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 net *net = sock_net(&sp->inet.sk);
355 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
358 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
360 !sp->inet.freebind &&
361 !net->ipv4.sysctl_ip_nonlocal_bind)
364 if (ipv6_only_sock(sctp_opt2sk(sp)))
370 /* Checking the loopback, private and other address scopes as defined in
371 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
372 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
374 * Level 0 - unusable SCTP addresses
375 * Level 1 - loopback address
376 * Level 2 - link-local addresses
377 * Level 3 - private addresses.
378 * Level 4 - global addresses
379 * For INIT and INIT-ACK address list, let L be the level of
380 * requested destination address, sender and receiver
381 * SHOULD include all of its addresses with level greater
382 * than or equal to L.
384 * IPv4 scoping can be controlled through sysctl option
385 * net.sctp.addr_scope_policy
387 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
389 enum sctp_scope retval;
391 /* Check for unusable SCTP addresses. */
392 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
393 retval = SCTP_SCOPE_UNUSABLE;
394 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
395 retval = SCTP_SCOPE_LOOPBACK;
396 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
397 retval = SCTP_SCOPE_LINK;
398 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
399 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
400 ipv4_is_private_192(addr->v4.sin_addr.s_addr) ||
401 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) {
402 retval = SCTP_SCOPE_PRIVATE;
404 retval = SCTP_SCOPE_GLOBAL;
410 /* Returns a valid dst cache entry for the given source and destination ip
411 * addresses. If an association is passed, trys to get a dst entry with a
412 * source address that matches an address in the bind address list.
414 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
415 struct flowi *fl, struct sock *sk)
417 struct sctp_association *asoc = t->asoc;
420 struct flowi4 *fl4 = &_fl.u.ip4;
421 struct sctp_bind_addr *bp;
422 struct sctp_sockaddr_entry *laddr;
423 struct dst_entry *dst = NULL;
424 union sctp_addr *daddr = &t->ipaddr;
425 union sctp_addr dst_saddr;
426 __u8 tos = inet_sk(sk)->tos;
428 if (t->dscp & SCTP_DSCP_SET_MASK)
429 tos = t->dscp & SCTP_DSCP_VAL_MASK;
430 memset(&_fl, 0x0, sizeof(_fl));
431 fl4->daddr = daddr->v4.sin_addr.s_addr;
432 fl4->fl4_dport = daddr->v4.sin_port;
433 fl4->flowi4_proto = IPPROTO_SCTP;
435 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
436 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
437 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
440 fl4->saddr = saddr->v4.sin_addr.s_addr;
442 fl4->fl4_sport = saddr->v4.sin_port;
445 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
448 rt = ip_route_output_key(sock_net(sk), fl4);
452 memcpy(fl, &_fl, sizeof(_fl));
455 /* If there is no association or if a source address is passed, no
456 * more validation is required.
461 bp = &asoc->base.bind_addr;
464 /* Walk through the bind address list and look for a bind
465 * address that matches the source address of the returned dst.
467 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
469 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
470 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
471 (laddr->state != SCTP_ADDR_SRC &&
472 !asoc->src_out_of_asoc_ok))
474 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
479 /* None of the bound addresses match the source address of the
480 * dst. So release it.
486 /* Walk through the bind address list and try to get a dst that
487 * matches a bind address as the source address.
490 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
491 struct net_device *odev;
495 if (laddr->state != SCTP_ADDR_SRC ||
496 AF_INET != laddr->a.sa.sa_family)
499 fl4->fl4_sport = laddr->a.v4.sin_port;
500 flowi4_update_output(fl4,
501 asoc->base.sk->sk_bound_dev_if,
502 RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
503 daddr->v4.sin_addr.s_addr,
504 laddr->a.v4.sin_addr.s_addr);
506 rt = ip_route_output_key(sock_net(sk), fl4);
510 /* Ensure the src address belongs to the output
513 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
515 if (!odev || odev->ifindex != fl4->flowi4_oif) {
519 memcpy(fl, &_fl, sizeof(_fl));
521 dst_release(&rt->dst);
529 memcpy(fl, &_fl, sizeof(_fl));
537 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
538 &fl->u.ip4.daddr, &fl->u.ip4.saddr);
541 pr_debug("no route\n");
545 /* For v4, the source address is cached in the route entry(dst). So no need
546 * to cache it separately and hence this is an empty routine.
