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, write to
27  * the Free Software Foundation, 59 Temple Place - Suite 330,
28  * Boston, MA 02111-1307, USA.
29  *
30  * Please send any bug reports or fixes you make to the
31  * email address(es):
32  *    lksctp developers <lksctp-developers@lists.sourceforge.net>
33  *
34  * Or submit a bug report through the following website:
35  *    http://www.sf.net/projects/lksctp
36  *
37  * Written or modified by:
38  *    La Monte H.P. Yarroll <piggy@acm.org>
39  *    Karl Knutson <karl@athena.chicago.il.us>
40  *    Jon Grimm <jgrimm@us.ibm.com>
41  *    Sridhar Samudrala <sri@us.ibm.com>
42  *    Daisy Chang <daisyc@us.ibm.com>
43  *    Ardelle Fan <ardelle.fan@intel.com>
44  *
45  * Any bugs reported given to us we will try to fix... any fixes shared will
46  * be incorporated into the next SCTP release.
47  */
48 
49 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
50 
51 #include <linux/module.h>
52 #include <linux/init.h>
53 #include <linux/netdevice.h>
54 #include <linux/inetdevice.h>
55 #include <linux/seq_file.h>
56 #include <linux/bootmem.h>
57 #include <linux/highmem.h>
58 #include <linux/swap.h>
59 #include <linux/slab.h>
60 #include <net/net_namespace.h>
61 #include <net/protocol.h>
62 #include <net/ip.h>
63 #include <net/ipv6.h>
64 #include <net/route.h>
65 #include <net/sctp/sctp.h>
66 #include <net/addrconf.h>
67 #include <net/inet_common.h>
68 #include <net/inet_ecn.h>
69 
70 /* Global data structures. */
71 struct sctp_globals sctp_globals __read_mostly;
72 DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
73 
74 #ifdef CONFIG_PROC_FS
75 struct proc_dir_entry	*proc_net_sctp;
76 #endif
77 
78 struct idr sctp_assocs_id;
79 DEFINE_SPINLOCK(sctp_assocs_id_lock);
80 
81 /* This is the global socket data structure used for responding to
82  * the Out-of-the-blue (OOTB) packets.  A control sock will be created
83  * for this socket at the initialization time.
84  */
85 static struct sock *sctp_ctl_sock;
86 
87 static struct sctp_pf *sctp_pf_inet6_specific;
88 static struct sctp_pf *sctp_pf_inet_specific;
89 static struct sctp_af *sctp_af_v4_specific;
90 static struct sctp_af *sctp_af_v6_specific;
91 
92 struct kmem_cache *sctp_chunk_cachep __read_mostly;
93 struct kmem_cache *sctp_bucket_cachep __read_mostly;
94 
95 long sysctl_sctp_mem[3];
96 int sysctl_sctp_rmem[3];
97 int sysctl_sctp_wmem[3];
98 
99 /* Return the address of the control sock. */
sctp_get_ctl_sock(void)100 struct sock *sctp_get_ctl_sock(void)
101 {
102 	return sctp_ctl_sock;
103 }
104 
105 /* Set up the proc fs entry for the SCTP protocol. */
sctp_proc_init(void)106 static __init int sctp_proc_init(void)
107 {
108 	if (percpu_counter_init(&sctp_sockets_allocated, 0))
109 		goto out_nomem;
110 #ifdef CONFIG_PROC_FS
111 	if (!proc_net_sctp) {
112 		proc_net_sctp = proc_mkdir("sctp", init_net.proc_net);
113 		if (!proc_net_sctp)
114 			goto out_free_percpu;
115 	}
116 
117 	if (sctp_snmp_proc_init())
118 		goto out_snmp_proc_init;
119 	if (sctp_eps_proc_init())
120 		goto out_eps_proc_init;
121 	if (sctp_assocs_proc_init())
122 		goto out_assocs_proc_init;
123 	if (sctp_remaddr_proc_init())
124 		goto out_remaddr_proc_init;
125 
126 	return 0;
127 
128 out_remaddr_proc_init:
129 	sctp_assocs_proc_exit();
130 out_assocs_proc_init:
131 	sctp_eps_proc_exit();
132 out_eps_proc_init:
133 	sctp_snmp_proc_exit();
134 out_snmp_proc_init:
135 	if (proc_net_sctp) {
136 		proc_net_sctp = NULL;
137 		remove_proc_entry("sctp", init_net.proc_net);
138 	}
139 out_free_percpu:
140 	percpu_counter_destroy(&sctp_sockets_allocated);
141 #else
142 	return 0;
143 #endif /* CONFIG_PROC_FS */
144 
145 out_nomem:
146 	return -ENOMEM;
147 }
148 
149 /* Clean up the proc fs entry for the SCTP protocol.
150  * Note: Do not make this __exit as it is used in the init error
151  * path.
152  */
sctp_proc_exit(void)153 static void sctp_proc_exit(void)
154 {
155 #ifdef CONFIG_PROC_FS
156 	sctp_snmp_proc_exit();
157 	sctp_eps_proc_exit();
158 	sctp_assocs_proc_exit();
159 	sctp_remaddr_proc_exit();
160 
161 	if (proc_net_sctp) {
162 		proc_net_sctp = NULL;
163 		remove_proc_entry("sctp", init_net.proc_net);
164 	}
165 #endif
166 	percpu_counter_destroy(&sctp_sockets_allocated);
167 }
168 
169 /* Private helper to extract ipv4 address and stash them in
170  * the protocol structure.
171  */
sctp_v4_copy_addrlist(struct list_head * addrlist,struct net_device * dev)172 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
173 				  struct net_device *dev)
174 {
175 	struct in_device *in_dev;
176 	struct in_ifaddr *ifa;
177 	struct sctp_sockaddr_entry *addr;
178 
179 	rcu_read_lock();
180 	if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
181 		rcu_read_unlock();
182 		return;
183 	}
184 
185 	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
186 		/* Add the address to the local list.  */
187 		addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
188 		if (addr) {
189 			addr->a.v4.sin_family = AF_INET;
190 			addr->a.v4.sin_port = 0;
191 			addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
192 			addr->valid = 1;
193 			INIT_LIST_HEAD(&addr->list);
194 			list_add_tail(&addr->list, addrlist);
195 		}
196 	}
197 
198 	rcu_read_unlock();
199 }
200 
201 /* Extract our IP addresses from the system and stash them in the
202  * protocol structure.
