1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * Initialization/cleanup for SCTP protocol support.
13 * This SCTP implementation is free software;
14 * you can redistribute it and/or modify it under the terms of
15 * the GNU General Public License as published by
16 * the Free Software Foundation; either version 2, or (at your option)
19 * This SCTP implementation is distributed in the hope that it
20 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
21 * ************************
22 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
23 * See the GNU General Public License for more details.
25 * You should have received a copy of the GNU General Public License
26 * along with GNU CC; see the file COPYING. If not, see
27 * <http://www.gnu.org/licenses/>.
29 * Please send any bug reports or fixes you make to the
31 * lksctp developers <linux-sctp@vger.kernel.org>
33 * Written or modified by:
34 * La Monte H.P. Yarroll <piggy@acm.org>
35 * Karl Knutson <karl@athena.chicago.il.us>
36 * Jon Grimm <jgrimm@us.ibm.com>
37 * Sridhar Samudrala <sri@us.ibm.com>
38 * Daisy Chang <daisyc@us.ibm.com>
39 * Ardelle Fan <ardelle.fan@intel.com>
42 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
44 #include <linux/module.h>
45 #include <linux/init.h>
46 #include <linux/netdevice.h>
47 #include <linux/inetdevice.h>
48 #include <linux/seq_file.h>
49 #include <linux/bootmem.h>
50 #include <linux/highmem.h>
51 #include <linux/swap.h>
52 #include <linux/slab.h>
53 #include <net/net_namespace.h>
54 #include <net/protocol.h>
57 #include <net/route.h>
58 #include <net/sctp/sctp.h>
59 #include <net/addrconf.h>
60 #include <net/inet_common.h>
61 #include <net/inet_ecn.h>
63 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
65 /* Global data structures. */
66 struct sctp_globals sctp_globals __read_mostly
;
68 struct idr sctp_assocs_id
;
69 DEFINE_SPINLOCK(sctp_assocs_id_lock
);
71 static struct sctp_pf
*sctp_pf_inet6_specific
;
72 static struct sctp_pf
*sctp_pf_inet_specific
;
73 static struct sctp_af
*sctp_af_v4_specific
;
74 static struct sctp_af
*sctp_af_v6_specific
;
76 struct kmem_cache
*sctp_chunk_cachep __read_mostly
;
77 struct kmem_cache
*sctp_bucket_cachep __read_mostly
;
79 long sysctl_sctp_mem
[3];
80 int sysctl_sctp_rmem
[3];
81 int sysctl_sctp_wmem
[3];
83 /* Set up the proc fs entry for the SCTP protocol. */
84 static int __net_init
sctp_proc_init(struct net
*net
)
87 net
->sctp
.proc_net_sctp
= proc_net_mkdir(net
, "sctp", net
->proc_net
);
88 if (!net
->sctp
.proc_net_sctp
)
89 goto out_proc_net_sctp
;
90 if (sctp_snmp_proc_init(net
))
91 goto out_snmp_proc_init
;
92 if (sctp_eps_proc_init(net
))
93 goto out_eps_proc_init
;
94 if (sctp_assocs_proc_init(net
))
95 goto out_assocs_proc_init
;
96 if (sctp_remaddr_proc_init(net
))
97 goto out_remaddr_proc_init
;
101 out_remaddr_proc_init
:
102 sctp_assocs_proc_exit(net
);
103 out_assocs_proc_init
:
104 sctp_eps_proc_exit(net
);
106 sctp_snmp_proc_exit(net
);
108 remove_proc_entry("sctp", net
->proc_net
);
109 net
->sctp
.proc_net_sctp
= NULL
;
112 #endif /* CONFIG_PROC_FS */
116 /* Clean up the proc fs entry for the SCTP protocol.
117 * Note: Do not make this __exit as it is used in the init error
120 static void sctp_proc_exit(struct net
*net
)
122 #ifdef CONFIG_PROC_FS
123 sctp_snmp_proc_exit(net
);
124 sctp_eps_proc_exit(net
);
125 sctp_assocs_proc_exit(net
);
126 sctp_remaddr_proc_exit(net
);
128 remove_proc_entry("sctp", net
->proc_net
);
129 net
->sctp
.proc_net_sctp
= NULL
;
133 /* Private helper to extract ipv4 address and stash them in
134 * the protocol structure.
136 static void sctp_v4_copy_addrlist(struct list_head
*addrlist
,
137 struct net_device
*dev
)
139 struct in_device
*in_dev
;
140 struct in_ifaddr
*ifa
;
141 struct sctp_sockaddr_entry
*addr
;
144 if ((in_dev
= __in_dev_get_rcu(dev
)) == NULL
) {
149 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
) {
150 /* Add the address to the local list. */
151 addr
= kzalloc(sizeof(*addr
), GFP_ATOMIC
);
153 addr
->a
.v4
.sin_family
= AF_INET
;
154 addr
->a
.v4
.sin_addr
.s_addr
= ifa
->ifa_local
;
156 INIT_LIST_HEAD(&addr
->list
);
157 list_add_tail(&addr
->list
, addrlist
);
164 /* Extract our IP addresses from the system and stash them in the
165 * protocol structure.
167 static void sctp_get_local_addr_list(struct net
*net
)
169 struct net_device
*dev
;
170 struct list_head
*pos
;
174 for_each_netdev_rcu(net
, dev
) {
175 list_for_each(pos
, &sctp_address_families
) {
176 af
= list_entry(pos
, struct sctp_af
, list
);
177 af
->copy_addrlist(&net
->sctp
.local_addr_list
, dev
);
183 /* Free the existing local addresses. */
184 static void sctp_free_local_addr_list(struct net
*net
)
186 struct sctp_sockaddr_entry
*addr
;
187 struct list_head
*pos
, *temp
;
189 list_for_each_safe(pos
, temp
, &net
->sctp
.local_addr_list
) {
190 addr
= list_entry(pos
, struct sctp_sockaddr_entry
, list
);
196 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
197 int sctp_copy_local_addr_list(struct net
*net
, struct sctp_bind_addr
*bp
,
198 sctp_scope_t scope
, gfp_t gfp
, int copy_flags
)
200 struct sctp_sockaddr_entry
*addr
;
204 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
207 if (sctp_in_scope(net
, &addr
->a
, scope
)) {
208 /* Now that the address is in scope, check to see if
209 * the address type is really supported by the local
210 * sock as well as the remote peer.