548 static void sctp_v4_get_saddr(struct sctp_sock *sk,
549 struct sctp_transport *t,
552 union sctp_addr *saddr = &t->saddr;
553 struct rtable *rt = (struct rtable *)t->dst;
556 saddr->v4.sin_family = AF_INET;
557 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
561 /* What interface did this skb arrive on? */
562 static int sctp_v4_skb_iif(const struct sk_buff *skb)
564 return inet_iif(skb);
567 /* Was this packet marked by Explicit Congestion Notification? */
568 static int sctp_v4_is_ce(const struct sk_buff *skb)
570 return INET_ECN_is_ce(ip_hdr(skb)->tos);
573 /* Create and initialize a new sk for the socket returned by accept(). */
574 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
575 struct sctp_association *asoc,
578 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
580 struct inet_sock *newinet;
585 sock_init_data(NULL, newsk);
587 sctp_copy_sock(newsk, sk, asoc);
588 sock_reset_flag(newsk, SOCK_ZAPPED);
590 sctp_v4_copy_ip_options(sk, newsk);
592 newinet = inet_sk(newsk);
594 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
596 sk_refcnt_debug_inc(newsk);
598 if (newsk->sk_prot->init(newsk)) {
599 sk_common_release(newsk);
607 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
609 /* No address mapping for V4 sockets */
610 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
611 return sizeof(struct sockaddr_in);
614 /* Dump the v4 addr to the seq file. */
615 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
617 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
620 static void sctp_v4_ecn_capable(struct sock *sk)
625 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
627 struct net *net = from_timer(net, t, sctp.addr_wq_timer);
628 struct sctp_sockaddr_entry *addrw, *temp;
629 struct sctp_sock *sp;
631 spin_lock_bh(&net->sctp.addr_wq_lock);
633 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
634 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
635 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
636 addrw->state, addrw);
638 #if IS_ENABLED(CONFIG_IPV6)
639 /* Now we send an ASCONF for each association */
640 /* Note. we currently don't handle link local IPv6 addressees */
641 if (addrw->a.sa.sa_family == AF_INET6) {
642 struct in6_addr *in6;
644 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
648 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
649 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
650 addrw->state == SCTP_ADDR_NEW) {
651 unsigned long timeo_val;
653 pr_debug("%s: this is on DAD, trying %d sec "
655 SCTP_ADDRESS_TICK_DELAY);
658 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
659 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
664 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
667 sk = sctp_opt2sk(sp);
668 /* ignore bound-specific endpoints */
669 if (!sctp_is_ep_boundall(sk))
672 if (sctp_asconf_mgmt(sp, addrw) < 0)
673 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
676 #if IS_ENABLED(CONFIG_IPV6)
679 list_del(&addrw->list);
682 spin_unlock_bh(&net->sctp.addr_wq_lock);
685 static void sctp_free_addr_wq(struct net *net)
687 struct sctp_sockaddr_entry *addrw;
688 struct sctp_sockaddr_entry *temp;
690 spin_lock_bh(&net->sctp.addr_wq_lock);
691 del_timer(&net->sctp.addr_wq_timer);
692 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
693 list_del(&addrw->list);
696 spin_unlock_bh(&net->sctp.addr_wq_lock);
699 /* lookup the entry for the same address in the addr_waitq
700 * sctp_addr_wq MUST be locked
702 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
703 struct sctp_sockaddr_entry *addr)
705 struct sctp_sockaddr_entry *addrw;
707 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
708 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
710 if (addrw->a.sa.sa_family == AF_INET) {
711 if (addrw->a.v4.sin_addr.s_addr ==
712 addr->a.v4.sin_addr.s_addr)
714 } else if (addrw->a.sa.sa_family == AF_INET6) {
715 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
716 &addr->a.v6.sin6_addr))
723 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
725 struct sctp_sockaddr_entry *addrw;
726 unsigned long timeo_val;
728 /* first, we check if an opposite message already exist in the queue.
729 * If we found such message, it is removed.
730 * This operation is a bit stupid, but the DHCP client attaches the
731 * new address after a couple of addition and deletion of that address
734 spin_lock_bh(&net->sctp.addr_wq_lock);
735 /* Offsets existing events in addr_wq */
736 addrw = sctp_addr_wq_lookup(net, addr);
738 if (addrw->state != cmd) {
739 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
740 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
741 &net->sctp.addr_waitq);
743 list_del(&addrw->list);
746 spin_unlock_bh(&net->sctp.addr_wq_lock);
750 /* OK, we have to add the new address to the wait queue */
751 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
753 spin_unlock_bh(&net->sctp.addr_wq_lock);
757 list_add_tail(&addrw->list, &net->sctp.addr_waitq);
759 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
760 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
762 if (!timer_pending(&net->sctp.addr_wq_timer)) {
764 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
765 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
767 spin_unlock_bh(&net->sctp.addr_wq_lock);
770 /* Event handler for inet address addition/deletion events.
