Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/* SCTP kernel implementation
3 * (C) Copyright IBM Corp. 2001, 2004
4 * Copyright (c) 1999-2000 Cisco, Inc.
5 * Copyright (c) 1999-2001 Motorola, Inc.
6 * Copyright (c) 2001 Intel Corp.
7 * Copyright (c) 2001 Nokia, Inc.
8 * Copyright (c) 2001 La Monte H.P. Yarroll
9 *
10 * This file is part of the SCTP kernel implementation
11 *
12 * Initialization/cleanup for SCTP protocol support.
13 *
14 * Please send any bug reports or fixes you make to the
15 * email address(es):
16 * lksctp developers <linux-sctp@vger.kernel.org>
17 *
18 * Written or modified by:
19 * La Monte H.P. Yarroll <piggy@acm.org>
20 * Karl Knutson <karl@athena.chicago.il.us>
21 * Jon Grimm <jgrimm@us.ibm.com>
22 * Sridhar Samudrala <sri@us.ibm.com>
23 * Daisy Chang <daisyc@us.ibm.com>
24 * Ardelle Fan <ardelle.fan@intel.com>
25 */
26
27#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
28
29#include <linux/module.h>
30#include <linux/init.h>
31#include <linux/netdevice.h>
32#include <linux/inetdevice.h>
33#include <linux/seq_file.h>
34#include <linux/memblock.h>
35#include <linux/highmem.h>
36#include <linux/swap.h>
37#include <linux/slab.h>
38#include <net/net_namespace.h>
39#include <net/protocol.h>
40#include <net/ip.h>
41#include <net/ipv6.h>
42#include <net/route.h>
43#include <net/sctp/sctp.h>
44#include <net/addrconf.h>
45#include <net/inet_common.h>
46#include <net/inet_ecn.h>
47
48#define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
49
50/* Global data structures. */
51struct sctp_globals sctp_globals __read_mostly;
52
53struct idr sctp_assocs_id;
54DEFINE_SPINLOCK(sctp_assocs_id_lock);
55
56static struct sctp_pf *sctp_pf_inet6_specific;
57static struct sctp_pf *sctp_pf_inet_specific;
58static struct sctp_af *sctp_af_v4_specific;
59static struct sctp_af *sctp_af_v6_specific;
60
61struct kmem_cache *sctp_chunk_cachep __read_mostly;
62struct kmem_cache *sctp_bucket_cachep __read_mostly;
63
64long sysctl_sctp_mem[3];
65int sysctl_sctp_rmem[3];
66int sysctl_sctp_wmem[3];
67
68/* Private helper to extract ipv4 address and stash them in
69 * the protocol structure.
70 */
71static void sctp_v4_copy_addrlist(struct list_head *addrlist,
72 struct net_device *dev)
73{
74 struct in_device *in_dev;
75 struct in_ifaddr *ifa;
76 struct sctp_sockaddr_entry *addr;
77
78 rcu_read_lock();
79 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
80 rcu_read_unlock();
81 return;
82 }
83
84 in_dev_for_each_ifa_rcu(ifa, in_dev) {
85 /* Add the address to the local list. */
86 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
87 if (addr) {
88 addr->a.v4.sin_family = AF_INET;
89 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
90 addr->valid = 1;
91 INIT_LIST_HEAD(&addr->list);
92 list_add_tail(&addr->list, addrlist);
93 }
94 }
95
96 rcu_read_unlock();
97}
98
99/* Extract our IP addresses from the system and stash them in the
100 * protocol structure.
101 */
102static void sctp_get_local_addr_list(struct net *net)
103{
104 struct net_device *dev;
105 struct list_head *pos;
106 struct sctp_af *af;
107
108 rcu_read_lock();
109 for_each_netdev_rcu(net, dev) {
110 list_for_each(pos, &sctp_address_families) {
111 af = list_entry(pos, struct sctp_af, list);
112 af->copy_addrlist(&net->sctp.local_addr_list, dev);
113 }
114 }
115 rcu_read_unlock();
116}
117
118/* Free the existing local addresses. */
119static void sctp_free_local_addr_list(struct net *net)
120{
121 struct sctp_sockaddr_entry *addr;
122 struct list_head *pos, *temp;
123
124 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
125 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
126 list_del(pos);
127 kfree(addr);
128 }
129}
130
131/* Copy the local addresses which are valid for 'scope' into 'bp'. */
132int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
133 enum sctp_scope scope, gfp_t gfp, int copy_flags)
134{
135 struct sctp_sockaddr_entry *addr;
136 union sctp_addr laddr;
137 int error = 0;
138
139 rcu_read_lock();
140 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
141 if (!addr->valid)
142 continue;
143 if (!sctp_in_scope(net, &addr->a, scope))
144 continue;
145
146 /* Now that the address is in scope, check to see if
147 * the address type is really supported by the local
148 * sock as well as the remote peer.
149 */
150 if (addr->a.sa.sa_family == AF_INET &&
151 !(copy_flags & SCTP_ADDR4_PEERSUPP))
152 continue;
153 if (addr->a.sa.sa_family == AF_INET6 &&
154 (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
155 !(copy_flags & SCTP_ADDR6_PEERSUPP)))
156 continue;
157
158 laddr = addr->a;
159 /* also works for setting ipv6 address port */
160 laddr.v4.sin_port = htons(bp->port);
161 if (sctp_bind_addr_state(bp, &laddr) != -1)
162 continue;
163
164 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
165 SCTP_ADDR_SRC, GFP_ATOMIC);
166 if (error)
167 break;
168 }
169
170 rcu_read_unlock();
171 return error;
172}
173
174/* Copy over any ip options */
175static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
176{
177 struct inet_sock *newinet, *inet = inet_sk(sk);
178 struct ip_options_rcu *inet_opt, *newopt = NULL;
179
180 newinet = inet_sk(newsk);
181
182 rcu_read_lock();
183 inet_opt = rcu_dereference(inet->inet_opt);
184 if (inet_opt) {
185 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
186 inet_opt->opt.optlen, GFP_ATOMIC);
187 if (newopt)
188 memcpy(newopt, inet_opt, sizeof(*inet_opt) +
189 inet_opt->opt.optlen);
190 else
191 pr_err("%s: Failed to copy ip options\n", __func__);
192 }
193 RCU_INIT_POINTER(newinet->inet_opt, newopt);
194 rcu_read_unlock();
195}
196
197/* Account for the IP options */
198static int sctp_v4_ip_options_len(struct sock *sk)
199{
200 struct inet_sock *inet = inet_sk(sk);
201 struct ip_options_rcu *inet_opt;
202 int len = 0;
203
204 rcu_read_lock();
205 inet_opt = rcu_dereference(inet->inet_opt);
206 if (inet_opt)
207 len = inet_opt->opt.optlen;
208
209 rcu_read_unlock();
210 return len;
211}
212
213/* Initialize a sctp_addr from in incoming skb. */
214static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
215 int is_saddr)
216{
217 /* Always called on head skb, so this is safe */
218 struct sctphdr *sh = sctp_hdr(skb);
219 struct sockaddr_in *sa = &addr->v4;
220
221 addr->v4.sin_family = AF_INET;
222
223 if (is_saddr) {
224 sa->sin_port = sh->source;
225 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
226 } else {
227 sa->sin_port = sh->dest;
228 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
229 }
230}
231
232/* Initialize an sctp_addr from a socket. */
233static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
234{
235 addr->v4.sin_family = AF_INET;
236 addr->v4.sin_port = 0;
237 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
238}
239
240/* Initialize sk->sk_rcv_saddr from sctp_addr. */
241static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
242{
243 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
244}
245
246/* Initialize sk->sk_daddr from sctp_addr. */
247static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
248{
249 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
250}
251
252/* Initialize a sctp_addr from an address parameter. */
253static void sctp_v4_from_addr_param(union sctp_addr *addr,
254 union sctp_addr_param *param,
255 __be16 port, int iif)
256{
257 addr->v4.sin_family = AF_INET;
258 addr->v4.sin_port = port;
259 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
260}
261
262/* Initialize an address parameter from a sctp_addr and return the length
263 * of the address parameter.
264 */
265static int sctp_v4_to_addr_param(const union sctp_addr *addr,
266 union sctp_addr_param *param)
267{
268 int length = sizeof(struct sctp_ipv4addr_param);
269
270 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
271 param->v4.param_hdr.length = htons(length);
272 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
273
274 return length;
275}
276
277/* Initialize a sctp_addr from a dst_entry. */
278static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
279 __be16 port)
280{
281 saddr->v4.sin_family = AF_INET;
282 saddr->v4.sin_port = port;
283 saddr->v4.sin_addr.s_addr = fl4->saddr;
284}
285
286/* Compare two addresses exactly. */
287static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
288 const union sctp_addr *addr2)
289{
290 if (addr1->sa.sa_family != addr2->sa.sa_family)
291 return 0;
292 if (addr1->v4.sin_port != addr2->v4.sin_port)
293 return 0;
294 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
295 return 0;
296
297 return 1;
298}
299
300/* Initialize addr struct to INADDR_ANY. */
301static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
302{
303 addr->v4.sin_family = AF_INET;
304 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
305 addr->v4.sin_port = port;
306}
307
308/* Is this a wildcard address? */
309static int sctp_v4_is_any(const union sctp_addr *addr)
310{
311 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
312}
313
314/* This function checks if the address is a valid address to be used for
315 * SCTP binding.
316 *
317 * Output:
318 * Return 0 - If the address is a non-unicast or an illegal address.
319 * Return 1 - If the address is a unicast.
320 */
321static int sctp_v4_addr_valid(union sctp_addr *addr,
322 struct sctp_sock *sp,
323 const struct sk_buff *skb)
324{
325 /* IPv4 addresses not allowed */
326 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
327 return 0;
328
329 /* Is this a non-unicast address or a unusable SCTP address? */
330 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
331 return 0;
332
333 /* Is this a broadcast address? */
334 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
335 return 0;
336
337 return 1;
338}
339
340/* Should this be available for binding? */
341static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
342{
343 struct net *net = sock_net(&sp->inet.sk);
344 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
345
346
347 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
348 ret != RTN_LOCAL &&
349 !sp->inet.freebind &&
350 !net->ipv4.sysctl_ip_nonlocal_bind)
351 return 0;
352
353 if (ipv6_only_sock(sctp_opt2sk(sp)))
354 return 0;
355
356 return 1;
357}
358
359/* Checking the loopback, private and other address scopes as defined in
360 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
361 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
362 *
363 * Level 0 - unusable SCTP addresses
364 * Level 1 - loopback address
365 * Level 2 - link-local addresses
366 * Level 3 - private addresses.
367 * Level 4 - global addresses
368 * For INIT and INIT-ACK address list, let L be the level of
369 * of requested destination address, sender and receiver
370 * SHOULD include all of its addresses with level greater
371 * than or equal to L.
372 *
373 * IPv4 scoping can be controlled through sysctl option
374 * net.sctp.addr_scope_policy
375 */
376static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
377{
378 enum sctp_scope retval;
379
380 /* Check for unusable SCTP addresses. */
381 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
382 retval = SCTP_SCOPE_UNUSABLE;
383 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
384 retval = SCTP_SCOPE_LOOPBACK;
385 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
386 retval = SCTP_SCOPE_LINK;
387 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
388 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
389 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
390 retval = SCTP_SCOPE_PRIVATE;
391 } else {
392 retval = SCTP_SCOPE_GLOBAL;
393 }
394
395 return retval;
396}
397
398/* Returns a valid dst cache entry for the given source and destination ip
399 * addresses. If an association is passed, trys to get a dst entry with a
400 * source address that matches an address in the bind address list.
