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/slab.h>
37 #include <net/net_namespace.h>
38 #include <net/protocol.h>
39 #include <net/ip.h>
40 #include <net/ipv6.h>
41 #include <net/route.h>
42 #include <net/sctp/sctp.h>
43 #include <net/addrconf.h>
44 #include <net/inet_common.h>
45 #include <net/inet_ecn.h>
46 #include <net/udp_tunnel.h>
47
48 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
49
50 /* Global data structures. */
51 struct sctp_globals sctp_globals __read_mostly;
52
53 struct idr sctp_assocs_id;
54 DEFINE_SPINLOCK(sctp_assocs_id_lock);
55
56 static struct sctp_pf *sctp_pf_inet6_specific;
57 static struct sctp_pf *sctp_pf_inet_specific;
58 static struct sctp_af *sctp_af_v4_specific;
59 static struct sctp_af *sctp_af_v6_specific;
60
61 struct kmem_cache *sctp_chunk_cachep __read_mostly;
62 struct kmem_cache *sctp_bucket_cachep __read_mostly;
63
64 long sysctl_sctp_mem[3];
65 int sysctl_sctp_rmem[3];
66 int sysctl_sctp_wmem[3];
67
68 /* Private helper to extract ipv4 address and stash them in
69 * the protocol structure.
70 */
71 static 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 */
102 static 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. */
119 static 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'. */
132 int 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_ALLOWED) ||
152 !(copy_flags & SCTP_ADDR4_PEERSUPP)))
153 continue;
154 if (addr->a.sa.sa_family == AF_INET6 &&
155 (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
156 !(copy_flags & SCTP_ADDR6_PEERSUPP)))
157 continue;
158
159 laddr = addr->a;
160 /* also works for setting ipv6 address port */
161 laddr.v4.sin_port = htons(bp->port);
162 if (sctp_bind_addr_state(bp, &laddr) != -1)
163 continue;
164
165 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
166 SCTP_ADDR_SRC, GFP_ATOMIC);
167 if (error)
168 break;
169 }
170
171 rcu_read_unlock();
172 return error;
173 }
174
175 /* Copy over any ip options */
176 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
177 {
178 struct inet_sock *newinet, *inet = inet_sk(sk);
179 struct ip_options_rcu *inet_opt, *newopt = NULL;
180
181 newinet = inet_sk(newsk);
182
183 rcu_read_lock();
184 inet_opt = rcu_dereference(inet->inet_opt);
185 if (inet_opt) {
186 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
187 inet_opt->opt.optlen, GFP_ATOMIC);
188 if (newopt)
189 memcpy(newopt, inet_opt, sizeof(*inet_opt) +
190 inet_opt->opt.optlen);
191 else
192 pr_err("%s: Failed to copy ip options\n", __func__);
193 }
194 RCU_INIT_POINTER(newinet->inet_opt, newopt);
195 rcu_read_unlock();
196 }
197
198 /* Account for the IP options */
199 static int sctp_v4_ip_options_len(struct sock *sk)
200 {
201 struct inet_sock *inet = inet_sk(sk);
202 struct ip_options_rcu *inet_opt;
203 int len = 0;
204
205 rcu_read_lock();
206 inet_opt = rcu_dereference(inet->inet_opt);
207 if (inet_opt)
208 len = inet_opt->opt.optlen;
209
210 rcu_read_unlock();
211 return len;
212 }
213
214 /* Initialize a sctp_addr from in incoming skb. */
215 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
216 int is_saddr)
217 {
218 /* Always called on head skb, so this is safe */
219 struct sctphdr *sh = sctp_hdr(skb);
220 struct sockaddr_in *sa = &addr->v4;
221
222 addr->v4.sin_family = AF_INET;
223
224 if (is_saddr) {
225 sa->sin_port = sh->source;
226 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
227 } else {
228 sa->sin_port = sh->dest;
229 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
230 }
231 memset(sa->sin_zero, 0, sizeof(sa->sin_zero));
232 }
233
234 /* Initialize an sctp_addr from a socket. */
235 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
236 {
237 addr->v4.sin_family = AF_INET;
238 addr->v4.sin_port = 0;
239 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
240 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
241 }
242
243 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
244 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
245 {
246 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
247 }
248
249 /* Initialize sk->sk_daddr from sctp_addr. */
250 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
251 {
252 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
253 }
254
255 /* Initialize a sctp_addr from an address parameter. */
256 static bool sctp_v4_from_addr_param(union sctp_addr *addr,
257 union sctp_addr_param *param,
258 __be16 port, int iif)
259 {
260 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param))
261 return false;
262
263 addr->v4.sin_family = AF_INET;
264 addr->v4.sin_port = port;
265 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
266 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
267
268 return true;
269 }
270
271 /* Initialize an address parameter from a sctp_addr and return the length
272 * of the address parameter.
273 */
274 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
275 union sctp_addr_param *param)
276 {
277 int length = sizeof(struct sctp_ipv4addr_param);
278
279 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
280 param->v4.param_hdr.length = htons(length);
281 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
282
283 return length;
284 }
285
286 /* Initialize a sctp_addr from a dst_entry. */
287 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
288 __be16 port)
289 {
290 saddr->v4.sin_family = AF_INET;
291 saddr->v4.sin_port = port;
292 saddr->v4.sin_addr.s_addr = fl4->saddr;
293 memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
294 }
295
296 /* Compare two addresses exactly. */
297 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
298 const union sctp_addr *addr2)
299 {
300 if (addr1->sa.sa_family != addr2->sa.sa_family)
301 return 0;
302 if (addr1->v4.sin_port != addr2->v4.sin_port)
303 return 0;
304 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
305 return 0;
306
307 return 1;
308 }
309
310 /* Initialize addr struct to INADDR_ANY. */
311 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
312 {
313 addr->v4.sin_family = AF_INET;
314 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
315 addr->v4.sin_port = port;
316 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
317 }
318
319 /* Is this a wildcard address? */
320 static int sctp_v4_is_any(const union sctp_addr *addr)
321 {
322 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
323 }
324
325 /* This function checks if the address is a valid address to be used for
326 * SCTP binding.
327 *
328 * Output:
329 * Return 0 - If the address is a non-unicast or an illegal address.
330 * Return 1 - If the address is a unicast.
