1135 if (rc)
1136 return rc;
1137
1138 /* Register SCTP(UDP and TCP style) with socket layer. */
1139 inet_register_protosw(&sctp_seqpacket_protosw);
1140 inet_register_protosw(&sctp_stream_protosw);
1141
1142 return 0;
1143}
1144
1145static void sctp_v4_protosw_exit(void)
1146{
1147 inet_unregister_protosw(&sctp_stream_protosw);
1148 inet_unregister_protosw(&sctp_seqpacket_protosw);
1149 proto_unregister(&sctp_prot);
1150}
1151
1152static int sctp_v4_add_protocol(void)
1153{
1154 /* Register notifier for inet address additions/deletions. */
1155 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1156
1157 /* Register SCTP with inet layer. */
1158 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1159 return -EAGAIN;
1160
1161 return 0;
1162}
1163
1164static void sctp_v4_del_protocol(void)
1165{
1166 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1167 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1168}
1169
1170static int __net_init sctp_net_init(struct net *net)
1171{
1172 int status;
1173
1174 /*
1175 * 14. Suggested SCTP Protocol Parameter Values
1176 */
1177 /* The following protocol parameters are RECOMMENDED: */
1178 /* RTO.Initial - 3 seconds */
1179 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1180 /* RTO.Min - 1 second */
1181 net->sctp.rto_min = SCTP_RTO_MIN;
1182 /* RTO.Max - 60 seconds */
1183 net->sctp.rto_max = SCTP_RTO_MAX;
1184 /* RTO.Alpha - 1/8 */
1185 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1186 /* RTO.Beta - 1/4 */
1187 net->sctp.rto_beta = SCTP_RTO_BETA;
1188
1189 /* Valid.Cookie.Life - 60 seconds */
1190 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1191
1192 /* Whether Cookie Preservative is enabled(1) or not(0) */
1193 net->sctp.cookie_preserve_enable = 1;
1194
1195 /* Default sctp sockets to use md5 as their hmac alg */
1196#if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1197 net->sctp.sctp_hmac_alg = "md5";
1198#elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1199 net->sctp.sctp_hmac_alg = "sha1";
1200#else
1201 net->sctp.sctp_hmac_alg = NULL;
1202#endif
1203
1204 /* Max.Burst - 4 */
1205 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1206
1207 /* Association.Max.Retrans - 10 attempts
1208 * Path.Max.Retrans - 5 attempts (per destination address)
1209 * Max.Init.Retransmits - 8 attempts
1210 */
1211 net->sctp.max_retrans_association = 10;
1212 net->sctp.max_retrans_path = 5;
1213 net->sctp.max_retrans_init = 8;
1214
1215 /* Sendbuffer growth - do per-socket accounting */
1216 net->sctp.sndbuf_policy = 0;
1217
1218 /* Rcvbuffer growth - do per-socket accounting */
1219 net->sctp.rcvbuf_policy = 0;
1220
1221 /* HB.interval - 30 seconds */
1222 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1223
1224 /* delayed SACK timeout */
1225 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1226
1227 /* Disable ADDIP by default. */
1228 net->sctp.addip_enable = 0;
1229 net->sctp.addip_noauth = 0;
1230 net->sctp.default_auto_asconf = 0;
1231
1232 /* Enable PR-SCTP by default. */
1233 net->sctp.prsctp_enable = 1;
1234
1235 /* Disable AUTH by default. */
1236 net->sctp.auth_enable = 0;
1237
1238 /* Set SCOPE policy to enabled */
1239 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1240
1241 /* Set the default rwnd update threshold */
1242 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1243
1244 /* Initialize maximum autoclose timeout. */
1245 net->sctp.max_autoclose = INT_MAX / HZ;
1246
1247 status = sctp_sysctl_net_register(net);
1248 if (status)
1249 goto err_sysctl_register;
1250
1251 /* Allocate and initialise sctp mibs. */
1252 status = init_sctp_mibs(net);
1253 if (status)
1254 goto err_init_mibs;
1255
1256 /* Initialize proc fs directory. */
1257 status = sctp_proc_init(net);
1258 if (status)
1259 goto err_init_proc;
1260
1261 sctp_dbg_objcnt_init(net);
1262
1263 /* Initialize the control inode/socket for handling OOTB packets. */
1264 if ((status = sctp_ctl_sock_init(net))) {
1265 pr_err("Failed to initialize the SCTP control sock\n");
1266 goto err_ctl_sock_init;
1267 }
1268
1269 /* Initialize the local address list. */
1270 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1271 spin_lock_init(&net->sctp.local_addr_lock);
1272 sctp_get_local_addr_list(net);
1273
1274 /* Initialize the address event list */
1275 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1276 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1277 spin_lock_init(&net->sctp.addr_wq_lock);
1278 net->sctp.addr_wq_timer.expires = 0;
1279 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1280 (unsigned long)net);
1281
1282 return 0;
1283
1284err_ctl_sock_init:
1285 sctp_dbg_objcnt_exit(net);
1286 sctp_proc_exit(net);
1287err_init_proc:
1288 cleanup_sctp_mibs(net);
1289err_init_mibs:
1290 sctp_sysctl_net_unregister(net);
1291err_sysctl_register:
1292 return status;
1293}
1294
1295static void __net_exit sctp_net_exit(struct net *net)
1296{
1297 /* Free the local address list */
1298 sctp_free_addr_wq(net);
1299 sctp_free_local_addr_list(net);
1300
1301 /* Free the control endpoint. */
1302 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1303
1304 sctp_dbg_objcnt_exit(net);
1305
1306 sctp_proc_exit(net);
1307 cleanup_sctp_mibs(net);
1308 sctp_sysctl_net_unregister(net);
1309}
1310
1311static struct pernet_operations sctp_net_ops = {
1312 .init = sctp_net_init,
1313 .exit = sctp_net_exit,
1314};
1315
1316/* Initialize the universe into something sensible. */
1317static __init int sctp_init(void)
1318{
1319 int i;
1320 int status = -EINVAL;
1321 unsigned long goal;
1322 unsigned long limit;
1323 int max_share;
1324 int order;
1325
1326 BUILD_BUG_ON(sizeof(struct sctp_ulpevent) >
1327 sizeof(((struct sk_buff *) 0)->cb));
1328
1329 /* Allocate bind_bucket and chunk caches. */
1330 status = -ENOBUFS;
1331 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1332 sizeof(struct sctp_bind_bucket),
1333 0, SLAB_HWCACHE_ALIGN,
1334 NULL);
1335 if (!sctp_bucket_cachep)
1336 goto out;
1337
1338 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1339 sizeof(struct sctp_chunk),
1340 0, SLAB_HWCACHE_ALIGN,
1341 NULL);
1342 if (!sctp_chunk_cachep)
1343 goto err_chunk_cachep;
1344
1345 status = percpu_counter_init(&sctp_sockets_allocated, 0);
1346 if (status)
1347 goto err_percpu_counter_init;
1348
1349 /* Implementation specific variables. */
1350
1351 /* Initialize default stream count setup information. */
1352 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1353 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1354
1355 /* Initialize handle used for association ids. */
1356 idr_init(&sctp_assocs_id);
1357
1358 limit = nr_free_buffer_pages() / 8;
1359 limit = max(limit, 128UL);
1360 sysctl_sctp_mem[0] = limit / 4 * 3;
1361 sysctl_sctp_mem[1] = limit;
1362 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1363
1364 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1365 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1366 max_share = min(4UL*1024*1024, limit);
1367
1368 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1369 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1370 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1371
1372 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1373 sysctl_sctp_wmem[1] = 16*1024;
1374 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1375
1376 /* Size and allocate the association hash table.
