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/swap.h>
37 #include <linux/slab.h>
38 #include <net/net_namespace.h>
39 #include <net/protocol.h>
42 #include <net/route.h>
43 #include <net/sctp/sctp.h>
44 #include <net/addrconf.h>
45 #include <net/inet_common.h>
46 #include <net/inet_ecn.h>
47 #include <net/udp_tunnel.h>
49 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
51 /* Global data structures. */
52 struct sctp_globals sctp_globals __read_mostly;
54 struct idr sctp_assocs_id;
55 DEFINE_SPINLOCK(sctp_assocs_id_lock);
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;
62 struct kmem_cache *sctp_chunk_cachep __read_mostly;
63 struct kmem_cache *sctp_bucket_cachep __read_mostly;
65 long sysctl_sctp_mem[3];
66 int sysctl_sctp_rmem[3];
67 int sysctl_sctp_wmem[3];
69 /* Private helper to extract ipv4 address and stash them in
70 * the protocol structure.
72 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
73 struct net_device *dev)
75 struct in_device *in_dev;
76 struct in_ifaddr *ifa;
77 struct sctp_sockaddr_entry *addr;
80 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
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);
89 addr->a.v4.sin_family = AF_INET;
90 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
92 INIT_LIST_HEAD(&addr->list);
93 list_add_tail(&addr->list, addrlist);
100 /* Extract our IP addresses from the system and stash them in the
101 * protocol structure.
103 static void sctp_get_local_addr_list(struct net *net)
105 struct net_device *dev;
106 struct list_head *pos;
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);
119 /* Free the existing local addresses. */
120 static void sctp_free_local_addr_list(struct net *net)
122 struct sctp_sockaddr_entry *addr;
123 struct list_head *pos, *temp;
125 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
126 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
132 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
133 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
134 enum sctp_scope scope, gfp_t gfp, int copy_flags)
136 struct sctp_sockaddr_entry *addr;
137 union sctp_addr laddr;
141 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
144 if (!sctp_in_scope(net, &addr->a, scope))
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.
151 if (addr->a.sa.sa_family == AF_INET &&
152 (!(copy_flags & SCTP_ADDR4_ALLOWED) ||
153 !(copy_flags & SCTP_ADDR4_PEERSUPP)))
155 if (addr->a.sa.sa_family == AF_INET6 &&
156 (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
157 !(copy_flags & SCTP_ADDR6_PEERSUPP)))
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)
166 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
167 SCTP_ADDR_SRC, GFP_ATOMIC);
176 /* Copy over any ip options */
177 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
179 struct inet_sock *newinet, *inet = inet_sk(sk);
180 struct ip_options_rcu *inet_opt, *newopt = NULL;
182 newinet = inet_sk(newsk);
185 inet_opt = rcu_dereference(inet->inet_opt);
187 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
188 inet_opt->opt.optlen, GFP_ATOMIC);
190 memcpy(newopt, inet_opt, sizeof(*inet_opt) +
191 inet_opt->opt.optlen);
193 pr_err("%s: Failed to copy ip options\n", __func__);
195 RCU_INIT_POINTER(newinet->inet_opt, newopt);
199 /* Account for the IP options */
200 static int sctp_v4_ip_options_len(struct sock *sk)
202 struct inet_sock *inet = inet_sk(sk);
203 struct ip_options_rcu *inet_opt;
207 inet_opt = rcu_dereference(inet->inet_opt);
209 len = inet_opt->opt.optlen;
215 /* Initialize a sctp_addr from in incoming skb. */
216 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
219 /* Always called on head skb, so this is safe */
220 struct sctphdr *sh = sctp_hdr(skb);
221 struct sockaddr_in *sa = &addr->v4;
223 addr->v4.sin_family = AF_INET;
226 sa->sin_port = sh->source;
227 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
229 sa->sin_port = sh->dest;
230 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
232 memset(sa->sin_zero, 0, sizeof(sa->sin_zero));
235 /* Initialize an sctp_addr from a socket. */
236 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
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));
244 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
245 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
247 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
250 /* Initialize sk->sk_daddr from sctp_addr. */
251 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
253 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
256 /* Initialize a sctp_addr from an address parameter. */
257 static bool sctp_v4_from_addr_param(union sctp_addr *addr,
258 union sctp_addr_param *param,
259 __be16 port, int iif)
261 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param))
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));
272 /* Initialize an address parameter from a sctp_addr and return the length
273 * of the address parameter.
275 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
276 union sctp_addr_param *param)
278 int length = sizeof(struct sctp_ipv4addr_param);
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;
287 /* Initialize a sctp_addr from a dst_entry. */
288 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
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));
297 /* Compare two addresses exactly. */
298 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
299 const union sctp_addr *addr2)
301 if (addr1->sa.sa_family != addr2->sa.sa_family)
303 if (addr1->v4.sin_port != addr2->v4.sin_port)
305 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
311 /* Initialize addr struct to INADDR_ANY. */
312 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
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));
320 /* Is this a wildcard address? */
321 static int sctp_v4_is_any(const union sctp_addr *addr)
323 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
326 /* This function checks if the address is a valid address to be used for
330 * Return 0 - If the address is a non-unicast or an illegal address.