331 */
332 static int sctp_v4_addr_valid(union sctp_addr *addr,
333 struct sctp_sock *sp,
334 const struct sk_buff *skb)
335 {
336 /* IPv4 addresses not allowed */
337 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
338 return 0;
339
340 /* Is this a non-unicast address or a unusable SCTP address? */
341 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
342 return 0;
343
344 /* Is this a broadcast address? */
345 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
346 return 0;
347
348 return 1;
349 }
350
351 /* Should this be available for binding? */
352 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
353 {
354 struct sock *sk = &sp->inet.sk;
355 struct net *net = sock_net(sk);
356 int tb_id = RT_TABLE_LOCAL;
357 int ret;
358
359 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ?: tb_id;
360 ret = inet_addr_type_table(net, addr->v4.sin_addr.s_addr, tb_id);
361 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
362 ret != RTN_LOCAL &&
363 !inet_test_bit(FREEBIND, sk) &&
364 !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind))
365 return 0;
366
367 if (ipv6_only_sock(sctp_opt2sk(sp)))
368 return 0;
369
370 return 1;
371 }
372
373 /* Checking the loopback, private and other address scopes as defined in
374 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
375 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
376 *
377 * Level 0 - unusable SCTP addresses
378 * Level 1 - loopback address
379 * Level 2 - link-local addresses
380 * Level 3 - private addresses.
381 * Level 4 - global addresses
382 * For INIT and INIT-ACK address list, let L be the level of
383 * requested destination address, sender and receiver
384 * SHOULD include all of its addresses with level greater
385 * than or equal to L.
386 *
387 * IPv4 scoping can be controlled through sysctl option
388 * net.sctp.addr_scope_policy
389 */
390 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
391 {
392 enum sctp_scope retval;
393
394 /* Check for unusable SCTP addresses. */
395 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
396 retval = SCTP_SCOPE_UNUSABLE;
397 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
398 retval = SCTP_SCOPE_LOOPBACK;
399 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
400 retval = SCTP_SCOPE_LINK;
401 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
402 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
403 ipv4_is_private_192(addr->v4.sin_addr.s_addr) ||
404 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) {
405 retval = SCTP_SCOPE_PRIVATE;
406 } else {
407 retval = SCTP_SCOPE_GLOBAL;
408 }
409
410 return retval;
411 }
412
413 /* Returns a valid dst cache entry for the given source and destination ip
414 * addresses. If an association is passed, trys to get a dst entry with a
415 * source address that matches an address in the bind address list.
416 */
417 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
418 struct flowi *fl, struct sock *sk)
419 {
420 struct sctp_association *asoc = t->asoc;
421 struct rtable *rt;
422 struct flowi _fl;
423 struct flowi4 *fl4 = &_fl.u.ip4;
424 struct sctp_bind_addr *bp;
425 struct sctp_sockaddr_entry *laddr;
426 struct dst_entry *dst = NULL;
427 union sctp_addr *daddr = &t->ipaddr;
428 union sctp_addr dst_saddr;
429 u8 tos = READ_ONCE(inet_sk(sk)->tos);
430
431 if (t->dscp & SCTP_DSCP_SET_MASK)
432 tos = t->dscp & SCTP_DSCP_VAL_MASK;
433 memset(&_fl, 0x0, sizeof(_fl));
434 fl4->daddr = daddr->v4.sin_addr.s_addr;
435 fl4->fl4_dport = daddr->v4.sin_port;
436 fl4->flowi4_proto = IPPROTO_SCTP;
437 if (asoc) {
438 fl4->flowi4_tos = RT_TOS(tos);
439 fl4->flowi4_scope = ip_sock_rt_scope(asoc->base.sk);
440 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
441 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
442 }
443 if (saddr) {
444 fl4->saddr = saddr->v4.sin_addr.s_addr;
445 if (!fl4->fl4_sport)
446 fl4->fl4_sport = saddr->v4.sin_port;
447 }
448
449 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
450 &fl4->saddr);
451
452 rt = ip_route_output_key(sock_net(sk), fl4);
453 if (!IS_ERR(rt)) {
454 dst = &rt->dst;
455 t->dst = dst;
456 memcpy(fl, &_fl, sizeof(_fl));
457 }
458
459 /* If there is no association or if a source address is passed, no
460 * more validation is required.
461 */
462 if (!asoc || saddr)
463 goto out;
464
465 bp = &asoc->base.bind_addr;
466
467 if (dst) {
468 /* Walk through the bind address list and look for a bind
469 * address that matches the source address of the returned dst.
470 */
471 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
472 rcu_read_lock();
473 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
474 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
475 (laddr->state != SCTP_ADDR_SRC &&
476 !asoc->src_out_of_asoc_ok))
477 continue;
478 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
479 goto out_unlock;
480 }
481 rcu_read_unlock();
482
483 /* None of the bound addresses match the source address of the
484 * dst. So release it.
485 */
486 dst_release(dst);
487 dst = NULL;
488 }
489
490 /* Walk through the bind address list and try to get a dst that
491 * matches a bind address as the source address.
492 */
493 rcu_read_lock();
494 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
495 struct net_device *odev;
496
497 if (!laddr->valid)
498 continue;
499 if (laddr->state != SCTP_ADDR_SRC ||
500 AF_INET != laddr->a.sa.sa_family)
501 continue;
502
503 fl4->fl4_sport = laddr->a.v4.sin_port;
504 flowi4_update_output(fl4, asoc->base.sk->sk_bound_dev_if,
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 /* Ensure the src address belongs to the output
513 * interface.
514 */
515 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
516 false);
517 if (!odev || odev->ifindex != fl4->flowi4_oif) {
518 if (!dst) {
519 dst = &rt->dst;
520 t->dst = dst;
521 memcpy(fl, &_fl, sizeof(_fl));
522 } else {
523 dst_release(&rt->dst);
524 }
525 continue;
526 }
527
528 dst_release(dst);
529 dst = &rt->dst;
530 t->dst = dst;
531 memcpy(fl, &_fl, sizeof(_fl));
532 break;
533 }
534
535 out_unlock:
536 rcu_read_unlock();
537 out:
538 if (dst) {
539 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
540 &fl->u.ip4.daddr, &fl->u.ip4.saddr);
541 } else {
542 t->dst = NULL;
543 pr_debug("no route\n");
544 }
545 }
546
547 /* For v4, the source address is cached in the route entry(dst). So no need
548 * to cache it separately and hence this is an empty routine.
