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
10 * This file is part of the SCTP kernel implementation
12 * Initialization/cleanup for SCTP protocol support.
14 * Please send any bug reports or fixes you make to the
16 * lksctp developers <linux-sctp@vger.kernel.org>
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>
27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
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>
48 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
50 /* Global data structures. */
51 struct sctp_globals sctp_globals __read_mostly;
53 struct idr sctp_assocs_id;
54 DEFINE_SPINLOCK(sctp_assocs_id_lock);
56 static struct sctp_pf *sctp_pf_inet6_specific;
57 static struct sctp_pf *sctp_pf_inet_specific;
58 static struct sctp_af *sctp_af_v4_specific;
59 static struct sctp_af *sctp_af_v6_specific;
61 struct kmem_cache *sctp_chunk_cachep __read_mostly;
62 struct kmem_cache *sctp_bucket_cachep __read_mostly;
64 long sysctl_sctp_mem[3];
65 int sysctl_sctp_rmem[3];
66 int sysctl_sctp_wmem[3];
68 /* Private helper to extract ipv4 address and stash them in
69 * the protocol structure.
71 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
72 struct net_device *dev)
74 struct in_device *in_dev;
75 struct in_ifaddr *ifa;
76 struct sctp_sockaddr_entry *addr;
79 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
84 in_dev_for_each_ifa_rcu(ifa, in_dev) {
85 /* Add the address to the local list. */
86 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
88 addr->a.v4.sin_family = AF_INET;
89 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
91 INIT_LIST_HEAD(&addr->list);
92 list_add_tail(&addr->list, addrlist);
99 /* Extract our IP addresses from the system and stash them in the
100 * protocol structure.
102 static void sctp_get_local_addr_list(struct net *net)
104 struct net_device *dev;
105 struct list_head *pos;
109 for_each_netdev_rcu(net, dev) {
110 list_for_each(pos, &sctp_address_families) {
111 af = list_entry(pos, struct sctp_af, list);
112 af->copy_addrlist(&net->sctp.local_addr_list, dev);
118 /* Free the existing local addresses. */
119 static void sctp_free_local_addr_list(struct net *net)
121 struct sctp_sockaddr_entry *addr;
122 struct list_head *pos, *temp;
124 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
125 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
131 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
132 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
133 enum sctp_scope scope, gfp_t gfp, int copy_flags)
135 struct sctp_sockaddr_entry *addr;
136 union sctp_addr laddr;
140 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
143 if (!sctp_in_scope(net, &addr->a, scope))
146 /* Now that the address is in scope, check to see if
147 * the address type is really supported by the local
148 * sock as well as the remote peer.
150 if (addr->a.sa.sa_family == AF_INET &&
151 (!(copy_flags & SCTP_ADDR4_ALLOWED) ||
152 !(copy_flags & SCTP_ADDR4_PEERSUPP)))
154 if (addr->a.sa.sa_family == AF_INET6 &&
155 (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
156 !(copy_flags & SCTP_ADDR6_PEERSUPP)))
160 /* also works for setting ipv6 address port */
161 laddr.v4.sin_port = htons(bp->port);
162 if (sctp_bind_addr_state(bp, &laddr) != -1)
165 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
166 SCTP_ADDR_SRC, GFP_ATOMIC);
175 /* Copy over any ip options */
176 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
178 struct inet_sock *newinet, *inet = inet_sk(sk);
179 struct ip_options_rcu *inet_opt, *newopt = NULL;
181 newinet = inet_sk(newsk);
184 inet_opt = rcu_dereference(inet->inet_opt);
186 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
187 inet_opt->opt.optlen, GFP_ATOMIC);
189 memcpy(newopt, inet_opt, sizeof(*inet_opt) +
190 inet_opt->opt.optlen);
192 pr_err("%s: Failed to copy ip options\n", __func__);
194 RCU_INIT_POINTER(newinet->inet_opt, newopt);
198 /* Account for the IP options */
199 static int sctp_v4_ip_options_len(struct sock *sk)
201 struct inet_sock *inet = inet_sk(sk);
202 struct ip_options_rcu *inet_opt;
206 inet_opt = rcu_dereference(inet->inet_opt);
208 len = inet_opt->opt.optlen;
214 /* Initialize a sctp_addr from in incoming skb. */
215 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
218 /* Always called on head skb, so this is safe */
219 struct sctphdr *sh = sctp_hdr(skb);
220 struct sockaddr_in *sa = &addr->v4;
222 addr->v4.sin_family = AF_INET;
225 sa->sin_port = sh->source;
226 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
228 sa->sin_port = sh->dest;
229 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
231 memset(sa->sin_zero, 0, sizeof(sa->sin_zero));
234 /* Initialize an sctp_addr from a socket. */
235 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
237 addr->v4.sin_family = AF_INET;
238 addr->v4.sin_port = 0;
239 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
240 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
243 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
244 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
246 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
249 /* Initialize sk->sk_daddr from sctp_addr. */
250 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
252 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
255 /* Initialize a sctp_addr from an address parameter. */
256 static bool sctp_v4_from_addr_param(union sctp_addr *addr,
257 union sctp_addr_param *param,
258 __be16 port, int iif)
260 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param))
263 addr->v4.sin_family = AF_INET;
264 addr->v4.sin_port = port;
265 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
266 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
271 /* Initialize an address parameter from a sctp_addr and return the length
272 * of the address parameter.
