Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* SCTP kernel reference 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) 2001-2003 Intel Corp. |
| 6 | * Copyright (c) 2001-2002 Nokia, Inc. |
| 7 | * Copyright (c) 2001 La Monte H.P. Yarroll |
| 8 | * |
| 9 | * This file is part of the SCTP kernel reference Implementation |
| 10 | * |
| 11 | * These functions interface with the sockets layer to implement the |
| 12 | * SCTP Extensions for the Sockets API. |
| 13 | * |
| 14 | * Note that the descriptions from the specification are USER level |
| 15 | * functions--this file is the functions which populate the struct proto |
| 16 | * for SCTP which is the BOTTOM of the sockets interface. |
| 17 | * |
| 18 | * The SCTP reference implementation is free software; |
| 19 | * you can redistribute it and/or modify it under the terms of |
| 20 | * the GNU General Public License as published by |
| 21 | * the Free Software Foundation; either version 2, or (at your option) |
| 22 | * any later version. |
| 23 | * |
| 24 | * The SCTP reference implementation is distributed in the hope that it |
| 25 | * will be useful, but WITHOUT ANY WARRANTY; without even the implied |
| 26 | * ************************ |
| 27 | * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. |
| 28 | * See the GNU General Public License for more details. |
| 29 | * |
| 30 | * You should have received a copy of the GNU General Public License |
| 31 | * along with GNU CC; see the file COPYING. If not, write to |
| 32 | * the Free Software Foundation, 59 Temple Place - Suite 330, |
| 33 | * Boston, MA 02111-1307, USA. |
| 34 | * |
| 35 | * Please send any bug reports or fixes you make to the |
| 36 | * email address(es): |
| 37 | * lksctp developers <lksctp-developers@lists.sourceforge.net> |
| 38 | * |
| 39 | * Or submit a bug report through the following website: |
| 40 | * http://www.sf.net/projects/lksctp |
| 41 | * |
| 42 | * Written or modified by: |
| 43 | * La Monte H.P. Yarroll <piggy@acm.org> |
| 44 | * Narasimha Budihal <narsi@refcode.org> |
| 45 | * Karl Knutson <karl@athena.chicago.il.us> |
| 46 | * Jon Grimm <jgrimm@us.ibm.com> |
| 47 | * Xingang Guo <xingang.guo@intel.com> |
| 48 | * Daisy Chang <daisyc@us.ibm.com> |
| 49 | * Sridhar Samudrala <samudrala@us.ibm.com> |
| 50 | * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com> |
| 51 | * Ardelle Fan <ardelle.fan@intel.com> |
| 52 | * Ryan Layer <rmlayer@us.ibm.com> |
| 53 | * Anup Pemmaiah <pemmaiah@cc.usu.edu> |
| 54 | * Kevin Gao <kevin.gao@intel.com> |
| 55 | * |
| 56 | * Any bugs reported given to us we will try to fix... any fixes shared will |
| 57 | * be incorporated into the next SCTP release. |
| 58 | */ |
| 59 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 60 | #include <linux/types.h> |
| 61 | #include <linux/kernel.h> |
| 62 | #include <linux/wait.h> |
| 63 | #include <linux/time.h> |
| 64 | #include <linux/ip.h> |
Randy Dunlap | 4fc268d | 2006-01-11 12:17:47 -0800 | [diff] [blame] | 65 | #include <linux/capability.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 66 | #include <linux/fcntl.h> |
| 67 | #include <linux/poll.h> |
| 68 | #include <linux/init.h> |
| 69 | #include <linux/crypto.h> |
| 70 | |
| 71 | #include <net/ip.h> |
| 72 | #include <net/icmp.h> |
| 73 | #include <net/route.h> |
| 74 | #include <net/ipv6.h> |
| 75 | #include <net/inet_common.h> |
| 76 | |
| 77 | #include <linux/socket.h> /* for sa_family_t */ |
| 78 | #include <net/sock.h> |
| 79 | #include <net/sctp/sctp.h> |
| 80 | #include <net/sctp/sm.h> |
| 81 | |
| 82 | /* WARNING: Please do not remove the SCTP_STATIC attribute to |
| 83 | * any of the functions below as they are used to export functions |
| 84 | * used by a project regression testsuite. |
| 85 | */ |
| 86 | |
| 87 | /* Forward declarations for internal helper functions. */ |
| 88 | static int sctp_writeable(struct sock *sk); |
| 89 | static void sctp_wfree(struct sk_buff *skb); |
| 90 | static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p, |
| 91 | size_t msg_len); |
| 92 | static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p); |
| 93 | static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p); |
| 94 | static int sctp_wait_for_accept(struct sock *sk, long timeo); |
| 95 | static void sctp_wait_for_close(struct sock *sk, long timeo); |
| 96 | static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt, |
| 97 | union sctp_addr *addr, int len); |
| 98 | static int sctp_bindx_add(struct sock *, struct sockaddr *, int); |
| 99 | static int sctp_bindx_rem(struct sock *, struct sockaddr *, int); |
| 100 | static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int); |
| 101 | static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int); |
| 102 | static int sctp_send_asconf(struct sctp_association *asoc, |
| 103 | struct sctp_chunk *chunk); |
| 104 | static int sctp_do_bind(struct sock *, union sctp_addr *, int); |
| 105 | static int sctp_autobind(struct sock *sk); |
| 106 | static void sctp_sock_migrate(struct sock *, struct sock *, |
| 107 | struct sctp_association *, sctp_socket_type_t); |
| 108 | static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG; |
| 109 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 110 | /* Get the sndbuf space available at the time on the association. */ |
| 111 | static inline int sctp_wspace(struct sctp_association *asoc) |
| 112 | { |
| 113 | struct sock *sk = asoc->base.sk; |
| 114 | int amt = 0; |
| 115 | |
Neil Horman | 4eb701d | 2005-04-28 12:02:04 -0700 | [diff] [blame] | 116 | if (asoc->ep->sndbuf_policy) { |
| 117 | /* make sure that no association uses more than sk_sndbuf */ |
| 118 | amt = sk->sk_sndbuf - asoc->sndbuf_used; |
| 119 | } else { |
| 120 | /* do socket level accounting */ |
| 121 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); |
| 122 | } |
| 123 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 124 | if (amt < 0) |
| 125 | amt = 0; |
Neil Horman | 4eb701d | 2005-04-28 12:02:04 -0700 | [diff] [blame] | 126 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 127 | return amt; |
| 128 | } |
| 129 | |
| 130 | /* Increment the used sndbuf space count of the corresponding association by |
| 131 | * the size of the outgoing data chunk. |
| 132 | * Also, set the skb destructor for sndbuf accounting later. |
| 133 | * |
| 134 | * Since it is always 1-1 between chunk and skb, and also a new skb is always |
| 135 | * allocated for chunk bundling in sctp_packet_transmit(), we can use the |
| 136 | * destructor in the data chunk skb for the purpose of the sndbuf space |
| 137 | * tracking. |
| 138 | */ |
| 139 | static inline void sctp_set_owner_w(struct sctp_chunk *chunk) |
| 140 | { |
| 141 | struct sctp_association *asoc = chunk->asoc; |
| 142 | struct sock *sk = asoc->base.sk; |
| 143 | |
| 144 | /* The sndbuf space is tracked per association. */ |
| 145 | sctp_association_hold(asoc); |
| 146 | |
Neil Horman | 4eb701d | 2005-04-28 12:02:04 -0700 | [diff] [blame] | 147 | skb_set_owner_w(chunk->skb, sk); |
| 148 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 149 | chunk->skb->destructor = sctp_wfree; |
| 150 | /* Save the chunk pointer in skb for sctp_wfree to use later. */ |
| 151 | *((struct sctp_chunk **)(chunk->skb->cb)) = chunk; |
| 152 | |
Neil Horman | 4eb701d | 2005-04-28 12:02:04 -0700 | [diff] [blame] | 153 | asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) + |
| 154 | sizeof(struct sk_buff) + |
| 155 | sizeof(struct sctp_chunk); |
| 156 | |
Neil Horman | 4eb701d | 2005-04-28 12:02:04 -0700 | [diff] [blame] | 157 | atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 158 | } |
| 159 | |
| 160 | /* Verify that this is a valid address. */ |
| 161 | static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr, |
| 162 | int len) |
| 163 | { |
| 164 | struct sctp_af *af; |
| 165 | |
| 166 | /* Verify basic sockaddr. */ |
| 167 | af = sctp_sockaddr_af(sctp_sk(sk), addr, len); |
| 168 | if (!af) |
| 169 | return -EINVAL; |
| 170 | |
| 171 | /* Is this a valid SCTP address? */ |
Vlad Yasevich | 5636bef | 2006-06-17 22:55:35 -0700 | [diff] [blame] | 172 | if (!af->addr_valid(addr, sctp_sk(sk), NULL)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 173 | return -EINVAL; |
| 174 | |
| 175 | if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr))) |
| 176 | return -EINVAL; |
| 177 | |
| 178 | return 0; |
| 179 | } |
| 180 | |
| 181 | /* Look up the association by its id. If this is not a UDP-style |
| 182 | * socket, the ID field is always ignored. |
| 183 | */ |
| 184 | struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id) |
| 185 | { |
| 186 | struct sctp_association *asoc = NULL; |
| 187 | |
| 188 | /* If this is not a UDP-style socket, assoc id should be ignored. */ |
| 189 | if (!sctp_style(sk, UDP)) { |
| 190 | /* Return NULL if the socket state is not ESTABLISHED. It |
| 191 | * could be a TCP-style listening socket or a socket which |
| 192 | * hasn't yet called connect() to establish an association. |
| 193 | */ |
| 194 | if (!sctp_sstate(sk, ESTABLISHED)) |
| 195 | return NULL; |
| 196 | |
| 197 | /* Get the first and the only association from the list. */ |
| 198 | if (!list_empty(&sctp_sk(sk)->ep->asocs)) |
| 199 | asoc = list_entry(sctp_sk(sk)->ep->asocs.next, |
| 200 | struct sctp_association, asocs); |
| 201 | return asoc; |
| 202 | } |
| 203 | |
| 204 | /* Otherwise this is a UDP-style socket. */ |
| 205 | if (!id || (id == (sctp_assoc_t)-1)) |
| 206 | return NULL; |
| 207 | |
| 208 | spin_lock_bh(&sctp_assocs_id_lock); |
| 209 | asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id); |
| 210 | spin_unlock_bh(&sctp_assocs_id_lock); |
| 211 | |
| 212 | if (!asoc || (asoc->base.sk != sk) || asoc->base.dead) |
| 213 | return NULL; |
| 214 | |
| 215 | return asoc; |
| 216 | } |
| 217 | |
| 218 | /* Look up the transport from an address and an assoc id. If both address and |
| 219 | * id are specified, the associations matching the address and the id should be |
| 220 | * the same. |
| 221 | */ |
| 222 | static struct sctp_transport *sctp_addr_id2transport(struct sock *sk, |
| 223 | struct sockaddr_storage *addr, |
| 224 | sctp_assoc_t id) |
| 225 | { |
| 226 | struct sctp_association *addr_asoc = NULL, *id_asoc = NULL; |
| 227 | struct sctp_transport *transport; |
| 228 | union sctp_addr *laddr = (union sctp_addr *)addr; |
| 229 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 230 | addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep, |
Al Viro | cd4ff03 | 2006-11-20 17:11:33 -0800 | [diff] [blame] | 231 | laddr, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 232 | &transport); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 233 | |
| 234 | if (!addr_asoc) |
| 235 | return NULL; |
| 236 | |
| 237 | id_asoc = sctp_id2assoc(sk, id); |
| 238 | if (id_asoc && (id_asoc != addr_asoc)) |
| 239 | return NULL; |
| 240 | |
| 241 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk), |
| 242 | (union sctp_addr *)addr); |
| 243 | |
| 244 | return transport; |
| 245 | } |
| 246 | |
| 247 | /* API 3.1.2 bind() - UDP Style Syntax |
| 248 | * The syntax of bind() is, |
| 249 | * |
| 250 | * ret = bind(int sd, struct sockaddr *addr, int addrlen); |
| 251 | * |
| 252 | * sd - the socket descriptor returned by socket(). |
| 253 | * addr - the address structure (struct sockaddr_in or struct |
| 254 | * sockaddr_in6 [RFC 2553]), |
| 255 | * addr_len - the size of the address structure. |
| 256 | */ |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 257 | SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 258 | { |
| 259 | int retval = 0; |
| 260 | |
| 261 | sctp_lock_sock(sk); |
| 262 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 263 | SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n", |
| 264 | sk, addr, addr_len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 265 | |
| 266 | /* Disallow binding twice. */ |
| 267 | if (!sctp_sk(sk)->ep->base.bind_addr.port) |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 268 | retval = sctp_do_bind(sk, (union sctp_addr *)addr, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 269 | addr_len); |
| 270 | else |
| 271 | retval = -EINVAL; |
| 272 | |
| 273 | sctp_release_sock(sk); |
| 274 | |
| 275 | return retval; |
| 276 | } |
| 277 | |
| 278 | static long sctp_get_port_local(struct sock *, union sctp_addr *); |
| 279 | |
| 280 | /* Verify this is a valid sockaddr. */ |
| 281 | static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt, |
| 282 | union sctp_addr *addr, int len) |
| 283 | { |
| 284 | struct sctp_af *af; |
| 285 | |
| 286 | /* Check minimum size. */ |
| 287 | if (len < sizeof (struct sockaddr)) |
| 288 | return NULL; |
| 289 | |
| 290 | /* Does this PF support this AF? */ |
| 291 | if (!opt->pf->af_supported(addr->sa.sa_family, opt)) |
| 292 | return NULL; |
| 293 | |
| 294 | /* If we get this far, af is valid. */ |
| 295 | af = sctp_get_af_specific(addr->sa.sa_family); |
| 296 | |
| 297 | if (len < af->sockaddr_len) |
| 298 | return NULL; |
| 299 | |
| 300 | return af; |
| 301 | } |
| 302 | |
| 303 | /* Bind a local address either to an endpoint or to an association. */ |
| 304 | SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len) |
| 305 | { |
| 306 | struct sctp_sock *sp = sctp_sk(sk); |
| 307 | struct sctp_endpoint *ep = sp->ep; |
| 308 | struct sctp_bind_addr *bp = &ep->base.bind_addr; |
| 309 | struct sctp_af *af; |
| 310 | unsigned short snum; |
| 311 | int ret = 0; |
| 312 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 313 | /* Common sockaddr verification. */ |
| 314 | af = sctp_sockaddr_af(sp, addr, len); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 315 | if (!af) { |
| 316 | SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n", |
| 317 | sk, addr, len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 318 | return -EINVAL; |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 319 | } |
| 320 | |
| 321 | snum = ntohs(addr->v4.sin_port); |
| 322 | |
| 323 | SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ", |
| 324 | ", port: %d, new port: %d, len: %d)\n", |
| 325 | sk, |
| 326 | addr, |
| 327 | bp->port, snum, |
| 328 | len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 329 | |
| 330 | /* PF specific bind() address verification. */ |
| 331 | if (!sp->pf->bind_verify(sp, addr)) |
| 332 | return -EADDRNOTAVAIL; |
| 333 | |
Vlad Yasevich | 8b35805 | 2007-05-15 17:14:58 -0400 | [diff] [blame] | 334 | /* We must either be unbound, or bind to the same port. |
| 335 | * It's OK to allow 0 ports if we are already bound. |
| 336 | * We'll just inhert an already bound port in this case |
| 337 | */ |
| 338 | if (bp->port) { |
| 339 | if (!snum) |
| 340 | snum = bp->port; |
| 341 | else if (snum != bp->port) { |
| 342 | SCTP_DEBUG_PRINTK("sctp_do_bind:" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 343 | " New port %d does not match existing port " |
| 344 | "%d.\n", snum, bp->port); |
Vlad Yasevich | 8b35805 | 2007-05-15 17:14:58 -0400 | [diff] [blame] | 345 | return -EINVAL; |
| 346 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 347 | } |
| 348 | |
| 349 | if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE)) |
| 350 | return -EACCES; |
| 351 | |
| 352 | /* Make sure we are allowed to bind here. |
| 353 | * The function sctp_get_port_local() does duplicate address |
| 354 | * detection. |
| 355 | */ |
Vlad Yasevich | 2772b49 | 2007-08-21 14:24:30 +0900 | [diff] [blame] | 356 | addr->v4.sin_port = htons(snum); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 357 | if ((ret = sctp_get_port_local(sk, addr))) { |
| 358 | if (ret == (long) sk) { |
| 359 | /* This endpoint has a conflicting address. */ |
| 360 | return -EINVAL; |
| 361 | } else { |
| 362 | return -EADDRINUSE; |
| 363 | } |
| 364 | } |
| 365 | |
| 366 | /* Refresh ephemeral port. */ |
| 367 | if (!bp->port) |
| 368 | bp->port = inet_sk(sk)->num; |
| 369 | |
| 370 | /* Add the address to the bind address list. */ |
| 371 | sctp_local_bh_disable(); |
| 372 | sctp_write_lock(&ep->base.addr_lock); |
| 373 | |
| 374 | /* Use GFP_ATOMIC since BHs are disabled. */ |
Al Viro | 5ab7b85 | 2006-11-20 17:10:38 -0800 | [diff] [blame] | 375 | ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 376 | sctp_write_unlock(&ep->base.addr_lock); |
| 377 | sctp_local_bh_enable(); |
| 378 | |
| 379 | /* Copy back into socket for getsockname() use. */ |
| 380 | if (!ret) { |
| 381 | inet_sk(sk)->sport = htons(inet_sk(sk)->num); |
| 382 | af->to_sk_saddr(addr, sk); |
| 383 | } |
| 384 | |
| 385 | return ret; |
| 386 | } |
| 387 | |
| 388 | /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks |
| 389 | * |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 390 | * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 391 | * at any one time. If a sender, after sending an ASCONF chunk, decides |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 392 | * it needs to transfer another ASCONF Chunk, it MUST wait until the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 393 | * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 394 | * subsequent ASCONF. Note this restriction binds each side, so at any |
| 395 | * time two ASCONF may be in-transit on any given association (one sent |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 396 | * from each endpoint). |
| 397 | */ |
| 398 | static int sctp_send_asconf(struct sctp_association *asoc, |
| 399 | struct sctp_chunk *chunk) |
| 400 | { |
| 401 | int retval = 0; |
| 402 | |
| 403 | /* If there is an outstanding ASCONF chunk, queue it for later |
| 404 | * transmission. |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 405 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 406 | if (asoc->addip_last_asconf) { |
David S. Miller | 79af02c | 2005-07-08 21:47:49 -0700 | [diff] [blame] | 407 | list_add_tail(&chunk->list, &asoc->addip_chunk_list); |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 408 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 409 | } |
| 410 | |
| 411 | /* Hold the chunk until an ASCONF_ACK is received. */ |
| 412 | sctp_chunk_hold(chunk); |
| 413 | retval = sctp_primitive_ASCONF(asoc, chunk); |
| 414 | if (retval) |
| 415 | sctp_chunk_free(chunk); |
| 416 | else |
| 417 | asoc->addip_last_asconf = chunk; |
| 418 | |
| 419 | out: |
| 420 | return retval; |
| 421 | } |
| 422 | |
| 423 | /* Add a list of addresses as bind addresses to local endpoint or |
| 424 | * association. |
| 425 | * |
| 426 | * Basically run through each address specified in the addrs/addrcnt |
| 427 | * array/length pair, determine if it is IPv6 or IPv4 and call |
| 428 | * sctp_do_bind() on it. |
| 429 | * |
| 430 | * If any of them fails, then the operation will be reversed and the |
| 431 | * ones that were added will be removed. |
| 432 | * |
| 433 | * Only sctp_setsockopt_bindx() is supposed to call this function. |
| 434 | */ |
sebastian@breakpoint.cc | 0467521 | 2007-07-26 23:21:31 +0200 | [diff] [blame] | 435 | static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 436 | { |
| 437 | int cnt; |
| 438 | int retval = 0; |
| 439 | void *addr_buf; |
| 440 | struct sockaddr *sa_addr; |
| 441 | struct sctp_af *af; |
| 442 | |
| 443 | SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n", |
| 444 | sk, addrs, addrcnt); |
| 445 | |
| 446 | addr_buf = addrs; |
| 447 | for (cnt = 0; cnt < addrcnt; cnt++) { |
| 448 | /* The list may contain either IPv4 or IPv6 address; |
| 449 | * determine the address length for walking thru the list. |
| 450 | */ |
| 451 | sa_addr = (struct sockaddr *)addr_buf; |
| 452 | af = sctp_get_af_specific(sa_addr->sa_family); |
| 453 | if (!af) { |
| 454 | retval = -EINVAL; |
| 455 | goto err_bindx_add; |
| 456 | } |
| 457 | |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 458 | retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 459 | af->sockaddr_len); |
| 460 | |
| 461 | addr_buf += af->sockaddr_len; |
| 462 | |
| 463 | err_bindx_add: |
| 464 | if (retval < 0) { |
| 465 | /* Failed. Cleanup the ones that have been added */ |
| 466 | if (cnt > 0) |
| 467 | sctp_bindx_rem(sk, addrs, cnt); |
| 468 | return retval; |
| 469 | } |
| 470 | } |
| 471 | |
| 472 | return retval; |
| 473 | } |
| 474 | |
| 475 | /* Send an ASCONF chunk with Add IP address parameters to all the peers of the |
| 476 | * associations that are part of the endpoint indicating that a list of local |
| 477 | * addresses are added to the endpoint. |
| 478 | * |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 479 | * If any of the addresses is already in the bind address list of the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 480 | * association, we do not send the chunk for that association. But it will not |
| 481 | * affect other associations. |
| 482 | * |
| 483 | * Only sctp_setsockopt_bindx() is supposed to call this function. |
| 484 | */ |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 485 | static int sctp_send_asconf_add_ip(struct sock *sk, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 486 | struct sockaddr *addrs, |
| 487 | int addrcnt) |
| 488 | { |
| 489 | struct sctp_sock *sp; |
| 490 | struct sctp_endpoint *ep; |
| 491 | struct sctp_association *asoc; |
| 492 | struct sctp_bind_addr *bp; |
| 493 | struct sctp_chunk *chunk; |
| 494 | struct sctp_sockaddr_entry *laddr; |
| 495 | union sctp_addr *addr; |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 496 | union sctp_addr saveaddr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 497 | void *addr_buf; |