549 */
550 static void sctp_v4_get_saddr(struct sctp_sock *sk,
551 struct sctp_transport *t,
552 struct flowi *fl)
553 {
554 union sctp_addr *saddr = &t->saddr;
555 struct rtable *rt = dst_rtable(t->dst);
556
557 if (rt) {
558 saddr->v4.sin_family = AF_INET;
559 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
560 }
561 }
562
563 /* What interface did this skb arrive on? */
564 static int sctp_v4_skb_iif(const struct sk_buff *skb)
565 {
566 return inet_iif(skb);
567 }
568
569 static int sctp_v4_skb_sdif(const struct sk_buff *skb)
570 {
571 return inet_sdif(skb);
572 }
573
574 /* Was this packet marked by Explicit Congestion Notification? */
575 static int sctp_v4_is_ce(const struct sk_buff *skb)
576 {
577 return INET_ECN_is_ce(ip_hdr(skb)->tos);
578 }
579
580 /* Create and initialize a new sk for the socket returned by accept(). */
581 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
582 struct sctp_association *asoc,
583 bool kern)
584 {
585 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
586 sk->sk_prot, kern);
587 struct inet_sock *newinet;
588
589 if (!newsk)
590 goto out;
591
592 sock_init_data(NULL, newsk);
593
594 sctp_copy_sock(newsk, sk, asoc);
595 sock_reset_flag(newsk, SOCK_ZAPPED);
596
597 sctp_v4_copy_ip_options(sk, newsk);
598
599 newinet = inet_sk(newsk);
600
601 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
602
603 if (newsk->sk_prot->init(newsk)) {
604 sk_common_release(newsk);
605 newsk = NULL;
606 }
607
608 out:
609 return newsk;
610 }
611
612 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
613 {
614 /* No address mapping for V4 sockets */
615 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
616 return sizeof(struct sockaddr_in);
617 }
618
619 /* Dump the v4 addr to the seq file. */
620 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
621 {
622 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
623 }
624
625 static void sctp_v4_ecn_capable(struct sock *sk)
626 {
627 INET_ECN_xmit(sk);
628 }
629
630 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
631 {
632 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
633 struct sctp_sockaddr_entry *addrw, *temp;
634 struct sctp_sock *sp;
635
636 spin_lock_bh(&net->sctp.addr_wq_lock);
637
638 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
639 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
640 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
641 addrw->state, addrw);
642
643 #if IS_ENABLED(CONFIG_IPV6)
644 /* Now we send an ASCONF for each association */
645 /* Note. we currently don't handle link local IPv6 addressees */
646 if (addrw->a.sa.sa_family == AF_INET6) {
647 struct in6_addr *in6;
648
649 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
650 IPV6_ADDR_LINKLOCAL)
651 goto free_next;
652
653 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
654 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
655 addrw->state == SCTP_ADDR_NEW) {
656 unsigned long timeo_val;
657
658 pr_debug("%s: this is on DAD, trying %d sec "
659 "later\n", __func__,
660 SCTP_ADDRESS_TICK_DELAY);
661
662 timeo_val = jiffies;
663 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
664 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
665 break;
666 }
667 }
668 #endif
669 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
670 struct sock *sk;
671
672 sk = sctp_opt2sk(sp);
673 /* ignore bound-specific endpoints */
674 if (!sctp_is_ep_boundall(sk))
675 continue;
676 bh_lock_sock(sk);
677 if (sctp_asconf_mgmt(sp, addrw) < 0)
678 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
679 bh_unlock_sock(sk);
680 }
681 #if IS_ENABLED(CONFIG_IPV6)
682 free_next:
683 #endif
684 list_del(&addrw->list);
685 kfree(addrw);
686 }
687 spin_unlock_bh(&net->sctp.addr_wq_lock);
688 }
689
690 static void sctp_free_addr_wq(struct net *net)
691 {
692 struct sctp_sockaddr_entry *addrw;
693 struct sctp_sockaddr_entry *temp;
694
695 spin_lock_bh(&net->sctp.addr_wq_lock);
696 del_timer(&net->sctp.addr_wq_timer);
697 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
698 list_del(&addrw->list);
699 kfree(addrw);
700 }
701 spin_unlock_bh(&net->sctp.addr_wq_lock);
702 }
703
704 /* lookup the entry for the same address in the addr_waitq
705 * sctp_addr_wq MUST be locked
706 */
707 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
708 struct sctp_sockaddr_entry *addr)
709 {
710 struct sctp_sockaddr_entry *addrw;
711
712 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
713 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
714 continue;
715 if (addrw->a.sa.sa_family == AF_INET) {
716 if (addrw->a.v4.sin_addr.s_addr ==
717 addr->a.v4.sin_addr.s_addr)
718 return addrw;
719 } else if (addrw->a.sa.sa_family == AF_INET6) {
720 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
721 &addr->a.v6.sin6_addr))
722 return addrw;
723 }
724 }
725 return NULL;
726 }
727
728 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
729 {
730 struct sctp_sockaddr_entry *addrw;
731 unsigned long timeo_val;
732
733 /* first, we check if an opposite message already exist in the queue.
734 * If we found such message, it is removed.
735 * This operation is a bit stupid, but the DHCP client attaches the
736 * new address after a couple of addition and deletion of that address
737 */
738
739 spin_lock_bh(&net->sctp.addr_wq_lock);
740 /* Offsets existing events in addr_wq */
741 addrw = sctp_addr_wq_lookup(net, addr);
742 if (addrw) {
743 if (addrw->state != cmd) {
744 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
745 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
746 &net->sctp.addr_waitq);
747
748 list_del(&addrw->list);
749 kfree(addrw);
750 }
751 spin_unlock_bh(&net->sctp.addr_wq_lock);
752 return;
753 }
754
755 /* OK, we have to add the new address to the wait queue */
756 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
757 if (addrw == NULL) {
758 spin_unlock_bh(&net->sctp.addr_wq_lock);
759 return;
760 }
761 addrw->state = cmd;
762 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
763
764 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
765 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
766
767 if (!timer_pending(&net->sctp.addr_wq_timer)) {
768 timeo_val = jiffies;
769 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
770 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
771 }
772 spin_unlock_bh(&net->sctp.addr_wq_lock);
773 }
774
775 /* Event handler for inet address addition/deletion events.