274 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
275 union sctp_addr_param *param)
277 int length = sizeof(struct sctp_ipv4addr_param);
279 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
280 param->v4.param_hdr.length = htons(length);
281 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
286 /* Initialize a sctp_addr from a dst_entry. */
287 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
290 saddr->v4.sin_family = AF_INET;
291 saddr->v4.sin_port = port;
292 saddr->v4.sin_addr.s_addr = fl4->saddr;
293 memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
296 /* Compare two addresses exactly. */
297 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
298 const union sctp_addr *addr2)
300 if (addr1->sa.sa_family != addr2->sa.sa_family)
302 if (addr1->v4.sin_port != addr2->v4.sin_port)
304 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
310 /* Initialize addr struct to INADDR_ANY. */
311 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
313 addr->v4.sin_family = AF_INET;
314 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
315 addr->v4.sin_port = port;
316 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
319 /* Is this a wildcard address? */
320 static int sctp_v4_is_any(const union sctp_addr *addr)
322 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
325 /* This function checks if the address is a valid address to be used for
329 * Return 0 - If the address is a non-unicast or an illegal address.
330 * Return 1 - If the address is a unicast.
332 static int sctp_v4_addr_valid(union sctp_addr *addr,
333 struct sctp_sock *sp,
334 const struct sk_buff *skb)
336 /* IPv4 addresses not allowed */
337 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
340 /* Is this a non-unicast address or a unusable SCTP address? */
341 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
344 /* Is this a broadcast address? */
345 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
351 /* Should this be available for binding? */
352 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
354 struct sock *sk = &sp->inet.sk;
355 struct net *net = sock_net(sk);
356 int tb_id = RT_TABLE_LOCAL;
359 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ?: tb_id;
360 ret = inet_addr_type_table(net, addr->v4.sin_addr.s_addr, tb_id);
361 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
363 !inet_test_bit(FREEBIND, sk) &&
364 !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind))
367 if (ipv6_only_sock(sctp_opt2sk(sp)))
373 /* Checking the loopback, private and other address scopes as defined in
374 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
375 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
377 * Level 0 - unusable SCTP addresses
378 * Level 1 - loopback address
379 * Level 2 - link-local addresses
380 * Level 3 - private addresses.
381 * Level 4 - global addresses
382 * For INIT and INIT-ACK address list, let L be the level of
383 * requested destination address, sender and receiver
384 * SHOULD include all of its addresses with level greater
385 * than or equal to L.
387 * IPv4 scoping can be controlled through sysctl option
388 * net.sctp.addr_scope_policy
390 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
392 enum sctp_scope retval;
394 /* Check for unusable SCTP addresses. */
395 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
396 retval = SCTP_SCOPE_UNUSABLE;
397 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
398 retval = SCTP_SCOPE_LOOPBACK;
399 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
400 retval = SCTP_SCOPE_LINK;
401 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
402 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
403 ipv4_is_private_192(addr->v4.sin_addr.s_addr) ||
404 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) {
405 retval = SCTP_SCOPE_PRIVATE;
407 retval = SCTP_SCOPE_GLOBAL;
413 /* Returns a valid dst cache entry for the given source and destination ip
414 * addresses. If an association is passed, trys to get a dst entry with a
415 * source address that matches an address in the bind address list.
417 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
418 struct flowi *fl, struct sock *sk)
420 struct sctp_association *asoc = t->asoc;
423 struct flowi4 *fl4 = &_fl.u.ip4;
424 struct sctp_bind_addr *bp;
425 struct sctp_sockaddr_entry *laddr;
426 struct dst_entry *dst = NULL;
427 union sctp_addr *daddr = &t->ipaddr;
428 union sctp_addr dst_saddr;
429 __u8 tos = inet_sk(sk)->tos;
431 if (t->dscp & SCTP_DSCP_SET_MASK)
432 tos = t->dscp & SCTP_DSCP_VAL_MASK;
433 memset(&_fl, 0x0, sizeof(_fl));
434 fl4->daddr = daddr->v4.sin_addr.s_addr;
435 fl4->fl4_dport = daddr->v4.sin_port;
436 fl4->flowi4_proto = IPPROTO_SCTP;
438 fl4->flowi4_tos = RT_TOS(tos);
439 fl4->flowi4_scope = ip_sock_rt_scope(asoc->base.sk);
440 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
441 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
444 fl4->saddr = saddr->v4.sin_addr.s_addr;
446 fl4->fl4_sport = saddr->v4.sin_port;
449 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
452 rt = ip_route_output_key(sock_net(sk), fl4);
456 memcpy(fl, &_fl, sizeof(_fl));
459 /* If there is no association or if a source address is passed, no
460 * more validation is required.
465 bp = &asoc->base.bind_addr;
468 /* Walk through the bind address list and look for a bind
469 * address that matches the source address of the returned dst.
471 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
473 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
474 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
475 (laddr->state != SCTP_ADDR_SRC &&
476 !asoc->src_out_of_asoc_ok))
478 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
483 /* None of the bound addresses match the source address of the
484 * dst. So release it.
490 /* Walk through the bind address list and try to get a dst that
491 * matches a bind address as the source address.