771 * The sctp_local_addr_list needs to be protocted by a spin lock since
772 * multiple notifiers (say IPv4 and IPv6) may be running at the same
773 * time and thus corrupt the list.
774 * The reader side is protected with RCU.
776 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
779 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
780 struct sctp_sockaddr_entry *addr = NULL;
781 struct sctp_sockaddr_entry *temp;
782 struct net *net = dev_net(ifa->ifa_dev->dev);
787 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
789 addr->a.v4.sin_family = AF_INET;
790 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
792 spin_lock_bh(&net->sctp.local_addr_lock);
793 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
794 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
795 spin_unlock_bh(&net->sctp.local_addr_lock);
799 spin_lock_bh(&net->sctp.local_addr_lock);
800 list_for_each_entry_safe(addr, temp,
801 &net->sctp.local_addr_list, list) {
802 if (addr->a.sa.sa_family == AF_INET &&
803 addr->a.v4.sin_addr.s_addr ==
805 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
808 list_del_rcu(&addr->list);
812 spin_unlock_bh(&net->sctp.local_addr_lock);
814 kfree_rcu(addr, rcu);
822 * Initialize the control inode/socket with a control endpoint data
823 * structure. This endpoint is reserved exclusively for the OOTB processing.
825 static int sctp_ctl_sock_init(struct net *net)
828 sa_family_t family = PF_INET;
830 if (sctp_get_pf_specific(PF_INET6))
833 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
834 SOCK_SEQPACKET, IPPROTO_SCTP, net);
836 /* If IPv6 socket could not be created, try the IPv4 socket */
837 if (err < 0 && family == PF_INET6)
838 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
839 SOCK_SEQPACKET, IPPROTO_SCTP,
843 pr_err("Failed to create the SCTP control socket\n");
849 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb)
851 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source;
853 skb_set_transport_header(skb, sizeof(struct udphdr));
858 int sctp_udp_sock_start(struct net *net)
860 struct udp_tunnel_sock_cfg tuncfg = {NULL};
861 struct udp_port_cfg udp_conf = {0};
865 udp_conf.family = AF_INET;
866 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
867 udp_conf.local_udp_port = htons(net->sctp.udp_port);
868 err = udp_sock_create(net, &udp_conf, &sock);
870 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n");
874 tuncfg.encap_type = 1;
875 tuncfg.encap_rcv = sctp_udp_rcv;
876 tuncfg.encap_err_lookup = sctp_udp_v4_err;
877 setup_udp_tunnel_sock(net, sock, &tuncfg);
878 net->sctp.udp4_sock = sock->sk;
880 #if IS_ENABLED(CONFIG_IPV6)
881 memset(&udp_conf, 0, sizeof(udp_conf));
883 udp_conf.family = AF_INET6;
884 udp_conf.local_ip6 = in6addr_any;
885 udp_conf.local_udp_port = htons(net->sctp.udp_port);
886 udp_conf.use_udp6_rx_checksums = true;
887 udp_conf.ipv6_v6only = true;
888 err = udp_sock_create(net, &udp_conf, &sock);
890 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n");
891 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
892 net->sctp.udp4_sock = NULL;
896 tuncfg.encap_type = 1;
897 tuncfg.encap_rcv = sctp_udp_rcv;
898 tuncfg.encap_err_lookup = sctp_udp_v6_err;
899 setup_udp_tunnel_sock(net, sock, &tuncfg);
900 net->sctp.udp6_sock = sock->sk;
906 void sctp_udp_sock_stop(struct net *net)
908 if (net->sctp.udp4_sock) {
909 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
910 net->sctp.udp4_sock = NULL;
912 if (net->sctp.udp6_sock) {
913 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket);
914 net->sctp.udp6_sock = NULL;
918 /* Register address family specific functions. */
919 int sctp_register_af(struct sctp_af *af)
921 switch (af->sa_family) {
923 if (sctp_af_v4_specific)
925 sctp_af_v4_specific = af;
928 if (sctp_af_v6_specific)
930 sctp_af_v6_specific = af;
936 INIT_LIST_HEAD(&af->list);
937 list_add_tail(&af->list, &sctp_address_families);
941 /* Get the table of functions for manipulating a particular address
944 struct sctp_af *sctp_get_af_specific(sa_family_t family)