331 * Return 1 - If the address is a unicast.
333 static int sctp_v4_addr_valid(union sctp_addr *addr,
334 struct sctp_sock *sp,
335 const struct sk_buff *skb)
337 /* IPv4 addresses not allowed */
338 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
341 /* Is this a non-unicast address or a unusable SCTP address? */
342 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
345 /* Is this a broadcast address? */
346 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
352 /* Should this be available for binding? */
353 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
355 struct net *net = sock_net(&sp->inet.sk);
356 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
359 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
361 !sp->inet.freebind &&
362 !net->ipv4.sysctl_ip_nonlocal_bind)
365 if (ipv6_only_sock(sctp_opt2sk(sp)))
371 /* Checking the loopback, private and other address scopes as defined in
372 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
373 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
375 * Level 0 - unusable SCTP addresses
376 * Level 1 - loopback address
377 * Level 2 - link-local addresses
378 * Level 3 - private addresses.
379 * Level 4 - global addresses
380 * For INIT and INIT-ACK address list, let L be the level of
381 * requested destination address, sender and receiver
382 * SHOULD include all of its addresses with level greater
383 * than or equal to L.
385 * IPv4 scoping can be controlled through sysctl option
386 * net.sctp.addr_scope_policy
388 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
390 enum sctp_scope retval;
392 /* Check for unusable SCTP addresses. */
393 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
394 retval = SCTP_SCOPE_UNUSABLE;
395 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
396 retval = SCTP_SCOPE_LOOPBACK;
397 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
398 retval = SCTP_SCOPE_LINK;
399 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
400 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
401 ipv4_is_private_192(addr->v4.sin_addr.s_addr) ||
402 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) {
403 retval = SCTP_SCOPE_PRIVATE;
405 retval = SCTP_SCOPE_GLOBAL;
411 /* Returns a valid dst cache entry for the given source and destination ip
412 * addresses. If an association is passed, trys to get a dst entry with a
413 * source address that matches an address in the bind address list.
415 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
416 struct flowi *fl, struct sock *sk)
418 struct sctp_association *asoc = t->asoc;
421 struct flowi4 *fl4 = &_fl.u.ip4;
422 struct sctp_bind_addr *bp;
423 struct sctp_sockaddr_entry *laddr;
424 struct dst_entry *dst = NULL;
425 union sctp_addr *daddr = &t->ipaddr;
426 union sctp_addr dst_saddr;
427 __u8 tos = inet_sk(sk)->tos;
429 if (t->dscp & SCTP_DSCP_SET_MASK)
430 tos = t->dscp & SCTP_DSCP_VAL_MASK;
431 memset(&_fl, 0x0, sizeof(_fl));
432 fl4->daddr = daddr->v4.sin_addr.s_addr;
433 fl4->fl4_dport = daddr->v4.sin_port;
434 fl4->flowi4_proto = IPPROTO_SCTP;
436 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
437 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
438 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
441 fl4->saddr = saddr->v4.sin_addr.s_addr;
443 fl4->fl4_sport = saddr->v4.sin_port;
446 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
449 rt = ip_route_output_key(sock_net(sk), fl4);
453 memcpy(fl, &_fl, sizeof(_fl));
456 /* If there is no association or if a source address is passed, no
457 * more validation is required.
462 bp = &asoc->base.bind_addr;
465 /* Walk through the bind address list and look for a bind
466 * address that matches the source address of the returned dst.
468 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
470 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
471 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
472 (laddr->state != SCTP_ADDR_SRC &&
473 !asoc->src_out_of_asoc_ok))
475 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
480 /* None of the bound addresses match the source address of the
481 * dst. So release it.
487 /* Walk through the bind address list and try to get a dst that
488 * matches a bind address as the source address.
491 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
492 struct net_device *odev;
496 if (laddr->state != SCTP_ADDR_SRC ||
497 AF_INET != laddr->a.sa.sa_family)
500 fl4->fl4_sport = laddr->a.v4.sin_port;
501 flowi4_update_output(fl4,
502 asoc->base.sk->sk_bound_dev_if,
503 RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
504 daddr->v4.sin_addr.s_addr,
505 laddr->a.v4.sin_addr.s_addr);
507 rt = ip_route_output_key(sock_net(sk), fl4);
511 /* Ensure the src address belongs to the output
514 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
516 if (!odev || odev->ifindex != fl4->flowi4_oif) {
520 memcpy(fl, &_fl, sizeof(_fl));
522 dst_release(&rt->dst);
530 memcpy(fl, &_fl, sizeof(_fl));
538 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
539 &fl->u.ip4.daddr, &fl->u.ip4.saddr);
542 pr_debug("no route\n");
546 /* For v4, the source address is cached in the route entry(dst). So no need
547 * to cache it separately and hence this is an empty routine.
