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