Back to home page

OSCL-LXR
 
 

     

0001 // SPDX-License-Identifier: GPL-2.0-or-later
0002 /* SCTP kernel implementation
0003  * (C) Copyright IBM Corp. 2001, 2004
0004  * Copyright (c) 1999-2000 Cisco, Inc.
0005  * Copyright (c) 1999-2001 Motorola, Inc.
0006  * Copyright (c) 2001 Intel Corp.
0007  * Copyright (c) 2001 Nokia, Inc.
0008  * Copyright (c) 2001 La Monte H.P. Yarroll
0009  *
0010  * This file is part of the SCTP kernel implementation
0011  *
0012  * Initialization/cleanup for SCTP protocol support.
0013  *
0014  * Please send any bug reports or fixes you make to the
0015  * email address(es):
0016  *    lksctp developers <linux-sctp@vger.kernel.org>
0017  *
0018  * Written or modified by:
0019  *    La Monte H.P. Yarroll <piggy@acm.org>
0020  *    Karl Knutson <karl@athena.chicago.il.us>
0021  *    Jon Grimm <jgrimm@us.ibm.com>
0022  *    Sridhar Samudrala <sri@us.ibm.com>
0023  *    Daisy Chang <daisyc@us.ibm.com>
0024  *    Ardelle Fan <ardelle.fan@intel.com>
0025  */
0026 
0027 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
0028 
0029 #include <linux/module.h>
0030 #include <linux/init.h>
0031 #include <linux/netdevice.h>
0032 #include <linux/inetdevice.h>
0033 #include <linux/seq_file.h>
0034 #include <linux/memblock.h>
0035 #include <linux/highmem.h>
0036 #include <linux/slab.h>
0037 #include <net/net_namespace.h>
0038 #include <net/protocol.h>
0039 #include <net/ip.h>
0040 #include <net/ipv6.h>
0041 #include <net/route.h>
0042 #include <net/sctp/sctp.h>
0043 #include <net/addrconf.h>
0044 #include <net/inet_common.h>
0045 #include <net/inet_ecn.h>
0046 #include <net/udp_tunnel.h>
0047 
0048 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
0049 
0050 /* Global data structures. */
0051 struct sctp_globals sctp_globals __read_mostly;
0052 
0053 struct idr sctp_assocs_id;
0054 DEFINE_SPINLOCK(sctp_assocs_id_lock);
0055 
0056 static struct sctp_pf *sctp_pf_inet6_specific;
0057 static struct sctp_pf *sctp_pf_inet_specific;
0058 static struct sctp_af *sctp_af_v4_specific;
0059 static struct sctp_af *sctp_af_v6_specific;
0060 
0061 struct kmem_cache *sctp_chunk_cachep __read_mostly;
0062 struct kmem_cache *sctp_bucket_cachep __read_mostly;
0063 
0064 long sysctl_sctp_mem[3];
0065 int sysctl_sctp_rmem[3];
0066 int sysctl_sctp_wmem[3];
0067 
0068 /* Private helper to extract ipv4 address and stash them in
0069  * the protocol structure.
0070  */
0071 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
0072                   struct net_device *dev)
0073 {
0074     struct in_device *in_dev;
0075     struct in_ifaddr *ifa;
0076     struct sctp_sockaddr_entry *addr;
0077 
0078     rcu_read_lock();
0079     if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
0080         rcu_read_unlock();
0081         return;
0082     }
0083 
0084     in_dev_for_each_ifa_rcu(ifa, in_dev) {
0085         /* Add the address to the local list.  */
0086         addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
0087         if (addr) {
0088             addr->a.v4.sin_family = AF_INET;
0089             addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
0090             addr->valid = 1;
0091             INIT_LIST_HEAD(&addr->list);
0092             list_add_tail(&addr->list, addrlist);
0093         }
0094     }
0095 
0096     rcu_read_unlock();
0097 }
0098 
0099 /* Extract our IP addresses from the system and stash them in the
0100  * protocol structure.
0101  */
0102 static void sctp_get_local_addr_list(struct net *net)
0103 {
0104     struct net_device *dev;
0105     struct list_head *pos;
0106     struct sctp_af *af;
0107 
0108     rcu_read_lock();
0109     for_each_netdev_rcu(net, dev) {
0110         list_for_each(pos, &sctp_address_families) {
0111             af = list_entry(pos, struct sctp_af, list);
0112             af->copy_addrlist(&net->sctp.local_addr_list, dev);
0113         }
0114     }
0115     rcu_read_unlock();
0116 }
0117 
0118 /* Free the existing local addresses.  */
0119 static void sctp_free_local_addr_list(struct net *net)
0120 {
0121     struct sctp_sockaddr_entry *addr;
0122     struct list_head *pos, *temp;
0123 
0124     list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
0125         addr = list_entry(pos, struct sctp_sockaddr_entry, list);
0126         list_del(pos);
0127         kfree(addr);
0128     }
0129 }
0130 
0131 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
0132 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
0133                   enum sctp_scope scope, gfp_t gfp, int copy_flags)
0134 {
0135     struct sctp_sockaddr_entry *addr;
0136     union sctp_addr laddr;
0137     int error = 0;
0138 
0139     rcu_read_lock();
0140     list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
0141         if (!addr->valid)
0142             continue;
0143         if (!sctp_in_scope(net, &addr->a, scope))
0144             continue;
0145 
0146         /* Now that the address is in scope, check to see if
0147          * the address type is really supported by the local
0148          * sock as well as the remote peer.