401 */
402static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
403 struct flowi *fl, struct sock *sk)
404{
405 struct sctp_association *asoc = t->asoc;
406 struct rtable *rt;
407 struct flowi4 *fl4 = &fl->u.ip4;
408 struct sctp_bind_addr *bp;
409 struct sctp_sockaddr_entry *laddr;
410 struct dst_entry *dst = NULL;
411 union sctp_addr *daddr = &t->ipaddr;
412 union sctp_addr dst_saddr;
413 __u8 tos = inet_sk(sk)->tos;
414
415 if (t->dscp & SCTP_DSCP_SET_MASK)
416 tos = t->dscp & SCTP_DSCP_VAL_MASK;
417 memset(fl4, 0x0, sizeof(struct flowi4));
418 fl4->daddr = daddr->v4.sin_addr.s_addr;
419 fl4->fl4_dport = daddr->v4.sin_port;
420 fl4->flowi4_proto = IPPROTO_SCTP;
421 if (asoc) {
422 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
423 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
424 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
425 }
426 if (saddr) {
427 fl4->saddr = saddr->v4.sin_addr.s_addr;
428 if (!fl4->fl4_sport)
429 fl4->fl4_sport = saddr->v4.sin_port;
430 }
431
432 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
433 &fl4->saddr);
434
435 rt = ip_route_output_key(sock_net(sk), fl4);
436 if (!IS_ERR(rt))
437 dst = &rt->dst;
438
439 /* If there is no association or if a source address is passed, no
440 * more validation is required.
441 */
442 if (!asoc || saddr)
443 goto out;
444
445 bp = &asoc->base.bind_addr;
446
447 if (dst) {
448 /* Walk through the bind address list and look for a bind
449 * address that matches the source address of the returned dst.
450 */
451 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
452 rcu_read_lock();
453 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
454 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
455 (laddr->state != SCTP_ADDR_SRC &&
456 !asoc->src_out_of_asoc_ok))
457 continue;
458 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
459 goto out_unlock;
460 }
461 rcu_read_unlock();
462
463 /* None of the bound addresses match the source address of the
464 * dst. So release it.
465 */
466 dst_release(dst);
467 dst = NULL;
468 }
469
470 /* Walk through the bind address list and try to get a dst that
471 * matches a bind address as the source address.
472 */
473 rcu_read_lock();
474 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
475 struct net_device *odev;
476
477 if (!laddr->valid)
478 continue;
479 if (laddr->state != SCTP_ADDR_SRC ||
480 AF_INET != laddr->a.sa.sa_family)
481 continue;
482
483 fl4->fl4_sport = laddr->a.v4.sin_port;
484 flowi4_update_output(fl4,
485 asoc->base.sk->sk_bound_dev_if,
486 RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
487 daddr->v4.sin_addr.s_addr,
488 laddr->a.v4.sin_addr.s_addr);
489
490 rt = ip_route_output_key(sock_net(sk), fl4);
491 if (IS_ERR(rt))
492 continue;
493
494 /* Ensure the src address belongs to the output
495 * interface.
496 */
497 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
498 false);
499 if (!odev || odev->ifindex != fl4->flowi4_oif) {
500 if (!dst)
501 dst = &rt->dst;
502 else
503 dst_release(&rt->dst);
504 continue;
505 }
506
507 dst_release(dst);
508 dst = &rt->dst;
509 break;
510 }
511
512out_unlock:
513 rcu_read_unlock();
514out:
515 t->dst = dst;
516 if (dst)
517 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
518 &fl4->daddr, &fl4->saddr);
519 else
520 pr_debug("no route\n");
521}
522
523/* For v4, the source address is cached in the route entry(dst). So no need
524 * to cache it separately and hence this is an empty routine.
525 */
526static void sctp_v4_get_saddr(struct sctp_sock *sk,
527 struct sctp_transport *t,
528 struct flowi *fl)
529{
530 union sctp_addr *saddr = &t->saddr;
531 struct rtable *rt = (struct rtable *)t->dst;
532
533 if (rt) {
534 saddr->v4.sin_family = AF_INET;
535 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
536 }
537}
538
539/* What interface did this skb arrive on? */
540static int sctp_v4_skb_iif(const struct sk_buff *skb)
541{
542 return inet_iif(skb);
543}
544
545/* Was this packet marked by Explicit Congestion Notification? */
546static int sctp_v4_is_ce(const struct sk_buff *skb)
547{
548 return INET_ECN_is_ce(ip_hdr(skb)->tos);
549}
550
551/* Create and initialize a new sk for the socket returned by accept(). */
552static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
553 struct sctp_association *asoc,
554 bool kern)
555{
556 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
557 sk->sk_prot, kern);
558 struct inet_sock *newinet;
559
560 if (!newsk)
561 goto out;
562
563 sock_init_data(NULL, newsk);
564
565 sctp_copy_sock(newsk, sk, asoc);
566 sock_reset_flag(newsk, SOCK_ZAPPED);
567
568 sctp_v4_copy_ip_options(sk, newsk);
569
570 newinet = inet_sk(newsk);
571
572 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
573
574 sk_refcnt_debug_inc(newsk);
575
576 if (newsk->sk_prot->init(newsk)) {
577 sk_common_release(newsk);
578 newsk = NULL;
579 }
580
581out:
582 return newsk;
583}
584
585static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
586{
587 /* No address mapping for V4 sockets */
588 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
589 return sizeof(struct sockaddr_in);
590}
591
592/* Dump the v4 addr to the seq file. */
593static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
594{
595 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
596}
597
598static void sctp_v4_ecn_capable(struct sock *sk)
599{
600 INET_ECN_xmit(sk);
601}
602
603static void sctp_addr_wq_timeout_handler(struct timer_list *t)
604{
605 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
606 struct sctp_sockaddr_entry *addrw, *temp;
607 struct sctp_sock *sp;
608
609 spin_lock_bh(&net->sctp.addr_wq_lock);
610
611 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
612 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
613 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
614 addrw->state, addrw);
615
616#if IS_ENABLED(CONFIG_IPV6)
617 /* Now we send an ASCONF for each association */
618 /* Note. we currently don't handle link local IPv6 addressees */
619 if (addrw->a.sa.sa_family == AF_INET6) {
620 struct in6_addr *in6;
621
622 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
623 IPV6_ADDR_LINKLOCAL)
624 goto free_next;
625
626 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
627 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
628 addrw->state == SCTP_ADDR_NEW) {
629 unsigned long timeo_val;
630
631 pr_debug("%s: this is on DAD, trying %d sec "
632 "later\n", __func__,
633 SCTP_ADDRESS_TICK_DELAY);
634
635 timeo_val = jiffies;
636 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
637 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
638 break;
639 }
640 }
641#endif
642 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
643 struct sock *sk;
644
645 sk = sctp_opt2sk(sp);
646 /* ignore bound-specific endpoints */
647 if (!sctp_is_ep_boundall(sk))
648 continue;
649 bh_lock_sock(sk);
650 if (sctp_asconf_mgmt(sp, addrw) < 0)
651 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
652 bh_unlock_sock(sk);
653 }
654#if IS_ENABLED(CONFIG_IPV6)
655free_next:
656#endif
657 list_del(&addrw->list);
658 kfree(addrw);
659 }
660 spin_unlock_bh(&net->sctp.addr_wq_lock);
661}
662
663static void sctp_free_addr_wq(struct net *net)
664{
665 struct sctp_sockaddr_entry *addrw;
666 struct sctp_sockaddr_entry *temp;
667
668 spin_lock_bh(&net->sctp.addr_wq_lock);
669 del_timer(&net->sctp.addr_wq_timer);
670 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
671 list_del(&addrw->list);
672 kfree(addrw);
673 }
674 spin_unlock_bh(&net->sctp.addr_wq_lock);
675}
676
677/* lookup the entry for the same address in the addr_waitq
678 * sctp_addr_wq MUST be locked
679 */
680static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
681 struct sctp_sockaddr_entry *addr)
682{
683 struct sctp_sockaddr_entry *addrw;
684
685 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
686 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
687 continue;
688 if (addrw->a.sa.sa_family == AF_INET) {
689 if (addrw->a.v4.sin_addr.s_addr ==
690 addr->a.v4.sin_addr.s_addr)
691 return addrw;
692 } else if (addrw->a.sa.sa_family == AF_INET6) {
693 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
694 &addr->a.v6.sin6_addr))
695 return addrw;
696 }
697 }
698 return NULL;
699}
700
701void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
702{
703 struct sctp_sockaddr_entry *addrw;
704 unsigned long timeo_val;
705
706 /* first, we check if an opposite message already exist in the queue.
707 * If we found such message, it is removed.
708 * This operation is a bit stupid, but the DHCP client attaches the
709 * new address after a couple of addition and deletion of that address
710 */
711
712 spin_lock_bh(&net->sctp.addr_wq_lock);
713 /* Offsets existing events in addr_wq */
714 addrw = sctp_addr_wq_lookup(net, addr);
715 if (addrw) {
716 if (addrw->state != cmd) {
717 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
718 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
719 &net->sctp.addr_waitq);
720
721 list_del(&addrw->list);
722 kfree(addrw);
723 }
724 spin_unlock_bh(&net->sctp.addr_wq_lock);
725 return;
726 }
727
728 /* OK, we have to add the new address to the wait queue */
729 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
730 if (addrw == NULL) {
731 spin_unlock_bh(&net->sctp.addr_wq_lock);
732 return;
733 }
734 addrw->state = cmd;
735 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
736
737 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
738 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
739
740 if (!timer_pending(&net->sctp.addr_wq_timer)) {
741 timeo_val = jiffies;
742 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
743 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
744 }
745 spin_unlock_bh(&net->sctp.addr_wq_lock);
746}
747
748/* Event handler for inet address addition/deletion events.
749 * The sctp_local_addr_list needs to be protocted by a spin lock since
750 * multiple notifiers (say IPv4 and IPv6) may be running at the same
751 * time and thus corrupt the list.
752 * The reader side is protected with RCU.
753 */
754static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
755 void *ptr)
756{
757 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
758 struct sctp_sockaddr_entry *addr = NULL;
759 struct sctp_sockaddr_entry *temp;
760 struct net *net = dev_net(ifa->ifa_dev->dev);
761 int found = 0;
762
763 switch (ev) {
764 case NETDEV_UP:
765 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
766 if (addr) {
767 addr->a.v4.sin_family = AF_INET;
768 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
769 addr->valid = 1;
770 spin_lock_bh(&net->sctp.local_addr_lock);
771 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
772 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
773 spin_unlock_bh(&net->sctp.local_addr_lock);
774 }
775 break;
776 case NETDEV_DOWN:
777 spin_lock_bh(&net->sctp.local_addr_lock);
778 list_for_each_entry_safe(addr, temp,
779 &net->sctp.local_addr_list, list) {
780 if (addr->a.sa.sa_family == AF_INET &&
781 addr->a.v4.sin_addr.s_addr ==
782 ifa->ifa_local) {
783 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
784 found = 1;
785 addr->valid = 0;
786 list_del_rcu(&addr->list);
787 break;
788 }
789 }
790 spin_unlock_bh(&net->sctp.local_addr_lock);
791 if (found)
792 kfree_rcu(addr, rcu);
793 break;
794 }
795
796 return NOTIFY_DONE;
797}
798
799/*
800 * Initialize the control inode/socket with a control endpoint data
801 * structure. This endpoint is reserved exclusively for the OOTB processing.
