1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* SCTP kernel implementation
3 * (C) Copyright IBM Corp. 2001, 2004
4 * Copyright (c) 1999-2000 Cisco, Inc.
5 * Copyright (c) 1999-2001 Motorola, Inc.
6 * Copyright (c) 2001 Intel Corp.
7 * Copyright (c) 2001 Nokia, Inc.
8 * Copyright (c) 2001 La Monte H.P. Yarroll
9 *
10 * This file is part of the SCTP kernel implementation
11 *
12 * Initialization/cleanup for SCTP protocol support.
13 *
14 * Please send any bug reports or fixes you make to the
15 * email address(es):
16 * lksctp developers <linux-sctp@vger.kernel.org>
17 *
18 * Written or modified by:
19 * La Monte H.P. Yarroll <piggy@acm.org>
20 * Karl Knutson <karl@athena.chicago.il.us>
21 * Jon Grimm <jgrimm@us.ibm.com>
22 * Sridhar Samudrala <sri@us.ibm.com>
23 * Daisy Chang <daisyc@us.ibm.com>
24 * Ardelle Fan <ardelle.fan@intel.com>
25 */
26
27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
28
29 #include <linux/module.h>
30 #include <linux/init.h>
31 #include <linux/netdevice.h>
32 #include <linux/inetdevice.h>
33 #include <linux/seq_file.h>
34 #include <linux/memblock.h>
35 #include <linux/highmem.h>
36 #include <linux/slab.h>
37 #include <net/net_namespace.h>
38 #include <net/protocol.h>
39 #include <net/ip.h>
40 #include <net/ipv6.h>
41 #include <net/route.h>
42 #include <net/sctp/sctp.h>
43 #include <net/addrconf.h>
44 #include <net/inet_common.h>
45 #include <net/inet_ecn.h>
46 #include <net/udp_tunnel.h>
47 #include <net/inet_dscp.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 */
sctp_v4_copy_addrlist(struct list_head * addrlist,struct net_device * dev)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 */
sctp_get_local_addr_list(struct net * net)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. */
sctp_free_local_addr_list(struct net * net)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'. */
sctp_copy_local_addr_list(struct net * net,struct sctp_bind_addr * bp,enum sctp_scope scope,gfp_t gfp,int copy_flags)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 */
sctp_v4_copy_ip_options(struct sock * sk,struct sock * newsk)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 */
sctp_v4_ip_options_len(struct sock * sk)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. */
sctp_v4_from_skb(union sctp_addr * addr,struct sk_buff * skb,int is_saddr)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. */
sctp_v4_from_sk(union sctp_addr * addr,struct sock * sk)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. */
sctp_v4_to_sk_saddr(union sctp_addr * addr,struct sock * sk)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. */
sctp_v4_to_sk_daddr(union sctp_addr * addr,struct sock * sk)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. */
sctp_v4_from_addr_param(union sctp_addr * addr,union sctp_addr_param * param,__be16 port,int iif)257 static bool sctp_v4_from_addr_param(union sctp_addr *addr,
258 union sctp_addr_param *param,
259 __be16 port, int iif)
260 {
261 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param))
262 return false;
263
264 addr->v4.sin_family = AF_INET;
265 addr->v4.sin_port = port;
266 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
267 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
268
269 return true;
270 }
271
272 /* Initialize an address parameter from a sctp_addr and return the length
273 * of the address parameter.
274 */
sctp_v4_to_addr_param(const union sctp_addr * addr,union sctp_addr_param * param)275 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
276 union sctp_addr_param *param)
277 {
278 int length = sizeof(struct sctp_ipv4addr_param);
279
280 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
281 param->v4.param_hdr.length = htons(length);
282 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
283
284 return length;
285 }
286
287 /* Initialize a sctp_addr from a dst_entry. */
sctp_v4_dst_saddr(union sctp_addr * saddr,struct flowi4 * fl4,__be16 port)288 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
289 __be16 port)
290 {
291 saddr->v4.sin_family = AF_INET;
292 saddr->v4.sin_port = port;
293 saddr->v4.sin_addr.s_addr = fl4->saddr;
294 memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
295 }
296
297 /* Compare two addresses exactly. */
sctp_v4_cmp_addr(const union sctp_addr * addr1,const union sctp_addr * addr2)298 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
299 const union sctp_addr *addr2)
300 {
301 if (addr1->sa.sa_family != addr2->sa.sa_family)
302 return 0;
303 if (addr1->v4.sin_port != addr2->v4.sin_port)
304 return 0;
305 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
306 return 0;
307
308 return 1;
309 }
310
311 /* Initialize addr struct to INADDR_ANY. */
sctp_v4_inaddr_any(union sctp_addr * addr,__be16 port)312 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
313 {
314 addr->v4.sin_family = AF_INET;
315 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
316 addr->v4.sin_port = port;
317 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
318 }
319
320 /* Is this a wildcard address? */
sctp_v4_is_any(const union sctp_addr * addr)321 static int sctp_v4_is_any(const union sctp_addr *addr)
322 {
323 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
324 }
325
326 /* This function checks if the address is a valid address to be used for
327 * SCTP binding.
328 *
329 * Output:
330 * Return 0 - If the address is a non-unicast or an illegal address.
331 * Return 1 - If the address is a unicast.
332 */
sctp_v4_addr_valid(union sctp_addr * addr,struct sctp_sock * sp,const struct sk_buff * skb)333 static int sctp_v4_addr_valid(union sctp_addr *addr,
334 struct sctp_sock *sp,
335 const struct sk_buff *skb)
336 {
337 /* IPv4 addresses not allowed */
338 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
339 return 0;
340
341 /* Is this a non-unicast address or a unusable SCTP address? */
342 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
343 return 0;
344
345 /* Is this a broadcast address? */
346 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
347 return 0;
348
349 return 1;
350 }
351
352 /* Should this be available for binding? */
sctp_v4_available(union sctp_addr * addr,struct sctp_sock * sp)353 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
354 {
355 struct sock *sk = &sp->inet.sk;
356 struct net *net = sock_net(sk);
357 int tb_id = RT_TABLE_LOCAL;
358 int ret;
359
360 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ?: tb_id;
361 ret = inet_addr_type_table(net, addr->v4.sin_addr.s_addr, tb_id);
362 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
363 ret != RTN_LOCAL &&
364 !inet_test_bit(FREEBIND, sk) &&
365 !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind))
366 return 0;
367
368 if (ipv6_only_sock(sctp_opt2sk(sp)))
369 return 0;
370
371 return 1;
372 }
373
374 /* Checking the loopback, private and other address scopes as defined in
375 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
376 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
377 *
378 * Level 0 - unusable SCTP addresses
379 * Level 1 - loopback address
380 * Level 2 - link-local addresses
381 * Level 3 - private addresses.