776 * The sctp_local_addr_list needs to be protocted by a spin lock since
777 * multiple notifiers (say IPv4 and IPv6) may be running at the same
778 * time and thus corrupt the list.
779 * The reader side is protected with RCU.
780 */
781 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
782 void *ptr)
783 {
784 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
785 struct sctp_sockaddr_entry *addr = NULL;
786 struct sctp_sockaddr_entry *temp;
787 struct net *net = dev_net(ifa->ifa_dev->dev);
788 int found = 0;
789
790 switch (ev) {
791 case NETDEV_UP:
792 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
793 if (addr) {
794 addr->a.v4.sin_family = AF_INET;
795 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
796 addr->valid = 1;
797 spin_lock_bh(&net->sctp.local_addr_lock);
798 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
799 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
800 spin_unlock_bh(&net->sctp.local_addr_lock);
801 }
802 break;
803 case NETDEV_DOWN:
804 spin_lock_bh(&net->sctp.local_addr_lock);
805 list_for_each_entry_safe(addr, temp,
806 &net->sctp.local_addr_list, list) {
807 if (addr->a.sa.sa_family == AF_INET &&
808 addr->a.v4.sin_addr.s_addr ==
809 ifa->ifa_local) {
810 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
811 found = 1;
812 addr->valid = 0;
813 list_del_rcu(&addr->list);
814 break;
815 }
816 }
817 spin_unlock_bh(&net->sctp.local_addr_lock);
818 if (found)
819 kfree_rcu(addr, rcu);
820 break;
821 }
822
823 return NOTIFY_DONE;
824 }
825
826 /*
827 * Initialize the control inode/socket with a control endpoint data
828 * structure. This endpoint is reserved exclusively for the OOTB processing.
829 */
830 static int sctp_ctl_sock_init(struct net *net)
831 {
832 int err;
833 sa_family_t family = PF_INET;
834
835 if (sctp_get_pf_specific(PF_INET6))
836 family = PF_INET6;
837
838 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
839 SOCK_SEQPACKET, IPPROTO_SCTP, net);
840
841 /* If IPv6 socket could not be created, try the IPv4 socket */
842 if (err < 0 && family == PF_INET6)
843 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
844 SOCK_SEQPACKET, IPPROTO_SCTP,
845 net);
846
847 if (err < 0) {
848 pr_err("Failed to create the SCTP control socket\n");
849 return err;
850 }
851 return 0;
852 }
853
854 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb)
855 {
856 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source;
857
858 skb_set_transport_header(skb, sizeof(struct udphdr));
859 sctp_rcv(skb);
860 return 0;
861 }
862
863 int sctp_udp_sock_start(struct net *net)
864 {
865 struct udp_tunnel_sock_cfg tuncfg = {NULL};
866 struct udp_port_cfg udp_conf = {0};
867 struct socket *sock;
868 int err;
869
870 udp_conf.family = AF_INET;
871 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
872 udp_conf.local_udp_port = htons(net->sctp.udp_port);
873 err = udp_sock_create(net, &udp_conf, &sock);
874 if (err) {
875 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n");
876 return err;
877 }
878
879 tuncfg.encap_type = 1;
880 tuncfg.encap_rcv = sctp_udp_rcv;
881 tuncfg.encap_err_lookup = sctp_udp_v4_err;
882 setup_udp_tunnel_sock(net, sock, &tuncfg);
883 net->sctp.udp4_sock = sock->sk;
884
885 #if IS_ENABLED(CONFIG_IPV6)
886 memset(&udp_conf, 0, sizeof(udp_conf));
887
888 udp_conf.family = AF_INET6;
889 udp_conf.local_ip6 = in6addr_any;
890 udp_conf.local_udp_port = htons(net->sctp.udp_port);
891 udp_conf.use_udp6_rx_checksums = true;
892 udp_conf.ipv6_v6only = true;
893 err = udp_sock_create(net, &udp_conf, &sock);
894 if (err) {
895 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n");
896 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
897 net->sctp.udp4_sock = NULL;
898 return err;
899 }
900
901 tuncfg.encap_type = 1;
902 tuncfg.encap_rcv = sctp_udp_rcv;
903 tuncfg.encap_err_lookup = sctp_udp_v6_err;
904 setup_udp_tunnel_sock(net, sock, &tuncfg);
905 net->sctp.udp6_sock = sock->sk;
906 #endif
907
908 return 0;
909 }
910
911 void sctp_udp_sock_stop(struct net *net)
912 {
913 if (net->sctp.udp4_sock) {
914 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
915 net->sctp.udp4_sock = NULL;
916 }
917 if (net->sctp.udp6_sock) {
918 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket);
919 net->sctp.udp6_sock = NULL;
920 }
921 }
922
923 /* Register address family specific functions. */
924 int sctp_register_af(struct sctp_af *af)
925 {
926 switch (af->sa_family) {
927 case AF_INET:
928 if (sctp_af_v4_specific)
929 return 0;
930 sctp_af_v4_specific = af;
931 break;
932 case AF_INET6:
933 if (sctp_af_v6_specific)
934 return 0;
935 sctp_af_v6_specific = af;
936 break;
937 default:
938 return 0;
939 }
940
941 INIT_LIST_HEAD(&af->list);
942 list_add_tail(&af->list, &sctp_address_families);
943 return 1;
944 }
945
946 /* Get the table of functions for manipulating a particular address
947 * family.