494 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
495 struct net_device *odev;
499 if (laddr->state != SCTP_ADDR_SRC ||
500 AF_INET != laddr->a.sa.sa_family)
503 fl4->fl4_sport = laddr->a.v4.sin_port;
504 flowi4_update_output(fl4, asoc->base.sk->sk_bound_dev_if,
505 daddr->v4.sin_addr.s_addr,
506 laddr->a.v4.sin_addr.s_addr);
508 rt = ip_route_output_key(sock_net(sk), fl4);
512 /* Ensure the src address belongs to the output
515 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
517 if (!odev || odev->ifindex != fl4->flowi4_oif) {
521 memcpy(fl, &_fl, sizeof(_fl));
523 dst_release(&rt->dst);
531 memcpy(fl, &_fl, sizeof(_fl));
539 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
540 &fl->u.ip4.daddr, &fl->u.ip4.saddr);
543 pr_debug("no route\n");
547 /* For v4, the source address is cached in the route entry(dst). So no need
548 * to cache it separately and hence this is an empty routine.
550 static void sctp_v4_get_saddr(struct sctp_sock *sk,
551 struct sctp_transport *t,
554 union sctp_addr *saddr = &t->saddr;
555 struct rtable *rt = (struct rtable *)t->dst;
558 saddr->v4.sin_family = AF_INET;
559 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
563 /* What interface did this skb arrive on? */
564 static int sctp_v4_skb_iif(const struct sk_buff *skb)
566 return inet_iif(skb);
569 static int sctp_v4_skb_sdif(const struct sk_buff *skb)
571 return inet_sdif(skb);
574 /* Was this packet marked by Explicit Congestion Notification? */
575 static int sctp_v4_is_ce(const struct sk_buff *skb)
577 return INET_ECN_is_ce(ip_hdr(skb)->tos);
580 /* Create and initialize a new sk for the socket returned by accept(). */
581 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
582 struct sctp_association *asoc,
585 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
587 struct inet_sock *newinet;
592 sock_init_data(NULL, newsk);
594 sctp_copy_sock(newsk, sk, asoc);
595 sock_reset_flag(newsk, SOCK_ZAPPED);
597 sctp_v4_copy_ip_options(sk, newsk);
599 newinet = inet_sk(newsk);
601 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
603 if (newsk->sk_prot->init(newsk)) {
604 sk_common_release(newsk);
612 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
614 /* No address mapping for V4 sockets */
615 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
616 return sizeof(struct sockaddr_in);
619 /* Dump the v4 addr to the seq file. */
620 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
622 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
625 static void sctp_v4_ecn_capable(struct sock *sk)
630 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
632 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
633 struct sctp_sockaddr_entry *addrw, *temp;
634 struct sctp_sock *sp;
636 spin_lock_bh(&net->sctp.addr_wq_lock);
638 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
639 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
640 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
641 addrw->state, addrw);
643 #if IS_ENABLED(CONFIG_IPV6)
644 /* Now we send an ASCONF for each association */
645 /* Note. we currently don't handle link local IPv6 addressees */
646 if (addrw->a.sa.sa_family == AF_INET6) {
647 struct in6_addr *in6;
649 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
653 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
654 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
655 addrw->state == SCTP_ADDR_NEW) {
656 unsigned long timeo_val;
658 pr_debug("%s: this is on DAD, trying %d sec "
660 SCTP_ADDRESS_TICK_DELAY);
663 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
664 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
669 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
672 sk = sctp_opt2sk(sp);
673 /* ignore bound-specific endpoints */
674 if (!sctp_is_ep_boundall(sk))
677 if (sctp_asconf_mgmt(sp, addrw) < 0)
678 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
681 #if IS_ENABLED(CONFIG_IPV6)
684 list_del(&addrw->list);
687 spin_unlock_bh(&net->sctp.addr_wq_lock);
690 static void sctp_free_addr_wq(struct net *net)
692 struct sctp_sockaddr_entry *addrw;
693 struct sctp_sockaddr_entry *temp;
695 spin_lock_bh(&net->sctp.addr_wq_lock);
696 del_timer(&net->sctp.addr_wq_timer);
697 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
698 list_del(&addrw->list);
701 spin_unlock_bh(&net->sctp.addr_wq_lock);
704 /* lookup the entry for the same address in the addr_waitq
705 * sctp_addr_wq MUST be locked
707 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
708 struct sctp_sockaddr_entry *addr)
710 struct sctp_sockaddr_entry *addrw;
712 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
713 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
715 if (addrw->a.sa.sa_family == AF_INET) {
716 if (addrw->a.v4.sin_addr.s_addr ==
717 addr->a.v4.sin_addr.s_addr)
719 } else if (addrw->a.sa.sa_family == AF_INET6) {
720 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
721 &addr->a.v6.sin6_addr))
728 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
730 struct sctp_sockaddr_entry *addrw;
731 unsigned long timeo_val;
733 /* first, we check if an opposite message already exist in the queue.
734 * If we found such message, it is removed.