382 * Level 4 - global addresses
383 * For INIT and INIT-ACK address list, let L be the level of
384 * requested destination address, sender and receiver
385 * SHOULD include all of its addresses with level greater
386 * than or equal to L.
387 *
388 * IPv4 scoping can be controlled through sysctl option
389 * net.sctp.addr_scope_policy
390 */
sctp_v4_scope(union sctp_addr * addr)391 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
392 {
393 enum sctp_scope retval;
394
395 /* Check for unusable SCTP addresses. */
396 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
397 retval = SCTP_SCOPE_UNUSABLE;
398 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
399 retval = SCTP_SCOPE_LOOPBACK;
400 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
401 retval = SCTP_SCOPE_LINK;
402 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
403 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
404 ipv4_is_private_192(addr->v4.sin_addr.s_addr) ||
405 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) {
406 retval = SCTP_SCOPE_PRIVATE;
407 } else {
408 retval = SCTP_SCOPE_GLOBAL;
409 }
410
411 return retval;
412 }
413
414 /* Returns a valid dst cache entry for the given source and destination ip
415 * addresses. If an association is passed, trys to get a dst entry with a
416 * source address that matches an address in the bind address list.
417 */
sctp_v4_get_dst(struct sctp_transport * t,union sctp_addr * saddr,struct flowi * fl,struct sock * sk)418 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
419 struct flowi *fl, struct sock *sk)
420 {
421 struct sctp_association *asoc = t->asoc;
422 struct rtable *rt;
423 struct flowi _fl;
424 struct flowi4 *fl4 = &_fl.u.ip4;
425 struct sctp_bind_addr *bp;
426 struct sctp_sockaddr_entry *laddr;
427 struct dst_entry *dst = NULL;
428 union sctp_addr *daddr = &t->ipaddr;
429 union sctp_addr dst_saddr;
430 u8 tos = READ_ONCE(inet_sk(sk)->tos);
431
432 if (t->dscp & SCTP_DSCP_SET_MASK)
433 tos = t->dscp & SCTP_DSCP_VAL_MASK;
434 memset(&_fl, 0x0, sizeof(_fl));
435 fl4->daddr = daddr->v4.sin_addr.s_addr;
436 fl4->fl4_dport = daddr->v4.sin_port;
437 fl4->flowi4_proto = IPPROTO_SCTP;
438 if (asoc) {
439 fl4->flowi4_tos = tos & INET_DSCP_MASK;
440 fl4->flowi4_scope = ip_sock_rt_scope(asoc->base.sk);
441 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
442 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
443 }
444 if (saddr) {
445 fl4->saddr = saddr->v4.sin_addr.s_addr;
446 if (!fl4->fl4_sport)
447 fl4->fl4_sport = saddr->v4.sin_port;
448 }
449
450 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
451 &fl4->saddr);
452
453 rt = ip_route_output_key(sock_net(sk), fl4);
454 if (!IS_ERR(rt)) {
455 dst = &rt->dst;
456 t->dst = dst;
457 memcpy(fl, &_fl, sizeof(_fl));
458 }
459
460 /* If there is no association or if a source address is passed, no
461 * more validation is required.
462 */
463 if (!asoc || saddr)
464 goto out;
465
466 bp = &asoc->base.bind_addr;
467
468 if (dst) {
469 /* Walk through the bind address list and look for a bind
470 * address that matches the source address of the returned dst.
471 */
472 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
473 rcu_read_lock();
474 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
475 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
476 (laddr->state != SCTP_ADDR_SRC &&
477 !asoc->src_out_of_asoc_ok))
478 continue;
479 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
480 goto out_unlock;
481 }
482 rcu_read_unlock();
483
484 /* None of the bound addresses match the source address of the
485 * dst. So release it.
486 */
487 dst_release(dst);
488 dst = NULL;
489 }
490
491 /* Walk through the bind address list and try to get a dst that
492 * matches a bind address as the source address.
493 */
494 rcu_read_lock();
495 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
496 struct net_device *odev;
497
498 if (!laddr->valid)
499 continue;
500 if (laddr->state != SCTP_ADDR_SRC ||
501 AF_INET != laddr->a.sa.sa_family)
502 continue;
503
504 fl4->fl4_sport = laddr->a.v4.sin_port;
505 flowi4_update_output(fl4, asoc->base.sk->sk_bound_dev_if,
506 daddr->v4.sin_addr.s_addr,
507 laddr->a.v4.sin_addr.s_addr);
508
509 rt = ip_route_output_key(sock_net(sk), fl4);
510 if (IS_ERR(rt))
511 continue;
512
513 /* Ensure the src address belongs to the output
514 * interface.
515 */
516 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
517 false);
518 if (!odev || odev->ifindex != fl4->flowi4_oif) {
519 if (!dst) {
520 dst = &rt->dst;
521 t->dst = dst;
522 memcpy(fl, &_fl, sizeof(_fl));
523 } else {
524 dst_release(&rt->dst);
525 }
526 continue;
527 }
528
529 dst_release(dst);
530 dst = &rt->dst;
531 t->dst = dst;
532 memcpy(fl, &_fl, sizeof(_fl));
533 break;
534 }
535
536 out_unlock:
537 rcu_read_unlock();
538 out:
539 if (dst) {
540 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
541 &fl->u.ip4.daddr, &fl->u.ip4.saddr);
542 } else {
543 t->dst = NULL;
544 pr_debug("no route\n");
545 }
546 }
547
548 /* For v4, the source address is cached in the route entry(dst). So no need
549 * to cache it separately and hence this is an empty routine.