549 static void sctp_v4_get_saddr(struct sctp_sock *sk,
550 struct sctp_transport *t,
553 union sctp_addr *saddr = &t->saddr;
554 struct rtable *rt = (struct rtable *)t->dst;
557 saddr->v4.sin_family = AF_INET;
558 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
562 /* What interface did this skb arrive on? */
563 static int sctp_v4_skb_iif(const struct sk_buff *skb)
565 return inet_iif(skb);
568 /* Was this packet marked by Explicit Congestion Notification? */
569 static int sctp_v4_is_ce(const struct sk_buff *skb)
571 return INET_ECN_is_ce(ip_hdr(skb)->tos);
574 /* Create and initialize a new sk for the socket returned by accept(). */
575 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
576 struct sctp_association *asoc,
579 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
581 struct inet_sock *newinet;
586 sock_init_data(NULL, newsk);
588 sctp_copy_sock(newsk, sk, asoc);
589 sock_reset_flag(newsk, SOCK_ZAPPED);
591 sctp_v4_copy_ip_options(sk, newsk);
593 newinet = inet_sk(newsk);
595 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
597 sk_refcnt_debug_inc(newsk);
599 if (newsk->sk_prot->init(newsk)) {
600 sk_common_release(newsk);
608 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
610 /* No address mapping for V4 sockets */
611 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
612 return sizeof(struct sockaddr_in);
615 /* Dump the v4 addr to the seq file. */
616 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
618 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
621 static void sctp_v4_ecn_capable(struct sock *sk)
626 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
628 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
629 struct sctp_sockaddr_entry *addrw, *temp;
630 struct sctp_sock *sp;
632 spin_lock_bh(&net->sctp.addr_wq_lock);
634 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
635 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
636 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
637 addrw->state, addrw);
639 #if IS_ENABLED(CONFIG_IPV6)
640 /* Now we send an ASCONF for each association */
641 /* Note. we currently don't handle link local IPv6 addressees */
642 if (addrw->a.sa.sa_family == AF_INET6) {
643 struct in6_addr *in6;
645 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
649 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
650 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
651 addrw->state == SCTP_ADDR_NEW) {
652 unsigned long timeo_val;
654 pr_debug("%s: this is on DAD, trying %d sec "
656 SCTP_ADDRESS_TICK_DELAY);
659 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
660 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
665 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
668 sk = sctp_opt2sk(sp);
669 /* ignore bound-specific endpoints */
670 if (!sctp_is_ep_boundall(sk))
673 if (sctp_asconf_mgmt(sp, addrw) < 0)
674 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
677 #if IS_ENABLED(CONFIG_IPV6)
680 list_del(&addrw->list);
683 spin_unlock_bh(&net->sctp.addr_wq_lock);
686 static void sctp_free_addr_wq(struct net *net)
688 struct sctp_sockaddr_entry *addrw;
689 struct sctp_sockaddr_entry *temp;
691 spin_lock_bh(&net->sctp.addr_wq_lock);
692 del_timer(&net->sctp.addr_wq_timer);
693 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
694 list_del(&addrw->list);
697 spin_unlock_bh(&net->sctp.addr_wq_lock);
700 /* lookup the entry for the same address in the addr_waitq
701 * sctp_addr_wq MUST be locked
703 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
704 struct sctp_sockaddr_entry *addr)
706 struct sctp_sockaddr_entry *addrw;
708 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
709 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
711 if (addrw->a.sa.sa_family == AF_INET) {
712 if (addrw->a.v4.sin_addr.s_addr ==
713 addr->a.v4.sin_addr.s_addr)
715 } else if (addrw->a.sa.sa_family == AF_INET6) {
716 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
717 &addr->a.v6.sin6_addr))
724 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
726 struct sctp_sockaddr_entry *addrw;
727 unsigned long timeo_val;
729 /* first, we check if an opposite message already exist in the queue.
730 * If we found such message, it is removed.
731 * This operation is a bit stupid, but the DHCP client attaches the
732 * new address after a couple of addition and deletion of that address
735 spin_lock_bh(&net->sctp.addr_wq_lock);
736 /* Offsets existing events in addr_wq */
737 addrw = sctp_addr_wq_lookup(net, addr);
739 if (addrw->state != cmd) {
740 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
741 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
742 &net->sctp.addr_waitq);
744 list_del(&addrw->list);
747 spin_unlock_bh(&net->sctp.addr_wq_lock);
751 /* OK, we have to add the new address to the wait queue */
752 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
754 spin_unlock_bh(&net->sctp.addr_wq_lock);
758 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
760 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
761 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
763 if (!timer_pending(&net->sctp.addr_wq_timer)) {
765 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
766 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
768 spin_unlock_bh(&net->sctp.addr_wq_lock);
771 /* Event handler for inet address addition/deletion events.
772 * The sctp_local_addr_list needs to be protocted by a spin lock since
773 * multiple notifiers (say IPv4 and IPv6) may be running at the same
774 * time and thus corrupt the list.
775 * The reader side is protected with RCU.