0149          */
0150         if (addr->a.sa.sa_family == AF_INET &&
0151             (!(copy_flags & SCTP_ADDR4_ALLOWED) ||
0152              !(copy_flags & SCTP_ADDR4_PEERSUPP)))
0153             continue;
0154         if (addr->a.sa.sa_family == AF_INET6 &&
0155             (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
0156              !(copy_flags & SCTP_ADDR6_PEERSUPP)))
0157             continue;
0158 
0159         laddr = addr->a;
0160         /* also works for setting ipv6 address port */
0161         laddr.v4.sin_port = htons(bp->port);
0162         if (sctp_bind_addr_state(bp, &laddr) != -1)
0163             continue;
0164 
0165         error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
0166                        SCTP_ADDR_SRC, GFP_ATOMIC);
0167         if (error)
0168             break;
0169     }
0170 
0171     rcu_read_unlock();
0172     return error;
0173 }
0174 
0175 /* Copy over any ip options */
0176 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
0177 {
0178     struct inet_sock *newinet, *inet = inet_sk(sk);
0179     struct ip_options_rcu *inet_opt, *newopt = NULL;
0180 
0181     newinet = inet_sk(newsk);
0182 
0183     rcu_read_lock();
0184     inet_opt = rcu_dereference(inet->inet_opt);
0185     if (inet_opt) {
0186         newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
0187                       inet_opt->opt.optlen, GFP_ATOMIC);
0188         if (newopt)
0189             memcpy(newopt, inet_opt, sizeof(*inet_opt) +
0190                    inet_opt->opt.optlen);
0191         else
0192             pr_err("%s: Failed to copy ip options\n", __func__);
0193     }
0194     RCU_INIT_POINTER(newinet->inet_opt, newopt);
0195     rcu_read_unlock();
0196 }
0197 
0198 /* Account for the IP options */
0199 static int sctp_v4_ip_options_len(struct sock *sk)
0200 {
0201     struct inet_sock *inet = inet_sk(sk);
0202     struct ip_options_rcu *inet_opt;
0203     int len = 0;
0204 
0205     rcu_read_lock();
0206     inet_opt = rcu_dereference(inet->inet_opt);
0207     if (inet_opt)
0208         len = inet_opt->opt.optlen;
0209 
0210     rcu_read_unlock();
0211     return len;
0212 }
0213 
0214 /* Initialize a sctp_addr from in incoming skb.  */
0215 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
0216                  int is_saddr)
0217 {
0218     /* Always called on head skb, so this is safe */
0219     struct sctphdr *sh = sctp_hdr(skb);
0220     struct sockaddr_in *sa = &addr->v4;
0221 
0222     addr->v4.sin_family = AF_INET;
0223 
0224     if (is_saddr) {
0225         sa->sin_port = sh->source;
0226         sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
0227     } else {
0228         sa->sin_port = sh->dest;
0229         sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
0230     }
0231     memset(sa->sin_zero, 0, sizeof(sa->sin_zero));
0232 }
0233 
0234 /* Initialize an sctp_addr from a socket. */
0235 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
0236 {
0237     addr->v4.sin_family = AF_INET;
0238     addr->v4.sin_port = 0;
0239     addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
0240     memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
0241 }
0242 
0243 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
0244 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
0245 {
0246     inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
0247 }
0248 
0249 /* Initialize sk->sk_daddr from sctp_addr. */
0250 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
0251 {
0252     inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
0253 }
0254 
0255 /* Initialize a sctp_addr from an address parameter. */
0256 static bool sctp_v4_from_addr_param(union sctp_addr *addr,
0257                     union sctp_addr_param *param,
0258                     __be16 port, int iif)
0259 {
0260     if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param))
0261         return false;
0262 
0263     addr->v4.sin_family = AF_INET;
0264     addr->v4.sin_port = port;
0265     addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
0266     memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
0267 
0268     return true;
0269 }
0270 
0271 /* Initialize an address parameter from a sctp_addr and return the length
0272  * of the address parameter.
0273  */
0274 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
0275                  union sctp_addr_param *param)
0276 {
0277     int length = sizeof(struct sctp_ipv4addr_param);
0278 
0279     param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
0280     param->v4.param_hdr.length = htons(length);
0281     param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
0282 
0283     return length;
0284 }
0285 
0286 /* Initialize a sctp_addr from a dst_entry. */
0287 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
0288                   __be16 port)
0289 {
0290     saddr->v4.sin_family = AF_INET;
0291     saddr->v4.sin_port = port;
0292     saddr->v4.sin_addr.s_addr = fl4->saddr;
0293     memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
0294 }
0295 
0296 /* Compare two addresses exactly. */
0297 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
0298                 const union sctp_addr *addr2)
0299 {
0300     if (addr1->sa.sa_family != addr2->sa.sa_family)
0301         return 0;
0302     if (addr1->v4.sin_port != addr2->v4.sin_port)
0303         return 0;
0304     if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
0305         return 0;
0306 
0307     return 1;
0308 }
0309 
0310 /* Initialize addr struct to INADDR_ANY. */
0311 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
0312 {
0313     addr->v4.sin_family = AF_INET;
0314     addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
0315     addr->v4.sin_port = port;
0316     memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
0317 }
0318 
0319 /* Is this a wildcard address? */
0320 static int sctp_v4_is_any(const union sctp_addr *addr)
0321 {
0322     return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
0323 }
0324 
0325 /* This function checks if the address is a valid address to be used for
0326  * SCTP binding.
0327  *
0328  * Output:
0329  * Return 0 - If the address is a non-unicast or an illegal address.
0330  * Return 1 - If the address is a unicast.
0331  */
0332 static int sctp_v4_addr_valid(union sctp_addr *addr,
0333                   struct sctp_sock *sp,
0334                   const struct sk_buff *skb)
0335 {
0336     /* IPv4 addresses not allowed */
0337     if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
0338         return 0;
0339 
0340     /* Is this a non-unicast address or a unusable SCTP address? */
0341     if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
0342         return 0;
0343 
0344     /* Is this a broadcast address? */
0345     if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
0346         return 0;
0347 
0348     return 1;
0349 }
0350 
0351 /* Should this be available for binding?   */
0352 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
0353 {
0354     struct net *net = sock_net(&sp->inet.sk);
0355     int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
0356 
0357 
0358     if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
0359        ret != RTN_LOCAL &&
0360        !sp->inet.freebind &&
0361         !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind))
0362         return 0;
0363 
0364     if (ipv6_only_sock(sctp_opt2sk(sp)))
0365         return 0;
0366 
0367     return 1;
0368 }
0369 
0370 /* Checking the loopback, private and other address scopes as defined in
0371  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
0372  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
0373  *
0374  * Level 0 - unusable SCTP addresses
0375  * Level 1 - loopback address
0376  * Level 2 - link-local addresses
0377  * Level 3 - private addresses.
0378  * Level 4 - global addresses
0379  * For INIT and INIT-ACK address list, let L be the level of
0380  * requested destination address, sender and receiver
0381  * SHOULD include all of its addresses with level greater
0382  * than or equal to L.
0383  *
0384  * IPv4 scoping can be controlled through sysctl option
0385  * net.sctp.addr_scope_policy
0386  */
0387 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
0388 {
0389     enum sctp_scope retval;
0390 
0391     /* Check for unusable SCTP addresses. */
0392     if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
0393         retval =  SCTP_SCOPE_UNUSABLE;
0394     } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
0395         retval = SCTP_SCOPE_LOOPBACK;
0396     } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
0397         retval = SCTP_SCOPE_LINK;
0398     } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
0399            ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
0400            ipv4_is_private_192(addr->v4.sin_addr.s_addr) ||
0401            ipv4_is_test_198(addr->v4.sin_addr.s_addr)) {
0402         retval = SCTP_SCOPE_PRIVATE;
0403     } else {
0404         retval = SCTP_SCOPE_GLOBAL;
0405     }
0406 
0407     return retval;
0408 }
0409 
0410 /* Returns a valid dst cache entry for the given source and destination ip
0411  * addresses. If an association is passed, trys to get a dst entry with a
0412  * source address that matches an address in the bind address list.