802 */
803static int sctp_ctl_sock_init(struct net *net)
804{
805 int err;
806 sa_family_t family = PF_INET;
807
808 if (sctp_get_pf_specific(PF_INET6))
809 family = PF_INET6;
810
811 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
812 SOCK_SEQPACKET, IPPROTO_SCTP, net);
813
814 /* If IPv6 socket could not be created, try the IPv4 socket */
815 if (err < 0 && family == PF_INET6)
816 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
817 SOCK_SEQPACKET, IPPROTO_SCTP,
818 net);
819
820 if (err < 0) {
821 pr_err("Failed to create the SCTP control socket\n");
822 return err;
823 }
824 return 0;
825}
826
827/* Register address family specific functions. */
828int sctp_register_af(struct sctp_af *af)
829{
830 switch (af->sa_family) {
831 case AF_INET:
832 if (sctp_af_v4_specific)
833 return 0;
834 sctp_af_v4_specific = af;
835 break;
836 case AF_INET6:
837 if (sctp_af_v6_specific)
838 return 0;
839 sctp_af_v6_specific = af;
840 break;
841 default:
842 return 0;
843 }
844
845 INIT_LIST_HEAD(&af->list);
846 list_add_tail(&af->list, &sctp_address_families);
847 return 1;
848}
849
850/* Get the table of functions for manipulating a particular address
851 * family.
852 */
853struct sctp_af *sctp_get_af_specific(sa_family_t family)
854{
855 switch (family) {
856 case AF_INET:
857 return sctp_af_v4_specific;
858 case AF_INET6:
859 return sctp_af_v6_specific;
860 default:
861 return NULL;
862 }
863}
864
865/* Common code to initialize a AF_INET msg_name. */
866static void sctp_inet_msgname(char *msgname, int *addr_len)
867{
868 struct sockaddr_in *sin;
869
870 sin = (struct sockaddr_in *)msgname;
871 *addr_len = sizeof(struct sockaddr_in);
872 sin->sin_family = AF_INET;
873 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
874}
875
876/* Copy the primary address of the peer primary address as the msg_name. */
877static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
878 int *addr_len)
879{
880 struct sockaddr_in *sin, *sinfrom;
881
882 if (msgname) {
883 struct sctp_association *asoc;
884
885 asoc = event->asoc;
886 sctp_inet_msgname(msgname, addr_len);
887 sin = (struct sockaddr_in *)msgname;
888 sinfrom = &asoc->peer.primary_addr.v4;
889 sin->sin_port = htons(asoc->peer.port);
890 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
891 }
892}
893
894/* Initialize and copy out a msgname from an inbound skb. */
895static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
896{
897 if (msgname) {
898 struct sctphdr *sh = sctp_hdr(skb);
899 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
900
901 sctp_inet_msgname(msgname, len);
902 sin->sin_port = sh->source;
903 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
904 }
905}
906
907/* Do we support this AF? */
908static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
909{
910 /* PF_INET only supports AF_INET addresses. */
911 return AF_INET == family;
912}
913
914/* Address matching with wildcards allowed. */
915static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
916 const union sctp_addr *addr2,
917 struct sctp_sock *opt)
918{
919 /* PF_INET only supports AF_INET addresses. */
920 if (addr1->sa.sa_family != addr2->sa.sa_family)
921 return 0;
922 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
923 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
924 return 1;
925 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
926 return 1;
927
928 return 0;
929}
930
931/* Verify that provided sockaddr looks bindable. Common verification has
932 * already been taken care of.
933 */
934static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
935{
936 return sctp_v4_available(addr, opt);
937}
938
939/* Verify that sockaddr looks sendable. Common verification has already
940 * been taken care of.
941 */
942static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
943{
944 return 1;
945}
946
947/* Fill in Supported Address Type information for INIT and INIT-ACK
948 * chunks. Returns number of addresses supported.
949 */
950static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
951 __be16 *types)
952{
953 types[0] = SCTP_PARAM_IPV4_ADDRESS;
954 return 1;
955}
956
957/* Wrapper routine that calls the ip transmit routine. */
958static inline int sctp_v4_xmit(struct sk_buff *skb,
959 struct sctp_transport *transport)
960{
961 struct inet_sock *inet = inet_sk(skb->sk);
962 __u8 dscp = inet->tos;
963
964 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
965 skb->len, &transport->fl.u.ip4.saddr,
966 &transport->fl.u.ip4.daddr);
967
968 if (transport->dscp & SCTP_DSCP_SET_MASK)
969 dscp = transport->dscp & SCTP_DSCP_VAL_MASK;
970
971 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
972 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
973
974 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
975
976 return __ip_queue_xmit(&inet->sk, skb, &transport->fl, dscp);
977}
978
979static struct sctp_af sctp_af_inet;
980
981static struct sctp_pf sctp_pf_inet = {
982 .event_msgname = sctp_inet_event_msgname,
983 .skb_msgname = sctp_inet_skb_msgname,
984 .af_supported = sctp_inet_af_supported,
985 .cmp_addr = sctp_inet_cmp_addr,
986 .bind_verify = sctp_inet_bind_verify,
987 .send_verify = sctp_inet_send_verify,
988 .supported_addrs = sctp_inet_supported_addrs,
989 .create_accept_sk = sctp_v4_create_accept_sk,
990 .addr_to_user = sctp_v4_addr_to_user,
991 .to_sk_saddr = sctp_v4_to_sk_saddr,
992 .to_sk_daddr = sctp_v4_to_sk_daddr,
993 .copy_ip_options = sctp_v4_copy_ip_options,
994 .af = &sctp_af_inet
995};
996
997/* Notifier for inetaddr addition/deletion events. */
998static struct notifier_block sctp_inetaddr_notifier = {
999 .notifier_call = sctp_inetaddr_event,
1000};
1001
1002/* Socket operations. */
1003static const struct proto_ops inet_seqpacket_ops = {
1004 .family = PF_INET,
1005 .owner = THIS_MODULE,
1006 .release = inet_release, /* Needs to be wrapped... */
1007 .bind = inet_bind,
1008 .connect = sctp_inet_connect,
1009 .socketpair = sock_no_socketpair,
1010 .accept = inet_accept,
1011 .getname = inet_getname, /* Semantics are different. */
1012 .poll = sctp_poll,
1013 .ioctl = inet_ioctl,
1014 .gettstamp = sock_gettstamp,
1015 .listen = sctp_inet_listen,
1016 .shutdown = inet_shutdown, /* Looks harmless. */
1017 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1018 .getsockopt = sock_common_getsockopt,
1019 .sendmsg = inet_sendmsg,
1020 .recvmsg = inet_recvmsg,
1021 .mmap = sock_no_mmap,
1022 .sendpage = sock_no_sendpage,
1023#ifdef CONFIG_COMPAT
1024 .compat_setsockopt = compat_sock_common_setsockopt,
1025 .compat_getsockopt = compat_sock_common_getsockopt,
1026#endif
1027};
1028
1029/* Registration with AF_INET family. */
1030static struct inet_protosw sctp_seqpacket_protosw = {
1031 .type = SOCK_SEQPACKET,
1032 .protocol = IPPROTO_SCTP,
1033 .prot = &sctp_prot,
1034 .ops = &inet_seqpacket_ops,
1035 .flags = SCTP_PROTOSW_FLAG
1036};
1037static struct inet_protosw sctp_stream_protosw = {
1038 .type = SOCK_STREAM,
1039 .protocol = IPPROTO_SCTP,
1040 .prot = &sctp_prot,
1041 .ops = &inet_seqpacket_ops,
1042 .flags = SCTP_PROTOSW_FLAG
1043};
1044
1045/* Register with IP layer. */
1046static const struct net_protocol sctp_protocol = {
1047 .handler = sctp_rcv,
1048 .err_handler = sctp_v4_err,
1049 .no_policy = 1,
1050 .netns_ok = 1,
1051 .icmp_strict_tag_validation = 1,
1052};
1053
1054/* IPv4 address related functions. */
1055static struct sctp_af sctp_af_inet = {
1056 .sa_family = AF_INET,
1057 .sctp_xmit = sctp_v4_xmit,
1058 .setsockopt = ip_setsockopt,
1059 .getsockopt = ip_getsockopt,
1060 .get_dst = sctp_v4_get_dst,
1061 .get_saddr = sctp_v4_get_saddr,
1062 .copy_addrlist = sctp_v4_copy_addrlist,
1063 .from_skb = sctp_v4_from_skb,
1064 .from_sk = sctp_v4_from_sk,
1065 .from_addr_param = sctp_v4_from_addr_param,
1066 .to_addr_param = sctp_v4_to_addr_param,
1067 .cmp_addr = sctp_v4_cmp_addr,
1068 .addr_valid = sctp_v4_addr_valid,
1069 .inaddr_any = sctp_v4_inaddr_any,
1070 .is_any = sctp_v4_is_any,
1071 .available = sctp_v4_available,
1072 .scope = sctp_v4_scope,
1073 .skb_iif = sctp_v4_skb_iif,
1074 .is_ce = sctp_v4_is_ce,
1075 .seq_dump_addr = sctp_v4_seq_dump_addr,
1076 .ecn_capable = sctp_v4_ecn_capable,
1077 .net_header_len = sizeof(struct iphdr),
1078 .sockaddr_len = sizeof(struct sockaddr_in),
1079 .ip_options_len = sctp_v4_ip_options_len,
1080#ifdef CONFIG_COMPAT
1081 .compat_setsockopt = compat_ip_setsockopt,
1082 .compat_getsockopt = compat_ip_getsockopt,
1083#endif
1084};
1085
1086struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1087{
1088 switch (family) {
1089 case PF_INET:
1090 return sctp_pf_inet_specific;
1091 case PF_INET6:
1092 return sctp_pf_inet6_specific;
1093 default:
1094 return NULL;
1095 }
1096}
1097
1098/* Register the PF specific function table. */
1099int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1100{
1101 switch (family) {
1102 case PF_INET:
1103 if (sctp_pf_inet_specific)
1104 return 0;
1105 sctp_pf_inet_specific = pf;
1106 break;
1107 case PF_INET6:
1108 if (sctp_pf_inet6_specific)
1109 return 0;
1110 sctp_pf_inet6_specific = pf;
1111 break;
1112 default:
1113 return 0;
1114 }
1115 return 1;
1116}
1117
1118static inline int init_sctp_mibs(struct net *net)
1119{
1120 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1121 if (!net->sctp.sctp_statistics)
1122 return -ENOMEM;
1123 return 0;
1124}
1125
1126static inline void cleanup_sctp_mibs(struct net *net)
1127{
1128 free_percpu(net->sctp.sctp_statistics);
1129}
1130
1131static void sctp_v4_pf_init(void)
1132{
1133 /* Initialize the SCTP specific PF functions. */
1134 sctp_register_pf(&sctp_pf_inet, PF_INET);
1135 sctp_register_af(&sctp_af_inet);
1136}
1137
1138static void sctp_v4_pf_exit(void)
1139{
1140 list_del(&sctp_af_inet.list);
1141}
1142
1143static int sctp_v4_protosw_init(void)
1144{
1145 int rc;
1146
1147 rc = proto_register(&sctp_prot, 1);
1148 if (rc)
1149 return rc;
1150
1151 /* Register SCTP(UDP and TCP style) with socket layer. */
1152 inet_register_protosw(&sctp_seqpacket_protosw);
1153 inet_register_protosw(&sctp_stream_protosw);
1154
1155 return 0;
1156}
1157
1158static void sctp_v4_protosw_exit(void)
1159{
1160 inet_unregister_protosw(&sctp_stream_protosw);
1161 inet_unregister_protosw(&sctp_seqpacket_protosw);
1162 proto_unregister(&sctp_prot);
1163}
1164
1165static int sctp_v4_add_protocol(void)
1166{
1167 /* Register notifier for inet address additions/deletions. */
1168 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1169
1170 /* Register SCTP with inet layer. */
1171 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1172 return -EAGAIN;
1173
1174 return 0;
1175}
1176
1177static void sctp_v4_del_protocol(void)
1178{
1179 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1180 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1181}
1182
1183static int __net_init sctp_defaults_init(struct net *net)
1184{
1185 int status;
1186
1187 /*
1188 * 14. Suggested SCTP Protocol Parameter Values
1189 */
1190 /* The following protocol parameters are RECOMMENDED: */
1191 /* RTO.Initial - 3 seconds */
1192 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1193 /* RTO.Min - 1 second */
1194 net->sctp.rto_min = SCTP_RTO_MIN;
1195 /* RTO.Max - 60 seconds */
1196 net->sctp.rto_max = SCTP_RTO_MAX;