948 return sctp_af_v4_specific;
950 return sctp_af_v6_specific;
956 /* Common code to initialize a AF_INET msg_name. */
957 static void sctp_inet_msgname(char *msgname, int *addr_len)
959 struct sockaddr_in *sin;
961 sin = (struct sockaddr_in *)msgname;
962 *addr_len = sizeof(struct sockaddr_in);
963 sin->sin_family = AF_INET;
964 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
967 /* Copy the primary address of the peer primary address as the msg_name. */
968 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
971 struct sockaddr_in *sin, *sinfrom;
974 struct sctp_association *asoc;
977 sctp_inet_msgname(msgname, addr_len);
978 sin = (struct sockaddr_in *)msgname;
979 sinfrom = &asoc->peer.primary_addr.v4;
980 sin->sin_port = htons(asoc->peer.port);
981 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
985 /* Initialize and copy out a msgname from an inbound skb. */
986 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
989 struct sctphdr *sh = sctp_hdr(skb);
990 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
992 sctp_inet_msgname(msgname, len);
993 sin->sin_port = sh->source;
994 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
998 /* Do we support this AF? */
999 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
1001 /* PF_INET only supports AF_INET addresses. */
1002 return AF_INET == family;
1005 /* Address matching with wildcards allowed. */
1006 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
1007 const union sctp_addr *addr2,
1008 struct sctp_sock *opt)
1010 /* PF_INET only supports AF_INET addresses. */
1011 if (addr1->sa.sa_family != addr2->sa.sa_family)
1013 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
1014 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
1016 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
1022 /* Verify that provided sockaddr looks bindable. Common verification has
1023 * already been taken care of.
1025 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
1027 return sctp_v4_available(addr, opt);
1030 /* Verify that sockaddr looks sendable. Common verification has already
1031 * been taken care of.
1033 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
1038 /* Fill in Supported Address Type information for INIT and INIT-ACK
1039 * chunks. Returns number of addresses supported.
1041 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
1044 types[0] = SCTP_PARAM_IPV4_ADDRESS;
1048 /* Wrapper routine that calls the ip transmit routine. */
1049 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t)
1051 struct dst_entry *dst = dst_clone(t->dst);
1052 struct flowi4 *fl4 = &t->fl.u.ip4;
1053 struct sock *sk = skb->sk;
1054 struct inet_sock *inet = inet_sk(sk);
1055 __u8 dscp = inet->tos;
1058 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
1059 skb->len, &fl4->saddr, &fl4->daddr);
1061 if (t->dscp & SCTP_DSCP_SET_MASK)
1062 dscp = t->dscp & SCTP_DSCP_VAL_MASK;
1064 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO
1066 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS);
1068 if (!t->encap_port || !sctp_sk(sk)->udp_port) {
1069 skb_dst_set(skb, dst);
1070 return __ip_queue_xmit(sk, skb, &t->fl, dscp);
1073 if (skb_is_gso(skb))
1074 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
1076 if (ip_dont_fragment(sk, dst) && !skb->ignore_df)
1079 skb->encapsulation = 1;
1080 skb_reset_inner_mac_header(skb);
1081 skb_reset_inner_transport_header(skb);
1082 skb_set_inner_ipproto(skb, IPPROTO_SCTP);
1083 udp_tunnel_xmit_skb((struct rtable *)dst, sk, skb, fl4->saddr,
1084 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df,
1085 sctp_sk(sk)->udp_port, t->encap_port, false, false);
1089 static struct sctp_af sctp_af_inet;
1091 static struct sctp_pf sctp_pf_inet = {
1092 .event_msgname = sctp_inet_event_msgname,
1093 .skb_msgname = sctp_inet_skb_msgname,
1094 .af_supported = sctp_inet_af_supported,
1095 .cmp_addr = sctp_inet_cmp_addr,
1096 .bind_verify = sctp_inet_bind_verify,
1097 .send_verify = sctp_inet_send_verify,
1098 .supported_addrs = sctp_inet_supported_addrs,