1377 * The methodology is similar to that of the tcp hash tables.
1378 */
1379 if (totalram_pages >= (128 * 1024))
1380 goal = totalram_pages >> (22 - PAGE_SHIFT);
1381 else
1382 goal = totalram_pages >> (24 - PAGE_SHIFT);
1383
1384 for (order = 0; (1UL << order) < goal; order++)
1385 ;
1386
1387 do {
1388 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1389 sizeof(struct sctp_hashbucket);
1390 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1391 continue;
1392 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1393 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1394 } while (!sctp_assoc_hashtable && --order > 0);
1395 if (!sctp_assoc_hashtable) {
1396 pr_err("Failed association hash alloc\n");
1397 status = -ENOMEM;
1398 goto err_ahash_alloc;
1399 }
1400 for (i = 0; i < sctp_assoc_hashsize; i++) {
1401 rwlock_init(&sctp_assoc_hashtable[i].lock);
1402 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1403 }
1404
1405 /* Allocate and initialize the endpoint hash table. */
1406 sctp_ep_hashsize = 64;
1407 sctp_ep_hashtable =
1408 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1409 if (!sctp_ep_hashtable) {
1410 pr_err("Failed endpoint_hash alloc\n");
1411 status = -ENOMEM;
1412 goto err_ehash_alloc;
1413 }
1414 for (i = 0; i < sctp_ep_hashsize; i++) {
1415 rwlock_init(&sctp_ep_hashtable[i].lock);
1416 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1417 }
1418
1419 /* Allocate and initialize the SCTP port hash table. */
1420 do {
1421 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1422 sizeof(struct sctp_bind_hashbucket);
1423 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1424 continue;
1425 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1426 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1427 } while (!sctp_port_hashtable && --order > 0);
1428 if (!sctp_port_hashtable) {
1429 pr_err("Failed bind hash alloc\n");
1430 status = -ENOMEM;
1431 goto err_bhash_alloc;
1432 }
1433 for (i = 0; i < sctp_port_hashsize; i++) {
1434 spin_lock_init(&sctp_port_hashtable[i].lock);
1435 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1436 }
1437
1438 pr_info("Hash tables configured (established %d bind %d)\n",
1439 sctp_assoc_hashsize, sctp_port_hashsize);
1440
1441 sctp_sysctl_register();
1442
1443 INIT_LIST_HEAD(&sctp_address_families);
1444 sctp_v4_pf_init();
1445 sctp_v6_pf_init();
1446
1447 status = sctp_v4_protosw_init();
1448
1449 if (status)
1450 goto err_protosw_init;
1451
1452 status = sctp_v6_protosw_init();
1453 if (status)
1454 goto err_v6_protosw_init;
1455
1456 status = register_pernet_subsys(&sctp_net_ops);
1457 if (status)
1458 goto err_register_pernet_subsys;
1459
1460 status = sctp_v4_add_protocol();
1461 if (status)
1462 goto err_add_protocol;
1463
1464 /* Register SCTP with inet6 layer. */
1465 status = sctp_v6_add_protocol();
1466 if (status)
1467 goto err_v6_add_protocol;
1468
1469out:
1470 return status;
1471err_v6_add_protocol:
1472 sctp_v4_del_protocol();
1473err_add_protocol:
1474 unregister_pernet_subsys(&sctp_net_ops);
1475err_register_pernet_subsys:
1476 sctp_v6_protosw_exit();
1477err_v6_protosw_init:
1478 sctp_v4_protosw_exit();
1479err_protosw_init:
1480 sctp_v4_pf_exit();
1481 sctp_v6_pf_exit();
1482 sctp_sysctl_unregister();
1483 free_pages((unsigned long)sctp_port_hashtable,
1484 get_order(sctp_port_hashsize *
1485 sizeof(struct sctp_bind_hashbucket)));
1486err_bhash_alloc:
1487 kfree(sctp_ep_hashtable);
1488err_ehash_alloc:
1489 free_pages((unsigned long)sctp_assoc_hashtable,
1490 get_order(sctp_assoc_hashsize *
1491 sizeof(struct sctp_hashbucket)));
1492err_ahash_alloc:
1493 percpu_counter_destroy(&sctp_sockets_allocated);
1494err_percpu_counter_init:
1495 kmem_cache_destroy(sctp_chunk_cachep);
1496err_chunk_cachep:
1497 kmem_cache_destroy(sctp_bucket_cachep);
1498 goto out;
1499}
1500
1501/* Exit handler for the SCTP protocol. */
1502static __exit void sctp_exit(void)
1503{
1504 /* BUG. This should probably do something useful like clean
1505 * up all the remaining associations and all that memory.
1506 */
1507
1508 /* Unregister with inet6/inet layers. */
1509 sctp_v6_del_protocol();
1510 sctp_v4_del_protocol();
1511
1512 unregister_pernet_subsys(&sctp_net_ops);
1513
1514 /* Free protosw registrations */
1515 sctp_v6_protosw_exit();
1516 sctp_v4_protosw_exit();
1517
1518 /* Unregister with socket layer. */
1519 sctp_v6_pf_exit();
1520 sctp_v4_pf_exit();
1521
1522 sctp_sysctl_unregister();
1523
1524 free_pages((unsigned long)sctp_assoc_hashtable,
1525 get_order(sctp_assoc_hashsize *
1526 sizeof(struct sctp_hashbucket)));
1527 kfree(sctp_ep_hashtable);
1528 free_pages((unsigned long)sctp_port_hashtable,
1529 get_order(sctp_port_hashsize *
1530 sizeof(struct sctp_bind_hashbucket)));
1531
1532 percpu_counter_destroy(&sctp_sockets_allocated);
1533
1534 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1535
1536 kmem_cache_destroy(sctp_chunk_cachep);
1537 kmem_cache_destroy(sctp_bucket_cachep);
1538}
1539
1540module_init(sctp_init);
1541module_exit(sctp_exit);
1542
1543/*
1544 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1545 */
1546MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1547MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1548MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1549MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1550module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1551MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1552MODULE_LICENSE("GPL");
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
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * Initialization/cleanup for SCTP protocol support.
12 *
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.
18 *
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.
24 *
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/>.
28 *
29 * Please send any bug reports or fixes you make to the
30 * email address(es):
31 * lksctp developers <linux-sctp@vger.kernel.org>
32 *
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>
40 */
41
42#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
43
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>
62
63#define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
64
65/* Global data structures. */
66struct sctp_globals sctp_globals __read_mostly;
67
68struct idr sctp_assocs_id;
69DEFINE_SPINLOCK(sctp_assocs_id_lock);
70
71static struct sctp_pf *sctp_pf_inet6_specific;
72static struct sctp_pf *sctp_pf_inet_specific;
73static struct sctp_af *sctp_af_v4_specific;
74static struct sctp_af *sctp_af_v6_specific;
75
76struct kmem_cache *sctp_chunk_cachep __read_mostly;
77struct kmem_cache *sctp_bucket_cachep __read_mostly;
78
79long sysctl_sctp_mem[3];
80int sysctl_sctp_rmem[3];
81int sysctl_sctp_wmem[3];
82
83/* Private helper to extract ipv4 address and stash them in
84 * the protocol structure.