1197 /* RTO.Alpha - 1/8 */
1198 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1199 /* RTO.Beta - 1/4 */
1200 net->sctp.rto_beta = SCTP_RTO_BETA;
1201
1202 /* Valid.Cookie.Life - 60 seconds */
1203 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1204
1205 /* Whether Cookie Preservative is enabled(1) or not(0) */
1206 net->sctp.cookie_preserve_enable = 1;
1207
1208 /* Default sctp sockets to use md5 as their hmac alg */
1209#if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1210 net->sctp.sctp_hmac_alg = "md5";
1211#elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1212 net->sctp.sctp_hmac_alg = "sha1";
1213#else
1214 net->sctp.sctp_hmac_alg = NULL;
1215#endif
1216
1217 /* Max.Burst - 4 */
1218 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1219
1220 /* Enable pf state by default */
1221 net->sctp.pf_enable = 1;
1222
1223 /* Association.Max.Retrans - 10 attempts
1224 * Path.Max.Retrans - 5 attempts (per destination address)
1225 * Max.Init.Retransmits - 8 attempts
1226 */
1227 net->sctp.max_retrans_association = 10;
1228 net->sctp.max_retrans_path = 5;
1229 net->sctp.max_retrans_init = 8;
1230
1231 /* Sendbuffer growth - do per-socket accounting */
1232 net->sctp.sndbuf_policy = 0;
1233
1234 /* Rcvbuffer growth - do per-socket accounting */
1235 net->sctp.rcvbuf_policy = 0;
1236
1237 /* HB.interval - 30 seconds */
1238 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1239
1240 /* delayed SACK timeout */
1241 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1242
1243 /* Disable ADDIP by default. */
1244 net->sctp.addip_enable = 0;
1245 net->sctp.addip_noauth = 0;
1246 net->sctp.default_auto_asconf = 0;
1247
1248 /* Enable PR-SCTP by default. */
1249 net->sctp.prsctp_enable = 1;
1250
1251 /* Disable RECONF by default. */
1252 net->sctp.reconf_enable = 0;
1253
1254 /* Disable AUTH by default. */
1255 net->sctp.auth_enable = 0;
1256
1257 /* Enable ECN by default. */
1258 net->sctp.ecn_enable = 1;
1259
1260 /* Set SCOPE policy to enabled */
1261 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1262
1263 /* Set the default rwnd update threshold */
1264 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1265
1266 /* Initialize maximum autoclose timeout. */
1267 net->sctp.max_autoclose = INT_MAX / HZ;
1268
1269 status = sctp_sysctl_net_register(net);
1270 if (status)
1271 goto err_sysctl_register;
1272
1273 /* Allocate and initialise sctp mibs. */
1274 status = init_sctp_mibs(net);
1275 if (status)
1276 goto err_init_mibs;
1277
1278#ifdef CONFIG_PROC_FS
1279 /* Initialize proc fs directory. */
1280 status = sctp_proc_init(net);
1281 if (status)
1282 goto err_init_proc;
1283#endif
1284
1285 sctp_dbg_objcnt_init(net);
1286
1287 /* Initialize the local address list. */
1288 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1289 spin_lock_init(&net->sctp.local_addr_lock);
1290 sctp_get_local_addr_list(net);
1291
1292 /* Initialize the address event list */
1293 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1294 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1295 spin_lock_init(&net->sctp.addr_wq_lock);
1296 net->sctp.addr_wq_timer.expires = 0;
1297 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1298
1299 return 0;
1300
1301#ifdef CONFIG_PROC_FS
1302err_init_proc:
1303 cleanup_sctp_mibs(net);
1304#endif
1305err_init_mibs:
1306 sctp_sysctl_net_unregister(net);
1307err_sysctl_register:
1308 return status;
1309}
1310
1311static void __net_exit sctp_defaults_exit(struct net *net)
1312{
1313 /* Free the local address list */
1314 sctp_free_addr_wq(net);
1315 sctp_free_local_addr_list(net);
1316
1317#ifdef CONFIG_PROC_FS
1318 remove_proc_subtree("sctp", net->proc_net);
1319 net->sctp.proc_net_sctp = NULL;
1320#endif
1321 cleanup_sctp_mibs(net);
1322 sctp_sysctl_net_unregister(net);
1323}
1324
1325static struct pernet_operations sctp_defaults_ops = {
1326 .init = sctp_defaults_init,
1327 .exit = sctp_defaults_exit,
1328};
1329
1330static int __net_init sctp_ctrlsock_init(struct net *net)
1331{
1332 int status;
1333
1334 /* Initialize the control inode/socket for handling OOTB packets. */
1335 status = sctp_ctl_sock_init(net);
1336 if (status)
1337 pr_err("Failed to initialize the SCTP control sock\n");
1338
1339 return status;
1340}
1341
1342static void __net_exit sctp_ctrlsock_exit(struct net *net)
1343{
1344 /* Free the control endpoint. */
1345 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1346}
1347
1348static struct pernet_operations sctp_ctrlsock_ops = {
1349 .init = sctp_ctrlsock_init,
1350 .exit = sctp_ctrlsock_exit,
1351};
1352
1353/* Initialize the universe into something sensible. */
1354static __init int sctp_init(void)
1355{
1356 int i;
1357 int status = -EINVAL;
1358 unsigned long goal;
1359 unsigned long limit;
1360 unsigned long nr_pages = totalram_pages();
1361 int max_share;
1362 int order;
1363 int num_entries;
1364 int max_entry_order;
1365
1366 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1367
1368 /* Allocate bind_bucket and chunk caches. */
1369 status = -ENOBUFS;
1370 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1371 sizeof(struct sctp_bind_bucket),
1372 0, SLAB_HWCACHE_ALIGN,
1373 NULL);
1374 if (!sctp_bucket_cachep)
1375 goto out;
1376
1377 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1378 sizeof(struct sctp_chunk),
1379 0, SLAB_HWCACHE_ALIGN,
1380 NULL);
1381 if (!sctp_chunk_cachep)
1382 goto err_chunk_cachep;
1383
1384 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1385 if (status)
1386 goto err_percpu_counter_init;
1387
1388 /* Implementation specific variables. */
1389
1390 /* Initialize default stream count setup information. */
1391 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1392 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1393
1394 /* Initialize handle used for association ids. */
1395 idr_init(&sctp_assocs_id);
1396
1397 limit = nr_free_buffer_pages() / 8;
1398 limit = max(limit, 128UL);
1399 sysctl_sctp_mem[0] = limit / 4 * 3;
1400 sysctl_sctp_mem[1] = limit;
1401 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1402
1403 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1404 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1405 max_share = min(4UL*1024*1024, limit);
1406
1407 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1408 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1409 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1410
1411 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1412 sysctl_sctp_wmem[1] = 16*1024;
1413 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1414
1415 /* Size and allocate the association hash table.
1416 * The methodology is similar to that of the tcp hash tables.
1417 * Though not identical. Start by getting a goal size
1418 */
1419 if (nr_pages >= (128 * 1024))
1420 goal = nr_pages >> (22 - PAGE_SHIFT);
1421 else
1422 goal = nr_pages >> (24 - PAGE_SHIFT);
1423
1424 /* Then compute the page order for said goal */
1425 order = get_order(goal);
1426
1427 /* Now compute the required page order for the maximum sized table we
1428 * want to create
1429 */
1430 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1431 sizeof(struct sctp_bind_hashbucket));
1432
1433 /* Limit the page order by that maximum hash table size */
1434 order = min(order, max_entry_order);
1435
1436 /* Allocate and initialize the endpoint hash table. */
1437 sctp_ep_hashsize = 64;
1438 sctp_ep_hashtable =
1439 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1440 if (!sctp_ep_hashtable) {
1441 pr_err("Failed endpoint_hash alloc\n");
1442 status = -ENOMEM;
1443 goto err_ehash_alloc;
1444 }
1445 for (i = 0; i < sctp_ep_hashsize; i++) {
1446 rwlock_init(&sctp_ep_hashtable[i].lock);
1447 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1448 }
1449
1450 /* Allocate and initialize the SCTP port hash table.
1451 * Note that order is initalized to start at the max sized
1452 * table we want to support. If we can't get that many pages
1453 * reduce the order and try again
1454 */
1455 do {
1456 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1457 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1458 } while (!sctp_port_hashtable && --order > 0);
1459
1460 if (!sctp_port_hashtable) {
1461 pr_err("Failed bind hash alloc\n");
1462 status = -ENOMEM;
1463 goto err_bhash_alloc;
1464 }
1465
1466 /* Now compute the number of entries that will fit in the
1467 * port hash space we allocated
1468 */
1469 num_entries = (1UL << order) * PAGE_SIZE /
1470 sizeof(struct sctp_bind_hashbucket);
1471
1472 /* And finish by rounding it down to the nearest power of two
1473 * this wastes some memory of course, but its needed because
1474 * the hash function operates based on the assumption that
1475 * that the number of entries is a power of two
1476 */
1477 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1478
1479 for (i = 0; i < sctp_port_hashsize; i++) {
1480 spin_lock_init(&sctp_port_hashtable[i].lock);
1481 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1482 }
1483
1484 status = sctp_transport_hashtable_init();
1485 if (status)
1486 goto err_thash_alloc;
1487
1488 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1489 num_entries);
1490
1491 sctp_sysctl_register();
1492
1493 INIT_LIST_HEAD(&sctp_address_families);
1494 sctp_v4_pf_init();
1495 sctp_v6_pf_init();
1496 sctp_sched_ops_init();
1497
1498 status = register_pernet_subsys(&sctp_defaults_ops);
1499 if (status)
1500 goto err_register_defaults;
1501
1502 status = sctp_v4_protosw_init();
1503 if (status)
1504 goto err_protosw_init;
1505
1506 status = sctp_v6_protosw_init();
1507 if (status)
1508 goto err_v6_protosw_init;
1509
1510 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1511 if (status)
1512 goto err_register_ctrlsock;
1513
1514 status = sctp_v4_add_protocol();
1515 if (status)
1516 goto err_add_protocol;
1517
1518 /* Register SCTP with inet6 layer. */
1519 status = sctp_v6_add_protocol();
1520 if (status)
1521 goto err_v6_add_protocol;
1522
1523 if (sctp_offload_init() < 0)
1524 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1525
1526out:
1527 return status;
1528err_v6_add_protocol:
1529 sctp_v4_del_protocol();
1530err_add_protocol:
1531 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1532err_register_ctrlsock:
1533 sctp_v6_protosw_exit();
1534err_v6_protosw_init:
1535 sctp_v4_protosw_exit();
1536err_protosw_init:
1537 unregister_pernet_subsys(&sctp_defaults_ops);
1538err_register_defaults:
1539 sctp_v4_pf_exit();
1540 sctp_v6_pf_exit();
1541 sctp_sysctl_unregister();
1542 free_pages((unsigned long)sctp_port_hashtable,
1543 get_order(sctp_port_hashsize *
1544 sizeof(struct sctp_bind_hashbucket)));
1545err_bhash_alloc:
1546 sctp_transport_hashtable_destroy();
1547err_thash_alloc:
1548 kfree(sctp_ep_hashtable);
1549err_ehash_alloc:
1550 percpu_counter_destroy(&sctp_sockets_allocated);
1551err_percpu_counter_init:
1552 kmem_cache_destroy(sctp_chunk_cachep);
1553err_chunk_cachep:
1554 kmem_cache_destroy(sctp_bucket_cachep);
1555 goto out;
1556}
1557
1558/* Exit handler for the SCTP protocol. */
1559static __exit void sctp_exit(void)
1560{
1561 /* BUG. This should probably do something useful like clean
1562 * up all the remaining associations and all that memory.