550 */
sctp_v4_get_saddr(struct sctp_sock * sk,struct sctp_transport * t,struct flowi * fl)551 static void sctp_v4_get_saddr(struct sctp_sock *sk,
552 struct sctp_transport *t,
553 struct flowi *fl)
554 {
555 union sctp_addr *saddr = &t->saddr;
556 struct rtable *rt = dst_rtable(t->dst);
557
558 if (rt) {
559 saddr->v4.sin_family = AF_INET;
560 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
561 }
562 }
563
564 /* What interface did this skb arrive on? */
sctp_v4_skb_iif(const struct sk_buff * skb)565 static int sctp_v4_skb_iif(const struct sk_buff *skb)
566 {
567 return inet_iif(skb);
568 }
569
sctp_v4_skb_sdif(const struct sk_buff * skb)570 static int sctp_v4_skb_sdif(const struct sk_buff *skb)
571 {
572 return inet_sdif(skb);
573 }
574
575 /* Was this packet marked by Explicit Congestion Notification? */
sctp_v4_is_ce(const struct sk_buff * skb)576 static int sctp_v4_is_ce(const struct sk_buff *skb)
577 {
578 return INET_ECN_is_ce(ip_hdr(skb)->tos);
579 }
580
581 /* Create and initialize a new sk for the socket returned by accept(). */
sctp_v4_create_accept_sk(struct sock * sk,struct sctp_association * asoc,bool kern)582 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
583 struct sctp_association *asoc,
584 bool kern)
585 {
586 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
587 sk->sk_prot, kern);
588 struct inet_sock *newinet;
589
590 if (!newsk)
591 goto out;
592
593 sock_init_data(NULL, newsk);
594
595 sctp_copy_sock(newsk, sk, asoc);
596 sock_reset_flag(newsk, SOCK_ZAPPED);
597
598 sctp_v4_copy_ip_options(sk, newsk);
599
600 newinet = inet_sk(newsk);
601
602 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
603
604 if (newsk->sk_prot->init(newsk)) {
605 sk_common_release(newsk);
606 newsk = NULL;
607 }
608
609 out:
610 return newsk;
611 }
612
sctp_v4_addr_to_user(struct sctp_sock * sp,union sctp_addr * addr)613 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
614 {
615 /* No address mapping for V4 sockets */
616 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
617 return sizeof(struct sockaddr_in);
618 }
619
620 /* Dump the v4 addr to the seq file. */
sctp_v4_seq_dump_addr(struct seq_file * seq,union sctp_addr * addr)621 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
622 {
623 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
624 }
625
sctp_v4_ecn_capable(struct sock * sk)626 static void sctp_v4_ecn_capable(struct sock *sk)
627 {
628 INET_ECN_xmit(sk);
629 }
630
sctp_addr_wq_timeout_handler(struct timer_list * t)631 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
632 {
633 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
634 struct sctp_sockaddr_entry *addrw, *temp;
635 struct sctp_sock *sp;
636
637 spin_lock_bh(&net->sctp.addr_wq_lock);
638
639 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
640 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
641 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
642 addrw->state, addrw);
643
644 #if IS_ENABLED(CONFIG_IPV6)
645 /* Now we send an ASCONF for each association */
646 /* Note. we currently don't handle link local IPv6 addressees */
647 if (addrw->a.sa.sa_family == AF_INET6) {
648 struct in6_addr *in6;
649
650 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
651 IPV6_ADDR_LINKLOCAL)
652 goto free_next;
653
654 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
655 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
656 addrw->state == SCTP_ADDR_NEW) {
657 unsigned long timeo_val;
658
659 pr_debug("%s: this is on DAD, trying %d sec "
660 "later\n", __func__,
661 SCTP_ADDRESS_TICK_DELAY);
662
663 timeo_val = jiffies;
664 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
665 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
666 break;
667 }
668 }
669 #endif
670 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
671 struct sock *sk;
672
673 sk = sctp_opt2sk(sp);
674 /* ignore bound-specific endpoints */
675 if (!sctp_is_ep_boundall(sk))
676 continue;
677 bh_lock_sock(sk);
678 if (sctp_asconf_mgmt(sp, addrw) < 0)
679 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
680 bh_unlock_sock(sk);
681 }
682 #if IS_ENABLED(CONFIG_IPV6)
683 free_next:
684 #endif
685 list_del(&addrw->list);
686 kfree(addrw);
687 }
688 spin_unlock_bh(&net->sctp.addr_wq_lock);
689 }
690
sctp_free_addr_wq(struct net * net)691 static void sctp_free_addr_wq(struct net *net)
692 {
693 struct sctp_sockaddr_entry *addrw;
694 struct sctp_sockaddr_entry *temp;
695
696 spin_lock_bh(&net->sctp.addr_wq_lock);
697 del_timer(&net->sctp.addr_wq_timer);
698 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
699 list_del(&addrw->list);
700 kfree(addrw);
701 }
702 spin_unlock_bh(&net->sctp.addr_wq_lock);
703 }
704
705 /* lookup the entry for the same address in the addr_waitq
706 * sctp_addr_wq MUST be locked
707 */
sctp_addr_wq_lookup(struct net * net,struct sctp_sockaddr_entry * addr)708 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
709 struct sctp_sockaddr_entry *addr)
710 {
711 struct sctp_sockaddr_entry *addrw;
712
713 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
714 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
715 continue;
716 if (addrw->a.sa.sa_family == AF_INET) {
717 if (addrw->a.v4.sin_addr.s_addr ==
718 addr->a.v4.sin_addr.s_addr)
719 return addrw;
720 } else if (addrw->a.sa.sa_family == AF_INET6) {
721 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
722 &addr->a.v6.sin6_addr))
723 return addrw;
724 }
725 }
726 return NULL;
727 }
728
sctp_addr_wq_mgmt(struct net * net,struct sctp_sockaddr_entry * addr,int cmd)729 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
730 {
731 struct sctp_sockaddr_entry *addrw;
732 unsigned long timeo_val;
733
734 /* first, we check if an opposite message already exist in the queue.
735 * If we found such message, it is removed.
736 * This operation is a bit stupid, but the DHCP client attaches the
737 * new address after a couple of addition and deletion of that address
738 */
739
740 spin_lock_bh(&net->sctp.addr_wq_lock);
741
742 /* Avoid searching the queue or modifying it if there are no consumers,
743 * as it can lead to performance degradation if addresses are modified
744 * en-masse.
745 *
746 * If the queue already contains some events, update it anyway to avoid
747 * ugly races between new sessions and new address events.
748 */
749 if (list_empty(&net->sctp.auto_asconf_splist) &&
750 list_empty(&net->sctp.addr_waitq)) {
751 spin_unlock_bh(&net->sctp.addr_wq_lock);
752 return;
753 }
754
755 /* Offsets existing events in addr_wq */
756 addrw = sctp_addr_wq_lookup(net, addr);
757 if (addrw) {
758 if (addrw->state != cmd) {
759 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
760 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
761 &net->sctp.addr_waitq);
762
763 list_del(&addrw->list);
764 kfree(addrw);
765 }
766 spin_unlock_bh(&net->sctp.addr_wq_lock);
767 return;
768 }
769
770 /* OK, we have to add the new address to the wait queue */
771 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
772 if (addrw == NULL) {
773 spin_unlock_bh(&net->sctp.addr_wq_lock);
774 return;
775 }
776 addrw->state = cmd;
777 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
778
779 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
780 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
781
782 if (!timer_pending(&net->sctp.addr_wq_timer)) {
783 timeo_val = jiffies;
784 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
785 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
786 }
787 spin_unlock_bh(&net->sctp.addr_wq_lock);
788 }
789
790 /* Event handler for inet address addition/deletion events.