212 if ((((AF_INET
== addr
->a
.sa
.sa_family
) &&
213 (copy_flags
& SCTP_ADDR4_PEERSUPP
))) ||
214 (((AF_INET6
== addr
->a
.sa
.sa_family
) &&
215 (copy_flags
& SCTP_ADDR6_ALLOWED
) &&
216 (copy_flags
& SCTP_ADDR6_PEERSUPP
)))) {
217 error
= sctp_add_bind_addr(bp
, &addr
->a
,
219 SCTP_ADDR_SRC
, GFP_ATOMIC
);
231 /* Initialize a sctp_addr from in incoming skb. */
232 static void sctp_v4_from_skb(union sctp_addr
*addr
, struct sk_buff
*skb
,
239 port
= &addr
->v4
.sin_port
;
240 addr
->v4
.sin_family
= AF_INET
;
242 /* Always called on head skb, so this is safe */
246 from
= &ip_hdr(skb
)->saddr
;
249 from
= &ip_hdr(skb
)->daddr
;
251 memcpy(&addr
->v4
.sin_addr
.s_addr
, from
, sizeof(struct in_addr
));
254 /* Initialize an sctp_addr from a socket. */
255 static void sctp_v4_from_sk(union sctp_addr
*addr
, struct sock
*sk
)
257 addr
->v4
.sin_family
= AF_INET
;
258 addr
->v4
.sin_port
= 0;
259 addr
->v4
.sin_addr
.s_addr
= inet_sk(sk
)->inet_rcv_saddr
;
262 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
263 static void sctp_v4_to_sk_saddr(union sctp_addr
*addr
, struct sock
*sk
)
265 inet_sk(sk
)->inet_rcv_saddr
= addr
->v4
.sin_addr
.s_addr
;
268 /* Initialize sk->sk_daddr from sctp_addr. */
269 static void sctp_v4_to_sk_daddr(union sctp_addr
*addr
, struct sock
*sk
)
271 inet_sk(sk
)->inet_daddr
= addr
->v4
.sin_addr
.s_addr
;
274 /* Initialize a sctp_addr from an address parameter. */
275 static void sctp_v4_from_addr_param(union sctp_addr
*addr
,
276 union sctp_addr_param
*param
,
277 __be16 port
, int iif
)
279 addr
->v4
.sin_family
= AF_INET
;
280 addr
->v4
.sin_port
= port
;
281 addr
->v4
.sin_addr
.s_addr
= param
->v4
.addr
.s_addr
;
284 /* Initialize an address parameter from a sctp_addr and return the length
285 * of the address parameter.
287 static int sctp_v4_to_addr_param(const union sctp_addr
*addr
,
288 union sctp_addr_param
*param
)
290 int length
= sizeof(sctp_ipv4addr_param_t
);
292 param
->v4
.param_hdr
.type
= SCTP_PARAM_IPV4_ADDRESS
;
293 param
->v4
.param_hdr
.length
= htons(length
);
294 param
->v4
.addr
.s_addr
= addr
->v4
.sin_addr
.s_addr
;
299 /* Initialize a sctp_addr from a dst_entry. */
300 static void sctp_v4_dst_saddr(union sctp_addr
*saddr
, struct flowi4
*fl4
,
303 saddr
->v4
.sin_family
= AF_INET
;
304 saddr
->v4
.sin_port
= port
;
305 saddr
->v4
.sin_addr
.s_addr
= fl4
->saddr
;
308 /* Compare two addresses exactly. */
309 static int sctp_v4_cmp_addr(const union sctp_addr
*addr1
,
310 const union sctp_addr
*addr2
)
312 if (addr1
->sa
.sa_family
!= addr2
->sa
.sa_family
)
314 if (addr1
->v4
.sin_port
!= addr2
->v4
.sin_port
)
316 if (addr1
->v4
.sin_addr
.s_addr
!= addr2
->v4
.sin_addr
.s_addr
)
322 /* Initialize addr struct to INADDR_ANY. */
323 static void sctp_v4_inaddr_any(union sctp_addr
*addr
, __be16 port
)
325 addr
->v4
.sin_family
= AF_INET
;
326 addr
->v4
.sin_addr
.s_addr
= htonl(INADDR_ANY
);
327 addr
->v4
.sin_port
= port
;
330 /* Is this a wildcard address? */
331 static int sctp_v4_is_any(const union sctp_addr
*addr
)
333 return htonl(INADDR_ANY
) == addr
->v4
.sin_addr
.s_addr
;
336 /* This function checks if the address is a valid address to be used for
340 * Return 0 - If the address is a non-unicast or an illegal address.
341 * Return 1 - If the address is a unicast.
343 static int sctp_v4_addr_valid(union sctp_addr
*addr
,
344 struct sctp_sock
*sp
,
345 const struct sk_buff
*skb
)
347 /* IPv4 addresses not allowed */
348 if (sp
&& ipv6_only_sock(sctp_opt2sk(sp
)))
351 /* Is this a non-unicast address or a unusable SCTP address? */
352 if (IS_IPV4_UNUSABLE_ADDRESS(addr
->v4
.sin_addr
.s_addr
))
355 /* Is this a broadcast address? */
356 if (skb
&& skb_rtable(skb
)->rt_flags
& RTCF_BROADCAST
)
362 /* Should this be available for binding? */
363 static int sctp_v4_available(union sctp_addr
*addr
, struct sctp_sock
*sp
)
365 struct net
*net
= sock_net(&sp
->inet
.sk
);
366 int ret
= inet_addr_type(net
, addr
->v4
.sin_addr
.s_addr
);
369 if (addr
->v4
.sin_addr
.s_addr
!= htonl(INADDR_ANY
) &&
371 !sp
->inet
.freebind
&&
372 !net
->ipv4
.sysctl_ip_nonlocal_bind
)
375 if (ipv6_only_sock(sctp_opt2sk(sp
)))
381 /* Checking the loopback, private and other address scopes as defined in
382 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
383 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
385 * Level 0 - unusable SCTP addresses
386 * Level 1 - loopback address
387 * Level 2 - link-local addresses
388 * Level 3 - private addresses.
389 * Level 4 - global addresses
390 * For INIT and INIT-ACK address list, let L be the level of
391 * of requested destination address, sender and receiver
392 * SHOULD include all of its addresses with level greater
393 * than or equal to L.
395 * IPv4 scoping can be controlled through sysctl option
396 * net.sctp.addr_scope_policy
398 static sctp_scope_t
sctp_v4_scope(union sctp_addr
*addr
)
402 /* Check for unusable SCTP addresses. */
403 if (IS_IPV4_UNUSABLE_ADDRESS(addr
->v4
.sin_addr
.s_addr
)) {
404 retval
= SCTP_SCOPE_UNUSABLE
;
405 } else if (ipv4_is_loopback(addr
->v4
.sin_addr
.s_addr
)) {
406 retval
= SCTP_SCOPE_LOOPBACK
;
407 } else if (ipv4_is_linklocal_169(addr
->v4
.sin_addr
.s_addr
)) {
408 retval
= SCTP_SCOPE_LINK
;
409 } else if (ipv4_is_private_10(addr
->v4
.sin_addr
.s_addr
) ||
410 ipv4_is_private_172(addr
->v4
.sin_addr
.s_addr
) ||
411 ipv4_is_private_192(addr
->v4
.sin_addr
.s_addr
)) {
412 retval
= SCTP_SCOPE_PRIVATE
;
414 retval
= SCTP_SCOPE_GLOBAL
;
420 /* Returns a valid dst cache entry for the given source and destination ip
421 * addresses. If an association is passed, trys to get a dst entry with a
422 * source address that matches an address in the bind address list.