777 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
780 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
781 struct sctp_sockaddr_entry *addr = NULL;
782 struct sctp_sockaddr_entry *temp;
783 struct net *net = dev_net(ifa->ifa_dev->dev);
788 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
790 addr->a.v4.sin_family = AF_INET;
791 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
793 spin_lock_bh(&net->sctp.local_addr_lock);
794 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
795 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
796 spin_unlock_bh(&net->sctp.local_addr_lock);
800 spin_lock_bh(&net->sctp.local_addr_lock);
801 list_for_each_entry_safe(addr, temp,
802 &net->sctp.local_addr_list, list) {
803 if (addr->a.sa.sa_family == AF_INET &&
804 addr->a.v4.sin_addr.s_addr ==
806 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
809 list_del_rcu(&addr->list);
813 spin_unlock_bh(&net->sctp.local_addr_lock);
815 kfree_rcu(addr, rcu);
823 * Initialize the control inode/socket with a control endpoint data
824 * structure. This endpoint is reserved exclusively for the OOTB processing.
826 static int sctp_ctl_sock_init(struct net *net)
829 sa_family_t family = PF_INET;
831 if (sctp_get_pf_specific(PF_INET6))
834 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
835 SOCK_SEQPACKET, IPPROTO_SCTP, net);
837 /* If IPv6 socket could not be created, try the IPv4 socket */
838 if (err < 0 && family == PF_INET6)
839 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
840 SOCK_SEQPACKET, IPPROTO_SCTP,
844 pr_err("Failed to create the SCTP control socket\n");
850 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb)
852 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source;
854 skb_set_transport_header(skb, sizeof(struct udphdr));
859 int sctp_udp_sock_start(struct net *net)
861 struct udp_tunnel_sock_cfg tuncfg = {NULL};
862 struct udp_port_cfg udp_conf = {0};
866 udp_conf.family = AF_INET;
867 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
868 udp_conf.local_udp_port = htons(net->sctp.udp_port);
869 err = udp_sock_create(net, &udp_conf, &sock);
871 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n");
875 tuncfg.encap_type = 1;
876 tuncfg.encap_rcv = sctp_udp_rcv;
877 tuncfg.encap_err_lookup = sctp_udp_v4_err;
878 setup_udp_tunnel_sock(net, sock, &tuncfg);
879 net->sctp.udp4_sock = sock->sk;
881 #if IS_ENABLED(CONFIG_IPV6)
882 memset(&udp_conf, 0, sizeof(udp_conf));
884 udp_conf.family = AF_INET6;
885 udp_conf.local_ip6 = in6addr_any;
886 udp_conf.local_udp_port = htons(net->sctp.udp_port);
887 udp_conf.use_udp6_rx_checksums = true;
888 udp_conf.ipv6_v6only = true;
889 err = udp_sock_create(net, &udp_conf, &sock);
891 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n");
892 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
893 net->sctp.udp4_sock = NULL;
897 tuncfg.encap_type = 1;
898 tuncfg.encap_rcv = sctp_udp_rcv;
899 tuncfg.encap_err_lookup = sctp_udp_v6_err;
900 setup_udp_tunnel_sock(net, sock, &tuncfg);
901 net->sctp.udp6_sock = sock->sk;
907 void sctp_udp_sock_stop(struct net *net)
909 if (net->sctp.udp4_sock) {
910 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
911 net->sctp.udp4_sock = NULL;
913 if (net->sctp.udp6_sock) {
914 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket);
915 net->sctp.udp6_sock = NULL;
919 /* Register address family specific functions. */
920 int sctp_register_af(struct sctp_af *af)
922 switch (af->sa_family) {
924 if (sctp_af_v4_specific)
926 sctp_af_v4_specific = af;
929 if (sctp_af_v6_specific)
931 sctp_af_v6_specific = af;
937 INIT_LIST_HEAD(&af->list);
938 list_add_tail(&af->list, &sctp_address_families);
942 /* Get the table of functions for manipulating a particular address
945 struct sctp_af *sctp_get_af_specific(sa_family_t family)
949 return sctp_af_v4_specific;
951 return sctp_af_v6_specific;
957 /* Common code to initialize a AF_INET msg_name. */
958 static void sctp_inet_msgname(char *msgname, int *addr_len)
960 struct sockaddr_in *sin;
962 sin = (struct sockaddr_in *)msgname;
963 *addr_len = sizeof(struct sockaddr_in);
964 sin->sin_family = AF_INET;
965 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
968 /* Copy the primary address of the peer primary address as the msg_name. */
969 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
972 struct sockaddr_in *sin, *sinfrom;
975 struct sctp_association *asoc;
978 sctp_inet_msgname(msgname, addr_len);
979 sin = (struct sockaddr_in *)msgname;
980 sinfrom = &asoc->peer.primary_addr.v4;
981 sin->sin_port = htons(asoc->peer.port);
982 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
986 /* Initialize and copy out a msgname from an inbound skb. */
987 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
990 struct sctphdr *sh = sctp_hdr(skb);
991 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
993 sctp_inet_msgname(msgname, len);
994 sin->sin_port = sh->source;
995 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
999 /* Do we support this AF? */
1000 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
1002 /* PF_INET only supports AF_INET addresses. */
1003 return AF_INET == family;
1006 /* Address matching with wildcards allowed. */
1007 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
1008 const union sctp_addr *addr2,
1009 struct sctp_sock *opt)
1011 /* PF_INET only supports AF_INET addresses. */
1012 if (addr1->sa.sa_family != addr2->sa.sa_family)
1014 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
1015 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
1017 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
1023 /* Verify that provided sockaddr looks bindable. Common verification has
1024 * already been taken care of.