85 */
86static void sctp_v4_copy_addrlist(struct list_head *addrlist,
87 struct net_device *dev)
88{
89 struct in_device *in_dev;
90 struct in_ifaddr *ifa;
91 struct sctp_sockaddr_entry *addr;
92
93 rcu_read_lock();
94 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
95 rcu_read_unlock();
96 return;
97 }
98
99 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
100 /* Add the address to the local list. */
101 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
102 if (addr) {
103 addr->a.v4.sin_family = AF_INET;
104 addr->a.v4.sin_port = 0;
105 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
106 addr->valid = 1;
107 INIT_LIST_HEAD(&addr->list);
108 list_add_tail(&addr->list, addrlist);
109 }
110 }
111
112 rcu_read_unlock();
113}
114
115/* Extract our IP addresses from the system and stash them in the
116 * protocol structure.
117 */
118static void sctp_get_local_addr_list(struct net *net)
119{
120 struct net_device *dev;
121 struct list_head *pos;
122 struct sctp_af *af;
123
124 rcu_read_lock();
125 for_each_netdev_rcu(net, dev) {
126 list_for_each(pos, &sctp_address_families) {
127 af = list_entry(pos, struct sctp_af, list);
128 af->copy_addrlist(&net->sctp.local_addr_list, dev);
129 }
130 }
131 rcu_read_unlock();
132}
133
134/* Free the existing local addresses. */
135static void sctp_free_local_addr_list(struct net *net)
136{
137 struct sctp_sockaddr_entry *addr;
138 struct list_head *pos, *temp;
139
140 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
141 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
142 list_del(pos);
143 kfree(addr);
144 }
145}
146
147/* Copy the local addresses which are valid for 'scope' into 'bp'. */
148int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
149 enum sctp_scope scope, gfp_t gfp, int copy_flags)
150{
151 struct sctp_sockaddr_entry *addr;
152 union sctp_addr laddr;
153 int error = 0;
154
155 rcu_read_lock();
156 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
157 if (!addr->valid)
158 continue;
159 if (!sctp_in_scope(net, &addr->a, scope))
160 continue;
161
162 /* Now that the address is in scope, check to see if
163 * the address type is really supported by the local
164 * sock as well as the remote peer.
165 */
166 if (addr->a.sa.sa_family == AF_INET &&
167 !(copy_flags & SCTP_ADDR4_PEERSUPP))
168 continue;
169 if (addr->a.sa.sa_family == AF_INET6 &&
170 (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
171 !(copy_flags & SCTP_ADDR6_PEERSUPP)))
172 continue;
173
174 laddr = addr->a;
175 /* also works for setting ipv6 address port */
176 laddr.v4.sin_port = htons(bp->port);
177 if (sctp_bind_addr_state(bp, &laddr) != -1)
178 continue;
179
180 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
181 SCTP_ADDR_SRC, GFP_ATOMIC);
182 if (error)
183 break;
184 }
185
186 rcu_read_unlock();
187 return error;
188}
189
190/* Copy over any ip options */
191static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
192{
193 struct inet_sock *newinet, *inet = inet_sk(sk);
194 struct ip_options_rcu *inet_opt, *newopt = NULL;
195
196 newinet = inet_sk(newsk);
197
198 rcu_read_lock();
199 inet_opt = rcu_dereference(inet->inet_opt);
200 if (inet_opt) {
201 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
202 inet_opt->opt.optlen, GFP_ATOMIC);
203 if (newopt)
204 memcpy(newopt, inet_opt, sizeof(*inet_opt) +
205 inet_opt->opt.optlen);
206 else
207 pr_err("%s: Failed to copy ip options\n", __func__);
208 }
209 RCU_INIT_POINTER(newinet->inet_opt, newopt);
210 rcu_read_unlock();
211}
212
213/* Account for the IP options */
214static int sctp_v4_ip_options_len(struct sock *sk)
215{
216 struct inet_sock *inet = inet_sk(sk);
217 struct ip_options_rcu *inet_opt;
218 int len = 0;
219
220 rcu_read_lock();
221 inet_opt = rcu_dereference(inet->inet_opt);
222 if (inet_opt)
223 len = inet_opt->opt.optlen;
224
225 rcu_read_unlock();
226 return len;
227}
228
229/* Initialize a sctp_addr from in incoming skb. */
230static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
231 int is_saddr)
232{
233 /* Always called on head skb, so this is safe */
234 struct sctphdr *sh = sctp_hdr(skb);
235 struct sockaddr_in *sa = &addr->v4;
236
237 addr->v4.sin_family = AF_INET;
238
239 if (is_saddr) {
240 sa->sin_port = sh->source;
241 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
242 } else {
243 sa->sin_port = sh->dest;
244 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
245 }
246}
247
248/* Initialize an sctp_addr from a socket. */
249static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
250{
251 addr->v4.sin_family = AF_INET;
252 addr->v4.sin_port = 0;
253 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
254}
255
256/* Initialize sk->sk_rcv_saddr from sctp_addr. */
257static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
258{
259 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
260}
261
262/* Initialize sk->sk_daddr from sctp_addr. */
263static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
264{
265 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
266}
267
268/* Initialize a sctp_addr from an address parameter. */
269static void sctp_v4_from_addr_param(union sctp_addr *addr,
270 union sctp_addr_param *param,
271 __be16 port, int iif)
272{
273 addr->v4.sin_family = AF_INET;
274 addr->v4.sin_port = port;
275 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
276}
277
278/* Initialize an address parameter from a sctp_addr and return the length
279 * of the address parameter.
280 */
281static int sctp_v4_to_addr_param(const union sctp_addr *addr,
282 union sctp_addr_param *param)
283{
284 int length = sizeof(struct sctp_ipv4addr_param);
285
286 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
287 param->v4.param_hdr.length = htons(length);
288 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
289
290 return length;
291}
292
293/* Initialize a sctp_addr from a dst_entry. */
294static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
295 __be16 port)
296{
297 saddr->v4.sin_family = AF_INET;
298 saddr->v4.sin_port = port;
299 saddr->v4.sin_addr.s_addr = fl4->saddr;
300}
301
302/* Compare two addresses exactly. */
303static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
304 const union sctp_addr *addr2)
305{
306 if (addr1->sa.sa_family != addr2->sa.sa_family)
307 return 0;
308 if (addr1->v4.sin_port != addr2->v4.sin_port)
309 return 0;
310 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
311 return 0;
312
313 return 1;
314}
315
316/* Initialize addr struct to INADDR_ANY. */
317static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
318{
319 addr->v4.sin_family = AF_INET;
320 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
321 addr->v4.sin_port = port;
322}
323
324/* Is this a wildcard address? */
325static int sctp_v4_is_any(const union sctp_addr *addr)
326{
327 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
328}
329
330/* This function checks if the address is a valid address to be used for
331 * SCTP binding.
332 *
333 * Output:
334 * Return 0 - If the address is a non-unicast or an illegal address.
335 * Return 1 - If the address is a unicast.
336 */
337static int sctp_v4_addr_valid(union sctp_addr *addr,
338 struct sctp_sock *sp,
339 const struct sk_buff *skb)
340{
341 /* IPv4 addresses not allowed */
342 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
343 return 0;
344
345 /* Is this a non-unicast address or a unusable SCTP address? */
346 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
347 return 0;
348
349 /* Is this a broadcast address? */
350 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
351 return 0;
352
353 return 1;
354}
355
356/* Should this be available for binding? */
357static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
358{
359 struct net *net = sock_net(&sp->inet.sk);
360 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
361
362
363 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
364 ret != RTN_LOCAL &&
365 !sp->inet.freebind &&
366 !net->ipv4.sysctl_ip_nonlocal_bind)
367 return 0;
368
369 if (ipv6_only_sock(sctp_opt2sk(sp)))
370 return 0;
371
372 return 1;
373}
374
375/* Checking the loopback, private and other address scopes as defined in
376 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
377 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
378 *
379 * Level 0 - unusable SCTP addresses
380 * Level 1 - loopback address
381 * Level 2 - link-local addresses
382 * Level 3 - private addresses.