424 static void sctp_v4_get_dst(struct sctp_transport
*t
, union sctp_addr
*saddr
,
425 struct flowi
*fl
, struct sock
*sk
)
427 struct sctp_association
*asoc
= t
->asoc
;
429 struct flowi4
*fl4
= &fl
->u
.ip4
;
430 struct sctp_bind_addr
*bp
;
431 struct sctp_sockaddr_entry
*laddr
;
432 struct dst_entry
*dst
= NULL
;
433 union sctp_addr
*daddr
= &t
->ipaddr
;
434 union sctp_addr dst_saddr
;
436 memset(fl4
, 0x0, sizeof(struct flowi4
));
437 fl4
->daddr
= daddr
->v4
.sin_addr
.s_addr
;
438 fl4
->fl4_dport
= daddr
->v4
.sin_port
;
439 fl4
->flowi4_proto
= IPPROTO_SCTP
;
441 fl4
->flowi4_tos
= RT_CONN_FLAGS(asoc
->base
.sk
);
442 fl4
->flowi4_oif
= asoc
->base
.sk
->sk_bound_dev_if
;
443 fl4
->fl4_sport
= htons(asoc
->base
.bind_addr
.port
);
446 fl4
->saddr
= saddr
->v4
.sin_addr
.s_addr
;
447 fl4
->fl4_sport
= saddr
->v4
.sin_port
;
450 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__
, &fl4
->daddr
,
453 rt
= ip_route_output_key(sock_net(sk
), fl4
);
457 /* If there is no association or if a source address is passed, no
458 * more validation is required.
463 bp
= &asoc
->base
.bind_addr
;
466 /* Walk through the bind address list and look for a bind
467 * address that matches the source address of the returned dst.
469 sctp_v4_dst_saddr(&dst_saddr
, fl4
, htons(bp
->port
));
471 list_for_each_entry_rcu(laddr
, &bp
->address_list
, list
) {
472 if (!laddr
->valid
|| (laddr
->state
== SCTP_ADDR_DEL
) ||
473 (laddr
->state
!= SCTP_ADDR_SRC
&&
474 !asoc
->src_out_of_asoc_ok
))
476 if (sctp_v4_cmp_addr(&dst_saddr
, &laddr
->a
))
481 /* None of the bound addresses match the source address of the
482 * dst. So release it.
488 /* Walk through the bind address list and try to get a dst that
489 * matches a bind address as the source address.
492 list_for_each_entry_rcu(laddr
, &bp
->address_list
, list
) {
493 struct net_device
*odev
;
497 if (laddr
->state
!= SCTP_ADDR_SRC
||
498 AF_INET
!= laddr
->a
.sa
.sa_family
)
501 fl4
->fl4_sport
= laddr
->a
.v4
.sin_port
;
502 flowi4_update_output(fl4
,
503 asoc
->base
.sk
->sk_bound_dev_if
,
504 RT_CONN_FLAGS(asoc
->base
.sk
),
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 dst_release(&rt
->dst
);
535 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
536 &fl4
->daddr
, &fl4
->saddr
);
538 pr_debug("no route\n");
541 /* For v4, the source address is cached in the route entry(dst). So no need
542 * to cache it separately and hence this is an empty routine.
544 static void sctp_v4_get_saddr(struct sctp_sock
*sk
,
545 struct sctp_transport
*t
,
548 union sctp_addr
*saddr
= &t
->saddr
;
549 struct rtable
*rt
= (struct rtable
*)t
->dst
;
552 saddr
->v4
.sin_family
= AF_INET
;
553 saddr
->v4
.sin_addr
.s_addr
= fl
->u
.ip4
.saddr
;
557 /* What interface did this skb arrive on? */
558 static int sctp_v4_skb_iif(const struct sk_buff
*skb
)
560 return inet_iif(skb
);
563 /* Was this packet marked by Explicit Congestion Notification? */
564 static int sctp_v4_is_ce(const struct sk_buff
*skb
)
566 return INET_ECN_is_ce(ip_hdr(skb
)->tos
);
569 /* Create and initialize a new sk for the socket returned by accept(). */
570 static struct sock
*sctp_v4_create_accept_sk(struct sock
*sk
,
571 struct sctp_association
*asoc
)
573 struct sock
*newsk
= sk_alloc(sock_net(sk
), PF_INET
, GFP_KERNEL
,
575 struct inet_sock
*newinet
;
580 sock_init_data(NULL
, newsk
);
582 sctp_copy_sock(newsk
, sk
, asoc
);
583 sock_reset_flag(newsk
, SOCK_ZAPPED
);
585 newinet
= inet_sk(newsk
);
587 newinet
->inet_daddr
= asoc
->peer
.primary_addr
.v4
.sin_addr
.s_addr
;
589 sk_refcnt_debug_inc(newsk
);
591 if (newsk
->sk_prot
->init(newsk
)) {
592 sk_common_release(newsk
);
600 static int sctp_v4_addr_to_user(struct sctp_sock
*sp
, union sctp_addr
*addr
)
602 /* No address mapping for V4 sockets */
603 memset(addr
->v4
.sin_zero
, 0, sizeof(addr
->v4
.sin_zero
));
604 return sizeof(struct sockaddr_in
);
607 /* Dump the v4 addr to the seq file. */
608 static void sctp_v4_seq_dump_addr(struct seq_file
*seq
, union sctp_addr
*addr
)
610 seq_printf(seq
, "%pI4 ", &addr
->v4
.sin_addr
);
613 static void sctp_v4_ecn_capable(struct sock
*sk
)
618 static void sctp_addr_wq_timeout_handler(unsigned long arg
)
620 struct net
*net
= (struct net
*)arg
;
621 struct sctp_sockaddr_entry
*addrw
, *temp
;
622 struct sctp_sock
*sp
;
624 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
626 list_for_each_entry_safe(addrw
, temp
, &net
->sctp
.addr_waitq
, list
) {
627 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
628 "entry:%p\n", __func__
, &net
->sctp
.addr_waitq
, &addrw
->a
.sa
,
629 addrw
->state
, addrw
);
631 #if IS_ENABLED(CONFIG_IPV6)
632 /* Now we send an ASCONF for each association */
633 /* Note. we currently don't handle link local IPv6 addressees */
634 if (addrw
->a
.sa
.sa_family
== AF_INET6
) {
635 struct in6_addr
*in6
;
637 if (ipv6_addr_type(&addrw