1099 .create_accept_sk = sctp_v4_create_accept_sk,
1100 .addr_to_user = sctp_v4_addr_to_user,
1101 .to_sk_saddr = sctp_v4_to_sk_saddr,
1102 .to_sk_daddr = sctp_v4_to_sk_daddr,
1103 .copy_ip_options = sctp_v4_copy_ip_options,
1107 /* Notifier for inetaddr addition/deletion events. */
1108 static struct notifier_block sctp_inetaddr_notifier = {
1109 .notifier_call = sctp_inetaddr_event,
1112 /* Socket operations. */
1113 static const struct proto_ops inet_seqpacket_ops = {
1115 .owner = THIS_MODULE,
1116 .release = inet_release, /* Needs to be wrapped... */
1118 .connect = sctp_inet_connect,
1119 .socketpair = sock_no_socketpair,
1120 .accept = inet_accept,
1121 .getname = inet_getname, /* Semantics are different. */
1123 .ioctl = inet_ioctl,
1124 .gettstamp = sock_gettstamp,
1125 .listen = sctp_inet_listen,
1126 .shutdown = inet_shutdown, /* Looks harmless. */
1127 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1128 .getsockopt = sock_common_getsockopt,
1129 .sendmsg = inet_sendmsg,
1130 .recvmsg = inet_recvmsg,
1131 .mmap = sock_no_mmap,
1132 .sendpage = sock_no_sendpage,
1135 /* Registration with AF_INET family. */
1136 static struct inet_protosw sctp_seqpacket_protosw = {
1137 .type = SOCK_SEQPACKET,
1138 .protocol = IPPROTO_SCTP,
1140 .ops = &inet_seqpacket_ops,
1141 .flags = SCTP_PROTOSW_FLAG
1143 static struct inet_protosw sctp_stream_protosw = {
1144 .type = SOCK_STREAM,
1145 .protocol = IPPROTO_SCTP,
1147 .ops = &inet_seqpacket_ops,
1148 .flags = SCTP_PROTOSW_FLAG
1151 static int sctp4_rcv(struct sk_buff *skb)
1153 SCTP_INPUT_CB(skb)->encap_port = 0;
1154 return sctp_rcv(skb);
1157 /* Register with IP layer. */
1158 static const struct net_protocol sctp_protocol = {
1159 .handler = sctp4_rcv,
1160 .err_handler = sctp_v4_err,
1162 .icmp_strict_tag_validation = 1,
1165 /* IPv4 address related functions. */
1166 static struct sctp_af sctp_af_inet = {
1167 .sa_family = AF_INET,
1168 .sctp_xmit = sctp_v4_xmit,
1169 .setsockopt = ip_setsockopt,
1170 .getsockopt = ip_getsockopt,
1171 .get_dst = sctp_v4_get_dst,
1172 .get_saddr = sctp_v4_get_saddr,
1173 .copy_addrlist = sctp_v4_copy_addrlist,
1174 .from_skb = sctp_v4_from_skb,
1175 .from_sk = sctp_v4_from_sk,
1176 .from_addr_param = sctp_v4_from_addr_param,
1177 .to_addr_param = sctp_v4_to_addr_param,
1178 .cmp_addr = sctp_v4_cmp_addr,
1179 .addr_valid = sctp_v4_addr_valid,
1180 .inaddr_any = sctp_v4_inaddr_any,
1181 .is_any = sctp_v4_is_any,
1182 .available = sctp_v4_available,
1183 .scope = sctp_v4_scope,
1184 .skb_iif = sctp_v4_skb_iif,
1185 .is_ce = sctp_v4_is_ce,
1186 .seq_dump_addr = sctp_v4_seq_dump_addr,
1187 .ecn_capable = sctp_v4_ecn_capable,
1188 .net_header_len = sizeof(struct iphdr),
1189 .sockaddr_len = sizeof(struct sockaddr_in),
1190 .ip_options_len = sctp_v4_ip_options_len,
1193 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1197 return sctp_pf_inet_specific;
1199 return sctp_pf_inet6_specific;
1205 /* Register the PF specific function table. */
1206 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1210 if (sctp_pf_inet_specific)
1212 sctp_pf_inet_specific = pf;
1215 if (sctp_pf_inet6_specific)
1217 sctp_pf_inet6_specific = pf;
1225 static inline int init_sctp_mibs(struct net *net)
1227 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1228 if (!net->sctp.sctp_statistics)
1233 static inline void cleanup_sctp_mibs(struct net *net)
1235 free_percpu(net->sctp.sctp_statistics);
1238 static void sctp_v4_pf_init(void)
1240 /* Initialize the SCTP specific PF functions. */
1241 sctp_register_pf(&sctp_pf_inet, PF_INET);
1242 sctp_register_af(&sctp_af_inet);
1245 static void sctp_v4_pf_exit(void)
1247 list_del(&sctp_af_inet.list);
1250 static int sctp_v4_protosw_init(void)
1254 rc = proto_register(&sctp_prot, 1);
1258 /* Register SCTP(UDP and TCP style) with socket layer. */
1259 inet_register_protosw(&sctp_seqpacket_protosw);
1260 inet_register_protosw(&sctp_stream_protosw);
1265 static void sctp_v4_protosw_exit(void)
1267 inet_unregister_protosw(&sctp_stream_protosw);
1268 inet_unregister_protosw(&sctp_seqpacket_protosw);