203  */
sctp_get_local_addr_list(void)204 static void sctp_get_local_addr_list(void)
205 {
206 	struct net_device *dev;
207 	struct list_head *pos;
208 	struct sctp_af *af;
209 
210 	rcu_read_lock();
211 	for_each_netdev_rcu(&init_net, dev) {
212 		__list_for_each(pos, &sctp_address_families) {
213 			af = list_entry(pos, struct sctp_af, list);
214 			af->copy_addrlist(&sctp_local_addr_list, dev);
215 		}
216 	}
217 	rcu_read_unlock();
218 }
219 
220 /* Free the existing local addresses.  */
sctp_free_local_addr_list(void)221 static void sctp_free_local_addr_list(void)
222 {
223 	struct sctp_sockaddr_entry *addr;
224 	struct list_head *pos, *temp;
225 
226 	list_for_each_safe(pos, temp, &sctp_local_addr_list) {
227 		addr = list_entry(pos, struct sctp_sockaddr_entry, list);
228 		list_del(pos);
229 		kfree(addr);
230 	}
231 }
232 
233 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
sctp_copy_local_addr_list(struct sctp_bind_addr * bp,sctp_scope_t scope,gfp_t gfp,int copy_flags)234 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
235 			      gfp_t gfp, int copy_flags)
236 {
237 	struct sctp_sockaddr_entry *addr;
238 	int error = 0;
239 
240 	rcu_read_lock();
241 	list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
242 		if (!addr->valid)
243 			continue;
244 		if (sctp_in_scope(&addr->a, scope)) {
245 			/* Now that the address is in scope, check to see if
246 			 * the address type is really supported by the local
247 			 * sock as well as the remote peer.
248 			 */
249 			if ((((AF_INET == addr->a.sa.sa_family) &&
250 			      (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
251 			    (((AF_INET6 == addr->a.sa.sa_family) &&
252 			      (copy_flags & SCTP_ADDR6_ALLOWED) &&
253 			      (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
254 				error = sctp_add_bind_addr(bp, &addr->a,
255 						    SCTP_ADDR_SRC, GFP_ATOMIC);
256 				if (error)
257 					goto end_copy;
258 			}
259 		}
260 	}
261 
262 end_copy:
263 	rcu_read_unlock();
264 	return error;
265 }
266 
267 /* Initialize a sctp_addr from in incoming skb.  */
sctp_v4_from_skb(union sctp_addr * addr,struct sk_buff * skb,int is_saddr)268 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
269 			     int is_saddr)
270 {
271 	void *from;
272 	__be16 *port;
273 	struct sctphdr *sh;
274 
275 	port = &addr->v4.sin_port;
276 	addr->v4.sin_family = AF_INET;
277 
278 	sh = sctp_hdr(skb);
279 	if (is_saddr) {
280 		*port  = sh->source;
281 		from = &ip_hdr(skb)->saddr;
282 	} else {
283 		*port = sh->dest;
284 		from = &ip_hdr(skb)->daddr;
285 	}
286 	memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
287 }
288 
289 /* Initialize an sctp_addr from a socket. */
sctp_v4_from_sk(union sctp_addr * addr,struct sock * sk)290 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
291 {
292 	addr->v4.sin_family = AF_INET;
293 	addr->v4.sin_port = 0;
294 	addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
295 }
296 
297 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
sctp_v4_to_sk_saddr(union sctp_addr * addr,struct sock * sk)298 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
299 {
300 	inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
301 }
302 
303 /* Initialize sk->sk_daddr from sctp_addr. */
sctp_v4_to_sk_daddr(union sctp_addr * addr,struct sock * sk)304 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
305 {
306 	inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
307 }
308 
309 /* 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)310 static void sctp_v4_from_addr_param(union sctp_addr *addr,
311 				    union sctp_addr_param *param,
312 				    __be16 port, int iif)
313 {
314 	addr->v4.sin_family = AF_INET;
315 	addr->v4.sin_port = port;
316 	addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
317 }
318 
319 /* Initialize an address parameter from a sctp_addr and return the length
320  * of the address parameter.
321  */
sctp_v4_to_addr_param(const union sctp_addr * addr,union sctp_addr_param * param)322 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
323 				 union sctp_addr_param *param)
324 {
325 	int length = sizeof(sctp_ipv4addr_param_t);
326 
327 	param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
328 	param->v4.param_hdr.length = htons(length);
329 	param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
330 
331 	return length;
332 }
333 
334 /* Initialize a sctp_addr from a dst_entry. */
sctp_v4_dst_saddr(union sctp_addr * saddr,struct flowi4 * fl4,__be16 port)335 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
336 			      __be16 port)
337 {
338 	saddr->v4.sin_family = AF_INET;
339 	saddr->v4.sin_port = port;
340 	saddr->v4.sin_addr.s_addr = fl4->saddr;
341 }
342 
343 /* Compare two addresses exactly. */
sctp_v4_cmp_addr(const union sctp_addr * addr1,const union sctp_addr * addr2)344 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
345 			    const union sctp_addr *addr2)
346 {
347 	if (addr1->sa.sa_family != addr2->sa.sa_family)
348 		return 0;
349 	if (addr1->v4.sin_port != addr2->v4.sin_port)
350 		return 0;
351 	if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
352 		return 0;
353 
354 	return 1;
355 }
356 
357 /* Initialize addr struct to INADDR_ANY. */
sctp_v4_inaddr_any(union sctp_addr * addr,__be16 port)358 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
359 {
360 	addr->v4.sin_family = AF_INET;
361 	addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
362 	addr->v4.sin_port = port;
363 }
364 
365 /* Is this a wildcard address? */
sctp_v4_is_any(const union sctp_addr * addr)366 static int sctp_v4_is_any(const union sctp_addr *addr)
367 {
368 	return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
369 }
370 
371 /* This function checks if the address is a valid address to be used for
372  * SCTP binding.
373  *
374  * Output:
375  * Return 0 - If the address is a non-unicast or an illegal address.
376  * Return 1 - If the address is a unicast.
377  */
sctp_v4_addr_valid(union sctp_addr * addr,struct sctp_sock * sp,const struct sk_buff * skb)378 static int sctp_v4_addr_valid(union sctp_addr *addr,
379 			      struct sctp_sock *sp,
380 			      const struct sk_buff *skb)
381 {
382 	/* IPv4 addresses not allowed */
383 	if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
384 		return 0;
385 
386 	/* Is this a non-unicast address or a unusable SCTP address? */
387 	if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
388 		return 0;
389 
390 	/* Is this a broadcast address? */
391 	if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
392 		return 0;
393 
394 	return 1;
395 }
396 
397 /* Should this be available for binding?   */
sctp_v4_available(union sctp_addr * addr,struct sctp_sock * sp)398 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
399 {
400 	int ret = inet_addr_type(&init_net, addr->v4.sin_addr.s_addr);
401 
402 
403 	if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
404 	   ret != RTN_LOCAL &&
405 	   !sp->inet.freebind &&
406 	   !sysctl_ip_nonlocal_bind)
407 		return 0;
408 
409 	if (ipv6_only_sock(sctp_opt2sk(sp)))
410 		return 0;
411 
412 	return 1;
413 }
414 
415 /* Checking the loopback, private and other address scopes as defined in
416  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
417  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
418  *
419  * Level 0 - unusable SCTP addresses
420  * Level 1 - loopback address
421  * Level 2 - link-local addresses
422  * Level 3 - private addresses.
423  * Level 4 - global addresses
424  * For INIT and INIT-ACK address list, let L be the level of
425  * of requested destination address, sender and receiver
426  * SHOULD include all of its addresses with level greater
427  * than or equal to L.