735 * This operation is a bit stupid, but the DHCP client attaches the
736 * new address after a couple of addition and deletion of that address
739 spin_lock_bh(&net->sctp.addr_wq_lock);
740 /* Offsets existing events in addr_wq */
741 addrw = sctp_addr_wq_lookup(net, addr);
743 if (addrw->state != cmd) {
744 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
745 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
746 &net->sctp.addr_waitq);
748 list_del(&addrw->list);
751 spin_unlock_bh(&net->sctp.addr_wq_lock);
755 /* OK, we have to add the new address to the wait queue */
756 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
758 spin_unlock_bh(&net->sctp.addr_wq_lock);
762 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
764 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
765 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
767 if (!timer_pending(&net->sctp.addr_wq_timer)) {
769 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
770 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
772 spin_unlock_bh(&net->sctp.addr_wq_lock);
775 /* Event handler for inet address addition/deletion events.
776 * The sctp_local_addr_list needs to be protocted by a spin lock since
777 * multiple notifiers (say IPv4 and IPv6) may be running at the same
778 * time and thus corrupt the list.
779 * The reader side is protected with RCU.
781 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
784 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
785 struct sctp_sockaddr_entry *addr = NULL;
786 struct sctp_sockaddr_entry *temp;
787 struct net *net = dev_net(ifa->ifa_dev->dev);
792 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
794 addr->a.v4.sin_family = AF_INET;
795 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
797 spin_lock_bh(&net->sctp.local_addr_lock);
798 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
799 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
800 spin_unlock_bh(&net->sctp.local_addr_lock);
804 spin_lock_bh(&net->sctp.local_addr_lock);
805 list_for_each_entry_safe(addr, temp,
806 &net->sctp.local_addr_list, list) {
807 if (addr->a.sa.sa_family == AF_INET &&
808 addr->a.v4.sin_addr.s_addr ==
810 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
813 list_del_rcu(&addr->list);
817 spin_unlock_bh(&net->sctp.local_addr_lock);
819 kfree_rcu(addr, rcu);
827 * Initialize the control inode/socket with a control endpoint data
828 * structure. This endpoint is reserved exclusively for the OOTB processing.
830 static int sctp_ctl_sock_init(struct net *net)
833 sa_family_t family = PF_INET;
835 if (sctp_get_pf_specific(PF_INET6))
838 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
839 SOCK_SEQPACKET, IPPROTO_SCTP, net);
841 /* If IPv6 socket could not be created, try the IPv4 socket */
842 if (err < 0 && family == PF_INET6)
843 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
844 SOCK_SEQPACKET, IPPROTO_SCTP,
848 pr_err("Failed to create the SCTP control socket\n");
854 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb)
856 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source;
858 skb_set_transport_header(skb, sizeof(struct udphdr));
863 int sctp_udp_sock_start(struct net *net)
865 struct udp_tunnel_sock_cfg tuncfg = {NULL};
866 struct udp_port_cfg udp_conf = {0};
870 udp_conf.family = AF_INET;
871 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
872 udp_conf.local_udp_port = htons(net->sctp.udp_port);
873 err = udp_sock_create(net, &udp_conf, &sock);
875 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n");
879 tuncfg.encap_type = 1;
880 tuncfg.encap_rcv = sctp_udp_rcv;
881 tuncfg.encap_err_lookup = sctp_udp_v4_err;
882 setup_udp_tunnel_sock(net, sock, &tuncfg);
883 net->sctp.udp4_sock = sock->sk;
885 #if IS_ENABLED(CONFIG_IPV6)
886 memset(&udp_conf, 0, sizeof(udp_conf));
888 udp_conf.family = AF_INET6;
889 udp_conf.local_ip6 = in6addr_any;
890 udp_conf.local_udp_port = htons(net->sctp.udp_port);
891 udp_conf.use_udp6_rx_checksums = true;
892 udp_conf.ipv6_v6only = true;
893 err = udp_sock_create(net, &udp_conf, &sock);
895 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n");
896 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
897 net->sctp.udp4_sock = NULL;
901 tuncfg.encap_type = 1;
902 tuncfg.encap_rcv = sctp_udp_rcv;
903 tuncfg.encap_err_lookup = sctp_udp_v6_err;
904 setup_udp_tunnel_sock(net, sock, &tuncfg);
905 net->sctp.udp6_sock = sock->sk;
911 void sctp_udp_sock_stop(struct net *net)
913 if (net->sctp.udp4_sock) {
914 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
915 net->sctp.udp4_sock = NULL;
917 if (net->sctp.udp6_sock) {
918 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket);
919 net->sctp.udp6_sock = NULL;
923 /* Register address family specific functions. */
924 int sctp_register_af(struct sctp_af *af)
926 switch (af->sa_family) {
928 if (sctp_af_v4_specific)
930 sctp_af_v4_specific = af;
933 if (sctp_af_v6_specific)
935 sctp_af_v6_specific = af;
941 INIT_LIST_HEAD(&af->list);
942 list_add_tail(&af->list, &sctp_address_families);
946 /* Get the table of functions for manipulating a particular address
949 struct sctp_af *sctp_get_af_specific(sa_family_t family)
953 return sctp_af_v4_specific;
955 return sctp_af_v6_specific;
961 /* Common code to initialize a AF_INET msg_name. */
962 static void sctp_inet_msgname(char *msgname, int *addr_len)
964 struct sockaddr_in *sin;
966 sin = (struct sockaddr_in *)msgname;
967 *addr_len = sizeof(struct sockaddr_in);
968 sin->sin_family = AF_INET;
969 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
972 /* Copy the primary address of the peer primary address as the msg_name. */
973 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
976 struct sockaddr_in *sin, *sinfrom;
979 struct sctp_association *asoc;
982 sctp_inet_msgname(msgname, addr_len);
983 sin = (struct sockaddr_in *)msgname;
984 sinfrom = &asoc->peer.primary_addr.v4;
985 sin->sin_port = htons(asoc->peer.port);
986 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
990 /* Initialize and copy out a msgname from an inbound skb. */
991 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
994 struct sctphdr *sh = sctp_hdr(skb);
995 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
997 sctp_inet_msgname(msgname, len);
998 sin->sin_port = sh->source;
999 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1003 /* Do we support this AF? */
1004 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
1006 /* PF_INET only supports AF_INET addresses. */
1007 return AF_INET == family;
1010 /* Address matching with wildcards allowed. */
1011 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
1012 const union sctp_addr *addr2,
1013 struct sctp_sock *opt)
1015 /* PF_INET only supports AF_INET addresses. */
1016 if (addr1->sa.sa_family != addr2->sa.sa_family)
1018 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
1019 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
1021 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
1027 /* Verify that provided sockaddr looks bindable. Common verification has
1028 * already been taken care of.