| 498 | struct sctp_af *af; |
| 499 | struct list_head *pos; |
| 500 | struct list_head *p; |
| 501 | int i; |
| 502 | int retval = 0; |
| 503 | |
| 504 | if (!sctp_addip_enable) |
| 505 | return retval; |
| 506 | |
| 507 | sp = sctp_sk(sk); |
| 508 | ep = sp->ep; |
| 509 | |
| 510 | SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n", |
| 511 | __FUNCTION__, sk, addrs, addrcnt); |
| 512 | |
| 513 | list_for_each(pos, &ep->asocs) { |
| 514 | asoc = list_entry(pos, struct sctp_association, asocs); |
| 515 | |
| 516 | if (!asoc->peer.asconf_capable) |
| 517 | continue; |
| 518 | |
| 519 | if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP) |
| 520 | continue; |
| 521 | |
| 522 | if (!sctp_state(asoc, ESTABLISHED)) |
| 523 | continue; |
| 524 | |
| 525 | /* Check if any address in the packed array of addresses is |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 526 | * in the bind address list of the association. If so, |
| 527 | * do not send the asconf chunk to its peer, but continue with |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 528 | * other associations. |
| 529 | */ |
| 530 | addr_buf = addrs; |
| 531 | for (i = 0; i < addrcnt; i++) { |
| 532 | addr = (union sctp_addr *)addr_buf; |
| 533 | af = sctp_get_af_specific(addr->v4.sin_family); |
| 534 | if (!af) { |
| 535 | retval = -EINVAL; |
| 536 | goto out; |
| 537 | } |
| 538 | |
| 539 | if (sctp_assoc_lookup_laddr(asoc, addr)) |
| 540 | break; |
| 541 | |
| 542 | addr_buf += af->sockaddr_len; |
| 543 | } |
| 544 | if (i < addrcnt) |
| 545 | continue; |
| 546 | |
| 547 | /* Use the first address in bind addr list of association as |
| 548 | * Address Parameter of ASCONF CHUNK. |
| 549 | */ |
| 550 | sctp_read_lock(&asoc->base.addr_lock); |
| 551 | bp = &asoc->base.bind_addr; |
| 552 | p = bp->address_list.next; |
| 553 | laddr = list_entry(p, struct sctp_sockaddr_entry, list); |
| 554 | sctp_read_unlock(&asoc->base.addr_lock); |
| 555 | |
Al Viro | 5ae955c | 2006-11-20 17:22:08 -0800 | [diff] [blame] | 556 | chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 557 | addrcnt, SCTP_PARAM_ADD_IP); |
| 558 | if (!chunk) { |
| 559 | retval = -ENOMEM; |
| 560 | goto out; |
| 561 | } |
| 562 | |
| 563 | retval = sctp_send_asconf(asoc, chunk); |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 564 | if (retval) |
| 565 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 566 | |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 567 | /* Add the new addresses to the bind address list with |
| 568 | * use_as_src set to 0. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 569 | */ |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 570 | sctp_local_bh_disable(); |
| 571 | sctp_write_lock(&asoc->base.addr_lock); |
| 572 | addr_buf = addrs; |
| 573 | for (i = 0; i < addrcnt; i++) { |
| 574 | addr = (union sctp_addr *)addr_buf; |
| 575 | af = sctp_get_af_specific(addr->v4.sin_family); |
| 576 | memcpy(&saveaddr, addr, af->sockaddr_len); |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 577 | retval = sctp_add_bind_addr(bp, &saveaddr, 0, |
| 578 | GFP_ATOMIC); |
| 579 | addr_buf += af->sockaddr_len; |
| 580 | } |
| 581 | sctp_write_unlock(&asoc->base.addr_lock); |
| 582 | sctp_local_bh_enable(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 583 | } |
| 584 | |
| 585 | out: |
| 586 | return retval; |
| 587 | } |
| 588 | |
| 589 | /* Remove a list of addresses from bind addresses list. Do not remove the |
| 590 | * last address. |
| 591 | * |
| 592 | * Basically run through each address specified in the addrs/addrcnt |
| 593 | * array/length pair, determine if it is IPv6 or IPv4 and call |
| 594 | * sctp_del_bind() on it. |
| 595 | * |
| 596 | * If any of them fails, then the operation will be reversed and the |
| 597 | * ones that were removed will be added back. |
| 598 | * |
| 599 | * At least one address has to be left; if only one address is |
| 600 | * available, the operation will return -EBUSY. |
| 601 | * |
| 602 | * Only sctp_setsockopt_bindx() is supposed to call this function. |
| 603 | */ |
sebastian@breakpoint.cc | 0467521 | 2007-07-26 23:21:31 +0200 | [diff] [blame] | 604 | static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 605 | { |
| 606 | struct sctp_sock *sp = sctp_sk(sk); |
| 607 | struct sctp_endpoint *ep = sp->ep; |
| 608 | int cnt; |
| 609 | struct sctp_bind_addr *bp = &ep->base.bind_addr; |
| 610 | int retval = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 611 | void *addr_buf; |
Al Viro | c9a0850 | 2006-11-20 17:07:48 -0800 | [diff] [blame] | 612 | union sctp_addr *sa_addr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 613 | struct sctp_af *af; |
| 614 | |
| 615 | SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n", |
| 616 | sk, addrs, addrcnt); |
| 617 | |
| 618 | addr_buf = addrs; |
| 619 | for (cnt = 0; cnt < addrcnt; cnt++) { |
| 620 | /* If the bind address list is empty or if there is only one |
| 621 | * bind address, there is nothing more to be removed (we need |
| 622 | * at least one address here). |
| 623 | */ |
| 624 | if (list_empty(&bp->address_list) || |
| 625 | (sctp_list_single_entry(&bp->address_list))) { |
| 626 | retval = -EBUSY; |
| 627 | goto err_bindx_rem; |
| 628 | } |
| 629 | |
Al Viro | c9a0850 | 2006-11-20 17:07:48 -0800 | [diff] [blame] | 630 | sa_addr = (union sctp_addr *)addr_buf; |
| 631 | af = sctp_get_af_specific(sa_addr->sa.sa_family); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 632 | if (!af) { |
| 633 | retval = -EINVAL; |
| 634 | goto err_bindx_rem; |
| 635 | } |
Paolo Galtieri | 0304ff8a | 2007-04-17 12:52:36 -0700 | [diff] [blame] | 636 | |
| 637 | if (!af->addr_valid(sa_addr, sp, NULL)) { |
| 638 | retval = -EADDRNOTAVAIL; |
| 639 | goto err_bindx_rem; |
| 640 | } |
| 641 | |
Al Viro | c9a0850 | 2006-11-20 17:07:48 -0800 | [diff] [blame] | 642 | if (sa_addr->v4.sin_port != htons(bp->port)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 643 | retval = -EINVAL; |
| 644 | goto err_bindx_rem; |
| 645 | } |
| 646 | |
| 647 | /* FIXME - There is probably a need to check if sk->sk_saddr and |
| 648 | * sk->sk_rcv_addr are currently set to one of the addresses to |
| 649 | * be removed. This is something which needs to be looked into |
| 650 | * when we are fixing the outstanding issues with multi-homing |
| 651 | * socket routing and failover schemes. Refer to comments in |
| 652 | * sctp_do_bind(). -daisy |
| 653 | */ |
| 654 | sctp_local_bh_disable(); |
| 655 | sctp_write_lock(&ep->base.addr_lock); |
| 656 | |
Al Viro | c9a0850 | 2006-11-20 17:07:48 -0800 | [diff] [blame] | 657 | retval = sctp_del_bind_addr(bp, sa_addr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 658 | |
| 659 | sctp_write_unlock(&ep->base.addr_lock); |
| 660 | sctp_local_bh_enable(); |
| 661 | |
| 662 | addr_buf += af->sockaddr_len; |
| 663 | err_bindx_rem: |
| 664 | if (retval < 0) { |
| 665 | /* Failed. Add the ones that has been removed back */ |
| 666 | if (cnt > 0) |
| 667 | sctp_bindx_add(sk, addrs, cnt); |
| 668 | return retval; |
| 669 | } |
| 670 | } |
| 671 | |
| 672 | return retval; |
| 673 | } |
| 674 | |
| 675 | /* Send an ASCONF chunk with Delete IP address parameters to all the peers of |
| 676 | * the associations that are part of the endpoint indicating that a list of |
| 677 | * local addresses are removed from the endpoint. |
| 678 | * |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 679 | * If any of the addresses is already in the bind address list of the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 680 | * association, we do not send the chunk for that association. But it will not |
| 681 | * affect other associations. |
| 682 | * |
| 683 | * Only sctp_setsockopt_bindx() is supposed to call this function. |
| 684 | */ |
| 685 | static int sctp_send_asconf_del_ip(struct sock *sk, |
| 686 | struct sockaddr *addrs, |
| 687 | int addrcnt) |
| 688 | { |
| 689 | struct sctp_sock *sp; |
| 690 | struct sctp_endpoint *ep; |
| 691 | struct sctp_association *asoc; |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 692 | struct sctp_transport *transport; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 693 | struct sctp_bind_addr *bp; |
| 694 | struct sctp_chunk *chunk; |
| 695 | union sctp_addr *laddr; |
| 696 | void *addr_buf; |
| 697 | struct sctp_af *af; |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 698 | struct list_head *pos, *pos1; |
| 699 | struct sctp_sockaddr_entry *saddr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 700 | int i; |
| 701 | int retval = 0; |
| 702 | |
| 703 | if (!sctp_addip_enable) |
| 704 | return retval; |
| 705 | |
| 706 | sp = sctp_sk(sk); |
| 707 | ep = sp->ep; |
| 708 | |
| 709 | SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n", |
| 710 | __FUNCTION__, sk, addrs, addrcnt); |
| 711 | |
| 712 | list_for_each(pos, &ep->asocs) { |
| 713 | asoc = list_entry(pos, struct sctp_association, asocs); |
| 714 | |
| 715 | if (!asoc->peer.asconf_capable) |
| 716 | continue; |
| 717 | |
| 718 | if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP) |
| 719 | continue; |
| 720 | |
| 721 | if (!sctp_state(asoc, ESTABLISHED)) |
| 722 | continue; |
| 723 | |
| 724 | /* Check if any address in the packed array of addresses is |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 725 | * not present in the bind address list of the association. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 726 | * If so, do not send the asconf chunk to its peer, but |
| 727 | * continue with other associations. |
| 728 | */ |
| 729 | addr_buf = addrs; |
| 730 | for (i = 0; i < addrcnt; i++) { |
| 731 | laddr = (union sctp_addr *)addr_buf; |
| 732 | af = sctp_get_af_specific(laddr->v4.sin_family); |
| 733 | if (!af) { |
| 734 | retval = -EINVAL; |
| 735 | goto out; |
| 736 | } |
| 737 | |
| 738 | if (!sctp_assoc_lookup_laddr(asoc, laddr)) |
| 739 | break; |
| 740 | |
| 741 | addr_buf += af->sockaddr_len; |
| 742 | } |
| 743 | if (i < addrcnt) |
| 744 | continue; |
| 745 | |
| 746 | /* Find one address in the association's bind address list |
| 747 | * that is not in the packed array of addresses. This is to |
| 748 | * make sure that we do not delete all the addresses in the |
| 749 | * association. |
| 750 | */ |
| 751 | sctp_read_lock(&asoc->base.addr_lock); |
| 752 | bp = &asoc->base.bind_addr; |
| 753 | laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs, |
| 754 | addrcnt, sp); |
| 755 | sctp_read_unlock(&asoc->base.addr_lock); |
| 756 | if (!laddr) |
| 757 | continue; |
| 758 | |
| 759 | chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt, |
| 760 | SCTP_PARAM_DEL_IP); |
| 761 | if (!chunk) { |
| 762 | retval = -ENOMEM; |
| 763 | goto out; |
| 764 | } |
| 765 | |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 766 | /* Reset use_as_src flag for the addresses in the bind address |
| 767 | * list that are to be deleted. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 768 | */ |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 769 | sctp_local_bh_disable(); |
| 770 | sctp_write_lock(&asoc->base.addr_lock); |
| 771 | addr_buf = addrs; |
| 772 | for (i = 0; i < addrcnt; i++) { |
| 773 | laddr = (union sctp_addr *)addr_buf; |
| 774 | af = sctp_get_af_specific(laddr->v4.sin_family); |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 775 | list_for_each(pos1, &bp->address_list) { |
| 776 | saddr = list_entry(pos1, |
| 777 | struct sctp_sockaddr_entry, |
| 778 | list); |
Al Viro | 5f242a1 | 2006-11-20 17:05:23 -0800 | [diff] [blame] | 779 | if (sctp_cmp_addr_exact(&saddr->a, laddr)) |
Sridhar Samudrala | dc022a9 | 2006-07-21 14:49:25 -0700 | [diff] [blame] | 780 | saddr->use_as_src = 0; |
| 781 | } |
| 782 | addr_buf += af->sockaddr_len; |
| 783 | } |
| 784 | sctp_write_unlock(&asoc->base.addr_lock); |
| 785 | sctp_local_bh_enable(); |
| 786 | |
| 787 | /* Update the route and saddr entries for all the transports |
| 788 | * as some of the addresses in the bind address list are |
| 789 | * about to be deleted and cannot be used as source addresses. |
| 790 | */ |
| 791 | list_for_each(pos1, &asoc->peer.transport_addr_list) { |
| 792 | transport = list_entry(pos1, struct sctp_transport, |
| 793 | transports); |
| 794 | dst_release(transport->dst); |
| 795 | sctp_transport_route(transport, NULL, |
| 796 | sctp_sk(asoc->base.sk)); |
| 797 | } |
| 798 | |
| 799 | retval = sctp_send_asconf(asoc, chunk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 800 | } |
| 801 | out: |
| 802 | return retval; |
| 803 | } |
| 804 | |
| 805 | /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt() |
| 806 | * |
| 807 | * API 8.1 |
| 808 | * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt, |
| 809 | * int flags); |
| 810 | * |
| 811 | * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses. |
| 812 | * If the sd is an IPv6 socket, the addresses passed can either be IPv4 |
| 813 | * or IPv6 addresses. |
| 814 | * |
| 815 | * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see |
| 816 | * Section 3.1.2 for this usage. |
| 817 | * |
| 818 | * addrs is a pointer to an array of one or more socket addresses. Each |
| 819 | * address is contained in its appropriate structure (i.e. struct |
| 820 | * sockaddr_in or struct sockaddr_in6) the family of the address type |
Ville Nuorvala | 23c435f | 2006-10-16 22:08:28 -0700 | [diff] [blame] | 821 | * must be used to distinguish the address length (note that this |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 822 | * representation is termed a "packed array" of addresses). The caller |
| 823 | * specifies the number of addresses in the array with addrcnt. |
| 824 | * |
| 825 | * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns |
| 826 | * -1, and sets errno to the appropriate error code. |
| 827 | * |
| 828 | * For SCTP, the port given in each socket address must be the same, or |
| 829 | * sctp_bindx() will fail, setting errno to EINVAL. |
| 830 | * |
| 831 | * The flags parameter is formed from the bitwise OR of zero or more of |
| 832 | * the following currently defined flags: |
| 833 | * |
| 834 | * SCTP_BINDX_ADD_ADDR |
| 835 | * |
| 836 | * SCTP_BINDX_REM_ADDR |
| 837 | * |
| 838 | * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the |
| 839 | * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given |
| 840 | * addresses from the association. The two flags are mutually exclusive; |
| 841 | * if both are given, sctp_bindx() will fail with EINVAL. A caller may |
| 842 | * not remove all addresses from an association; sctp_bindx() will |
| 843 | * reject such an attempt with EINVAL. |
| 844 | * |
| 845 | * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate |
| 846 | * additional addresses with an endpoint after calling bind(). Or use |
| 847 | * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening |
| 848 | * socket is associated with so that no new association accepted will be |
| 849 | * associated with those addresses. If the endpoint supports dynamic |
| 850 | * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a |
| 851 | * endpoint to send the appropriate message to the peer to change the |
| 852 | * peers address lists. |
| 853 | * |
| 854 | * Adding and removing addresses from a connected association is |
| 855 | * optional functionality. Implementations that do not support this |
| 856 | * functionality should return EOPNOTSUPP. |
| 857 | * |
| 858 | * Basically do nothing but copying the addresses from user to kernel |
| 859 | * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk. |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 860 | * This is used for tunneling the sctp_bindx() request through sctp_setsockopt() |
| 861 | * from userspace. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 862 | * |
| 863 | * We don't use copy_from_user() for optimization: we first do the |
| 864 | * sanity checks (buffer size -fast- and access check-healthy |
| 865 | * pointer); if all of those succeed, then we can alloc the memory |
| 866 | * (expensive operation) needed to copy the data to kernel. Then we do |
| 867 | * the copying without checking the user space area |
| 868 | * (__copy_from_user()). |
| 869 | * |
| 870 | * On exit there is no need to do sockfd_put(), sys_setsockopt() does |
| 871 | * it. |
| 872 | * |
| 873 | * sk The sk of the socket |
| 874 | * addrs The pointer to the addresses in user land |
| 875 | * addrssize Size of the addrs buffer |
| 876 | * op Operation to perform (add or remove, see the flags of |
| 877 | * sctp_bindx) |
| 878 | * |
| 879 | * Returns 0 if ok, <0 errno code on error. |
| 880 | */ |
| 881 | SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk, |
| 882 | struct sockaddr __user *addrs, |
| 883 | int addrs_size, int op) |
| 884 | { |
| 885 | struct sockaddr *kaddrs; |
| 886 | int err; |
| 887 | int addrcnt = 0; |
| 888 | int walk_size = 0; |
| 889 | struct sockaddr *sa_addr; |
| 890 | void *addr_buf; |
| 891 | struct sctp_af *af; |
| 892 | |
| 893 | SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p" |
| 894 | " addrs_size %d opt %d\n", sk, addrs, addrs_size, op); |
| 895 | |
| 896 | if (unlikely(addrs_size <= 0)) |
| 897 | return -EINVAL; |
| 898 | |
| 899 | /* Check the user passed a healthy pointer. */ |
| 900 | if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size))) |
| 901 | return -EFAULT; |
| 902 | |
| 903 | /* Alloc space for the address array in kernel memory. */ |
Kris Katterjohn | 8b3a700 | 2006-01-11 15:56:43 -0800 | [diff] [blame] | 904 | kaddrs = kmalloc(addrs_size, GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 905 | if (unlikely(!kaddrs)) |
| 906 | return -ENOMEM; |
| 907 | |
| 908 | if (__copy_from_user(kaddrs, addrs, addrs_size)) { |
| 909 | kfree(kaddrs); |
| 910 | return -EFAULT; |
| 911 | } |
| 912 | |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 913 | /* Walk through the addrs buffer and count the number of addresses. */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 914 | addr_buf = kaddrs; |
| 915 | while (walk_size < addrs_size) { |
| 916 | sa_addr = (struct sockaddr *)addr_buf; |
| 917 | af = sctp_get_af_specific(sa_addr->sa_family); |
| 918 | |
| 919 | /* If the address family is not supported or if this address |
| 920 | * causes the address buffer to overflow return EINVAL. |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 921 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 922 | if (!af || (walk_size + af->sockaddr_len) > addrs_size) { |
| 923 | kfree(kaddrs); |
| 924 | return -EINVAL; |
| 925 | } |
| 926 | addrcnt++; |
| 927 | addr_buf += af->sockaddr_len; |
| 928 | walk_size += af->sockaddr_len; |
| 929 | } |
| 930 | |
| 931 | /* Do the work. */ |
| 932 | switch (op) { |
| 933 | case SCTP_BINDX_ADD_ADDR: |
| 934 | err = sctp_bindx_add(sk, kaddrs, addrcnt); |
| 935 | if (err) |
| 936 | goto out; |
| 937 | err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt); |
| 938 | break; |
| 939 | |
| 940 | case SCTP_BINDX_REM_ADDR: |
| 941 | err = sctp_bindx_rem(sk, kaddrs, addrcnt); |
| 942 | if (err) |
| 943 | goto out; |
| 944 | err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt); |
| 945 | break; |
| 946 | |
| 947 | default: |
| 948 | err = -EINVAL; |
| 949 | break; |
Stephen Hemminger | 3ff50b7 | 2007-04-20 17:09:22 -0700 | [diff] [blame] | 950 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 951 | |
| 952 | out: |
| 953 | kfree(kaddrs); |
| 954 | |
| 955 | return err; |
| 956 | } |
| 957 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 958 | /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size) |
| 959 | * |
| 960 | * Common routine for handling connect() and sctp_connectx(). |
| 961 | * Connect will come in with just a single address. |
| 962 | */ |
| 963 | static int __sctp_connect(struct sock* sk, |
| 964 | struct sockaddr *kaddrs, |
| 965 | int addrs_size) |
| 966 | { |
| 967 | struct sctp_sock *sp; |
| 968 | struct sctp_endpoint *ep; |
| 969 | struct sctp_association *asoc = NULL; |
| 970 | struct sctp_association *asoc2; |
| 971 | struct sctp_transport *transport; |
| 972 | union sctp_addr to; |
| 973 | struct sctp_af *af; |
| 974 | sctp_scope_t scope; |
| 975 | long timeo; |
| 976 | int err = 0; |
| 977 | int addrcnt = 0; |
| 978 | int walk_size = 0; |
Vlad Yasevich | e4d1fea | 2007-08-01 10:56:43 -0400 | [diff] [blame] | 979 | union sctp_addr *sa_addr = NULL; |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 980 | void *addr_buf; |
Vlad Yasevich | 16d00fb | 2007-05-04 13:34:09 -0700 | [diff] [blame] | 981 | unsigned short port; |
Vlad Yasevich | f50f95c | 2007-07-03 12:47:40 -0400 | [diff] [blame] | 982 | unsigned int f_flags = 0; |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 983 | |
| 984 | sp = sctp_sk(sk); |
| 985 | ep = sp->ep; |
| 986 | |
| 987 | /* connect() cannot be done on a socket that is already in ESTABLISHED |
| 988 | * state - UDP-style peeled off socket or a TCP-style socket that |
| 989 | * is already connected. |
| 990 | * It cannot be done even on a TCP-style listening socket. |
| 991 | */ |
| 992 | if (sctp_sstate(sk, ESTABLISHED) || |
| 993 | (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) { |
| 994 | err = -EISCONN; |
| 995 | goto out_free; |
| 996 | } |
| 997 | |
| 998 | /* Walk through the addrs buffer and count the number of addresses. */ |
| 999 | addr_buf = kaddrs; |
| 1000 | while (walk_size < addrs_size) { |