428  *
429  * IPv4 scoping can be controlled through sysctl option
430  * net.sctp.addr_scope_policy
431  */
sctp_v4_scope(union sctp_addr * addr)432 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
433 {
434 	sctp_scope_t retval;
435 
436 	/* Check for unusable SCTP addresses. */
437 	if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
438 		retval =  SCTP_SCOPE_UNUSABLE;
439 	} else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
440 		retval = SCTP_SCOPE_LOOPBACK;
441 	} else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
442 		retval = SCTP_SCOPE_LINK;
443 	} else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
444 		   ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
445 		   ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
446 		retval = SCTP_SCOPE_PRIVATE;
447 	} else {
448 		retval = SCTP_SCOPE_GLOBAL;
449 	}
450 
451 	return retval;
452 }
453 
454 /* Returns a valid dst cache entry for the given source and destination ip
455  * addresses. If an association is passed, trys to get a dst entry with a
456  * source address that matches an address in the bind address list.
457  */
sctp_v4_get_dst(struct sctp_transport * t,union sctp_addr * saddr,struct flowi * fl,struct sock * sk)458 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
459 				struct flowi *fl, struct sock *sk)
460 {
461 	struct sctp_association *asoc = t->asoc;
462 	struct rtable *rt;
463 	struct flowi4 *fl4 = &fl->u.ip4;
464 	struct sctp_bind_addr *bp;
465 	struct sctp_sockaddr_entry *laddr;
466 	struct dst_entry *dst = NULL;
467 	union sctp_addr *daddr = &t->ipaddr;
468 	union sctp_addr dst_saddr;
469 
470 	memset(fl4, 0x0, sizeof(struct flowi4));
471 	fl4->daddr  = daddr->v4.sin_addr.s_addr;
472 	fl4->fl4_dport = daddr->v4.sin_port;
473 	fl4->flowi4_proto = IPPROTO_SCTP;
474 	if (asoc) {
475 		fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
476 		fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
477 		fl4->fl4_sport = htons(asoc->base.bind_addr.port);
478 	}
479 	if (saddr) {
480 		fl4->saddr = saddr->v4.sin_addr.s_addr;
481 		fl4->fl4_sport = saddr->v4.sin_port;
482 	}
483 
484 	SCTP_DEBUG_PRINTK("%s: DST:%pI4, SRC:%pI4 - ",
485 			  __func__, &fl4->daddr, &fl4->saddr);
486 
487 	rt = ip_route_output_key(&init_net, fl4);
488 	if (!IS_ERR(rt))
489 		dst = &rt->dst;
490 
491 	/* If there is no association or if a source address is passed, no
492 	 * more validation is required.
493 	 */
494 	if (!asoc || saddr)
495 		goto out;
496 
497 	bp = &asoc->base.bind_addr;
498 
499 	if (dst) {
500 		/* Walk through the bind address list and look for a bind
501 		 * address that matches the source address of the returned dst.
502 		 */
503 		sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
504 		rcu_read_lock();
505 		list_for_each_entry_rcu(laddr, &bp->address_list, list) {
506 			if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
507 			    (laddr->state != SCTP_ADDR_SRC &&
508 			    !asoc->src_out_of_asoc_ok))
509 				continue;
510 			if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
511 				goto out_unlock;
512 		}
513 		rcu_read_unlock();
514 
515 		/* None of the bound addresses match the source address of the
516 		 * dst. So release it.
517 		 */
518 		dst_release(dst);
519 		dst = NULL;
520 	}
521 
522 	/* Walk through the bind address list and try to get a dst that
523 	 * matches a bind address as the source address.
524 	 */
525 	rcu_read_lock();
526 	list_for_each_entry_rcu(laddr, &bp->address_list, list) {
527 		if (!laddr->valid)
528 			continue;
529 		if ((laddr->state == SCTP_ADDR_SRC) &&
530 		    (AF_INET == laddr->a.sa.sa_family)) {
531 			fl4->saddr = laddr->a.v4.sin_addr.s_addr;
532 			fl4->fl4_sport = laddr->a.v4.sin_port;
533 			rt = ip_route_output_key(&init_net, fl4);
534 			if (!IS_ERR(rt)) {
535 				dst = &rt->dst;
536 				goto out_unlock;
537 			}
538 		}
539 	}
540 
541 out_unlock:
542 	rcu_read_unlock();
543 out:
544 	t->dst = dst;
545 	if (dst)
546 		SCTP_DEBUG_PRINTK("rt_dst:%pI4, rt_src:%pI4\n",
547 				  &fl4->daddr, &fl4->saddr);
548 	else
549 		SCTP_DEBUG_PRINTK("NO ROUTE\n");
550 }
551 
552 /* For v4, the source address is cached in the route entry(dst). So no need
553  * to cache it separately and hence this is an empty routine.
554  */
sctp_v4_get_saddr(struct sctp_sock * sk,struct sctp_transport * t,struct flowi * fl)555 static void sctp_v4_get_saddr(struct sctp_sock *sk,
556 			      struct sctp_transport *t,
557 			      struct flowi *fl)
558 {
559 	union sctp_addr *saddr = &t->saddr;
560 	struct rtable *rt = (struct rtable *)t->dst;
561 
562 	if (rt) {
563 		saddr->v4.sin_family = AF_INET;
564 		saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
565 	}
566 }
567 
568 /* What interface did this skb arrive on? */
sctp_v4_skb_iif(const struct sk_buff * skb)569 static int sctp_v4_skb_iif(const struct sk_buff *skb)
570 {
571 	return skb_rtable(skb)->rt_iif;
572 }
573 
574 /* Was this packet marked by Explicit Congestion Notification? */
sctp_v4_is_ce(const struct sk_buff * skb)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(). */
sctp_v4_create_accept_sk(struct sock * sk,struct sctp_association * asoc)581 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
582 					     struct sctp_association *asoc)
583 {
584 	struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
585 			sk->sk_prot);
586 	struct inet_sock *newinet;
587 
588 	if (!newsk)
589 		goto out;
590 
591 	sock_init_data(NULL, newsk);
592 
593 	sctp_copy_sock(newsk, sk, asoc);
594 	sock_reset_flag(newsk, SOCK_ZAPPED);
595 
596 	newinet = inet_sk(newsk);
597 
598 	newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
599 
600 	sk_refcnt_debug_inc(newsk);
601 
602 	if (newsk->sk_prot->init(newsk)) {
603 		sk_common_release(newsk);
604 		newsk = NULL;
605 	}
606 
607 out:
608 	return newsk;
609 }
610 
611 /* Map address, empty for v4 family */
sctp_v4_addr_v4map(struct sctp_sock * sp,union sctp_addr * addr)612 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
613 {
614 	/* Empty */
615 }
616 
617 /* Dump the v4 addr to the seq file. */
sctp_v4_seq_dump_addr(struct seq_file * seq,union sctp_addr * addr)618 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
619 {
620 	seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
621 }
622 
sctp_v4_ecn_capable(struct sock * sk)623 static void sctp_v4_ecn_capable(struct sock *sk)
624 {
625 	INET_ECN_xmit(sk);
626 }
627 
sctp_addr_wq_timeout_handler(unsigned long arg)628 void sctp_addr_wq_timeout_handler(unsigned long arg)
629 {
630 	struct sctp_sockaddr_entry *addrw, *temp;
631 	struct sctp_sock *sp;