1030 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
1032 return sctp_v4_available(addr, opt);
1035 /* Verify that sockaddr looks sendable. Common verification has already
1036 * been taken care of.
1038 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
1043 /* Fill in Supported Address Type information for INIT and INIT-ACK
1044 * chunks. Returns number of addresses supported.
1046 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
1049 types[0] = SCTP_PARAM_IPV4_ADDRESS;
1053 /* Wrapper routine that calls the ip transmit routine. */
1054 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t)
1056 struct dst_entry *dst = dst_clone(t->dst);
1057 struct flowi4 *fl4 = &t->fl.u.ip4;
1058 struct sock *sk = skb->sk;
1059 struct inet_sock *inet = inet_sk(sk);
1060 __u8 dscp = inet->tos;
1063 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
1064 skb->len, &fl4->saddr, &fl4->daddr);
1066 if (t->dscp & SCTP_DSCP_SET_MASK)
1067 dscp = t->dscp & SCTP_DSCP_VAL_MASK;
1069 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO
1071 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS);
1073 if (!t->encap_port || !sctp_sk(sk)->udp_port) {
1074 skb_dst_set(skb, dst);
1075 return __ip_queue_xmit(sk, skb, &t->fl, dscp);
1078 if (skb_is_gso(skb))
1079 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
1081 if (ip_dont_fragment(sk, dst) && !skb->ignore_df)
1084 skb->encapsulation = 1;
1085 skb_reset_inner_mac_header(skb);
1086 skb_reset_inner_transport_header(skb);
1087 skb_set_inner_ipproto(skb, IPPROTO_SCTP);
1088 udp_tunnel_xmit_skb((struct rtable *)dst, sk, skb, fl4->saddr,
1089 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df,
1090 sctp_sk(sk)->udp_port, t->encap_port, false, false);
1094 static struct sctp_af sctp_af_inet;
1096 static struct sctp_pf sctp_pf_inet = {
1097 .event_msgname = sctp_inet_event_msgname,
1098 .skb_msgname = sctp_inet_skb_msgname,
1099 .af_supported = sctp_inet_af_supported,
1100 .cmp_addr = sctp_inet_cmp_addr,
1101 .bind_verify = sctp_inet_bind_verify,
1102 .send_verify = sctp_inet_send_verify,
1103 .supported_addrs = sctp_inet_supported_addrs,
1104 .create_accept_sk = sctp_v4_create_accept_sk,
1105 .addr_to_user = sctp_v4_addr_to_user,
1106 .to_sk_saddr = sctp_v4_to_sk_saddr,
1107 .to_sk_daddr = sctp_v4_to_sk_daddr,
1108 .copy_ip_options = sctp_v4_copy_ip_options,
1112 /* Notifier for inetaddr addition/deletion events. */
1113 static struct notifier_block sctp_inetaddr_notifier = {
1114 .notifier_call = sctp_inetaddr_event,
1117 /* Socket operations. */
1118 static const struct proto_ops inet_seqpacket_ops = {
1120 .owner = THIS_MODULE,
1121 .release = inet_release, /* Needs to be wrapped... */
1123 .connect = sctp_inet_connect,
1124 .socketpair = sock_no_socketpair,
1125 .accept = inet_accept,
1126 .getname = inet_getname, /* Semantics are different. */
1128 .ioctl = inet_ioctl,
1129 .gettstamp = sock_gettstamp,
1130 .listen = sctp_inet_listen,
1131 .shutdown = inet_shutdown, /* Looks harmless. */
1132 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1133 .getsockopt = sock_common_getsockopt,
1134 .sendmsg = inet_sendmsg,
1135 .recvmsg = inet_recvmsg,
1136 .mmap = sock_no_mmap,
1139 /* Registration with AF_INET family. */
1140 static struct inet_protosw sctp_seqpacket_protosw = {
1141 .type = SOCK_SEQPACKET,
1142 .protocol = IPPROTO_SCTP,
1144 .ops = &inet_seqpacket_ops,
1145 .flags = SCTP_PROTOSW_FLAG
1147 static struct inet_protosw sctp_stream_protosw = {
1148 .type = SOCK_STREAM,
1149 .protocol = IPPROTO_SCTP,
1151 .ops = &inet_seqpacket_ops,
1152 .flags = SCTP_PROTOSW_FLAG
1155 static int sctp4_rcv(struct sk_buff *skb)
1157 SCTP_INPUT_CB(skb)->encap_port = 0;
1158 return sctp_rcv(skb);
1161 /* Register with IP layer. */
1162 static const struct net_protocol sctp_protocol = {
1163 .handler = sctp4_rcv,
1164 .err_handler = sctp_v4_err,
1166 .icmp_strict_tag_validation = 1,
1169 /* IPv4 address related functions. */
1170 static struct sctp_af sctp_af_inet = {
1171 .sa_family = AF_INET,
1172 .sctp_xmit = sctp_v4_xmit,
1173 .setsockopt = ip_setsockopt,
1174 .getsockopt = ip_getsockopt,
1175 .get_dst = sctp_v4_get_dst,
1176 .get_saddr = sctp_v4_get_saddr,
1177 .copy_addrlist = sctp_v4_copy_addrlist,
1178 .from_skb = sctp_v4_from_skb,
1179 .from_sk = sctp_v4_from_sk,
1180 .from_addr_param = sctp_v4_from_addr_param,
1181 .to_addr_param = sctp_v4_to_addr_param,
1182 .cmp_addr = sctp_v4_cmp_addr,
1183 .addr_valid = sctp_v4_addr_valid,
