xref: /openbmc/linux/net/sctp/protocol.c (revision 29d97219)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* SCTP kernel implementation
3  * (C) Copyright IBM Corp. 2001, 2004
4  * Copyright (c) 1999-2000 Cisco, Inc.
5  * Copyright (c) 1999-2001 Motorola, Inc.
6  * Copyright (c) 2001 Intel Corp.
7  * Copyright (c) 2001 Nokia, Inc.
8  * Copyright (c) 2001 La Monte H.P. Yarroll
9  *
10  * This file is part of the SCTP kernel implementation
11  *
12  * Initialization/cleanup for SCTP protocol support.
13  *
14  * Please send any bug reports or fixes you make to the
15  * email address(es):
16  *    lksctp developers <linux-sctp@vger.kernel.org>
17  *
18  * Written or modified by:
19  *    La Monte H.P. Yarroll <piggy@acm.org>
20  *    Karl Knutson <karl@athena.chicago.il.us>
21  *    Jon Grimm <jgrimm@us.ibm.com>
22  *    Sridhar Samudrala <sri@us.ibm.com>
23  *    Daisy Chang <daisyc@us.ibm.com>
24  *    Ardelle Fan <ardelle.fan@intel.com>
25  */
26 
27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
28 
29 #include <linux/module.h>
30 #include <linux/init.h>
31 #include <linux/netdevice.h>
32 #include <linux/inetdevice.h>
33 #include <linux/seq_file.h>
34 #include <linux/memblock.h>
35 #include <linux/highmem.h>
36 #include <linux/swap.h>
37 #include <linux/slab.h>
38 #include <net/net_namespace.h>
39 #include <net/protocol.h>
40 #include <net/ip.h>
41 #include <net/ipv6.h>
42 #include <net/route.h>
43 #include <net/sctp/sctp.h>
44 #include <net/addrconf.h>
45 #include <net/inet_common.h>
46 #include <net/inet_ecn.h>
47 #include <net/udp_tunnel.h>
48 
49 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
50 
51 /* Global data structures. */
52 struct sctp_globals sctp_globals __read_mostly;
53 
54 struct idr sctp_assocs_id;
55 DEFINE_SPINLOCK(sctp_assocs_id_lock);
56 
57 static struct sctp_pf *sctp_pf_inet6_specific;
58 static struct sctp_pf *sctp_pf_inet_specific;
59 static struct sctp_af *sctp_af_v4_specific;
60 static struct sctp_af *sctp_af_v6_specific;
61 
62 struct kmem_cache *sctp_chunk_cachep __read_mostly;
63 struct kmem_cache *sctp_bucket_cachep __read_mostly;
64 
65 long sysctl_sctp_mem[3];
66 int sysctl_sctp_rmem[3];
67 int sysctl_sctp_wmem[3];
68 
69 /* Private helper to extract ipv4 address and stash them in
70  * the protocol structure.
71  */
72 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
73 				  struct net_device *dev)
74 {
75 	struct in_device *in_dev;
76 	struct in_ifaddr *ifa;
77 	struct sctp_sockaddr_entry *addr;
78 
79 	rcu_read_lock();
80 	if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
81 		rcu_read_unlock();
82 		return;
83 	}
84 
85 	in_dev_for_each_ifa_rcu(ifa, in_dev) {
86 		/* Add the address to the local list.  */
87 		addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
88 		if (addr) {
89 			addr->a.v4.sin_family = AF_INET;
90 			addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
91 			addr->valid = 1;
92 			INIT_LIST_HEAD(&addr->list);
93 			list_add_tail(&addr->list, addrlist);
94 		}
95 	}
96 
97 	rcu_read_unlock();
98 }
99 
100 /* Extract our IP addresses from the system and stash them in the
101  * protocol structure.
102  */
103 static void sctp_get_local_addr_list(struct net *net)
104 {
105 	struct net_device *dev;
106 	struct list_head *pos;
107 	struct sctp_af *af;
108 
109 	rcu_read_lock();
110 	for_each_netdev_rcu(net, dev) {
111 		list_for_each(pos, &sctp_address_families) {
112 			af = list_entry(pos, struct sctp_af, list);
113 			af->copy_addrlist(&net->sctp.local_addr_list, dev);
114 		}
115 	}
116 	rcu_read_unlock();
117 }
118 
119 /* Free the existing local addresses.  */
120 static void sctp_free_local_addr_list(struct net *net)
121 {
122 	struct sctp_sockaddr_entry *addr;
123 	struct list_head *pos, *temp;
124 
125 	list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
126 		addr = list_entry(pos, struct sctp_sockaddr_entry, list);
127 		list_del(pos);
128 		kfree(addr);
129 	}
130 }
131 
132 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
133 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
134 			      enum sctp_scope scope, gfp_t gfp, int copy_flags)
135 {
136 	struct sctp_sockaddr_entry *addr;
137 	union sctp_addr laddr;
138 	int error = 0;
139 
140 	rcu_read_lock();
141 	list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
142 		if (!addr->valid)
143 			continue;
144 		if (!sctp_in_scope(net, &addr->a, scope))
145 			continue;
146 
147 		/* Now that the address is in scope, check to see if
148 		 * the address type is really supported by the local
149 		 * sock as well as the remote peer.
150 		 */
151 		if (addr->a.sa.sa_family == AF_INET &&
152 		    (!(copy_flags & SCTP_ADDR4_ALLOWED) ||
153 		     !(copy_flags & SCTP_ADDR4_PEERSUPP)))
154 			continue;
155 		if (addr->a.sa.sa_family == AF_INET6 &&
156 		    (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
157 		     !(copy_flags & SCTP_ADDR6_PEERSUPP)))
158 			continue;
159 
160 		laddr = addr->a;
161 		/* also works for setting ipv6 address port */
162 		laddr.v4.sin_port = htons(bp->port);
163 		if (sctp_bind_addr_state(bp, &laddr) != -1)
164 			continue;
165 
166 		error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
167 					   SCTP_ADDR_SRC, GFP_ATOMIC);
168 		if (error)
169 			break;
170 	}
171 
172 	rcu_read_unlock();
173 	return error;
174 }
175 
176 /* Copy over any ip options */
177 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
178 {
179 	struct inet_sock *newinet, *inet = inet_sk(sk);
180 	struct ip_options_rcu *inet_opt, *newopt = NULL;
181 
182 	newinet = inet_sk(newsk);
183 
184 	rcu_read_lock();
185 	inet_opt = rcu_dereference(inet->inet_opt);
186 	if (inet_opt) {
187 		newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
188 				      inet_opt->opt.optlen, GFP_ATOMIC);
189 		if (newopt)
190 			memcpy(newopt, inet_opt, sizeof(*inet_opt) +
191 			       inet_opt->opt.optlen);
192 		else
193 			pr_err("%s: Failed to copy ip options\n", __func__);
194 	}
195 	RCU_INIT_POINTER(newinet->inet_opt, newopt);
196 	rcu_read_unlock();
197 }
198 
199 /* Account for the IP options */
200 static int sctp_v4_ip_options_len(struct sock *sk)
201 {
202 	struct inet_sock *inet = inet_sk(sk);
203 	struct ip_options_rcu *inet_opt;
204 	int len = 0;
205 
206 	rcu_read_lock();
207 	inet_opt = rcu_dereference(inet->inet_opt);
208 	if (inet_opt)
209 		len = inet_opt->opt.optlen;
210 
211 	rcu_read_unlock();
212 	return len;
213 }
214 
215 /* Initialize a sctp_addr from in incoming skb.  */
216 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
217 			     int is_saddr)
218 {
219 	/* Always called on head skb, so this is safe */
220 	struct sctphdr *sh = sctp_hdr(skb);
221 	struct sockaddr_in *sa = &addr->v4;
222 
223 	addr->v4.sin_family = AF_INET;
224 
225 	if (is_saddr) {
226 		sa->sin_port = sh->source;
227 		sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
228 	} else {
229 		sa->sin_port = sh->dest;
230 		sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
231 	}
232 	memset(sa->sin_zero, 0, sizeof(sa->sin_zero));
233 }
234 
235 /* Initialize an sctp_addr from a socket. */
236 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
237 {
238 	addr->v4.sin_family = AF_INET;
239 	addr->v4.sin_port = 0;
240 	addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
241 	memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
242 }
243 
244 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
245 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
246 {
247 	inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
248 }
249 
250 /* Initialize sk->sk_daddr from sctp_addr. */
251 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
252 {
253 	inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
254 }
255 
256 /* Initialize a sctp_addr from an address parameter. */
257 static void sctp_v4_from_addr_param(union sctp_addr *addr,
258 				    union sctp_addr_param *param,
259 				    __be16 port, int iif)
260 {
261 	addr->v4.sin_family = AF_INET;
262 	addr->v4.sin_port = port;
263 	addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
264 	memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
265 }
266 
267 /* Initialize an address parameter from a sctp_addr and return the length
268  * of the address parameter.