1563 */
1564
1565 /* Unregister with inet6/inet layers. */
1566 sctp_v6_del_protocol();
1567 sctp_v4_del_protocol();
1568
1569 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1570
1571 /* Free protosw registrations */
1572 sctp_v6_protosw_exit();
1573 sctp_v4_protosw_exit();
1574
1575 unregister_pernet_subsys(&sctp_defaults_ops);
1576
1577 /* Unregister with socket layer. */
1578 sctp_v6_pf_exit();
1579 sctp_v4_pf_exit();
1580
1581 sctp_sysctl_unregister();
1582
1583 free_pages((unsigned long)sctp_port_hashtable,
1584 get_order(sctp_port_hashsize *
1585 sizeof(struct sctp_bind_hashbucket)));
1586 kfree(sctp_ep_hashtable);
1587 sctp_transport_hashtable_destroy();
1588
1589 percpu_counter_destroy(&sctp_sockets_allocated);
1590
1591 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1592
1593 kmem_cache_destroy(sctp_chunk_cachep);
1594 kmem_cache_destroy(sctp_bucket_cachep);
1595}
1596
1597module_init(sctp_init);
1598module_exit(sctp_exit);
1599
1600/*
1601 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1602 */
1603MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1604MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1605MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1606MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1607module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1608MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1609MODULE_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/* Set up the proc fs entry for the SCTP protocol. */
84static int __net_init sctp_proc_init(struct net *net)
85{
86#ifdef CONFIG_PROC_FS
87 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
88 if (!net->sctp.proc_net_sctp)
89 goto out_proc_net_sctp;
90 if (sctp_snmp_proc_init(net))
91 goto out_snmp_proc_init;
92 if (sctp_eps_proc_init(net))
93 goto out_eps_proc_init;
94 if (sctp_assocs_proc_init(net))
95 goto out_assocs_proc_init;
96 if (sctp_remaddr_proc_init(net))
97 goto out_remaddr_proc_init;
98
99 return 0;
100
101out_remaddr_proc_init:
102 sctp_assocs_proc_exit(net);
103out_assocs_proc_init:
104 sctp_eps_proc_exit(net);
105out_eps_proc_init:
106 sctp_snmp_proc_exit(net);
107out_snmp_proc_init:
108 remove_proc_entry("sctp", net->proc_net);
109 net->sctp.proc_net_sctp = NULL;
110out_proc_net_sctp:
111 return -ENOMEM;
112#endif /* CONFIG_PROC_FS */
113 return 0;
114}
115
116/* Clean up the proc fs entry for the SCTP protocol.
117 * Note: Do not make this __exit as it is used in the init error
118 * path.
119 */
120static void sctp_proc_exit(struct net *net)
121{
122#ifdef CONFIG_PROC_FS
123 sctp_snmp_proc_exit(net);
124 sctp_eps_proc_exit(net);
125 sctp_assocs_proc_exit(net);
126 sctp_remaddr_proc_exit(net);
127
128 remove_proc_entry("sctp", net->proc_net);
129 net->sctp.proc_net_sctp = NULL;
130#endif
131}
132
133/* Private helper to extract ipv4 address and stash them in
134 * the protocol structure.
135 */
136static void sctp_v4_copy_addrlist(struct list_head *addrlist,
137 struct net_device *dev)
138{
139 struct in_device *in_dev;
140 struct in_ifaddr *ifa;
141 struct sctp_sockaddr_entry *addr;
142
143 rcu_read_lock();
144 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
145 rcu_read_unlock();
146 return;
147 }
148
149 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
150 /* Add the address to the local list. */
151 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
152 if (addr) {
153 addr->a.v4.sin_family = AF_INET;
154 addr->a.v4.sin_port = 0;
155 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
156 addr->valid = 1;
157 INIT_LIST_HEAD(&addr->list);
158 list_add_tail(&addr->list, addrlist);
159 }
160 }
161
162 rcu_read_unlock();
163}
164
165/* Extract our IP addresses from the system and stash them in the
166 * protocol structure.
167 */
168static void sctp_get_local_addr_list(struct net *net)
169{
170 struct net_device *dev;
171 struct list_head *pos;
172 struct sctp_af *af;
173
174 rcu_read_lock();
175 for_each_netdev_rcu(net, dev) {
176 list_for_each(pos, &sctp_address_families) {
177 af = list_entry(pos, struct sctp_af, list);
178 af->copy_addrlist(&net->sctp.local_addr_list, dev);
179 }
180 }
181 rcu_read_unlock();
182}
183
184/* Free the existing local addresses. */
185static void sctp_free_local_addr_list(struct net *net)
186{
187 struct sctp_sockaddr_entry *addr;
188 struct list_head *pos, *temp;
189
190 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
191 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
192 list_del(pos);
193 kfree(addr);
194 }
195}
196
197/* Copy the local addresses which are valid for 'scope' into 'bp'. */
198int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
199 sctp_scope_t scope, gfp_t gfp, int copy_flags)
200{
201 struct sctp_sockaddr_entry *addr;
202 int error = 0;
203
204 rcu_read_lock();
205 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
206 if (!addr->valid)
207 continue;
208 if (sctp_in_scope(net, &addr->a, scope)) {
209 /* Now that the address is in scope, check to see if
210 * the address type is really supported by the local
211 * sock as well as the remote peer.
212 */
213 if ((((AF_INET == addr->a.sa.sa_family) &&
214 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
215 (((AF_INET6 == addr->a.sa.sa_family) &&
216 (copy_flags & SCTP_ADDR6_ALLOWED) &&
217 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
218 error = sctp_add_bind_addr(bp, &addr->a,
219 sizeof(addr->a),
220 SCTP_ADDR_SRC, GFP_ATOMIC);
221 if (error)
222 goto end_copy;
223 }
224 }
225 }
226
227end_copy:
228 rcu_read_unlock();
229 return error;
230}
231
232/* Initialize a sctp_addr from in incoming skb. */
233static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
234 int is_saddr)
235{
236 void *from;
237 __be16 *port;
238 struct sctphdr *sh;
239
240 port = &addr->v4.sin_port;
241 addr->v4.sin_family = AF_INET;
242
243 sh = sctp_hdr(skb);
244 if (is_saddr) {
245 *port = sh->source;
246 from = &ip_hdr(skb)->saddr;
247 } else {
248 *port = sh->dest;
249 from = &ip_hdr(skb)->daddr;
250 }
251 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
252}
253
254/* Initialize an sctp_addr from a socket. */
255static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
256{
257 addr->v4.sin_family = AF_INET;
258 addr->v4.sin_port = 0;
259 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
260}
261
262/* Initialize sk->sk_rcv_saddr from sctp_addr. */
263static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
264{
265 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
266}
267
268/* Initialize sk->sk_daddr from sctp_addr. */
269static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
270{
271 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
272}
273
274/* Initialize a sctp_addr from an address parameter. */
275static void sctp_v4_from_addr_param(union sctp_addr *addr,
276 union sctp_addr_param *param,
277 __be16 port, int iif)
278{
279 addr->v4.sin_family = AF_INET;
280 addr->v4.sin_port = port;
281 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
282}
283
284/* Initialize an address parameter from a sctp_addr and return the length
285 * of the address parameter.
286 */
287static int sctp_v4_to_addr_param(const union sctp_addr *addr,
288 union sctp_addr_param *param)
289{
290 int length = sizeof(sctp_ipv4addr_param_t);
291
292 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
293 param->v4.param_hdr.length = htons(length);
294 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
295
296 return length;
297}
298
299/* Initialize a sctp_addr from a dst_entry. */
300static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
301 __be16 port)
302{
303 saddr->v4.sin_family = AF_INET;
304 saddr->v4.sin_port = port;
305 saddr->v4.sin_addr.s_addr = fl4->saddr;
306}
307
308/* Compare two addresses exactly. */
309static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
310 const union sctp_addr *addr2)
311{
312 if (addr1->sa.sa_family != addr2->sa.sa_family)
313 return 0;
314 if (addr1->v4.sin_port != addr2->v4.sin_port)
315 return 0;
316 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
317 return 0;
318
319 return 1;
320}
321
322/* Initialize addr struct to INADDR_ANY. */
323static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
324{
325 addr->v4.sin_family = AF_INET;
326 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
327 addr->v4.sin_port = port;
328}
329
330/* Is this a wildcard address? */
331static int sctp_v4_is_any(const union sctp_addr *addr)
332{
333 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
334}
335
336/* This function checks if the address is a valid address to be used for
337 * SCTP binding.
338 *
339 * Output:
340 * Return 0 - If the address is a non-unicast or an illegal address.
341 * Return 1 - If the address is a unicast.
342 */
343static int sctp_v4_addr_valid(union sctp_addr *addr,
344 struct sctp_sock *sp,
345 const struct sk_buff *skb)
346{
347 /* IPv4 addresses not allowed */
348 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
349 return 0;
350
351 /* Is this a non-unicast address or a unusable SCTP address? */
352 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
353 return 0;
354
355 /* Is this a broadcast address? */
356 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
357 return 0;
358
359 return 1;
360}
361
362/* Should this be available for binding? */
363static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
364{
365 struct net *net = sock_net(&sp->inet.sk);
366 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
367
368
369 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
370 ret != RTN_LOCAL &&
371 !sp->inet.freebind &&
372 !net->ipv4.sysctl_ip_nonlocal_bind)
373 return 0;
374
375 if (ipv6_only_sock(sctp_opt2sk(sp)))
376 return 0;
377
378 return 1;
379}
380
381/* Checking the loopback, private and other address scopes as defined in
382 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
383 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
384 *
385 * Level 0 - unusable SCTP addresses
386 * Level 1 - loopback address
387 * Level 2 - link-local addresses
388 * Level 3 - private addresses.
389 * Level 4 - global addresses
390 * For INIT and INIT-ACK address list, let L be the level of
391 * of requested destination address, sender and receiver
392 * SHOULD include all of its addresses with level greater
393 * than or equal to L.
394 *
395 * IPv4 scoping can be controlled through sysctl option
396 * net.sctp.addr_scope_policy
397 */
398static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
399{
400 sctp_scope_t retval;
401
402 /* Check for unusable SCTP addresses. */
403 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
404 retval = SCTP_SCOPE_UNUSABLE;
405 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
406 retval = SCTP_SCOPE_LOOPBACK;
407 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
408 retval = SCTP_SCOPE_LINK;
409 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
410 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
411 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
412 retval = SCTP_SCOPE_PRIVATE;
413 } else {
414 retval = SCTP_SCOPE_GLOBAL;
415 }
416
417 return retval;
418}
419
420/* Returns a valid dst cache entry for the given source and destination ip
421 * addresses. If an association is passed, trys to get a dst entry with a
422 * source address that matches an address in the bind address list.