1184 .inaddr_any = sctp_v4_inaddr_any,
1185 .is_any = sctp_v4_is_any,
1186 .available = sctp_v4_available,
1187 .scope = sctp_v4_scope,
1188 .skb_iif = sctp_v4_skb_iif,
1189 .skb_sdif = sctp_v4_skb_sdif,
1190 .is_ce = sctp_v4_is_ce,
1191 .seq_dump_addr = sctp_v4_seq_dump_addr,
1192 .ecn_capable = sctp_v4_ecn_capable,
1193 .net_header_len = sizeof(struct iphdr),
1194 .sockaddr_len = sizeof(struct sockaddr_in),
1195 .ip_options_len = sctp_v4_ip_options_len,
1198 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1202 return sctp_pf_inet_specific;
1204 return sctp_pf_inet6_specific;
1210 /* Register the PF specific function table. */
1211 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1215 if (sctp_pf_inet_specific)
1217 sctp_pf_inet_specific = pf;
1220 if (sctp_pf_inet6_specific)
1222 sctp_pf_inet6_specific = pf;
1230 static inline int init_sctp_mibs(struct net *net)
1232 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1233 if (!net->sctp.sctp_statistics)
1238 static inline void cleanup_sctp_mibs(struct net *net)
1240 free_percpu(net->sctp.sctp_statistics);
1243 static void sctp_v4_pf_init(void)
1245 /* Initialize the SCTP specific PF functions. */
1246 sctp_register_pf(&sctp_pf_inet, PF_INET);
1247 sctp_register_af(&sctp_af_inet);
1250 static void sctp_v4_pf_exit(void)
1252 list_del(&sctp_af_inet.list);
1255 static int sctp_v4_protosw_init(void)
1259 rc = proto_register(&sctp_prot, 1);
1263 /* Register SCTP(UDP and TCP style) with socket layer. */
1264 inet_register_protosw(&sctp_seqpacket_protosw);
1265 inet_register_protosw(&sctp_stream_protosw);
1270 static void sctp_v4_protosw_exit(void)
1272 inet_unregister_protosw(&sctp_stream_protosw);
1273 inet_unregister_protosw(&sctp_seqpacket_protosw);
1274 proto_unregister(&sctp_prot);
1277 static int sctp_v4_add_protocol(void)
1279 /* Register notifier for inet address additions/deletions. */
1280 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1282 /* Register SCTP with inet layer. */
1283 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1289 static void sctp_v4_del_protocol(void)
1291 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1292 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1295 static int __net_init sctp_defaults_init(struct net *net)
1300 * 14. Suggested SCTP Protocol Parameter Values
1302 /* The following protocol parameters are RECOMMENDED: */
1303 /* RTO.Initial - 3 seconds */
1304 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1305 /* RTO.Min - 1 second */
1306 net->sctp.rto_min = SCTP_RTO_MIN;
1307 /* RTO.Max - 60 seconds */
1308 net->sctp.rto_max = SCTP_RTO_MAX;
1309 /* RTO.Alpha - 1/8 */
1310 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1311 /* RTO.Beta - 1/4 */
1312 net->sctp.rto_beta = SCTP_RTO_BETA;
1314 /* Valid.Cookie.Life - 60 seconds */
1315 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1317 /* Whether Cookie Preservative is enabled(1) or not(0) */
1318 net->sctp.cookie_preserve_enable = 1;
1320 /* Default sctp sockets to use md5 as their hmac alg */
1321 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1322 net->sctp.sctp_hmac_alg = "md5";
1323 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1324 net->sctp.sctp_hmac_alg = "sha1";
1326 net->sctp.sctp_hmac_alg = NULL;
1330 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1332 /* Disable of Primary Path Switchover by default */
1333 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1335 /* Enable pf state by default */
1336 net->sctp.pf_enable = 1;
1338 /* Ignore pf exposure feature by default */
1339 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1341 /* Association.Max.Retrans - 10 attempts
1342 * Path.Max.Retrans - 5 attempts (per destination address)
1343 * Max.Init.Retransmits - 8 attempts
1345 net->sctp.max_retrans_association = 10;
1346 net->sctp.max_retrans_path = 5;
1347 net->sctp.max_retrans_init = 8;
1349 /* Sendbuffer growth - do per-socket accounting */
1350 net->sctp.sndbuf_policy = 0;
1352 /* Rcvbuffer growth - do per-socket accounting */
1353 net->sctp.rcvbuf_policy = 0;
1355 /* HB.interval - 30 seconds */
1356 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1358 /* delayed SACK timeout */
1359 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1361 /* Disable ADDIP by default. */
1362 net->sctp.addip_enable = 0;
1363 net->sctp.addip_noauth = 0;
1364 net->sctp.default_auto_asconf = 0;
1366 /* Enable PR-SCTP by default. */