0413  */
0414 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
0415                 struct flowi *fl, struct sock *sk)
0416 {
0417     struct sctp_association *asoc = t->asoc;
0418     struct rtable *rt;
0419     struct flowi _fl;
0420     struct flowi4 *fl4 = &_fl.u.ip4;
0421     struct sctp_bind_addr *bp;
0422     struct sctp_sockaddr_entry *laddr;
0423     struct dst_entry *dst = NULL;
0424     union sctp_addr *daddr = &t->ipaddr;
0425     union sctp_addr dst_saddr;
0426     __u8 tos = inet_sk(sk)->tos;
0427 
0428     if (t->dscp & SCTP_DSCP_SET_MASK)
0429         tos = t->dscp & SCTP_DSCP_VAL_MASK;
0430     memset(&_fl, 0x0, sizeof(_fl));
0431     fl4->daddr  = daddr->v4.sin_addr.s_addr;
0432     fl4->fl4_dport = daddr->v4.sin_port;
0433     fl4->flowi4_proto = IPPROTO_SCTP;
0434     if (asoc) {
0435         fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
0436         fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
0437         fl4->fl4_sport = htons(asoc->base.bind_addr.port);
0438     }
0439     if (saddr) {
0440         fl4->saddr = saddr->v4.sin_addr.s_addr;
0441         if (!fl4->fl4_sport)
0442             fl4->fl4_sport = saddr->v4.sin_port;
0443     }
0444 
0445     pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
0446          &fl4->saddr);
0447 
0448     rt = ip_route_output_key(sock_net(sk), fl4);
0449     if (!IS_ERR(rt)) {
0450         dst = &rt->dst;
0451         t->dst = dst;
0452         memcpy(fl, &_fl, sizeof(_fl));
0453     }
0454 
0455     /* If there is no association or if a source address is passed, no
0456      * more validation is required.
0457      */
0458     if (!asoc || saddr)
0459         goto out;
0460 
0461     bp = &asoc->base.bind_addr;
0462 
0463     if (dst) {
0464         /* Walk through the bind address list and look for a bind
0465          * address that matches the source address of the returned dst.
0466          */
0467         sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
0468         rcu_read_lock();
0469         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
0470             if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
0471                 (laddr->state != SCTP_ADDR_SRC &&
0472                 !asoc->src_out_of_asoc_ok))
0473                 continue;
0474             if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
0475                 goto out_unlock;
0476         }
0477         rcu_read_unlock();
0478 
0479         /* None of the bound addresses match the source address of the
0480          * dst. So release it.
0481          */
0482         dst_release(dst);
0483         dst = NULL;
0484     }
0485 
0486     /* Walk through the bind address list and try to get a dst that
0487      * matches a bind address as the source address.
0488      */
0489     rcu_read_lock();
0490     list_for_each_entry_rcu(laddr, &bp->address_list, list) {
0491         struct net_device *odev;
0492 
0493         if (!laddr->valid)
0494             continue;
0495         if (laddr->state != SCTP_ADDR_SRC ||
0496             AF_INET != laddr->a.sa.sa_family)
0497             continue;
0498 
0499         fl4->fl4_sport = laddr->a.v4.sin_port;
0500         flowi4_update_output(fl4,
0501                      asoc->base.sk->sk_bound_dev_if,
0502                      RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
0503                      daddr->v4.sin_addr.s_addr,
0504                      laddr->a.v4.sin_addr.s_addr);
0505 
0506         rt = ip_route_output_key(sock_net(sk), fl4);
0507         if (IS_ERR(rt))
0508             continue;
0509 
0510         /* Ensure the src address belongs to the output
0511          * interface.
0512          */
0513         odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
0514                      false);
0515         if (!odev || odev->ifindex != fl4->flowi4_oif) {
0516             if (!dst) {
0517                 dst = &rt->dst;
0518                 t->dst = dst;
0519                 memcpy(fl, &_fl, sizeof(_fl));
0520             } else {
0521                 dst_release(&rt->dst);
0522             }
0523             continue;
0524         }
0525 
0526         dst_release(dst);
0527         dst = &rt->dst;
0528         t->dst = dst;
0529         memcpy(fl, &_fl, sizeof(_fl));
0530         break;
0531     }
0532 
0533 out_unlock:
0534     rcu_read_unlock();
0535 out:
0536     if (dst) {
0537         pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
0538              &fl->u.ip4.daddr, &fl->u.ip4.saddr);
0539     } else {
0540         t->dst = NULL;
0541         pr_debug("no route\n");
0542     }
0543 }
0544 
0545 /* For v4, the source address is cached in the route entry(dst). So no need
0546  * to cache it separately and hence this is an empty routine.