1269 proto_unregister(&sctp_prot);
1272 static int sctp_v4_add_protocol(void)
1274 /* Register notifier for inet address additions/deletions. */
1275 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1277 /* Register SCTP with inet layer. */
1278 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1284 static void sctp_v4_del_protocol(void)
1286 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1287 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1290 static int __net_init sctp_defaults_init(struct net *net)
1295 * 14. Suggested SCTP Protocol Parameter Values
1297 /* The following protocol parameters are RECOMMENDED: */
1298 /* RTO.Initial - 3 seconds */
1299 net->sctp.rto_initial = SCTP_RTO_INITIAL;
1300 /* RTO.Min - 1 second */
1301 net->sctp.rto_min = SCTP_RTO_MIN;
1302 /* RTO.Max - 60 seconds */
1303 net->sctp.rto_max = SCTP_RTO_MAX;
1304 /* RTO.Alpha - 1/8 */
1305 net->sctp.rto_alpha = SCTP_RTO_ALPHA;
1306 /* RTO.Beta - 1/4 */
1307 net->sctp.rto_beta = SCTP_RTO_BETA;
1309 /* Valid.Cookie.Life - 60 seconds */
1310 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1312 /* Whether Cookie Preservative is enabled(1) or not(0) */
1313 net->sctp.cookie_preserve_enable = 1;
1315 /* Default sctp sockets to use md5 as their hmac alg */
1316 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1317 net->sctp.sctp_hmac_alg = "md5";
1318 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1319 net->sctp.sctp_hmac_alg = "sha1";
1321 net->sctp.sctp_hmac_alg = NULL;
1325 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST;
1327 /* Disable of Primary Path Switchover by default */
1328 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1330 /* Enable pf state by default */
1331 net->sctp.pf_enable = 1;
1333 /* Ignore pf exposure feature by default */
1334 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1336 /* Association.Max.Retrans - 10 attempts
1337 * Path.Max.Retrans - 5 attempts (per destination address)
1338 * Max.Init.Retransmits - 8 attempts
1340 net->sctp.max_retrans_association = 10;
1341 net->sctp.max_retrans_path = 5;
1342 net->sctp.max_retrans_init = 8;
1344 /* Sendbuffer growth - do per-socket accounting */
1345 net->sctp.sndbuf_policy = 0;
1347 /* Rcvbuffer growth - do per-socket accounting */
1348 net->sctp.rcvbuf_policy = 0;
1350 /* HB.interval - 30 seconds */
1351 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1353 /* delayed SACK timeout */
1354 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1356 /* Disable ADDIP by default. */
1357 net->sctp.addip_enable = 0;
1358 net->sctp.addip_noauth = 0;
1359 net->sctp.default_auto_asconf = 0;
1361 /* Enable PR-SCTP by default. */
1362 net->sctp.prsctp_enable = 1;
1364 /* Disable RECONF by default. */
1365 net->sctp.reconf_enable = 0;
1367 /* Disable AUTH by default. */
1368 net->sctp.auth_enable = 0;
1370 /* Enable ECN by default. */
1371 net->sctp.ecn_enable = 1;
1373 /* Set UDP tunneling listening port to 0 by default */
1374 net->sctp.udp_port = 0;
1376 /* Set remote encap port to 0 by default */
1377 net->sctp.encap_port = 0;
1379 /* Set SCOPE policy to enabled */
1380 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1382 /* Set the default rwnd update threshold */
1383 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1385 /* Initialize maximum autoclose timeout. */
1386 net->sctp.max_autoclose = INT_MAX / HZ;
1388 status = sctp_sysctl_net_register(net);
1390 goto err_sysctl_register;
1392 /* Allocate and initialise sctp mibs. */
1393 status = init_sctp_mibs(net);
1397 #ifdef CONFIG_PROC_FS
1398 /* Initialize proc fs directory. */
1399 status = sctp_proc_init(net);
1404 sctp_dbg_objcnt_init(net);
1406 /* Initialize the local address list. */
1407 INIT_LIST_HEAD(&net->sctp.local_addr_list);
1408 spin_lock_init(&net->sctp.local_addr_lock);
1409 sctp_get_local_addr_list(net);
1411 /* Initialize the address event list */
1412 INIT_LIST_HEAD(&net->sctp.addr_waitq);
1413 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1414 spin_lock_init(&net->sctp.addr_wq_lock);