1367 net->sctp.prsctp_enable = 1;
1369 /* Disable RECONF by default. */
1370 net->sctp.reconf_enable = 0;
1372 /* Disable AUTH by default. */
1373 net->sctp.auth_enable = 0;
1375 /* Enable ECN by default. */
1376 net->sctp.ecn_enable = 1;
1378 /* Set UDP tunneling listening port to 0 by default */
1379 net->sctp.udp_port = 0;
1381 /* Set remote encap port to 0 by default */
1382 net->sctp.encap_port = 0;
1384 /* Set SCOPE policy to enabled */
1385 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1387 /* Set the default rwnd update threshold */
1388 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1390 /* Initialize maximum autoclose timeout. */
1391 net->sctp.max_autoclose = INT_MAX / HZ;
1393 #ifdef CONFIG_NET_L3_MASTER_DEV
1394 net->sctp.l3mdev_accept = 1;
1397 status = sctp_sysctl_net_register(net);
1399 goto err_sysctl_register;
1401 /* Allocate and initialise sctp mibs. */
1402 status = init_sctp_mibs(net);
1406 #ifdef CONFIG_PROC_FS
1407 /* Initialize proc fs directory. */
1408 status = sctp_proc_init(net);
1413 sctp_dbg_objcnt_init(net);
1415 /* Initialize the local address list. */
1416 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1417 spin_lock_init(&net->sctp.local_addr_lock);
1418 sctp_get_local_addr_list(net);
1420 /* Initialize the address event list */
1421 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1422 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1423 spin_lock_init(&net->sctp.addr_wq_lock);
1424 net->sctp.addr_wq_timer.expires = 0;
1425 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1429 #ifdef CONFIG_PROC_FS
1431 cleanup_sctp_mibs(net);
1434 sctp_sysctl_net_unregister(net);
1435 err_sysctl_register:
1439 static void __net_exit sctp_defaults_exit(struct net *net)
1441 /* Free the local address list */
1442 sctp_free_addr_wq(net);
1443 sctp_free_local_addr_list(net);
1445 #ifdef CONFIG_PROC_FS
1446 remove_proc_subtree("sctp", net->proc_net);
1447 net->sctp.proc_net_sctp = NULL;
1449 cleanup_sctp_mibs(net);
1450 sctp_sysctl_net_unregister(net);
1453 static struct pernet_operations sctp_defaults_ops = {
1454 .init = sctp_defaults_init,
1455 .exit = sctp_defaults_exit,
1458 static int __net_init sctp_ctrlsock_init(struct net *net)
1462 /* Initialize the control inode/socket for handling OOTB packets. */
1463 status = sctp_ctl_sock_init(net);
1465 pr_err("Failed to initialize the SCTP control sock\n");
1470 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1472 /* Free the control endpoint. */
1473 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1476 static struct pernet_operations sctp_ctrlsock_ops = {
1477 .init = sctp_ctrlsock_init,
1478 .exit = sctp_ctrlsock_exit,
1481 /* Initialize the universe into something sensible. */
1482 static __init int sctp_init(void)
1484 unsigned long nr_pages = totalram_pages();
1485 unsigned long limit;
1487 int max_entry_order;
1494 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1496 /* Allocate bind_bucket and chunk caches. */
1498 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1499 sizeof(struct sctp_bind_bucket),
1500 0, SLAB_HWCACHE_ALIGN,
1502 if (!sctp_bucket_cachep)
1505 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1506 sizeof(struct sctp_chunk),
1507 0, SLAB_HWCACHE_ALIGN,
1509 if (!sctp_chunk_cachep)
1510 goto err_chunk_cachep;
1512 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1514 goto err_percpu_counter_init;
1516 /* Implementation specific variables. */
1518 /* Initialize default stream count setup information. */
1519 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1520 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1522 /* Initialize handle used for association ids. */
1523 idr_init(&sctp_assocs_id);
1525 limit = nr_free_buffer_pages() / 8;
1526 limit = max(limit, 128UL);
1527 sysctl_sctp_mem[0] = limit / 4 * 3;
1528 sysctl_sctp_mem[1] = limit;
1529 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1531 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1532 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1533 max_share = min(4UL*1024*1024, limit);
1535 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
1536 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1537 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1539 sysctl_sctp_wmem[0] = PAGE_SIZE;
1540 sysctl_sctp_wmem[1] = 16*1024;
1541 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1543 /* Size and allocate the association hash table.