791 * The sctp_local_addr_list needs to be protocted by a spin lock since
792 * multiple notifiers (say IPv4 and IPv6) may be running at the same
793 * time and thus corrupt the list.
794 * The reader side is protected with RCU.
795 */
sctp_inetaddr_event(struct notifier_block * this,unsigned long ev,void * ptr)796 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
797 void *ptr)
798 {
799 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
800 struct sctp_sockaddr_entry *addr = NULL;
801 struct sctp_sockaddr_entry *temp;
802 struct net *net = dev_net(ifa->ifa_dev->dev);
803 int found = 0;
804
805 switch (ev) {
806 case NETDEV_UP:
807 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
808 if (addr) {
809 addr->a.v4.sin_family = AF_INET;
810 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
811 addr->valid = 1;
812 spin_lock_bh(&net->sctp.local_addr_lock);
813 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
814 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
815 spin_unlock_bh(&net->sctp.local_addr_lock);
816 }
817 break;
818 case NETDEV_DOWN:
819 spin_lock_bh(&net->sctp.local_addr_lock);
820 list_for_each_entry_safe(addr, temp,
821 &net->sctp.local_addr_list, list) {
822 if (addr->a.sa.sa_family == AF_INET &&
823 addr->a.v4.sin_addr.s_addr ==
824 ifa->ifa_local) {
825 found = 1;
826 addr->valid = 0;
827 list_del_rcu(&addr->list);
828 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
829 break;
830 }
831 }
832 spin_unlock_bh(&net->sctp.local_addr_lock);
833 if (found)
834 kfree_rcu(addr, rcu);
835 break;
836 }
837
838 return NOTIFY_DONE;
839 }
840
841 /*
842 * Initialize the control inode/socket with a control endpoint data
843 * structure. This endpoint is reserved exclusively for the OOTB processing.
844 */
sctp_ctl_sock_init(struct net * net)845 static int sctp_ctl_sock_init(struct net *net)
846 {
847 int err;
848 sa_family_t family = PF_INET;
849
850 if (sctp_get_pf_specific(PF_INET6))
851 family = PF_INET6;
852
853 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
854 SOCK_SEQPACKET, IPPROTO_SCTP, net);
855
856 /* If IPv6 socket could not be created, try the IPv4 socket */
857 if (err < 0 && family == PF_INET6)
858 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
859 SOCK_SEQPACKET, IPPROTO_SCTP,
860 net);
861
862 if (err < 0) {
863 pr_err("Failed to create the SCTP control socket\n");
864 return err;
865 }
866 return 0;
867 }
868
sctp_udp_rcv(struct sock * sk,struct sk_buff * skb)869 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb)
870 {
871 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source;
872
873 skb_set_transport_header(skb, sizeof(struct udphdr));
874 sctp_rcv(skb);
875 return 0;
876 }
877
sctp_udp_sock_start(struct net * net)878 int sctp_udp_sock_start(struct net *net)
879 {
880 struct udp_tunnel_sock_cfg tuncfg = {NULL};
881 struct udp_port_cfg udp_conf = {0};
882 struct socket *sock;
883 int err;
884
885 udp_conf.family = AF_INET;
886 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
887 udp_conf.local_udp_port = htons(net->sctp.udp_port);
888 err = udp_sock_create(net, &udp_conf, &sock);
889 if (err) {
890 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n");
891 return err;
892 }
893
894 tuncfg.encap_type = 1;
895 tuncfg.encap_rcv = sctp_udp_rcv;
896 tuncfg.encap_err_lookup = sctp_udp_v4_err;
897 setup_udp_tunnel_sock(net, sock, &tuncfg);
898 net->sctp.udp4_sock = sock->sk;
899
900 #if IS_ENABLED(CONFIG_IPV6)
901 memset(&udp_conf, 0, sizeof(udp_conf));
902
903 udp_conf.family = AF_INET6;
904 udp_conf.local_ip6 = in6addr_any;
905 udp_conf.local_udp_port = htons(net->sctp.udp_port);
906 udp_conf.use_udp6_rx_checksums = true;
907 udp_conf.ipv6_v6only = true;
908 err = udp_sock_create(net, &udp_conf, &sock);
909 if (err) {
910 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n");
911 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
912 net->sctp.udp4_sock = NULL;
913 return err;
914 }
915
916 tuncfg.encap_type = 1;
917 tuncfg.encap_rcv = sctp_udp_rcv;
918 tuncfg.encap_err_lookup = sctp_udp_v6_err;
919 setup_udp_tunnel_sock(net, sock, &tuncfg);
920 net->sctp.udp6_sock = sock->sk;
921 #endif
922
923 return 0;
924 }
925
sctp_udp_sock_stop(struct net * net)926 void sctp_udp_sock_stop(struct net *net)
927 {
928 if (net->sctp.udp4_sock) {
929 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
930 net->sctp.udp4_sock = NULL;
931 }
932 if (net->sctp.udp6_sock) {
933 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket);
934 net->sctp.udp6_sock = NULL;
935 }
936 }
937
938 /* Register address family specific functions. */
sctp_register_af(struct sctp_af * af)939 int sctp_register_af(struct sctp_af *af)
940 {
941 switch (af->sa_family) {
942 case AF_INET:
943 if (sctp_af_v4_specific)
944 return 0;
945 sctp_af_v4_specific = af;
946 break;
947 case AF_INET6:
948 if (sctp_af_v6_specific)
949 return 0;
950 sctp_af_v6_specific = af;
951 break;
952 default:
953 return 0;
954 }
955
956 INIT_LIST_HEAD(&af->list);
957 list_add_tail(&af->list, &sctp_address_families);
958 return 1;
959 }
960
961 /* Get the table of functions for manipulating a particular address
962 * family.