1415 net->sctp.addr_wq_timer.expires = 0;
1416 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1420 #ifdef CONFIG_PROC_FS
1422 cleanup_sctp_mibs(net);
1425 sctp_sysctl_net_unregister(net);
1426 err_sysctl_register:
1430 static void __net_exit sctp_defaults_exit(struct net *net)
1432 /* Free the local address list */
1433 sctp_free_addr_wq(net);
1434 sctp_free_local_addr_list(net);
1436 #ifdef CONFIG_PROC_FS
1437 remove_proc_subtree("sctp", net->proc_net);
1438 net->sctp.proc_net_sctp = NULL;
1440 cleanup_sctp_mibs(net);
1441 sctp_sysctl_net_unregister(net);
1444 static struct pernet_operations sctp_defaults_ops = {
1445 .init = sctp_defaults_init,
1446 .exit = sctp_defaults_exit,
1449 static int __net_init sctp_ctrlsock_init(struct net *net)
1453 /* Initialize the control inode/socket for handling OOTB packets. */
1454 status = sctp_ctl_sock_init(net);
1456 pr_err("Failed to initialize the SCTP control sock\n");
1461 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1463 /* Free the control endpoint. */
1464 inet_ctl_sock_destroy(net->sctp.ctl_sock);
1467 static struct pernet_operations sctp_ctrlsock_ops = {
1468 .init = sctp_ctrlsock_init,
1469 .exit = sctp_ctrlsock_exit,
1472 /* Initialize the universe into something sensible. */
1473 static __init int sctp_init(void)
1475 unsigned long nr_pages = totalram_pages();
1476 unsigned long limit;
1478 int max_entry_order;
1485 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1487 /* Allocate bind_bucket and chunk caches. */
1489 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1490 sizeof(struct sctp_bind_bucket),
1491 0, SLAB_HWCACHE_ALIGN,
1493 if (!sctp_bucket_cachep)
1496 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1497 sizeof(struct sctp_chunk),
1498 0, SLAB_HWCACHE_ALIGN,
1500 if (!sctp_chunk_cachep)
1501 goto err_chunk_cachep;
1503 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1505 goto err_percpu_counter_init;
1507 /* Implementation specific variables. */
1509 /* Initialize default stream count setup information. */
1510 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1511 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1513 /* Initialize handle used for association ids. */
1514 idr_init(&sctp_assocs_id);
1516 limit = nr_free_buffer_pages() / 8;
1517 limit = max(limit, 128UL);
1518 sysctl_sctp_mem[0] = limit / 4 * 3;
1519 sysctl_sctp_mem[1] = limit;
1520 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1522 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1523 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1524 max_share = min(4UL*1024*1024, limit);
1526 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1527 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1528 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1530 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1531 sysctl_sctp_wmem[1] = 16*1024;
1532 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1534 /* Size and allocate the association hash table.
1535 * The methodology is similar to that of the tcp hash tables.
1536 * Though not identical. Start by getting a goal size
1538 if (nr_pages >= (128 * 1024))
1539 goal = nr_pages >> (22 - PAGE_SHIFT);
1541 goal = nr_pages >> (24 - PAGE_SHIFT);
1543 /* Then compute the page order for said goal */
1544 order = get_order(goal);
1546 /* Now compute the required page order for the maximum sized table we
1549 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1550 sizeof(struct sctp_bind_hashbucket));
1552 /* Limit the page order by that maximum hash table size */
1553 order = min(order, max_entry_order);
1555 /* Allocate and initialize the endpoint hash table. */
1556 sctp_ep_hashsize = 64;
1558 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1559 if (!sctp_ep_hashtable) {
1560 pr_err("Failed endpoint_hash alloc\n");
1562 goto err_ehash_alloc;
1564 for (i = 0; i < sctp_ep_hashsize; i++) {
1565 rwlock_init(&sctp_ep_hashtable[i].lock);
1566 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1569 /* Allocate and initialize the SCTP port hash table.