948 */
949 struct sctp_af *sctp_get_af_specific(sa_family_t family)
950 {
951 switch (family) {
952 case AF_INET:
953 return sctp_af_v4_specific;
954 case AF_INET6:
955 return sctp_af_v6_specific;
956 default:
957 return NULL;
958 }
959 }
960
961 /* Common code to initialize a AF_INET msg_name. */
962 static void sctp_inet_msgname(char *msgname, int *addr_len)
963 {
964 struct sockaddr_in *sin;
965
966 sin = (struct sockaddr_in *)msgname;
967 *addr_len = sizeof(struct sockaddr_in);
968 sin->sin_family = AF_INET;
969 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
970 }
971
972 /* Copy the primary address of the peer primary address as the msg_name. */
973 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
974 int *addr_len)
975 {
976 struct sockaddr_in *sin, *sinfrom;
977
978 if (msgname) {
979 struct sctp_association *asoc;
980
981 asoc = event->asoc;
982 sctp_inet_msgname(msgname, addr_len);
983 sin = (struct sockaddr_in *)msgname;
984 sinfrom = &asoc->peer.primary_addr.v4;
985 sin->sin_port = htons(asoc->peer.port);
986 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
987 }
988 }
989
990 /* Initialize and copy out a msgname from an inbound skb. */
991 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
992 {
993 if (msgname) {
994 struct sctphdr *sh = sctp_hdr(skb);
995 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
996
997 sctp_inet_msgname(msgname, len);
998 sin->sin_port = sh->source;
999 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1000 }
1001 }
1002
1003 /* Do we support this AF? */
1004 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
1005 {
1006 /* PF_INET only supports AF_INET addresses. */
1007 return AF_INET == family;
1008 }
1009
1010 /* Address matching with wildcards allowed. */
1011 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
1012 const union sctp_addr *addr2,
1013 struct sctp_sock *opt)
1014 {
1015 /* PF_INET only supports AF_INET addresses. */
1016 if (addr1->sa.sa_family != addr2->sa.sa_family)
1017 return 0;
1018 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
1019 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
1020 return 1;
1021 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
1022 return 1;
1023
1024 return 0;
1025 }
1026
1027 /* Verify that provided sockaddr looks bindable. Common verification has
1028 * already been taken care of.
1029 */
1030 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
1031 {
1032 return sctp_v4_available(addr, opt);
1033 }
1034
1035 /* Verify that sockaddr looks sendable. Common verification has already
1036 * been taken care of.
1037 */
1038 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
1039 {
1040 return 1;
1041 }
1042
1043 /* Fill in Supported Address Type information for INIT and INIT-ACK
1044 * chunks. Returns number of addresses supported.
1045 */
1046 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
1047 __be16 *types)
1048 {
1049 types[0] = SCTP_PARAM_IPV4_ADDRESS;
1050 return 1;
1051 }
1052
1053 /* Wrapper routine that calls the ip transmit routine. */
1054 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t)
1055 {
1056 struct dst_entry *dst = dst_clone(t->dst);
1057 struct flowi4 *fl4 = &t->fl.u.ip4;
1058 struct sock *sk = skb->sk;
1059 struct inet_sock *inet = inet_sk(sk);
1060 __u8 dscp = READ_ONCE(inet->tos);
1061 __be16 df = 0;
1062
1063 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
1064 skb->len, &fl4->saddr, &fl4->daddr);
1065
1066 if (t->dscp & SCTP_DSCP_SET_MASK)
1067 dscp = t->dscp & SCTP_DSCP_VAL_MASK;
1068
1069 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO
1070 : IP_PMTUDISC_DONT;
1071 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS);
1072
1073 if (!t->encap_port || !sctp_sk(sk)->udp_port) {
1074 skb_dst_set(skb, dst);
1075 return __ip_queue_xmit(sk, skb, &t->fl, dscp);
1076 }
1077
1078 if (skb_is_gso(skb))
1079 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
1080
1081 if (ip_dont_fragment(sk, dst) && !skb->ignore_df)
1082 df = htons(IP_DF);
1083
1084 skb->encapsulation = 1;
1085 skb_reset_inner_mac_header(skb);
1086 skb_reset_inner_transport_header(skb);
1087 skb_set_inner_ipproto(skb, IPPROTO_SCTP);
1088 udp_tunnel_xmit_skb(dst_rtable(dst), sk, skb, fl4->saddr,
1089 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df,
1090 sctp_sk(sk)->udp_port, t->encap_port, false, false);
1091 return 0;
1092 }
1093
1094 static struct sctp_af sctp_af_inet;
1095
1096 static struct sctp_pf sctp_pf_inet = {
1097 .event_msgname = sctp_inet_event_msgname,
1098 .skb_msgname = sctp_inet_skb_msgname,
1099 .af_supported = sctp_inet_af_supported,
1100 .cmp_addr = sctp_inet_cmp_addr,
1101 .bind_verify = sctp_inet_bind_verify,
1102 .send_verify = sctp_inet_send_verify,
1103 .supported_addrs = sctp_inet_supported_addrs,
1104 .create_accept_sk = sctp_v4_create_accept_sk,
1105 .addr_to_user = sctp_v4_addr_to_user,
1106 .to_sk_saddr = sctp_v4_to_sk_saddr,
1107 .to_sk_daddr = sctp_v4_to_sk_daddr,
1108 .copy_ip_options = sctp_v4_copy_ip_options,
1109 .af = &sctp_af_inet
1110 };
1111
1112 /* Notifier for inetaddr addition/deletion events. */
1113 static struct notifier_block sctp_inetaddr_notifier = {
1114 .notifier_call = sctp_inetaddr_event,
1115 };
1116
1117 /* Socket operations. */
1118 static const struct proto_ops inet_seqpacket_ops = {
1119 .family = PF_INET,
1120 .owner = THIS_MODULE,
1121 .release = inet_release, /* Needs to be wrapped... */
1122 .bind = inet_bind,
1123 .connect = sctp_inet_connect,
1124 .socketpair = sock_no_socketpair,
1125 .accept = inet_accept,
1126 .getname = inet_getname, /* Semantics are different. */
1127 .poll = sctp_poll,
1128 .ioctl = inet_ioctl,
1129 .gettstamp = sock_gettstamp,
1130 .listen = sctp_inet_listen,
1131 .shutdown = inet_shutdown, /* Looks harmless. */
1132 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1133 .getsockopt = sock_common_getsockopt,