Al Viro | 4bdf4b5 | 2006-11-20 17:10:20 -0800 | [diff] [blame] | 1001 | sa_addr = (union sctp_addr *)addr_buf; |
| 1002 | af = sctp_get_af_specific(sa_addr->sa.sa_family); |
Vlad Yasevich | 16d00fb | 2007-05-04 13:34:09 -0700 | [diff] [blame] | 1003 | port = ntohs(sa_addr->v4.sin_port); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1004 | |
| 1005 | /* If the address family is not supported or if this address |
| 1006 | * causes the address buffer to overflow return EINVAL. |
| 1007 | */ |
| 1008 | if (!af || (walk_size + af->sockaddr_len) > addrs_size) { |
| 1009 | err = -EINVAL; |
| 1010 | goto out_free; |
| 1011 | } |
| 1012 | |
Vlad Yasevich | e4d1fea | 2007-08-01 10:56:43 -0400 | [diff] [blame] | 1013 | /* Save current address so we can work with it */ |
| 1014 | memcpy(&to, sa_addr, af->sockaddr_len); |
| 1015 | |
| 1016 | err = sctp_verify_addr(sk, &to, af->sockaddr_len); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1017 | if (err) |
| 1018 | goto out_free; |
| 1019 | |
Vlad Yasevich | 16d00fb | 2007-05-04 13:34:09 -0700 | [diff] [blame] | 1020 | /* Make sure the destination port is correctly set |
| 1021 | * in all addresses. |
| 1022 | */ |
| 1023 | if (asoc && asoc->peer.port && asoc->peer.port != port) |
| 1024 | goto out_free; |
| 1025 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1026 | |
| 1027 | /* Check if there already is a matching association on the |
| 1028 | * endpoint (other than the one created here). |
| 1029 | */ |
Vlad Yasevich | e4d1fea | 2007-08-01 10:56:43 -0400 | [diff] [blame] | 1030 | asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1031 | if (asoc2 && asoc2 != asoc) { |
| 1032 | if (asoc2->state >= SCTP_STATE_ESTABLISHED) |
| 1033 | err = -EISCONN; |
| 1034 | else |
| 1035 | err = -EALREADY; |
| 1036 | goto out_free; |
| 1037 | } |
| 1038 | |
| 1039 | /* If we could not find a matching association on the endpoint, |
| 1040 | * make sure that there is no peeled-off association matching |
| 1041 | * the peer address even on another socket. |
| 1042 | */ |
Vlad Yasevich | e4d1fea | 2007-08-01 10:56:43 -0400 | [diff] [blame] | 1043 | if (sctp_endpoint_is_peeled_off(ep, &to)) { |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1044 | err = -EADDRNOTAVAIL; |
| 1045 | goto out_free; |
| 1046 | } |
| 1047 | |
| 1048 | if (!asoc) { |
| 1049 | /* If a bind() or sctp_bindx() is not called prior to |
| 1050 | * an sctp_connectx() call, the system picks an |
| 1051 | * ephemeral port and will choose an address set |
| 1052 | * equivalent to binding with a wildcard address. |
| 1053 | */ |
| 1054 | if (!ep->base.bind_addr.port) { |
| 1055 | if (sctp_autobind(sk)) { |
| 1056 | err = -EAGAIN; |
| 1057 | goto out_free; |
| 1058 | } |
Ivan Skytte Jorgensen | 64a0c1c | 2005-10-28 15:39:02 -0700 | [diff] [blame] | 1059 | } else { |
| 1060 | /* |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 1061 | * If an unprivileged user inherits a 1-many |
| 1062 | * style socket with open associations on a |
| 1063 | * privileged port, it MAY be permitted to |
| 1064 | * accept new associations, but it SHOULD NOT |
Ivan Skytte Jorgensen | 64a0c1c | 2005-10-28 15:39:02 -0700 | [diff] [blame] | 1065 | * be permitted to open new associations. |
| 1066 | */ |
| 1067 | if (ep->base.bind_addr.port < PROT_SOCK && |
| 1068 | !capable(CAP_NET_BIND_SERVICE)) { |
| 1069 | err = -EACCES; |
| 1070 | goto out_free; |
| 1071 | } |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1072 | } |
| 1073 | |
Vlad Yasevich | e4d1fea | 2007-08-01 10:56:43 -0400 | [diff] [blame] | 1074 | scope = sctp_scope(&to); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1075 | asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL); |
| 1076 | if (!asoc) { |
| 1077 | err = -ENOMEM; |
| 1078 | goto out_free; |
| 1079 | } |
| 1080 | } |
| 1081 | |
| 1082 | /* Prime the peer's transport structures. */ |
Vlad Yasevich | e4d1fea | 2007-08-01 10:56:43 -0400 | [diff] [blame] | 1083 | transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1084 | SCTP_UNKNOWN); |
| 1085 | if (!transport) { |
| 1086 | err = -ENOMEM; |
| 1087 | goto out_free; |
| 1088 | } |
| 1089 | |
| 1090 | addrcnt++; |
| 1091 | addr_buf += af->sockaddr_len; |
| 1092 | walk_size += af->sockaddr_len; |
| 1093 | } |
| 1094 | |
| 1095 | err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL); |
| 1096 | if (err < 0) { |
| 1097 | goto out_free; |
| 1098 | } |
| 1099 | |
| 1100 | err = sctp_primitive_ASSOCIATE(asoc, NULL); |
| 1101 | if (err < 0) { |
| 1102 | goto out_free; |
| 1103 | } |
| 1104 | |
| 1105 | /* Initialize sk's dport and daddr for getpeername() */ |
| 1106 | inet_sk(sk)->dport = htons(asoc->peer.port); |
Vlad Yasevich | e4d1fea | 2007-08-01 10:56:43 -0400 | [diff] [blame] | 1107 | af = sctp_get_af_specific(sa_addr->sa.sa_family); |
| 1108 | af->to_sk_daddr(sa_addr, sk); |
Sridhar Samudrala | 8de8c87 | 2006-05-19 10:58:12 -0700 | [diff] [blame] | 1109 | sk->sk_err = 0; |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1110 | |
Vlad Yasevich | f50f95c | 2007-07-03 12:47:40 -0400 | [diff] [blame] | 1111 | /* in-kernel sockets don't generally have a file allocated to them |
| 1112 | * if all they do is call sock_create_kern(). |
| 1113 | */ |
| 1114 | if (sk->sk_socket->file) |
| 1115 | f_flags = sk->sk_socket->file->f_flags; |
| 1116 | |
| 1117 | timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK); |
| 1118 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1119 | err = sctp_wait_for_connect(asoc, &timeo); |
| 1120 | |
| 1121 | /* Don't free association on exit. */ |
| 1122 | asoc = NULL; |
| 1123 | |
| 1124 | out_free: |
| 1125 | |
| 1126 | SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p" |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 1127 | " kaddrs: %p err: %d\n", |
| 1128 | asoc, kaddrs, err); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1129 | if (asoc) |
| 1130 | sctp_association_free(asoc); |
| 1131 | return err; |
| 1132 | } |
| 1133 | |
| 1134 | /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt() |
| 1135 | * |
| 1136 | * API 8.9 |
| 1137 | * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt); |
| 1138 | * |
| 1139 | * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses. |
| 1140 | * If the sd is an IPv6 socket, the addresses passed can either be IPv4 |
| 1141 | * or IPv6 addresses. |
| 1142 | * |
| 1143 | * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see |
| 1144 | * Section 3.1.2 for this usage. |
| 1145 | * |
| 1146 | * addrs is a pointer to an array of one or more socket addresses. Each |
| 1147 | * address is contained in its appropriate structure (i.e. struct |
| 1148 | * sockaddr_in or struct sockaddr_in6) the family of the address type |
| 1149 | * must be used to distengish the address length (note that this |
| 1150 | * representation is termed a "packed array" of addresses). The caller |
| 1151 | * specifies the number of addresses in the array with addrcnt. |
| 1152 | * |
| 1153 | * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns |
| 1154 | * -1, and sets errno to the appropriate error code. |
| 1155 | * |
| 1156 | * For SCTP, the port given in each socket address must be the same, or |
| 1157 | * sctp_connectx() will fail, setting errno to EINVAL. |
| 1158 | * |
| 1159 | * An application can use sctp_connectx to initiate an association with |
| 1160 | * an endpoint that is multi-homed. Much like sctp_bindx() this call |
| 1161 | * allows a caller to specify multiple addresses at which a peer can be |
| 1162 | * reached. The way the SCTP stack uses the list of addresses to set up |
| 1163 | * the association is implementation dependant. This function only |
| 1164 | * specifies that the stack will try to make use of all the addresses in |
| 1165 | * the list when needed. |
| 1166 | * |
| 1167 | * Note that the list of addresses passed in is only used for setting up |
| 1168 | * the association. It does not necessarily equal the set of addresses |
| 1169 | * the peer uses for the resulting association. If the caller wants to |
| 1170 | * find out the set of peer addresses, it must use sctp_getpaddrs() to |
| 1171 | * retrieve them after the association has been set up. |
| 1172 | * |
| 1173 | * Basically do nothing but copying the addresses from user to kernel |
| 1174 | * land and invoking either sctp_connectx(). This is used for tunneling |
| 1175 | * the sctp_connectx() request through sctp_setsockopt() from userspace. |
| 1176 | * |
| 1177 | * We don't use copy_from_user() for optimization: we first do the |
| 1178 | * sanity checks (buffer size -fast- and access check-healthy |
| 1179 | * pointer); if all of those succeed, then we can alloc the memory |
| 1180 | * (expensive operation) needed to copy the data to kernel. Then we do |
| 1181 | * the copying without checking the user space area |
| 1182 | * (__copy_from_user()). |
| 1183 | * |
| 1184 | * On exit there is no need to do sockfd_put(), sys_setsockopt() does |
| 1185 | * it. |
| 1186 | * |
| 1187 | * sk The sk of the socket |
| 1188 | * addrs The pointer to the addresses in user land |
| 1189 | * addrssize Size of the addrs buffer |
| 1190 | * |
| 1191 | * Returns 0 if ok, <0 errno code on error. |
| 1192 | */ |
| 1193 | SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk, |
| 1194 | struct sockaddr __user *addrs, |
| 1195 | int addrs_size) |
| 1196 | { |
| 1197 | int err = 0; |
| 1198 | struct sockaddr *kaddrs; |
| 1199 | |
| 1200 | SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n", |
| 1201 | __FUNCTION__, sk, addrs, addrs_size); |
| 1202 | |
| 1203 | if (unlikely(addrs_size <= 0)) |
| 1204 | return -EINVAL; |
| 1205 | |
| 1206 | /* Check the user passed a healthy pointer. */ |
| 1207 | if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size))) |
| 1208 | return -EFAULT; |
| 1209 | |
| 1210 | /* Alloc space for the address array in kernel memory. */ |
Kris Katterjohn | 8b3a700 | 2006-01-11 15:56:43 -0800 | [diff] [blame] | 1211 | kaddrs = kmalloc(addrs_size, GFP_KERNEL); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1212 | if (unlikely(!kaddrs)) |
| 1213 | return -ENOMEM; |
| 1214 | |
| 1215 | if (__copy_from_user(kaddrs, addrs, addrs_size)) { |
| 1216 | err = -EFAULT; |
| 1217 | } else { |
| 1218 | err = __sctp_connect(sk, kaddrs, addrs_size); |
| 1219 | } |
| 1220 | |
| 1221 | kfree(kaddrs); |
| 1222 | return err; |
| 1223 | } |
| 1224 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1225 | /* API 3.1.4 close() - UDP Style Syntax |
| 1226 | * Applications use close() to perform graceful shutdown (as described in |
| 1227 | * Section 10.1 of [SCTP]) on ALL the associations currently represented |
| 1228 | * by a UDP-style socket. |
| 1229 | * |
| 1230 | * The syntax is |
| 1231 | * |
| 1232 | * ret = close(int sd); |
| 1233 | * |
| 1234 | * sd - the socket descriptor of the associations to be closed. |
| 1235 | * |
| 1236 | * To gracefully shutdown a specific association represented by the |
| 1237 | * UDP-style socket, an application should use the sendmsg() call, |
| 1238 | * passing no user data, but including the appropriate flag in the |
| 1239 | * ancillary data (see Section xxxx). |
| 1240 | * |
| 1241 | * If sd in the close() call is a branched-off socket representing only |
| 1242 | * one association, the shutdown is performed on that association only. |
| 1243 | * |
| 1244 | * 4.1.6 close() - TCP Style Syntax |
| 1245 | * |
| 1246 | * Applications use close() to gracefully close down an association. |
| 1247 | * |
| 1248 | * The syntax is: |
| 1249 | * |
| 1250 | * int close(int sd); |
| 1251 | * |
| 1252 | * sd - the socket descriptor of the association to be closed. |
| 1253 | * |
| 1254 | * After an application calls close() on a socket descriptor, no further |
| 1255 | * socket operations will succeed on that descriptor. |
| 1256 | * |
| 1257 | * API 7.1.4 SO_LINGER |
| 1258 | * |
| 1259 | * An application using the TCP-style socket can use this option to |
| 1260 | * perform the SCTP ABORT primitive. The linger option structure is: |
| 1261 | * |
| 1262 | * struct linger { |
| 1263 | * int l_onoff; // option on/off |
| 1264 | * int l_linger; // linger time |
| 1265 | * }; |
| 1266 | * |
| 1267 | * To enable the option, set l_onoff to 1. If the l_linger value is set |
| 1268 | * to 0, calling close() is the same as the ABORT primitive. If the |
| 1269 | * value is set to a negative value, the setsockopt() call will return |
| 1270 | * an error. If the value is set to a positive value linger_time, the |
| 1271 | * close() can be blocked for at most linger_time ms. If the graceful |
| 1272 | * shutdown phase does not finish during this period, close() will |
| 1273 | * return but the graceful shutdown phase continues in the system. |
| 1274 | */ |
| 1275 | SCTP_STATIC void sctp_close(struct sock *sk, long timeout) |
| 1276 | { |
| 1277 | struct sctp_endpoint *ep; |
| 1278 | struct sctp_association *asoc; |
| 1279 | struct list_head *pos, *temp; |
| 1280 | |
| 1281 | SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout); |
| 1282 | |
| 1283 | sctp_lock_sock(sk); |
| 1284 | sk->sk_shutdown = SHUTDOWN_MASK; |
| 1285 | |
| 1286 | ep = sctp_sk(sk)->ep; |
| 1287 | |
Vladislav Yasevich | 61c9fed | 2006-05-19 11:01:18 -0700 | [diff] [blame] | 1288 | /* Walk all associations on an endpoint. */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1289 | list_for_each_safe(pos, temp, &ep->asocs) { |
| 1290 | asoc = list_entry(pos, struct sctp_association, asocs); |
| 1291 | |
| 1292 | if (sctp_style(sk, TCP)) { |
| 1293 | /* A closed association can still be in the list if |
| 1294 | * it belongs to a TCP-style listening socket that is |
| 1295 | * not yet accepted. If so, free it. If not, send an |
| 1296 | * ABORT or SHUTDOWN based on the linger options. |
| 1297 | */ |
| 1298 | if (sctp_state(asoc, CLOSED)) { |
| 1299 | sctp_unhash_established(asoc); |
| 1300 | sctp_association_free(asoc); |
Vladislav Yasevich | b89498a | 2006-05-19 14:32:06 -0700 | [diff] [blame] | 1301 | continue; |
| 1302 | } |
| 1303 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1304 | |
Sridhar Samudrala | b9ac867 | 2006-08-28 13:53:01 -0700 | [diff] [blame] | 1305 | if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) { |
| 1306 | struct sctp_chunk *chunk; |
| 1307 | |
| 1308 | chunk = sctp_make_abort_user(asoc, NULL, 0); |
| 1309 | if (chunk) |
| 1310 | sctp_primitive_ABORT(asoc, chunk); |
| 1311 | } else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1312 | sctp_primitive_SHUTDOWN(asoc, NULL); |
| 1313 | } |
| 1314 | |
| 1315 | /* Clean up any skbs sitting on the receive queue. */ |
| 1316 | sctp_queue_purge_ulpevents(&sk->sk_receive_queue); |
| 1317 | sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby); |
| 1318 | |
| 1319 | /* On a TCP-style socket, block for at most linger_time if set. */ |
| 1320 | if (sctp_style(sk, TCP) && timeout) |
| 1321 | sctp_wait_for_close(sk, timeout); |
| 1322 | |
| 1323 | /* This will run the backlog queue. */ |
| 1324 | sctp_release_sock(sk); |
| 1325 | |
| 1326 | /* Supposedly, no process has access to the socket, but |
| 1327 | * the net layers still may. |
| 1328 | */ |
| 1329 | sctp_local_bh_disable(); |
| 1330 | sctp_bh_lock_sock(sk); |
| 1331 | |
| 1332 | /* Hold the sock, since sk_common_release() will put sock_put() |
| 1333 | * and we have just a little more cleanup. |
| 1334 | */ |
| 1335 | sock_hold(sk); |
| 1336 | sk_common_release(sk); |
| 1337 | |
| 1338 | sctp_bh_unlock_sock(sk); |
| 1339 | sctp_local_bh_enable(); |
| 1340 | |
| 1341 | sock_put(sk); |
| 1342 | |
| 1343 | SCTP_DBG_OBJCNT_DEC(sock); |
| 1344 | } |
| 1345 | |
| 1346 | /* Handle EPIPE error. */ |
| 1347 | static int sctp_error(struct sock *sk, int flags, int err) |
| 1348 | { |
| 1349 | if (err == -EPIPE) |
| 1350 | err = sock_error(sk) ? : -EPIPE; |
| 1351 | if (err == -EPIPE && !(flags & MSG_NOSIGNAL)) |
| 1352 | send_sig(SIGPIPE, current, 0); |
| 1353 | return err; |
| 1354 | } |
| 1355 | |
| 1356 | /* API 3.1.3 sendmsg() - UDP Style Syntax |
| 1357 | * |
| 1358 | * An application uses sendmsg() and recvmsg() calls to transmit data to |
| 1359 | * and receive data from its peer. |
| 1360 | * |
| 1361 | * ssize_t sendmsg(int socket, const struct msghdr *message, |
| 1362 | * int flags); |
| 1363 | * |
| 1364 | * socket - the socket descriptor of the endpoint. |
| 1365 | * message - pointer to the msghdr structure which contains a single |
| 1366 | * user message and possibly some ancillary data. |
| 1367 | * |
| 1368 | * See Section 5 for complete description of the data |
| 1369 | * structures. |
| 1370 | * |
| 1371 | * flags - flags sent or received with the user message, see Section |
| 1372 | * 5 for complete description of the flags. |
| 1373 | * |
| 1374 | * Note: This function could use a rewrite especially when explicit |
| 1375 | * connect support comes in. |
| 1376 | */ |
| 1377 | /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */ |
| 1378 | |
| 1379 | SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *); |
| 1380 | |
| 1381 | SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk, |
| 1382 | struct msghdr *msg, size_t msg_len) |
| 1383 | { |
| 1384 | struct sctp_sock *sp; |
| 1385 | struct sctp_endpoint *ep; |
| 1386 | struct sctp_association *new_asoc=NULL, *asoc=NULL; |
| 1387 | struct sctp_transport *transport, *chunk_tp; |
| 1388 | struct sctp_chunk *chunk; |
Al Viro | dce116a | 2006-11-20 17:25:15 -0800 | [diff] [blame] | 1389 | union sctp_addr to; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1390 | struct sockaddr *msg_name = NULL; |
| 1391 | struct sctp_sndrcvinfo default_sinfo = { 0 }; |
| 1392 | struct sctp_sndrcvinfo *sinfo; |
| 1393 | struct sctp_initmsg *sinit; |
| 1394 | sctp_assoc_t associd = 0; |
| 1395 | sctp_cmsgs_t cmsgs = { NULL }; |
| 1396 | int err; |
| 1397 | sctp_scope_t scope; |
| 1398 | long timeo; |
| 1399 | __u16 sinfo_flags = 0; |
| 1400 | struct sctp_datamsg *datamsg; |
| 1401 | struct list_head *pos; |
| 1402 | int msg_flags = msg->msg_flags; |
| 1403 | |
| 1404 | SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n", |
| 1405 | sk, msg, msg_len); |
| 1406 | |
| 1407 | err = 0; |
| 1408 | sp = sctp_sk(sk); |
| 1409 | ep = sp->ep; |
| 1410 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 1411 | SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1412 | |
| 1413 | /* We cannot send a message over a TCP-style listening socket. */ |
| 1414 | if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) { |
| 1415 | err = -EPIPE; |
| 1416 | goto out_nounlock; |
| 1417 | } |
| 1418 | |
| 1419 | /* Parse out the SCTP CMSGs. */ |
| 1420 | err = sctp_msghdr_parse(msg, &cmsgs); |
| 1421 | |
| 1422 | if (err) { |
| 1423 | SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err); |
| 1424 | goto out_nounlock; |
| 1425 | } |
| 1426 | |
| 1427 | /* Fetch the destination address for this packet. This |
| 1428 | * address only selects the association--it is not necessarily |
| 1429 | * the address we will send to. |
| 1430 | * For a peeled-off socket, msg_name is ignored. |
| 1431 | */ |
| 1432 | if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) { |
| 1433 | int msg_namelen = msg->msg_namelen; |
| 1434 | |
| 1435 | err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name, |
| 1436 | msg_namelen); |
| 1437 | if (err) |
| 1438 | return err; |
| 1439 | |
| 1440 | if (msg_namelen > sizeof(to)) |
| 1441 | msg_namelen = sizeof(to); |
| 1442 | memcpy(&to, msg->msg_name, msg_namelen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1443 | msg_name = msg->msg_name; |
| 1444 | } |
| 1445 | |
| 1446 | sinfo = cmsgs.info; |
| 1447 | sinit = cmsgs.init; |
| 1448 | |
| 1449 | /* Did the user specify SNDRCVINFO? */ |
| 1450 | if (sinfo) { |
| 1451 | sinfo_flags = sinfo->sinfo_flags; |
| 1452 | associd = sinfo->sinfo_assoc_id; |
| 1453 | } |
| 1454 | |
| 1455 | SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n", |
| 1456 | msg_len, sinfo_flags); |
| 1457 | |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1458 | /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */ |
| 1459 | if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1460 | err = -EINVAL; |
| 1461 | goto out_nounlock; |
| 1462 | } |
| 1463 | |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1464 | /* If SCTP_EOF is set, no data can be sent. Disallow sending zero |
| 1465 | * length messages when SCTP_EOF|SCTP_ABORT is not set. |
| 1466 | * If SCTP_ABORT is set, the message length could be non zero with |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1467 | * the msg_iov set to the user abort reason. |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 1468 | */ |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1469 | if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) || |
| 1470 | (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1471 | err = -EINVAL; |
| 1472 | goto out_nounlock; |
| 1473 | } |
| 1474 | |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1475 | /* If SCTP_ADDR_OVER is set, there must be an address |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1476 | * specified in msg_name. |
| 1477 | */ |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1478 | if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1479 | err = -EINVAL; |
| 1480 | goto out_nounlock; |
| 1481 | } |
| 1482 | |
| 1483 | transport = NULL; |
| 1484 | |
| 1485 | SCTP_DEBUG_PRINTK("About to look up association.\n"); |