632 
633 	spin_lock_bh(&sctp_addr_wq_lock);
634 
635 	list_for_each_entry_safe(addrw, temp, &sctp_addr_waitq, list) {
636 		SCTP_DEBUG_PRINTK_IPADDR("sctp_addrwq_timo_handler: the first ent in wq %p is ",
637 		    " for cmd %d at entry %p\n", &sctp_addr_waitq, &addrw->a, addrw->state,
638 		    addrw);
639 
640 #if IS_ENABLED(CONFIG_IPV6)
641 		/* Now we send an ASCONF for each association */
642 		/* Note. we currently don't handle link local IPv6 addressees */
643 		if (addrw->a.sa.sa_family == AF_INET6) {
644 			struct in6_addr *in6;
645 
646 			if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
647 			    IPV6_ADDR_LINKLOCAL)
648 				goto free_next;
649 
650 			in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
651 			if (ipv6_chk_addr(&init_net, in6, NULL, 0) == 0 &&
652 			    addrw->state == SCTP_ADDR_NEW) {
653 				unsigned long timeo_val;
654 
655 				SCTP_DEBUG_PRINTK("sctp_timo_handler: this is on DAD, trying %d sec later\n",
656 				    SCTP_ADDRESS_TICK_DELAY);
657 				timeo_val = jiffies;
658 				timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
659 				mod_timer(&sctp_addr_wq_timer, timeo_val);
660 				break;
661 			}
662 		}
663 #endif
664 		list_for_each_entry(sp, &sctp_auto_asconf_splist, auto_asconf_list) {
665 			struct sock *sk;
666 
667 			sk = sctp_opt2sk(sp);
668 			/* ignore bound-specific endpoints */
669 			if (!sctp_is_ep_boundall(sk))
670 				continue;
671 			sctp_bh_lock_sock(sk);
672 			if (sctp_asconf_mgmt(sp, addrw) < 0)
673 				SCTP_DEBUG_PRINTK("sctp_addrwq_timo_handler: sctp_asconf_mgmt failed\n");
674 			sctp_bh_unlock_sock(sk);
675 		}
676 free_next:
677 		list_del(&addrw->list);
678 		kfree(addrw);
679 	}
680 	spin_unlock_bh(&sctp_addr_wq_lock);
681 }
682 
sctp_free_addr_wq(void)683 static void sctp_free_addr_wq(void)
684 {
685 	struct sctp_sockaddr_entry *addrw;
686 	struct sctp_sockaddr_entry *temp;
687 
688 	spin_lock_bh(&sctp_addr_wq_lock);
689 	del_timer(&sctp_addr_wq_timer);
690 	list_for_each_entry_safe(addrw, temp, &sctp_addr_waitq, list) {
691 		list_del(&addrw->list);
692 		kfree(addrw);
693 	}
694 	spin_unlock_bh(&sctp_addr_wq_lock);
695 }
696 
697 /* lookup the entry for the same address in the addr_waitq
698  * sctp_addr_wq MUST be locked
699  */
sctp_addr_wq_lookup(struct sctp_sockaddr_entry * addr)700 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct sctp_sockaddr_entry *addr)
701 {
702 	struct sctp_sockaddr_entry *addrw;
703 
704 	list_for_each_entry(addrw, &sctp_addr_waitq, list) {
705 		if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
706 			continue;
707 		if (addrw->a.sa.sa_family == AF_INET) {
708 			if (addrw->a.v4.sin_addr.s_addr ==
709 			    addr->a.v4.sin_addr.s_addr)
710 				return addrw;
711 		} else if (addrw->a.sa.sa_family == AF_INET6) {
712 			if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
713 			    &addr->a.v6.sin6_addr))
714 				return addrw;
715 		}
716 	}
717 	return NULL;
718 }
719 
sctp_addr_wq_mgmt(struct sctp_sockaddr_entry * addr,int cmd)720 void sctp_addr_wq_mgmt(struct sctp_sockaddr_entry *addr, int cmd)
721 {
722 	struct sctp_sockaddr_entry *addrw;
723 	unsigned long timeo_val;
724 
725 	/* first, we check if an opposite message already exist in the queue.
726 	 * If we found such message, it is removed.
727 	 * This operation is a bit stupid, but the DHCP client attaches the
728 	 * new address after a couple of addition and deletion of that address
729 	 */
730 
731 	spin_lock_bh(&sctp_addr_wq_lock);
732 	/* Offsets existing events in addr_wq */
733 	addrw = sctp_addr_wq_lookup(addr);
734 	if (addrw) {
735 		if (addrw->state != cmd) {
736 			SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt offsets existing entry for %d ",
737 			    " in wq %p\n", addrw->state, &addrw->a,
738 			    &sctp_addr_waitq);
739 			list_del(&addrw->list);
740 			kfree(addrw);
741 		}
742 		spin_unlock_bh(&sctp_addr_wq_lock);
743 		return;
744 	}
745 
746 	/* OK, we have to add the new address to the wait queue */
747 	addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
748 	if (addrw == NULL) {
749 		spin_unlock_bh(&sctp_addr_wq_lock);
750 		return;
751 	}
752 	addrw->state = cmd;
753 	list_add_tail(&addrw->list, &sctp_addr_waitq);
754 	SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt add new entry for cmd:%d ",
755 	    " in wq %p\n", addrw->state, &addrw->a, &sctp_addr_waitq);
756 
757 	if (!timer_pending(&sctp_addr_wq_timer)) {
758 		timeo_val = jiffies;
759 		timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
760 		mod_timer(&sctp_addr_wq_timer, timeo_val);
761 	}
762 	spin_unlock_bh(&sctp_addr_wq_lock);
763 }
764 
765 /* Event handler for inet address addition/deletion events.
766  * The sctp_local_addr_list needs to be protocted by a spin lock since
767  * multiple notifiers (say IPv4 and IPv6) may be running at the same
768  * time and thus corrupt the list.
769  * The reader side is protected with RCU.
770  */
sctp_inetaddr_event(struct notifier_block * this,unsigned long ev,void * ptr)771 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
772 			       void *ptr)
773 {
774 	struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
775 	struct sctp_sockaddr_entry *addr = NULL;
776 	struct sctp_sockaddr_entry *temp;
777 	int found = 0;
778 
779 	if (!net_eq(dev_net(ifa->ifa_dev->dev), &init_net))
780 		return NOTIFY_DONE;
781 
782 	switch (ev) {
783 	case NETDEV_UP:
784 		addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
785 		if (addr) {
786 			addr->a.v4.sin_family = AF_INET;
787 			addr->a.v4.sin_port = 0;
788 			addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
789 			addr->valid = 1;
790 			spin_lock_bh(&sctp_local_addr_lock);
791 			list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
792 			sctp_addr_wq_mgmt(addr, SCTP_ADDR_NEW);
793 			spin_unlock_bh(&sctp_local_addr_lock);
794 		}
795 		break;
796 	case NETDEV_DOWN:
797 		spin_lock_bh(&sctp_local_addr_lock);
798 		list_for_each_entry_safe(addr, temp,
799 					&sctp_local_addr_list, list) {
800 			if (addr->a.sa.sa_family == AF_INET &&
801 					addr->a.v4.sin_addr.s_addr ==
802 					ifa->ifa_local) {
803 				sctp_addr_wq_mgmt(addr, SCTP_ADDR_DEL);
804 				found = 1;
805 				addr->valid = 0;
806 				list_del_rcu(&addr->list);
807 				break;
808 			}
809 		}
810 		spin_unlock_bh(&sctp_local_addr_lock);
811 		if (found)
812 			kfree_rcu(addr, rcu);
813 		break;
814 	}
815 
816 	return NOTIFY_DONE;
817 }
818 
819 /*
820  * Initialize the control inode/socket with a control endpoint data
821  * structure.  This endpoint is reserved exclusively for the OOTB processing.