->a
.v6
.sin6_addr
) &
641 in6
= (struct in6_addr
*)&addrw
->a
.v6
.sin6_addr
;
642 if (ipv6_chk_addr(net
, in6
, NULL
, 0) == 0 &&
643 addrw
->state
== SCTP_ADDR_NEW
) {
644 unsigned long timeo_val
;
646 pr_debug("%s: this is on DAD, trying %d sec "
648 SCTP_ADDRESS_TICK_DELAY
);
651 timeo_val
+= msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY
);
652 mod_timer(&net
->sctp
.addr_wq_timer
, timeo_val
);
657 list_for_each_entry(sp
, &net
->sctp
.auto_asconf_splist
, auto_asconf_list
) {
660 sk
= sctp_opt2sk(sp
);
661 /* ignore bound-specific endpoints */
662 if (!sctp_is_ep_boundall(sk
))
665 if (sctp_asconf_mgmt(sp
, addrw
) < 0)
666 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__
);
669 #if IS_ENABLED(CONFIG_IPV6)
672 list_del(&addrw
->list
);
675 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
678 static void sctp_free_addr_wq(struct net
*net
)
680 struct sctp_sockaddr_entry
*addrw
;
681 struct sctp_sockaddr_entry
*temp
;
683 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
684 del_timer(&net
->sctp
.addr_wq_timer
);
685 list_for_each_entry_safe(addrw
, temp
, &net
->sctp
.addr_waitq
, list
) {
686 list_del(&addrw
->list
);
689 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
692 /* lookup the entry for the same address in the addr_waitq
693 * sctp_addr_wq MUST be locked
695 static struct sctp_sockaddr_entry
*sctp_addr_wq_lookup(struct net
*net
,
696 struct sctp_sockaddr_entry
*addr
)
698 struct sctp_sockaddr_entry
*addrw
;
700 list_for_each_entry(addrw
, &net
->sctp
.addr_waitq
, list
) {
701 if (addrw
->a
.sa
.sa_family
!= addr
->a
.sa
.sa_family
)
703 if (addrw
->a
.sa
.sa_family
== AF_INET
) {
704 if (addrw
->a
.v4
.sin_addr
.s_addr
==
705 addr
->a
.v4
.sin_addr
.s_addr
)
707 } else if (addrw
->a
.sa
.sa_family
== AF_INET6
) {
708 if (ipv6_addr_equal(&addrw
->a
.v6
.sin6_addr
,
709 &addr
->a
.v6
.sin6_addr
))
716 void sctp_addr_wq_mgmt(struct net
*net
, struct sctp_sockaddr_entry
*addr
, int cmd
)
718 struct sctp_sockaddr_entry
*addrw
;
719 unsigned long timeo_val
;
721 /* first, we check if an opposite message already exist in the queue.
722 * If we found such message, it is removed.
723 * This operation is a bit stupid, but the DHCP client attaches the
724 * new address after a couple of addition and deletion of that address
727 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
728 /* Offsets existing events in addr_wq */
729 addrw
= sctp_addr_wq_lookup(net
, addr
);
731 if (addrw
->state
!= cmd
) {
732 pr_debug("%s: offsets existing entry for %d, addr:%pISc "
733 "in wq:%p\n", __func__
, addrw
->state
, &addrw
->a
.sa
,
734 &net
->sctp
.addr_waitq
);
736 list_del(&addrw
->list
);
739 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
743 /* OK, we have to add the new address to the wait queue */
744 addrw
= kmemdup(addr
, sizeof(struct sctp_sockaddr_entry
), GFP_ATOMIC
);
746 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
750 list_add_tail(&addrw
->list
, &net
->sctp
.addr_waitq
);
752 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
753 __func__
, addrw
->state
, &addrw
->a
.sa
, &net
->sctp
.addr_waitq
);
755 if (!timer_pending(&net
->sctp
.addr_wq_timer
)) {
757 timeo_val
+= msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY
);
758 mod_timer(&net
->sctp
.addr_wq_timer
, timeo_val
);
760 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
763 /* Event handler for inet address addition/deletion events.
764 * The sctp_local_addr_list needs to be protocted by a spin lock since
765 * multiple notifiers (say IPv4 and IPv6) may be running at the same
766 * time and thus corrupt the list.
767 * The reader side is protected with RCU.
769 static int sctp_inetaddr_event(struct notifier_block
*this, unsigned long ev
,
772 struct in_ifaddr
*ifa
= (struct in_ifaddr
*)ptr
;
773 struct sctp_sockaddr_entry
*addr
= NULL
;
774 struct sctp_sockaddr_entry
*temp
;
775 struct net
*net
= dev_net(ifa
->ifa_dev
->dev
);
780 addr
= kzalloc(sizeof(*addr
), GFP_ATOMIC
);
782 addr
->a
.v4
.sin_family
= AF_INET
;
783 addr
->a
.v4
.sin_addr
.s_addr
= ifa
->ifa_local
;
785 spin_lock_bh(&net
->sctp
.local_addr_lock
);
786 list_add_tail_rcu(&addr
->list
, &net
->sctp
.local_addr_list
);
787 sctp_addr_wq_mgmt(net
, addr
, SCTP_ADDR_NEW
);
788 spin_unlock_bh(&net
->sctp
.local_addr_lock
);
792 spin_lock_bh(&net
->sctp
.local_addr_lock
);
793 list_for_each_entry_safe(addr
, temp
,
794 &net
->sctp
.local_addr_list
, list
) {
795 if (addr
->a
.sa
.sa_family
== AF_INET
&&
796 addr
->a
.v4
.sin_addr
.s_addr
==
798 sctp_addr_wq_mgmt(net
, addr
, SCTP_ADDR_DEL
);
801 list_del_rcu(&addr
->list
);
805 spin_unlock_bh(&net
->sctp
.local_addr_lock
);
807 kfree_rcu(addr
, rcu
);
815 * Initialize the control inode/socket with a control endpoint data
816 * structure. This endpoint is reserved exclusively for the OOTB processing.