963 */
sctp_get_af_specific(sa_family_t family)964 struct sctp_af *sctp_get_af_specific(sa_family_t family)
965 {
966 switch (family) {
967 case AF_INET:
968 return sctp_af_v4_specific;
969 case AF_INET6:
970 return sctp_af_v6_specific;
971 default:
972 return NULL;
973 }
974 }
975
976 /* Common code to initialize a AF_INET msg_name. */
sctp_inet_msgname(char * msgname,int * addr_len)977 static void sctp_inet_msgname(char *msgname, int *addr_len)
978 {
979 struct sockaddr_in *sin;
980
981 sin = (struct sockaddr_in *)msgname;
982 *addr_len = sizeof(struct sockaddr_in);
983 sin->sin_family = AF_INET;
984 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
985 }
986
987 /* 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)988 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
989 int *addr_len)
990 {
991 struct sockaddr_in *sin, *sinfrom;
992
993 if (msgname) {
994 struct sctp_association *asoc;
995
996 asoc = event->asoc;
997 sctp_inet_msgname(msgname, addr_len);
998 sin = (struct sockaddr_in *)msgname;
999 sinfrom = &asoc->peer.primary_addr.v4;
1000 sin->sin_port = htons(asoc->peer.port);
1001 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
1002 }
1003 }
1004
1005 /* Initialize and copy out a msgname from an inbound skb. */
sctp_inet_skb_msgname(struct sk_buff * skb,char * msgname,int * len)1006 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
1007 {
1008 if (msgname) {
1009 struct sctphdr *sh = sctp_hdr(skb);
1010 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
1011
1012 sctp_inet_msgname(msgname, len);
1013 sin->sin_port = sh->source;
1014 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1015 }
1016 }
1017
1018 /* Do we support this AF? */
sctp_inet_af_supported(sa_family_t family,struct sctp_sock * sp)1019 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
1020 {
1021 /* PF_INET only supports AF_INET addresses. */
1022 return AF_INET == family;
1023 }
1024
1025 /* Address matching with wildcards allowed. */
sctp_inet_cmp_addr(const union sctp_addr * addr1,const union sctp_addr * addr2,struct sctp_sock * opt)1026 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
1027 const union sctp_addr *addr2,
1028 struct sctp_sock *opt)
1029 {
1030 /* PF_INET only supports AF_INET addresses. */
1031 if (addr1->sa.sa_family != addr2->sa.sa_family)
1032 return 0;
1033 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
1034 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
1035 return 1;
1036 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
1037 return 1;
1038
1039 return 0;
1040 }
1041
1042 /* Verify that provided sockaddr looks bindable. Common verification has
1043 * already been taken care of.
1044 */
sctp_inet_bind_verify(struct sctp_sock * opt,union sctp_addr * addr)1045 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
1046 {
1047 return sctp_v4_available(addr, opt);
1048 }
1049
1050 /* Verify that sockaddr looks sendable. Common verification has already
1051 * been taken care of.
1052 */
sctp_inet_send_verify(struct sctp_sock * opt,union sctp_addr * addr)1053 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
1054 {
1055 return 1;
1056 }
1057
1058 /* Fill in Supported Address Type information for INIT and INIT-ACK
1059 * chunks. Returns number of addresses supported.
1060 */
sctp_inet_supported_addrs(const struct sctp_sock * opt,__be16 * types)1061 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
1062 __be16 *types)
1063 {
1064 types[0] = SCTP_PARAM_IPV4_ADDRESS;
1065 return 1;
1066 }
1067
1068 /* Wrapper routine that calls the ip transmit routine. */
sctp_v4_xmit(struct sk_buff * skb,struct sctp_transport * t)1069 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t)
1070 {
1071 struct dst_entry *dst = dst_clone(t->dst);
1072 struct flowi4 *fl4 = &t->fl.u.ip4;
1073 struct sock *sk = skb->sk;
1074 struct inet_sock *inet = inet_sk(sk);
1075 __u8 dscp = READ_ONCE(inet->tos);
1076 __be16 df = 0;
1077
1078 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
1079 skb->len, &fl4->saddr, &fl4->daddr);
1080
1081 if (t->dscp & SCTP_DSCP_SET_MASK)
1082 dscp = t->dscp & SCTP_DSCP_VAL_MASK;
1083
1084 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO
1085 : IP_PMTUDISC_DONT;
1086 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS);
1087
1088 if (!t->encap_port || !sctp_sk(sk)->udp_port) {
1089 skb_dst_set(skb, dst);
1090 return __ip_queue_xmit(sk, skb, &t->fl, dscp);
1091 }
1092
1093 if (skb_is_gso(skb))
1094 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
1095
1096 if (ip_dont_fragment(sk, dst) && !skb->ignore_df)
1097 df = htons(IP_DF);
1098
1099 skb->encapsulation = 1;
1100 skb_reset_inner_mac_header(skb);
1101 skb_reset_inner_transport_header(skb);
1102 skb_set_inner_ipproto(skb, IPPROTO_SCTP);
1103 udp_tunnel_xmit_skb(dst_rtable(dst), sk, skb, fl4->saddr,
1104 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df,
1105 sctp_sk(sk)->udp_port, t->encap_port, false, false);
1106 return 0;
1107 }
1108
1109 static struct sctp_af sctp_af_inet;
1110
1111 static struct sctp_pf sctp_pf_inet = {
1112 .event_msgname = sctp_inet_event_msgname,
1113 .skb_msgname = sctp_inet_skb_msgname,
1114 .af_supported = sctp_inet_af_supported,
1115 .cmp_addr = sctp_inet_cmp_addr,
1116 .bind_verify = sctp_inet_bind_verify,
1117 .send_verify = sctp_inet_send_verify,
1118 .supported_addrs = sctp_inet_supported_addrs,
1119 .create_accept_sk = sctp_v4_create_accept_sk,
1120 .addr_to_user = sctp_v4_addr_to_user,
1121 .to_sk_saddr = sctp_v4_to_sk_saddr,
1122 .to_sk_daddr = sctp_v4_to_sk_daddr,
1123 .copy_ip_options = sctp_v4_copy_ip_options,
1124 .af = &sctp_af_inet
1125 };
1126
1127 /* Notifier for inetaddr addition/deletion events. */
1128 static struct notifier_block sctp_inetaddr_notifier = {
1129 .notifier_call = sctp_inetaddr_event,
1130 };
1131
1132 /* Socket operations. */
1133 static const struct proto_ops inet_seqpacket_ops = {
1134 .family = PF_INET,
1135 .owner = THIS_MODULE,
1136 .release = inet_release, /* Needs to be wrapped... */
1137 .bind = inet_bind,
1138 .connect = sctp_inet_connect,
1139 .socketpair = sock_no_socketpair,
1140 .accept = inet_accept,
1141 .getname = inet_getname, /* Semantics are different. */
1142 .poll = sctp_poll,
1143 .ioctl = inet_ioctl,