423 */
424static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
425 struct flowi *fl, struct sock *sk)
426{
427 struct sctp_association *asoc = t->asoc;
428 struct rtable *rt;
429 struct flowi4 *fl4 = &fl->u.ip4;
430 struct sctp_bind_addr *bp;
431 struct sctp_sockaddr_entry *laddr;
432 struct dst_entry *dst = NULL;
433 union sctp_addr *daddr = &t->ipaddr;
434 union sctp_addr dst_saddr;
435
436 memset(fl4, 0x0, sizeof(struct flowi4));
437 fl4->daddr = daddr->v4.sin_addr.s_addr;
438 fl4->fl4_dport = daddr->v4.sin_port;
439 fl4->flowi4_proto = IPPROTO_SCTP;
440 if (asoc) {
441 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
442 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
443 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
444 }
445 if (saddr) {
446 fl4->saddr = saddr->v4.sin_addr.s_addr;
447 fl4->fl4_sport = saddr->v4.sin_port;
448 }
449
450 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
451 &fl4->saddr);
452
453 rt = ip_route_output_key(sock_net(sk), fl4);
454 if (!IS_ERR(rt))
455 dst = &rt->dst;
456
457 /* If there is no association or if a source address is passed, no
458 * more validation is required.
459 */
460 if (!asoc || saddr)
461 goto out;
462
463 bp = &asoc->base.bind_addr;
464
465 if (dst) {
466 /* Walk through the bind address list and look for a bind
467 * address that matches the source address of the returned dst.
468 */
469 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
470 rcu_read_lock();
471 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
472 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
473 (laddr->state != SCTP_ADDR_SRC &&
474 !asoc->src_out_of_asoc_ok))
475 continue;
476 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
477 goto out_unlock;
478 }
479 rcu_read_unlock();
480
481 /* None of the bound addresses match the source address of the
482 * dst. So release it.
483 */
484 dst_release(dst);
485 dst = NULL;
486 }
487
488 /* Walk through the bind address list and try to get a dst that
489 * matches a bind address as the source address.
490 */
491 rcu_read_lock();
492 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
493 struct net_device *odev;
494
495 if (!laddr->valid)
496 continue;
497 if (laddr->state != SCTP_ADDR_SRC ||
498 AF_INET != laddr->a.sa.sa_family)
499 continue;
500
501 fl4->fl4_sport = laddr->a.v4.sin_port;
502 flowi4_update_output(fl4,
503 asoc->base.sk->sk_bound_dev_if,
504 RT_CONN_FLAGS(asoc->base.sk),
505 daddr->v4.sin_addr.s_addr,
506 laddr->a.v4.sin_addr.s_addr);
507
508 rt = ip_route_output_key(sock_net(sk), fl4);
509 if (IS_ERR(rt))
510 continue;
511
512 if (!dst)
513 dst = &rt->dst;
514
515 /* Ensure the src address belongs to the output
516 * interface.
517 */
518 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
519 false);
520 if (!odev || odev->ifindex != fl4->flowi4_oif) {
521 if (&rt->dst != dst)
522 dst_release(&rt->dst);
523 continue;
524 }
525
526 if (dst != &rt->dst)
527 dst_release(dst);
528 dst = &rt->dst;
529 break;
530 }
531
532out_unlock:
533 rcu_read_unlock();
534out:
535 t->dst = dst;
536 if (dst)
537 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
538 &fl4->daddr, &fl4->saddr);
539 else
540 pr_debug("no route\n");
541}
542
543/* For v4, the source address is cached in the route entry(dst). So no need
544 * to cache it separately and hence this is an empty routine.
545 */
546static void sctp_v4_get_saddr(struct sctp_sock *sk,
547 struct sctp_transport *t,
548 struct flowi *fl)
549{
550 union sctp_addr *saddr = &t->saddr;
551 struct rtable *rt = (struct rtable *)t->dst;
552
553 if (rt) {
554 saddr->v4.sin_family = AF_INET;
555 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
556 }
557}
558
559/* What interface did this skb arrive on? */
560static int sctp_v4_skb_iif(const struct sk_buff *skb)
561{
562 return inet_iif(skb);
563}
564
565/* Was this packet marked by Explicit Congestion Notification? */
566static int sctp_v4_is_ce(const struct sk_buff *skb)
567{
568 return INET_ECN_is_ce(ip_hdr(skb)->tos);
569}
570
571/* Create and initialize a new sk for the socket returned by accept(). */
572static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
573 struct sctp_association *asoc)
574{
575 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
576 sk->sk_prot, 0);
577 struct inet_sock *newinet;
578
579 if (!newsk)
580 goto out;
581
582 sock_init_data(NULL, newsk);
583
584 sctp_copy_sock(newsk, sk, asoc);
585 sock_reset_flag(newsk, SOCK_ZAPPED);
586
587 newinet = inet_sk(newsk);
588
589 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
590
591 sk_refcnt_debug_inc(newsk);
592
593 if (newsk->sk_prot->init(newsk)) {
594 sk_common_release(newsk);
595 newsk = NULL;
596 }
597
598out:
599 return newsk;
600}
601
602static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
603{
604 /* No address mapping for V4 sockets */
605 return sizeof(struct sockaddr_in);
606}
607
608/* Dump the v4 addr to the seq file. */
609static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
610{
611 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
612}
613
614static void sctp_v4_ecn_capable(struct sock *sk)
615{
616 INET_ECN_xmit(sk);
617}
618
619static void sctp_addr_wq_timeout_handler(unsigned long arg)
620{
621 struct net *net = (struct net *)arg;
622 struct sctp_sockaddr_entry *addrw, *temp;
623 struct sctp_sock *sp;
624
625 spin_lock_bh(&net->sctp.addr_wq_lock);
626
627 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
628 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
629 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
630 addrw->state, addrw);
631
632#if IS_ENABLED(CONFIG_IPV6)
633 /* Now we send an ASCONF for each association */
634 /* Note. we currently don't handle link local IPv6 addressees */
635 if (addrw->a.sa.sa_family == AF_INET6) {
636 struct in6_addr *in6;
637
638 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
639 IPV6_ADDR_LINKLOCAL)
640 goto free_next;
641
642 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
643 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
644 addrw->state == SCTP_ADDR_NEW) {
645 unsigned long timeo_val;
646
647 pr_debug("%s: this is on DAD, trying %d sec "
648 "later\n", __func__,
649 SCTP_ADDRESS_TICK_DELAY);
650
651 timeo_val = jiffies;
652 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
653 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
654 break;
655 }
656 }
657#endif
658 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
659 struct sock *sk;
660
661 sk = sctp_opt2sk(sp);
662 /* ignore bound-specific endpoints */
663 if (!sctp_is_ep_boundall(sk))
664 continue;
665 bh_lock_sock(sk);
666 if (sctp_asconf_mgmt(sp, addrw) < 0)
667 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
668 bh_unlock_sock(sk);
669 }
670#if IS_ENABLED(CONFIG_IPV6)
671free_next:
672#endif
673 list_del(&addrw->list);
674 kfree(addrw);
675 }
676 spin_unlock_bh(&net->sctp.addr_wq_lock);
677}
678
679static void sctp_free_addr_wq(struct net *net)
680{
681 struct sctp_sockaddr_entry *addrw;
682 struct sctp_sockaddr_entry *temp;
683
684 spin_lock_bh(&net->sctp.addr_wq_lock);
685 del_timer(&net->sctp.addr_wq_timer);
686 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
687 list_del(&addrw->list);
688 kfree(addrw);
689 }
690 spin_unlock_bh(&net->sctp.addr_wq_lock);
691}
692
693/* lookup the entry for the same address in the addr_waitq
694 * sctp_addr_wq MUST be locked
695 */
696static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
697 struct sctp_sockaddr_entry *addr)
698{
699 struct sctp_sockaddr_entry *addrw;
700
701 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
702 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
703 continue;
704 if (addrw->a.sa.sa_family == AF_INET) {
705 if (addrw->a.v4.sin_addr.s_addr ==
706 addr->a.v4.sin_addr.s_addr)
707 return addrw;
708 } else if (addrw->a.sa.sa_family == AF_INET6) {
709 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
710 &addr->a.v6.sin6_addr))
711 return addrw;
712 }
713 }
714 return NULL;
715}
716
717void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
718{
719 struct sctp_sockaddr_entry *addrw;
720 unsigned long timeo_val;
721
722 /* first, we check if an opposite message already exist in the queue.
723 * If we found such message, it is removed.
724 * This operation is a bit stupid, but the DHCP client attaches the
725 * new address after a couple of addition and deletion of that address
726 */
727
728 spin_lock_bh(&net->sctp.addr_wq_lock);
729 /* Offsets existing events in addr_wq */
730 addrw = sctp_addr_wq_lookup(net, addr);
731 if (addrw) {
732 if (addrw->state != cmd) {
733 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
734 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
735 &net->sctp.addr_waitq);
736
737 list_del(&addrw->list);
738 kfree(addrw);
739 }
740 spin_unlock_bh(&net->sctp.addr_wq_lock);
741 return;
742 }
743
744 /* OK, we have to add the new address to the wait queue */
745 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
746 if (addrw == NULL) {
747 spin_unlock_bh(&net->sctp.addr_wq_lock);
748 return;
749 }
750 addrw->state = cmd;
751 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
752
753 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
754 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
755
756 if (!timer_pending(&net->sctp.addr_wq_timer)) {
757 timeo_val = jiffies;
758 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
759 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
760 }
761 spin_unlock_bh(&net->sctp.addr_wq_lock);
762}
763
764/* Event handler for inet address addition/deletion events.
765 * The sctp_local_addr_list needs to be protocted by a spin lock since
766 * multiple notifiers (say IPv4 and IPv6) may be running at the same
767 * time and thus corrupt the list.
768 * The reader side is protected with RCU.
769 */
770static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
771 void *ptr)
772{
773 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
774 struct sctp_sockaddr_entry *addr = NULL;
775 struct sctp_sockaddr_entry *temp;
776 struct net *net = dev_net(ifa->ifa_dev->dev);
777 int found = 0;
778
779 switch (ev) {
780 case NETDEV_UP:
781 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
782 if (addr) {
783 addr->a.v4.sin_family = AF_INET;
784 addr->a.v4.sin_port = 0;
785 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
786 addr->valid = 1;
787 spin_lock_bh(&net->sctp.local_addr_lock);
788 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
789 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
790 spin_unlock_bh(&net->sctp.local_addr_lock);
791 }
792 break;
793 case NETDEV_DOWN:
794 spin_lock_bh(&net->sctp.local_addr_lock);
795 list_for_each_entry_safe(addr, temp,
796 &net->sctp.local_addr_list, list) {
797 if (addr->a.sa.sa_family == AF_INET &&
798 addr->a.v4.sin_addr.s_addr ==
799 ifa->ifa_local) {
800 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
801 found = 1;
802 addr->valid = 0;
803 list_del_rcu(&addr->list);
804 break;
805 }
806 }
807 spin_unlock_bh(&net->sctp.local_addr_lock);
808 if (found)
809 kfree_rcu(addr, rcu);
810 break;
811 }
812
813 return NOTIFY_DONE;
814}
815
816/*
817 * Initialize the control inode/socket with a control endpoint data
818 * structure. This endpoint is reserved exclusively for the OOTB processing.