822  */
sctp_ctl_sock_init(void)823 static int sctp_ctl_sock_init(void)
824 {
825 	int err;
826 	sa_family_t family = PF_INET;
827 
828 	if (sctp_get_pf_specific(PF_INET6))
829 		family = PF_INET6;
830 
831 	err = inet_ctl_sock_create(&sctp_ctl_sock, family,
832 				   SOCK_SEQPACKET, IPPROTO_SCTP, &init_net);
833 
834 	/* If IPv6 socket could not be created, try the IPv4 socket */
835 	if (err < 0 && family == PF_INET6)
836 		err = inet_ctl_sock_create(&sctp_ctl_sock, AF_INET,
837 					   SOCK_SEQPACKET, IPPROTO_SCTP,
838 					   &init_net);
839 
840 	if (err < 0) {
841 		pr_err("Failed to create the SCTP control socket\n");
842 		return err;
843 	}
844 	return 0;
845 }
846 
847 /* Register address family specific functions. */
sctp_register_af(struct sctp_af * af)848 int sctp_register_af(struct sctp_af *af)
849 {
850 	switch (af->sa_family) {
851 	case AF_INET:
852 		if (sctp_af_v4_specific)
853 			return 0;
854 		sctp_af_v4_specific = af;
855 		break;
856 	case AF_INET6:
857 		if (sctp_af_v6_specific)
858 			return 0;
859 		sctp_af_v6_specific = af;
860 		break;
861 	default:
862 		return 0;
863 	}
864 
865 	INIT_LIST_HEAD(&af->list);
866 	list_add_tail(&af->list, &sctp_address_families);
867 	return 1;
868 }
869 
870 /* Get the table of functions for manipulating a particular address
871  * family.
872  */
sctp_get_af_specific(sa_family_t family)873 struct sctp_af *sctp_get_af_specific(sa_family_t family)
874 {
875 	switch (family) {
876 	case AF_INET:
877 		return sctp_af_v4_specific;
878 	case AF_INET6:
879 		return sctp_af_v6_specific;
880 	default:
881 		return NULL;
882 	}
883 }
884 
885 /* Common code to initialize a AF_INET msg_name. */
sctp_inet_msgname(char * msgname,int * addr_len)886 static void sctp_inet_msgname(char *msgname, int *addr_len)
887 {
888 	struct sockaddr_in *sin;
889 
890 	sin = (struct sockaddr_in *)msgname;
891 	*addr_len = sizeof(struct sockaddr_in);
892 	sin->sin_family = AF_INET;
893 	memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
894 }
895 
896 /* 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)897 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
898 				    int *addr_len)
899 {
900 	struct sockaddr_in *sin, *sinfrom;
901 
902 	if (msgname) {
903 		struct sctp_association *asoc;
904 
905 		asoc = event->asoc;
906 		sctp_inet_msgname(msgname, addr_len);
907 		sin = (struct sockaddr_in *)msgname;
908 		sinfrom = &asoc->peer.primary_addr.v4;
909 		sin->sin_port = htons(asoc->peer.port);
910 		sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
911 	}
912 }
913 
914 /* Initialize and copy out a msgname from an inbound skb. */
sctp_inet_skb_msgname(struct sk_buff * skb,char * msgname,int * len)915 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
916 {
917 	if (msgname) {
918 		struct sctphdr *sh = sctp_hdr(skb);
919 		struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
920 
921 		sctp_inet_msgname(msgname, len);
922 		sin->sin_port = sh->source;
923 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
924 	}
925 }
926 
927 /* Do we support this AF? */
sctp_inet_af_supported(sa_family_t family,struct sctp_sock * sp)928 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
929 {
930 	/* PF_INET only supports AF_INET addresses. */
931 	return AF_INET == family;
932 }
933 
934 /* Address matching with wildcards allowed. */
sctp_inet_cmp_addr(const union sctp_addr * addr1,const union sctp_addr * addr2,struct sctp_sock * opt)935 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
936 			      const union sctp_addr *addr2,
937 			      struct sctp_sock *opt)
938 {
939 	/* PF_INET only supports AF_INET addresses. */
940 	if (addr1->sa.sa_family != addr2->sa.sa_family)
941 		return 0;
942 	if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
943 	    htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
944 		return 1;
945 	if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
946 		return 1;
947 
948 	return 0;
949 }
950 
951 /* Verify that provided sockaddr looks bindable.  Common verification has
952  * already been taken care of.
953  */
sctp_inet_bind_verify(struct sctp_sock * opt,union sctp_addr * addr)954 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
955 {
956 	return sctp_v4_available(addr, opt);
957 }
958 
959 /* Verify that sockaddr looks sendable.  Common verification has already
960  * been taken care of.
961  */
sctp_inet_send_verify(struct sctp_sock * opt,union sctp_addr * addr)962 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
963 {
964 	return 1;
965 }
966 
967 /* Fill in Supported Address Type information for INIT and INIT-ACK
968  * chunks.  Returns number of addresses supported.