1144 .gettstamp = sock_gettstamp,
1145 .listen = sctp_inet_listen,
1146 .shutdown = inet_shutdown, /* Looks harmless. */
1147 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1148 .getsockopt = sock_common_getsockopt,
1149 .sendmsg = inet_sendmsg,
1150 .recvmsg = inet_recvmsg,
1151 .mmap = sock_no_mmap,
1152 };
1153
1154 /* Registration with AF_INET family. */
1155 static struct inet_protosw sctp_seqpacket_protosw = {
1156 .type = SOCK_SEQPACKET,
1157 .protocol = IPPROTO_SCTP,
1158 .prot = &sctp_prot,
1159 .ops = &inet_seqpacket_ops,
1160 .flags = SCTP_PROTOSW_FLAG
1161 };
1162 static struct inet_protosw sctp_stream_protosw = {
1163 .type = SOCK_STREAM,
1164 .protocol = IPPROTO_SCTP,
1165 .prot = &sctp_prot,
1166 .ops = &inet_seqpacket_ops,
1167 .flags = SCTP_PROTOSW_FLAG
1168 };
1169
sctp4_rcv(struct sk_buff * skb)1170 static int sctp4_rcv(struct sk_buff *skb)
1171 {
1172 SCTP_INPUT_CB(skb)->encap_port = 0;
1173 return sctp_rcv(skb);
1174 }
1175
1176 /* Register with IP layer. */
1177 static const struct net_protocol sctp_protocol = {
1178 .handler = sctp4_rcv,
1179 .err_handler = sctp_v4_err,
1180 .no_policy = 1,
1181 .icmp_strict_tag_validation = 1,
1182 };
1183
1184 /* IPv4 address related functions. */
1185 static struct sctp_af sctp_af_inet = {
1186 .sa_family = AF_INET,
1187 .sctp_xmit = sctp_v4_xmit,
1188 .setsockopt = ip_setsockopt,
1189 .getsockopt = ip_getsockopt,
1190 .get_dst = sctp_v4_get_dst,
1191 .get_saddr = sctp_v4_get_saddr,
1192 .copy_addrlist = sctp_v4_copy_addrlist,
1193 .from_skb = sctp_v4_from_skb,
1194 .from_sk = sctp_v4_from_sk,
1195 .from_addr_param = sctp_v4_from_addr_param,
1196 .to_addr_param = sctp_v4_to_addr_param,
1197 .cmp_addr = sctp_v4_cmp_addr,
1198 .addr_valid = sctp_v4_addr_valid,
1199 .inaddr_any = sctp_v4_inaddr_any,
1200 .is_any = sctp_v4_is_any,
1201 .available = sctp_v4_available,
1202 .scope = sctp_v4_scope,
1203 .skb_iif = sctp_v4_skb_iif,
1204 .skb_sdif = sctp_v4_skb_sdif,
1205 .is_ce = sctp_v4_is_ce,
1206 .seq_dump_addr = sctp_v4_seq_dump_addr,
1207 .ecn_capable = sctp_v4_ecn_capable,
1208 .net_header_len = sizeof(struct iphdr),
1209 .sockaddr_len = sizeof(struct sockaddr_in),
1210 .ip_options_len = sctp_v4_ip_options_len,
1211 };
1212
sctp_get_pf_specific(sa_family_t family)1213 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1214 {
1215 switch (family) {
1216 case PF_INET:
1217 return sctp_pf_inet_specific;
1218 case PF_INET6:
1219 return sctp_pf_inet6_specific;
1220 default:
1221 return NULL;
1222 }
1223 }
1224
1225 /* Register the PF specific function table. */
sctp_register_pf(struct sctp_pf * pf,sa_family_t family)1226 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1227 {
1228 switch (family) {
1229 case PF_INET:
1230 if (sctp_pf_inet_specific)
1231 return 0;
1232 sctp_pf_inet_specific = pf;
1233 break;
1234 case PF_INET6:
1235 if (sctp_pf_inet6_specific)
1236 return 0;
1237 sctp_pf_inet6_specific = pf;
1238 break;
1239 default:
1240 return 0;
1241 }
1242 return 1;
1243 }
1244
init_sctp_mibs(struct net * net)1245 static inline int init_sctp_mibs(struct net *net)
1246 {
1247 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1248 if (!net->sctp.sctp_statistics)
1249 return -ENOMEM;
1250 return 0;
1251 }
1252
cleanup_sctp_mibs(struct net * net)1253 static inline void cleanup_sctp_mibs(struct net *net)
1254 {
1255 free_percpu(net->sctp.sctp_statistics);
1256 }
1257
sctp_v4_pf_init(void)1258 static void sctp_v4_pf_init(void)
1259 {
1260 /* Initialize the SCTP specific PF functions. */
1261 sctp_register_pf(&sctp_pf_inet, PF_INET);
1262 sctp_register_af(&sctp_af_inet);
1263 }
1264
sctp_v4_pf_exit(void)1265 static void sctp_v4_pf_exit(void)
1266 {
1267 list_del(&sctp_af_inet.list);
1268 }
1269
sctp_v4_protosw_init(void)1270 static int sctp_v4_protosw_init(void)
1271 {
1272 int rc;
1273
1274 rc = proto_register(&sctp_prot, 1);
1275 if (rc)
1276 return rc;
1277
1278 /* Register SCTP(UDP and TCP style) with socket layer. */
1279 inet_register_protosw(&sctp_seqpacket_protosw);
1280 inet_register_protosw(&sctp_stream_protosw);
1281
1282 return 0;
1283 }
1284
sctp_v4_protosw_exit(void)1285 static void sctp_v4_protosw_exit(void)
1286 {
1287 inet_unregister_protosw(&sctp_stream_protosw);
1288 inet_unregister_protosw(&sctp_seqpacket_protosw);
1289 proto_unregister(&sctp_prot);
1290 }
1291
sctp_v4_add_protocol(void)1292 static int sctp_v4_add_protocol(void)
1293 {
1294 /* Register notifier for inet address additions/deletions. */
1295 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1296
1297 /* Register SCTP with inet layer. */
1298 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1299 return -EAGAIN;
1300
1301 return 0;
1302 }
1303
sctp_v4_del_protocol(void)1304 static void sctp_v4_del_protocol(void)
1305 {
1306 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1307 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1308 }
1309
sctp_defaults_init(struct net * net)1310 static int __net_init sctp_defaults_init(struct net *net)
1311 {
1312 int status;
1313
1314 /*
1315 * 14. Suggested SCTP Protocol Parameter Values
1316 */
1317 /* The following protocol parameters are RECOMMENDED: */
1318 /* RTO.Initial - 3 seconds */
1319 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1320 /* RTO.Min - 1 second */
1321 net->sctp.rto_min = SCTP_RTO_MIN;
1322 /* RTO.Max - 60 seconds */
1323 net->sctp.rto_max = SCTP_RTO_MAX;
1324 /* RTO.Alpha - 1/8 */
1325 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1326 /* RTO.Beta - 1/4 */
1327 net->sctp.rto_beta = SCTP_RTO_BETA;
1328
1329 /* Valid.Cookie.Life - 60 seconds */
1330 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1331
1332 /* Whether Cookie Preservative is enabled(1) or not(0) */
1333 net->sctp.cookie_preserve_enable = 1;
1334
1335 /* Default sctp sockets to use md5 as their hmac alg */
1336 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1337 net->sctp.sctp_hmac_alg = "md5";
1338 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1339 net->sctp.sctp_hmac_alg = "sha1";
1340 #else
1341 net->sctp.sctp_hmac_alg = NULL;
1342 #endif
1343
1344 /* Max.Burst - 4 */
1345 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1346
1347 /* Disable of Primary Path Switchover by default */
1348 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1349