269  */
270 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
271 				 union sctp_addr_param *param)
272 {
273 	int length = sizeof(struct sctp_ipv4addr_param);
274 
275 	param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
276 	param->v4.param_hdr.length = htons(length);
277 	param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
278 
279 	return length;
280 }
281 
282 /* Initialize a sctp_addr from a dst_entry. */
283 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
284 			      __be16 port)
285 {
286 	saddr->v4.sin_family = AF_INET;
287 	saddr->v4.sin_port = port;
288 	saddr->v4.sin_addr.s_addr = fl4->saddr;
289 	memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
290 }
291 
292 /* Compare two addresses exactly. */
293 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
294 			    const union sctp_addr *addr2)
295 {
296 	if (addr1->sa.sa_family != addr2->sa.sa_family)
297 		return 0;
298 	if (addr1->v4.sin_port != addr2->v4.sin_port)
299 		return 0;
300 	if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
301 		return 0;
302 
303 	return 1;
304 }
305 
306 /* Initialize addr struct to INADDR_ANY. */
307 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
308 {
309 	addr->v4.sin_family = AF_INET;
310 	addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
311 	addr->v4.sin_port = port;
312 	memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
313 }
314 
315 /* Is this a wildcard address? */
316 static int sctp_v4_is_any(const union sctp_addr *addr)
317 {
318 	return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
319 }
320 
321 /* This function checks if the address is a valid address to be used for
322  * SCTP binding.
323  *
324  * Output:
325  * Return 0 - If the address is a non-unicast or an illegal address.
326  * Return 1 - If the address is a unicast.
327  */
328 static int sctp_v4_addr_valid(union sctp_addr *addr,
329 			      struct sctp_sock *sp,
330 			      const struct sk_buff *skb)
331 {
332 	/* IPv4 addresses not allowed */
333 	if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
334 		return 0;
335 
336 	/* Is this a non-unicast address or a unusable SCTP address? */
337 	if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
338 		return 0;
339 
340 	/* Is this a broadcast address? */
341 	if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
342 		return 0;
343 
344 	return 1;
345 }
346 
347 /* Should this be available for binding?   */
348 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
349 {
350 	struct net *net = sock_net(&sp->inet.sk);
351 	int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
352 
353 
354 	if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
355 	   ret != RTN_LOCAL &&
356 	   !sp->inet.freebind &&
357 	   !net->ipv4.sysctl_ip_nonlocal_bind)
358 		return 0;
359 
360 	if (ipv6_only_sock(sctp_opt2sk(sp)))
361 		return 0;
362 
363 	return 1;
364 }
365 
366 /* Checking the loopback, private and other address scopes as defined in
367  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
368  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
369  *
370  * Level 0 - unusable SCTP addresses
371  * Level 1 - loopback address
372  * Level 2 - link-local addresses
373  * Level 3 - private addresses.
374  * Level 4 - global addresses
375  * For INIT and INIT-ACK address list, let L be the level of
376  * requested destination address, sender and receiver
377  * SHOULD include all of its addresses with level greater
378  * than or equal to L.
379  *
380  * IPv4 scoping can be controlled through sysctl option
381  * net.sctp.addr_scope_policy
382  */
383 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
384 {
385 	enum sctp_scope retval;
386 
387 	/* Check for unusable SCTP addresses. */
388 	if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
389 		retval =  SCTP_SCOPE_UNUSABLE;
390 	} else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
391 		retval = SCTP_SCOPE_LOOPBACK;
392 	} else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
393 		retval = SCTP_SCOPE_LINK;
394 	} else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
395 		   ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
396 		   ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
397 		retval = SCTP_SCOPE_PRIVATE;
398 	} else {
399 		retval = SCTP_SCOPE_GLOBAL;
400 	}
401 
402 	return retval;
403 }
404 
405 /* Returns a valid dst cache entry for the given source and destination ip
406  * addresses. If an association is passed, trys to get a dst entry with a
407  * source address that matches an address in the bind address list.
408  */
409 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
410 				struct flowi *fl, struct sock *sk)
411 {
412 	struct sctp_association *asoc = t->asoc;
413 	struct rtable *rt;
414 	struct flowi _fl;
415 	struct flowi4 *fl4 = &_fl.u.ip4;
416 	struct sctp_bind_addr *bp;
417 	struct sctp_sockaddr_entry *laddr;
418 	struct dst_entry *dst = NULL;
419 	union sctp_addr *daddr = &t->ipaddr;
420 	union sctp_addr dst_saddr;
421 	__u8 tos = inet_sk(sk)->tos;
422 
423 	if (t->dscp & SCTP_DSCP_SET_MASK)
424 		tos = t->dscp & SCTP_DSCP_VAL_MASK;
425 	memset(&_fl, 0x0, sizeof(_fl));
426 	fl4->daddr  = daddr->v4.sin_addr.s_addr;
427 	fl4->fl4_dport = daddr->v4.sin_port;
428 	fl4->flowi4_proto = IPPROTO_SCTP;
429 	if (asoc) {
430 		fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
431 		fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
432 		fl4->fl4_sport = htons(asoc->base.bind_addr.port);
433 	}
434 	if (saddr) {
435 		fl4->saddr = saddr->v4.sin_addr.s_addr;
436 		if (!fl4->fl4_sport)
437 			fl4->fl4_sport = saddr->v4.sin_port;
438 	}
439 
440 	pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
441 		 &fl4->saddr);
442 
443 	rt = ip_route_output_key(sock_net(sk), fl4);
444 	if (!IS_ERR(rt)) {
445 		dst = &rt->dst;
446 		t->dst = dst;
447 		memcpy(fl, &_fl, sizeof(_fl));
448 	}
449 
450 	/* If there is no association or if a source address is passed, no
451 	 * more validation is required.
452 	 */
453 	if (!asoc || saddr)
454 		goto out;
455 
456 	bp = &asoc->base.bind_addr;
457 
458 	if (dst) {
459 		/* Walk through the bind address list and look for a bind
460 		 * address that matches the source address of the returned dst.