969  */
sctp_inet_supported_addrs(const struct sctp_sock * opt,__be16 * types)970 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
971 				     __be16 *types)
972 {
973 	types[0] = SCTP_PARAM_IPV4_ADDRESS;
974 	return 1;
975 }
976 
977 /* Wrapper routine that calls the ip transmit routine. */
sctp_v4_xmit(struct sk_buff * skb,struct sctp_transport * transport)978 static inline int sctp_v4_xmit(struct sk_buff *skb,
979 			       struct sctp_transport *transport)
980 {
981 	struct inet_sock *inet = inet_sk(skb->sk);
982 
983 	SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n",
984 			  __func__, skb, skb->len,
985 			  &transport->fl.u.ip4.saddr,
986 			  &transport->fl.u.ip4.daddr);
987 
988 	inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
989 			 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
990 
991 	SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
992 	return ip_queue_xmit(skb, &transport->fl);
993 }
994 
995 static struct sctp_af sctp_af_inet;
996 
997 static struct sctp_pf sctp_pf_inet = {
998 	.event_msgname = sctp_inet_event_msgname,
999 	.skb_msgname   = sctp_inet_skb_msgname,
1000 	.af_supported  = sctp_inet_af_supported,
1001 	.cmp_addr      = sctp_inet_cmp_addr,
1002 	.bind_verify   = sctp_inet_bind_verify,
1003 	.send_verify   = sctp_inet_send_verify,
1004 	.supported_addrs = sctp_inet_supported_addrs,
1005 	.create_accept_sk = sctp_v4_create_accept_sk,
1006 	.addr_v4map	= sctp_v4_addr_v4map,
1007 	.af            = &sctp_af_inet
1008 };
1009 
1010 /* Notifier for inetaddr addition/deletion events.  */
1011 static struct notifier_block sctp_inetaddr_notifier = {
1012 	.notifier_call = sctp_inetaddr_event,
1013 };
1014 
1015 /* Socket operations.  */
1016 static const struct proto_ops inet_seqpacket_ops = {
1017 	.family		   = PF_INET,
1018 	.owner		   = THIS_MODULE,
1019 	.release	   = inet_release,	/* Needs to be wrapped... */
1020 	.bind		   = inet_bind,
1021 	.connect	   = inet_dgram_connect,
1022 	.socketpair	   = sock_no_socketpair,
1023 	.accept		   = inet_accept,
1024 	.getname	   = inet_getname,	/* Semantics are different.  */
1025 	.poll		   = sctp_poll,
1026 	.ioctl		   = inet_ioctl,
1027 	.listen		   = sctp_inet_listen,
1028 	.shutdown	   = inet_shutdown,	/* Looks harmless.  */
1029 	.setsockopt	   = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1030 	.getsockopt	   = sock_common_getsockopt,
1031 	.sendmsg	   = inet_sendmsg,
1032 	.recvmsg	   = sock_common_recvmsg,
1033 	.mmap		   = sock_no_mmap,
1034 	.sendpage	   = sock_no_sendpage,
1035 #ifdef CONFIG_COMPAT
1036 	.compat_setsockopt = compat_sock_common_setsockopt,
1037 	.compat_getsockopt = compat_sock_common_getsockopt,
1038 #endif
1039 };
1040 
1041 /* Registration with AF_INET family.  */
1042 static struct inet_protosw sctp_seqpacket_protosw = {
1043 	.type       = SOCK_SEQPACKET,
1044 	.protocol   = IPPROTO_SCTP,
1045 	.prot       = &sctp_prot,
1046 	.ops        = &inet_seqpacket_ops,
1047 	.no_check   = 0,
1048 	.flags      = SCTP_PROTOSW_FLAG
1049 };
1050 static struct inet_protosw sctp_stream_protosw = {
1051 	.type       = SOCK_STREAM,
1052 	.protocol   = IPPROTO_SCTP,
1053 	.prot       = &sctp_prot,
1054 	.ops        = &inet_seqpacket_ops,
1055 	.no_check   = 0,
1056 	.flags      = SCTP_PROTOSW_FLAG
1057 };
1058 
1059 /* Register with IP layer.  */
1060 static const struct net_protocol sctp_protocol = {
1061 	.handler     = sctp_rcv,
1062 	.err_handler = sctp_v4_err,
1063 	.no_policy   = 1,
1064 };
1065 
1066 /* IPv4 address related functions.  */
1067 static struct sctp_af sctp_af_inet = {
1068 	.sa_family	   = AF_INET,
1069 	.sctp_xmit	   = sctp_v4_xmit,
1070 	.setsockopt	   = ip_setsockopt,
1071 	.getsockopt	   = ip_getsockopt,
1072 	.get_dst	   = sctp_v4_get_dst,
1073 	.get_saddr	   = sctp_v4_get_saddr,
1074 	.copy_addrlist	   = sctp_v4_copy_addrlist,
1075 	.from_skb	   = sctp_v4_from_skb,
1076 	.from_sk	   = sctp_v4_from_sk,
1077 	.to_sk_saddr	   = sctp_v4_to_sk_saddr,
1078 	.to_sk_daddr	   = sctp_v4_to_sk_daddr,
1079 	.from_addr_param   = sctp_v4_from_addr_param,
1080 	.to_addr_param	   = sctp_v4_to_addr_param,
1081 	.cmp_addr	   = sctp_v4_cmp_addr,
1082 	.addr_valid	   = sctp_v4_addr_valid,
1083 	.inaddr_any	   = sctp_v4_inaddr_any,
1084 	.is_any		   = sctp_v4_is_any,
1085 	.available	   = sctp_v4_available,
1086 	.scope		   = sctp_v4_scope,
1087 	.skb_iif	   = sctp_v4_skb_iif,
1088 	.is_ce		   = sctp_v4_is_ce,
1089 	.seq_dump_addr	   = sctp_v4_seq_dump_addr,
1090 	.ecn_capable	   = sctp_v4_ecn_capable,
1091 	.net_header_len	   = sizeof(struct iphdr),
1092 	.sockaddr_len	   = sizeof(struct sockaddr_in),
1093 #ifdef CONFIG_COMPAT
1094 	.compat_setsockopt = compat_ip_setsockopt,
1095 	.compat_getsockopt = compat_ip_getsockopt,
1096 #endif
1097 };
1098 
sctp_get_pf_specific(sa_family_t family)1099 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
1100 
1101 	switch (family) {
1102 	case PF_INET:
1103 		return sctp_pf_inet_specific;
1104 	case PF_INET6:
1105 		return sctp_pf_inet6_specific;
1106 	default:
1107 		return NULL;
1108 	}
1109 }
1110 
1111 /* Register the PF specific function table.  */
sctp_register_pf(struct sctp_pf * pf,sa_family_t family)1112 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1113 {
1114 	switch (family) {
1115 	case PF_INET:
1116 		if (sctp_pf_inet_specific)
1117 			return 0;
1118 		sctp_pf_inet_specific = pf;
1119 		break;
1120 	case PF_INET6:
1121 		if (sctp_pf_inet6_specific)
1122 			return 0;
1123 		sctp_pf_inet6_specific = pf;
1124 		break;
1125 	default:
1126 		return 0;
1127 	}
1128 	return 1;
1129 }
1130 
init_sctp_mibs(void)1131 static inline int init_sctp_mibs(void)
1132 {
1133 	return snmp_mib_init((void __percpu **)sctp_statistics,
1134 			     sizeof(struct sctp_mib),
1135 			     __alignof__(struct sctp_mib));
1136 }
1137 