383 * Level 4 - global addresses
384 * For INIT and INIT-ACK address list, let L be the level of
385 * of requested destination address, sender and receiver
386 * SHOULD include all of its addresses with level greater
387 * than or equal to L.
388 *
389 * IPv4 scoping can be controlled through sysctl option
390 * net.sctp.addr_scope_policy
391 */
392static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
393{
394 enum sctp_scope retval;
395
396 /* Check for unusable SCTP addresses. */
397 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
398 retval = SCTP_SCOPE_UNUSABLE;
399 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
400 retval = SCTP_SCOPE_LOOPBACK;
401 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
402 retval = SCTP_SCOPE_LINK;
403 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
404 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
405 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
406 retval = SCTP_SCOPE_PRIVATE;
407 } else {
408 retval = SCTP_SCOPE_GLOBAL;
409 }
410
411 return retval;
412}
413
414/* Returns a valid dst cache entry for the given source and destination ip
415 * addresses. If an association is passed, trys to get a dst entry with a
416 * source address that matches an address in the bind address list.
417 */
418static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
419 struct flowi *fl, struct sock *sk)
420{
421 struct sctp_association *asoc = t->asoc;
422 struct rtable *rt;
423 struct flowi4 *fl4 = &fl->u.ip4;
424 struct sctp_bind_addr *bp;
425 struct sctp_sockaddr_entry *laddr;
426 struct dst_entry *dst = NULL;
427 union sctp_addr *daddr = &t->ipaddr;
428 union sctp_addr dst_saddr;
429
430 memset(fl4, 0x0, sizeof(struct flowi4));
431 fl4->daddr = daddr->v4.sin_addr.s_addr;
432 fl4->fl4_dport = daddr->v4.sin_port;
433 fl4->flowi4_proto = IPPROTO_SCTP;
434 if (asoc) {
435 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
436 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
437 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
438 }
439 if (saddr) {
440 fl4->saddr = saddr->v4.sin_addr.s_addr;
441 fl4->fl4_sport = saddr->v4.sin_port;
442 }
443
444 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
445 &fl4->saddr);
446
447 rt = ip_route_output_key(sock_net(sk), fl4);
448 if (!IS_ERR(rt))
449 dst = &rt->dst;
450
451 /* If there is no association or if a source address is passed, no
452 * more validation is required.
453 */
454 if (!asoc || saddr)
455 goto out;
456
457 bp = &asoc->base.bind_addr;
458
459 if (dst) {
460 /* Walk through the bind address list and look for a bind
461 * address that matches the source address of the returned dst.
462 */
463 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
464 rcu_read_lock();
465 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
466 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
467 (laddr->state != SCTP_ADDR_SRC &&
468 !asoc->src_out_of_asoc_ok))
469 continue;
470 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
471 goto out_unlock;
472 }
473 rcu_read_unlock();
474
475 /* None of the bound addresses match the source address of the
476 * dst. So release it.
477 */
478 dst_release(dst);
479 dst = NULL;
480 }
481
482 /* Walk through the bind address list and try to get a dst that
483 * matches a bind address as the source address.
484 */
485 rcu_read_lock();
486 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
487 struct net_device *odev;
488
489 if (!laddr->valid)
490 continue;
491 if (laddr->state != SCTP_ADDR_SRC ||
492 AF_INET != laddr->a.sa.sa_family)
493 continue;
494
495 fl4->fl4_sport = laddr->a.v4.sin_port;
496 flowi4_update_output(fl4,
497 asoc->base.sk->sk_bound_dev_if,
498 RT_CONN_FLAGS(asoc->base.sk),
499 daddr->v4.sin_addr.s_addr,
500 laddr->a.v4.sin_addr.s_addr);
501
502 rt = ip_route_output_key(sock_net(sk), fl4);
503 if (IS_ERR(rt))
504 continue;
505
506 /* Ensure the src address belongs to the output
507 * interface.
508 */
509 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
510 false);
511 if (!odev || odev->ifindex != fl4->flowi4_oif) {
512 if (!dst)
513 dst = &rt->dst;
514 else
515 dst_release(&rt->dst);
516 continue;
517 }
518
519 dst_release(dst);
520 dst = &rt->dst;
521 break;
522 }
523
524out_unlock:
525 rcu_read_unlock();
526out:
527 t->dst = dst;
528 if (dst)
529 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
530 &fl4->daddr, &fl4->saddr);
531 else
532 pr_debug("no route\n");
533}
534
535/* For v4, the source address is cached in the route entry(dst). So no need
536 * to cache it separately and hence this is an empty routine.
537 */
538static void sctp_v4_get_saddr(struct sctp_sock *sk,
539 struct sctp_transport *t,
540 struct flowi *fl)
541{
542 union sctp_addr *saddr = &t->saddr;
543 struct rtable *rt = (struct rtable *)t->dst;
544
545 if (rt) {
546 saddr->v4.sin_family = AF_INET;
547 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
548 }
549}
550
551/* What interface did this skb arrive on? */
552static int sctp_v4_skb_iif(const struct sk_buff *skb)
553{
554 return inet_iif(skb);
555}
556
557/* Was this packet marked by Explicit Congestion Notification? */
558static int sctp_v4_is_ce(const struct sk_buff *skb)
559{
560 return INET_ECN_is_ce(ip_hdr(skb)->tos);
561}
562
563/* Create and initialize a new sk for the socket returned by accept(). */
564static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
565 struct sctp_association *asoc,
566 bool kern)
567{
568 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
569 sk->sk_prot, kern);
570 struct inet_sock *newinet;
571
572 if (!newsk)
573 goto out;
574
575 sock_init_data(NULL, newsk);
576
577 sctp_copy_sock(newsk, sk, asoc);
578 sock_reset_flag(newsk, SOCK_ZAPPED);
579
580 sctp_v4_copy_ip_options(sk, newsk);
581
582 newinet = inet_sk(newsk);
583
584 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
585
586 sk_refcnt_debug_inc(newsk);
587
588 if (newsk->sk_prot->init(newsk)) {
589 sk_common_release(newsk);
590 newsk = NULL;
591 }
592
593out:
594 return newsk;
595}
596
597static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
598{
599 /* No address mapping for V4 sockets */
600 return sizeof(struct sockaddr_in);
601}
602
603/* Dump the v4 addr to the seq file. */
604static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
605{
606 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
607}
608
609static void sctp_v4_ecn_capable(struct sock *sk)
610{
611 INET_ECN_xmit(sk);
612}
613
614static void sctp_addr_wq_timeout_handler(struct timer_list *t)
615{
616 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
617 struct sctp_sockaddr_entry *addrw, *temp;
618 struct sctp_sock *sp;
619
620 spin_lock_bh(&net->sctp.addr_wq_lock);
621
622 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
623 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
624 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
625 addrw->state, addrw);
626
627#if IS_ENABLED(CONFIG_IPV6)
628 /* Now we send an ASCONF for each association */
629 /* Note. we currently don't handle link local IPv6 addressees */
630 if (addrw->a.sa.sa_family == AF_INET6) {
631 struct in6_addr *in6;
632
633 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
634 IPV6_ADDR_LINKLOCAL)
635 goto free_next;
636
637 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
638 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
639 addrw->state == SCTP_ADDR_NEW) {
640 unsigned long timeo_val;
641
642 pr_debug("%s: this is on DAD, trying %d sec "
643 "later\n", __func__,
644 SCTP_ADDRESS_TICK_DELAY);
645
646 timeo_val = jiffies;
647 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
648 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
649 break;
650 }
651 }
652#endif
653 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
654 struct sock *sk;
655
656 sk = sctp_opt2sk(sp);
657 /* ignore bound-specific endpoints */
658 if (!sctp_is_ep_boundall(sk))
659 continue;
660 bh_lock_sock(sk);
661 if (sctp_asconf_mgmt(sp, addrw) < 0)
662 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
663 bh_unlock_sock(sk);
664 }
665#if IS_ENABLED(CONFIG_IPV6)
666free_next:
667#endif
668 list_del(&addrw->list);
669 kfree(addrw);
670 }
671 spin_unlock_bh(&net->sctp.addr_wq_lock);
672}
673
674static void sctp_free_addr_wq(struct net *net)
675{
676 struct sctp_sockaddr_entry *addrw;
677 struct sctp_sockaddr_entry *temp;
678
679 spin_lock_bh(&net->sctp.addr_wq_lock);
680 del_timer(&net->sctp.addr_wq_timer);
681 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
682 list_del(&addrw->list);
683 kfree(addrw);
684 }
685 spin_unlock_bh(&net->sctp.addr_wq_lock);
686}
687
688/* lookup the entry for the same address in the addr_waitq
689 * sctp_addr_wq MUST be locked
690 */
691static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
692 struct sctp_sockaddr_entry *addr)
693{
694 struct sctp_sockaddr_entry *addrw;
695
696 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
697 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
698 continue;
699 if (addrw->a.sa.sa_family == AF_INET) {
700 if (addrw->a.v4.sin_addr.s_addr ==
701 addr->a.v4.sin_addr.s_addr)
702 return addrw;
703 } else if (addrw->a.sa.sa_family == AF_INET6) {
704 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
705 &addr->a.v6.sin6_addr))
706 return addrw;
707 }
708 }
709 return NULL;
710}
711
712void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
713{
714 struct sctp_sockaddr_entry *addrw;
715 unsigned long timeo_val;
716
717 /* first, we check if an opposite message already exist in the queue.