461 		 */
462 		sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
463 		rcu_read_lock();
464 		list_for_each_entry_rcu(laddr, &bp->address_list, list) {
465 			if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
466 			    (laddr->state != SCTP_ADDR_SRC &&
467 			    !asoc->src_out_of_asoc_ok))
468 				continue;
469 			if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
470 				goto out_unlock;
471 		}
472 		rcu_read_unlock();
473 
474 		/* None of the bound addresses match the source address of the
475 		 * dst. So release it.
476 		 */
477 		dst_release(dst);
478 		dst = NULL;
479 	}
480 
481 	/* Walk through the bind address list and try to get a dst that
482 	 * matches a bind address as the source address.
483 	 */
484 	rcu_read_lock();
485 	list_for_each_entry_rcu(laddr, &bp->address_list, list) {
486 		struct net_device *odev;
487 
488 		if (!laddr->valid)
489 			continue;
490 		if (laddr->state != SCTP_ADDR_SRC ||
491 		    AF_INET != laddr->a.sa.sa_family)
492 			continue;
493 
494 		fl4->fl4_sport = laddr->a.v4.sin_port;
495 		flowi4_update_output(fl4,
496 				     asoc->base.sk->sk_bound_dev_if,
497 				     RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
498 				     daddr->v4.sin_addr.s_addr,
499 				     laddr->a.v4.sin_addr.s_addr);
500 
501 		rt = ip_route_output_key(sock_net(sk), fl4);
502 		if (IS_ERR(rt))
503 			continue;
504 
505 		/* Ensure the src address belongs to the output
506 		 * interface.
507 		 */
508 		odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
509 				     false);
510 		if (!odev || odev->ifindex != fl4->flowi4_oif) {
511 			if (!dst) {
512 				dst = &rt->dst;
513 				t->dst = dst;
514 				memcpy(fl, &_fl, sizeof(_fl));
515 			} else {
516 				dst_release(&rt->dst);
517 			}
518 			continue;
519 		}
520 
521 		dst_release(dst);
522 		dst = &rt->dst;
523 		t->dst = dst;
524 		memcpy(fl, &_fl, sizeof(_fl));
525 		break;
526 	}
527 
528 out_unlock:
529 	rcu_read_unlock();
530 out:
531 	if (dst) {
532 		pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
533 			 &fl->u.ip4.daddr, &fl->u.ip4.saddr);
534 	} else {
535 		t->dst = NULL;
536 		pr_debug("no route\n");
537 	}
538 }
539 
540 /* For v4, the source address is cached in the route entry(dst). So no need
541  * to cache it separately and hence this is an empty routine.
542  */
543 static void sctp_v4_get_saddr(struct sctp_sock *sk,
544 			      struct sctp_transport *t,
545 			      struct flowi *fl)
546 {
547 	union sctp_addr *saddr = &t->saddr;
548 	struct rtable *rt = (struct rtable *)t->dst;
549 
550 	if (rt) {
551 		saddr->v4.sin_family = AF_INET;
552 		saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
553 	}
554 }
555 
556 /* What interface did this skb arrive on? */
557 static int sctp_v4_skb_iif(const struct sk_buff *skb)
558 {
559 	return inet_iif(skb);
560 }
561 
562 /* Was this packet marked by Explicit Congestion Notification? */
563 static int sctp_v4_is_ce(const struct sk_buff *skb)
564 {
565 	return INET_ECN_is_ce(ip_hdr(skb)->tos);
566 }
567 
568 /* Create and initialize a new sk for the socket returned by accept(). */
569 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
570 					     struct sctp_association *asoc,
571 					     bool kern)
572 {
573 	struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
574 			sk->sk_prot, kern);
575 	struct inet_sock *newinet;
576 
577 	if (!newsk)
578 		goto out;
579 
580 	sock_init_data(NULL, newsk);
581 
582 	sctp_copy_sock(newsk, sk, asoc);
583 	sock_reset_flag(newsk, SOCK_ZAPPED);
584 
585 	sctp_v4_copy_ip_options(sk, newsk);
586 
587 	newinet = inet_sk(newsk);
588 
589 	newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
590 
591 	sk_refcnt_debug_inc(newsk);
592 
593 	if (newsk->sk_prot->init(newsk)) {
594 		sk_common_release(newsk);
595 		newsk = NULL;
596 	}
597 
598 out:
599 	return newsk;
600 }
601 
602 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
603 {
604 	/* No address mapping for V4 sockets */
605 	memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
606 	return sizeof(struct sockaddr_in);
607 }
608 
609 /* Dump the v4 addr to the seq file. */
610 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
611 {
612 	seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
613 }
614 
615 static void sctp_v4_ecn_capable(struct sock *sk)
616 {
617 	INET_ECN_xmit(sk);
618 }
619 
620 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
621 {
622 	struct net *net = from_timer(net, t, sctp.addr_wq_timer);
623 	struct sctp_sockaddr_entry *addrw, *temp;
624 	struct sctp_sock *sp;
625 
626 	spin_lock_bh(&net->sctp.addr_wq_lock);
627 
628 	list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
629 		pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
630 			 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
631 			 addrw->state, addrw);
632 
633 #if IS_ENABLED(CONFIG_IPV6)
634 		/* Now we send an ASCONF for each association */
635 		/* Note. we currently don't handle link local IPv6 addressees */
636 		if (addrw->a.sa.sa_family == AF_INET6) {
637 			struct in6_addr *in6;
638 
639 			if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
640 			    IPV6_ADDR_LINKLOCAL)
641 				goto free_next;
642 
643 			in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
644 			if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
645 			    addrw->state == SCTP_ADDR_NEW) {
646 				unsigned long timeo_val;
647 
648 				pr_debug("%s: this is on DAD, trying %d sec "
649 					 "later\n", __func__,
650 					 SCTP_ADDRESS_TICK_DELAY);
651 
652 				timeo_val = jiffies;
653 				timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
654 				mod_timer(&net->sctp.addr_wq_timer, timeo_val);
655 				break;
656 			}
657 		}
658 #endif
659 		list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
660 			struct sock *sk;
661 
662 			sk = sctp_opt2sk(sp);
663 			/* ignore bound-specific endpoints */
664 			if (!sctp_is_ep_boundall(sk))
665 				continue;
666 			bh_lock_sock(sk);
667 			if (sctp_asconf_mgmt(sp, addrw) < 0)
668 				pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
669 			bh_unlock_sock(sk);
670 		}
671 #if IS_ENABLED(CONFIG_IPV6)
672 free_next:
673 #endif
674 		list_del(&addrw->list);
675 		kfree(addrw);
676 	}
677 	spin_unlock_bh(&net->sctp.addr_wq_lock);
678 }
679 
680 static void sctp_free_addr_wq(struct net *net)
681 {
682 	struct sctp_sockaddr_entry *addrw;
683 	struct sctp_sockaddr_entry *temp;
684 
685 	spin_lock_bh(&net->sctp.addr_wq_lock);
686 	del_timer(&net->sctp.addr_wq_timer);
687 	list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
688 		list_del(&addrw->list);
689 		kfree(addrw);
690 	}
691 	spin_unlock_bh(&net->sctp.addr_wq_lock);
692 }
693 
694 /* lookup the entry for the same address in the addr_waitq
695  * sctp_addr_wq MUST be locked
696  */
697 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
698 					struct sctp_sockaddr_entry *addr)
699 {
700 	struct sctp_sockaddr_entry *addrw;
701 
702 	list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
703 		if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
704 			continue;
705 		if (addrw->a.sa.sa_family == AF_INET) {
706 			if (addrw->a.v4.sin_addr.s_addr ==
707 			    addr->a.v4.sin_addr.s_addr)
708 				return addrw;
709 		} else if (addrw->a.sa.sa_family == AF_INET6) {
710 			if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
711 			    &addr->a.v6.sin6_addr))
712 				return addrw;
713 		}
714 	}
715 	return NULL;
716 }
717 
718 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
719 {
720 	struct sctp_sockaddr_entry *addrw;
721 	unsigned long timeo_val;
722 
723 	/* first, we check if an opposite message already exist in the queue.