1544 * The methodology is similar to that of the tcp hash tables.
1545 * Though not identical. Start by getting a goal size
1547 if (nr_pages >= (128 * 1024))
1548 goal = nr_pages >> (22 - PAGE_SHIFT);
1550 goal = nr_pages >> (24 - PAGE_SHIFT);
1552 /* Then compute the page order for said goal */
1553 order = get_order(goal);
1555 /* Now compute the required page order for the maximum sized table we
1558 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1559 sizeof(struct sctp_bind_hashbucket));
1561 /* Limit the page order by that maximum hash table size */
1562 order = min(order, max_entry_order);
1564 /* Allocate and initialize the endpoint hash table. */
1565 sctp_ep_hashsize = 64;
1567 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1568 if (!sctp_ep_hashtable) {
1569 pr_err("Failed endpoint_hash alloc\n");
1571 goto err_ehash_alloc;
1573 for (i = 0; i < sctp_ep_hashsize; i++) {
1574 rwlock_init(&sctp_ep_hashtable[i].lock);
1575 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1578 /* Allocate and initialize the SCTP port hash table.
1579 * Note that order is initalized to start at the max sized
1580 * table we want to support. If we can't get that many pages
1581 * reduce the order and try again
1584 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1585 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1586 } while (!sctp_port_hashtable && --order > 0);
1588 if (!sctp_port_hashtable) {
1589 pr_err("Failed bind hash alloc\n");
1591 goto err_bhash_alloc;
1594 /* Now compute the number of entries that will fit in the
1595 * port hash space we allocated
1597 num_entries = (1UL << order) * PAGE_SIZE /
1598 sizeof(struct sctp_bind_hashbucket);
1600 /* And finish by rounding it down to the nearest power of two.
1601 * This wastes some memory of course, but it's needed because
1602 * the hash function operates based on the assumption that
1603 * the number of entries is a power of two.
1605 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1607 for (i = 0; i < sctp_port_hashsize; i++) {
1608 spin_lock_init(&sctp_port_hashtable[i].lock);
1609 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1612 status = sctp_transport_hashtable_init();
1614 goto err_thash_alloc;
1616 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1619 sctp_sysctl_register();
1621 INIT_LIST_HEAD(&sctp_address_families);
1624 sctp_sched_ops_init();
1626 status = register_pernet_subsys(&sctp_defaults_ops);
1628 goto err_register_defaults;
1630 status = sctp_v4_protosw_init();
1632 goto err_protosw_init;
1634 status = sctp_v6_protosw_init();
1636 goto err_v6_protosw_init;
1638 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1640 goto err_register_ctrlsock;
1642 status = sctp_v4_add_protocol();
1644 goto err_add_protocol;
1646 /* Register SCTP with inet6 layer. */
1647 status = sctp_v6_add_protocol();
1649 goto err_v6_add_protocol;
1651 if (sctp_offload_init() < 0)
1652 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1656 err_v6_add_protocol:
1657 sctp_v4_del_protocol();
1659 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1660 err_register_ctrlsock:
1661 sctp_v6_protosw_exit();
1662 err_v6_protosw_init:
1663 sctp_v4_protosw_exit();
1665 unregister_pernet_subsys(&sctp_defaults_ops);
1666 err_register_defaults:
1669 sctp_sysctl_unregister();
1670 free_pages((unsigned long)sctp_port_hashtable,
1671 get_order(sctp_port_hashsize *
1672 sizeof(struct sctp_bind_hashbucket)));
1674 sctp_transport_hashtable_destroy();
1676 kfree(sctp_ep_hashtable);
1678 percpu_counter_destroy(&sctp_sockets_allocated);
1679 err_percpu_counter_init:
1680 kmem_cache_destroy(sctp_chunk_cachep);
1682 kmem_cache_destroy(sctp_bucket_cachep);
1686 /* Exit handler for the SCTP protocol. */
1687 static __exit void sctp_exit(void)
1689 /* BUG. This should probably do something useful like clean
1690 * up all the remaining associations and all that memory.
1693 /* Unregister with inet6/inet layers. */
1694 sctp_v6_del_protocol();
1695 sctp_v4_del_protocol();
1697 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1699 /* Free protosw registrations */
1700 sctp_v6_protosw_exit();
1701 sctp_v4_protosw_exit();
1703 unregister_pernet_subsys(&sctp_defaults_ops);
1705 /* Unregister with socket layer. */
1709 sctp_sysctl_unregister();
1711 free_pages((unsigned long)sctp_port_hashtable,
1712 get_order(sctp_port_hashsize *
1713 sizeof(struct sctp_bind_hashbucket)));
1714 kfree(sctp_ep_hashtable);
1715 sctp_transport_hashtable_destroy();
1717 percpu_counter_destroy(&sctp_sockets_allocated);
1719 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1721 kmem_cache_destroy(sctp_chunk_cachep);
1722 kmem_cache_destroy(sctp_bucket_cachep);
1725 module_init(sctp_init);
1726 module_exit(sctp_exit);
1729 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1731 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1732 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1733 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1734 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1735 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1736 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1737 MODULE_LICENSE("GPL");