cleanup_sctp_mibs(void)1138 static inline void cleanup_sctp_mibs(void)
1139 {
1140 	snmp_mib_free((void __percpu **)sctp_statistics);
1141 }
1142 
sctp_v4_pf_init(void)1143 static void sctp_v4_pf_init(void)
1144 {
1145 	/* Initialize the SCTP specific PF functions. */
1146 	sctp_register_pf(&sctp_pf_inet, PF_INET);
1147 	sctp_register_af(&sctp_af_inet);
1148 }
1149 
sctp_v4_pf_exit(void)1150 static void sctp_v4_pf_exit(void)
1151 {
1152 	list_del(&sctp_af_inet.list);
1153 }
1154 
sctp_v4_protosw_init(void)1155 static int sctp_v4_protosw_init(void)
1156 {
1157 	int rc;
1158 
1159 	rc = proto_register(&sctp_prot, 1);
1160 	if (rc)
1161 		return rc;
1162 
1163 	/* Register SCTP(UDP and TCP style) with socket layer.  */
1164 	inet_register_protosw(&sctp_seqpacket_protosw);
1165 	inet_register_protosw(&sctp_stream_protosw);
1166 
1167 	return 0;
1168 }
1169 
sctp_v4_protosw_exit(void)1170 static void sctp_v4_protosw_exit(void)
1171 {
1172 	inet_unregister_protosw(&sctp_stream_protosw);
1173 	inet_unregister_protosw(&sctp_seqpacket_protosw);
1174 	proto_unregister(&sctp_prot);
1175 }
1176 
sctp_v4_add_protocol(void)1177 static int sctp_v4_add_protocol(void)
1178 {
1179 	/* Register notifier for inet address additions/deletions. */
1180 	register_inetaddr_notifier(&sctp_inetaddr_notifier);
1181 
1182 	/* Register SCTP with inet layer.  */
1183 	if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1184 		return -EAGAIN;
1185 
1186 	return 0;
1187 }
1188 
sctp_v4_del_protocol(void)1189 static void sctp_v4_del_protocol(void)
1190 {
1191 	inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1192 	unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1193 }
1194 
1195 /* Initialize the universe into something sensible.  */
sctp_init(void)1196 SCTP_STATIC __init int sctp_init(void)
1197 {
1198 	int i;
1199 	int status = -EINVAL;
1200 	unsigned long goal;
1201 	unsigned long limit;
1202 	int max_share;
1203 	int order;
1204 
1205 	/* SCTP_DEBUG sanity check. */
1206 	if (!sctp_sanity_check())
1207 		goto out;
1208 
1209 	/* Allocate bind_bucket and chunk caches. */
1210 	status = -ENOBUFS;
1211 	sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1212 					       sizeof(struct sctp_bind_bucket),
1213 					       0, SLAB_HWCACHE_ALIGN,
1214 					       NULL);
1215 	if (!sctp_bucket_cachep)
1216 		goto out;
1217 
1218 	sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1219 					       sizeof(struct sctp_chunk),
1220 					       0, SLAB_HWCACHE_ALIGN,
1221 					       NULL);
1222 	if (!sctp_chunk_cachep)
1223 		goto err_chunk_cachep;
1224 
1225 	/* Allocate and initialise sctp mibs.  */
1226 	status = init_sctp_mibs();
1227 	if (status)
1228 		goto err_init_mibs;
1229 
1230 	/* Initialize proc fs directory.  */
1231 	status = sctp_proc_init();
1232 	if (status)
1233 		goto err_init_proc;
1234 
1235 	/* Initialize object count debugging.  */
1236 	sctp_dbg_objcnt_init();
1237 
1238 	/*
1239 	 * 14. Suggested SCTP Protocol Parameter Values
1240 	 */
1241 	/* The following protocol parameters are RECOMMENDED:  */
1242 	/* RTO.Initial              - 3  seconds */
1243 	sctp_rto_initial		= SCTP_RTO_INITIAL;
1244 	/* RTO.Min                  - 1  second */
1245 	sctp_rto_min	 		= SCTP_RTO_MIN;
1246 	/* RTO.Max                 -  60 seconds */
1247 	sctp_rto_max 			= SCTP_RTO_MAX;
1248 	/* RTO.Alpha                - 1/8 */
1249 	sctp_rto_alpha	        	= SCTP_RTO_ALPHA;
1250 	/* RTO.Beta                 - 1/4 */
1251 	sctp_rto_beta			= SCTP_RTO_BETA;
1252 
1253 	/* Valid.Cookie.Life        - 60  seconds */
1254 	sctp_valid_cookie_life		= SCTP_DEFAULT_COOKIE_LIFE;
1255 
1256 	/* Whether Cookie Preservative is enabled(1) or not(0) */
1257 	sctp_cookie_preserve_enable 	= 1;
1258 
1259 	/* Max.Burst		    - 4 */
1260 	sctp_max_burst 			= SCTP_DEFAULT_MAX_BURST;
1261 
1262 	/* Association.Max.Retrans  - 10 attempts
1263 	 * Path.Max.Retrans         - 5  attempts (per destination address)
1264 	 * Max.Init.Retransmits     - 8  attempts
1265 	 */
1266 	sctp_max_retrans_association 	= 10;
1267 	sctp_max_retrans_path		= 5;
1268 	sctp_max_retrans_init		= 8;
1269 
1270 	/* Sendbuffer growth	    - do per-socket accounting */
1271 	sctp_sndbuf_policy		= 0;
1272 
1273 	/* Rcvbuffer growth	    - do per-socket accounting */
1274 	sctp_rcvbuf_policy		= 0;
1275 
1276 	/* HB.interval              - 30 seconds */
1277 	sctp_hb_interval		= SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1278 
1279 	/* delayed SACK timeout */
1280 	sctp_sack_timeout		= SCTP_DEFAULT_TIMEOUT_SACK;
1281 
1282 	/* Implementation specific variables. */
1283 
1284 	/* Initialize default stream count setup information. */
1285 	sctp_max_instreams    		= SCTP_DEFAULT_INSTREAMS;
1286 	sctp_max_outstreams   		= SCTP_DEFAULT_OUTSTREAMS;
1287 
1288 	/* Initialize maximum autoclose timeout. */
1289 	sctp_max_autoclose		= INT_MAX / HZ;
1290 
1291 	/* Initialize handle used for association ids. */
1292 	idr_init(&sctp_assocs_id);
1293 
1294 	limit = nr_free_buffer_pages() / 8;
1295 	limit = max(limit, 128UL);
1296 	sysctl_sctp_mem[0] = limit / 4 * 3;
1297 	sysctl_sctp_mem[1] = limit;
1298 	sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1299 
1300 	/* Set per-socket limits to no more than 1/128 the pressure threshold*/
1301 	limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1302 	max_share = min(4UL*1024*1024, limit);
1303 
1304 	sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1305 	sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1306 	sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1307 
1308 	sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1309 	sysctl_sctp_wmem[1] = 16*1024;
1310 	sysctl_sctp_wmem[2] = max(64*1024, max_share);
1311 
1312 	/* Size and allocate the association hash table.
1313 	 * The methodology is similar to that of the tcp hash tables.