724 	 * If we found such message, it is removed.
725 	 * This operation is a bit stupid, but the DHCP client attaches the
726 	 * new address after a couple of addition and deletion of that address
727 	 */
728 
729 	spin_lock_bh(&net->sctp.addr_wq_lock);
730 	/* Offsets existing events in addr_wq */
731 	addrw = sctp_addr_wq_lookup(net, addr);
732 	if (addrw) {
733 		if (addrw->state != cmd) {
734 			pr_debug("%s: offsets existing entry for %d, addr:%pISc "
735 				 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
736 				 &net->sctp.addr_waitq);
737 
738 			list_del(&addrw->list);
739 			kfree(addrw);
740 		}
741 		spin_unlock_bh(&net->sctp.addr_wq_lock);
742 		return;
743 	}
744 
745 	/* OK, we have to add the new address to the wait queue */
746 	addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
747 	if (addrw == NULL) {
748 		spin_unlock_bh(&net->sctp.addr_wq_lock);
749 		return;
750 	}
751 	addrw->state = cmd;
752 	list_add_tail(&addrw->list, &net->sctp.addr_waitq);
753 
754 	pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
755 		 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
756 
757 	if (!timer_pending(&net->sctp.addr_wq_timer)) {
758 		timeo_val = jiffies;
759 		timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
760 		mod_timer(&net->sctp.addr_wq_timer, timeo_val);
761 	}
762 	spin_unlock_bh(&net->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  */
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 	struct net *net = dev_net(ifa->ifa_dev->dev);
778 	int found = 0;
779 
780 	switch (ev) {
781 	case NETDEV_UP:
782 		addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
783 		if (addr) {
784 			addr->a.v4.sin_family = AF_INET;
785 			addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
786 			addr->valid = 1;
787 			spin_lock_bh(&net->sctp.local_addr_lock);
788 			list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
789 			sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
790 			spin_unlock_bh(&net->sctp.local_addr_lock);
791 		}
792 		break;
793 	case NETDEV_DOWN:
794 		spin_lock_bh(&net->sctp.local_addr_lock);
795 		list_for_each_entry_safe(addr, temp,
796 					&net->sctp.local_addr_list, list) {
797 			if (addr->a.sa.sa_family == AF_INET &&
798 					addr->a.v4.sin_addr.s_addr ==
799 					ifa->ifa_local) {
800 				sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
801 				found = 1;
802 				addr->valid = 0;
803 				list_del_rcu(&addr->list);
804 				break;
805 			}
806 		}
807 		spin_unlock_bh(&net->sctp.local_addr_lock);
808 		if (found)
809 			kfree_rcu(addr, rcu);
810 		break;
811 	}
812 
813 	return NOTIFY_DONE;
814 }
815 
816 /*
817  * Initialize the control inode/socket with a control endpoint data
818  * structure.  This endpoint is reserved exclusively for the OOTB processing.
819  */
820 static int sctp_ctl_sock_init(struct net *net)
821 {
822 	int err;
823 	sa_family_t family = PF_INET;
824 
825 	if (sctp_get_pf_specific(PF_INET6))
826 		family = PF_INET6;
827 
828 	err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
829 				   SOCK_SEQPACKET, IPPROTO_SCTP, net);
830 
831 	/* If IPv6 socket could not be created, try the IPv4 socket */
832 	if (err < 0 && family == PF_INET6)
833 		err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
834 					   SOCK_SEQPACKET, IPPROTO_SCTP,
835 					   net);
836 
837 	if (err < 0) {
838 		pr_err("Failed to create the SCTP control socket\n");
839 		return err;
840 	}
841 	return 0;
842 }
843 
844 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb)
845 {
846 	SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source;
847 
848 	skb_set_transport_header(skb, sizeof(struct udphdr));
849 	sctp_rcv(skb);
850 	return 0;
851 }
852 
853 static int sctp_udp_err_lookup(struct sock *sk, struct sk_buff *skb)
854 {
855 	struct sctp_association *asoc;
856 	struct sctp_transport *t;
857 	int family;
858 
859 	skb->transport_header += sizeof(struct udphdr);
860 	family = (ip_hdr(skb)->version == 4) ? AF_INET : AF_INET6;
861 	sk = sctp_err_lookup(dev_net(skb->dev), family, skb, sctp_hdr(skb),
862 			     &asoc, &t);
863 	if (!sk)
864 		return -ENOENT;
865 
866 	sctp_err_finish(sk, t);
867 	return 0;
868 }
869 
870 int sctp_udp_sock_start(struct net *net)
871 {
872 	struct udp_tunnel_sock_cfg tuncfg = {NULL};
873 	struct udp_port_cfg udp_conf = {0};
874 	struct socket *sock;
875 	int err;
876 
877 	udp_conf.family = AF_INET;
878 	udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
879 	udp_conf.local_udp_port = htons(net->sctp.udp_port);
880 	err = udp_sock_create(net, &udp_conf, &sock);
881 	if (err) {
882 		pr_err("Failed to create the SCTP UDP tunneling v4 sock\n");
883 		return err;
884 	}
885 
886 	tuncfg.encap_type = 1;
887 	tuncfg.encap_rcv = sctp_udp_rcv;
888 	tuncfg.encap_err_lookup = sctp_udp_err_lookup;
889 	setup_udp_tunnel_sock(net, sock, &tuncfg);
890 	net->sctp.udp4_sock = sock->sk;
891 
892 #if IS_ENABLED(CONFIG_IPV6)
893 	memset(&udp_conf, 0, sizeof(udp_conf));
894 
895 	udp_conf.family = AF_INET6;
896 	udp_conf.local_ip6 = in6addr_any;
897 	udp_conf.local_udp_port = htons(net->sctp.udp_port);
898 	udp_conf.use_udp6_rx_checksums = true;
899 	udp_conf.ipv6_v6only = true;
900 	err = udp_sock_create(net, &udp_conf, &sock);
901 	if (err) {
902 		pr_err("Failed to create the SCTP UDP tunneling v6 sock\n");
903 		udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
904 		net->sctp.udp4_sock = NULL;
905 		return err;
906 	}
907 
908 	tuncfg.encap_type = 1;
909 	tuncfg.encap_rcv = sctp_udp_rcv;
910 	tuncfg.encap_err_lookup = sctp_udp_err_lookup;
911 	setup_udp_tunnel_sock(net, sock, &tuncfg);
912 	net->sctp.udp6_sock = sock->sk;
913 #endif
914 
915 	return 0;
916 }
917 
918 void sctp_udp_sock_stop(struct net *net)
919 {
920 	if (net->sctp.udp4_sock) {
921 		udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
922 		net->sctp.udp4_sock = NULL;
923 	}
924 	if (net->sctp.udp6_sock) {
925 		udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket);
926 		net->sctp.udp6_sock = NULL;
927 	}
928 }
929 
930 /* Register address family specific functions. */
931 int sctp_register_af(struct sctp_af *af)
932 {
933 	switch (af->sa_family) {
934 	case AF_INET:
935 		if (sctp_af_v4_specific)
936 			return 0;
937 		sctp_af_v4_specific = af;
938 		break;
939 	case AF_INET6:
940 		if (sctp_af_v6_specific)
941 			return 0;
942 		sctp_af_v6_specific = af;
943 		break;
944 	default:
945 		return 0;
946 	}
947 
948 	INIT_LIST_HEAD(&af->list);
949 	list_add_tail(&af->list, &sctp_address_families);
950 	return 1;
951 }
952 
953 /* Get the table of functions for manipulating a particular address
954  * family.