0547  */
0548 static void sctp_v4_get_saddr(struct sctp_sock *sk,
0549                   struct sctp_transport *t,
0550                   struct flowi *fl)
0551 {
0552     union sctp_addr *saddr = &t->saddr;
0553     struct rtable *rt = (struct rtable *)t->dst;
0554 
0555     if (rt) {
0556         saddr->v4.sin_family = AF_INET;
0557         saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
0558     }
0559 }
0560 
0561 /* What interface did this skb arrive on? */
0562 static int sctp_v4_skb_iif(const struct sk_buff *skb)
0563 {
0564     return inet_iif(skb);
0565 }
0566 
0567 /* Was this packet marked by Explicit Congestion Notification? */
0568 static int sctp_v4_is_ce(const struct sk_buff *skb)
0569 {
0570     return INET_ECN_is_ce(ip_hdr(skb)->tos);
0571 }
0572 
0573 /* Create and initialize a new sk for the socket returned by accept(). */
0574 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
0575                          struct sctp_association *asoc,
0576                          bool kern)
0577 {
0578     struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
0579             sk->sk_prot, kern);
0580     struct inet_sock *newinet;
0581 
0582     if (!newsk)
0583         goto out;
0584 
0585     sock_init_data(NULL, newsk);
0586 
0587     sctp_copy_sock(newsk, sk, asoc);
0588     sock_reset_flag(newsk, SOCK_ZAPPED);
0589 
0590     sctp_v4_copy_ip_options(sk, newsk);
0591 
0592     newinet = inet_sk(newsk);
0593 
0594     newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
0595 
0596     sk_refcnt_debug_inc(newsk);
0597 
0598     if (newsk->sk_prot->init(newsk)) {
0599         sk_common_release(newsk);
0600         newsk = NULL;
0601     }
0602 
0603 out:
0604     return newsk;
0605 }
0606 
0607 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
0608 {
0609     /* No address mapping for V4 sockets */
0610     memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
0611     return sizeof(struct sockaddr_in);
0612 }
0613 
0614 /* Dump the v4 addr to the seq file. */
0615 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
0616 {
0617     seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
0618 }
0619 
0620 static void sctp_v4_ecn_capable(struct sock *sk)
0621 {
0622     INET_ECN_xmit(sk);
0623 }
0624 
0625 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
0626 {
0627     struct net *net = from_timer(net, t, sctp.addr_wq_timer);
0628     struct sctp_sockaddr_entry *addrw, *temp;
0629     struct sctp_sock *sp;
0630 
0631     spin_lock_bh(&net->sctp.addr_wq_lock);
0632 
0633     list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
0634         pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
0635              "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
0636              addrw->state, addrw);
0637 
0638 #if IS_ENABLED(CONFIG_IPV6)
0639         /* Now we send an ASCONF for each association */
0640         /* Note. we currently don't handle link local IPv6 addressees */
0641         if (addrw->a.sa.sa_family == AF_INET6) {
0642             struct in6_addr *in6;
0643 
0644             if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
0645                 IPV6_ADDR_LINKLOCAL)
0646                 goto free_next;
0647 
0648             in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
0649             if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
0650                 addrw->state == SCTP_ADDR_NEW) {
0651                 unsigned long timeo_val;
0652 
0653                 pr_debug("%s: this is on DAD, trying %d sec "
0654                      "later\n", __func__,
0655                      SCTP_ADDRESS_TICK_DELAY);
0656 
0657                 timeo_val = jiffies;
0658                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
0659                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
0660                 break;
0661             }
0662         }
0663 #endif
0664         list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
0665             struct sock *sk;
0666 
0667             sk = sctp_opt2sk(sp);
0668             /* ignore bound-specific endpoints */
0669             if (!sctp_is_ep_boundall(sk))
0670                 continue;
0671             bh_lock_sock(sk);
0672             if (sctp_asconf_mgmt(sp, addrw) < 0)
0673                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
0674             bh_unlock_sock(sk);
0675         }
0676 #if IS_ENABLED(CONFIG_IPV6)
0677 free_next:
0678 #endif
0679         list_del(&addrw->list);
0680         kfree(addrw);
0681     }
0682     spin_unlock_bh(&net->sctp.addr_wq_lock);
0683 }
0684 
0685 static void sctp_free_addr_wq(struct net *net)
0686 {
0687     struct sctp_sockaddr_entry *addrw;
0688     struct sctp_sockaddr_entry *temp;
0689 
0690     spin_lock_bh(&net->sctp.addr_wq_lock);
0691     del_timer(&net->sctp.addr_wq_timer);
0692     list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
0693         list_del(&addrw->list);
0694         kfree(addrw);
0695     }
0696     spin_unlock_bh(&net->sctp.addr_wq_lock);
0697 }
0698 
0699 /* lookup the entry for the same address in the addr_waitq
0700  * sctp_addr_wq MUST be locked
0701  */
0702 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
0703                     struct sctp_sockaddr_entry *addr)
0704 {
0705     struct sctp_sockaddr_entry *addrw;
0706 
0707     list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
0708         if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
0709             continue;
0710         if (addrw->a.sa.sa_family == AF_INET) {
0711             if (addrw->a.v4.sin_addr.s_addr ==
0712                 addr->a.v4.sin_addr.s_addr)
0713                 return addrw;
0714         } else if (addrw->a.sa.sa_family == AF_INET6) {
0715             if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
0716                 &addr->a.v6.sin6_addr))
0717                 return addrw;
0718         }
0719     }
0720     return NULL;
0721 }
0722 
0723 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
0724 {
0725     struct sctp_sockaddr_entry *addrw;
0726     unsigned long timeo_val;
0727 
0728     /* first, we check if an opposite message already exist in the queue.
0729      * If we found such message, it is removed.
0730      * This operation is a bit stupid, but the DHCP client attaches the
0731      * new address after a couple of addition and deletion of that address
0732      */
0733 
0734     spin_lock_bh(&net->sctp.addr_wq_lock);
0735     /* Offsets existing events in addr_wq */
0736     addrw = sctp_addr_wq_lookup(net, addr);
0737     if (addrw) {
0738         if (addrw->state != cmd) {
0739             pr_debug("%s: offsets existing entry for %d, addr:%pISc "
0740                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
0741                  &net->sctp.addr_waitq);
0742 
0743             list_del(&addrw->list);
0744             kfree(addrw);
0745         }
0746         spin_unlock_bh(&net->sctp.addr_wq_lock);
0747         return;
0748     }
0749 
0750     /* OK, we have to add the new address to the wait queue */
0751     addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
0752     if (addrw == NULL) {
0753         spin_unlock_bh(&net->sctp.addr_wq_lock);
0754         return;
0755     }
0756     addrw->state = cmd;
0757     list_add_tail(&addrw->list, &net->sctp.addr_waitq);
0758 
0759     pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
0760          __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
0761 
0762     if (!timer_pending(&net->sctp.addr_wq_timer)) {
0763         timeo_val = jiffies;
0764         timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
0765         mod_timer(&net->sctp.addr_wq_timer, timeo_val);
0766     }
0767     spin_unlock_bh(&net->sctp.addr_wq_lock);
0768 }
0769 
0770 /* Event handler for inet address addition/deletion events.
0771  * The sctp_local_addr_list needs to be protocted by a spin lock since
0772  * multiple notifiers (say IPv4 and IPv6) may be running at the same
0773  * time and thus corrupt the list.
0774  * The reader side is protected with RCU.