718 * If we found such message, it is removed.
719 * This operation is a bit stupid, but the DHCP client attaches the
720 * new address after a couple of addition and deletion of that address
721 */
722
723 spin_lock_bh(&net->sctp.addr_wq_lock);
724 /* Offsets existing events in addr_wq */
725 addrw = sctp_addr_wq_lookup(net, addr);
726 if (addrw) {
727 if (addrw->state != cmd) {
728 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
729 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
730 &net->sctp.addr_waitq);
731
732 list_del(&addrw->list);
733 kfree(addrw);
734 }
735 spin_unlock_bh(&net->sctp.addr_wq_lock);
736 return;
737 }
738
739 /* OK, we have to add the new address to the wait queue */
740 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
741 if (addrw == NULL) {
742 spin_unlock_bh(&net->sctp.addr_wq_lock);
743 return;
744 }
745 addrw->state = cmd;
746 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
747
748 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
749 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
750
751 if (!timer_pending(&net->sctp.addr_wq_timer)) {
752 timeo_val = jiffies;
753 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
754 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
755 }
756 spin_unlock_bh(&net->sctp.addr_wq_lock);
757}
758
759/* Event handler for inet address addition/deletion events.
760 * The sctp_local_addr_list needs to be protocted by a spin lock since
761 * multiple notifiers (say IPv4 and IPv6) may be running at the same
762 * time and thus corrupt the list.
763 * The reader side is protected with RCU.
764 */
765static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
766 void *ptr)
767{
768 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
769 struct sctp_sockaddr_entry *addr = NULL;
770 struct sctp_sockaddr_entry *temp;
771 struct net *net = dev_net(ifa->ifa_dev->dev);
772 int found = 0;
773
774 switch (ev) {
775 case NETDEV_UP:
776 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
777 if (addr) {
778 addr->a.v4.sin_family = AF_INET;
779 addr->a.v4.sin_port = 0;
780 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
781 addr->valid = 1;
782 spin_lock_bh(&net->sctp.local_addr_lock);
783 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
784 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
785 spin_unlock_bh(&net->sctp.local_addr_lock);
786 }
787 break;
788 case NETDEV_DOWN:
789 spin_lock_bh(&net->sctp.local_addr_lock);
790 list_for_each_entry_safe(addr, temp,
791 &net->sctp.local_addr_list, list) {
792 if (addr->a.sa.sa_family == AF_INET &&
793 addr->a.v4.sin_addr.s_addr ==
794 ifa->ifa_local) {
795 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
796 found = 1;
797 addr->valid = 0;
798 list_del_rcu(&addr->list);
799 break;
800 }
801 }
802 spin_unlock_bh(&net->sctp.local_addr_lock);
803 if (found)
804 kfree_rcu(addr, rcu);
805 break;
806 }
807
808 return NOTIFY_DONE;
809}
810
811/*
812 * Initialize the control inode/socket with a control endpoint data
813 * structure. This endpoint is reserved exclusively for the OOTB processing.
814 */
815static int sctp_ctl_sock_init(struct net *net)
816{
817 int err;
818 sa_family_t family = PF_INET;
819
820 if (sctp_get_pf_specific(PF_INET6))
821 family = PF_INET6;
822
823 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
824 SOCK_SEQPACKET, IPPROTO_SCTP, net);
825
826 /* If IPv6 socket could not be created, try the IPv4 socket */
827 if (err < 0 && family == PF_INET6)
828 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
829 SOCK_SEQPACKET, IPPROTO_SCTP,
830 net);
831
832 if (err < 0) {
833 pr_err("Failed to create the SCTP control socket\n");
834 return err;
835 }
836 return 0;
837}
838
839/* Register address family specific functions. */
840int sctp_register_af(struct sctp_af *af)
841{
842 switch (af->sa_family) {
843 case AF_INET:
844 if (sctp_af_v4_specific)
845 return 0;
846 sctp_af_v4_specific = af;
847 break;
848 case AF_INET6:
849 if (sctp_af_v6_specific)
850 return 0;
851 sctp_af_v6_specific = af;
852 break;
853 default:
854 return 0;
855 }
856
857 INIT_LIST_HEAD(&af->list);
858 list_add_tail(&af->list, &sctp_address_families);
859 return 1;
860}
861
862/* Get the table of functions for manipulating a particular address
863 * family.
864 */
865struct sctp_af *sctp_get_af_specific(sa_family_t family)
866{
867 switch (family) {
868 case AF_INET:
869 return sctp_af_v4_specific;
870 case AF_INET6:
871 return sctp_af_v6_specific;
872 default:
873 return NULL;
874 }
875}
876
877/* Common code to initialize a AF_INET msg_name. */
878static void sctp_inet_msgname(char *msgname, int *addr_len)
879{
880 struct sockaddr_in *sin;
881
882 sin = (struct sockaddr_in *)msgname;
883 *addr_len = sizeof(struct sockaddr_in);
884 sin->sin_family = AF_INET;
885 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
886}
887
888/* Copy the primary address of the peer primary address as the msg_name. */
889static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
890 int *addr_len)
891{
892 struct sockaddr_in *sin, *sinfrom;
893
894 if (msgname) {
895 struct sctp_association *asoc;
896
897 asoc = event->asoc;
898 sctp_inet_msgname(msgname, addr_len);
899 sin = (struct sockaddr_in *)msgname;
900 sinfrom = &asoc->peer.primary_addr.v4;
901 sin->sin_port = htons(asoc->peer.port);
902 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
903 }
904}
905
906/* Initialize and copy out a msgname from an inbound skb. */
907static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
908{
909 if (msgname) {
910 struct sctphdr *sh = sctp_hdr(skb);
911 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
912
913 sctp_inet_msgname(msgname, len);
914 sin->sin_port = sh->source;
915 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
916 }
917}
918
919/* Do we support this AF? */
920static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
921{
922 /* PF_INET only supports AF_INET addresses. */
923 return AF_INET == family;
924}
925
926/* Address matching with wildcards allowed. */
927static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
928 const union sctp_addr *addr2,
929 struct sctp_sock *opt)
930{
931 /* PF_INET only supports AF_INET addresses. */
932 if (addr1->sa.sa_family != addr2->sa.sa_family)
933 return 0;
934 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
935 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
936 return 1;
937 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
938 return 1;
939
940 return 0;
941}
942
943/* Verify that provided sockaddr looks bindable. Common verification has
944 * already been taken care of.