1134 .sendmsg = inet_sendmsg,
1135 .recvmsg = inet_recvmsg,
1136 .mmap = sock_no_mmap,
1137 };
1138
1139 /* Registration with AF_INET family. */
1140 static struct inet_protosw sctp_seqpacket_protosw = {
1141 .type = SOCK_SEQPACKET,
1142 .protocol = IPPROTO_SCTP,
1143 .prot = &sctp_prot,
1144 .ops = &inet_seqpacket_ops,
1145 .flags = SCTP_PROTOSW_FLAG
1146 };
1147 static struct inet_protosw sctp_stream_protosw = {
1148 .type = SOCK_STREAM,
1149 .protocol = IPPROTO_SCTP,
1150 .prot = &sctp_prot,
1151 .ops = &inet_seqpacket_ops,
1152 .flags = SCTP_PROTOSW_FLAG
1153 };
1154
1155 static int sctp4_rcv(struct sk_buff *skb)
1156 {
1157 SCTP_INPUT_CB(skb)->encap_port = 0;
1158 return sctp_rcv(skb);
1159 }
1160
1161 /* Register with IP layer. */
1162 static const struct net_protocol sctp_protocol = {
1163 .handler = sctp4_rcv,
1164 .err_handler = sctp_v4_err,
1165 .no_policy = 1,
1166 .icmp_strict_tag_validation = 1,
1167 };
1168
1169 /* IPv4 address related functions. */
1170 static struct sctp_af sctp_af_inet = {
1171 .sa_family = AF_INET,
1172 .sctp_xmit = sctp_v4_xmit,
1173 .setsockopt = ip_setsockopt,
1174 .getsockopt = ip_getsockopt,
1175 .get_dst = sctp_v4_get_dst,
1176 .get_saddr = sctp_v4_get_saddr,
1177 .copy_addrlist = sctp_v4_copy_addrlist,
1178 .from_skb = sctp_v4_from_skb,
1179 .from_sk = sctp_v4_from_sk,
1180 .from_addr_param = sctp_v4_from_addr_param,
1181 .to_addr_param = sctp_v4_to_addr_param,
1182 .cmp_addr = sctp_v4_cmp_addr,
1183 .addr_valid = sctp_v4_addr_valid,
1184 .inaddr_any = sctp_v4_inaddr_any,
1185 .is_any = sctp_v4_is_any,
1186 .available = sctp_v4_available,
1187 .scope = sctp_v4_scope,
1188 .skb_iif = sctp_v4_skb_iif,
1189 .skb_sdif = sctp_v4_skb_sdif,
1190 .is_ce = sctp_v4_is_ce,
1191 .seq_dump_addr = sctp_v4_seq_dump_addr,
1192 .ecn_capable = sctp_v4_ecn_capable,
1193 .net_header_len = sizeof(struct iphdr),
1194 .sockaddr_len = sizeof(struct sockaddr_in),
1195 .ip_options_len = sctp_v4_ip_options_len,
1196 };
1197
1198 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1199 {
1200 switch (family) {
1201 case PF_INET:
1202 return sctp_pf_inet_specific;
1203 case PF_INET6:
1204 return sctp_pf_inet6_specific;
1205 default:
1206 return NULL;
1207 }
1208 }
1209
1210 /* Register the PF specific function table. */
1211 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1212 {
1213 switch (family) {
1214 case PF_INET:
1215 if (sctp_pf_inet_specific)
1216 return 0;
1217 sctp_pf_inet_specific = pf;
1218 break;
1219 case PF_INET6:
1220 if (sctp_pf_inet6_specific)
1221 return 0;
1222 sctp_pf_inet6_specific = pf;
1223 break;
1224 default:
1225 return 0;
1226 }
1227 return 1;
1228 }
1229
1230 static inline int init_sctp_mibs(struct net *net)
1231 {
1232 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1233 if (!net->sctp.sctp_statistics)
1234 return -ENOMEM;
1235 return 0;
1236 }
1237
1238 static inline void cleanup_sctp_mibs(struct net *net)
1239 {
1240 free_percpu(net->sctp.sctp_statistics);
1241 }
1242
1243 static void sctp_v4_pf_init(void)
1244 {
1245 /* Initialize the SCTP specific PF functions. */
1246 sctp_register_pf(&sctp_pf_inet, PF_INET);
1247 sctp_register_af(&sctp_af_inet);
1248 }
1249
1250 static void sctp_v4_pf_exit(void)
1251 {
1252 list_del(&sctp_af_inet.list);
1253 }
1254
1255 static int sctp_v4_protosw_init(void)
1256 {
1257 int rc;
1258
1259 rc = proto_register(&sctp_prot, 1);
1260 if (rc)
1261 return rc;
1262
1263 /* Register SCTP(UDP and TCP style) with socket layer. */
1264 inet_register_protosw(&sctp_seqpacket_protosw);
1265 inet_register_protosw(&sctp_stream_protosw);
1266
1267 return 0;
1268 }
1269
1270 static void sctp_v4_protosw_exit(void)
1271 {
1272 inet_unregister_protosw(&sctp_stream_protosw);
1273 inet_unregister_protosw(&sctp_seqpacket_protosw);
1274 proto_unregister(&sctp_prot);
1275 }
1276
1277 static int sctp_v4_add_protocol(void)
1278 {
1279 /* Register notifier for inet address additions/deletions. */
1280 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1281
1282 /* Register SCTP with inet layer. */
1283 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1284 return -EAGAIN;
1285
1286 return 0;
1287 }
1288
1289 static void sctp_v4_del_protocol(void)
1290 {
1291 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1292 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1293 }
1294
1295 static int __net_init sctp_defaults_init(struct net *net)
1296 {
1297 int status;
1298
1299 /*
1300 * 14. Suggested SCTP Protocol Parameter Values
1301 */
1302 /* The following protocol parameters are RECOMMENDED: */
1303 /* RTO.Initial - 3 seconds */
1304 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1305 /* RTO.Min - 1 second */
1306 net->sctp.rto_min = SCTP_RTO_MIN;
1307 /* RTO.Max - 60 seconds */
1308 net->sctp.rto_max = SCTP_RTO_MAX;
1309 /* RTO.Alpha - 1/8 */
1310 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1311 /* RTO.Beta - 1/4 */
1312 net->sctp.rto_beta = SCTP_RTO_BETA;
1313
1314 /* Valid.Cookie.Life - 60 seconds */
1315 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1316
1317 /* Whether Cookie Preservative is enabled(1) or not(0) */
1318 net->sctp.cookie_preserve_enable = 1;
1319
1320 /* Default sctp sockets to use md5 as their hmac alg */
1321 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1322 net->sctp.sctp_hmac_alg = "md5";
1323 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1324 net->sctp.sctp_hmac_alg = "sha1";
1325 #else
1326 net->sctp.sctp_hmac_alg = NULL;
1327 #endif
1328
1329 /* Max.Burst - 4 */
1330 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1331
1332 /* Disable of Primary Path Switchover by default */
1333 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1334
1335 /* Enable pf state by default */
1336 net->sctp.pf_enable = 1;
1337
1338 /* Ignore pf exposure feature by default */
1339 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1340
1341 /* Association.Max.Retrans - 10 attempts