819 */
820static int sctp_ctl_sock_init(struct net *net)
821{
822 int err;
823 sa_family_t family = PF_INET;
824
825 if (sctp_get_pf_specific(PF_INET6))
826 family = PF_INET6;
827
828 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
829 SOCK_SEQPACKET, IPPROTO_SCTP, net);
830
831 /* If IPv6 socket could not be created, try the IPv4 socket */
832 if (err < 0 && family == PF_INET6)
833 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
834 SOCK_SEQPACKET, IPPROTO_SCTP,
835 net);
836
837 if (err < 0) {
838 pr_err("Failed to create the SCTP control socket\n");
839 return err;
840 }
841 return 0;
842}
843
844/* Register address family specific functions. */
845int sctp_register_af(struct sctp_af *af)
846{
847 switch (af->sa_family) {
848 case AF_INET:
849 if (sctp_af_v4_specific)
850 return 0;
851 sctp_af_v4_specific = af;
852 break;
853 case AF_INET6:
854 if (sctp_af_v6_specific)
855 return 0;
856 sctp_af_v6_specific = af;
857 break;
858 default:
859 return 0;
860 }
861
862 INIT_LIST_HEAD(&af->list);
863 list_add_tail(&af->list, &sctp_address_families);
864 return 1;
865}
866
867/* Get the table of functions for manipulating a particular address
868 * family.
869 */
870struct sctp_af *sctp_get_af_specific(sa_family_t family)
871{
872 switch (family) {
873 case AF_INET:
874 return sctp_af_v4_specific;
875 case AF_INET6:
876 return sctp_af_v6_specific;
877 default:
878 return NULL;
879 }
880}
881
882/* Common code to initialize a AF_INET msg_name. */
883static void sctp_inet_msgname(char *msgname, int *addr_len)
884{
885 struct sockaddr_in *sin;
886
887 sin = (struct sockaddr_in *)msgname;
888 *addr_len = sizeof(struct sockaddr_in);
889 sin->sin_family = AF_INET;
890 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
891}
892
893/* Copy the primary address of the peer primary address as the msg_name. */
894static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
895 int *addr_len)
896{
897 struct sockaddr_in *sin, *sinfrom;
898
899 if (msgname) {
900 struct sctp_association *asoc;
901
902 asoc = event->asoc;
903 sctp_inet_msgname(msgname, addr_len);
904 sin = (struct sockaddr_in *)msgname;
905 sinfrom = &asoc->peer.primary_addr.v4;
906 sin->sin_port = htons(asoc->peer.port);
907 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
908 }
909}
910
911/* Initialize and copy out a msgname from an inbound skb. */
912static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
913{
914 if (msgname) {
915 struct sctphdr *sh = sctp_hdr(skb);
916 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
917
918 sctp_inet_msgname(msgname, len);
919 sin->sin_port = sh->source;
920 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
921 }
922}
923
924/* Do we support this AF? */
925static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
926{
927 /* PF_INET only supports AF_INET addresses. */
928 return AF_INET == family;
929}
930
931/* Address matching with wildcards allowed. */
932static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
933 const union sctp_addr *addr2,
934 struct sctp_sock *opt)
935{
936 /* PF_INET only supports AF_INET addresses. */
937 if (addr1->sa.sa_family != addr2->sa.sa_family)
938 return 0;
939 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
940 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
941 return 1;
942 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
943 return 1;
944
945 return 0;
946}
947
948/* Verify that provided sockaddr looks bindable. Common verification has
949 * already been taken care of.
950 */
951static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
952{
953 return sctp_v4_available(addr, opt);
954}
955
956/* Verify that sockaddr looks sendable. Common verification has already
957 * been taken care of.
958 */
959static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
960{
961 return 1;
962}
963
964/* Fill in Supported Address Type information for INIT and INIT-ACK
965 * chunks. Returns number of addresses supported.
966 */
967static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
968 __be16 *types)
969{
970 types[0] = SCTP_PARAM_IPV4_ADDRESS;
971 return 1;
972}
973
974/* Wrapper routine that calls the ip transmit routine. */
975static inline int sctp_v4_xmit(struct sk_buff *skb,
976 struct sctp_transport *transport)
977{
978 struct inet_sock *inet = inet_sk(skb->sk);
979
980 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
981 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
982
983 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
984 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
985
986 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
987
988 return ip_queue_xmit(&inet->sk, skb, &transport->fl);
989}
990
991static struct sctp_af sctp_af_inet;
992
993static struct sctp_pf sctp_pf_inet = {
994 .event_msgname = sctp_inet_event_msgname,
995 .skb_msgname = sctp_inet_skb_msgname,
996 .af_supported = sctp_inet_af_supported,
997 .cmp_addr = sctp_inet_cmp_addr,
998 .bind_verify = sctp_inet_bind_verify,
999 .send_verify = sctp_inet_send_verify,
1000 .supported_addrs = sctp_inet_supported_addrs,
1001 .create_accept_sk = sctp_v4_create_accept_sk,
1002 .addr_to_user = sctp_v4_addr_to_user,
1003 .to_sk_saddr = sctp_v4_to_sk_saddr,
1004 .to_sk_daddr = sctp_v4_to_sk_daddr,
1005 .af = &sctp_af_inet
1006};
1007
1008/* Notifier for inetaddr addition/deletion events. */
1009static struct notifier_block sctp_inetaddr_notifier = {
1010 .notifier_call = sctp_inetaddr_event,
1011};
1012
1013/* Socket operations. */
1014static const struct proto_ops inet_seqpacket_ops = {
1015 .family = PF_INET,
1016 .owner = THIS_MODULE,
1017 .release = inet_release, /* Needs to be wrapped... */
1018 .bind = inet_bind,
1019 .connect = inet_dgram_connect,
1020 .socketpair = sock_no_socketpair,
1021 .accept = inet_accept,
1022 .getname = inet_getname, /* Semantics are different. */
1023 .poll = sctp_poll,
1024 .ioctl = inet_ioctl,
1025 .listen = sctp_inet_listen,
1026 .shutdown = inet_shutdown, /* Looks harmless. */
1027 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1028 .getsockopt = sock_common_getsockopt,
1029 .sendmsg = inet_sendmsg,
1030 .recvmsg = sock_common_recvmsg,
1031 .mmap = sock_no_mmap,
1032 .sendpage = sock_no_sendpage,
1033#ifdef CONFIG_COMPAT
1034 .compat_setsockopt = compat_sock_common_setsockopt,
1035 .compat_getsockopt = compat_sock_common_getsockopt,
1036#endif
1037};
1038
1039/* Registration with AF_INET family. */
1040static struct inet_protosw sctp_seqpacket_protosw = {
1041 .type = SOCK_SEQPACKET,
1042 .protocol = IPPROTO_SCTP,
1043 .prot = &sctp_prot,
1044 .ops = &inet_seqpacket_ops,
1045 .flags = SCTP_PROTOSW_FLAG
1046};
1047static struct inet_protosw sctp_stream_protosw = {
1048 .type = SOCK_STREAM,
1049 .protocol = IPPROTO_SCTP,
1050 .prot = &sctp_prot,
1051 .ops = &inet_seqpacket_ops,
1052 .flags = SCTP_PROTOSW_FLAG
1053};
1054
1055/* Register with IP layer. */
1056static const struct net_protocol sctp_protocol = {
1057 .handler = sctp_rcv,
1058 .err_handler = sctp_v4_err,
1059 .no_policy = 1,
1060 .netns_ok = 1,
1061 .icmp_strict_tag_validation = 1,
1062};
1063
1064/* IPv4 address related functions. */
1065static struct sctp_af sctp_af_inet = {
1066 .sa_family = AF_INET,
1067 .sctp_xmit = sctp_v4_xmit,
1068 .setsockopt = ip_setsockopt,
1069 .getsockopt = ip_getsockopt,
1070 .get_dst = sctp_v4_get_dst,
1071 .get_saddr = sctp_v4_get_saddr,
1072 .copy_addrlist = sctp_v4_copy_addrlist,
1073 .from_skb = sctp_v4_from_skb,
1074 .from_sk = sctp_v4_from_sk,
1075 .from_addr_param = sctp_v4_from_addr_param,
1076 .to_addr_param = sctp_v4_to_addr_param,
1077 .cmp_addr = sctp_v4_cmp_addr,
1078 .addr_valid = sctp_v4_addr_valid,
1079 .inaddr_any = sctp_v4_inaddr_any,
1080 .is_any = sctp_v4_is_any,
1081 .available = sctp_v4_available,
1082 .scope = sctp_v4_scope,
1083 .skb_iif = sctp_v4_skb_iif,
1084 .is_ce = sctp_v4_is_ce,
1085 .seq_dump_addr = sctp_v4_seq_dump_addr,
1086 .ecn_capable = sctp_v4_ecn_capable,
1087 .net_header_len = sizeof(struct iphdr),
1088 .sockaddr_len = sizeof(struct sockaddr_in),
1089#ifdef CONFIG_COMPAT
1090 .compat_setsockopt = compat_ip_setsockopt,
1091 .compat_getsockopt = compat_ip_getsockopt,
1092#endif
1093};
1094
1095struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1096{
1097 switch (family) {
1098 case PF_INET:
1099 return sctp_pf_inet_specific;
1100 case PF_INET6:
1101 return sctp_pf_inet6_specific;
1102 default:
1103 return NULL;
1104 }
1105}
1106
1107/* Register the PF specific function table. */
1108int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1109{
1110 switch (family) {
1111 case PF_INET:
1112 if (sctp_pf_inet_specific)
1113 return 0;
1114 sctp_pf_inet_specific = pf;
1115 break;
1116 case PF_INET6:
1117 if (sctp_pf_inet6_specific)
1118 return 0;
1119 sctp_pf_inet6_specific = pf;
1120 break;
1121 default:
1122 return 0;
1123 }
1124 return 1;
1125}
1126
1127static inline int init_sctp_mibs(struct net *net)
1128{
1129 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1130 if (!net->sctp.sctp_statistics)
1131 return -ENOMEM;
1132 return 0;
1133}
1134
1135static inline void cleanup_sctp_mibs(struct net *net)
1136{
1137 free_percpu(net->sctp.sctp_statistics);
1138}
1139
1140static void sctp_v4_pf_init(void)
1141{
1142 /* Initialize the SCTP specific PF functions. */
1143 sctp_register_pf(&sctp_pf_inet, PF_INET);
1144 sctp_register_af(&sctp_af_inet);
1145}
1146
1147static void sctp_v4_pf_exit(void)
1148{
1149 list_del(&sctp_af_inet.list);
1150}
1151
1152static int sctp_v4_protosw_init(void)
1153{
1154 int rc;
1155
1156 rc = proto_register(&sctp_prot, 1);
1157 if (rc)
1158 return rc;
1159
1160 /* Register SCTP(UDP and TCP style) with socket layer. */
1161 inet_register_protosw(&sctp_seqpacket_protosw);
1162 inet_register_protosw(&sctp_stream_protosw);
1163
1164 return 0;
1165}
1166
1167static void sctp_v4_protosw_exit(void)
1168{
1169 inet_unregister_protosw(&sctp_stream_protosw);
1170 inet_unregister_protosw(&sctp_seqpacket_protosw);
1171 proto_unregister(&sctp_prot);
1172}
1173
1174static int sctp_v4_add_protocol(void)
1175{
1176 /* Register notifier for inet address additions/deletions. */
1177 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1178
1179 /* Register SCTP with inet layer. */
1180 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1181 return -EAGAIN;
1182
1183 return 0;
1184}
1185
1186static void sctp_v4_del_protocol(void)
1187{
1188 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1189 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1190}
1191
1192static int __net_init sctp_defaults_init(struct net *net)
1193{
1194 int status;
1195
1196 /*
1197 * 14. Suggested SCTP Protocol Parameter Values
1198 */
1199 /* The following protocol parameters are RECOMMENDED: */
1200 /* RTO.Initial - 3 seconds */
1201 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1202 /* RTO.Min - 1 second */
1203 net->sctp.rto_min = SCTP_RTO_MIN;
1204 /* RTO.Max - 60 seconds */
1205 net->sctp.rto_max = SCTP_RTO_MAX;
1206 /* RTO.Alpha - 1/8 */
1207 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1208 /* RTO.Beta - 1/4 */
1209 net->sctp.rto_beta = SCTP_RTO_BETA;
1210
1211 /* Valid.Cookie.Life - 60 seconds */
1212 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1213
1214 /* Whether Cookie Preservative is enabled(1) or not(0) */