945 */
946static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
947{
948 return sctp_v4_available(addr, opt);
949}
950
951/* Verify that sockaddr looks sendable. Common verification has already
952 * been taken care of.
953 */
954static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
955{
956 return 1;
957}
958
959/* Fill in Supported Address Type information for INIT and INIT-ACK
960 * chunks. Returns number of addresses supported.
961 */
962static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
963 __be16 *types)
964{
965 types[0] = SCTP_PARAM_IPV4_ADDRESS;
966 return 1;
967}
968
969/* Wrapper routine that calls the ip transmit routine. */
970static inline int sctp_v4_xmit(struct sk_buff *skb,
971 struct sctp_transport *transport)
972{
973 struct inet_sock *inet = inet_sk(skb->sk);
974
975 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
976 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
977
978 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
979 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
980
981 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
982
983 return ip_queue_xmit(&inet->sk, skb, &transport->fl);
984}
985
986static struct sctp_af sctp_af_inet;
987
988static struct sctp_pf sctp_pf_inet = {
989 .event_msgname = sctp_inet_event_msgname,
990 .skb_msgname = sctp_inet_skb_msgname,
991 .af_supported = sctp_inet_af_supported,
992 .cmp_addr = sctp_inet_cmp_addr,
993 .bind_verify = sctp_inet_bind_verify,
994 .send_verify = sctp_inet_send_verify,
995 .supported_addrs = sctp_inet_supported_addrs,
996 .create_accept_sk = sctp_v4_create_accept_sk,
997 .addr_to_user = sctp_v4_addr_to_user,
998 .to_sk_saddr = sctp_v4_to_sk_saddr,
999 .to_sk_daddr = sctp_v4_to_sk_daddr,
1000 .copy_ip_options = sctp_v4_copy_ip_options,
1001 .af = &sctp_af_inet
1002};
1003
1004/* Notifier for inetaddr addition/deletion events. */
1005static struct notifier_block sctp_inetaddr_notifier = {
1006 .notifier_call = sctp_inetaddr_event,
1007};
1008
1009/* Socket operations. */
1010static const struct proto_ops inet_seqpacket_ops = {
1011 .family = PF_INET,
1012 .owner = THIS_MODULE,
1013 .release = inet_release, /* Needs to be wrapped... */
1014 .bind = inet_bind,
1015 .connect = sctp_inet_connect,
1016 .socketpair = sock_no_socketpair,
1017 .accept = inet_accept,
1018 .getname = inet_getname, /* Semantics are different. */
1019 .poll = sctp_poll,
1020 .ioctl = inet_ioctl,
1021 .listen = sctp_inet_listen,
1022 .shutdown = inet_shutdown, /* Looks harmless. */
1023 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1024 .getsockopt = sock_common_getsockopt,
1025 .sendmsg = inet_sendmsg,
1026 .recvmsg = inet_recvmsg,
1027 .mmap = sock_no_mmap,
1028 .sendpage = sock_no_sendpage,
1029#ifdef CONFIG_COMPAT
1030 .compat_setsockopt = compat_sock_common_setsockopt,
1031 .compat_getsockopt = compat_sock_common_getsockopt,
1032#endif
1033};
1034
1035/* Registration with AF_INET family. */
1036static struct inet_protosw sctp_seqpacket_protosw = {
1037 .type = SOCK_SEQPACKET,
1038 .protocol = IPPROTO_SCTP,
1039 .prot = &sctp_prot,
1040 .ops = &inet_seqpacket_ops,
1041 .flags = SCTP_PROTOSW_FLAG
1042};
1043static struct inet_protosw sctp_stream_protosw = {
1044 .type = SOCK_STREAM,
1045 .protocol = IPPROTO_SCTP,
1046 .prot = &sctp_prot,
1047 .ops = &inet_seqpacket_ops,
1048 .flags = SCTP_PROTOSW_FLAG
1049};
1050
1051/* Register with IP layer. */
1052static const struct net_protocol sctp_protocol = {
1053 .handler = sctp_rcv,
1054 .err_handler = sctp_v4_err,
1055 .no_policy = 1,
1056 .netns_ok = 1,
1057 .icmp_strict_tag_validation = 1,
1058};
1059
1060/* IPv4 address related functions. */
1061static struct sctp_af sctp_af_inet = {
1062 .sa_family = AF_INET,
1063 .sctp_xmit = sctp_v4_xmit,
1064 .setsockopt = ip_setsockopt,
1065 .getsockopt = ip_getsockopt,
1066 .get_dst = sctp_v4_get_dst,
1067 .get_saddr = sctp_v4_get_saddr,
1068 .copy_addrlist = sctp_v4_copy_addrlist,
1069 .from_skb = sctp_v4_from_skb,
1070 .from_sk = sctp_v4_from_sk,
1071 .from_addr_param = sctp_v4_from_addr_param,
1072 .to_addr_param = sctp_v4_to_addr_param,
1073 .cmp_addr = sctp_v4_cmp_addr,
1074 .addr_valid = sctp_v4_addr_valid,
1075 .inaddr_any = sctp_v4_inaddr_any,
1076 .is_any = sctp_v4_is_any,
1077 .available = sctp_v4_available,
1078 .scope = sctp_v4_scope,
1079 .skb_iif = sctp_v4_skb_iif,
1080 .is_ce = sctp_v4_is_ce,
1081 .seq_dump_addr = sctp_v4_seq_dump_addr,
1082 .ecn_capable = sctp_v4_ecn_capable,
1083 .net_header_len = sizeof(struct iphdr),
1084 .sockaddr_len = sizeof(struct sockaddr_in),
1085 .ip_options_len = sctp_v4_ip_options_len,
1086#ifdef CONFIG_COMPAT
1087 .compat_setsockopt = compat_ip_setsockopt,
1088 .compat_getsockopt = compat_ip_getsockopt,
1089#endif
1090};
1091
1092struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1093{
1094 switch (family) {
1095 case PF_INET:
1096 return sctp_pf_inet_specific;
1097 case PF_INET6:
1098 return sctp_pf_inet6_specific;
1099 default:
1100 return NULL;
1101 }
1102}
1103
1104/* Register the PF specific function table. */
1105int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1106{
1107 switch (family) {
1108 case PF_INET:
1109 if (sctp_pf_inet_specific)
1110 return 0;
1111 sctp_pf_inet_specific = pf;
1112 break;
1113 case PF_INET6:
1114 if (sctp_pf_inet6_specific)
1115 return 0;
1116 sctp_pf_inet6_specific = pf;
1117 break;
1118 default:
1119 return 0;
1120 }
1121 return 1;
1122}
1123
1124static inline int init_sctp_mibs(struct net *net)
1125{
1126 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1127 if (!net->sctp.sctp_statistics)
1128 return -ENOMEM;
1129 return 0;
1130}
1131
1132static inline void cleanup_sctp_mibs(struct net *net)
1133{
1134 free_percpu(net->sctp.sctp_statistics);
1135}
1136
1137static void sctp_v4_pf_init(void)
1138{
1139 /* Initialize the SCTP specific PF functions. */
1140 sctp_register_pf(&sctp_pf_inet, PF_INET);
1141 sctp_register_af(&sctp_af_inet);
1142}
1143
1144static void sctp_v4_pf_exit(void)
1145{
1146 list_del(&sctp_af_inet.list);
1147}
1148
1149static int sctp_v4_protosw_init(void)
1150{
1151 int rc;
1152
1153 rc = proto_register(&sctp_prot, 1);
1154 if (rc)
1155 return rc;
1156
1157 /* Register SCTP(UDP and TCP style) with socket layer. */
1158 inet_register_protosw(&sctp_seqpacket_protosw);
1159 inet_register_protosw(&sctp_stream_protosw);
1160
1161 return 0;
1162}
1163
1164static void sctp_v4_protosw_exit(void)
1165{
1166 inet_unregister_protosw(&sctp_stream_protosw);
1167 inet_unregister_protosw(&sctp_seqpacket_protosw);
1168 proto_unregister(&sctp_prot);
1169}
1170
1171static int sctp_v4_add_protocol(void)
1172{
1173 /* Register notifier for inet address additions/deletions. */
1174 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1175
1176 /* Register SCTP with inet layer. */
1177 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1178 return -EAGAIN;
1179
1180 return 0;
1181}
1182
1183static void sctp_v4_del_protocol(void)
1184{
1185 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1186 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1187}