1350 /* Enable pf state by default */
1351 net->sctp.pf_enable = 1;
1352
1353 /* Ignore pf exposure feature by default */
1354 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1355
1356 /* Association.Max.Retrans - 10 attempts
1357 * Path.Max.Retrans - 5 attempts (per destination address)
1358 * Max.Init.Retransmits - 8 attempts
1359 */
1360 net->sctp.max_retrans_association = 10;
1361 net->sctp.max_retrans_path = 5;
1362 net->sctp.max_retrans_init = 8;
1363
1364 /* Sendbuffer growth - do per-socket accounting */
1365 net->sctp.sndbuf_policy = 0;
1366
1367 /* Rcvbuffer growth - do per-socket accounting */
1368 net->sctp.rcvbuf_policy = 0;
1369
1370 /* HB.interval - 30 seconds */
1371 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1372
1373 /* delayed SACK timeout */
1374 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1375
1376 /* Disable ADDIP by default. */
1377 net->sctp.addip_enable = 0;
1378 net->sctp.addip_noauth = 0;
1379 net->sctp.default_auto_asconf = 0;
1380
1381 /* Enable PR-SCTP by default. */
1382 net->sctp.prsctp_enable = 1;
1383
1384 /* Disable RECONF by default. */
1385 net->sctp.reconf_enable = 0;
1386
1387 /* Disable AUTH by default. */
1388 net->sctp.auth_enable = 0;
1389
1390 /* Enable ECN by default. */
1391 net->sctp.ecn_enable = 1;
1392
1393 /* Set UDP tunneling listening port to 0 by default */
1394 net->sctp.udp_port = 0;
1395
1396 /* Set remote encap port to 0 by default */
1397 net->sctp.encap_port = 0;
1398
1399 /* Set SCOPE policy to enabled */
1400 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1401
1402 /* Set the default rwnd update threshold */
1403 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1404
1405 /* Initialize maximum autoclose timeout. */
1406 net->sctp.max_autoclose = INT_MAX / HZ;
1407
1408 #ifdef CONFIG_NET_L3_MASTER_DEV
1409 net->sctp.l3mdev_accept = 1;
1410 #endif
1411
1412 status = sctp_sysctl_net_register(net);
1413 if (status)
1414 goto err_sysctl_register;
1415
1416 /* Allocate and initialise sctp mibs. */
1417 status = init_sctp_mibs(net);
1418 if (status)
1419 goto err_init_mibs;
1420
1421 #ifdef CONFIG_PROC_FS
1422 /* Initialize proc fs directory. */
1423 status = sctp_proc_init(net);
1424 if (status)
1425 goto err_init_proc;
1426 #endif
1427
1428 sctp_dbg_objcnt_init(net);
1429
1430 /* Initialize the local address list. */
1431 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1432 spin_lock_init(&net->sctp.local_addr_lock);
1433 sctp_get_local_addr_list(net);
1434
1435 /* Initialize the address event list */
1436 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1437 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1438 spin_lock_init(&net->sctp.addr_wq_lock);
1439 net->sctp.addr_wq_timer.expires = 0;
1440 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1441
1442 return 0;
1443
1444 #ifdef CONFIG_PROC_FS
1445 err_init_proc:
1446 cleanup_sctp_mibs(net);
1447 #endif
1448 err_init_mibs:
1449 sctp_sysctl_net_unregister(net);
1450 err_sysctl_register:
1451 return status;
1452 }
1453
sctp_defaults_exit(struct net * net)1454 static void __net_exit sctp_defaults_exit(struct net *net)
1455 {
1456 /* Free the local address list */
1457 sctp_free_addr_wq(net);
1458 sctp_free_local_addr_list(net);
1459
1460 #ifdef CONFIG_PROC_FS
1461 remove_proc_subtree("sctp", net->proc_net);
1462 net->sctp.proc_net_sctp = NULL;
1463 #endif
1464 cleanup_sctp_mibs(net);
1465 sctp_sysctl_net_unregister(net);
1466 }
1467
1468 static struct pernet_operations sctp_defaults_ops = {
1469 .init = sctp_defaults_init,
1470 .exit = sctp_defaults_exit,
1471 };
1472
sctp_ctrlsock_init(struct net * net)1473 static int __net_init sctp_ctrlsock_init(struct net *net)
1474 {
1475 int status;
1476
1477 /* Initialize the control inode/socket for handling OOTB packets. */
1478 status = sctp_ctl_sock_init(net);
1479 if (status)
1480 pr_err("Failed to initialize the SCTP control sock\n");
1481
1482 return status;
1483 }
1484
sctp_ctrlsock_exit(struct net * net)1485 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1486 {
1487 /* Free the control endpoint. */
1488 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1489 }
1490
1491 static struct pernet_operations sctp_ctrlsock_ops = {
1492 .init = sctp_ctrlsock_init,
1493 .exit = sctp_ctrlsock_exit,
1494 };
1495
1496 /* Initialize the universe into something sensible. */
sctp_init(void)1497 static __init int sctp_init(void)
1498 {
1499 unsigned long nr_pages = totalram_pages();
1500 unsigned long limit;
1501 unsigned long goal;
1502 int max_entry_order;
1503 int num_entries;
1504 int max_share;
1505 int status;
1506 int order;
1507 int i;
1508
1509 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1510
1511 /* Allocate bind_bucket and chunk caches. */
1512 status = -ENOBUFS;
1513 sctp_bucket_cachep = KMEM_CACHE(sctp_bind_bucket, SLAB_HWCACHE_ALIGN);
1514 if (!sctp_bucket_cachep)
1515 goto out;
1516
1517 sctp_chunk_cachep = KMEM_CACHE(sctp_chunk, SLAB_HWCACHE_ALIGN);
1518 if (!sctp_chunk_cachep)
1519 goto err_chunk_cachep;
1520
1521 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1522 if (status)
1523 goto err_percpu_counter_init;
1524
1525 /* Implementation specific variables. */
1526
1527 /* Initialize default stream count setup information. */
1528 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1529 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1530
1531 /* Initialize handle used for association ids. */
1532 idr_init(&sctp_assocs_id);
1533
1534 limit = nr_free_buffer_pages() / 8;
1535 limit = max(limit, 128UL);
1536 sysctl_sctp_mem[0] = limit / 4 * 3;
1537 sysctl_sctp_mem[1] = limit;
1538 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1539
1540 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1541 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1542 max_share = min(4UL*1024*1024, limit);
1543
1544 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
1545 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1546 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1547
1548 sysctl_sctp_wmem[0] = PAGE_SIZE;
1549 sysctl_sctp_wmem[1] = 16*1024;
1550 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1551
1552 /* Size and allocate the association hash table.