1026 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
1028 return sctp_v4_available(addr, opt);
1031 /* Verify that sockaddr looks sendable. Common verification has already
1032 * been taken care of.
1034 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
1039 /* Fill in Supported Address Type information for INIT and INIT-ACK
1040 * chunks. Returns number of addresses supported.
1042 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
1045 types[0] = SCTP_PARAM_IPV4_ADDRESS;
1049 /* Wrapper routine that calls the ip transmit routine. */
1050 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t)
1052 struct dst_entry *dst = dst_clone(t->dst);
1053 struct flowi4 *fl4 = &t->fl.u.ip4;
1054 struct sock *sk = skb->sk;
1055 struct inet_sock *inet = inet_sk(sk);
1056 __u8 dscp = inet->tos;
1059 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
1060 skb->len, &fl4->saddr, &fl4->daddr);
1062 if (t->dscp & SCTP_DSCP_SET_MASK)
1063 dscp = t->dscp & SCTP_DSCP_VAL_MASK;
1065 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO
1067 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS);
1069 if (!t->encap_port || !sctp_sk(sk)->udp_port) {
1070 skb_dst_set(skb, dst);
1071 return __ip_queue_xmit(sk, skb, &t->fl, dscp);
1074 if (skb_is_gso(skb))
1075 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
1077 if (ip_dont_fragment(sk, dst) && !skb->ignore_df)
1080 skb->encapsulation = 1;
1081 skb_reset_inner_mac_header(skb);
1082 skb_reset_inner_transport_header(skb);
1083 skb_set_inner_ipproto(skb, IPPROTO_SCTP);
1084 udp_tunnel_xmit_skb((struct rtable *)dst, sk, skb, fl4->saddr,
1085 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df,
1086 sctp_sk(sk)->udp_port, t->encap_port, false, false);
1090 static struct sctp_af sctp_af_inet;
1092 static struct sctp_pf sctp_pf_inet = {
1093 .event_msgname = sctp_inet_event_msgname,
1094 .skb_msgname = sctp_inet_skb_msgname,
1095 .af_supported = sctp_inet_af_supported,
1096 .cmp_addr = sctp_inet_cmp_addr,
1097 .bind_verify = sctp_inet_bind_verify,
1098 .send_verify = sctp_inet_send_verify,
1099 .supported_addrs = sctp_inet_supported_addrs,
1100 .create_accept_sk = sctp_v4_create_accept_sk,
1101 .addr_to_user = sctp_v4_addr_to_user,
1102 .to_sk_saddr = sctp_v4_to_sk_saddr,
1103 .to_sk_daddr = sctp_v4_to_sk_daddr,
1104 .copy_ip_options = sctp_v4_copy_ip_options,
1108 /* Notifier for inetaddr addition/deletion events. */
1109 static struct notifier_block sctp_inetaddr_notifier = {
1110 .notifier_call = sctp_inetaddr_event,
1113 /* Socket operations. */
1114 static const struct proto_ops inet_seqpacket_ops = {
1116 .owner = THIS_MODULE,
1117 .release = inet_release, /* Needs to be wrapped... */
1119 .connect = sctp_inet_connect,
1120 .socketpair = sock_no_socketpair,
1121 .accept = inet_accept,
1122 .getname = inet_getname, /* Semantics are different. */
1124 .ioctl = inet_ioctl,
1125 .gettstamp = sock_gettstamp,
1126 .listen = sctp_inet_listen,
1127 .shutdown = inet_shutdown, /* Looks harmless. */
1128 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1129 .getsockopt = sock_common_getsockopt,
1130 .sendmsg = inet_sendmsg,
1131 .recvmsg = inet_recvmsg,
1132 .mmap = sock_no_mmap,
1133 .sendpage = sock_no_sendpage,
1136 /* Registration with AF_INET family. */
1137 static struct inet_protosw sctp_seqpacket_protosw = {
1138 .type = SOCK_SEQPACKET,
1139 .protocol = IPPROTO_SCTP,
1141 .ops = &inet_seqpacket_ops,
1142 .flags = SCTP_PROTOSW_FLAG
1144 static struct inet_protosw sctp_stream_protosw = {
1145 .type = SOCK_STREAM,
1146 .protocol = IPPROTO_SCTP,
1148 .ops = &inet_seqpacket_ops,
1149 .flags = SCTP_PROTOSW_FLAG
1152 static int sctp4_rcv(struct sk_buff *skb)
1154 SCTP_INPUT_CB(skb)->encap_port = 0;
1155 return sctp_rcv(skb);
1158 /* Register with IP layer. */
1159 static const struct net_protocol sctp_protocol = {
1160 .handler = sctp4_rcv,
1161 .err_handler = sctp_v4_err,
1163 .icmp_strict_tag_validation = 1,
1166 /* IPv4 address related functions. */
1167 static struct sctp_af sctp_af_inet = {
1168 .sa_family = AF_INET,
1169 .sctp_xmit = sctp_v4_xmit,
1170 .setsockopt = ip_setsockopt,
1171 .getsockopt = ip_getsockopt,
1172 .get_dst = sctp_v4_get_dst,
1173 .get_saddr = sctp_v4_get_saddr,
1174 .copy_addrlist = sctp_v4_copy_addrlist,
1175 .from_skb = sctp_v4_from_skb,
1176 .from_sk = sctp_v4_from_sk,
1177 .from_addr_param = sctp_v4_from_addr_param,
1178 .to_addr_param = sctp_v4_to_addr_param,
1179 .cmp_addr = sctp_v4_cmp_addr,