818 static int sctp_ctl_sock_init(struct net
*net
)
821 sa_family_t family
= PF_INET
;
823 if (sctp_get_pf_specific(PF_INET6
))
826 err
= inet_ctl_sock_create(&net
->sctp
.ctl_sock
, family
,
827 SOCK_SEQPACKET
, IPPROTO_SCTP
, net
);
829 /* If IPv6 socket could not be created, try the IPv4 socket */
830 if (err
< 0 && family
== PF_INET6
)
831 err
= inet_ctl_sock_create(&net
->sctp
.ctl_sock
, AF_INET
,
832 SOCK_SEQPACKET
, IPPROTO_SCTP
,
836 pr_err("Failed to create the SCTP control socket\n");
842 /* Register address family specific functions. */
843 int sctp_register_af(struct sctp_af
*af
)
845 switch (af
->sa_family
) {
847 if (sctp_af_v4_specific
)
849 sctp_af_v4_specific
= af
;
852 if (sctp_af_v6_specific
)
854 sctp_af_v6_specific
= af
;
860 INIT_LIST_HEAD(&af
->list
);
861 list_add_tail(&af
->list
, &sctp_address_families
);
865 /* Get the table of functions for manipulating a particular address
868 struct sctp_af
*sctp_get_af_specific(sa_family_t family
)
872 return sctp_af_v4_specific
;
874 return sctp_af_v6_specific
;
880 /* Common code to initialize a AF_INET msg_name. */
881 static void sctp_inet_msgname(char *msgname
, int *addr_len
)
883 struct sockaddr_in
*sin
;
885 sin
= (struct sockaddr_in
*)msgname
;
886 *addr_len
= sizeof(struct sockaddr_in
);
887 sin
->sin_family
= AF_INET
;
888 memset(sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
891 /* Copy the primary address of the peer primary address as the msg_name. */
892 static void sctp_inet_event_msgname(struct sctp_ulpevent
*event
, char *msgname
,
895 struct sockaddr_in
*sin
, *sinfrom
;
898 struct sctp_association
*asoc
;
901 sctp_inet_msgname(msgname
, addr_len
);
902 sin
= (struct sockaddr_in
*)msgname
;
903 sinfrom
= &asoc
->peer
.primary_addr
.v4
;
904 sin
->sin_port
= htons(asoc
->peer
.port
);
905 sin
->sin_addr
.s_addr
= sinfrom
->sin_addr
.s_addr
;
909 /* Initialize and copy out a msgname from an inbound skb. */
910 static void sctp_inet_skb_msgname(struct sk_buff
*skb
, char *msgname
, int *len
)
913 struct sctphdr
*sh
= sctp_hdr(skb
);
914 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msgname
;
916 sctp_inet_msgname(msgname
, len
);
917 sin
->sin_port
= sh
->source
;
918 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
922 /* Do we support this AF? */
923 static int sctp_inet_af_supported(sa_family_t family
, struct sctp_sock
*sp
)
925 /* PF_INET only supports AF_INET addresses. */
926 return AF_INET
== family
;
929 /* Address matching with wildcards allowed. */
930 static int sctp_inet_cmp_addr(const union sctp_addr
*addr1
,
931 const union sctp_addr
*addr2
,
932 struct sctp_sock
*opt
)
934 /* PF_INET only supports AF_INET addresses. */
935 if (addr1
->sa
.sa_family
!= addr2
->sa
.sa_family
)
937 if (htonl(INADDR_ANY
) == addr1
->v4
.sin_addr
.s_addr
||
938 htonl(INADDR_ANY
) == addr2
->v4
.sin_addr
.s_addr
)
940 if (addr1
->v4
.sin_addr
.s_addr
== addr2
->v4
.sin_addr
.s_addr
)
946 /* Verify that provided sockaddr looks bindable. Common verification has
947 * already been taken care of.
949 static int sctp_inet_bind_verify(struct sctp_sock
*opt
, union sctp_addr
*addr
)
951 return sctp_v4_available(addr
, opt
);
954 /* Verify that sockaddr looks sendable. Common verification has already
955 * been taken care of.
957 static int sctp_inet_send_verify(struct sctp_sock
*opt
, union sctp_addr
*addr
)
962 /* Fill in Supported Address Type information for INIT and INIT-ACK
963 * chunks. Returns number of addresses supported.
965 static int sctp_inet_supported_addrs(const struct sctp_sock
*opt
,
968 types
[0] = SCTP_PARAM_IPV4_ADDRESS
;
972 /* Wrapper routine that calls the ip transmit routine. */
973 static inline int sctp_v4_xmit(struct sk_buff
*skb
,
974 struct sctp_transport
*transport
)
976 struct inet_sock
*inet
= inet_sk(skb
->sk
);
978 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__
, skb
,
979 skb
->len
, &transport
->fl
.u
.ip4
.saddr
, &transport
->fl
.u
.ip4
.daddr
);
981 inet
->pmtudisc
= transport
->param_flags
& SPP_PMTUD_ENABLE
?
982 IP_PMTUDISC_DO
: IP_PMTUDISC_DONT
;
984 SCTP_INC_STATS(sock_net(&inet
->sk
), SCTP_MIB_OUTSCTPPACKS
);
986 return ip_queue_xmit(&inet
->sk
, skb
, &transport
->fl
);
989 static struct sctp_af sctp_af_inet
;
991 static struct sctp_pf sctp_pf_inet
= {
992 .event_msgname
= sctp_inet_event_msgname
,
993 .skb_msgname
= sctp_inet_skb_msgname
,
994 .af_supported
= sctp_inet_af_supported
,
995 .cmp_addr
= sctp_inet_cmp_addr
,
996 .bind_verify
= sctp_inet_bind_verify
,
997 .send_verify
= sctp_inet_send_verify
,
998 .supported_addrs
= sctp_inet_supported_addrs
,
999 .create_accept_sk
= sctp_v4_create_accept_sk
,
1000 .addr_to_user
= sctp_v4_addr_to_user
,
1001 .to_sk_saddr
= sctp_v4_to_sk_saddr
,
1002 .to_sk_daddr
= sctp_v4_to_sk_daddr
,
1006 /* Notifier for inetaddr addition/deletion events. */
1007 static struct notifier_block sctp_inetaddr_notifier
= {
1008 .notifier_call
= sctp_inetaddr_event
,
1011 /* Socket operations. */
1012 static const struct proto_ops inet_seqpacket_ops
= {
1014 .owner
= THIS_MODULE
,
1015 .release
= inet_release
, /* Needs to be wrapped... */
1017 .connect