955  */
956 struct sctp_af *sctp_get_af_specific(sa_family_t family)
957 {
958 	switch (family) {
959 	case AF_INET:
960 		return sctp_af_v4_specific;
961 	case AF_INET6:
962 		return sctp_af_v6_specific;
963 	default:
964 		return NULL;
965 	}
966 }
967 
968 /* Common code to initialize a AF_INET msg_name. */
969 static void sctp_inet_msgname(char *msgname, int *addr_len)
970 {
971 	struct sockaddr_in *sin;
972 
973 	sin = (struct sockaddr_in *)msgname;
974 	*addr_len = sizeof(struct sockaddr_in);
975 	sin->sin_family = AF_INET;
976 	memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
977 }
978 
979 /* Copy the primary address of the peer primary address as the msg_name. */
980 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
981 				    int *addr_len)
982 {
983 	struct sockaddr_in *sin, *sinfrom;
984 
985 	if (msgname) {
986 		struct sctp_association *asoc;
987 
988 		asoc = event->asoc;
989 		sctp_inet_msgname(msgname, addr_len);
990 		sin = (struct sockaddr_in *)msgname;
991 		sinfrom = &asoc->peer.primary_addr.v4;
992 		sin->sin_port = htons(asoc->peer.port);
993 		sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
994 	}
995 }
996 
997 /* Initialize and copy out a msgname from an inbound skb. */
998 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
999 {
1000 	if (msgname) {
1001 		struct sctphdr *sh = sctp_hdr(skb);
1002 		struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
1003 
1004 		sctp_inet_msgname(msgname, len);
1005 		sin->sin_port = sh->source;
1006 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1007 	}
1008 }
1009 
1010 /* Do we support this AF? */
1011 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
1012 {
1013 	/* PF_INET only supports AF_INET addresses. */
1014 	return AF_INET == family;
1015 }
1016 
1017 /* Address matching with wildcards allowed. */
1018 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
1019 			      const union sctp_addr *addr2,
1020 			      struct sctp_sock *opt)
1021 {
1022 	/* PF_INET only supports AF_INET addresses. */
1023 	if (addr1->sa.sa_family != addr2->sa.sa_family)
1024 		return 0;
1025 	if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
1026 	    htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
1027 		return 1;
1028 	if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
1029 		return 1;
1030 
1031 	return 0;
1032 }
1033 
1034 /* Verify that provided sockaddr looks bindable.  Common verification has
1035  * already been taken care of.
1036  */
1037 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
1038 {
1039 	return sctp_v4_available(addr, opt);
1040 }
1041 
1042 /* Verify that sockaddr looks sendable.  Common verification has already
1043  * been taken care of.
1044  */
1045 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
1046 {
1047 	return 1;
1048 }
1049 
1050 /* Fill in Supported Address Type information for INIT and INIT-ACK
1051  * chunks.  Returns number of addresses supported.
1052  */
1053 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
1054 				     __be16 *types)
1055 {
1056 	types[0] = SCTP_PARAM_IPV4_ADDRESS;
1057 	return 1;
1058 }
1059 
1060 /* Wrapper routine that calls the ip transmit routine. */
1061 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t)
1062 {
1063 	struct dst_entry *dst = dst_clone(t->dst);
1064 	struct flowi4 *fl4 = &t->fl.u.ip4;
1065 	struct sock *sk = skb->sk;
1066 	struct inet_sock *inet = inet_sk(sk);
1067 	__u8 dscp = inet->tos;
1068 	__be16 df = 0;
1069 
1070 	pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
1071 		 skb->len, &fl4->saddr, &fl4->daddr);
1072 
1073 	if (t->dscp & SCTP_DSCP_SET_MASK)
1074 		dscp = t->dscp & SCTP_DSCP_VAL_MASK;
1075 
1076 	inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO
1077 							   : IP_PMTUDISC_DONT;
1078 	SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS);
1079 
1080 	if (!t->encap_port || !sctp_sk(sk)->udp_port) {
1081 		skb_dst_set(skb, dst);
1082 		return __ip_queue_xmit(sk, skb, &t->fl, dscp);
1083 	}
1084 
1085 	if (skb_is_gso(skb))
1086 		skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
1087 
1088 	if (ip_dont_fragment(sk, dst) && !skb->ignore_df)
1089 		df = htons(IP_DF);
1090 
1091 	skb->encapsulation = 1;
1092 	skb_reset_inner_mac_header(skb);
1093 	skb_reset_inner_transport_header(skb);
1094 	skb_set_inner_ipproto(skb, IPPROTO_SCTP);
1095 	udp_tunnel_xmit_skb((struct rtable *)dst, sk, skb, fl4->saddr,
1096 			    fl4->daddr, dscp, ip4_dst_hoplimit(dst), df,
1097 			    sctp_sk(sk)->udp_port, t->encap_port, false, false);
1098 	return 0;
1099 }
1100 
1101 static struct sctp_af sctp_af_inet;
1102 
1103 static struct sctp_pf sctp_pf_inet = {
1104 	.event_msgname = sctp_inet_event_msgname,
1105 	.skb_msgname   = sctp_inet_skb_msgname,
1106 	.af_supported  = sctp_inet_af_supported,
1107 	.cmp_addr      = sctp_inet_cmp_addr,
1108 	.bind_verify   = sctp_inet_bind_verify,
1109 	.send_verify   = sctp_inet_send_verify,
1110 	.supported_addrs = sctp_inet_supported_addrs,
1111 	.create_accept_sk = sctp_v4_create_accept_sk,
1112 	.addr_to_user  = sctp_v4_addr_to_user,
1113 	.to_sk_saddr   = sctp_v4_to_sk_saddr,
1114 	.to_sk_daddr   = sctp_v4_to_sk_daddr,
1115 	.copy_ip_options = sctp_v4_copy_ip_options,
1116 	.af            = &sctp_af_inet
1117 };
1118 
1119 /* Notifier for inetaddr addition/deletion events.  */
1120 static struct notifier_block sctp_inetaddr_notifier = {
1121 	.notifier_call = sctp_inetaddr_event,
1122 };
1123 
1124 /* Socket operations.  */
1125 static const struct proto_ops inet_seqpacket_ops = {
1126 	.family		   = PF_INET,
1127 	.owner		   = THIS_MODULE,
1128 	.release	   = inet_release,	/* Needs to be wrapped... */
1129 	.bind		   = inet_bind,
1130 	.connect	   = sctp_inet_connect,
1131 	.socketpair	   = sock_no_socketpair,
1132 	.accept		   = inet_accept,
1133 	.getname	   = inet_getname,	/* Semantics are different.  */
1134 	.poll		   = sctp_poll,
1135 	.ioctl		   = inet_ioctl,
1136 	.gettstamp	   = sock_gettstamp,
1137 	.listen		   = sctp_inet_listen,
1138 	.shutdown	   = inet_shutdown,	/* Looks harmless.  */
1139 	.setsockopt	   = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1140 	.getsockopt	   = sock_common_getsockopt,
1141 	.sendmsg	   = inet_sendmsg,
1142 	.recvmsg	   = inet_recvmsg,