0775  */
0776 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
0777                    void *ptr)
0778 {
0779     struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
0780     struct sctp_sockaddr_entry *addr = NULL;
0781     struct sctp_sockaddr_entry *temp;
0782     struct net *net = dev_net(ifa->ifa_dev->dev);
0783     int found = 0;
0784 
0785     switch (ev) {
0786     case NETDEV_UP:
0787         addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
0788         if (addr) {
0789             addr->a.v4.sin_family = AF_INET;
0790             addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
0791             addr->valid = 1;
0792             spin_lock_bh(&net->sctp.local_addr_lock);
0793             list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
0794             sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
0795             spin_unlock_bh(&net->sctp.local_addr_lock);
0796         }
0797         break;
0798     case NETDEV_DOWN:
0799         spin_lock_bh(&net->sctp.local_addr_lock);
0800         list_for_each_entry_safe(addr, temp,
0801                     &net->sctp.local_addr_list, list) {
0802             if (addr->a.sa.sa_family == AF_INET &&
0803                     addr->a.v4.sin_addr.s_addr ==
0804                     ifa->ifa_local) {
0805                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
0806                 found = 1;
0807                 addr->valid = 0;
0808                 list_del_rcu(&addr->list);
0809                 break;
0810             }
0811         }
0812         spin_unlock_bh(&net->sctp.local_addr_lock);
0813         if (found)
0814             kfree_rcu(addr, rcu);
0815         break;
0816     }
0817 
0818     return NOTIFY_DONE;
0819 }
0820 
0821 /*
0822  * Initialize the control inode/socket with a control endpoint data
0823  * structure.  This endpoint is reserved exclusively for the OOTB processing.
0824  */
0825 static int sctp_ctl_sock_init(struct net *net)
0826 {
0827     int err;
0828     sa_family_t family = PF_INET;
0829 
0830     if (sctp_get_pf_specific(PF_INET6))
0831         family = PF_INET6;
0832 
0833     err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
0834                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
0835 
0836     /* If IPv6 socket could not be created, try the IPv4 socket */
0837     if (err < 0 && family == PF_INET6)
0838         err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
0839                        SOCK_SEQPACKET, IPPROTO_SCTP,
0840                        net);
0841 
0842     if (err < 0) {
0843         pr_err("Failed to create the SCTP control socket\n");
0844         return err;
0845     }
0846     return 0;
0847 }
0848 
0849 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb)
0850 {
0851     SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source;
0852 
0853     skb_set_transport_header(skb, sizeof(struct udphdr));
0854     sctp_rcv(skb);
0855     return 0;
0856 }
0857 
0858 int sctp_udp_sock_start(struct net *net)
0859 {
0860     struct udp_tunnel_sock_cfg tuncfg = {NULL};
0861     struct udp_port_cfg udp_conf = {0};
0862     struct socket *sock;
0863     int err;
0864 
0865     udp_conf.family = AF_INET;
0866     udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
0867     udp_conf.local_udp_port = htons(net->sctp.udp_port);
0868     err = udp_sock_create(net, &udp_conf, &sock);
0869     if (err) {
0870         pr_err("Failed to create the SCTP UDP tunneling v4 sock\n");
0871         return err;
0872     }
0873 
0874     tuncfg.encap_type = 1;
0875     tuncfg.encap_rcv = sctp_udp_rcv;
0876     tuncfg.encap_err_lookup = sctp_udp_v4_err;
0877     setup_udp_tunnel_sock(net, sock, &tuncfg);
0878     net->sctp.udp4_sock = sock->sk;
0879 
0880 #if IS_ENABLED(CONFIG_IPV6)
0881     memset(&udp_conf, 0, sizeof(udp_conf));
0882 
0883     udp_conf.family = AF_INET6;
0884     udp_conf.local_ip6 = in6addr_any;
0885     udp_conf.local_udp_port = htons(net->sctp.udp_port);
0886     udp_conf.use_udp6_rx_checksums = true;
0887     udp_conf.ipv6_v6only = true;
0888     err = udp_sock_create(net, &udp_conf, &sock);
0889     if (err) {
0890         pr_err("Failed to create the SCTP UDP tunneling v6 sock\n");
0891         udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
0892         net->sctp.udp4_sock = NULL;
0893         return err;
0894     }
0895 
0896     tuncfg.encap_type = 1;
0897     tuncfg.encap_rcv = sctp_udp_rcv;
0898     tuncfg.encap_err_lookup = sctp_udp_v6_err;
0899     setup_udp_tunnel_sock(net, sock, &tuncfg);
0900     net->sctp.udp6_sock = sock->sk;
0901 #endif
0902 
0903     return 0;
0904 }
0905 
0906 void sctp_udp_sock_stop(struct net *net)
0907 {
0908     if (net->sctp.udp4_sock) {
0909         udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket);
0910         net->sctp.udp4_sock = NULL;
0911     }
0912     if (net->sctp.udp6_sock) {
0913         udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket);
0914         net->sctp.udp6_sock = NULL;
0915     }
0916 }
0917 
0918 /* Register address family specific functions. */
0919 int sctp_register_af(struct sctp_af *af)
0920 {
0921     switch (af->sa_family) {
0922     case AF_INET:
0923         if (sctp_af_v4_specific)
0924             return 0;
0925         sctp_af_v4_specific = af;
0926         break;
0927     case AF_INET6:
0928         if (sctp_af_v6_specific)
0929             return 0;
0930         sctp_af_v6_specific = af;
0931         break;
0932     default:
0933         return 0;
0934     }
0935 
0936     INIT_LIST_HEAD(&af->list);
0937     list_add_tail(&af->list, &sctp_address_families);
0938     return 1;
0939 }
0940 
0941 /* Get the table of functions for manipulating a particular address
0942  * family.
0943  */
0944 struct sctp_af *sctp_get_af_specific(sa_family_t family)
0945 {
0946     switch (family) {
0947     case AF_INET:
0948         return sctp_af_v4_specific;
0949     case AF_INET6:
0950         return sctp_af_v6_specific;
0951     default:
0952         return NULL;
0953     }
0954 }
0955 
0956 /* Common code to initialize a AF_INET msg_name. */
0957 static void sctp_inet_msgname(char *msgname, int *addr_len)
0958 {
0959     struct sockaddr_in *sin;
0960 
0961     sin = (struct sockaddr_in *)msgname;
0962     *addr_len = sizeof(struct sockaddr_in);
0963     sin->sin_family = AF_INET;
0964     memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
0965 }
0966 
0967 /* Copy the primary address of the peer primary address as the msg_name. */
0968 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
0969                     int *addr_len)
0970 {
0971     struct sockaddr_in *sin, *sinfrom;
0972 
0973     if (msgname) {
0974         struct sctp_association *asoc;
0975 
0976         asoc = event->asoc;
0977         sctp_inet_msgname(msgname, addr_len);
0978         sin = (struct sockaddr_in *)msgname;
0979         sinfrom = &asoc->peer.primary_addr.v4;
0980         sin->sin_port = htons(asoc->peer.port);
0981         sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
0982     }
0983 }
0984 
0985 /* Initialize and copy out a msgname from an inbound skb. */
0986 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
0987 {
0988     if (msgname) {
0989         struct sctphdr *sh = sctp_hdr(skb);
0990         struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
0991 
0992         sctp_inet_msgname(msgname, len);
0993         sin->sin_port = sh->source;
0994         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
0995     }
0996 }
0997 
0998 /* Do we support this AF? */
0999 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
1000 {
1001     /* PF_INET only supports AF_INET addresses. */
1002     return AF_INET == family;
1003 }
1004 
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)
1009 {
1010     /* PF_INET only supports AF_INET addresses. */
1011     if (addr1->sa.sa_family != addr2->sa.sa_family)
1012         return 0;
1013     if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
1014         htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
1015         return 1;
1016     if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
1017         return 1;
1018 
1019     return 0;
1020 }
1021 
1022 /* Verify that provided sockaddr looks bindable.  Common verification has
1023  * already been taken care of.