= sctp_inet_connect
,
1018 .socketpair
= sock_no_socketpair
,
1019 .accept
= inet_accept
,
1020 .getname
= inet_getname
, /* Semantics are different. */
1022 .ioctl
= inet_ioctl
,
1023 .listen
= sctp_inet_listen
,
1024 .shutdown
= inet_shutdown
, /* Looks harmless. */
1025 .setsockopt
= sock_common_setsockopt
, /* IP_SOL IP_OPTION is a problem */
1026 .getsockopt
= sock_common_getsockopt
,
1027 .sendmsg
= inet_sendmsg
,
1028 .recvmsg
= inet_recvmsg
,
1029 .mmap
= sock_no_mmap
,
1030 .sendpage
= sock_no_sendpage
,
1031 #ifdef CONFIG_COMPAT
1032 .compat_setsockopt
= compat_sock_common_setsockopt
,
1033 .compat_getsockopt
= compat_sock_common_getsockopt
,
1037 /* Registration with AF_INET family. */
1038 static struct inet_protosw sctp_seqpacket_protosw
= {
1039 .type
= SOCK_SEQPACKET
,
1040 .protocol
= IPPROTO_SCTP
,
1042 .ops
= &inet_seqpacket_ops
,
1043 .flags
= SCTP_PROTOSW_FLAG
1045 static struct inet_protosw sctp_stream_protosw
= {
1046 .type
= SOCK_STREAM
,
1047 .protocol
= IPPROTO_SCTP
,
1049 .ops
= &inet_seqpacket_ops
,
1050 .flags
= SCTP_PROTOSW_FLAG
1053 /* Register with IP layer. */
1054 static const struct net_protocol sctp_protocol
= {
1055 .handler
= sctp_rcv
,
1056 .err_handler
= sctp_v4_err
,
1059 .icmp_strict_tag_validation
= 1,
1062 /* IPv4 address related functions. */
1063 static struct sctp_af sctp_af_inet
= {
1064 .sa_family
= AF_INET
,
1065 .sctp_xmit
= sctp_v4_xmit
,
1066 .setsockopt
= ip_setsockopt
,
1067 .getsockopt
= ip_getsockopt
,
1068 .get_dst
= sctp_v4_get_dst
,
1069 .get_saddr
= sctp_v4_get_saddr
,
1070 .copy_addrlist
= sctp_v4_copy_addrlist
,
1071 .from_skb
= sctp_v4_from_skb
,
1072 .from_sk
= sctp_v4_from_sk
,
1073 .from_addr_param
= sctp_v4_from_addr_param
,
1074 .to_addr_param
= sctp_v4_to_addr_param
,
1075 .cmp_addr
= sctp_v4_cmp_addr
,
1076 .addr_valid
= sctp_v4_addr_valid
,
1077 .inaddr_any
= sctp_v4_inaddr_any
,
1078 .is_any
= sctp_v4_is_any
,
1079 .available
= sctp_v4_available
,
1080 .scope
= sctp_v4_scope
,
1081 .skb_iif
= sctp_v4_skb_iif
,
1082 .is_ce
= sctp_v4_is_ce
,
1083 .seq_dump_addr
= sctp_v4_seq_dump_addr
,
1084 .ecn_capable
= sctp_v4_ecn_capable
,
1085 .net_header_len
= sizeof(struct iphdr
),
1086 .sockaddr_len
= sizeof(struct sockaddr_in
),
1087 #ifdef CONFIG_COMPAT
1088 .compat_setsockopt
= compat_ip_setsockopt
,
1089 .compat_getsockopt
= compat_ip_getsockopt
,
1093 struct sctp_pf
*sctp_get_pf_specific(sa_family_t family
)
1097 return sctp_pf_inet_specific
;
1099 return sctp_pf_inet6_specific
;
1105 /* Register the PF specific function table. */
1106 int sctp_register_pf(struct sctp_pf
*pf
, sa_family_t family
)
1110 if (sctp_pf_inet_specific
)
1112 sctp_pf_inet_specific
= pf
;
1115 if (sctp_pf_inet6_specific
)
1117 sctp_pf_inet6_specific
= pf
;
1125 static inline int init_sctp_mibs(struct net
*net
)
1127 net
->sctp
.sctp_statistics
= alloc_percpu(struct sctp_mib
);
1128 if (!net
->sctp
.sctp_statistics
)
1133 static inline void cleanup_sctp_mibs(struct net
*net
)
1135 free_percpu(net
->sctp
.sctp_statistics
);
1138 static void sctp_v4_pf_init(void)
1140 /* Initialize the SCTP specific PF functions. */
1141 sctp_register_pf(&sctp_pf_inet
, PF_INET
);
1142 sctp_register_af(&sctp_af_inet
);
1145 static void sctp_v4_pf_exit(void)
1147 list_del(&sctp_af_inet
.list
);
1150 static int sctp_v4_protosw_init(void)
1154 rc
= proto_register(&sctp_prot
, 1);
1158 /* Register SCTP(UDP and TCP style) with socket layer. */
1159 inet_register_protosw(&sctp_seqpacket_protosw
);
1160 inet_register_protosw(&sctp_stream_protosw
);
1165 static void sctp_v4_protosw_exit(void)
1167 inet_unregister_protosw(&sctp_stream_protosw
);
1168 inet_unregister_protosw(&sctp_seqpacket_protosw
);
1169 proto_unregister(&sctp_prot
);
1172 static int sctp_v4_add_protocol(void)
1174 /* Register notifier for inet address additions/deletions. */
1175 register_inetaddr_notifier(&sctp_inetaddr_notifier
);
1177 /* Register SCTP with inet layer. */
1178 if (inet_add_protocol(&sctp_protocol
, IPPROTO_SCTP
) < 0)
1184 static void sctp_v4_del_protocol(void)
1186 inet_del_protocol(&sctp_protocol
, IPPROTO_SCTP
);
1187 unregister_inetaddr_notifier(&sctp_inetaddr_notifier
);
1190 static int __net_init
sctp_defaults_init(struct net
*net
)
1195 * 14. Suggested SCTP Protocol Parameter Values
1197 /* The following protocol parameters are RECOMMENDED: */
1198 /* RTO.Initial - 3 seconds */
1199 net
->sctp
.rto_initial
= SCTP_RTO_INITIAL
;
1200 /* RTO.Min - 1 second */
1201 net
->sctp
.rto_min
= SCTP_RTO_MIN
;
1202 /* RTO.Max - 60 seconds */
1203 net
->sctp
.rto_max
= SCTP_RTO_MAX
;
1204 /* RTO.Alpha - 1/8 */
1205 net
->sctp
.rto_alpha
= SCTP_RTO_ALPHA
;
1206 /* RTO.Beta - 1/4 */
1207 net
->sctp
.rto_beta
= SCTP_RTO_BETA
;
1209 /* Valid.Cookie.Life - 60 seconds */
1210 net
->sctp
.valid_cookie_life
= SCTP_DEFAULT_COOKIE_LIFE
;
1212 /* Whether Cookie Preservative is enabled(1) or not(0) */
1213 net
->sctp
.cookie_preserve_enable
= 1;
1215 /* Default sctp sockets to use md5 as their hmac alg */
1216 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1217 net
->sctp
.sctp_hmac_alg
= "md5";
1218 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1219 net