1143 	.mmap		   = sock_no_mmap,
1144 	.sendpage	   = sock_no_sendpage,
1145 };
1146 
1147 /* Registration with AF_INET family.  */
1148 static struct inet_protosw sctp_seqpacket_protosw = {
1149 	.type       = SOCK_SEQPACKET,
1150 	.protocol   = IPPROTO_SCTP,
1151 	.prot       = &sctp_prot,
1152 	.ops        = &inet_seqpacket_ops,
1153 	.flags      = SCTP_PROTOSW_FLAG
1154 };
1155 static struct inet_protosw sctp_stream_protosw = {
1156 	.type       = SOCK_STREAM,
1157 	.protocol   = IPPROTO_SCTP,
1158 	.prot       = &sctp_prot,
1159 	.ops        = &inet_seqpacket_ops,
1160 	.flags      = SCTP_PROTOSW_FLAG
1161 };
1162 
1163 static int sctp4_rcv(struct sk_buff *skb)
1164 {
1165 	SCTP_INPUT_CB(skb)->encap_port = 0;
1166 	return sctp_rcv(skb);
1167 }
1168 
1169 /* Register with IP layer.  */
1170 static const struct net_protocol sctp_protocol = {
1171 	.handler     = sctp4_rcv,
1172 	.err_handler = sctp_v4_err,
1173 	.no_policy   = 1,
1174 	.icmp_strict_tag_validation = 1,
1175 };
1176 
1177 /* IPv4 address related functions.  */
1178 static struct sctp_af sctp_af_inet = {
1179 	.sa_family	   = AF_INET,
1180 	.sctp_xmit	   = sctp_v4_xmit,
1181 	.setsockopt	   = ip_setsockopt,
1182 	.getsockopt	   = ip_getsockopt,
1183 	.get_dst	   = sctp_v4_get_dst,
1184 	.get_saddr	   = sctp_v4_get_saddr,
1185 	.copy_addrlist	   = sctp_v4_copy_addrlist,
1186 	.from_skb	   = sctp_v4_from_skb,
1187 	.from_sk	   = sctp_v4_from_sk,
1188 	.from_addr_param   = sctp_v4_from_addr_param,
1189 	.to_addr_param	   = sctp_v4_to_addr_param,
1190 	.cmp_addr	   = sctp_v4_cmp_addr,
1191 	.addr_valid	   = sctp_v4_addr_valid,
1192 	.inaddr_any	   = sctp_v4_inaddr_any,
1193 	.is_any		   = sctp_v4_is_any,
1194 	.available	   = sctp_v4_available,
1195 	.scope		   = sctp_v4_scope,
1196 	.skb_iif	   = sctp_v4_skb_iif,
1197 	.is_ce		   = sctp_v4_is_ce,
1198 	.seq_dump_addr	   = sctp_v4_seq_dump_addr,
1199 	.ecn_capable	   = sctp_v4_ecn_capable,
1200 	.net_header_len	   = sizeof(struct iphdr),
1201 	.sockaddr_len	   = sizeof(struct sockaddr_in),
1202 	.ip_options_len	   = sctp_v4_ip_options_len,
1203 };
1204 
1205 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1206 {
1207 	switch (family) {
1208 	case PF_INET:
1209 		return sctp_pf_inet_specific;
1210 	case PF_INET6:
1211 		return sctp_pf_inet6_specific;
1212 	default:
1213 		return NULL;
1214 	}
1215 }
1216 
1217 /* Register the PF specific function table.  */
1218 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1219 {
1220 	switch (family) {
1221 	case PF_INET:
1222 		if (sctp_pf_inet_specific)
1223 			return 0;
1224 		sctp_pf_inet_specific = pf;
1225 		break;
1226 	case PF_INET6:
1227 		if (sctp_pf_inet6_specific)
1228 			return 0;
1229 		sctp_pf_inet6_specific = pf;
1230 		break;
1231 	default:
1232 		return 0;
1233 	}
1234 	return 1;
1235 }
1236 
1237 static inline int init_sctp_mibs(struct net *net)
1238 {
1239 	net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1240 	if (!net->sctp.sctp_statistics)
1241 		return -ENOMEM;
1242 	return 0;
1243 }
1244 
1245 static inline void cleanup_sctp_mibs(struct net *net)
1246 {
1247 	free_percpu(net->sctp.sctp_statistics);
1248 }
1249 
1250 static void sctp_v4_pf_init(void)
1251 {
1252 	/* Initialize the SCTP specific PF functions. */
1253 	sctp_register_pf(&sctp_pf_inet, PF_INET);
1254 	sctp_register_af(&sctp_af_inet);
1255 }
1256 
1257 static void sctp_v4_pf_exit(void)
1258 {
1259 	list_del(&sctp_af_inet.list);
1260 }
1261 
1262 static int sctp_v4_protosw_init(void)
1263 {
1264 	int rc;
1265 
1266 	rc = proto_register(&sctp_prot, 1);
1267 	if (rc)
1268 		return rc;
1269 
1270 	/* Register SCTP(UDP and TCP style) with socket layer.  */
1271 	inet_register_protosw(&sctp_seqpacket_protosw);
1272 	inet_register_protosw(&sctp_stream_protosw);
1273 
1274 	return 0;
1275 }
1276 
1277 static void sctp_v4_protosw_exit(void)
1278 {
1279 	inet_unregister_protosw(&sctp_stream_protosw);
1280 	inet_unregister_protosw(&sctp_seqpacket_protosw);
1281 	proto_unregister(&sctp_prot);
1282 }
1283 
1284 static int sctp_v4_add_protocol(void)
1285 {
1286 	/* Register notifier for inet address additions/deletions. */
1287 	register_inetaddr_notifier(&sctp_inetaddr_notifier);
1288 
1289 	/* Register SCTP with inet layer.  */
1290 	if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1291 		return -EAGAIN;
1292 
1293 	return 0;
1294 }
1295 
1296 static void sctp_v4_del_protocol(void)
1297 {
1298 	inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1299 	unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1300 }
1301 
1302 static int __net_init sctp_defaults_init(struct net *net)
1303 {
1304 	int status;
1305 
1306 	/*
1307 	 * 14. Suggested SCTP Protocol Parameter Values
1308 	 */
1309 	/* The following protocol parameters are RECOMMENDED:  */
1310 	/* RTO.Initial              - 3  seconds */
1311 	net->sctp.rto_initial			= SCTP_RTO_INITIAL;
1312 	/* RTO.Min                  - 1  second */
1313 	net->sctp.rto_min	 		= SCTP_RTO_MIN;
1314 	/* RTO.Max                 -  60 seconds */
1315 	net->sctp.rto_max 			= SCTP_RTO_MAX;
1316 	/* RTO.Alpha                - 1/8 */
1317 	net->sctp.rto_alpha			= SCTP_RTO_ALPHA;
1318 	/* RTO.Beta                 - 1/4 */
1319 	net->sctp.rto_beta			= SCTP_RTO_BETA;
1320 
1321 	/* Valid.Cookie.Life        - 60  seconds */
1322 	net->sctp.valid_cookie_life		= SCTP_DEFAULT_COOKIE_LIFE;
1323 
1324 	/* Whether Cookie Preservative is enabled(1) or not(0) */
1325 	net->sctp.cookie_preserve_enable 	= 1;
1326 
1327 	/* Default sctp sockets to use md5 as their hmac alg */
1328 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1329 	net->sctp.sctp_hmac_alg			= "md5";
1330 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1331 	net->sctp.sctp_hmac_alg			= "sha1";
1332 #else
1333 	net->sctp.sctp_hmac_alg			= NULL;
1334 #endif
1335 
1336 	/* Max.Burst		    - 4 */
1337 	net->sctp.max_burst			= SCTP_DEFAULT_MAX_BURST;
1338 
1339 	/* Disable of Primary Path Switchover by default */
1340 	net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1341 
1342 	/* Enable pf state by default */
1343 	net->sctp.pf_enable = 1;
1344 
1345 	/* Ignore pf exposure feature by default */
1346 	net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1347 
1348 	/* Association.Max.Retrans  - 10 attempts