1180 .addr_valid = sctp_v4_addr_valid,
1181 .inaddr_any = sctp_v4_inaddr_any,
1182 .is_any = sctp_v4_is_any,
1183 .available = sctp_v4_available,
1184 .scope = sctp_v4_scope,
1185 .skb_iif = sctp_v4_skb_iif,
1186 .is_ce = sctp_v4_is_ce,
1187 .seq_dump_addr = sctp_v4_seq_dump_addr,
1188 .ecn_capable = sctp_v4_ecn_capable,
1189 .net_header_len = sizeof(struct iphdr),
1190 .sockaddr_len = sizeof(struct sockaddr_in),
1191 .ip_options_len = sctp_v4_ip_options_len,
1194 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1198 return sctp_pf_inet_specific;
1200 return sctp_pf_inet6_specific;
1206 /* Register the PF specific function table. */
1207 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1211 if (sctp_pf_inet_specific)
1213 sctp_pf_inet_specific = pf;
1216 if (sctp_pf_inet6_specific)
1218 sctp_pf_inet6_specific = pf;
1226 static inline int init_sctp_mibs(struct net *net)
1228 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1229 if (!net->sctp.sctp_statistics)
1234 static inline void cleanup_sctp_mibs(struct net *net)
1236 free_percpu(net->sctp.sctp_statistics);
1239 static void sctp_v4_pf_init(void)
1241 /* Initialize the SCTP specific PF functions. */
1242 sctp_register_pf(&sctp_pf_inet, PF_INET);
1243 sctp_register_af(&sctp_af_inet);
1246 static void sctp_v4_pf_exit(void)
1248 list_del(&sctp_af_inet.list);
1251 static int sctp_v4_protosw_init(void)
1255 rc = proto_register(&sctp_prot, 1);
1259 /* Register SCTP(UDP and TCP style) with socket layer. */
1260 inet_register_protosw(&sctp_seqpacket_protosw);
1261 inet_register_protosw(&sctp_stream_protosw);
1266 static void sctp_v4_protosw_exit(void)
1268 inet_unregister_protosw(&sctp_stream_protosw);
1269 inet_unregister_protosw(&sctp_seqpacket_protosw);
1270 proto_unregister(&sctp_prot);
1273 static int sctp_v4_add_protocol(void)
1275 /* Register notifier for inet address additions/deletions. */
1276 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1278 /* Register SCTP with inet layer. */
1279 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1285 static void sctp_v4_del_protocol(void)
1287 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1288 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1291 static int __net_init sctp_defaults_init(struct net *net)
1296 * 14. Suggested SCTP Protocol Parameter Values
1298 /* The following protocol parameters are RECOMMENDED: */
1299 /* RTO.Initial - 3 seconds */
1300 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1301 /* RTO.Min - 1 second */
1302 net->sctp.rto_min = SCTP_RTO_MIN;
1303 /* RTO.Max - 60 seconds */
1304 net->sctp.rto_max = SCTP_RTO_MAX;
1305 /* RTO.Alpha - 1/8 */
1306 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1307 /* RTO.Beta - 1/4 */
1308 net->sctp.rto_beta = SCTP_RTO_BETA;
1310 /* Valid.Cookie.Life - 60 seconds */
1311 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1313 /* Whether Cookie Preservative is enabled(1) or not(0) */
1314 net->sctp.cookie_preserve_enable = 1;
1316 /* Default sctp sockets to use md5 as their hmac alg */
1317 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1318 net->sctp.sctp_hmac_alg = "md5";
1319 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1320 net->sctp.sctp_hmac_alg = "sha1";
1322 net->sctp.sctp_hmac_alg = NULL;
1326 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1328 /* Disable of Primary Path Switchover by default */
1329 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1331 /* Enable pf state by default */
1332 net->sctp.pf_enable = 1;
1334 /* Ignore pf exposure feature by default */
1335 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1337 /* Association.Max.Retrans - 10 attempts
1338 * Path.Max.Retrans - 5 attempts (per destination address)
1339 * Max.Init.Retransmits - 8 attempts
1341 net->sctp.max_retrans_association = 10;
1342 net->sctp.max_retrans_path = 5;
1343 net->sctp.max_retrans_init = 8;
1345 /* Sendbuffer growth - do per-socket accounting */
1346 net->sctp.sndbuf_policy = 0;
1348 /* Rcvbuffer growth - do per-socket accounting */
1349 net->sctp.rcvbuf_policy = 0;
1351 /* HB.interval - 30 seconds */
1352 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1354 /* delayed SACK timeout */
1355 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1357 /* Disable ADDIP by default. */
1358 net->sctp.addip_enable = 0;
1359 net->sctp.addip_noauth = 0;
1360 net->sctp.default_auto_asconf = 0;
1362 /* Enable PR-SCTP by default. */