->sctp
.sctp_hmac_alg
= "sha1";
1221 net
->sctp
.sctp_hmac_alg
= NULL
;
1225 net
->sctp
.max_burst
= SCTP_DEFAULT_MAX_BURST
;
1227 /* Enable pf state by default */
1228 net
->sctp
.pf_enable
= 1;
1230 /* Association.Max.Retrans - 10 attempts
1231 * Path.Max.Retrans - 5 attempts (per destination address)
1232 * Max.Init.Retransmits - 8 attempts
1234 net
->sctp
.max_retrans_association
= 10;
1235 net
->sctp
.max_retrans_path
= 5;
1236 net
->sctp
.max_retrans_init
= 8;
1238 /* Sendbuffer growth - do per-socket accounting */
1239 net
->sctp
.sndbuf_policy
= 0;
1241 /* Rcvbuffer growth - do per-socket accounting */
1242 net
->sctp
.rcvbuf_policy
= 0;
1244 /* HB.interval - 30 seconds */
1245 net
->sctp
.hb_interval
= SCTP_DEFAULT_TIMEOUT_HEARTBEAT
;
1247 /* delayed SACK timeout */
1248 net
->sctp
.sack_timeout
= SCTP_DEFAULT_TIMEOUT_SACK
;
1250 /* Disable ADDIP by default. */
1251 net
->sctp
.addip_enable
= 0;
1252 net
->sctp
.addip_noauth
= 0;
1253 net
->sctp
.default_auto_asconf
= 0;
1255 /* Enable PR-SCTP by default. */
1256 net
->sctp
.prsctp_enable
= 1;
1258 /* Disable AUTH by default. */
1259 net
->sctp
.auth_enable
= 0;
1261 /* Set SCOPE policy to enabled */
1262 net
->sctp
.scope_policy
= SCTP_SCOPE_POLICY_ENABLE
;
1264 /* Set the default rwnd update threshold */
1265 net
->sctp
.rwnd_upd_shift
= SCTP_DEFAULT_RWND_SHIFT
;
1267 /* Initialize maximum autoclose timeout. */
1268 net
->sctp
.max_autoclose
= INT_MAX
/ HZ
;
1270 status
= sctp_sysctl_net_register(net
);
1272 goto err_sysctl_register
;
1274 /* Allocate and initialise sctp mibs. */
1275 status
= init_sctp_mibs(net
);
1279 /* Initialize proc fs directory. */
1280 status
= sctp_proc_init(net
);
1284 sctp_dbg_objcnt_init(net
);
1286 /* Initialize the local address list. */
1287 INIT_LIST_HEAD(&net
->sctp
.local_addr_list
);
1288 spin_lock_init(&net
->sctp
.local_addr_lock
);
1289 sctp_get_local_addr_list(net
);
1291 /* Initialize the address event list */
1292 INIT_LIST_HEAD(&net
->sctp
.addr_waitq
);
1293 INIT_LIST_HEAD(&net
->sctp
.auto_asconf_splist
);
1294 spin_lock_init(&net
->sctp
.addr_wq_lock
);
1295 net
->sctp
.addr_wq_timer
.expires
= 0;
1296 setup_timer(&net
->sctp
.addr_wq_timer
, sctp_addr_wq_timeout_handler
,
1297 (unsigned long)net
);
1302 cleanup_sctp_mibs(net
);
1304 sctp_sysctl_net_unregister(net
);
1305 err_sysctl_register
:
1309 static void __net_exit
sctp_defaults_exit(struct net
*net
)
1311 /* Free the local address list */
1312 sctp_free_addr_wq(net
);
1313 sctp_free_local_addr_list(net
);
1315 sctp_dbg_objcnt_exit(net
);
1317 sctp_proc_exit(net
);
1318 cleanup_sctp_mibs(net
);
1319 sctp_sysctl_net_unregister(net
);
1322 static struct pernet_operations sctp_defaults_ops
= {
1323 .init
= sctp_defaults_init
,
1324 .exit
= sctp_defaults_exit
,
1327 static int __net_init
sctp_ctrlsock_init(struct net
*net
)
1331 /* Initialize the control inode/socket for handling OOTB packets. */
1332 status
= sctp_ctl_sock_init(net
);
1334 pr_err("Failed to initialize the SCTP control sock\n");
1339 static void __net_exit
sctp_ctrlsock_exit(struct net
*net
)
1341 /* Free the control endpoint. */
1342 inet_ctl_sock_destroy(net
->sctp
.ctl_sock
);
1345 static struct pernet_operations sctp_ctrlsock_ops
= {
1346 .init
= sctp_ctrlsock_init
,
1347 .exit
= sctp_ctrlsock_exit
,
1350 /* Initialize the universe into something sensible. */
1351 static __init
int sctp_init(void)
1354 int status
= -EINVAL
;
1356 unsigned long limit
;
1360 int max_entry_order
;
1362 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent
));
1364 /* Allocate bind_bucket and chunk caches. */
1366 sctp_bucket_cachep
= kmem_cache_create("sctp_bind_bucket",
1367 sizeof(struct sctp_bind_bucket
),
1368 0, SLAB_HWCACHE_ALIGN
,
1370 if (!sctp_bucket_cachep
)
1373 sctp_chunk_cachep
= kmem_cache_create("sctp_chunk",
1374 sizeof(struct sctp_chunk
),
1375 0, SLAB_HWCACHE_ALIGN
,
1377 if (!sctp_chunk_cachep
)
1378 goto err_chunk_cachep
;
1380 status
= percpu_counter_init(&sctp_sockets_allocated
, 0, GFP_KERNEL
);
1382 goto err_percpu_counter_init
;
1384 /* Implementation specific variables. */
1386 /* Initialize default stream count setup information. */
1387 sctp_max_instreams
= SCTP_DEFAULT_INSTREAMS
;
1388 sctp_max_outstreams
= SCTP_DEFAULT_OUTSTREAMS
;
1390 /* Initialize handle used for association ids. */
1391 idr_init(&sctp_assocs_id
);
1393 limit
= nr_free_buffer_pages() / 8;
1394 limit
= max(limit
, 128UL);
1395 sysctl_sctp_mem
[0] = limit
/ 4 * 3;
1396 sysctl_sctp_mem
[1] = limit
;
1397 sysctl_sctp_mem
[2] = sysctl_sctp_mem
[0] * 2;
1399 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1400 limit
= (sysctl_sctp_mem
[1]) << (PAGE_SHIFT
- 7);
1401 max_share
= min(4UL*1024*1024, limit
);
1403 sysctl_sctp_rmem
[0] = SK_MEM_QUANTUM
; /* give each asoc 1 page min */
1404 sysctl_sctp_rmem
[1] = 1500 * SKB_TRUESIZE(1);
1405 sysctl_sctp_rmem
[2] = max(sysctl_sctp_rmem
[1], max_share
);
1407 sysctl_sctp_wmem
[0] = SK_MEM_QUANTUM
;
1408 sysctl_sctp_wmem
[1] = 16*1024;
1409 sysctl_sctp_wmem
[2] = max(64*1024, max_share
);
1411 /* Size and allocate the association hash table.