1349 	 * Path.Max.Retrans         - 5  attempts (per destination address)
1350 	 * Max.Init.Retransmits     - 8  attempts
1351 	 */
1352 	net->sctp.max_retrans_association	= 10;
1353 	net->sctp.max_retrans_path		= 5;
1354 	net->sctp.max_retrans_init		= 8;
1355 
1356 	/* Sendbuffer growth	    - do per-socket accounting */
1357 	net->sctp.sndbuf_policy			= 0;
1358 
1359 	/* Rcvbuffer growth	    - do per-socket accounting */
1360 	net->sctp.rcvbuf_policy			= 0;
1361 
1362 	/* HB.interval              - 30 seconds */
1363 	net->sctp.hb_interval			= SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1364 
1365 	/* delayed SACK timeout */
1366 	net->sctp.sack_timeout			= SCTP_DEFAULT_TIMEOUT_SACK;
1367 
1368 	/* Disable ADDIP by default. */
1369 	net->sctp.addip_enable = 0;
1370 	net->sctp.addip_noauth = 0;
1371 	net->sctp.default_auto_asconf = 0;
1372 
1373 	/* Enable PR-SCTP by default. */
1374 	net->sctp.prsctp_enable = 1;
1375 
1376 	/* Disable RECONF by default. */
1377 	net->sctp.reconf_enable = 0;
1378 
1379 	/* Disable AUTH by default. */
1380 	net->sctp.auth_enable = 0;
1381 
1382 	/* Enable ECN by default. */
1383 	net->sctp.ecn_enable = 1;
1384 
1385 	/* Set UDP tunneling listening port to 0 by default */
1386 	net->sctp.udp_port = 0;
1387 
1388 	/* Set remote encap port to 0 by default */
1389 	net->sctp.encap_port = 0;
1390 
1391 	/* Set SCOPE policy to enabled */
1392 	net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1393 
1394 	/* Set the default rwnd update threshold */
1395 	net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1396 
1397 	/* Initialize maximum autoclose timeout. */
1398 	net->sctp.max_autoclose		= INT_MAX / HZ;
1399 
1400 	status = sctp_sysctl_net_register(net);
1401 	if (status)
1402 		goto err_sysctl_register;
1403 
1404 	/* Allocate and initialise sctp mibs.  */
1405 	status = init_sctp_mibs(net);
1406 	if (status)
1407 		goto err_init_mibs;
1408 
1409 #ifdef CONFIG_PROC_FS
1410 	/* Initialize proc fs directory.  */
1411 	status = sctp_proc_init(net);
1412 	if (status)
1413 		goto err_init_proc;
1414 #endif
1415 
1416 	sctp_dbg_objcnt_init(net);
1417 
1418 	/* Initialize the local address list. */
1419 	INIT_LIST_HEAD(&net->sctp.local_addr_list);
1420 	spin_lock_init(&net->sctp.local_addr_lock);
1421 	sctp_get_local_addr_list(net);
1422 
1423 	/* Initialize the address event list */
1424 	INIT_LIST_HEAD(&net->sctp.addr_waitq);
1425 	INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1426 	spin_lock_init(&net->sctp.addr_wq_lock);
1427 	net->sctp.addr_wq_timer.expires = 0;
1428 	timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1429 
1430 	return 0;
1431 
1432 #ifdef CONFIG_PROC_FS
1433 err_init_proc:
1434 	cleanup_sctp_mibs(net);
1435 #endif
1436 err_init_mibs:
1437 	sctp_sysctl_net_unregister(net);
1438 err_sysctl_register:
1439 	return status;
1440 }
1441 
1442 static void __net_exit sctp_defaults_exit(struct net *net)
1443 {
1444 	/* Free the local address list */
1445 	sctp_free_addr_wq(net);
1446 	sctp_free_local_addr_list(net);
1447 
1448 #ifdef CONFIG_PROC_FS
1449 	remove_proc_subtree("sctp", net->proc_net);
1450 	net->sctp.proc_net_sctp = NULL;
1451 #endif
1452 	cleanup_sctp_mibs(net);
1453 	sctp_sysctl_net_unregister(net);
1454 }
1455 
1456 static struct pernet_operations sctp_defaults_ops = {
1457 	.init = sctp_defaults_init,
1458 	.exit = sctp_defaults_exit,
1459 };
1460 
1461 static int __net_init sctp_ctrlsock_init(struct net *net)
1462 {
1463 	int status;
1464 
1465 	/* Initialize the control inode/socket for handling OOTB packets.  */
1466 	status = sctp_ctl_sock_init(net);
1467 	if (status)
1468 		pr_err("Failed to initialize the SCTP control sock\n");
1469 
1470 	return status;
1471 }
1472 
1473 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1474 {
1475 	/* Free the control endpoint.  */
1476 	inet_ctl_sock_destroy(net->sctp.ctl_sock);
1477 }
1478 
1479 static struct pernet_operations sctp_ctrlsock_ops = {
1480 	.init = sctp_ctrlsock_init,
1481 	.exit = sctp_ctrlsock_exit,
1482 };
1483 
1484 /* Initialize the universe into something sensible.  */
1485 static __init int sctp_init(void)
1486 {
1487 	unsigned long nr_pages = totalram_pages();
1488 	unsigned long limit;
1489 	unsigned long goal;
1490 	int max_entry_order;
1491 	int num_entries;
1492 	int max_share;
1493 	int status;
1494 	int order;
1495 	int i;
1496 
1497 	sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1498 
1499 	/* Allocate bind_bucket and chunk caches. */
1500 	status = -ENOBUFS;
1501 	sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1502 					       sizeof(struct sctp_bind_bucket),
1503 					       0, SLAB_HWCACHE_ALIGN,
1504 					       NULL);
1505 	if (!sctp_bucket_cachep)
1506 		goto out;
1507 
1508 	sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1509 					       sizeof(struct sctp_chunk),
1510 					       0, SLAB_HWCACHE_ALIGN,
1511 					       NULL);
1512 	if (!sctp_chunk_cachep)
1513 		goto err_chunk_cachep;
1514 
1515 	status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1516 	if (status)
1517 		goto err_percpu_counter_init;
1518 
1519 	/* Implementation specific variables. */
1520 
1521 	/* Initialize default stream count setup information. */
1522 	sctp_max_instreams    		= SCTP_DEFAULT_INSTREAMS;
1523 	sctp_max_outstreams   		= SCTP_DEFAULT_OUTSTREAMS;
1524 
1525 	/* Initialize handle used for association ids. */
1526 	idr_init(&sctp_assocs_id);
1527 
1528 	limit = nr_free_buffer_pages() / 8;
1529 	limit = max(limit, 128UL);
1530 	sysctl_sctp_mem[0] = limit / 4 * 3;
1531 	sysctl_sctp_mem[1] = limit;
1532 	sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1533 
1534 	/* Set per-socket limits to no more than 1/128 the pressure threshold*/
1535 	limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1536 	max_share = min(4UL*1024*1024, limit);
1537 
1538 	sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1539 	sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1540 	sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1541 
1542 	sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1543 	sysctl_sctp_wmem[1] = 16*1024;
1544 	sysctl_sctp_wmem[2] = max(64*1024, max_share);
1545 
1546 	/* Size and allocate the association hash table.