| 1486 | |
| 1487 | sctp_lock_sock(sk); |
| 1488 | |
| 1489 | /* If a msg_name has been specified, assume this is to be used. */ |
| 1490 | if (msg_name) { |
| 1491 | /* Look for a matching association on the endpoint. */ |
Al Viro | dce116a | 2006-11-20 17:25:15 -0800 | [diff] [blame] | 1492 | asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1493 | if (!asoc) { |
| 1494 | /* If we could not find a matching association on the |
| 1495 | * endpoint, make sure that it is not a TCP-style |
| 1496 | * socket that already has an association or there is |
| 1497 | * no peeled-off association on another socket. |
| 1498 | */ |
| 1499 | if ((sctp_style(sk, TCP) && |
| 1500 | sctp_sstate(sk, ESTABLISHED)) || |
Al Viro | dce116a | 2006-11-20 17:25:15 -0800 | [diff] [blame] | 1501 | sctp_endpoint_is_peeled_off(ep, &to)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1502 | err = -EADDRNOTAVAIL; |
| 1503 | goto out_unlock; |
| 1504 | } |
| 1505 | } |
| 1506 | } else { |
| 1507 | asoc = sctp_id2assoc(sk, associd); |
| 1508 | if (!asoc) { |
| 1509 | err = -EPIPE; |
| 1510 | goto out_unlock; |
| 1511 | } |
| 1512 | } |
| 1513 | |
| 1514 | if (asoc) { |
| 1515 | SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc); |
| 1516 | |
| 1517 | /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED |
| 1518 | * socket that has an association in CLOSED state. This can |
| 1519 | * happen when an accepted socket has an association that is |
| 1520 | * already CLOSED. |
| 1521 | */ |
| 1522 | if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) { |
| 1523 | err = -EPIPE; |
| 1524 | goto out_unlock; |
| 1525 | } |
| 1526 | |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1527 | if (sinfo_flags & SCTP_EOF) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1528 | SCTP_DEBUG_PRINTK("Shutting down association: %p\n", |
| 1529 | asoc); |
| 1530 | sctp_primitive_SHUTDOWN(asoc, NULL); |
| 1531 | err = 0; |
| 1532 | goto out_unlock; |
| 1533 | } |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1534 | if (sinfo_flags & SCTP_ABORT) { |
Sridhar Samudrala | c164a9b | 2006-08-22 11:50:39 -0700 | [diff] [blame] | 1535 | |
| 1536 | chunk = sctp_make_abort_user(asoc, msg, msg_len); |
| 1537 | if (!chunk) { |
| 1538 | err = -ENOMEM; |
| 1539 | goto out_unlock; |
| 1540 | } |
| 1541 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1542 | SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc); |
Sridhar Samudrala | c164a9b | 2006-08-22 11:50:39 -0700 | [diff] [blame] | 1543 | sctp_primitive_ABORT(asoc, chunk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1544 | err = 0; |
| 1545 | goto out_unlock; |
| 1546 | } |
| 1547 | } |
| 1548 | |
| 1549 | /* Do we need to create the association? */ |
| 1550 | if (!asoc) { |
| 1551 | SCTP_DEBUG_PRINTK("There is no association yet.\n"); |
| 1552 | |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1553 | if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1554 | err = -EINVAL; |
| 1555 | goto out_unlock; |
| 1556 | } |
| 1557 | |
| 1558 | /* Check for invalid stream against the stream counts, |
| 1559 | * either the default or the user specified stream counts. |
| 1560 | */ |
| 1561 | if (sinfo) { |
| 1562 | if (!sinit || (sinit && !sinit->sinit_num_ostreams)) { |
| 1563 | /* Check against the defaults. */ |
| 1564 | if (sinfo->sinfo_stream >= |
| 1565 | sp->initmsg.sinit_num_ostreams) { |
| 1566 | err = -EINVAL; |
| 1567 | goto out_unlock; |
| 1568 | } |
| 1569 | } else { |
| 1570 | /* Check against the requested. */ |
| 1571 | if (sinfo->sinfo_stream >= |
| 1572 | sinit->sinit_num_ostreams) { |
| 1573 | err = -EINVAL; |
| 1574 | goto out_unlock; |
| 1575 | } |
| 1576 | } |
| 1577 | } |
| 1578 | |
| 1579 | /* |
| 1580 | * API 3.1.2 bind() - UDP Style Syntax |
| 1581 | * If a bind() or sctp_bindx() is not called prior to a |
| 1582 | * sendmsg() call that initiates a new association, the |
| 1583 | * system picks an ephemeral port and will choose an address |
| 1584 | * set equivalent to binding with a wildcard address. |
| 1585 | */ |
| 1586 | if (!ep->base.bind_addr.port) { |
| 1587 | if (sctp_autobind(sk)) { |
| 1588 | err = -EAGAIN; |
| 1589 | goto out_unlock; |
| 1590 | } |
Ivan Skytte Jorgensen | 64a0c1c | 2005-10-28 15:39:02 -0700 | [diff] [blame] | 1591 | } else { |
| 1592 | /* |
| 1593 | * If an unprivileged user inherits a one-to-many |
| 1594 | * style socket with open associations on a privileged |
| 1595 | * port, it MAY be permitted to accept new associations, |
| 1596 | * but it SHOULD NOT be permitted to open new |
| 1597 | * associations. |
| 1598 | */ |
| 1599 | if (ep->base.bind_addr.port < PROT_SOCK && |
| 1600 | !capable(CAP_NET_BIND_SERVICE)) { |
| 1601 | err = -EACCES; |
| 1602 | goto out_unlock; |
| 1603 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1604 | } |
| 1605 | |
| 1606 | scope = sctp_scope(&to); |
| 1607 | new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL); |
| 1608 | if (!new_asoc) { |
| 1609 | err = -ENOMEM; |
| 1610 | goto out_unlock; |
| 1611 | } |
| 1612 | asoc = new_asoc; |
| 1613 | |
| 1614 | /* If the SCTP_INIT ancillary data is specified, set all |
| 1615 | * the association init values accordingly. |
| 1616 | */ |
| 1617 | if (sinit) { |
| 1618 | if (sinit->sinit_num_ostreams) { |
| 1619 | asoc->c.sinit_num_ostreams = |
| 1620 | sinit->sinit_num_ostreams; |
| 1621 | } |
| 1622 | if (sinit->sinit_max_instreams) { |
| 1623 | asoc->c.sinit_max_instreams = |
| 1624 | sinit->sinit_max_instreams; |
| 1625 | } |
| 1626 | if (sinit->sinit_max_attempts) { |
| 1627 | asoc->max_init_attempts |
| 1628 | = sinit->sinit_max_attempts; |
| 1629 | } |
| 1630 | if (sinit->sinit_max_init_timeo) { |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 1631 | asoc->max_init_timeo = |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1632 | msecs_to_jiffies(sinit->sinit_max_init_timeo); |
| 1633 | } |
| 1634 | } |
| 1635 | |
| 1636 | /* Prime the peer's transport structures. */ |
Al Viro | dce116a | 2006-11-20 17:25:15 -0800 | [diff] [blame] | 1637 | transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1638 | if (!transport) { |
| 1639 | err = -ENOMEM; |
| 1640 | goto out_free; |
| 1641 | } |
| 1642 | err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL); |
| 1643 | if (err < 0) { |
| 1644 | err = -ENOMEM; |
| 1645 | goto out_free; |
| 1646 | } |
| 1647 | } |
| 1648 | |
| 1649 | /* ASSERT: we have a valid association at this point. */ |
| 1650 | SCTP_DEBUG_PRINTK("We have a valid association.\n"); |
| 1651 | |
| 1652 | if (!sinfo) { |
| 1653 | /* If the user didn't specify SNDRCVINFO, make up one with |
| 1654 | * some defaults. |
| 1655 | */ |
| 1656 | default_sinfo.sinfo_stream = asoc->default_stream; |
| 1657 | default_sinfo.sinfo_flags = asoc->default_flags; |
| 1658 | default_sinfo.sinfo_ppid = asoc->default_ppid; |
| 1659 | default_sinfo.sinfo_context = asoc->default_context; |
| 1660 | default_sinfo.sinfo_timetolive = asoc->default_timetolive; |
| 1661 | default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc); |
| 1662 | sinfo = &default_sinfo; |
| 1663 | } |
| 1664 | |
| 1665 | /* API 7.1.7, the sndbuf size per association bounds the |
| 1666 | * maximum size of data that can be sent in a single send call. |
| 1667 | */ |
| 1668 | if (msg_len > sk->sk_sndbuf) { |
| 1669 | err = -EMSGSIZE; |
| 1670 | goto out_free; |
| 1671 | } |
| 1672 | |
Vlad Yasevich | 8a47949 | 2007-06-07 14:21:05 -0400 | [diff] [blame] | 1673 | if (asoc->pmtu_pending) |
| 1674 | sctp_assoc_pending_pmtu(asoc); |
| 1675 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1676 | /* If fragmentation is disabled and the message length exceeds the |
| 1677 | * association fragmentation point, return EMSGSIZE. The I-D |
| 1678 | * does not specify what this error is, but this looks like |
| 1679 | * a great fit. |
| 1680 | */ |
| 1681 | if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) { |
| 1682 | err = -EMSGSIZE; |
| 1683 | goto out_free; |
| 1684 | } |
| 1685 | |
| 1686 | if (sinfo) { |
| 1687 | /* Check for invalid stream. */ |
| 1688 | if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) { |
| 1689 | err = -EINVAL; |
| 1690 | goto out_free; |
| 1691 | } |
| 1692 | } |
| 1693 | |
| 1694 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); |
| 1695 | if (!sctp_wspace(asoc)) { |
| 1696 | err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len); |
| 1697 | if (err) |
| 1698 | goto out_free; |
| 1699 | } |
| 1700 | |
| 1701 | /* If an address is passed with the sendto/sendmsg call, it is used |
| 1702 | * to override the primary destination address in the TCP model, or |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1703 | * when SCTP_ADDR_OVER flag is set in the UDP model. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1704 | */ |
| 1705 | if ((sctp_style(sk, TCP) && msg_name) || |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 1706 | (sinfo_flags & SCTP_ADDR_OVER)) { |
Al Viro | dce116a | 2006-11-20 17:25:15 -0800 | [diff] [blame] | 1707 | chunk_tp = sctp_assoc_lookup_paddr(asoc, &to); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1708 | if (!chunk_tp) { |
| 1709 | err = -EINVAL; |
| 1710 | goto out_free; |
| 1711 | } |
| 1712 | } else |
| 1713 | chunk_tp = NULL; |
| 1714 | |
| 1715 | /* Auto-connect, if we aren't connected already. */ |
| 1716 | if (sctp_state(asoc, CLOSED)) { |
| 1717 | err = sctp_primitive_ASSOCIATE(asoc, NULL); |
| 1718 | if (err < 0) |
| 1719 | goto out_free; |
| 1720 | SCTP_DEBUG_PRINTK("We associated primitively.\n"); |
| 1721 | } |
| 1722 | |
| 1723 | /* Break the message into multiple chunks of maximum size. */ |
| 1724 | datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len); |
| 1725 | if (!datamsg) { |
| 1726 | err = -ENOMEM; |
| 1727 | goto out_free; |
| 1728 | } |
| 1729 | |
| 1730 | /* Now send the (possibly) fragmented message. */ |
| 1731 | list_for_each(pos, &datamsg->chunks) { |
| 1732 | chunk = list_entry(pos, struct sctp_chunk, frag_list); |
| 1733 | sctp_datamsg_track(chunk); |
| 1734 | |
| 1735 | /* Do accounting for the write space. */ |
| 1736 | sctp_set_owner_w(chunk); |
| 1737 | |
| 1738 | chunk->transport = chunk_tp; |
| 1739 | |
| 1740 | /* Send it to the lower layers. Note: all chunks |
| 1741 | * must either fail or succeed. The lower layer |
| 1742 | * works that way today. Keep it that way or this |
| 1743 | * breaks. |
| 1744 | */ |
| 1745 | err = sctp_primitive_SEND(asoc, chunk); |
| 1746 | /* Did the lower layer accept the chunk? */ |
| 1747 | if (err) |
| 1748 | sctp_chunk_free(chunk); |
| 1749 | SCTP_DEBUG_PRINTK("We sent primitively.\n"); |
| 1750 | } |
| 1751 | |
| 1752 | sctp_datamsg_free(datamsg); |
| 1753 | if (err) |
| 1754 | goto out_free; |
| 1755 | else |
| 1756 | err = msg_len; |
| 1757 | |
| 1758 | /* If we are already past ASSOCIATE, the lower |
| 1759 | * layers are responsible for association cleanup. |
| 1760 | */ |
| 1761 | goto out_unlock; |
| 1762 | |
| 1763 | out_free: |
| 1764 | if (new_asoc) |
| 1765 | sctp_association_free(asoc); |
| 1766 | out_unlock: |
| 1767 | sctp_release_sock(sk); |
| 1768 | |
| 1769 | out_nounlock: |
| 1770 | return sctp_error(sk, msg_flags, err); |
| 1771 | |
| 1772 | #if 0 |
| 1773 | do_sock_err: |
| 1774 | if (msg_len) |
| 1775 | err = msg_len; |
| 1776 | else |
| 1777 | err = sock_error(sk); |
| 1778 | goto out; |
| 1779 | |
| 1780 | do_interrupted: |
| 1781 | if (msg_len) |
| 1782 | err = msg_len; |
| 1783 | goto out; |
| 1784 | #endif /* 0 */ |
| 1785 | } |
| 1786 | |
| 1787 | /* This is an extended version of skb_pull() that removes the data from the |
| 1788 | * start of a skb even when data is spread across the list of skb's in the |
| 1789 | * frag_list. len specifies the total amount of data that needs to be removed. |
| 1790 | * when 'len' bytes could be removed from the skb, it returns 0. |
| 1791 | * If 'len' exceeds the total skb length, it returns the no. of bytes that |
| 1792 | * could not be removed. |
| 1793 | */ |
| 1794 | static int sctp_skb_pull(struct sk_buff *skb, int len) |
| 1795 | { |
| 1796 | struct sk_buff *list; |
| 1797 | int skb_len = skb_headlen(skb); |
| 1798 | int rlen; |
| 1799 | |
| 1800 | if (len <= skb_len) { |
| 1801 | __skb_pull(skb, len); |
| 1802 | return 0; |
| 1803 | } |
| 1804 | len -= skb_len; |
| 1805 | __skb_pull(skb, skb_len); |
| 1806 | |
| 1807 | for (list = skb_shinfo(skb)->frag_list; list; list = list->next) { |
| 1808 | rlen = sctp_skb_pull(list, len); |
| 1809 | skb->len -= (len-rlen); |
| 1810 | skb->data_len -= (len-rlen); |
| 1811 | |
| 1812 | if (!rlen) |
| 1813 | return 0; |
| 1814 | |
| 1815 | len = rlen; |
| 1816 | } |
| 1817 | |
| 1818 | return len; |
| 1819 | } |
| 1820 | |
| 1821 | /* API 3.1.3 recvmsg() - UDP Style Syntax |
| 1822 | * |
| 1823 | * ssize_t recvmsg(int socket, struct msghdr *message, |
| 1824 | * int flags); |
| 1825 | * |
| 1826 | * socket - the socket descriptor of the endpoint. |
| 1827 | * message - pointer to the msghdr structure which contains a single |
| 1828 | * user message and possibly some ancillary data. |
| 1829 | * |
| 1830 | * See Section 5 for complete description of the data |
| 1831 | * structures. |
| 1832 | * |
| 1833 | * flags - flags sent or received with the user message, see Section |
| 1834 | * 5 for complete description of the flags. |
| 1835 | */ |
| 1836 | static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *); |
| 1837 | |
| 1838 | SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk, |
| 1839 | struct msghdr *msg, size_t len, int noblock, |
| 1840 | int flags, int *addr_len) |
| 1841 | { |
| 1842 | struct sctp_ulpevent *event = NULL; |
| 1843 | struct sctp_sock *sp = sctp_sk(sk); |
| 1844 | struct sk_buff *skb; |
| 1845 | int copied; |
| 1846 | int err = 0; |
| 1847 | int skb_len; |
| 1848 | |
| 1849 | SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: " |
| 1850 | "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg, |
| 1851 | "len", len, "knoblauch", noblock, |
| 1852 | "flags", flags, "addr_len", addr_len); |
| 1853 | |
| 1854 | sctp_lock_sock(sk); |
| 1855 | |
| 1856 | if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) { |
| 1857 | err = -ENOTCONN; |
| 1858 | goto out; |
| 1859 | } |
| 1860 | |
| 1861 | skb = sctp_skb_recv_datagram(sk, flags, noblock, &err); |
| 1862 | if (!skb) |
| 1863 | goto out; |
| 1864 | |
| 1865 | /* Get the total length of the skb including any skb's in the |
| 1866 | * frag_list. |
| 1867 | */ |
| 1868 | skb_len = skb->len; |
| 1869 | |
| 1870 | copied = skb_len; |
| 1871 | if (copied > len) |
| 1872 | copied = len; |
| 1873 | |
| 1874 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); |
| 1875 | |
| 1876 | event = sctp_skb2event(skb); |
| 1877 | |
| 1878 | if (err) |
| 1879 | goto out_free; |
| 1880 | |
| 1881 | sock_recv_timestamp(msg, sk, skb); |
| 1882 | if (sctp_ulpevent_is_notification(event)) { |
| 1883 | msg->msg_flags |= MSG_NOTIFICATION; |
| 1884 | sp->pf->event_msgname(event, msg->msg_name, addr_len); |
| 1885 | } else { |
| 1886 | sp->pf->skb_msgname(skb, msg->msg_name, addr_len); |
| 1887 | } |
| 1888 | |
| 1889 | /* Check if we allow SCTP_SNDRCVINFO. */ |
| 1890 | if (sp->subscribe.sctp_data_io_event) |
| 1891 | sctp_ulpevent_read_sndrcvinfo(event, msg); |
| 1892 | #if 0 |
| 1893 | /* FIXME: we should be calling IP/IPv6 layers. */ |
| 1894 | if (sk->sk_protinfo.af_inet.cmsg_flags) |
| 1895 | ip_cmsg_recv(msg, skb); |
| 1896 | #endif |
| 1897 | |
| 1898 | err = copied; |
| 1899 | |
| 1900 | /* If skb's length exceeds the user's buffer, update the skb and |
| 1901 | * push it back to the receive_queue so that the next call to |
| 1902 | * recvmsg() will return the remaining data. Don't set MSG_EOR. |
| 1903 | */ |
| 1904 | if (skb_len > copied) { |
| 1905 | msg->msg_flags &= ~MSG_EOR; |
| 1906 | if (flags & MSG_PEEK) |
| 1907 | goto out_free; |
| 1908 | sctp_skb_pull(skb, copied); |
| 1909 | skb_queue_head(&sk->sk_receive_queue, skb); |
| 1910 | |
| 1911 | /* When only partial message is copied to the user, increase |
| 1912 | * rwnd by that amount. If all the data in the skb is read, |
| 1913 | * rwnd is updated when the event is freed. |
| 1914 | */ |
| 1915 | sctp_assoc_rwnd_increase(event->asoc, copied); |
| 1916 | goto out; |
| 1917 | } else if ((event->msg_flags & MSG_NOTIFICATION) || |
| 1918 | (event->msg_flags & MSG_EOR)) |
| 1919 | msg->msg_flags |= MSG_EOR; |
| 1920 | else |
| 1921 | msg->msg_flags &= ~MSG_EOR; |
| 1922 | |
| 1923 | out_free: |
| 1924 | if (flags & MSG_PEEK) { |
| 1925 | /* Release the skb reference acquired after peeking the skb in |
| 1926 | * sctp_skb_recv_datagram(). |
| 1927 | */ |
| 1928 | kfree_skb(skb); |
| 1929 | } else { |
| 1930 | /* Free the event which includes releasing the reference to |
| 1931 | * the owner of the skb, freeing the skb and updating the |
| 1932 | * rwnd. |
| 1933 | */ |
| 1934 | sctp_ulpevent_free(event); |
| 1935 | } |
| 1936 | out: |
| 1937 | sctp_release_sock(sk); |
| 1938 | return err; |
| 1939 | } |
| 1940 | |
| 1941 | /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) |
| 1942 | * |
| 1943 | * This option is a on/off flag. If enabled no SCTP message |
| 1944 | * fragmentation will be performed. Instead if a message being sent |
| 1945 | * exceeds the current PMTU size, the message will NOT be sent and |
| 1946 | * instead a error will be indicated to the user. |
| 1947 | */ |
| 1948 | static int sctp_setsockopt_disable_fragments(struct sock *sk, |
| 1949 | char __user *optval, int optlen) |
| 1950 | { |
| 1951 | int val; |
| 1952 | |
| 1953 | if (optlen < sizeof(int)) |
| 1954 | return -EINVAL; |
| 1955 | |
| 1956 | if (get_user(val, (int __user *)optval)) |
| 1957 | return -EFAULT; |
| 1958 | |
| 1959 | sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1; |
| 1960 | |
| 1961 | return 0; |
| 1962 | } |
| 1963 | |
| 1964 | static int sctp_setsockopt_events(struct sock *sk, char __user *optval, |
| 1965 | int optlen) |
| 1966 | { |
| 1967 | if (optlen != sizeof(struct sctp_event_subscribe)) |
| 1968 | return -EINVAL; |
| 1969 | if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen)) |
| 1970 | return -EFAULT; |
| 1971 | return 0; |
| 1972 | } |
| 1973 | |
| 1974 | /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) |
| 1975 | * |
| 1976 | * This socket option is applicable to the UDP-style socket only. When |
| 1977 | * set it will cause associations that are idle for more than the |
| 1978 | * specified number of seconds to automatically close. An association |
| 1979 | * being idle is defined an association that has NOT sent or received |
| 1980 | * user data. The special value of '0' indicates that no automatic |
| 1981 | * close of any associations should be performed. The option expects an |
| 1982 | * integer defining the number of seconds of idle time before an |
| 1983 | * association is closed. |
| 1984 | */ |
| 1985 | static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval, |
| 1986 | int optlen) |
| 1987 | { |
| 1988 | struct sctp_sock *sp = sctp_sk(sk); |
| 1989 | |
| 1990 | /* Applicable to UDP-style socket only */ |
| 1991 | if (sctp_style(sk, TCP)) |
| 1992 | return -EOPNOTSUPP; |
| 1993 | if (optlen != sizeof(int)) |
| 1994 | return -EINVAL; |
| 1995 | if (copy_from_user(&sp->autoclose, optval, optlen)) |
| 1996 | return -EFAULT; |
| 1997 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1998 | return 0; |
| 1999 | } |
| 2000 | |
| 2001 | /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) |
| 2002 | * |
| 2003 | * Applications can enable or disable heartbeats for any peer address of |
| 2004 | * an association, modify an address's heartbeat interval, force a |
| 2005 | * heartbeat to be sent immediately, and adjust the address's maximum |
| 2006 | * number of retransmissions sent before an address is considered |
| 2007 | * unreachable. The following structure is used to access and modify an |
| 2008 | * address's parameters: |
| 2009 | * |
| 2010 | * struct sctp_paddrparams { |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2011 | * sctp_assoc_t spp_assoc_id; |
| 2012 | * struct sockaddr_storage spp_address; |
| 2013 | * uint32_t spp_hbinterval; |
| 2014 | * uint16_t spp_pathmaxrxt; |
| 2015 | * uint32_t spp_pathmtu; |
| 2016 | * uint32_t spp_sackdelay; |
| 2017 | * uint32_t spp_flags; |
| 2018 | * }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2019 | * |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2020 | * spp_assoc_id - (one-to-many style socket) This is filled in the |
| 2021 | * application, and identifies the association for |
| 2022 | * this query. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2023 | * spp_address - This specifies which address is of interest. |
| 2024 | * spp_hbinterval - This contains the value of the heartbeat interval, |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2025 | * in milliseconds. If a value of zero |
| 2026 | * is present in this field then no changes are to |
| 2027 | * be made to this parameter. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2028 | * spp_pathmaxrxt - This contains the maximum number of |
| 2029 | * retransmissions before this address shall be |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2030 | * considered unreachable. If a value of zero |
| 2031 | * is present in this field then no changes are to |
| 2032 | * be made to this parameter. |
| 2033 | * spp_pathmtu - When Path MTU discovery is disabled the value |