1412 * The methodology is similar to that of the tcp hash tables.
1413 * Though not identical. Start by getting a goal size
1415 if (totalram_pages
>= (128 * 1024))
1416 goal
= totalram_pages
>> (22 - PAGE_SHIFT
);
1418 goal
= totalram_pages
>> (24 - PAGE_SHIFT
);
1420 /* Then compute the page order for said goal */
1421 order
= get_order(goal
);
1423 /* Now compute the required page order for the maximum sized table we
1426 max_entry_order
= get_order(MAX_SCTP_PORT_HASH_ENTRIES
*
1427 sizeof(struct sctp_bind_hashbucket
));
1429 /* Limit the page order by that maximum hash table size */
1430 order
= min(order
, max_entry_order
);
1432 /* Allocate and initialize the endpoint hash table. */
1433 sctp_ep_hashsize
= 64;
1435 kmalloc(64 * sizeof(struct sctp_hashbucket
), GFP_KERNEL
);
1436 if (!sctp_ep_hashtable
) {
1437 pr_err("Failed endpoint_hash alloc\n");
1439 goto err_ehash_alloc
;
1441 for (i
= 0; i
< sctp_ep_hashsize
; i
++) {
1442 rwlock_init(&sctp_ep_hashtable
[i
].lock
);
1443 INIT_HLIST_HEAD(&sctp_ep_hashtable
[i
].chain
);
1446 /* Allocate and initialize the SCTP port hash table.
1447 * Note that order is initalized to start at the max sized
1448 * table we want to support. If we can't get that many pages
1449 * reduce the order and try again
1452 sctp_port_hashtable
= (struct sctp_bind_hashbucket
*)
1453 __get_free_pages(GFP_KERNEL
| __GFP_NOWARN
, order
);
1454 } while (!sctp_port_hashtable
&& --order
> 0);
1456 if (!sctp_port_hashtable
) {
1457 pr_err("Failed bind hash alloc\n");
1459 goto err_bhash_alloc
;
1462 /* Now compute the number of entries that will fit in the
1463 * port hash space we allocated
1465 num_entries
= (1UL << order
) * PAGE_SIZE
/
1466 sizeof(struct sctp_bind_hashbucket
);
1468 /* And finish by rounding it down to the nearest power of two
1469 * this wastes some memory of course, but its needed because
1470 * the hash function operates based on the assumption that
1471 * that the number of entries is a power of two
1473 sctp_port_hashsize
= rounddown_pow_of_two(num_entries
);
1475 for (i
= 0; i
< sctp_port_hashsize
; i
++) {
1476 spin_lock_init(&sctp_port_hashtable
[i
].lock
);
1477 INIT_HLIST_HEAD(&sctp_port_hashtable
[i
].chain
);
1480 status
= sctp_transport_hashtable_init();
1482 goto err_thash_alloc
;
1484 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize
,
1487 sctp_sysctl_register();
1489 INIT_LIST_HEAD(&sctp_address_families
);
1493 status
= register_pernet_subsys(&sctp_defaults_ops
);
1495 goto err_register_defaults
;
1497 status
= sctp_v4_protosw_init();
1499 goto err_protosw_init
;
1501 status
= sctp_v6_protosw_init();
1503 goto err_v6_protosw_init
;
1505 status
= register_pernet_subsys(&sctp_ctrlsock_ops
);
1507 goto err_register_ctrlsock
;
1509 status
= sctp_v4_add_protocol();
1511 goto err_add_protocol
;
1513 /* Register SCTP with inet6 layer. */
1514 status
= sctp_v6_add_protocol();
1516 goto err_v6_add_protocol
;
1518 if (sctp_offload_init() < 0)
1519 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__
);
1523 err_v6_add_protocol
:
1524 sctp_v4_del_protocol();
1526 unregister_pernet_subsys(&sctp_ctrlsock_ops
);
1527 err_register_ctrlsock
:
1528 sctp_v6_protosw_exit();
1529 err_v6_protosw_init
:
1530 sctp_v4_protosw_exit();
1532 unregister_pernet_subsys(&sctp_defaults_ops
);
1533 err_register_defaults
:
1536 sctp_sysctl_unregister();
1537 free_pages((unsigned long)sctp_port_hashtable
,
1538 get_order(sctp_port_hashsize
*
1539 sizeof(struct sctp_bind_hashbucket
)));
1541 sctp_transport_hashtable_destroy();
1543 kfree(sctp_ep_hashtable
);
1545 percpu_counter_destroy(&sctp_sockets_allocated
);
1546 err_percpu_counter_init
:
1547 kmem_cache_destroy(sctp_chunk_cachep
);
1549 kmem_cache_destroy(sctp_bucket_cachep
);
1553 /* Exit handler for the SCTP protocol. */
1554 static __exit
void sctp_exit(void)
1556 /* BUG. This should probably do something useful like clean
1557 * up all the remaining associations and all that memory.
1560 /* Unregister with inet6/inet layers. */
1561 sctp_v6_del_protocol();
1562 sctp_v4_del_protocol();
1564 unregister_pernet_subsys(&sctp_ctrlsock_ops
);
1566 /* Free protosw registrations */
1567 sctp_v6_protosw_exit();
1568 sctp_v4_protosw_exit();
1570 unregister_pernet_subsys(&sctp_defaults_ops
);
1572 /* Unregister with socket layer. */
1576 sctp_sysctl_unregister();
1578 free_pages((unsigned long)sctp_port_hashtable
,
1579 get_order(sctp_port_hashsize
*
1580 sizeof(struct sctp_bind_hashbucket
)));
1581 kfree(sctp_ep_hashtable
);
1582 sctp_transport_hashtable_destroy();
1584 percpu_counter_destroy(&sctp_sockets_allocated
);
1586 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1588 kmem_cache_destroy(sctp_chunk_cachep
);
1589 kmem_cache_destroy(sctp_bucket_cachep
);
1592 module_init(sctp_init
);
1593 module_exit(sctp_exit
);
1596 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1598 MODULE_ALIAS("net-pf-" __stringify(PF_INET
) "-proto-132");
1599 MODULE_ALIAS("net-pf-" __stringify(PF_INET6
) "-proto-132");
1600 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1601 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1602 module_param_named(no_checksums
, sctp_checksum_disable
, bool, 0644);
1603 MODULE_PARM_DESC(no_checksums
, "Disable checksums computing and verification");
1604 MODULE_LICENSE("GPL");