1215 net->sctp.cookie_preserve_enable = 1;
1216
1217 /* Default sctp sockets to use md5 as their hmac alg */
1218#if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1219 net->sctp.sctp_hmac_alg = "md5";
1220#elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1221 net->sctp.sctp_hmac_alg = "sha1";
1222#else
1223 net->sctp.sctp_hmac_alg = NULL;
1224#endif
1225
1226 /* Max.Burst - 4 */
1227 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1228
1229 /* Enable pf state by default */
1230 net->sctp.pf_enable = 1;
1231
1232 /* Association.Max.Retrans - 10 attempts
1233 * Path.Max.Retrans - 5 attempts (per destination address)
1234 * Max.Init.Retransmits - 8 attempts
1235 */
1236 net->sctp.max_retrans_association = 10;
1237 net->sctp.max_retrans_path = 5;
1238 net->sctp.max_retrans_init = 8;
1239
1240 /* Sendbuffer growth - do per-socket accounting */
1241 net->sctp.sndbuf_policy = 0;
1242
1243 /* Rcvbuffer growth - do per-socket accounting */
1244 net->sctp.rcvbuf_policy = 0;
1245
1246 /* HB.interval - 30 seconds */
1247 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1248
1249 /* delayed SACK timeout */
1250 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1251
1252 /* Disable ADDIP by default. */
1253 net->sctp.addip_enable = 0;
1254 net->sctp.addip_noauth = 0;
1255 net->sctp.default_auto_asconf = 0;
1256
1257 /* Enable PR-SCTP by default. */
1258 net->sctp.prsctp_enable = 1;
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 /* Initialize proc fs directory. */
1282 status = sctp_proc_init(net);
1283 if (status)
1284 goto err_init_proc;
1285
1286 sctp_dbg_objcnt_init(net);
1287
1288 /* Initialize the local address list. */
1289 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1290 spin_lock_init(&net->sctp.local_addr_lock);
1291 sctp_get_local_addr_list(net);
1292
1293 /* Initialize the address event list */
1294 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1295 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1296 spin_lock_init(&net->sctp.addr_wq_lock);
1297 net->sctp.addr_wq_timer.expires = 0;
1298 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1299 (unsigned long)net);
1300
1301 return 0;
1302
1303err_init_proc:
1304 cleanup_sctp_mibs(net);
1305err_init_mibs:
1306 sctp_sysctl_net_unregister(net);
1307err_sysctl_register:
1308 return status;
1309}
1310
1311static void __net_exit sctp_defaults_exit(struct net *net)
1312{
1313 /* Free the local address list */
1314 sctp_free_addr_wq(net);
1315 sctp_free_local_addr_list(net);
1316
1317 sctp_dbg_objcnt_exit(net);
1318
1319 sctp_proc_exit(net);
1320 cleanup_sctp_mibs(net);
1321 sctp_sysctl_net_unregister(net);
1322}
1323
1324static struct pernet_operations sctp_defaults_ops = {
1325 .init = sctp_defaults_init,
1326 .exit = sctp_defaults_exit,
1327};
1328
1329static int __net_init sctp_ctrlsock_init(struct net *net)
1330{
1331 int status;
1332
1333 /* Initialize the control inode/socket for handling OOTB packets. */
1334 status = sctp_ctl_sock_init(net);
1335 if (status)
1336 pr_err("Failed to initialize the SCTP control sock\n");
1337
1338 return status;
1339}
1340
1341static void __net_init sctp_ctrlsock_exit(struct net *net)
1342{
1343 /* Free the control endpoint. */
1344 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1345}
1346
1347static struct pernet_operations sctp_ctrlsock_ops = {
1348 .init = sctp_ctrlsock_init,
1349 .exit = sctp_ctrlsock_exit,
1350};
1351
1352/* Initialize the universe into something sensible. */
1353static __init int sctp_init(void)
1354{
1355 int i;
1356 int status = -EINVAL;
1357 unsigned long goal;
1358 unsigned long limit;
1359 int max_share;
1360 int order;
1361 int num_entries;
1362 int max_entry_order;
1363
1364 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1365
1366 /* Allocate bind_bucket and chunk caches. */
1367 status = -ENOBUFS;
1368 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1369 sizeof(struct sctp_bind_bucket),
1370 0, SLAB_HWCACHE_ALIGN,
1371 NULL);
1372 if (!sctp_bucket_cachep)
1373 goto out;
1374
1375 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1376 sizeof(struct sctp_chunk),
1377 0, SLAB_HWCACHE_ALIGN,
1378 NULL);
1379 if (!sctp_chunk_cachep)
1380 goto err_chunk_cachep;
1381
1382 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1383 if (status)
1384 goto err_percpu_counter_init;
1385
1386 /* Implementation specific variables. */
1387
1388 /* Initialize default stream count setup information. */
1389 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1390 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1391
1392 /* Initialize handle used for association ids. */
1393 idr_init(&sctp_assocs_id);
1394
1395 limit = nr_free_buffer_pages() / 8;
1396 limit = max(limit, 128UL);
1397 sysctl_sctp_mem[0] = limit / 4 * 3;
1398 sysctl_sctp_mem[1] = limit;
1399 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1400
1401 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1402 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1403 max_share = min(4UL*1024*1024, limit);
1404
1405 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1406 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1407 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1408
1409 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1410 sysctl_sctp_wmem[1] = 16*1024;
1411 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1412
1413 /* Size and allocate the association hash table.
1414 * The methodology is similar to that of the tcp hash tables.
1415 * Though not identical. Start by getting a goal size
1416 */
1417 if (totalram_pages >= (128 * 1024))
1418 goal = totalram_pages >> (22 - PAGE_SHIFT);
1419 else
1420 goal = totalram_pages >> (24 - PAGE_SHIFT);
1421
1422 /* Then compute the page order for said goal */
1423 order = get_order(goal);
1424
1425 /* Now compute the required page order for the maximum sized table we
1426 * want to create
1427 */
1428 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1429 sizeof(struct sctp_bind_hashbucket));
1430
1431 /* Limit the page order by that maximum hash table size */
1432 order = min(order, max_entry_order);
1433
1434 /* Allocate and initialize the endpoint hash table. */
1435 sctp_ep_hashsize = 64;
1436 sctp_ep_hashtable =
1437 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1438 if (!sctp_ep_hashtable) {
1439 pr_err("Failed endpoint_hash alloc\n");
1440 status = -ENOMEM;
1441 goto err_ehash_alloc;
1442 }
1443 for (i = 0; i < sctp_ep_hashsize; i++) {
1444 rwlock_init(&sctp_ep_hashtable[i].lock);
1445 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1446 }
1447
1448 /* Allocate and initialize the SCTP port hash table.
1449 * Note that order is initalized to start at the max sized
1450 * table we want to support. If we can't get that many pages
1451 * reduce the order and try again
1452 */
1453 do {
1454 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1455 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1456 } while (!sctp_port_hashtable && --order > 0);
1457
1458 if (!sctp_port_hashtable) {
1459 pr_err("Failed bind hash alloc\n");
1460 status = -ENOMEM;
1461 goto err_bhash_alloc;
1462 }
1463
1464 /* Now compute the number of entries that will fit in the
1465 * port hash space we allocated
1466 */
1467 num_entries = (1UL << order) * PAGE_SIZE /
1468 sizeof(struct sctp_bind_hashbucket);
1469
1470 /* And finish by rounding it down to the nearest power of two
1471 * this wastes some memory of course, but its needed because
1472 * the hash function operates based on the assumption that
1473 * that the number of entries is a power of two
1474 */
1475 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1476
1477 for (i = 0; i < sctp_port_hashsize; i++) {
1478 spin_lock_init(&sctp_port_hashtable[i].lock);
1479 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1480 }
1481
1482 if (sctp_transport_hashtable_init())
1483 goto err_thash_alloc;
1484
1485 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1486 num_entries);
1487
1488 sctp_sysctl_register();
1489
1490 INIT_LIST_HEAD(&sctp_address_families);
1491 sctp_v4_pf_init();
1492 sctp_v6_pf_init();
1493
1494 status = register_pernet_subsys(&sctp_defaults_ops);
1495 if (status)
1496 goto err_register_defaults;
1497
1498 status = sctp_v4_protosw_init();
1499 if (status)
1500 goto err_protosw_init;
1501
1502 status = sctp_v6_protosw_init();
1503 if (status)
1504 goto err_v6_protosw_init;
1505
1506 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1507 if (status)
1508 goto err_register_ctrlsock;
1509
1510 status = sctp_v4_add_protocol();
1511 if (status)
1512 goto err_add_protocol;
1513
1514 /* Register SCTP with inet6 layer. */
1515 status = sctp_v6_add_protocol();
1516 if (status)
1517 goto err_v6_add_protocol;
1518
1519out:
1520 return status;
1521err_v6_add_protocol:
1522 sctp_v4_del_protocol();
1523err_add_protocol:
1524 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1525err_register_ctrlsock:
1526 sctp_v6_protosw_exit();
1527err_v6_protosw_init:
1528 sctp_v4_protosw_exit();
1529err_protosw_init:
1530 unregister_pernet_subsys(&sctp_defaults_ops);
1531err_register_defaults:
1532 sctp_v4_pf_exit();
1533 sctp_v6_pf_exit();
1534 sctp_sysctl_unregister();
1535 free_pages((unsigned long)sctp_port_hashtable,
1536 get_order(sctp_port_hashsize *
1537 sizeof(struct sctp_bind_hashbucket)));
1538err_bhash_alloc:
1539 sctp_transport_hashtable_destroy();
1540err_thash_alloc:
1541 kfree(sctp_ep_hashtable);
1542err_ehash_alloc:
1543 percpu_counter_destroy(&sctp_sockets_allocated);
1544err_percpu_counter_init:
1545 kmem_cache_destroy(sctp_chunk_cachep);
1546err_chunk_cachep:
1547 kmem_cache_destroy(sctp_bucket_cachep);
1548 goto out;
1549}
1550
1551/* Exit handler for the SCTP protocol. */
1552static __exit void sctp_exit(void)
1553{
1554 /* BUG. This should probably do something useful like clean
1555 * up all the remaining associations and all that memory.
1556 */
1557
1558 /* Unregister with inet6/inet layers. */
1559 sctp_v6_del_protocol();
1560 sctp_v4_del_protocol();
1561
1562 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1563
1564 /* Free protosw registrations */
1565 sctp_v6_protosw_exit();
1566 sctp_v4_protosw_exit();
1567
1568 unregister_pernet_subsys(&sctp_defaults_ops);
1569
1570 /* Unregister with socket layer. */
1571 sctp_v6_pf_exit();
1572 sctp_v4_pf_exit();
1573
1574 sctp_sysctl_unregister();
1575
1576 free_pages((unsigned long)sctp_port_hashtable,
1577 get_order(sctp_port_hashsize *
1578 sizeof(struct sctp_bind_hashbucket)));
1579 kfree(sctp_ep_hashtable);
1580 sctp_transport_hashtable_destroy();
1581
1582 percpu_counter_destroy(&sctp_sockets_allocated);
1583
1584 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1585
1586 kmem_cache_destroy(sctp_chunk_cachep);
1587 kmem_cache_destroy(sctp_bucket_cachep);
1588}
1589
1590module_init(sctp_init);
1591module_exit(sctp_exit);
1592
1593/*
1594 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1595 */
1596MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1597MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1598MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1599MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1600module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1601MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1602MODULE_LICENSE("GPL");