| 2034 | * specified here will be the "fixed" path mtu. |
| 2035 | * Note that if the spp_address field is empty |
| 2036 | * then all associations on this address will |
| 2037 | * have this fixed path mtu set upon them. |
| 2038 | * |
| 2039 | * spp_sackdelay - When delayed sack is enabled, this value specifies |
| 2040 | * the number of milliseconds that sacks will be delayed |
| 2041 | * for. This value will apply to all addresses of an |
| 2042 | * association if the spp_address field is empty. Note |
| 2043 | * also, that if delayed sack is enabled and this |
| 2044 | * value is set to 0, no change is made to the last |
| 2045 | * recorded delayed sack timer value. |
| 2046 | * |
| 2047 | * spp_flags - These flags are used to control various features |
| 2048 | * on an association. The flag field may contain |
| 2049 | * zero or more of the following options. |
| 2050 | * |
| 2051 | * SPP_HB_ENABLE - Enable heartbeats on the |
| 2052 | * specified address. Note that if the address |
| 2053 | * field is empty all addresses for the association |
| 2054 | * have heartbeats enabled upon them. |
| 2055 | * |
| 2056 | * SPP_HB_DISABLE - Disable heartbeats on the |
| 2057 | * speicifed address. Note that if the address |
| 2058 | * field is empty all addresses for the association |
| 2059 | * will have their heartbeats disabled. Note also |
| 2060 | * that SPP_HB_ENABLE and SPP_HB_DISABLE are |
| 2061 | * mutually exclusive, only one of these two should |
| 2062 | * be specified. Enabling both fields will have |
| 2063 | * undetermined results. |
| 2064 | * |
| 2065 | * SPP_HB_DEMAND - Request a user initiated heartbeat |
| 2066 | * to be made immediately. |
| 2067 | * |
Vlad Yasevich | bdf3092 | 2007-03-23 11:33:12 -0700 | [diff] [blame] | 2068 | * SPP_HB_TIME_IS_ZERO - Specify's that the time for |
| 2069 | * heartbeat delayis to be set to the value of 0 |
| 2070 | * milliseconds. |
| 2071 | * |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2072 | * SPP_PMTUD_ENABLE - This field will enable PMTU |
| 2073 | * discovery upon the specified address. Note that |
| 2074 | * if the address feild is empty then all addresses |
| 2075 | * on the association are effected. |
| 2076 | * |
| 2077 | * SPP_PMTUD_DISABLE - This field will disable PMTU |
| 2078 | * discovery upon the specified address. Note that |
| 2079 | * if the address feild is empty then all addresses |
| 2080 | * on the association are effected. Not also that |
| 2081 | * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually |
| 2082 | * exclusive. Enabling both will have undetermined |
| 2083 | * results. |
| 2084 | * |
| 2085 | * SPP_SACKDELAY_ENABLE - Setting this flag turns |
| 2086 | * on delayed sack. The time specified in spp_sackdelay |
| 2087 | * is used to specify the sack delay for this address. Note |
| 2088 | * that if spp_address is empty then all addresses will |
| 2089 | * enable delayed sack and take on the sack delay |
| 2090 | * value specified in spp_sackdelay. |
| 2091 | * SPP_SACKDELAY_DISABLE - Setting this flag turns |
| 2092 | * off delayed sack. If the spp_address field is blank then |
| 2093 | * delayed sack is disabled for the entire association. Note |
| 2094 | * also that this field is mutually exclusive to |
| 2095 | * SPP_SACKDELAY_ENABLE, setting both will have undefined |
| 2096 | * results. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2097 | */ |
Adrian Bunk | 1616436 | 2006-09-18 00:40:38 -0700 | [diff] [blame] | 2098 | static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params, |
| 2099 | struct sctp_transport *trans, |
| 2100 | struct sctp_association *asoc, |
| 2101 | struct sctp_sock *sp, |
| 2102 | int hb_change, |
| 2103 | int pmtud_change, |
| 2104 | int sackdelay_change) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2105 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2106 | int error; |
| 2107 | |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2108 | if (params->spp_flags & SPP_HB_DEMAND && trans) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2109 | error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans); |
| 2110 | if (error) |
| 2111 | return error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2112 | } |
| 2113 | |
Vlad Yasevich | bdf3092 | 2007-03-23 11:33:12 -0700 | [diff] [blame] | 2114 | /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of |
| 2115 | * this field is ignored. Note also that a value of zero indicates |
| 2116 | * the current setting should be left unchanged. |
| 2117 | */ |
| 2118 | if (params->spp_flags & SPP_HB_ENABLE) { |
| 2119 | |
| 2120 | /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is |
| 2121 | * set. This lets us use 0 value when this flag |
| 2122 | * is set. |
| 2123 | */ |
| 2124 | if (params->spp_flags & SPP_HB_TIME_IS_ZERO) |
| 2125 | params->spp_hbinterval = 0; |
| 2126 | |
| 2127 | if (params->spp_hbinterval || |
| 2128 | (params->spp_flags & SPP_HB_TIME_IS_ZERO)) { |
| 2129 | if (trans) { |
| 2130 | trans->hbinterval = |
| 2131 | msecs_to_jiffies(params->spp_hbinterval); |
| 2132 | } else if (asoc) { |
| 2133 | asoc->hbinterval = |
| 2134 | msecs_to_jiffies(params->spp_hbinterval); |
| 2135 | } else { |
| 2136 | sp->hbinterval = params->spp_hbinterval; |
| 2137 | } |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2138 | } |
| 2139 | } |
| 2140 | |
| 2141 | if (hb_change) { |
| 2142 | if (trans) { |
| 2143 | trans->param_flags = |
| 2144 | (trans->param_flags & ~SPP_HB) | hb_change; |
| 2145 | } else if (asoc) { |
| 2146 | asoc->param_flags = |
| 2147 | (asoc->param_flags & ~SPP_HB) | hb_change; |
| 2148 | } else { |
| 2149 | sp->param_flags = |
| 2150 | (sp->param_flags & ~SPP_HB) | hb_change; |
| 2151 | } |
| 2152 | } |
| 2153 | |
Vlad Yasevich | bdf3092 | 2007-03-23 11:33:12 -0700 | [diff] [blame] | 2154 | /* When Path MTU discovery is disabled the value specified here will |
| 2155 | * be the "fixed" path mtu (i.e. the value of the spp_flags field must |
| 2156 | * include the flag SPP_PMTUD_DISABLE for this field to have any |
| 2157 | * effect). |
| 2158 | */ |
| 2159 | if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) { |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2160 | if (trans) { |
| 2161 | trans->pathmtu = params->spp_pathmtu; |
| 2162 | sctp_assoc_sync_pmtu(asoc); |
| 2163 | } else if (asoc) { |
| 2164 | asoc->pathmtu = params->spp_pathmtu; |
| 2165 | sctp_frag_point(sp, params->spp_pathmtu); |
| 2166 | } else { |
| 2167 | sp->pathmtu = params->spp_pathmtu; |
| 2168 | } |
| 2169 | } |
| 2170 | |
| 2171 | if (pmtud_change) { |
| 2172 | if (trans) { |
| 2173 | int update = (trans->param_flags & SPP_PMTUD_DISABLE) && |
| 2174 | (params->spp_flags & SPP_PMTUD_ENABLE); |
| 2175 | trans->param_flags = |
| 2176 | (trans->param_flags & ~SPP_PMTUD) | pmtud_change; |
| 2177 | if (update) { |
| 2178 | sctp_transport_pmtu(trans); |
| 2179 | sctp_assoc_sync_pmtu(asoc); |
| 2180 | } |
| 2181 | } else if (asoc) { |
| 2182 | asoc->param_flags = |
| 2183 | (asoc->param_flags & ~SPP_PMTUD) | pmtud_change; |
| 2184 | } else { |
| 2185 | sp->param_flags = |
| 2186 | (sp->param_flags & ~SPP_PMTUD) | pmtud_change; |
| 2187 | } |
| 2188 | } |
| 2189 | |
Vlad Yasevich | bdf3092 | 2007-03-23 11:33:12 -0700 | [diff] [blame] | 2190 | /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the |
| 2191 | * value of this field is ignored. Note also that a value of zero |
| 2192 | * indicates the current setting should be left unchanged. |
| 2193 | */ |
| 2194 | if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) { |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2195 | if (trans) { |
| 2196 | trans->sackdelay = |
| 2197 | msecs_to_jiffies(params->spp_sackdelay); |
| 2198 | } else if (asoc) { |
| 2199 | asoc->sackdelay = |
| 2200 | msecs_to_jiffies(params->spp_sackdelay); |
| 2201 | } else { |
| 2202 | sp->sackdelay = params->spp_sackdelay; |
| 2203 | } |
| 2204 | } |
| 2205 | |
| 2206 | if (sackdelay_change) { |
| 2207 | if (trans) { |
| 2208 | trans->param_flags = |
| 2209 | (trans->param_flags & ~SPP_SACKDELAY) | |
| 2210 | sackdelay_change; |
| 2211 | } else if (asoc) { |
| 2212 | asoc->param_flags = |
| 2213 | (asoc->param_flags & ~SPP_SACKDELAY) | |
| 2214 | sackdelay_change; |
| 2215 | } else { |
| 2216 | sp->param_flags = |
| 2217 | (sp->param_flags & ~SPP_SACKDELAY) | |
| 2218 | sackdelay_change; |
| 2219 | } |
| 2220 | } |
| 2221 | |
Vlad Yasevich | bdf3092 | 2007-03-23 11:33:12 -0700 | [diff] [blame] | 2222 | /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value |
| 2223 | * of this field is ignored. Note also that a value of zero |
| 2224 | * indicates the current setting should be left unchanged. |
| 2225 | */ |
| 2226 | if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) { |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2227 | if (trans) { |
| 2228 | trans->pathmaxrxt = params->spp_pathmaxrxt; |
| 2229 | } else if (asoc) { |
| 2230 | asoc->pathmaxrxt = params->spp_pathmaxrxt; |
| 2231 | } else { |
| 2232 | sp->pathmaxrxt = params->spp_pathmaxrxt; |
| 2233 | } |
| 2234 | } |
| 2235 | |
| 2236 | return 0; |
| 2237 | } |
| 2238 | |
| 2239 | static int sctp_setsockopt_peer_addr_params(struct sock *sk, |
| 2240 | char __user *optval, int optlen) |
| 2241 | { |
| 2242 | struct sctp_paddrparams params; |
| 2243 | struct sctp_transport *trans = NULL; |
| 2244 | struct sctp_association *asoc = NULL; |
| 2245 | struct sctp_sock *sp = sctp_sk(sk); |
| 2246 | int error; |
| 2247 | int hb_change, pmtud_change, sackdelay_change; |
| 2248 | |
| 2249 | if (optlen != sizeof(struct sctp_paddrparams)) |
| 2250 | return - EINVAL; |
| 2251 | |
| 2252 | if (copy_from_user(¶ms, optval, optlen)) |
| 2253 | return -EFAULT; |
| 2254 | |
| 2255 | /* Validate flags and value parameters. */ |
| 2256 | hb_change = params.spp_flags & SPP_HB; |
| 2257 | pmtud_change = params.spp_flags & SPP_PMTUD; |
| 2258 | sackdelay_change = params.spp_flags & SPP_SACKDELAY; |
| 2259 | |
| 2260 | if (hb_change == SPP_HB || |
| 2261 | pmtud_change == SPP_PMTUD || |
| 2262 | sackdelay_change == SPP_SACKDELAY || |
| 2263 | params.spp_sackdelay > 500 || |
| 2264 | (params.spp_pathmtu |
| 2265 | && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT)) |
| 2266 | return -EINVAL; |
| 2267 | |
| 2268 | /* If an address other than INADDR_ANY is specified, and |
| 2269 | * no transport is found, then the request is invalid. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2270 | */ |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 2271 | if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) { |
| 2272 | trans = sctp_addr_id2transport(sk, ¶ms.spp_address, |
| 2273 | params.spp_assoc_id); |
| 2274 | if (!trans) |
| 2275 | return -EINVAL; |
| 2276 | } |
| 2277 | |
| 2278 | /* Get association, if assoc_id != 0 and the socket is a one |
| 2279 | * to many style socket, and an association was not found, then |
| 2280 | * the id was invalid. |
| 2281 | */ |
| 2282 | asoc = sctp_id2assoc(sk, params.spp_assoc_id); |
| 2283 | if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) |
| 2284 | return -EINVAL; |
| 2285 | |
| 2286 | /* Heartbeat demand can only be sent on a transport or |
| 2287 | * association, but not a socket. |
| 2288 | */ |
| 2289 | if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc) |
| 2290 | return -EINVAL; |
| 2291 | |
| 2292 | /* Process parameters. */ |
| 2293 | error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp, |
| 2294 | hb_change, pmtud_change, |
| 2295 | sackdelay_change); |
| 2296 | |
| 2297 | if (error) |
| 2298 | return error; |
| 2299 | |
| 2300 | /* If changes are for association, also apply parameters to each |
| 2301 | * transport. |
| 2302 | */ |
| 2303 | if (!trans && asoc) { |
| 2304 | struct list_head *pos; |
| 2305 | |
| 2306 | list_for_each(pos, &asoc->peer.transport_addr_list) { |
| 2307 | trans = list_entry(pos, struct sctp_transport, |
| 2308 | transports); |
| 2309 | sctp_apply_peer_addr_params(¶ms, trans, asoc, sp, |
| 2310 | hb_change, pmtud_change, |
| 2311 | sackdelay_change); |
| 2312 | } |
| 2313 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2314 | |
| 2315 | return 0; |
| 2316 | } |
| 2317 | |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 2318 | /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME) |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2319 | * |
| 2320 | * This options will get or set the delayed ack timer. The time is set |
| 2321 | * in milliseconds. If the assoc_id is 0, then this sets or gets the |
| 2322 | * endpoints default delayed ack timer value. If the assoc_id field is |
| 2323 | * non-zero, then the set or get effects the specified association. |
| 2324 | * |
| 2325 | * struct sctp_assoc_value { |
| 2326 | * sctp_assoc_t assoc_id; |
| 2327 | * uint32_t assoc_value; |
| 2328 | * }; |
| 2329 | * |
| 2330 | * assoc_id - This parameter, indicates which association the |
| 2331 | * user is preforming an action upon. Note that if |
| 2332 | * this field's value is zero then the endpoints |
| 2333 | * default value is changed (effecting future |
| 2334 | * associations only). |
| 2335 | * |
| 2336 | * assoc_value - This parameter contains the number of milliseconds |
| 2337 | * that the user is requesting the delayed ACK timer |
| 2338 | * be set to. Note that this value is defined in |
| 2339 | * the standard to be between 200 and 500 milliseconds. |
| 2340 | * |
| 2341 | * Note: a value of zero will leave the value alone, |
| 2342 | * but disable SACK delay. A non-zero value will also |
| 2343 | * enable SACK delay. |
| 2344 | */ |
| 2345 | |
| 2346 | static int sctp_setsockopt_delayed_ack_time(struct sock *sk, |
| 2347 | char __user *optval, int optlen) |
| 2348 | { |
| 2349 | struct sctp_assoc_value params; |
| 2350 | struct sctp_transport *trans = NULL; |
| 2351 | struct sctp_association *asoc = NULL; |
| 2352 | struct sctp_sock *sp = sctp_sk(sk); |
| 2353 | |
| 2354 | if (optlen != sizeof(struct sctp_assoc_value)) |
| 2355 | return - EINVAL; |
| 2356 | |
| 2357 | if (copy_from_user(¶ms, optval, optlen)) |
| 2358 | return -EFAULT; |
| 2359 | |
| 2360 | /* Validate value parameter. */ |
| 2361 | if (params.assoc_value > 500) |
| 2362 | return -EINVAL; |
| 2363 | |
| 2364 | /* Get association, if assoc_id != 0 and the socket is a one |
| 2365 | * to many style socket, and an association was not found, then |
| 2366 | * the id was invalid. |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2367 | */ |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2368 | asoc = sctp_id2assoc(sk, params.assoc_id); |
| 2369 | if (!asoc && params.assoc_id && sctp_style(sk, UDP)) |
| 2370 | return -EINVAL; |
| 2371 | |
| 2372 | if (params.assoc_value) { |
| 2373 | if (asoc) { |
| 2374 | asoc->sackdelay = |
| 2375 | msecs_to_jiffies(params.assoc_value); |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2376 | asoc->param_flags = |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2377 | (asoc->param_flags & ~SPP_SACKDELAY) | |
| 2378 | SPP_SACKDELAY_ENABLE; |
| 2379 | } else { |
| 2380 | sp->sackdelay = params.assoc_value; |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2381 | sp->param_flags = |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2382 | (sp->param_flags & ~SPP_SACKDELAY) | |
| 2383 | SPP_SACKDELAY_ENABLE; |
| 2384 | } |
| 2385 | } else { |
| 2386 | if (asoc) { |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2387 | asoc->param_flags = |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2388 | (asoc->param_flags & ~SPP_SACKDELAY) | |
| 2389 | SPP_SACKDELAY_DISABLE; |
| 2390 | } else { |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2391 | sp->param_flags = |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2392 | (sp->param_flags & ~SPP_SACKDELAY) | |
| 2393 | SPP_SACKDELAY_DISABLE; |
| 2394 | } |
| 2395 | } |
| 2396 | |
| 2397 | /* If change is for association, also apply to each transport. */ |
| 2398 | if (asoc) { |
| 2399 | struct list_head *pos; |
| 2400 | |
| 2401 | list_for_each(pos, &asoc->peer.transport_addr_list) { |
| 2402 | trans = list_entry(pos, struct sctp_transport, |
| 2403 | transports); |
| 2404 | if (params.assoc_value) { |
| 2405 | trans->sackdelay = |
| 2406 | msecs_to_jiffies(params.assoc_value); |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2407 | trans->param_flags = |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2408 | (trans->param_flags & ~SPP_SACKDELAY) | |
| 2409 | SPP_SACKDELAY_ENABLE; |
| 2410 | } else { |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2411 | trans->param_flags = |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2412 | (trans->param_flags & ~SPP_SACKDELAY) | |
| 2413 | SPP_SACKDELAY_DISABLE; |
| 2414 | } |
| 2415 | } |
| 2416 | } |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2417 | |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 2418 | return 0; |
| 2419 | } |
| 2420 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2421 | /* 7.1.3 Initialization Parameters (SCTP_INITMSG) |
| 2422 | * |
| 2423 | * Applications can specify protocol parameters for the default association |
| 2424 | * initialization. The option name argument to setsockopt() and getsockopt() |
| 2425 | * is SCTP_INITMSG. |
| 2426 | * |
| 2427 | * Setting initialization parameters is effective only on an unconnected |
| 2428 | * socket (for UDP-style sockets only future associations are effected |
| 2429 | * by the change). With TCP-style sockets, this option is inherited by |
| 2430 | * sockets derived from a listener socket. |
| 2431 | */ |
| 2432 | static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen) |
| 2433 | { |
| 2434 | struct sctp_initmsg sinit; |
| 2435 | struct sctp_sock *sp = sctp_sk(sk); |
| 2436 | |
| 2437 | if (optlen != sizeof(struct sctp_initmsg)) |
| 2438 | return -EINVAL; |
| 2439 | if (copy_from_user(&sinit, optval, optlen)) |
| 2440 | return -EFAULT; |
| 2441 | |
| 2442 | if (sinit.sinit_num_ostreams) |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2443 | sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2444 | if (sinit.sinit_max_instreams) |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2445 | sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2446 | if (sinit.sinit_max_attempts) |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2447 | sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2448 | if (sinit.sinit_max_init_timeo) |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2449 | sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2450 | |
| 2451 | return 0; |
| 2452 | } |
| 2453 | |
| 2454 | /* |
| 2455 | * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM) |
| 2456 | * |
| 2457 | * Applications that wish to use the sendto() system call may wish to |
| 2458 | * specify a default set of parameters that would normally be supplied |
| 2459 | * through the inclusion of ancillary data. This socket option allows |
| 2460 | * such an application to set the default sctp_sndrcvinfo structure. |
| 2461 | * The application that wishes to use this socket option simply passes |
| 2462 | * in to this call the sctp_sndrcvinfo structure defined in Section |
| 2463 | * 5.2.2) The input parameters accepted by this call include |
| 2464 | * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context, |
| 2465 | * sinfo_timetolive. The user must provide the sinfo_assoc_id field in |
| 2466 | * to this call if the caller is using the UDP model. |
| 2467 | */ |
| 2468 | static int sctp_setsockopt_default_send_param(struct sock *sk, |
| 2469 | char __user *optval, int optlen) |
| 2470 | { |
| 2471 | struct sctp_sndrcvinfo info; |
| 2472 | struct sctp_association *asoc; |
| 2473 | struct sctp_sock *sp = sctp_sk(sk); |
| 2474 | |
| 2475 | if (optlen != sizeof(struct sctp_sndrcvinfo)) |
| 2476 | return -EINVAL; |
| 2477 | if (copy_from_user(&info, optval, optlen)) |
| 2478 | return -EFAULT; |
| 2479 | |
| 2480 | asoc = sctp_id2assoc(sk, info.sinfo_assoc_id); |
| 2481 | if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP)) |
| 2482 | return -EINVAL; |
| 2483 | |
| 2484 | if (asoc) { |
| 2485 | asoc->default_stream = info.sinfo_stream; |
| 2486 | asoc->default_flags = info.sinfo_flags; |
| 2487 | asoc->default_ppid = info.sinfo_ppid; |
| 2488 | asoc->default_context = info.sinfo_context; |
| 2489 | asoc->default_timetolive = info.sinfo_timetolive; |
| 2490 | } else { |
| 2491 | sp->default_stream = info.sinfo_stream; |
| 2492 | sp->default_flags = info.sinfo_flags; |
| 2493 | sp->default_ppid = info.sinfo_ppid; |
| 2494 | sp->default_context = info.sinfo_context; |
| 2495 | sp->default_timetolive = info.sinfo_timetolive; |
| 2496 | } |
| 2497 | |
| 2498 | return 0; |
| 2499 | } |
| 2500 | |
| 2501 | /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR) |
| 2502 | * |
| 2503 | * Requests that the local SCTP stack use the enclosed peer address as |
| 2504 | * the association primary. The enclosed address must be one of the |
| 2505 | * association peer's addresses. |
| 2506 | */ |
| 2507 | static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval, |
| 2508 | int optlen) |
| 2509 | { |
| 2510 | struct sctp_prim prim; |
| 2511 | struct sctp_transport *trans; |
| 2512 | |
| 2513 | if (optlen != sizeof(struct sctp_prim)) |
| 2514 | return -EINVAL; |
| 2515 | |
| 2516 | if (copy_from_user(&prim, optval, sizeof(struct sctp_prim))) |
| 2517 | return -EFAULT; |
| 2518 | |