1188
1189static int __net_init sctp_defaults_init(struct net *net)
1190{
1191 int status;
1192
1193 /*
1194 * 14. Suggested SCTP Protocol Parameter Values
1195 */
1196 /* The following protocol parameters are RECOMMENDED: */
1197 /* RTO.Initial - 3 seconds */
1198 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1199 /* RTO.Min - 1 second */
1200 net->sctp.rto_min = SCTP_RTO_MIN;
1201 /* RTO.Max - 60 seconds */
1202 net->sctp.rto_max = SCTP_RTO_MAX;
1203 /* RTO.Alpha - 1/8 */
1204 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1205 /* RTO.Beta - 1/4 */
1206 net->sctp.rto_beta = SCTP_RTO_BETA;
1207
1208 /* Valid.Cookie.Life - 60 seconds */
1209 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1210
1211 /* Whether Cookie Preservative is enabled(1) or not(0) */
1212 net->sctp.cookie_preserve_enable = 1;
1213
1214 /* Default sctp sockets to use md5 as their hmac alg */
1215#if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1216 net->sctp.sctp_hmac_alg = "md5";
1217#elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1218 net->sctp.sctp_hmac_alg = "sha1";
1219#else
1220 net->sctp.sctp_hmac_alg = NULL;
1221#endif
1222
1223 /* Max.Burst - 4 */
1224 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1225
1226 /* Enable pf state by default */
1227 net->sctp.pf_enable = 1;
1228
1229 /* Association.Max.Retrans - 10 attempts
1230 * Path.Max.Retrans - 5 attempts (per destination address)
1231 * Max.Init.Retransmits - 8 attempts
1232 */
1233 net->sctp.max_retrans_association = 10;
1234 net->sctp.max_retrans_path = 5;
1235 net->sctp.max_retrans_init = 8;
1236
1237 /* Sendbuffer growth - do per-socket accounting */
1238 net->sctp.sndbuf_policy = 0;
1239
1240 /* Rcvbuffer growth - do per-socket accounting */
1241 net->sctp.rcvbuf_policy = 0;
1242
1243 /* HB.interval - 30 seconds */
1244 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1245
1246 /* delayed SACK timeout */
1247 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1248
1249 /* Disable ADDIP by default. */
1250 net->sctp.addip_enable = 0;
1251 net->sctp.addip_noauth = 0;
1252 net->sctp.default_auto_asconf = 0;
1253
1254 /* Enable PR-SCTP by default. */
1255 net->sctp.prsctp_enable = 1;
1256
1257 /* Disable RECONF by default. */
1258 net->sctp.reconf_enable = 0;
1259
1260 /* Disable AUTH by default. */
1261 net->sctp.auth_enable = 0;
1262
1263 /* Set SCOPE policy to enabled */
1264 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1265
1266 /* Set the default rwnd update threshold */
1267 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1268
1269 /* Initialize maximum autoclose timeout. */
1270 net->sctp.max_autoclose = INT_MAX / HZ;
1271
1272 status = sctp_sysctl_net_register(net);
1273 if (status)
1274 goto err_sysctl_register;
1275
1276 /* Allocate and initialise sctp mibs. */
1277 status = init_sctp_mibs(net);
1278 if (status)
1279 goto err_init_mibs;
1280
1281#ifdef CONFIG_PROC_FS
1282 /* Initialize proc fs directory. */
1283 status = sctp_proc_init(net);
1284 if (status)
1285 goto err_init_proc;
1286#endif
1287
1288 sctp_dbg_objcnt_init(net);
1289
1290 /* Initialize the local address list. */
1291 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1292 spin_lock_init(&net->sctp.local_addr_lock);
1293 sctp_get_local_addr_list(net);
1294
1295 /* Initialize the address event list */
1296 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1297 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1298 spin_lock_init(&net->sctp.addr_wq_lock);
1299 net->sctp.addr_wq_timer.expires = 0;
1300 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1301
1302 return 0;
1303
1304#ifdef CONFIG_PROC_FS
1305err_init_proc:
1306 cleanup_sctp_mibs(net);
1307#endif
1308err_init_mibs:
1309 sctp_sysctl_net_unregister(net);
1310err_sysctl_register:
1311 return status;
1312}
1313
1314static void __net_exit sctp_defaults_exit(struct net *net)
1315{
1316 /* Free the local address list */
1317 sctp_free_addr_wq(net);
1318 sctp_free_local_addr_list(net);
1319
1320#ifdef CONFIG_PROC_FS
1321 remove_proc_subtree("sctp", net->proc_net);
1322 net->sctp.proc_net_sctp = NULL;
1323#endif
1324 cleanup_sctp_mibs(net);
1325 sctp_sysctl_net_unregister(net);
1326}
1327
1328static struct pernet_operations sctp_defaults_ops = {
1329 .init = sctp_defaults_init,
1330 .exit = sctp_defaults_exit,
1331};
1332
1333static int __net_init sctp_ctrlsock_init(struct net *net)
1334{
1335 int status;
1336
1337 /* Initialize the control inode/socket for handling OOTB packets. */
1338 status = sctp_ctl_sock_init(net);
1339 if (status)
1340 pr_err("Failed to initialize the SCTP control sock\n");
1341
1342 return status;
1343}
1344
1345static void __net_init sctp_ctrlsock_exit(struct net *net)
1346{
1347 /* Free the control endpoint. */
1348 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1349}
1350
1351static struct pernet_operations sctp_ctrlsock_ops = {
1352 .init = sctp_ctrlsock_init,
1353 .exit = sctp_ctrlsock_exit,
1354};
1355
1356/* Initialize the universe into something sensible. */
1357static __init int sctp_init(void)
1358{
1359 int i;
1360 int status = -EINVAL;
1361 unsigned long goal;
1362 unsigned long limit;
1363 int max_share;
1364 int order;
1365 int num_entries;
1366 int max_entry_order;
1367
1368 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1369
1370 /* Allocate bind_bucket and chunk caches. */
1371 status = -ENOBUFS;
1372 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1373 sizeof(struct sctp_bind_bucket),
1374 0, SLAB_HWCACHE_ALIGN,
1375 NULL);
1376 if (!sctp_bucket_cachep)
1377 goto out;
1378
1379 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1380 sizeof(struct sctp_chunk),
1381 0, SLAB_HWCACHE_ALIGN,
1382 NULL);
1383 if (!sctp_chunk_cachep)
1384 goto err_chunk_cachep;
1385
1386 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1387 if (status)
1388 goto err_percpu_counter_init;
1389
1390 /* Implementation specific variables. */
1391
1392 /* Initialize default stream count setup information. */
1393 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1394 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1395
1396 /* Initialize handle used for association ids. */
1397 idr_init(&sctp_assocs_id);
1398
1399 limit = nr_free_buffer_pages() / 8;
1400 limit = max(limit, 128UL);
1401 sysctl_sctp_mem[0] = limit / 4 * 3;
1402 sysctl_sctp_mem[1] = limit;
1403 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1404
1405 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1406 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1407 max_share = min(4UL*1024*1024, limit);
1408
1409 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1410 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1411 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1412
1413 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1414 sysctl_sctp_wmem[1] = 16*1024;
1415 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1416
1417 /* Size and allocate the association hash table.