1024  */
1025 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
1026 {
1027     return sctp_v4_available(addr, opt);
1028 }
1029 
1030 /* Verify that sockaddr looks sendable.  Common verification has already
1031  * been taken care of.
1032  */
1033 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
1034 {
1035     return 1;
1036 }
1037 
1038 /* Fill in Supported Address Type information for INIT and INIT-ACK
1039  * chunks.  Returns number of addresses supported.
1040  */
1041 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
1042                      __be16 *types)
1043 {
1044     types[0] = SCTP_PARAM_IPV4_ADDRESS;
1045     return 1;
1046 }
1047 
1048 /* Wrapper routine that calls the ip transmit routine. */
1049 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t)
1050 {
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;
1056     __be16 df = 0;
1057 
1058     pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
1059          skb->len, &fl4->saddr, &fl4->daddr);
1060 
1061     if (t->dscp & SCTP_DSCP_SET_MASK)
1062         dscp = t->dscp & SCTP_DSCP_VAL_MASK;
1063 
1064     inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO
1065                                : IP_PMTUDISC_DONT;
1066     SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS);
1067 
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);
1071     }
1072 
1073     if (skb_is_gso(skb))
1074         skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
1075 
1076     if (ip_dont_fragment(sk, dst) && !skb->ignore_df)
1077         df = htons(IP_DF);
1078 
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);
1086     return 0;
1087 }
1088 
1089 static struct sctp_af sctp_af_inet;
1090 
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,
1104     .af            = &sctp_af_inet
1105 };
1106 
1107 /* Notifier for inetaddr addition/deletion events.  */
1108 static struct notifier_block sctp_inetaddr_notifier = {
1109     .notifier_call = sctp_inetaddr_event,
1110 };
1111 
1112 /* Socket operations.  */
1113 static const struct proto_ops inet_seqpacket_ops = {
1114     .family        = PF_INET,
1115     .owner         = THIS_MODULE,
1116     .release       = inet_release,  /* Needs to be wrapped... */
1117     .bind          = inet_bind,
1118     .connect       = sctp_inet_connect,
1119     .socketpair    = sock_no_socketpair,
1120     .accept        = inet_accept,
1121     .getname       = inet_getname,  /* Semantics are different.  */
1122     .poll          = sctp_poll,
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,
1133 };
1134 
1135 /* Registration with AF_INET family.  */
1136 static struct inet_protosw sctp_seqpacket_protosw = {
1137     .type       = SOCK_SEQPACKET,
1138     .protocol   = IPPROTO_SCTP,
1139     .prot       = &sctp_prot,
1140     .ops        = &inet_seqpacket_ops,
1141     .flags      = SCTP_PROTOSW_FLAG
1142 };
1143 static struct inet_protosw sctp_stream_protosw = {
1144     .type       = SOCK_STREAM,
1145     .protocol   = IPPROTO_SCTP,
1146     .prot       = &sctp_prot,
1147     .ops        = &inet_seqpacket_ops,
1148     .flags      = SCTP_PROTOSW_FLAG
1149 };
1150 
1151 static int sctp4_rcv(struct sk_buff *skb)
1152 {
1153     SCTP_INPUT_CB(skb)->encap_port = 0;
1154     return sctp_rcv(skb);
1155 }
1156 
1157 /* Register with IP layer.  */
1158 static const struct net_protocol sctp_protocol = {
1159     .handler     = sctp4_rcv,
1160     .err_handler = sctp_v4_err,
1161     .no_policy   = 1,
1162     .icmp_strict_tag_validation = 1,
1163 };
1164 
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,
1191 };
1192 
1193 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1194 {
1195     switch (family) {
1196     case PF_INET:
1197         return sctp_pf_inet_specific;
1198     case PF_INET6:
1199         return sctp_pf_inet6_specific;
1200     default:
1201         return NULL;
1202     }
1203 }
1204 
1205 /* Register the PF specific function table.  */
1206 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1207 {
1208     switch (family) {
1209     case PF_INET:
1210         if (sctp_pf_inet_specific)
1211             return 0;
1212         sctp_pf_inet_specific = pf;
1213         break;
1214     case PF_INET6:
1215         if (sctp_pf_inet6_specific)
1216             return 0;
1217         sctp_pf_inet6_specific = pf;
1218         break;
1219     default:
1220         return 0;
1221     }
1222     return 1;
1223 }
1224 
1225 static inline int init_sctp_mibs(struct net *net)
1226 {
1227     net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1228     if (!net->sctp.sctp_statistics)
1229         return -ENOMEM;
1230     return 0;
1231 }
1232 
1233 static inline void cleanup_sctp_mibs(struct net *net)
1234 {
1235     free_percpu(net->sctp.sctp_statistics);
1236 }
1237 
1238 static void sctp_v4_pf_init(void)
1239 {
1240     /* Initialize the SCTP specific PF functions. */
1241     sctp_register_pf(&sctp_pf_inet, PF_INET);
1242     sctp_register_af(&sctp_af_inet);
1243 }
1244 
1245 static void sctp_v4_pf_exit(void)
1246 {
1247     list_del(&sctp_af_inet.list);
1248 }
1249 
1250 static int sctp_v4_protosw_init(void)
1251 {
1252     int rc;
1253 
1254     rc = proto_register(&sctp_prot, 1);
1255     if (rc)
1256         return rc;
1257 
1258     /* Register SCTP(UDP and TCP style) with socket layer.  */
1259     inet_register_protosw(&sctp_seqpacket_protosw);
1260     inet_register_protosw(&sctp_stream_protosw);
1261 
1262     return 0;
1263 }
1264 
1265 static void sctp_v4_protosw_exit(void)
1266 {
1267     inet_unregister_protosw(&sctp_stream_protosw);
1268     inet_unregister_protosw(&sctp_seqpacket_protosw);
1269     proto_unregister(&sctp_prot);
1270 }
1271 
1272 static int sctp_v4_add_protocol(void)
1273 {
1274     /* Register notifier for inet address additions/deletions. */
1275     register_inetaddr_notifier(&sctp_inetaddr_notifier);
1276 
1277     /* Register SCTP with inet layer.  */
1278     if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1279         return -EAGAIN;
1280 
1281     return 0;
1282 }
1283 
1284 static void sctp_v4_del_protocol(void)
1285 {
1286     inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1287     unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1288 }
1289 
1290 static int __net_init sctp_defaults_init(struct net *net)