1547 	 * The methodology is similar to that of the tcp hash tables.
1548 	 * Though not identical.  Start by getting a goal size
1549 	 */
1550 	if (nr_pages >= (128 * 1024))
1551 		goal = nr_pages >> (22 - PAGE_SHIFT);
1552 	else
1553 		goal = nr_pages >> (24 - PAGE_SHIFT);
1554 
1555 	/* Then compute the page order for said goal */
1556 	order = get_order(goal);
1557 
1558 	/* Now compute the required page order for the maximum sized table we
1559 	 * want to create
1560 	 */
1561 	max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1562 				    sizeof(struct sctp_bind_hashbucket));
1563 
1564 	/* Limit the page order by that maximum hash table size */
1565 	order = min(order, max_entry_order);
1566 
1567 	/* Allocate and initialize the endpoint hash table.  */
1568 	sctp_ep_hashsize = 64;
1569 	sctp_ep_hashtable =
1570 		kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1571 	if (!sctp_ep_hashtable) {
1572 		pr_err("Failed endpoint_hash alloc\n");
1573 		status = -ENOMEM;
1574 		goto err_ehash_alloc;
1575 	}
1576 	for (i = 0; i < sctp_ep_hashsize; i++) {
1577 		rwlock_init(&sctp_ep_hashtable[i].lock);
1578 		INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1579 	}
1580 
1581 	/* Allocate and initialize the SCTP port hash table.
1582 	 * Note that order is initalized to start at the max sized
1583 	 * table we want to support.  If we can't get that many pages
1584 	 * reduce the order and try again
1585 	 */
1586 	do {
1587 		sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1588 			__get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1589 	} while (!sctp_port_hashtable && --order > 0);
1590 
1591 	if (!sctp_port_hashtable) {
1592 		pr_err("Failed bind hash alloc\n");
1593 		status = -ENOMEM;
1594 		goto err_bhash_alloc;
1595 	}
1596 
1597 	/* Now compute the number of entries that will fit in the
1598 	 * port hash space we allocated
1599 	 */
1600 	num_entries = (1UL << order) * PAGE_SIZE /
1601 		      sizeof(struct sctp_bind_hashbucket);
1602 
1603 	/* And finish by rounding it down to the nearest power of two.
1604 	 * This wastes some memory of course, but it's needed because
1605 	 * the hash function operates based on the assumption that
1606 	 * the number of entries is a power of two.
1607 	 */
1608 	sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1609 
1610 	for (i = 0; i < sctp_port_hashsize; i++) {
1611 		spin_lock_init(&sctp_port_hashtable[i].lock);
1612 		INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1613 	}
1614 
1615 	status = sctp_transport_hashtable_init();
1616 	if (status)
1617 		goto err_thash_alloc;
1618 
1619 	pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1620 		num_entries);
1621 
1622 	sctp_sysctl_register();
1623 
1624 	INIT_LIST_HEAD(&sctp_address_families);
1625 	sctp_v4_pf_init();
1626 	sctp_v6_pf_init();
1627 	sctp_sched_ops_init();
1628 
1629 	status = register_pernet_subsys(&sctp_defaults_ops);
1630 	if (status)
1631 		goto err_register_defaults;
1632 
1633 	status = sctp_v4_protosw_init();
1634 	if (status)
1635 		goto err_protosw_init;
1636 
1637 	status = sctp_v6_protosw_init();
1638 	if (status)
1639 		goto err_v6_protosw_init;
1640 
1641 	status = register_pernet_subsys(&sctp_ctrlsock_ops);
1642 	if (status)
1643 		goto err_register_ctrlsock;
1644 
1645 	status = sctp_v4_add_protocol();
1646 	if (status)
1647 		goto err_add_protocol;
1648 
1649 	/* Register SCTP with inet6 layer.  */
1650 	status = sctp_v6_add_protocol();
1651 	if (status)
1652 		goto err_v6_add_protocol;
1653 
1654 	if (sctp_offload_init() < 0)
1655 		pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1656 
1657 out:
1658 	return status;
1659 err_v6_add_protocol:
1660 	sctp_v4_del_protocol();
1661 err_add_protocol:
1662 	unregister_pernet_subsys(&sctp_ctrlsock_ops);
1663 err_register_ctrlsock:
1664 	sctp_v6_protosw_exit();
1665 err_v6_protosw_init:
1666 	sctp_v4_protosw_exit();
1667 err_protosw_init:
1668 	unregister_pernet_subsys(&sctp_defaults_ops);
1669 err_register_defaults:
1670 	sctp_v4_pf_exit();
1671 	sctp_v6_pf_exit();
1672 	sctp_sysctl_unregister();
1673 	free_pages((unsigned long)sctp_port_hashtable,
1674 		   get_order(sctp_port_hashsize *
1675 			     sizeof(struct sctp_bind_hashbucket)));
1676 err_bhash_alloc:
1677 	sctp_transport_hashtable_destroy();
1678 err_thash_alloc:
1679 	kfree(sctp_ep_hashtable);
1680 err_ehash_alloc:
1681 	percpu_counter_destroy(&sctp_sockets_allocated);
1682 err_percpu_counter_init:
1683 	kmem_cache_destroy(sctp_chunk_cachep);
1684 err_chunk_cachep:
1685 	kmem_cache_destroy(sctp_bucket_cachep);
1686 	goto out;
1687 }
1688 
1689 /* Exit handler for the SCTP protocol.  */
1690 static __exit void sctp_exit(void)
1691 {
1692 	/* BUG.  This should probably do something useful like clean
1693 	 * up all the remaining associations and all that memory.
1694 	 */
1695 
1696 	/* Unregister with inet6/inet layers. */
1697 	sctp_v6_del_protocol();
1698 	sctp_v4_del_protocol();
1699 
1700 	unregister_pernet_subsys(&sctp_ctrlsock_ops);
1701 
1702 	/* Free protosw registrations */
1703 	sctp_v6_protosw_exit();
1704 	sctp_v4_protosw_exit();
1705 
1706 	unregister_pernet_subsys(&sctp_defaults_ops);
1707 
1708 	/* Unregister with socket layer. */
1709 	sctp_v6_pf_exit();
1710 	sctp_v4_pf_exit();
1711 
1712 	sctp_sysctl_unregister();
1713 
1714 	free_pages((unsigned long)sctp_port_hashtable,
1715 		   get_order(sctp_port_hashsize *
1716 			     sizeof(struct sctp_bind_hashbucket)));
1717 	kfree(sctp_ep_hashtable);
1718 	sctp_transport_hashtable_destroy();
1719 
1720 	percpu_counter_destroy(&sctp_sockets_allocated);
1721 
1722 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1723 
1724 	kmem_cache_destroy(sctp_chunk_cachep);
1725 	kmem_cache_destroy(sctp_bucket_cachep);
1726 }
1727 
1728 module_init(sctp_init);
1729 module_exit(sctp_exit);
1730 
1731 /*
1732  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1733  */
1734 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1735 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1736 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1737 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1738 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1739 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1740 MODULE_LICENSE("GPL");
1741