1418 * The methodology is similar to that of the tcp hash tables.
1419 * Though not identical. Start by getting a goal size
1420 */
1421 if (totalram_pages >= (128 * 1024))
1422 goal = totalram_pages >> (22 - PAGE_SHIFT);
1423 else
1424 goal = totalram_pages >> (24 - PAGE_SHIFT);
1425
1426 /* Then compute the page order for said goal */
1427 order = get_order(goal);
1428
1429 /* Now compute the required page order for the maximum sized table we
1430 * want to create
1431 */
1432 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1433 sizeof(struct sctp_bind_hashbucket));
1434
1435 /* Limit the page order by that maximum hash table size */
1436 order = min(order, max_entry_order);
1437
1438 /* Allocate and initialize the endpoint hash table. */
1439 sctp_ep_hashsize = 64;
1440 sctp_ep_hashtable =
1441 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1442 if (!sctp_ep_hashtable) {
1443 pr_err("Failed endpoint_hash alloc\n");
1444 status = -ENOMEM;
1445 goto err_ehash_alloc;
1446 }
1447 for (i = 0; i < sctp_ep_hashsize; i++) {
1448 rwlock_init(&sctp_ep_hashtable[i].lock);
1449 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1450 }
1451
1452 /* Allocate and initialize the SCTP port hash table.
1453 * Note that order is initalized to start at the max sized
1454 * table we want to support. If we can't get that many pages
1455 * reduce the order and try again
1456 */
1457 do {
1458 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1459 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1460 } while (!sctp_port_hashtable && --order > 0);
1461
1462 if (!sctp_port_hashtable) {
1463 pr_err("Failed bind hash alloc\n");
1464 status = -ENOMEM;
1465 goto err_bhash_alloc;
1466 }
1467
1468 /* Now compute the number of entries that will fit in the
1469 * port hash space we allocated
1470 */
1471 num_entries = (1UL << order) * PAGE_SIZE /
1472 sizeof(struct sctp_bind_hashbucket);
1473
1474 /* And finish by rounding it down to the nearest power of two
1475 * this wastes some memory of course, but its needed because
1476 * the hash function operates based on the assumption that
1477 * that the number of entries is a power of two
1478 */
1479 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1480
1481 for (i = 0; i < sctp_port_hashsize; i++) {
1482 spin_lock_init(&sctp_port_hashtable[i].lock);
1483 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1484 }
1485
1486 status = sctp_transport_hashtable_init();
1487 if (status)
1488 goto err_thash_alloc;
1489
1490 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1491 num_entries);
1492
1493 sctp_sysctl_register();
1494
1495 INIT_LIST_HEAD(&sctp_address_families);
1496 sctp_v4_pf_init();
1497 sctp_v6_pf_init();
1498 sctp_sched_ops_init();
1499
1500 status = register_pernet_subsys(&sctp_defaults_ops);
1501 if (status)
1502 goto err_register_defaults;
1503
1504 status = sctp_v4_protosw_init();
1505 if (status)
1506 goto err_protosw_init;
1507
1508 status = sctp_v6_protosw_init();
1509 if (status)
1510 goto err_v6_protosw_init;
1511
1512 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1513 if (status)
1514 goto err_register_ctrlsock;
1515
1516 status = sctp_v4_add_protocol();
1517 if (status)
1518 goto err_add_protocol;
1519
1520 /* Register SCTP with inet6 layer. */
1521 status = sctp_v6_add_protocol();
1522 if (status)
1523 goto err_v6_add_protocol;
1524
1525 if (sctp_offload_init() < 0)
1526 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1527
1528out:
1529 return status;
1530err_v6_add_protocol:
1531 sctp_v4_del_protocol();
1532err_add_protocol:
1533 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1534err_register_ctrlsock:
1535 sctp_v6_protosw_exit();
1536err_v6_protosw_init:
1537 sctp_v4_protosw_exit();
1538err_protosw_init:
1539 unregister_pernet_subsys(&sctp_defaults_ops);
1540err_register_defaults:
1541 sctp_v4_pf_exit();
1542 sctp_v6_pf_exit();
1543 sctp_sysctl_unregister();
1544 free_pages((unsigned long)sctp_port_hashtable,
1545 get_order(sctp_port_hashsize *
1546 sizeof(struct sctp_bind_hashbucket)));
1547err_bhash_alloc:
1548 sctp_transport_hashtable_destroy();
1549err_thash_alloc:
1550 kfree(sctp_ep_hashtable);
1551err_ehash_alloc:
1552 percpu_counter_destroy(&sctp_sockets_allocated);
1553err_percpu_counter_init:
1554 kmem_cache_destroy(sctp_chunk_cachep);
1555err_chunk_cachep:
1556 kmem_cache_destroy(sctp_bucket_cachep);
1557 goto out;
1558}
1559
1560/* Exit handler for the SCTP protocol. */
1561static __exit void sctp_exit(void)
1562{
1563 /* BUG. This should probably do something useful like clean
1564 * up all the remaining associations and all that memory.
1565 */
1566
1567 /* Unregister with inet6/inet layers. */
1568 sctp_v6_del_protocol();
1569 sctp_v4_del_protocol();
1570
1571 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1572
1573 /* Free protosw registrations */
1574 sctp_v6_protosw_exit();
1575 sctp_v4_protosw_exit();
1576
1577 unregister_pernet_subsys(&sctp_defaults_ops);
1578
1579 /* Unregister with socket layer. */
1580 sctp_v6_pf_exit();
1581 sctp_v4_pf_exit();
1582
1583 sctp_sysctl_unregister();
1584
1585 free_pages((unsigned long)sctp_port_hashtable,
1586 get_order(sctp_port_hashsize *
1587 sizeof(struct sctp_bind_hashbucket)));
1588 kfree(sctp_ep_hashtable);
1589 sctp_transport_hashtable_destroy();
1590
1591 percpu_counter_destroy(&sctp_sockets_allocated);
1592
1593 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1594
1595 kmem_cache_destroy(sctp_chunk_cachep);
1596 kmem_cache_destroy(sctp_bucket_cachep);
1597}
1598
1599module_init(sctp_init);
1600module_exit(sctp_exit);
1601
1602/*
1603 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1604 */
1605MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1606MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1607MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1608MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1609module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1610MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1611MODULE_LICENSE("GPL");