1291 {
1292     int status;
1293 
1294     /*
1295      * 14. Suggested SCTP Protocol Parameter Values
1296      */
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;
1308 
1309     /* Valid.Cookie.Life        - 60  seconds */
1310     net->sctp.valid_cookie_life     = SCTP_DEFAULT_COOKIE_LIFE;
1311 
1312     /* Whether Cookie Preservative is enabled(1) or not(0) */
1313     net->sctp.cookie_preserve_enable    = 1;
1314 
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";
1320 #else
1321     net->sctp.sctp_hmac_alg         = NULL;
1322 #endif
1323 
1324     /* Max.Burst            - 4 */
1325     net->sctp.max_burst         = SCTP_DEFAULT_MAX_BURST;
1326 
1327     /* Disable of Primary Path Switchover by default */
1328     net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1329 
1330     /* Enable pf state by default */
1331     net->sctp.pf_enable = 1;
1332 
1333     /* Ignore pf exposure feature by default */
1334     net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1335 
1336     /* Association.Max.Retrans  - 10 attempts
1337      * Path.Max.Retrans         - 5  attempts (per destination address)
1338      * Max.Init.Retransmits     - 8  attempts
1339      */
1340     net->sctp.max_retrans_association   = 10;
1341     net->sctp.max_retrans_path      = 5;
1342     net->sctp.max_retrans_init      = 8;
1343 
1344     /* Sendbuffer growth        - do per-socket accounting */
1345     net->sctp.sndbuf_policy         = 0;
1346 
1347     /* Rcvbuffer growth     - do per-socket accounting */
1348     net->sctp.rcvbuf_policy         = 0;
1349 
1350     /* HB.interval              - 30 seconds */
1351     net->sctp.hb_interval           = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1352 
1353     /* delayed SACK timeout */
1354     net->sctp.sack_timeout          = SCTP_DEFAULT_TIMEOUT_SACK;
1355 
1356     /* Disable ADDIP by default. */
1357     net->sctp.addip_enable = 0;
1358     net->sctp.addip_noauth = 0;
1359     net->sctp.default_auto_asconf = 0;
1360 
1361     /* Enable PR-SCTP by default. */
1362     net->sctp.prsctp_enable = 1;
1363 
1364     /* Disable RECONF by default. */
1365     net->sctp.reconf_enable = 0;
1366 
1367     /* Disable AUTH by default. */
1368     net->sctp.auth_enable = 0;
1369 
1370     /* Enable ECN by default. */
1371     net->sctp.ecn_enable = 1;
1372 
1373     /* Set UDP tunneling listening port to 0 by default */
1374     net->sctp.udp_port = 0;
1375 
1376     /* Set remote encap port to 0 by default */
1377     net->sctp.encap_port = 0;
1378 
1379     /* Set SCOPE policy to enabled */
1380     net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1381 
1382     /* Set the default rwnd update threshold */
1383     net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1384 
1385     /* Initialize maximum autoclose timeout. */
1386     net->sctp.max_autoclose     = INT_MAX / HZ;
1387 
1388     status = sctp_sysctl_net_register(net);
1389     if (status)
1390         goto err_sysctl_register;
1391 
1392     /* Allocate and initialise sctp mibs.  */
1393     status = init_sctp_mibs(net);
1394     if (status)
1395         goto err_init_mibs;
1396 
1397 #ifdef CONFIG_PROC_FS
1398     /* Initialize proc fs directory.  */
1399     status = sctp_proc_init(net);
1400     if (status)
1401         goto err_init_proc;
1402 #endif
1403 
1404     sctp_dbg_objcnt_init(net);
1405 
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);
1410 
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);
1417 
1418     return 0;
1419 
1420 #ifdef CONFIG_PROC_FS
1421 err_init_proc:
1422     cleanup_sctp_mibs(net);
1423 #endif
1424 err_init_mibs:
1425     sctp_sysctl_net_unregister(net);
1426 err_sysctl_register:
1427     return status;
1428 }
1429 
1430 static void __net_exit sctp_defaults_exit(struct net *net)
1431 {
1432     /* Free the local address list */
1433     sctp_free_addr_wq(net);
1434     sctp_free_local_addr_list(net);
1435 
1436 #ifdef CONFIG_PROC_FS
1437     remove_proc_subtree("sctp", net->proc_net);
1438     net->sctp.proc_net_sctp = NULL;
1439 #endif
1440     cleanup_sctp_mibs(net);
1441     sctp_sysctl_net_unregister(net);
1442 }
1443 
1444 static struct pernet_operations sctp_defaults_ops = {
1445     .init = sctp_defaults_init,
1446     .exit = sctp_defaults_exit,
1447 };
1448 
1449 static int __net_init sctp_ctrlsock_init(struct net *net)
1450 {
1451     int status;
1452 
1453     /* Initialize the control inode/socket for handling OOTB packets.  */
1454     status = sctp_ctl_sock_init(net);
1455     if (status)
1456         pr_err("Failed to initialize the SCTP control sock\n");
1457 
1458     return status;
1459 }
1460 
1461 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1462 {
1463     /* Free the control endpoint.  */
1464     inet_ctl_sock_destroy(net->sctp.ctl_sock);
1465 }
1466 
1467 static struct pernet_operations sctp_ctrlsock_ops = {
1468     .init = sctp_ctrlsock_init,
1469     .exit = sctp_ctrlsock_exit,
1470 };
1471 
1472 /* Initialize the universe into something sensible.  */
1473 static __init int sctp_init(void)
1474 {
1475     unsigned long nr_pages = totalram_pages();
1476     unsigned long limit;
1477     unsigned long goal;
1478     int max_entry_order;
1479     int num_entries;
1480     int max_share;
1481     int status;
1482     int order;
1483     int i;
1484 
1485     sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1486 
1487     /* Allocate bind_bucket and chunk caches. */
1488     status = -ENOBUFS;
1489     sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1490                            sizeof(struct sctp_bind_bucket),
1491                            0, SLAB_HWCACHE_ALIGN,
1492                            NULL);
1493     if (!sctp_bucket_cachep)
1494         goto out;
1495 
1496     sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1497                            sizeof(struct sctp_chunk),
1498                            0, SLAB_HWCACHE_ALIGN,
1499                            NULL);
1500     if (!sctp_chunk_cachep)
1501         goto err_chunk_cachep;
1502 
1503     status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1504     if (status)
1505         goto err_percpu_counter_init;
1506 
1507     /* Implementation specific variables. */
1508 