1314 	 */
1315 	if (totalram_pages >= (128 * 1024))
1316 		goal = totalram_pages >> (22 - PAGE_SHIFT);
1317 	else
1318 		goal = totalram_pages >> (24 - PAGE_SHIFT);
1319 
1320 	for (order = 0; (1UL << order) < goal; order++)
1321 		;
1322 
1323 	do {
1324 		sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1325 					sizeof(struct sctp_hashbucket);
1326 		if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1327 			continue;
1328 		sctp_assoc_hashtable = (struct sctp_hashbucket *)
1329 			__get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1330 	} while (!sctp_assoc_hashtable && --order > 0);
1331 	if (!sctp_assoc_hashtable) {
1332 		pr_err("Failed association hash alloc\n");
1333 		status = -ENOMEM;
1334 		goto err_ahash_alloc;
1335 	}
1336 	for (i = 0; i < sctp_assoc_hashsize; i++) {
1337 		rwlock_init(&sctp_assoc_hashtable[i].lock);
1338 		INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1339 	}
1340 
1341 	/* Allocate and initialize the endpoint hash table.  */
1342 	sctp_ep_hashsize = 64;
1343 	sctp_ep_hashtable = (struct sctp_hashbucket *)
1344 		kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1345 	if (!sctp_ep_hashtable) {
1346 		pr_err("Failed endpoint_hash alloc\n");
1347 		status = -ENOMEM;
1348 		goto err_ehash_alloc;
1349 	}
1350 	for (i = 0; i < sctp_ep_hashsize; i++) {
1351 		rwlock_init(&sctp_ep_hashtable[i].lock);
1352 		INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1353 	}
1354 
1355 	/* Allocate and initialize the SCTP port hash table.  */
1356 	do {
1357 		sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1358 					sizeof(struct sctp_bind_hashbucket);
1359 		if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1360 			continue;
1361 		sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1362 			__get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1363 	} while (!sctp_port_hashtable && --order > 0);
1364 	if (!sctp_port_hashtable) {
1365 		pr_err("Failed bind hash alloc\n");
1366 		status = -ENOMEM;
1367 		goto err_bhash_alloc;
1368 	}
1369 	for (i = 0; i < sctp_port_hashsize; i++) {
1370 		spin_lock_init(&sctp_port_hashtable[i].lock);
1371 		INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1372 	}
1373 
1374 	pr_info("Hash tables configured (established %d bind %d)\n",
1375 		sctp_assoc_hashsize, sctp_port_hashsize);
1376 
1377 	/* Disable ADDIP by default. */
1378 	sctp_addip_enable = 0;
1379 	sctp_addip_noauth = 0;
1380 	sctp_default_auto_asconf = 0;
1381 
1382 	/* Enable PR-SCTP by default. */
1383 	sctp_prsctp_enable = 1;
1384 
1385 	/* Disable AUTH by default. */
1386 	sctp_auth_enable = 0;
1387 
1388 	/* Set SCOPE policy to enabled */
1389 	sctp_scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1390 
1391 	/* Set the default rwnd update threshold */
1392 	sctp_rwnd_upd_shift		= SCTP_DEFAULT_RWND_SHIFT;
1393 
1394 	sctp_sysctl_register();
1395 
1396 	INIT_LIST_HEAD(&sctp_address_families);
1397 	sctp_v4_pf_init();
1398 	sctp_v6_pf_init();
1399 
1400 	/* Initialize the local address list. */
1401 	INIT_LIST_HEAD(&sctp_local_addr_list);
1402 	spin_lock_init(&sctp_local_addr_lock);
1403 	sctp_get_local_addr_list();
1404 
1405 	/* Initialize the address event list */
1406 	INIT_LIST_HEAD(&sctp_addr_waitq);
1407 	INIT_LIST_HEAD(&sctp_auto_asconf_splist);
1408 	spin_lock_init(&sctp_addr_wq_lock);
1409 	sctp_addr_wq_timer.expires = 0;
1410 	setup_timer(&sctp_addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1411 
1412 	status = sctp_v4_protosw_init();
1413 
1414 	if (status)
1415 		goto err_protosw_init;
1416 
1417 	status = sctp_v6_protosw_init();
1418 	if (status)
1419 		goto err_v6_protosw_init;
1420 
1421 	/* Initialize the control inode/socket for handling OOTB packets.  */
1422 	if ((status = sctp_ctl_sock_init())) {
1423 		pr_err("Failed to initialize the SCTP control sock\n");
1424 		goto err_ctl_sock_init;
1425 	}
1426 
1427 	status = sctp_v4_add_protocol();
1428 	if (status)
1429 		goto err_add_protocol;
1430 
1431 	/* Register SCTP with inet6 layer.  */
1432 	status = sctp_v6_add_protocol();
1433 	if (status)
1434 		goto err_v6_add_protocol;
1435 
1436 	status = 0;
1437 out:
1438 	return status;
1439 err_v6_add_protocol:
1440 	sctp_v4_del_protocol();
1441 err_add_protocol:
1442 	inet_ctl_sock_destroy(sctp_ctl_sock);
1443 err_ctl_sock_init:
1444 	sctp_v6_protosw_exit();
1445 err_v6_protosw_init:
1446 	sctp_v4_protosw_exit();
1447 err_protosw_init:
1448 	sctp_free_local_addr_list();
1449 	sctp_v4_pf_exit();
1450 	sctp_v6_pf_exit();
1451 	sctp_sysctl_unregister();
1452 	free_pages((unsigned long)sctp_port_hashtable,
1453 		   get_order(sctp_port_hashsize *
1454 			     sizeof(struct sctp_bind_hashbucket)));
1455 err_bhash_alloc:
1456 	kfree(sctp_ep_hashtable);
1457 err_ehash_alloc:
1458 	free_pages((unsigned long)sctp_assoc_hashtable,
1459 		   get_order(sctp_assoc_hashsize *
1460 			     sizeof(struct sctp_hashbucket)));
1461 err_ahash_alloc:
1462 	sctp_dbg_objcnt_exit();
1463 	sctp_proc_exit();
1464 err_init_proc:
1465 	cleanup_sctp_mibs();
1466 err_init_mibs:
1467 	kmem_cache_destroy(sctp_chunk_cachep);
1468 err_chunk_cachep:
1469 	kmem_cache_destroy(sctp_bucket_cachep);
1470 	goto out;
1471 }
1472 
1473 /* Exit handler for the SCTP protocol.  */
sctp_exit(void)1474 SCTP_STATIC __exit void sctp_exit(void)
1475 {
1476 	/* BUG.  This should probably do something useful like clean
1477 	 * up all the remaining associations and all that memory.
1478 	 */
1479 
1480 	/* Unregister with inet6/inet layers. */
1481 	sctp_v6_del_protocol();
1482 	sctp_v4_del_protocol();
1483 	sctp_free_addr_wq();
1484 
1485 	/* Free the control endpoint.  */
1486 	inet_ctl_sock_destroy(sctp_ctl_sock);
1487 
1488 	/* Free protosw registrations */
1489 	sctp_v6_protosw_exit();
1490 	sctp_v4_protosw_exit();
1491 
1492 	/* Free the local address list.  */
1493 	sctp_free_local_addr_list();
1494 
1495 	/* Unregister with socket layer. */
1496 	sctp_v6_pf_exit();
1497 	sctp_v4_pf_exit();
1498 
1499 	sctp_sysctl_unregister();
1500 
1501 	free_pages((unsigned long)sctp_assoc_hashtable,
1502 		   get_order(sctp_assoc_hashsize *
1503 			     sizeof(struct sctp_hashbucket)));
1504 	kfree(sctp_ep_hashtable);
1505 	free_pages((unsigned long)sctp_port_hashtable,
1506 		   get_order(sctp_port_hashsize *
1507 			     sizeof(struct sctp_bind_hashbucket)));
1508 
1509 	sctp_dbg_objcnt_exit();
1510 	sctp_proc_exit();
1511 	cleanup_sctp_mibs();
1512 
1513 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1514 
1515 	kmem_cache_destroy(sctp_chunk_cachep);
1516 	kmem_cache_destroy(sctp_bucket_cachep);
1517 }
1518 
1519 module_init(sctp_init);
1520 module_exit(sctp_exit);
1521 
1522 /*
1523  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1524  */
1525 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1526 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1527 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1528 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1529 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1530 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1531 MODULE_LICENSE("GPL");
1532