| 2519 | trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id); |
| 2520 | if (!trans) |
| 2521 | return -EINVAL; |
| 2522 | |
| 2523 | sctp_assoc_set_primary(trans->asoc, trans); |
| 2524 | |
| 2525 | return 0; |
| 2526 | } |
| 2527 | |
| 2528 | /* |
| 2529 | * 7.1.5 SCTP_NODELAY |
| 2530 | * |
| 2531 | * Turn on/off any Nagle-like algorithm. This means that packets are |
| 2532 | * generally sent as soon as possible and no unnecessary delays are |
| 2533 | * introduced, at the cost of more packets in the network. Expects an |
| 2534 | * integer boolean flag. |
| 2535 | */ |
| 2536 | static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval, |
| 2537 | int optlen) |
| 2538 | { |
| 2539 | int val; |
| 2540 | |
| 2541 | if (optlen < sizeof(int)) |
| 2542 | return -EINVAL; |
| 2543 | if (get_user(val, (int __user *)optval)) |
| 2544 | return -EFAULT; |
| 2545 | |
| 2546 | sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1; |
| 2547 | return 0; |
| 2548 | } |
| 2549 | |
| 2550 | /* |
| 2551 | * |
| 2552 | * 7.1.1 SCTP_RTOINFO |
| 2553 | * |
| 2554 | * The protocol parameters used to initialize and bound retransmission |
| 2555 | * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access |
| 2556 | * and modify these parameters. |
| 2557 | * All parameters are time values, in milliseconds. A value of 0, when |
| 2558 | * modifying the parameters, indicates that the current value should not |
| 2559 | * be changed. |
| 2560 | * |
| 2561 | */ |
| 2562 | static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) { |
| 2563 | struct sctp_rtoinfo rtoinfo; |
| 2564 | struct sctp_association *asoc; |
| 2565 | |
| 2566 | if (optlen != sizeof (struct sctp_rtoinfo)) |
| 2567 | return -EINVAL; |
| 2568 | |
| 2569 | if (copy_from_user(&rtoinfo, optval, optlen)) |
| 2570 | return -EFAULT; |
| 2571 | |
| 2572 | asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id); |
| 2573 | |
| 2574 | /* Set the values to the specific association */ |
| 2575 | if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP)) |
| 2576 | return -EINVAL; |
| 2577 | |
| 2578 | if (asoc) { |
| 2579 | if (rtoinfo.srto_initial != 0) |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2580 | asoc->rto_initial = |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2581 | msecs_to_jiffies(rtoinfo.srto_initial); |
| 2582 | if (rtoinfo.srto_max != 0) |
| 2583 | asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max); |
| 2584 | if (rtoinfo.srto_min != 0) |
| 2585 | asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min); |
| 2586 | } else { |
| 2587 | /* If there is no association or the association-id = 0 |
| 2588 | * set the values to the endpoint. |
| 2589 | */ |
| 2590 | struct sctp_sock *sp = sctp_sk(sk); |
| 2591 | |
| 2592 | if (rtoinfo.srto_initial != 0) |
| 2593 | sp->rtoinfo.srto_initial = rtoinfo.srto_initial; |
| 2594 | if (rtoinfo.srto_max != 0) |
| 2595 | sp->rtoinfo.srto_max = rtoinfo.srto_max; |
| 2596 | if (rtoinfo.srto_min != 0) |
| 2597 | sp->rtoinfo.srto_min = rtoinfo.srto_min; |
| 2598 | } |
| 2599 | |
| 2600 | return 0; |
| 2601 | } |
| 2602 | |
| 2603 | /* |
| 2604 | * |
| 2605 | * 7.1.2 SCTP_ASSOCINFO |
| 2606 | * |
Michael Opdenacker | 59c5159 | 2007-05-09 08:57:56 +0200 | [diff] [blame] | 2607 | * This option is used to tune the maximum retransmission attempts |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2608 | * of the association. |
| 2609 | * Returns an error if the new association retransmission value is |
| 2610 | * greater than the sum of the retransmission value of the peer. |
| 2611 | * See [SCTP] for more information. |
| 2612 | * |
| 2613 | */ |
| 2614 | static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen) |
| 2615 | { |
| 2616 | |
| 2617 | struct sctp_assocparams assocparams; |
| 2618 | struct sctp_association *asoc; |
| 2619 | |
| 2620 | if (optlen != sizeof(struct sctp_assocparams)) |
| 2621 | return -EINVAL; |
| 2622 | if (copy_from_user(&assocparams, optval, optlen)) |
| 2623 | return -EFAULT; |
| 2624 | |
| 2625 | asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id); |
| 2626 | |
| 2627 | if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP)) |
| 2628 | return -EINVAL; |
| 2629 | |
| 2630 | /* Set the values to the specific association */ |
| 2631 | if (asoc) { |
Vlad Yasevich | 402d68c | 2006-06-17 22:54:51 -0700 | [diff] [blame] | 2632 | if (assocparams.sasoc_asocmaxrxt != 0) { |
| 2633 | __u32 path_sum = 0; |
| 2634 | int paths = 0; |
| 2635 | struct list_head *pos; |
| 2636 | struct sctp_transport *peer_addr; |
| 2637 | |
| 2638 | list_for_each(pos, &asoc->peer.transport_addr_list) { |
| 2639 | peer_addr = list_entry(pos, |
| 2640 | struct sctp_transport, |
| 2641 | transports); |
| 2642 | path_sum += peer_addr->pathmaxrxt; |
| 2643 | paths++; |
| 2644 | } |
| 2645 | |
| 2646 | /* Only validate asocmaxrxt if we have more then |
| 2647 | * one path/transport. We do this because path |
| 2648 | * retransmissions are only counted when we have more |
| 2649 | * then one path. |
| 2650 | */ |
| 2651 | if (paths > 1 && |
| 2652 | assocparams.sasoc_asocmaxrxt > path_sum) |
| 2653 | return -EINVAL; |
| 2654 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2655 | asoc->max_retrans = assocparams.sasoc_asocmaxrxt; |
Vlad Yasevich | 402d68c | 2006-06-17 22:54:51 -0700 | [diff] [blame] | 2656 | } |
| 2657 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2658 | if (assocparams.sasoc_cookie_life != 0) { |
| 2659 | asoc->cookie_life.tv_sec = |
| 2660 | assocparams.sasoc_cookie_life / 1000; |
| 2661 | asoc->cookie_life.tv_usec = |
| 2662 | (assocparams.sasoc_cookie_life % 1000) |
| 2663 | * 1000; |
| 2664 | } |
| 2665 | } else { |
| 2666 | /* Set the values to the endpoint */ |
| 2667 | struct sctp_sock *sp = sctp_sk(sk); |
| 2668 | |
| 2669 | if (assocparams.sasoc_asocmaxrxt != 0) |
| 2670 | sp->assocparams.sasoc_asocmaxrxt = |
| 2671 | assocparams.sasoc_asocmaxrxt; |
| 2672 | if (assocparams.sasoc_cookie_life != 0) |
| 2673 | sp->assocparams.sasoc_cookie_life = |
| 2674 | assocparams.sasoc_cookie_life; |
| 2675 | } |
| 2676 | return 0; |
| 2677 | } |
| 2678 | |
| 2679 | /* |
| 2680 | * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR) |
| 2681 | * |
| 2682 | * This socket option is a boolean flag which turns on or off mapped V4 |
| 2683 | * addresses. If this option is turned on and the socket is type |
| 2684 | * PF_INET6, then IPv4 addresses will be mapped to V6 representation. |
| 2685 | * If this option is turned off, then no mapping will be done of V4 |
| 2686 | * addresses and a user will receive both PF_INET6 and PF_INET type |
| 2687 | * addresses on the socket. |
| 2688 | */ |
| 2689 | static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen) |
| 2690 | { |
| 2691 | int val; |
| 2692 | struct sctp_sock *sp = sctp_sk(sk); |
| 2693 | |
| 2694 | if (optlen < sizeof(int)) |
| 2695 | return -EINVAL; |
| 2696 | if (get_user(val, (int __user *)optval)) |
| 2697 | return -EFAULT; |
| 2698 | if (val) |
| 2699 | sp->v4mapped = 1; |
| 2700 | else |
| 2701 | sp->v4mapped = 0; |
| 2702 | |
| 2703 | return 0; |
| 2704 | } |
| 2705 | |
| 2706 | /* |
| 2707 | * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG) |
| 2708 | * |
| 2709 | * This socket option specifies the maximum size to put in any outgoing |
| 2710 | * SCTP chunk. If a message is larger than this size it will be |
| 2711 | * fragmented by SCTP into the specified size. Note that the underlying |
| 2712 | * SCTP implementation may fragment into smaller sized chunks when the |
| 2713 | * PMTU of the underlying association is smaller than the value set by |
| 2714 | * the user. |
| 2715 | */ |
| 2716 | static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen) |
| 2717 | { |
| 2718 | struct sctp_association *asoc; |
| 2719 | struct list_head *pos; |
| 2720 | struct sctp_sock *sp = sctp_sk(sk); |
| 2721 | int val; |
| 2722 | |
| 2723 | if (optlen < sizeof(int)) |
| 2724 | return -EINVAL; |
| 2725 | if (get_user(val, (int __user *)optval)) |
| 2726 | return -EFAULT; |
Ivan Skytte Jorgensen | 96a3399 | 2005-10-28 15:36:12 -0700 | [diff] [blame] | 2727 | if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2728 | return -EINVAL; |
| 2729 | sp->user_frag = val; |
| 2730 | |
Ivan Skytte Jorgensen | 96a3399 | 2005-10-28 15:36:12 -0700 | [diff] [blame] | 2731 | /* Update the frag_point of the existing associations. */ |
| 2732 | list_for_each(pos, &(sp->ep->asocs)) { |
| 2733 | asoc = list_entry(pos, struct sctp_association, asocs); |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2734 | asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2735 | } |
| 2736 | |
| 2737 | return 0; |
| 2738 | } |
| 2739 | |
| 2740 | |
| 2741 | /* |
| 2742 | * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR) |
| 2743 | * |
| 2744 | * Requests that the peer mark the enclosed address as the association |
| 2745 | * primary. The enclosed address must be one of the association's |
| 2746 | * locally bound addresses. The following structure is used to make a |
| 2747 | * set primary request: |
| 2748 | */ |
| 2749 | static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval, |
| 2750 | int optlen) |
| 2751 | { |
| 2752 | struct sctp_sock *sp; |
| 2753 | struct sctp_endpoint *ep; |
| 2754 | struct sctp_association *asoc = NULL; |
| 2755 | struct sctp_setpeerprim prim; |
| 2756 | struct sctp_chunk *chunk; |
| 2757 | int err; |
| 2758 | |
| 2759 | sp = sctp_sk(sk); |
| 2760 | ep = sp->ep; |
| 2761 | |
| 2762 | if (!sctp_addip_enable) |
| 2763 | return -EPERM; |
| 2764 | |
| 2765 | if (optlen != sizeof(struct sctp_setpeerprim)) |
| 2766 | return -EINVAL; |
| 2767 | |
| 2768 | if (copy_from_user(&prim, optval, optlen)) |
| 2769 | return -EFAULT; |
| 2770 | |
| 2771 | asoc = sctp_id2assoc(sk, prim.sspp_assoc_id); |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 2772 | if (!asoc) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2773 | return -EINVAL; |
| 2774 | |
| 2775 | if (!asoc->peer.asconf_capable) |
| 2776 | return -EPERM; |
| 2777 | |
| 2778 | if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY) |
| 2779 | return -EPERM; |
| 2780 | |
| 2781 | if (!sctp_state(asoc, ESTABLISHED)) |
| 2782 | return -ENOTCONN; |
| 2783 | |
| 2784 | if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr)) |
| 2785 | return -EADDRNOTAVAIL; |
| 2786 | |
| 2787 | /* Create an ASCONF chunk with SET_PRIMARY parameter */ |
| 2788 | chunk = sctp_make_asconf_set_prim(asoc, |
| 2789 | (union sctp_addr *)&prim.sspp_addr); |
| 2790 | if (!chunk) |
| 2791 | return -ENOMEM; |
| 2792 | |
| 2793 | err = sctp_send_asconf(asoc, chunk); |
| 2794 | |
| 2795 | SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n"); |
| 2796 | |
| 2797 | return err; |
| 2798 | } |
| 2799 | |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 2800 | static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2801 | int optlen) |
| 2802 | { |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 2803 | struct sctp_setadaptation adaptation; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2804 | |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 2805 | if (optlen != sizeof(struct sctp_setadaptation)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2806 | return -EINVAL; |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 2807 | if (copy_from_user(&adaptation, optval, optlen)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2808 | return -EFAULT; |
| 2809 | |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 2810 | sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2811 | |
| 2812 | return 0; |
| 2813 | } |
| 2814 | |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 2815 | /* |
| 2816 | * 7.1.29. Set or Get the default context (SCTP_CONTEXT) |
| 2817 | * |
| 2818 | * The context field in the sctp_sndrcvinfo structure is normally only |
| 2819 | * used when a failed message is retrieved holding the value that was |
| 2820 | * sent down on the actual send call. This option allows the setting of |
| 2821 | * a default context on an association basis that will be received on |
| 2822 | * reading messages from the peer. This is especially helpful in the |
| 2823 | * one-2-many model for an application to keep some reference to an |
| 2824 | * internal state machine that is processing messages on the |
| 2825 | * association. Note that the setting of this value only effects |
| 2826 | * received messages from the peer and does not effect the value that is |
| 2827 | * saved with outbound messages. |
| 2828 | */ |
| 2829 | static int sctp_setsockopt_context(struct sock *sk, char __user *optval, |
| 2830 | int optlen) |
| 2831 | { |
| 2832 | struct sctp_assoc_value params; |
| 2833 | struct sctp_sock *sp; |
| 2834 | struct sctp_association *asoc; |
| 2835 | |
| 2836 | if (optlen != sizeof(struct sctp_assoc_value)) |
| 2837 | return -EINVAL; |
| 2838 | if (copy_from_user(¶ms, optval, optlen)) |
| 2839 | return -EFAULT; |
| 2840 | |
| 2841 | sp = sctp_sk(sk); |
| 2842 | |
| 2843 | if (params.assoc_id != 0) { |
| 2844 | asoc = sctp_id2assoc(sk, params.assoc_id); |
| 2845 | if (!asoc) |
| 2846 | return -EINVAL; |
| 2847 | asoc->default_rcv_context = params.assoc_value; |
| 2848 | } else { |
| 2849 | sp->default_rcv_context = params.assoc_value; |
| 2850 | } |
| 2851 | |
| 2852 | return 0; |
| 2853 | } |
| 2854 | |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 2855 | /* |
| 2856 | * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE) |
| 2857 | * |
| 2858 | * This options will at a minimum specify if the implementation is doing |
| 2859 | * fragmented interleave. Fragmented interleave, for a one to many |
| 2860 | * socket, is when subsequent calls to receive a message may return |
| 2861 | * parts of messages from different associations. Some implementations |
| 2862 | * may allow you to turn this value on or off. If so, when turned off, |
| 2863 | * no fragment interleave will occur (which will cause a head of line |
| 2864 | * blocking amongst multiple associations sharing the same one to many |
| 2865 | * socket). When this option is turned on, then each receive call may |
| 2866 | * come from a different association (thus the user must receive data |
| 2867 | * with the extended calls (e.g. sctp_recvmsg) to keep track of which |
| 2868 | * association each receive belongs to. |
| 2869 | * |
| 2870 | * This option takes a boolean value. A non-zero value indicates that |
| 2871 | * fragmented interleave is on. A value of zero indicates that |
| 2872 | * fragmented interleave is off. |
| 2873 | * |
| 2874 | * Note that it is important that an implementation that allows this |
| 2875 | * option to be turned on, have it off by default. Otherwise an unaware |
| 2876 | * application using the one to many model may become confused and act |
| 2877 | * incorrectly. |
| 2878 | */ |
| 2879 | static int sctp_setsockopt_fragment_interleave(struct sock *sk, |
| 2880 | char __user *optval, |
| 2881 | int optlen) |
| 2882 | { |
| 2883 | int val; |
| 2884 | |
| 2885 | if (optlen != sizeof(int)) |
| 2886 | return -EINVAL; |
| 2887 | if (get_user(val, (int __user *)optval)) |
| 2888 | return -EFAULT; |
| 2889 | |
| 2890 | sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1; |
| 2891 | |
| 2892 | return 0; |
| 2893 | } |
| 2894 | |
Vlad Yasevich | d49d91d | 2007-03-23 11:32:00 -0700 | [diff] [blame] | 2895 | /* |
| 2896 | * 7.1.25. Set or Get the sctp partial delivery point |
| 2897 | * (SCTP_PARTIAL_DELIVERY_POINT) |
| 2898 | * This option will set or get the SCTP partial delivery point. This |
| 2899 | * point is the size of a message where the partial delivery API will be |
| 2900 | * invoked to help free up rwnd space for the peer. Setting this to a |
| 2901 | * lower value will cause partial delivery's to happen more often. The |
| 2902 | * calls argument is an integer that sets or gets the partial delivery |
| 2903 | * point. |
| 2904 | */ |
| 2905 | static int sctp_setsockopt_partial_delivery_point(struct sock *sk, |
| 2906 | char __user *optval, |
| 2907 | int optlen) |
| 2908 | { |
| 2909 | u32 val; |
| 2910 | |
| 2911 | if (optlen != sizeof(u32)) |
| 2912 | return -EINVAL; |
| 2913 | if (get_user(val, (int __user *)optval)) |
| 2914 | return -EFAULT; |
| 2915 | |
| 2916 | sctp_sk(sk)->pd_point = val; |
| 2917 | |
| 2918 | return 0; /* is this the right error code? */ |
| 2919 | } |
| 2920 | |
Vlad Yasevich | 7033157 | 2007-03-23 11:34:36 -0700 | [diff] [blame] | 2921 | /* |
| 2922 | * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST) |
| 2923 | * |
| 2924 | * This option will allow a user to change the maximum burst of packets |
| 2925 | * that can be emitted by this association. Note that the default value |
| 2926 | * is 4, and some implementations may restrict this setting so that it |
| 2927 | * can only be lowered. |
| 2928 | * |
| 2929 | * NOTE: This text doesn't seem right. Do this on a socket basis with |
| 2930 | * future associations inheriting the socket value. |
| 2931 | */ |
| 2932 | static int sctp_setsockopt_maxburst(struct sock *sk, |
| 2933 | char __user *optval, |
| 2934 | int optlen) |
| 2935 | { |
| 2936 | int val; |
| 2937 | |
| 2938 | if (optlen != sizeof(int)) |
| 2939 | return -EINVAL; |
| 2940 | if (get_user(val, (int __user *)optval)) |
| 2941 | return -EFAULT; |
| 2942 | |
| 2943 | if (val < 0) |
| 2944 | return -EINVAL; |
| 2945 | |
| 2946 | sctp_sk(sk)->max_burst = val; |
| 2947 | |
| 2948 | return 0; |
| 2949 | } |
| 2950 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2951 | /* API 6.2 setsockopt(), getsockopt() |
| 2952 | * |
| 2953 | * Applications use setsockopt() and getsockopt() to set or retrieve |
| 2954 | * socket options. Socket options are used to change the default |
| 2955 | * behavior of sockets calls. They are described in Section 7. |
| 2956 | * |
| 2957 | * The syntax is: |
| 2958 | * |
| 2959 | * ret = getsockopt(int sd, int level, int optname, void __user *optval, |
| 2960 | * int __user *optlen); |
| 2961 | * ret = setsockopt(int sd, int level, int optname, const void __user *optval, |
| 2962 | * int optlen); |
| 2963 | * |
| 2964 | * sd - the socket descript. |
| 2965 | * level - set to IPPROTO_SCTP for all SCTP options. |
| 2966 | * optname - the option name. |
| 2967 | * optval - the buffer to store the value of the option. |
| 2968 | * optlen - the size of the buffer. |
| 2969 | */ |
| 2970 | SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname, |
| 2971 | char __user *optval, int optlen) |
| 2972 | { |
| 2973 | int retval = 0; |
| 2974 | |
| 2975 | SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n", |
| 2976 | sk, optname); |
| 2977 | |
| 2978 | /* I can hardly begin to describe how wrong this is. This is |
| 2979 | * so broken as to be worse than useless. The API draft |
| 2980 | * REALLY is NOT helpful here... I am not convinced that the |
| 2981 | * semantics of setsockopt() with a level OTHER THAN SOL_SCTP |
| 2982 | * are at all well-founded. |
| 2983 | */ |
| 2984 | if (level != SOL_SCTP) { |
| 2985 | struct sctp_af *af = sctp_sk(sk)->pf->af; |
| 2986 | retval = af->setsockopt(sk, level, optname, optval, optlen); |
| 2987 | goto out_nounlock; |
| 2988 | } |
| 2989 | |
| 2990 | sctp_lock_sock(sk); |
| 2991 | |
| 2992 | switch (optname) { |
| 2993 | case SCTP_SOCKOPT_BINDX_ADD: |
| 2994 | /* 'optlen' is the size of the addresses buffer. */ |
| 2995 | retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, |
| 2996 | optlen, SCTP_BINDX_ADD_ADDR); |
| 2997 | break; |
| 2998 | |
| 2999 | case SCTP_SOCKOPT_BINDX_REM: |
| 3000 | /* 'optlen' is the size of the addresses buffer. */ |
| 3001 | retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, |
| 3002 | optlen, SCTP_BINDX_REM_ADDR); |
| 3003 | break; |
| 3004 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3005 | case SCTP_SOCKOPT_CONNECTX: |
| 3006 | /* 'optlen' is the size of the addresses buffer. */ |
| 3007 | retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval, |
| 3008 | optlen); |
| 3009 | break; |
| 3010 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3011 | case SCTP_DISABLE_FRAGMENTS: |
| 3012 | retval = sctp_setsockopt_disable_fragments(sk, optval, optlen); |
| 3013 | break; |
| 3014 | |
| 3015 | case SCTP_EVENTS: |
| 3016 | retval = sctp_setsockopt_events(sk, optval, optlen); |
| 3017 | break; |
| 3018 | |
| 3019 | case SCTP_AUTOCLOSE: |
| 3020 | retval = sctp_setsockopt_autoclose(sk, optval, optlen); |
| 3021 | break; |
| 3022 | |
| 3023 | case SCTP_PEER_ADDR_PARAMS: |
| 3024 | retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen); |
| 3025 | break; |
| 3026 | |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 3027 | case SCTP_DELAYED_ACK_TIME: |
| 3028 | retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen); |
| 3029 | break; |
Vlad Yasevich | d49d91d | 2007-03-23 11:32:00 -0700 | [diff] [blame] | 3030 | case SCTP_PARTIAL_DELIVERY_POINT: |
| 3031 | retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen); |
| 3032 | break; |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 3033 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3034 | case SCTP_INITMSG: |
| 3035 | retval = sctp_setsockopt_initmsg(sk, optval, optlen); |
| 3036 | break; |
| 3037 | case SCTP_DEFAULT_SEND_PARAM: |
| 3038 | retval = sctp_setsockopt_default_send_param(sk, optval, |
| 3039 | optlen); |
| 3040 | break; |
| 3041 | case SCTP_PRIMARY_ADDR: |
| 3042 | retval = sctp_setsockopt_primary_addr(sk, optval, optlen); |