1342 * Path.Max.Retrans - 5 attempts (per destination address)
1343 * Max.Init.Retransmits - 8 attempts
1344 */
1345 net->sctp.max_retrans_association = 10;
1346 net->sctp.max_retrans_path = 5;
1347 net->sctp.max_retrans_init = 8;
1348
1349 /* Sendbuffer growth - do per-socket accounting */
1350 net->sctp.sndbuf_policy = 0;
1351
1352 /* Rcvbuffer growth - do per-socket accounting */
1353 net->sctp.rcvbuf_policy = 0;
1354
1355 /* HB.interval - 30 seconds */
1356 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1357
1358 /* delayed SACK timeout */
1359 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1360
1361 /* Disable ADDIP by default. */
1362 net->sctp.addip_enable = 0;
1363 net->sctp.addip_noauth = 0;
1364 net->sctp.default_auto_asconf = 0;
1365
1366 /* Enable PR-SCTP by default. */
1367 net->sctp.prsctp_enable = 1;
1368
1369 /* Disable RECONF by default. */
1370 net->sctp.reconf_enable = 0;
1371
1372 /* Disable AUTH by default. */
1373 net->sctp.auth_enable = 0;
1374
1375 /* Enable ECN by default. */
1376 net->sctp.ecn_enable = 1;
1377
1378 /* Set UDP tunneling listening port to 0 by default */
1379 net->sctp.udp_port = 0;
1380
1381 /* Set remote encap port to 0 by default */
1382 net->sctp.encap_port = 0;
1383
1384 /* Set SCOPE policy to enabled */
1385 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1386
1387 /* Set the default rwnd update threshold */
1388 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1389
1390 /* Initialize maximum autoclose timeout. */
1391 net->sctp.max_autoclose = INT_MAX / HZ;
1392
1393 #ifdef CONFIG_NET_L3_MASTER_DEV
1394 net->sctp.l3mdev_accept = 1;
1395 #endif
1396
1397 status = sctp_sysctl_net_register(net);
1398 if (status)
1399 goto err_sysctl_register;
1400
1401 /* Allocate and initialise sctp mibs. */
1402 status = init_sctp_mibs(net);
1403 if (status)
1404 goto err_init_mibs;
1405
1406 #ifdef CONFIG_PROC_FS
1407 /* Initialize proc fs directory. */
1408 status = sctp_proc_init(net);
1409 if (status)
1410 goto err_init_proc;
1411 #endif
1412
1413 sctp_dbg_objcnt_init(net);
1414
1415 /* Initialize the local address list. */
1416 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1417 spin_lock_init(&net->sctp.local_addr_lock);
1418 sctp_get_local_addr_list(net);
1419
1420 /* Initialize the address event list */
1421 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1422 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1423 spin_lock_init(&net->sctp.addr_wq_lock);
1424 net->sctp.addr_wq_timer.expires = 0;
1425 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1426
1427 return 0;
1428
1429 #ifdef CONFIG_PROC_FS
1430 err_init_proc:
1431 cleanup_sctp_mibs(net);
1432 #endif
1433 err_init_mibs:
1434 sctp_sysctl_net_unregister(net);
1435 err_sysctl_register:
1436 return status;
1437 }
1438
1439 static void __net_exit sctp_defaults_exit(struct net *net)
1440 {
1441 /* Free the local address list */
1442 sctp_free_addr_wq(net);
1443 sctp_free_local_addr_list(net);
1444
1445 #ifdef CONFIG_PROC_FS
1446 remove_proc_subtree("sctp", net->proc_net);
1447 net->sctp.proc_net_sctp = NULL;
1448 #endif
1449 cleanup_sctp_mibs(net);
1450 sctp_sysctl_net_unregister(net);
1451 }
1452
1453 static struct pernet_operations sctp_defaults_ops = {
1454 .init = sctp_defaults_init,
1455 .exit = sctp_defaults_exit,
1456 };
1457
1458 static int __net_init sctp_ctrlsock_init(struct net *net)
1459 {
1460 int status;
1461
1462 /* Initialize the control inode/socket for handling OOTB packets. */
1463 status = sctp_ctl_sock_init(net);
1464 if (status)
1465 pr_err("Failed to initialize the SCTP control sock\n");
1466
1467 return status;
1468 }
1469
1470 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1471 {
1472 /* Free the control endpoint. */
1473 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1474 }
1475
1476 static struct pernet_operations sctp_ctrlsock_ops = {
1477 .init = sctp_ctrlsock_init,
1478 .exit = sctp_ctrlsock_exit,
1479 };
1480
1481 /* Initialize the universe into something sensible. */
1482 static __init int sctp_init(void)
1483 {
1484 unsigned long nr_pages = totalram_pages();
1485 unsigned long limit;
1486 unsigned long goal;
1487 int max_entry_order;
1488 int num_entries;
1489 int max_share;
1490 int status;
1491 int order;
1492 int i;
1493
1494 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1495
1496 /* Allocate bind_bucket and chunk caches. */
1497 status = -ENOBUFS;
1498 sctp_bucket_cachep = KMEM_CACHE(sctp_bind_bucket, SLAB_HWCACHE_ALIGN);
1499 if (!sctp_bucket_cachep)
1500 goto out;
1501
1502 sctp_chunk_cachep = KMEM_CACHE(sctp_chunk, SLAB_HWCACHE_ALIGN);
1503 if (!sctp_chunk_cachep)
1504 goto err_chunk_cachep;
1505
1506 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1507 if (status)
1508 goto err_percpu_counter_init;
1509
1510 /* Implementation specific variables. */
1511
1512 /* Initialize default stream count setup information. */
1513 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1514 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1515
1516 /* Initialize handle used for association ids. */
1517 idr_init(&sctp_assocs_id);
1518
1519 limit = nr_free_buffer_pages() / 8;
1520 limit = max(limit, 128UL);
1521 sysctl_sctp_mem[0] = limit / 4 * 3;
1522 sysctl_sctp_mem[1] = limit;
1523 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1524
1525 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1526 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1527 max_share = min(4UL*1024*1024, limit);
1528
1529 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
1530 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1531 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1532
1533 sysctl_sctp_wmem[0] = PAGE_SIZE;
1534 sysctl_sctp_wmem[1] = 16*1024;
1535 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1536
1537 /* Size and allocate the association hash table.