1570 * Note that order is initalized to start at the max sized
1571 * table we want to support. If we can't get that many pages
1572 * reduce the order and try again
1575 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1576 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1577 } while (!sctp_port_hashtable && --order > 0);
1579 if (!sctp_port_hashtable) {
1580 pr_err("Failed bind hash alloc\n");
1582 goto err_bhash_alloc;
1585 /* Now compute the number of entries that will fit in the
1586 * port hash space we allocated
1588 num_entries = (1UL << order) * PAGE_SIZE /
1589 sizeof(struct sctp_bind_hashbucket);
1591 /* And finish by rounding it down to the nearest power of two.
1592 * This wastes some memory of course, but it's needed because
1593 * the hash function operates based on the assumption that
1594 * the number of entries is a power of two.
1596 sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1598 for (i = 0; i < sctp_port_hashsize; i++) {
1599 spin_lock_init(&sctp_port_hashtable[i].lock);
1600 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1603 status = sctp_transport_hashtable_init();
1605 goto err_thash_alloc;
1607 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1610 sctp_sysctl_register();
1612 INIT_LIST_HEAD(&sctp_address_families);
1615 sctp_sched_ops_init();
1617 status = register_pernet_subsys(&sctp_defaults_ops);
1619 goto err_register_defaults;
1621 status = sctp_v4_protosw_init();
1623 goto err_protosw_init;
1625 status = sctp_v6_protosw_init();
1627 goto err_v6_protosw_init;
1629 status = register_pernet_subsys(&sctp_ctrlsock_ops);
1631 goto err_register_ctrlsock;
1633 status = sctp_v4_add_protocol();
1635 goto err_add_protocol;
1637 /* Register SCTP with inet6 layer. */
1638 status = sctp_v6_add_protocol();
1640 goto err_v6_add_protocol;
1642 if (sctp_offload_init() < 0)
1643 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1647 err_v6_add_protocol:
1648 sctp_v4_del_protocol();
1650 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1651 err_register_ctrlsock:
1652 sctp_v6_protosw_exit();
1653 err_v6_protosw_init:
1654 sctp_v4_protosw_exit();
1656 unregister_pernet_subsys(&sctp_defaults_ops);
1657 err_register_defaults:
1660 sctp_sysctl_unregister();
1661 free_pages((unsigned long)sctp_port_hashtable,
1662 get_order(sctp_port_hashsize *
1663 sizeof(struct sctp_bind_hashbucket)));
1665 sctp_transport_hashtable_destroy();
1667 kfree(sctp_ep_hashtable);
1669 percpu_counter_destroy(&sctp_sockets_allocated);
1670 err_percpu_counter_init:
1671 kmem_cache_destroy(sctp_chunk_cachep);
1673 kmem_cache_destroy(sctp_bucket_cachep);
1677 /* Exit handler for the SCTP protocol. */
1678 static __exit void sctp_exit(void)
1680 /* BUG. This should probably do something useful like clean
1681 * up all the remaining associations and all that memory.
1684 /* Unregister with inet6/inet layers. */
1685 sctp_v6_del_protocol();
1686 sctp_v4_del_protocol();
1688 unregister_pernet_subsys(&sctp_ctrlsock_ops);
1690 /* Free protosw registrations */
1691 sctp_v6_protosw_exit();
1692 sctp_v4_protosw_exit();
1694 unregister_pernet_subsys(&sctp_defaults_ops);
1696 /* Unregister with socket layer. */
1700 sctp_sysctl_unregister();
1702 free_pages((unsigned long)sctp_port_hashtable,
1703 get_order(sctp_port_hashsize *
1704 sizeof(struct sctp_bind_hashbucket)));
1705 kfree(sctp_ep_hashtable);
1706 sctp_transport_hashtable_destroy();
1708 percpu_counter_destroy(&sctp_sockets_allocated);
1710 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1712 kmem_cache_destroy(sctp_chunk_cachep);
1713 kmem_cache_destroy(sctp_bucket_cachep);
1716 module_init(sctp_init);
1717 module_exit(sctp_exit);
1720 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1722 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1723 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1724 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1725 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1726 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1727 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1728 MODULE_LICENSE("GPL");