| 3043 | break; |
| 3044 | case SCTP_SET_PEER_PRIMARY_ADDR: |
| 3045 | retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen); |
| 3046 | break; |
| 3047 | case SCTP_NODELAY: |
| 3048 | retval = sctp_setsockopt_nodelay(sk, optval, optlen); |
| 3049 | break; |
| 3050 | case SCTP_RTOINFO: |
| 3051 | retval = sctp_setsockopt_rtoinfo(sk, optval, optlen); |
| 3052 | break; |
| 3053 | case SCTP_ASSOCINFO: |
| 3054 | retval = sctp_setsockopt_associnfo(sk, optval, optlen); |
| 3055 | break; |
| 3056 | case SCTP_I_WANT_MAPPED_V4_ADDR: |
| 3057 | retval = sctp_setsockopt_mappedv4(sk, optval, optlen); |
| 3058 | break; |
| 3059 | case SCTP_MAXSEG: |
| 3060 | retval = sctp_setsockopt_maxseg(sk, optval, optlen); |
| 3061 | break; |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 3062 | case SCTP_ADAPTATION_LAYER: |
| 3063 | retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3064 | break; |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 3065 | case SCTP_CONTEXT: |
| 3066 | retval = sctp_setsockopt_context(sk, optval, optlen); |
| 3067 | break; |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 3068 | case SCTP_FRAGMENT_INTERLEAVE: |
| 3069 | retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen); |
| 3070 | break; |
Vlad Yasevich | 7033157 | 2007-03-23 11:34:36 -0700 | [diff] [blame] | 3071 | case SCTP_MAX_BURST: |
| 3072 | retval = sctp_setsockopt_maxburst(sk, optval, optlen); |
| 3073 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3074 | default: |
| 3075 | retval = -ENOPROTOOPT; |
| 3076 | break; |
Stephen Hemminger | 3ff50b7 | 2007-04-20 17:09:22 -0700 | [diff] [blame] | 3077 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3078 | |
| 3079 | sctp_release_sock(sk); |
| 3080 | |
| 3081 | out_nounlock: |
| 3082 | return retval; |
| 3083 | } |
| 3084 | |
| 3085 | /* API 3.1.6 connect() - UDP Style Syntax |
| 3086 | * |
| 3087 | * An application may use the connect() call in the UDP model to initiate an |
| 3088 | * association without sending data. |
| 3089 | * |
| 3090 | * The syntax is: |
| 3091 | * |
| 3092 | * ret = connect(int sd, const struct sockaddr *nam, socklen_t len); |
| 3093 | * |
| 3094 | * sd: the socket descriptor to have a new association added to. |
| 3095 | * |
| 3096 | * nam: the address structure (either struct sockaddr_in or struct |
| 3097 | * sockaddr_in6 defined in RFC2553 [7]). |
| 3098 | * |
| 3099 | * len: the size of the address. |
| 3100 | */ |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3101 | SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3102 | int addr_len) |
| 3103 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3104 | int err = 0; |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3105 | struct sctp_af *af; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3106 | |
| 3107 | sctp_lock_sock(sk); |
| 3108 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3109 | SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n", |
| 3110 | __FUNCTION__, sk, addr, addr_len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3111 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3112 | /* Validate addr_len before calling common connect/connectx routine. */ |
| 3113 | af = sctp_get_af_specific(addr->sa_family); |
| 3114 | if (!af || addr_len < af->sockaddr_len) { |
| 3115 | err = -EINVAL; |
| 3116 | } else { |
| 3117 | /* Pass correct addr len to common routine (so it knows there |
| 3118 | * is only one address being passed. |
| 3119 | */ |
| 3120 | err = __sctp_connect(sk, addr, af->sockaddr_len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3121 | } |
| 3122 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3123 | sctp_release_sock(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3124 | return err; |
| 3125 | } |
| 3126 | |
| 3127 | /* FIXME: Write comments. */ |
| 3128 | SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags) |
| 3129 | { |
| 3130 | return -EOPNOTSUPP; /* STUB */ |
| 3131 | } |
| 3132 | |
| 3133 | /* 4.1.4 accept() - TCP Style Syntax |
| 3134 | * |
| 3135 | * Applications use accept() call to remove an established SCTP |
| 3136 | * association from the accept queue of the endpoint. A new socket |
| 3137 | * descriptor will be returned from accept() to represent the newly |
| 3138 | * formed association. |
| 3139 | */ |
| 3140 | SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err) |
| 3141 | { |
| 3142 | struct sctp_sock *sp; |
| 3143 | struct sctp_endpoint *ep; |
| 3144 | struct sock *newsk = NULL; |
| 3145 | struct sctp_association *asoc; |
| 3146 | long timeo; |
| 3147 | int error = 0; |
| 3148 | |
| 3149 | sctp_lock_sock(sk); |
| 3150 | |
| 3151 | sp = sctp_sk(sk); |
| 3152 | ep = sp->ep; |
| 3153 | |
| 3154 | if (!sctp_style(sk, TCP)) { |
| 3155 | error = -EOPNOTSUPP; |
| 3156 | goto out; |
| 3157 | } |
| 3158 | |
| 3159 | if (!sctp_sstate(sk, LISTENING)) { |
| 3160 | error = -EINVAL; |
| 3161 | goto out; |
| 3162 | } |
| 3163 | |
Sridhar Samudrala | 8abfedd | 2006-08-22 00:24:09 -0700 | [diff] [blame] | 3164 | timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3165 | |
| 3166 | error = sctp_wait_for_accept(sk, timeo); |
| 3167 | if (error) |
| 3168 | goto out; |
| 3169 | |
| 3170 | /* We treat the list of associations on the endpoint as the accept |
| 3171 | * queue and pick the first association on the list. |
| 3172 | */ |
| 3173 | asoc = list_entry(ep->asocs.next, struct sctp_association, asocs); |
| 3174 | |
| 3175 | newsk = sp->pf->create_accept_sk(sk, asoc); |
| 3176 | if (!newsk) { |
| 3177 | error = -ENOMEM; |
| 3178 | goto out; |
| 3179 | } |
| 3180 | |
| 3181 | /* Populate the fields of the newsk from the oldsk and migrate the |
| 3182 | * asoc to the newsk. |
| 3183 | */ |
| 3184 | sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP); |
| 3185 | |
| 3186 | out: |
| 3187 | sctp_release_sock(sk); |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 3188 | *err = error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3189 | return newsk; |
| 3190 | } |
| 3191 | |
| 3192 | /* The SCTP ioctl handler. */ |
| 3193 | SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
| 3194 | { |
| 3195 | return -ENOIOCTLCMD; |
| 3196 | } |
| 3197 | |
| 3198 | /* This is the function which gets called during socket creation to |
| 3199 | * initialized the SCTP-specific portion of the sock. |
| 3200 | * The sock structure should already be zero-filled memory. |
| 3201 | */ |
| 3202 | SCTP_STATIC int sctp_init_sock(struct sock *sk) |
| 3203 | { |
| 3204 | struct sctp_endpoint *ep; |
| 3205 | struct sctp_sock *sp; |
| 3206 | |
| 3207 | SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk); |
| 3208 | |
| 3209 | sp = sctp_sk(sk); |
| 3210 | |
| 3211 | /* Initialize the SCTP per socket area. */ |
| 3212 | switch (sk->sk_type) { |
| 3213 | case SOCK_SEQPACKET: |
| 3214 | sp->type = SCTP_SOCKET_UDP; |
| 3215 | break; |
| 3216 | case SOCK_STREAM: |
| 3217 | sp->type = SCTP_SOCKET_TCP; |
| 3218 | break; |
| 3219 | default: |
| 3220 | return -ESOCKTNOSUPPORT; |
| 3221 | } |
| 3222 | |
| 3223 | /* Initialize default send parameters. These parameters can be |
| 3224 | * modified with the SCTP_DEFAULT_SEND_PARAM socket option. |
| 3225 | */ |
| 3226 | sp->default_stream = 0; |
| 3227 | sp->default_ppid = 0; |
| 3228 | sp->default_flags = 0; |
| 3229 | sp->default_context = 0; |
| 3230 | sp->default_timetolive = 0; |
| 3231 | |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 3232 | sp->default_rcv_context = 0; |
Vlad Yasevich | 7033157 | 2007-03-23 11:34:36 -0700 | [diff] [blame] | 3233 | sp->max_burst = sctp_max_burst; |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 3234 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3235 | /* Initialize default setup parameters. These parameters |
| 3236 | * can be modified with the SCTP_INITMSG socket option or |
| 3237 | * overridden by the SCTP_INIT CMSG. |
| 3238 | */ |
| 3239 | sp->initmsg.sinit_num_ostreams = sctp_max_outstreams; |
| 3240 | sp->initmsg.sinit_max_instreams = sctp_max_instreams; |
| 3241 | sp->initmsg.sinit_max_attempts = sctp_max_retrans_init; |
Vladislav Yasevich | 3fd091e | 2006-08-22 13:29:17 -0700 | [diff] [blame] | 3242 | sp->initmsg.sinit_max_init_timeo = sctp_rto_max; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3243 | |
| 3244 | /* Initialize default RTO related parameters. These parameters can |
| 3245 | * be modified for with the SCTP_RTOINFO socket option. |
| 3246 | */ |
Vladislav Yasevich | 3fd091e | 2006-08-22 13:29:17 -0700 | [diff] [blame] | 3247 | sp->rtoinfo.srto_initial = sctp_rto_initial; |
| 3248 | sp->rtoinfo.srto_max = sctp_rto_max; |
| 3249 | sp->rtoinfo.srto_min = sctp_rto_min; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3250 | |
| 3251 | /* Initialize default association related parameters. These parameters |
| 3252 | * can be modified with the SCTP_ASSOCINFO socket option. |
| 3253 | */ |
| 3254 | sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association; |
| 3255 | sp->assocparams.sasoc_number_peer_destinations = 0; |
| 3256 | sp->assocparams.sasoc_peer_rwnd = 0; |
| 3257 | sp->assocparams.sasoc_local_rwnd = 0; |
Vladislav Yasevich | 3fd091e | 2006-08-22 13:29:17 -0700 | [diff] [blame] | 3258 | sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3259 | |
| 3260 | /* Initialize default event subscriptions. By default, all the |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 3261 | * options are off. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3262 | */ |
| 3263 | memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe)); |
| 3264 | |
| 3265 | /* Default Peer Address Parameters. These defaults can |
| 3266 | * be modified via SCTP_PEER_ADDR_PARAMS |
| 3267 | */ |
Vladislav Yasevich | 3fd091e | 2006-08-22 13:29:17 -0700 | [diff] [blame] | 3268 | sp->hbinterval = sctp_hb_interval; |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3269 | sp->pathmaxrxt = sctp_max_retrans_path; |
| 3270 | sp->pathmtu = 0; // allow default discovery |
Vladislav Yasevich | 3fd091e | 2006-08-22 13:29:17 -0700 | [diff] [blame] | 3271 | sp->sackdelay = sctp_sack_timeout; |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3272 | sp->param_flags = SPP_HB_ENABLE | |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 3273 | SPP_PMTUD_ENABLE | |
| 3274 | SPP_SACKDELAY_ENABLE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3275 | |
| 3276 | /* If enabled no SCTP message fragmentation will be performed. |
| 3277 | * Configure through SCTP_DISABLE_FRAGMENTS socket option. |
| 3278 | */ |
| 3279 | sp->disable_fragments = 0; |
| 3280 | |
Sridhar Samudrala | 208edef | 2006-09-29 17:08:01 -0700 | [diff] [blame] | 3281 | /* Enable Nagle algorithm by default. */ |
| 3282 | sp->nodelay = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3283 | |
| 3284 | /* Enable by default. */ |
| 3285 | sp->v4mapped = 1; |
| 3286 | |
| 3287 | /* Auto-close idle associations after the configured |
| 3288 | * number of seconds. A value of 0 disables this |
| 3289 | * feature. Configure through the SCTP_AUTOCLOSE socket option, |
| 3290 | * for UDP-style sockets only. |
| 3291 | */ |
| 3292 | sp->autoclose = 0; |
| 3293 | |
| 3294 | /* User specified fragmentation limit. */ |
| 3295 | sp->user_frag = 0; |
| 3296 | |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 3297 | sp->adaptation_ind = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3298 | |
| 3299 | sp->pf = sctp_get_pf_specific(sk->sk_family); |
| 3300 | |
| 3301 | /* Control variables for partial data delivery. */ |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 3302 | atomic_set(&sp->pd_mode, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3303 | skb_queue_head_init(&sp->pd_lobby); |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 3304 | sp->frag_interleave = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3305 | |
| 3306 | /* Create a per socket endpoint structure. Even if we |
| 3307 | * change the data structure relationships, this may still |
| 3308 | * be useful for storing pre-connect address information. |
| 3309 | */ |
| 3310 | ep = sctp_endpoint_new(sk, GFP_KERNEL); |
| 3311 | if (!ep) |
| 3312 | return -ENOMEM; |
| 3313 | |
| 3314 | sp->ep = ep; |
| 3315 | sp->hmac = NULL; |
| 3316 | |
| 3317 | SCTP_DBG_OBJCNT_INC(sock); |
| 3318 | return 0; |
| 3319 | } |
| 3320 | |
| 3321 | /* Cleanup any SCTP per socket resources. */ |
| 3322 | SCTP_STATIC int sctp_destroy_sock(struct sock *sk) |
| 3323 | { |
| 3324 | struct sctp_endpoint *ep; |
| 3325 | |
| 3326 | SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk); |
| 3327 | |
| 3328 | /* Release our hold on the endpoint. */ |
| 3329 | ep = sctp_sk(sk)->ep; |
| 3330 | sctp_endpoint_free(ep); |
| 3331 | |
| 3332 | return 0; |
| 3333 | } |
| 3334 | |
| 3335 | /* API 4.1.7 shutdown() - TCP Style Syntax |
| 3336 | * int shutdown(int socket, int how); |
| 3337 | * |
| 3338 | * sd - the socket descriptor of the association to be closed. |
| 3339 | * how - Specifies the type of shutdown. The values are |
| 3340 | * as follows: |
| 3341 | * SHUT_RD |
| 3342 | * Disables further receive operations. No SCTP |
| 3343 | * protocol action is taken. |
| 3344 | * SHUT_WR |
| 3345 | * Disables further send operations, and initiates |
| 3346 | * the SCTP shutdown sequence. |
| 3347 | * SHUT_RDWR |
| 3348 | * Disables further send and receive operations |
| 3349 | * and initiates the SCTP shutdown sequence. |
| 3350 | */ |
| 3351 | SCTP_STATIC void sctp_shutdown(struct sock *sk, int how) |
| 3352 | { |
| 3353 | struct sctp_endpoint *ep; |
| 3354 | struct sctp_association *asoc; |
| 3355 | |
| 3356 | if (!sctp_style(sk, TCP)) |
| 3357 | return; |
| 3358 | |
| 3359 | if (how & SEND_SHUTDOWN) { |
| 3360 | ep = sctp_sk(sk)->ep; |
| 3361 | if (!list_empty(&ep->asocs)) { |
| 3362 | asoc = list_entry(ep->asocs.next, |
| 3363 | struct sctp_association, asocs); |
| 3364 | sctp_primitive_SHUTDOWN(asoc, NULL); |
| 3365 | } |
| 3366 | } |
| 3367 | } |
| 3368 | |
| 3369 | /* 7.2.1 Association Status (SCTP_STATUS) |
| 3370 | |
| 3371 | * Applications can retrieve current status information about an |
| 3372 | * association, including association state, peer receiver window size, |
| 3373 | * number of unacked data chunks, and number of data chunks pending |
| 3374 | * receipt. This information is read-only. |
| 3375 | */ |
| 3376 | static int sctp_getsockopt_sctp_status(struct sock *sk, int len, |
| 3377 | char __user *optval, |
| 3378 | int __user *optlen) |
| 3379 | { |
| 3380 | struct sctp_status status; |
| 3381 | struct sctp_association *asoc = NULL; |
| 3382 | struct sctp_transport *transport; |
| 3383 | sctp_assoc_t associd; |
| 3384 | int retval = 0; |
| 3385 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3386 | if (len < sizeof(status)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3387 | retval = -EINVAL; |
| 3388 | goto out; |
| 3389 | } |
| 3390 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3391 | len = sizeof(status); |
| 3392 | if (copy_from_user(&status, optval, len)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3393 | retval = -EFAULT; |
| 3394 | goto out; |
| 3395 | } |
| 3396 | |
| 3397 | associd = status.sstat_assoc_id; |
| 3398 | asoc = sctp_id2assoc(sk, associd); |
| 3399 | if (!asoc) { |
| 3400 | retval = -EINVAL; |
| 3401 | goto out; |
| 3402 | } |
| 3403 | |
| 3404 | transport = asoc->peer.primary_path; |
| 3405 | |
| 3406 | status.sstat_assoc_id = sctp_assoc2id(asoc); |
| 3407 | status.sstat_state = asoc->state; |
| 3408 | status.sstat_rwnd = asoc->peer.rwnd; |
| 3409 | status.sstat_unackdata = asoc->unack_data; |
| 3410 | |
| 3411 | status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map); |
| 3412 | status.sstat_instrms = asoc->c.sinit_max_instreams; |
| 3413 | status.sstat_outstrms = asoc->c.sinit_num_ostreams; |
| 3414 | status.sstat_fragmentation_point = asoc->frag_point; |
| 3415 | status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc); |
Al Viro | 8cec6b8 | 2006-11-20 17:23:01 -0800 | [diff] [blame] | 3416 | memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr, |
| 3417 | transport->af_specific->sockaddr_len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3418 | /* Map ipv4 address into v4-mapped-on-v6 address. */ |
| 3419 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk), |
| 3420 | (union sctp_addr *)&status.sstat_primary.spinfo_address); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3421 | status.sstat_primary.spinfo_state = transport->state; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3422 | status.sstat_primary.spinfo_cwnd = transport->cwnd; |
| 3423 | status.sstat_primary.spinfo_srtt = transport->srtt; |
| 3424 | status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto); |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3425 | status.sstat_primary.spinfo_mtu = transport->pathmtu; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3426 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3427 | if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN) |
| 3428 | status.sstat_primary.spinfo_state = SCTP_ACTIVE; |
| 3429 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3430 | if (put_user(len, optlen)) { |
| 3431 | retval = -EFAULT; |
| 3432 | goto out; |
| 3433 | } |
| 3434 | |
| 3435 | SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n", |
| 3436 | len, status.sstat_state, status.sstat_rwnd, |
| 3437 | status.sstat_assoc_id); |
| 3438 | |
| 3439 | if (copy_to_user(optval, &status, len)) { |
| 3440 | retval = -EFAULT; |
| 3441 | goto out; |
| 3442 | } |
| 3443 | |
| 3444 | out: |
| 3445 | return (retval); |
| 3446 | } |
| 3447 | |
| 3448 | |
| 3449 | /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO) |
| 3450 | * |
| 3451 | * Applications can retrieve information about a specific peer address |
| 3452 | * of an association, including its reachability state, congestion |
| 3453 | * window, and retransmission timer values. This information is |
| 3454 | * read-only. |
| 3455 | */ |
| 3456 | static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len, |
| 3457 | char __user *optval, |
| 3458 | int __user *optlen) |
| 3459 | { |
| 3460 | struct sctp_paddrinfo pinfo; |
| 3461 | struct sctp_transport *transport; |
| 3462 | int retval = 0; |
| 3463 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3464 | if (len < sizeof(pinfo)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3465 | retval = -EINVAL; |
| 3466 | goto out; |
| 3467 | } |
| 3468 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3469 | len = sizeof(pinfo); |
| 3470 | if (copy_from_user(&pinfo, optval, len)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3471 | retval = -EFAULT; |
| 3472 | goto out; |
| 3473 | } |
| 3474 | |
| 3475 | transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address, |
| 3476 | pinfo.spinfo_assoc_id); |
| 3477 | if (!transport) |
| 3478 | return -EINVAL; |
| 3479 | |
| 3480 | pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc); |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3481 | pinfo.spinfo_state = transport->state; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3482 | pinfo.spinfo_cwnd = transport->cwnd; |
| 3483 | pinfo.spinfo_srtt = transport->srtt; |
| 3484 | pinfo.spinfo_rto = jiffies_to_msecs(transport->rto); |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3485 | pinfo.spinfo_mtu = transport->pathmtu; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3486 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 3487 | if (pinfo.spinfo_state == SCTP_UNKNOWN) |
| 3488 | pinfo.spinfo_state = SCTP_ACTIVE; |
| 3489 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3490 | if (put_user(len, optlen)) { |
| 3491 | retval = -EFAULT; |
| 3492 | goto out; |
| 3493 | } |
| 3494 | |
| 3495 | if (copy_to_user(optval, &pinfo, len)) { |
| 3496 | retval = -EFAULT; |
| 3497 | goto out; |
| 3498 | } |
| 3499 | |
| 3500 | out: |
| 3501 | return (retval); |
| 3502 | } |
| 3503 | |
| 3504 | /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) |
| 3505 | * |
| 3506 | * This option is a on/off flag. If enabled no SCTP message |
| 3507 | * fragmentation will be performed. Instead if a message being sent |
| 3508 | * exceeds the current PMTU size, the message will NOT be sent and |
| 3509 | * instead a error will be indicated to the user. |
| 3510 | */ |
| 3511 | static int sctp_getsockopt_disable_fragments(struct sock *sk, int len, |
| 3512 | char __user *optval, int __user *optlen) |
| 3513 | { |
| 3514 | int val; |
| 3515 | |
| 3516 | if (len < sizeof(int)) |
| 3517 | return -EINVAL; |
| 3518 | |
| 3519 | len = sizeof(int); |
| 3520 | val = (sctp_sk(sk)->disable_fragments == 1); |
| 3521 | if (put_user(len, optlen)) |
| 3522 | return -EFAULT; |
| 3523 | if (copy_to_user(optval, &val, len)) |
| 3524 | return -EFAULT; |
| 3525 | return 0; |
| 3526 | } |
| 3527 | |
| 3528 | /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS) |
| 3529 | * |
| 3530 | * This socket option is used to specify various notifications and |
| 3531 | * ancillary data the user wishes to receive. |
| 3532 | */ |
| 3533 | static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval, |
| 3534 | int __user *optlen) |
| 3535 | { |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3536 | if (len < sizeof(struct sctp_event_subscribe)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3537 | return -EINVAL; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3538 | len = sizeof(struct sctp_event_subscribe); |