1363 net->sctp.prsctp_enable = 1;
1365 /* Disable RECONF by default. */
1366 net->sctp.reconf_enable = 0;
1368 /* Disable AUTH by default. */
1369 net->sctp.auth_enable = 0;
1371 /* Enable ECN by default. */
1372 net->sctp.ecn_enable = 1;
1374 /* Set UDP tunneling listening port to 0 by default */
1375 net->sctp.udp_port = 0;
1377 /* Set remote encap port to 0 by default */
1378 net->sctp.encap_port = 0;
1380 /* Set SCOPE policy to enabled */
1381 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1383 /* Set the default rwnd update threshold */
1384 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1386 /* Initialize maximum autoclose timeout. */
1387 net->sctp.max_autoclose = INT_MAX / HZ;
1389 status = sctp_sysctl_net_register(net);
1391 goto err_sysctl_register;
1393 /* Allocate and initialise sctp mibs. */
1394 status = init_sctp_mibs(net);
1398 #ifdef CONFIG_PROC_FS
1399 /* Initialize proc fs directory. */
1400 status = sctp_proc_init(net);
1405 sctp_dbg_objcnt_init(net);
1407 /* Initialize the local address list. */
1408 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1409 spin_lock_init(&net->sctp.local_addr_lock);
1410 sctp_get_local_addr_list(net);
1412 /* Initialize the address event list */
1413 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1414 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1415 spin_lock_init(&net->sctp.addr_wq_lock);
1416 net->sctp.addr_wq_timer.expires = 0;
1417 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1421 #ifdef CONFIG_PROC_FS
1423 cleanup_sctp_mibs(net);
1426 sctp_sysctl_net_unregister(net);
1427 err_sysctl_register:
1431 static void __net_exit sctp_defaults_exit(struct net *net)
1433 /* Free the local address list */
1434 sctp_free_addr_wq(net);
1435 sctp_free_local_addr_list(net);
1437 #ifdef CONFIG_PROC_FS
1438 remove_proc_subtree("sctp", net->proc_net);
1439 net->sctp.proc_net_sctp = NULL;
1441 cleanup_sctp_mibs(net);
1442 sctp_sysctl_net_unregister(net);
1445 static struct pernet_operations sctp_defaults_ops = {
1446 .init = sctp_defaults_init,
1447 .exit = sctp_defaults_exit,
1450 static int __net_init sctp_ctrlsock_init(struct net *net)
1454 /* Initialize the control inode/socket for handling OOTB packets. */
1455 status = sctp_ctl_sock_init(net);
1457 pr_err("Failed to initialize the SCTP control sock\n");
1462 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1464 /* Free the control endpoint. */
1465 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1468 static struct pernet_operations sctp_ctrlsock_ops = {
1469 .init = sctp_ctrlsock_init,
1470 .exit = sctp_ctrlsock_exit,
1473 /* Initialize the universe into something sensible. */
1474 static __init int sctp_init(void)
1476 unsigned long nr_pages = totalram_pages();
1477 unsigned long limit;
1479 int max_entry_order;
1486 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1488 /* Allocate bind_bucket and chunk caches. */
1490 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1491 sizeof(struct sctp_bind_bucket),
1492 0, SLAB_HWCACHE_ALIGN,
1494 if (!sctp_bucket_cachep)
1497 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1498 sizeof(struct sctp_chunk),
1499 0, SLAB_HWCACHE_ALIGN,
1501 if (!sctp_chunk_cachep)
1502 goto err_chunk_cachep;
1504 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1506 goto err_percpu_counter_init;
1508 /* Implementation specific variables. */
1510 /* Initialize default stream count setup information. */
1511 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1512 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1514 /* Initialize handle used for association ids. */
1515 idr_init(&sctp_assocs_id);
1517 limit = nr_free_buffer_pages() / 8;
1518 limit = max(limit, 128UL);
1519 sysctl_sctp_mem[0] = limit / 4 * 3;
1520 sysctl_sctp_mem[1] = limit;
1521 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1523 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1524 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1525 max_share = min(4UL*1024*1024, limit);
1527 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1528 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1529 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1531 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1532 sysctl_sctp_wmem[1] = 16*1024;
1533 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1535 /* Size and allocate the association hash table.
1536 * The methodology is similar to that of the tcp hash tables.