1509     /* Initialize default stream count setup information. */
1510     sctp_max_instreams          = SCTP_DEFAULT_INSTREAMS;
1511     sctp_max_outstreams         = SCTP_DEFAULT_OUTSTREAMS;
1512 
1513     /* Initialize handle used for association ids. */
1514     idr_init(&sctp_assocs_id);
1515 
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;
1521 
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);
1525 
1526     sysctl_sctp_rmem[0] = PAGE_SIZE; /* 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);
1529 
1530     sysctl_sctp_wmem[0] = PAGE_SIZE;
1531     sysctl_sctp_wmem[1] = 16*1024;
1532     sysctl_sctp_wmem[2] = max(64*1024, max_share);
1533 
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
1537      */
1538     if (nr_pages >= (128 * 1024))
1539         goal = nr_pages >> (22 - PAGE_SHIFT);
1540     else
1541         goal = nr_pages >> (24 - PAGE_SHIFT);
1542 
1543     /* Then compute the page order for said goal */
1544     order = get_order(goal);
1545 
1546     /* Now compute the required page order for the maximum sized table we
1547      * want to create
1548      */
1549     max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1550                     sizeof(struct sctp_bind_hashbucket));
1551 
1552     /* Limit the page order by that maximum hash table size */
1553     order = min(order, max_entry_order);
1554 
1555     /* Allocate and initialize the endpoint hash table.  */
1556     sctp_ep_hashsize = 64;
1557     sctp_ep_hashtable =
1558         kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1559     if (!sctp_ep_hashtable) {
1560         pr_err("Failed endpoint_hash alloc\n");
1561         status = -ENOMEM;
1562         goto err_ehash_alloc;
1563     }
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);
1567     }
1568 
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
1573      */
1574     do {
1575         sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1576             __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1577     } while (!sctp_port_hashtable && --order > 0);
1578 
1579     if (!sctp_port_hashtable) {
1580         pr_err("Failed bind hash alloc\n");
1581         status = -ENOMEM;
1582         goto err_bhash_alloc;
1583     }
1584 
1585     /* Now compute the number of entries that will fit in the
1586      * port hash space we allocated
1587      */
1588     num_entries = (1UL << order) * PAGE_SIZE /
1589               sizeof(struct sctp_bind_hashbucket);
1590 
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.
1595      */
1596     sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1597 
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);
1601     }
1602 
1603     status = sctp_transport_hashtable_init();
1604     if (status)
1605         goto err_thash_alloc;
1606 
1607     pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1608         num_entries);
1609 
1610     sctp_sysctl_register();
1611 
1612     INIT_LIST_HEAD(&sctp_address_families);
1613     sctp_v4_pf_init();
1614     sctp_v6_pf_init();
1615     sctp_sched_ops_init();
1616 
1617     status = register_pernet_subsys(&sctp_defaults_ops);
1618     if (status)
1619         goto err_register_defaults;
1620 
1621     status = sctp_v4_protosw_init();
1622     if (status)
1623         goto err_protosw_init;
1624 
1625     status = sctp_v6_protosw_init();
1626     if (status)
1627         goto err_v6_protosw_init;
1628 
1629     status = register_pernet_subsys(&sctp_ctrlsock_ops);
1630     if (status)
1631         goto err_register_ctrlsock;
1632 
1633     status = sctp_v4_add_protocol();
1634     if (status)
1635         goto err_add_protocol;
1636 
1637     /* Register SCTP with inet6 layer.  */
1638     status = sctp_v6_add_protocol();
1639     if (status)
1640         goto err_v6_add_protocol;
1641 
1642     if (sctp_offload_init() < 0)
1643         pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1644 
1645 out:
1646     return status;
1647 err_v6_add_protocol:
1648     sctp_v4_del_protocol();
1649 err_add_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();
1655 err_protosw_init:
1656     unregister_pernet_subsys(&sctp_defaults_ops);
1657 err_register_defaults:
1658     sctp_v4_pf_exit();
1659     sctp_v6_pf_exit();
1660     sctp_sysctl_unregister();
1661     free_pages((unsigned long)sctp_port_hashtable,
1662            get_order(sctp_port_hashsize *
1663                  sizeof(struct sctp_bind_hashbucket)));
1664 err_bhash_alloc:
1665     sctp_transport_hashtable_destroy();
1666 err_thash_alloc:
1667     kfree(sctp_ep_hashtable);
1668 err_ehash_alloc:
1669     percpu_counter_destroy(&sctp_sockets_allocated);
1670 err_percpu_counter_init:
1671     kmem_cache_destroy(sctp_chunk_cachep);
1672 err_chunk_cachep:
1673     kmem_cache_destroy(sctp_bucket_cachep);
1674     goto out;
1675 }
1676 
1677 /* Exit handler for the SCTP protocol.  */
1678 static __exit void sctp_exit(void)
1679 {
1680     /* BUG.  This should probably do something useful like clean
1681      * up all the remaining associations and all that memory.
1682      */
1683 
1684     /* Unregister with inet6/inet layers. */
1685     sctp_v6_del_protocol();
1686     sctp_v4_del_protocol();
1687 
1688     unregister_pernet_subsys(&sctp_ctrlsock_ops);
1689 
1690     /* Free protosw registrations */
1691     sctp_v6_protosw_exit();
1692     sctp_v4_protosw_exit();
1693 
1694     unregister_pernet_subsys(&sctp_defaults_ops);
1695 
1696     /* Unregister with socket layer. */
1697     sctp_v6_pf_exit();
1698     sctp_v4_pf_exit();
1699 
1700     sctp_sysctl_unregister();
1701 
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();
1707 
1708     percpu_counter_destroy(&sctp_sockets_allocated);
1709 
1710     rcu_barrier(); /* Wait for completion of call_rcu()'s */
1711 
1712     kmem_cache_destroy(sctp_chunk_cachep);
1713     kmem_cache_destroy(sctp_bucket_cachep);
1714 }
1715 
1716 module_init(sctp_init);
1717 module_exit(sctp_exit);
1718 
1719 /*
1720  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1721  */
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");
[ Source navigation ] [ Diff markup ] [ Identifier search ] [ general search ]

This page was automatically generated by the 2.1.0 LXR engine.

The LXR team
Valid CSS 2.1! Valid HTML 4.01!