1538 * The methodology is similar to that of the tcp hash tables.
1539 * Though not identical. Start by getting a goal size
1540 */
1541 if (nr_pages >= (128 * 1024))
1542 goal = nr_pages >> (22 - PAGE_SHIFT);
1543 else
1544 goal = nr_pages >> (24 - PAGE_SHIFT);
1545
1546 /* Then compute the page order for said goal */
1547 order = get_order(goal);
1548
1549 /* Now compute the required page order for the maximum sized table we
1550 * want to create
1551 */
1552 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1553 sizeof(struct sctp_bind_hashbucket));
1554
1555 /* Limit the page order by that maximum hash table size */
1556 order = min(order, max_entry_order);
1557
1558 /* Allocate and initialize the endpoint hash table. */
1559 sctp_ep_hashsize = 64;
1560 sctp_ep_hashtable =
1561 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1562 if (!sctp_ep_hashtable) {
1563 pr_err("Failed endpoint_hash alloc\n");
1564 status = -ENOMEM;
1565 goto err_ehash_alloc;
1566 }
1567 for (i = 0; i < sctp_ep_hashsize; i++) {
1568 rwlock_init(&sctp_ep_hashtable[i].lock);
1569 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1570 }
1571
1572 /* Allocate and initialize the SCTP port hash table.
1573 * Note that order is initalized to start at the max sized
1574 * table we want to support. If we can't get that many pages
1575 * reduce the order and try again
1576 */
1577 do {
1578 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1579 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1580 } while (!sctp_port_hashtable && --order > 0);
1581
1582 if (!sctp_port_hashtable) {
1583 pr_err("Failed bind hash alloc\n");
1584 status = -ENOMEM;
1585 goto err_bhash_alloc;
1586 }
1587
1588 /* Now compute the number of entries that will fit in the
1589 * port hash space we allocated
1590 */
1591 num_entries = (1UL << order) * PAGE_SIZE /
1592 sizeof(struct sctp_bind_hashbucket);
1593
1594 /* And finish by rounding it down to the nearest power of two.
1595 * This wastes some memory of course, but it's needed because
1596 * the hash function operates based on the assumption that
1597 * the number of entries is a power of two.
1598 */
1599 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1600
1601 for (i = 0; i < sctp_port_hashsize; i++) {
1602 spin_lock_init(&sctp_port_hashtable[i].lock);
1603 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1604 }
1605
1606 status = sctp_transport_hashtable_init();
1607 if (status)
1608 goto err_thash_alloc;
1609
1610 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1611 num_entries);
1612
1613 sctp_sysctl_register();
1614
1615 INIT_LIST_HEAD(&sctp_address_families);
1616 sctp_v4_pf_init();
1617 sctp_v6_pf_init();
1618 sctp_sched_ops_init();
1619
1620 status = register_pernet_subsys(&sctp_defaults_ops);
1621 if (status)
1622 goto err_register_defaults;
1623
1624 status = sctp_v4_protosw_init();
1625 if (status)
1626 goto err_protosw_init;
1627
1628 status = sctp_v6_protosw_init();
1629 if (status)
1630 goto err_v6_protosw_init;
1631
1632 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1633 if (status)
1634 goto err_register_ctrlsock;
1635
1636 status = sctp_v4_add_protocol();
1637 if (status)
1638 goto err_add_protocol;
1639
1640 /* Register SCTP with inet6 layer. */
1641 status = sctp_v6_add_protocol();
1642 if (status)
1643 goto err_v6_add_protocol;
1644
1645 if (sctp_offload_init() < 0)
1646 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1647
1648 out:
1649 return status;
1650 err_v6_add_protocol:
1651 sctp_v4_del_protocol();
1652 err_add_protocol:
1653 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1654 err_register_ctrlsock:
1655 sctp_v6_protosw_exit();
1656 err_v6_protosw_init:
1657 sctp_v4_protosw_exit();
1658 err_protosw_init:
1659 unregister_pernet_subsys(&sctp_defaults_ops);
1660 err_register_defaults:
1661 sctp_v4_pf_exit();
1662 sctp_v6_pf_exit();
1663 sctp_sysctl_unregister();
1664 free_pages((unsigned long)sctp_port_hashtable,
1665 get_order(sctp_port_hashsize *
1666 sizeof(struct sctp_bind_hashbucket)));
1667 err_bhash_alloc:
1668 sctp_transport_hashtable_destroy();
1669 err_thash_alloc:
1670 kfree(sctp_ep_hashtable);
1671 err_ehash_alloc:
1672 percpu_counter_destroy(&sctp_sockets_allocated);
1673 err_percpu_counter_init:
1674 kmem_cache_destroy(sctp_chunk_cachep);
1675 err_chunk_cachep:
1676 kmem_cache_destroy(sctp_bucket_cachep);
1677 goto out;
1678 }
1679
1680 /* Exit handler for the SCTP protocol. */
1681 static __exit void sctp_exit(void)
1682 {
1683 /* BUG. This should probably do something useful like clean
1684 * up all the remaining associations and all that memory.
1685 */
1686
1687 /* Unregister with inet6/inet layers. */
1688 sctp_v6_del_protocol();
1689 sctp_v4_del_protocol();
1690
1691 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1692
1693 /* Free protosw registrations */
1694 sctp_v6_protosw_exit();
1695 sctp_v4_protosw_exit();
1696
1697 unregister_pernet_subsys(&sctp_defaults_ops);
1698
1699 /* Unregister with socket layer. */
1700 sctp_v6_pf_exit();
1701 sctp_v4_pf_exit();
1702
1703 sctp_sysctl_unregister();
1704
1705 free_pages((unsigned long)sctp_port_hashtable,
1706 get_order(sctp_port_hashsize *
1707 sizeof(struct sctp_bind_hashbucket)));
1708 kfree(sctp_ep_hashtable);
1709 sctp_transport_hashtable_destroy();
1710
1711 percpu_counter_destroy(&sctp_sockets_allocated);
1712
1713 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1714
1715 kmem_cache_destroy(sctp_chunk_cachep);
1716 kmem_cache_destroy(sctp_bucket_cachep);
1717 }
1718
1719 module_init(sctp_init);
1720 module_exit(sctp_exit);
1721
1722 /*
1723 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1724 */
1725 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1726 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1727 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1728 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1729 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1730 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1731 MODULE_LICENSE("GPL");
1732
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.