1537 * Though not identical. Start by getting a goal size
1539 if (nr_pages >= (128 * 1024))
1540 goal = nr_pages >> (22 - PAGE_SHIFT);
1542 goal = nr_pages >> (24 - PAGE_SHIFT);
1544 /* Then compute the page order for said goal */
1545 order = get_order(goal);
1547 /* Now compute the required page order for the maximum sized table we
1550 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1551 sizeof(struct sctp_bind_hashbucket));
1553 /* Limit the page order by that maximum hash table size */
1554 order = min(order, max_entry_order);
1556 /* Allocate and initialize the endpoint hash table. */
1557 sctp_ep_hashsize = 64;
1559 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1560 if (!sctp_ep_hashtable) {
1561 pr_err("Failed endpoint_hash alloc\n");
1563 goto err_ehash_alloc;
1565 for (i = 0; i < sctp_ep_hashsize; i++) {
1566 rwlock_init(&sctp_ep_hashtable[i].lock);
1567 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1570 /* Allocate and initialize the SCTP port hash table.
1571 * Note that order is initalized to start at the max sized
1572 * table we want to support. If we can't get that many pages
1573 * reduce the order and try again
1576 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1577 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1578 } while (!sctp_port_hashtable && --order > 0);
1580 if (!sctp_port_hashtable) {
1581 pr_err("Failed bind hash alloc\n");
1583 goto err_bhash_alloc;
1586 /* Now compute the number of entries that will fit in the
1587 * port hash space we allocated
1589 num_entries = (1UL << order) * PAGE_SIZE /
1590 sizeof(struct sctp_bind_hashbucket);
1592 /* And finish by rounding it down to the nearest power of two.
1593 * This wastes some memory of course, but it's needed because
1594 * the hash function operates based on the assumption that
1595 * the number of entries is a power of two.
1597 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1599 for (i = 0; i < sctp_port_hashsize; i++) {
1600 spin_lock_init(&sctp_port_hashtable[i].lock);
1601 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1604 status = sctp_transport_hashtable_init();
1606 goto err_thash_alloc;
1608 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1611 sctp_sysctl_register();
1613 INIT_LIST_HEAD(&sctp_address_families);
1616 sctp_sched_ops_init();
1618 status = register_pernet_subsys(&sctp_defaults_ops);
1620 goto err_register_defaults;
1622 status = sctp_v4_protosw_init();
1624 goto err_protosw_init;
1626 status = sctp_v6_protosw_init();
1628 goto err_v6_protosw_init;
1630 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1632 goto err_register_ctrlsock;
1634 status = sctp_v4_add_protocol();
1636 goto err_add_protocol;
1638 /* Register SCTP with inet6 layer. */
1639 status = sctp_v6_add_protocol();
1641 goto err_v6_add_protocol;
1643 if (sctp_offload_init() < 0)
1644 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1648 err_v6_add_protocol:
1649 sctp_v4_del_protocol();
1651 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1652 err_register_ctrlsock:
1653 sctp_v6_protosw_exit();
1654 err_v6_protosw_init:
1655 sctp_v4_protosw_exit();
1657 unregister_pernet_subsys(&sctp_defaults_ops);
1658 err_register_defaults:
1661 sctp_sysctl_unregister();
1662 free_pages((unsigned long)sctp_port_hashtable,
1663 get_order(sctp_port_hashsize *
1664 sizeof(struct sctp_bind_hashbucket)));
1666 sctp_transport_hashtable_destroy();
1668 kfree(sctp_ep_hashtable);
1670 percpu_counter_destroy(&sctp_sockets_allocated);
1671 err_percpu_counter_init:
1672 kmem_cache_destroy(sctp_chunk_cachep);
1674 kmem_cache_destroy(sctp_bucket_cachep);
1678 /* Exit handler for the SCTP protocol. */
1679 static __exit void sctp_exit(void)
1681 /* BUG. This should probably do something useful like clean
1682 * up all the remaining associations and all that memory.
1685 /* Unregister with inet6/inet layers. */
1686 sctp_v6_del_protocol();
1687 sctp_v4_del_protocol();
1689 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1691 /* Free protosw registrations */
1692 sctp_v6_protosw_exit();
1693 sctp_v4_protosw_exit();
1695 unregister_pernet_subsys(&sctp_defaults_ops);
1697 /* Unregister with socket layer. */
1701 sctp_sysctl_unregister();
1703 free_pages((unsigned long)sctp_port_hashtable,
1704 get_order(sctp_port_hashsize *
1705 sizeof(struct sctp_bind_hashbucket)));
1706 kfree(sctp_ep_hashtable);
1707 sctp_transport_hashtable_destroy();
1709 percpu_counter_destroy(&sctp_sockets_allocated);
1711 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1713 kmem_cache_destroy(sctp_chunk_cachep);
1714 kmem_cache_destroy(sctp_bucket_cachep);
1717 module_init(sctp_init);
1718 module_exit(sctp_exit);
1721 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1723 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1724 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1725 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1726 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1727 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1728 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1729 MODULE_LICENSE("GPL");