1553 * The methodology is similar to that of the tcp hash tables.
1554 * Though not identical. Start by getting a goal size
1555 */
1556 if (nr_pages >= (128 * 1024))
1557 goal = nr_pages >> (22 - PAGE_SHIFT);
1558 else
1559 goal = nr_pages >> (24 - PAGE_SHIFT);
1560
1561 /* Then compute the page order for said goal */
1562 order = get_order(goal);
1563
1564 /* Now compute the required page order for the maximum sized table we
1565 * want to create
1566 */
1567 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1568 sizeof(struct sctp_bind_hashbucket));
1569
1570 /* Limit the page order by that maximum hash table size */
1571 order = min(order, max_entry_order);
1572
1573 /* Allocate and initialize the endpoint hash table. */
1574 sctp_ep_hashsize = 64;
1575 sctp_ep_hashtable =
1576 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1577 if (!sctp_ep_hashtable) {
1578 pr_err("Failed endpoint_hash alloc\n");
1579 status = -ENOMEM;
1580 goto err_ehash_alloc;
1581 }
1582 for (i = 0; i < sctp_ep_hashsize; i++) {
1583 rwlock_init(&sctp_ep_hashtable[i].lock);
1584 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1585 }
1586
1587 /* Allocate and initialize the SCTP port hash table.
1588 * Note that order is initalized to start at the max sized
1589 * table we want to support. If we can't get that many pages
1590 * reduce the order and try again
1591 */
1592 do {
1593 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1594 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1595 } while (!sctp_port_hashtable && --order > 0);
1596
1597 if (!sctp_port_hashtable) {
1598 pr_err("Failed bind hash alloc\n");
1599 status = -ENOMEM;
1600 goto err_bhash_alloc;
1601 }
1602
1603 /* Now compute the number of entries that will fit in the
1604 * port hash space we allocated
1605 */
1606 num_entries = (1UL << order) * PAGE_SIZE /
1607 sizeof(struct sctp_bind_hashbucket);
1608
1609 /* And finish by rounding it down to the nearest power of two.
1610 * This wastes some memory of course, but it's needed because
1611 * the hash function operates based on the assumption that
1612 * the number of entries is a power of two.
1613 */
1614 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1615
1616 for (i = 0; i < sctp_port_hashsize; i++) {
1617 spin_lock_init(&sctp_port_hashtable[i].lock);
1618 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1619 }
1620
1621 status = sctp_transport_hashtable_init();
1622 if (status)
1623 goto err_thash_alloc;
1624
1625 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1626 num_entries);
1627
1628 sctp_sysctl_register();
1629
1630 INIT_LIST_HEAD(&sctp_address_families);
1631 sctp_v4_pf_init();
1632 sctp_v6_pf_init();
1633 sctp_sched_ops_init();
1634
1635 status = register_pernet_subsys(&sctp_defaults_ops);
1636 if (status)
1637 goto err_register_defaults;
1638
1639 status = sctp_v4_protosw_init();
1640 if (status)
1641 goto err_protosw_init;
1642
1643 status = sctp_v6_protosw_init();
1644 if (status)
1645 goto err_v6_protosw_init;
1646
1647 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1648 if (status)
1649 goto err_register_ctrlsock;
1650
1651 status = sctp_v4_add_protocol();
1652 if (status)
1653 goto err_add_protocol;
1654
1655 /* Register SCTP with inet6 layer. */
1656 status = sctp_v6_add_protocol();
1657 if (status)
1658 goto err_v6_add_protocol;
1659
1660 if (sctp_offload_init() < 0)
1661 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1662
1663 out:
1664 return status;
1665 err_v6_add_protocol:
1666 sctp_v4_del_protocol();
1667 err_add_protocol:
1668 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1669 err_register_ctrlsock:
1670 sctp_v6_protosw_exit();
1671 err_v6_protosw_init:
1672 sctp_v4_protosw_exit();
1673 err_protosw_init:
1674 unregister_pernet_subsys(&sctp_defaults_ops);
1675 err_register_defaults:
1676 sctp_v4_pf_exit();
1677 sctp_v6_pf_exit();
1678 sctp_sysctl_unregister();
1679 free_pages((unsigned long)sctp_port_hashtable,
1680 get_order(sctp_port_hashsize *
1681 sizeof(struct sctp_bind_hashbucket)));
1682 err_bhash_alloc:
1683 sctp_transport_hashtable_destroy();
1684 err_thash_alloc:
1685 kfree(sctp_ep_hashtable);
1686 err_ehash_alloc:
1687 percpu_counter_destroy(&sctp_sockets_allocated);
1688 err_percpu_counter_init:
1689 kmem_cache_destroy(sctp_chunk_cachep);
1690 err_chunk_cachep:
1691 kmem_cache_destroy(sctp_bucket_cachep);
1692 goto out;
1693 }
1694
1695 /* Exit handler for the SCTP protocol. */
sctp_exit(void)1696 static __exit void sctp_exit(void)
1697 {
1698 /* BUG. This should probably do something useful like clean
1699 * up all the remaining associations and all that memory.
1700 */
1701
1702 /* Unregister with inet6/inet layers. */
1703 sctp_v6_del_protocol();
1704 sctp_v4_del_protocol();
1705
1706 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1707
1708 /* Free protosw registrations */
1709 sctp_v6_protosw_exit();
1710 sctp_v4_protosw_exit();
1711
1712 unregister_pernet_subsys(&sctp_defaults_ops);
1713
1714 /* Unregister with socket layer. */
1715 sctp_v6_pf_exit();
1716 sctp_v4_pf_exit();
1717
1718 sctp_sysctl_unregister();
1719
1720 free_pages((unsigned long)sctp_port_hashtable,
1721 get_order(sctp_port_hashsize *
1722 sizeof(struct sctp_bind_hashbucket)));
1723 kfree(sctp_ep_hashtable);
1724 sctp_transport_hashtable_destroy();
1725
1726 percpu_counter_destroy(&sctp_sockets_allocated);
1727
1728 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1729
1730 kmem_cache_destroy(sctp_chunk_cachep);
1731 kmem_cache_destroy(sctp_bucket_cachep);
1732 }
1733
1734 module_init(sctp_init);
1735 module_exit(sctp_exit);
1736
1737 /*
1738 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1739 */
1740 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1741 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1742 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1743 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1744 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1745 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1746 MODULE_LICENSE("GPL");
1747