| 3539 | if (put_user(len, optlen)) |
| 3540 | return -EFAULT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3541 | if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len)) |
| 3542 | return -EFAULT; |
| 3543 | return 0; |
| 3544 | } |
| 3545 | |
| 3546 | /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) |
| 3547 | * |
| 3548 | * This socket option is applicable to the UDP-style socket only. When |
| 3549 | * set it will cause associations that are idle for more than the |
| 3550 | * specified number of seconds to automatically close. An association |
| 3551 | * being idle is defined an association that has NOT sent or received |
| 3552 | * user data. The special value of '0' indicates that no automatic |
| 3553 | * close of any associations should be performed. The option expects an |
| 3554 | * integer defining the number of seconds of idle time before an |
| 3555 | * association is closed. |
| 3556 | */ |
| 3557 | static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen) |
| 3558 | { |
| 3559 | /* Applicable to UDP-style socket only */ |
| 3560 | if (sctp_style(sk, TCP)) |
| 3561 | return -EOPNOTSUPP; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3562 | if (len < sizeof(int)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3563 | return -EINVAL; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3564 | len = sizeof(int); |
| 3565 | if (put_user(len, optlen)) |
| 3566 | return -EFAULT; |
| 3567 | if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3568 | return -EFAULT; |
| 3569 | return 0; |
| 3570 | } |
| 3571 | |
| 3572 | /* Helper routine to branch off an association to a new socket. */ |
| 3573 | SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc, |
| 3574 | struct socket **sockp) |
| 3575 | { |
| 3576 | struct sock *sk = asoc->base.sk; |
| 3577 | struct socket *sock; |
Vlad Yasevich | 4f44430 | 2006-10-30 18:54:32 -0800 | [diff] [blame] | 3578 | struct inet_sock *inetsk; |
Vlad Yasevich | d570ee4 | 2007-05-15 16:32:39 -0400 | [diff] [blame] | 3579 | struct sctp_af *af; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3580 | int err = 0; |
| 3581 | |
| 3582 | /* An association cannot be branched off from an already peeled-off |
| 3583 | * socket, nor is this supported for tcp style sockets. |
| 3584 | */ |
| 3585 | if (!sctp_style(sk, UDP)) |
| 3586 | return -EINVAL; |
| 3587 | |
| 3588 | /* Create a new socket. */ |
| 3589 | err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock); |
| 3590 | if (err < 0) |
| 3591 | return err; |
| 3592 | |
| 3593 | /* Populate the fields of the newsk from the oldsk and migrate the |
| 3594 | * asoc to the newsk. |
| 3595 | */ |
| 3596 | sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH); |
Vlad Yasevich | 4f44430 | 2006-10-30 18:54:32 -0800 | [diff] [blame] | 3597 | |
| 3598 | /* Make peeled-off sockets more like 1-1 accepted sockets. |
| 3599 | * Set the daddr and initialize id to something more random |
| 3600 | */ |
Vlad Yasevich | d570ee4 | 2007-05-15 16:32:39 -0400 | [diff] [blame] | 3601 | af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family); |
| 3602 | af->to_sk_daddr(&asoc->peer.primary_addr, sk); |
Vlad Yasevich | 4f44430 | 2006-10-30 18:54:32 -0800 | [diff] [blame] | 3603 | inetsk = inet_sk(sock->sk); |
Vlad Yasevich | 4f44430 | 2006-10-30 18:54:32 -0800 | [diff] [blame] | 3604 | inetsk->id = asoc->next_tsn ^ jiffies; |
| 3605 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3606 | *sockp = sock; |
| 3607 | |
| 3608 | return err; |
| 3609 | } |
| 3610 | |
| 3611 | static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen) |
| 3612 | { |
| 3613 | sctp_peeloff_arg_t peeloff; |
| 3614 | struct socket *newsock; |
| 3615 | int retval = 0; |
| 3616 | struct sctp_association *asoc; |
| 3617 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3618 | if (len < sizeof(sctp_peeloff_arg_t)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3619 | return -EINVAL; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3620 | len = sizeof(sctp_peeloff_arg_t); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3621 | if (copy_from_user(&peeloff, optval, len)) |
| 3622 | return -EFAULT; |
| 3623 | |
| 3624 | asoc = sctp_id2assoc(sk, peeloff.associd); |
| 3625 | if (!asoc) { |
| 3626 | retval = -EINVAL; |
| 3627 | goto out; |
| 3628 | } |
| 3629 | |
| 3630 | SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc); |
| 3631 | |
| 3632 | retval = sctp_do_peeloff(asoc, &newsock); |
| 3633 | if (retval < 0) |
| 3634 | goto out; |
| 3635 | |
| 3636 | /* Map the socket to an unused fd that can be returned to the user. */ |
| 3637 | retval = sock_map_fd(newsock); |
| 3638 | if (retval < 0) { |
| 3639 | sock_release(newsock); |
| 3640 | goto out; |
| 3641 | } |
| 3642 | |
| 3643 | SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n", |
| 3644 | __FUNCTION__, sk, asoc, newsock->sk, retval); |
| 3645 | |
| 3646 | /* Return the fd mapped to the new socket. */ |
| 3647 | peeloff.sd = retval; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3648 | if (put_user(len, optlen)) |
| 3649 | return -EFAULT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3650 | if (copy_to_user(optval, &peeloff, len)) |
| 3651 | retval = -EFAULT; |
| 3652 | |
| 3653 | out: |
| 3654 | return retval; |
| 3655 | } |
| 3656 | |
| 3657 | /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) |
| 3658 | * |
| 3659 | * Applications can enable or disable heartbeats for any peer address of |
| 3660 | * an association, modify an address's heartbeat interval, force a |
| 3661 | * heartbeat to be sent immediately, and adjust the address's maximum |
| 3662 | * number of retransmissions sent before an address is considered |
| 3663 | * unreachable. The following structure is used to access and modify an |
| 3664 | * address's parameters: |
| 3665 | * |
| 3666 | * struct sctp_paddrparams { |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3667 | * sctp_assoc_t spp_assoc_id; |
| 3668 | * struct sockaddr_storage spp_address; |
| 3669 | * uint32_t spp_hbinterval; |
| 3670 | * uint16_t spp_pathmaxrxt; |
| 3671 | * uint32_t spp_pathmtu; |
| 3672 | * uint32_t spp_sackdelay; |
| 3673 | * uint32_t spp_flags; |
| 3674 | * }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3675 | * |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3676 | * spp_assoc_id - (one-to-many style socket) This is filled in the |
| 3677 | * application, and identifies the association for |
| 3678 | * this query. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3679 | * spp_address - This specifies which address is of interest. |
| 3680 | * spp_hbinterval - This contains the value of the heartbeat interval, |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3681 | * in milliseconds. If a value of zero |
| 3682 | * is present in this field then no changes are to |
| 3683 | * be made to this parameter. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3684 | * spp_pathmaxrxt - This contains the maximum number of |
| 3685 | * retransmissions before this address shall be |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3686 | * considered unreachable. If a value of zero |
| 3687 | * is present in this field then no changes are to |
| 3688 | * be made to this parameter. |
| 3689 | * spp_pathmtu - When Path MTU discovery is disabled the value |
| 3690 | * specified here will be the "fixed" path mtu. |
| 3691 | * Note that if the spp_address field is empty |
| 3692 | * then all associations on this address will |
| 3693 | * have this fixed path mtu set upon them. |
| 3694 | * |
| 3695 | * spp_sackdelay - When delayed sack is enabled, this value specifies |
| 3696 | * the number of milliseconds that sacks will be delayed |
| 3697 | * for. This value will apply to all addresses of an |
| 3698 | * association if the spp_address field is empty. Note |
| 3699 | * also, that if delayed sack is enabled and this |
| 3700 | * value is set to 0, no change is made to the last |
| 3701 | * recorded delayed sack timer value. |
| 3702 | * |
| 3703 | * spp_flags - These flags are used to control various features |
| 3704 | * on an association. The flag field may contain |
| 3705 | * zero or more of the following options. |
| 3706 | * |
| 3707 | * SPP_HB_ENABLE - Enable heartbeats on the |
| 3708 | * specified address. Note that if the address |
| 3709 | * field is empty all addresses for the association |
| 3710 | * have heartbeats enabled upon them. |
| 3711 | * |
| 3712 | * SPP_HB_DISABLE - Disable heartbeats on the |
| 3713 | * speicifed address. Note that if the address |
| 3714 | * field is empty all addresses for the association |
| 3715 | * will have their heartbeats disabled. Note also |
| 3716 | * that SPP_HB_ENABLE and SPP_HB_DISABLE are |
| 3717 | * mutually exclusive, only one of these two should |
| 3718 | * be specified. Enabling both fields will have |
| 3719 | * undetermined results. |
| 3720 | * |
| 3721 | * SPP_HB_DEMAND - Request a user initiated heartbeat |
| 3722 | * to be made immediately. |
| 3723 | * |
| 3724 | * SPP_PMTUD_ENABLE - This field will enable PMTU |
| 3725 | * discovery upon the specified address. Note that |
| 3726 | * if the address feild is empty then all addresses |
| 3727 | * on the association are effected. |
| 3728 | * |
| 3729 | * SPP_PMTUD_DISABLE - This field will disable PMTU |
| 3730 | * discovery upon the specified address. Note that |
| 3731 | * if the address feild is empty then all addresses |
| 3732 | * on the association are effected. Not also that |
| 3733 | * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually |
| 3734 | * exclusive. Enabling both will have undetermined |
| 3735 | * results. |
| 3736 | * |
| 3737 | * SPP_SACKDELAY_ENABLE - Setting this flag turns |
| 3738 | * on delayed sack. The time specified in spp_sackdelay |
| 3739 | * is used to specify the sack delay for this address. Note |
| 3740 | * that if spp_address is empty then all addresses will |
| 3741 | * enable delayed sack and take on the sack delay |
| 3742 | * value specified in spp_sackdelay. |
| 3743 | * SPP_SACKDELAY_DISABLE - Setting this flag turns |
| 3744 | * off delayed sack. If the spp_address field is blank then |
| 3745 | * delayed sack is disabled for the entire association. Note |
| 3746 | * also that this field is mutually exclusive to |
| 3747 | * SPP_SACKDELAY_ENABLE, setting both will have undefined |
| 3748 | * results. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3749 | */ |
| 3750 | static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len, |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3751 | char __user *optval, int __user *optlen) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3752 | { |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3753 | struct sctp_paddrparams params; |
| 3754 | struct sctp_transport *trans = NULL; |
| 3755 | struct sctp_association *asoc = NULL; |
| 3756 | struct sctp_sock *sp = sctp_sk(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3757 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3758 | if (len < sizeof(struct sctp_paddrparams)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3759 | return -EINVAL; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3760 | len = sizeof(struct sctp_paddrparams); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3761 | if (copy_from_user(¶ms, optval, len)) |
| 3762 | return -EFAULT; |
| 3763 | |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3764 | /* If an address other than INADDR_ANY is specified, and |
| 3765 | * no transport is found, then the request is invalid. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3766 | */ |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3767 | if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) { |
| 3768 | trans = sctp_addr_id2transport(sk, ¶ms.spp_address, |
| 3769 | params.spp_assoc_id); |
| 3770 | if (!trans) { |
| 3771 | SCTP_DEBUG_PRINTK("Failed no transport\n"); |
| 3772 | return -EINVAL; |
| 3773 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3774 | } |
| 3775 | |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3776 | /* Get association, if assoc_id != 0 and the socket is a one |
| 3777 | * to many style socket, and an association was not found, then |
| 3778 | * the id was invalid. |
| 3779 | */ |
| 3780 | asoc = sctp_id2assoc(sk, params.spp_assoc_id); |
| 3781 | if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) { |
| 3782 | SCTP_DEBUG_PRINTK("Failed no association\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3783 | return -EINVAL; |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3784 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3785 | |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3786 | if (trans) { |
| 3787 | /* Fetch transport values. */ |
| 3788 | params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval); |
| 3789 | params.spp_pathmtu = trans->pathmtu; |
| 3790 | params.spp_pathmaxrxt = trans->pathmaxrxt; |
| 3791 | params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3792 | |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3793 | /*draft-11 doesn't say what to return in spp_flags*/ |
| 3794 | params.spp_flags = trans->param_flags; |
| 3795 | } else if (asoc) { |
| 3796 | /* Fetch association values. */ |
| 3797 | params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval); |
| 3798 | params.spp_pathmtu = asoc->pathmtu; |
| 3799 | params.spp_pathmaxrxt = asoc->pathmaxrxt; |
| 3800 | params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3801 | |
Frank Filz | 52ccb8e | 2005-12-22 11:36:46 -0800 | [diff] [blame] | 3802 | /*draft-11 doesn't say what to return in spp_flags*/ |
| 3803 | params.spp_flags = asoc->param_flags; |
| 3804 | } else { |
| 3805 | /* Fetch socket values. */ |
| 3806 | params.spp_hbinterval = sp->hbinterval; |
| 3807 | params.spp_pathmtu = sp->pathmtu; |
| 3808 | params.spp_sackdelay = sp->sackdelay; |
| 3809 | params.spp_pathmaxrxt = sp->pathmaxrxt; |
| 3810 | |
| 3811 | /*draft-11 doesn't say what to return in spp_flags*/ |
| 3812 | params.spp_flags = sp->param_flags; |
| 3813 | } |
| 3814 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3815 | if (copy_to_user(optval, ¶ms, len)) |
| 3816 | return -EFAULT; |
| 3817 | |
| 3818 | if (put_user(len, optlen)) |
| 3819 | return -EFAULT; |
| 3820 | |
| 3821 | return 0; |
| 3822 | } |
| 3823 | |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 3824 | /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME) |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 3825 | * |
| 3826 | * This options will get or set the delayed ack timer. The time is set |
| 3827 | * in milliseconds. If the assoc_id is 0, then this sets or gets the |
| 3828 | * endpoints default delayed ack timer value. If the assoc_id field is |
| 3829 | * non-zero, then the set or get effects the specified association. |
| 3830 | * |
| 3831 | * struct sctp_assoc_value { |
| 3832 | * sctp_assoc_t assoc_id; |
| 3833 | * uint32_t assoc_value; |
| 3834 | * }; |
| 3835 | * |
| 3836 | * assoc_id - This parameter, indicates which association the |
| 3837 | * user is preforming an action upon. Note that if |
| 3838 | * this field's value is zero then the endpoints |
| 3839 | * default value is changed (effecting future |
| 3840 | * associations only). |
| 3841 | * |
| 3842 | * assoc_value - This parameter contains the number of milliseconds |
| 3843 | * that the user is requesting the delayed ACK timer |
| 3844 | * be set to. Note that this value is defined in |
| 3845 | * the standard to be between 200 and 500 milliseconds. |
| 3846 | * |
| 3847 | * Note: a value of zero will leave the value alone, |
| 3848 | * but disable SACK delay. A non-zero value will also |
| 3849 | * enable SACK delay. |
| 3850 | */ |
| 3851 | static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len, |
| 3852 | char __user *optval, |
| 3853 | int __user *optlen) |
| 3854 | { |
| 3855 | struct sctp_assoc_value params; |
| 3856 | struct sctp_association *asoc = NULL; |
| 3857 | struct sctp_sock *sp = sctp_sk(sk); |
| 3858 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3859 | if (len < sizeof(struct sctp_assoc_value)) |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 3860 | return - EINVAL; |
| 3861 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3862 | len = sizeof(struct sctp_assoc_value); |
| 3863 | |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 3864 | if (copy_from_user(¶ms, optval, len)) |
| 3865 | return -EFAULT; |
| 3866 | |
| 3867 | /* Get association, if assoc_id != 0 and the socket is a one |
| 3868 | * to many style socket, and an association was not found, then |
| 3869 | * the id was invalid. |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 3870 | */ |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 3871 | asoc = sctp_id2assoc(sk, params.assoc_id); |
| 3872 | if (!asoc && params.assoc_id && sctp_style(sk, UDP)) |
| 3873 | return -EINVAL; |
| 3874 | |
| 3875 | if (asoc) { |
| 3876 | /* Fetch association values. */ |
| 3877 | if (asoc->param_flags & SPP_SACKDELAY_ENABLE) |
| 3878 | params.assoc_value = jiffies_to_msecs( |
| 3879 | asoc->sackdelay); |
| 3880 | else |
| 3881 | params.assoc_value = 0; |
| 3882 | } else { |
| 3883 | /* Fetch socket values. */ |
| 3884 | if (sp->param_flags & SPP_SACKDELAY_ENABLE) |
| 3885 | params.assoc_value = sp->sackdelay; |
| 3886 | else |
| 3887 | params.assoc_value = 0; |
| 3888 | } |
| 3889 | |
| 3890 | if (copy_to_user(optval, ¶ms, len)) |
| 3891 | return -EFAULT; |
| 3892 | |
| 3893 | if (put_user(len, optlen)) |
| 3894 | return -EFAULT; |
| 3895 | |
| 3896 | return 0; |
| 3897 | } |
| 3898 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3899 | /* 7.1.3 Initialization Parameters (SCTP_INITMSG) |
| 3900 | * |
| 3901 | * Applications can specify protocol parameters for the default association |
| 3902 | * initialization. The option name argument to setsockopt() and getsockopt() |
| 3903 | * is SCTP_INITMSG. |
| 3904 | * |
| 3905 | * Setting initialization parameters is effective only on an unconnected |
| 3906 | * socket (for UDP-style sockets only future associations are effected |
| 3907 | * by the change). With TCP-style sockets, this option is inherited by |
| 3908 | * sockets derived from a listener socket. |
| 3909 | */ |
| 3910 | static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen) |
| 3911 | { |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3912 | if (len < sizeof(struct sctp_initmsg)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3913 | return -EINVAL; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3914 | len = sizeof(struct sctp_initmsg); |
| 3915 | if (put_user(len, optlen)) |
| 3916 | return -EFAULT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3917 | if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len)) |
| 3918 | return -EFAULT; |
| 3919 | return 0; |
| 3920 | } |
| 3921 | |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 3922 | static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len, |
| 3923 | char __user *optval, |
| 3924 | int __user *optlen) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3925 | { |
| 3926 | sctp_assoc_t id; |
| 3927 | struct sctp_association *asoc; |
| 3928 | struct list_head *pos; |
| 3929 | int cnt = 0; |
| 3930 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3931 | if (len < sizeof(sctp_assoc_t)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3932 | return -EINVAL; |
| 3933 | |
| 3934 | if (copy_from_user(&id, optval, sizeof(sctp_assoc_t))) |
| 3935 | return -EFAULT; |
| 3936 | |
| 3937 | /* For UDP-style sockets, id specifies the association to query. */ |
| 3938 | asoc = sctp_id2assoc(sk, id); |
| 3939 | if (!asoc) |
| 3940 | return -EINVAL; |
| 3941 | |
| 3942 | list_for_each(pos, &asoc->peer.transport_addr_list) { |
| 3943 | cnt ++; |
| 3944 | } |
| 3945 | |
| 3946 | return cnt; |
| 3947 | } |
| 3948 | |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 3949 | /* |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 3950 | * Old API for getting list of peer addresses. Does not work for 32-bit |
| 3951 | * programs running on a 64-bit kernel |
| 3952 | */ |
| 3953 | static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len, |
| 3954 | char __user *optval, |
| 3955 | int __user *optlen) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3956 | { |
| 3957 | struct sctp_association *asoc; |
| 3958 | struct list_head *pos; |
| 3959 | int cnt = 0; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 3960 | struct sctp_getaddrs_old getaddrs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3961 | struct sctp_transport *from; |
| 3962 | void __user *to; |
| 3963 | union sctp_addr temp; |
| 3964 | struct sctp_sock *sp = sctp_sk(sk); |
| 3965 | int addrlen; |
| 3966 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3967 | if (len < sizeof(struct sctp_getaddrs_old)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3968 | return -EINVAL; |
| 3969 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3970 | len = sizeof(struct sctp_getaddrs_old); |
| 3971 | |
| 3972 | if (copy_from_user(&getaddrs, optval, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3973 | return -EFAULT; |
| 3974 | |
| 3975 | if (getaddrs.addr_num <= 0) return -EINVAL; |
| 3976 | |
| 3977 | /* For UDP-style sockets, id specifies the association to query. */ |
| 3978 | asoc = sctp_id2assoc(sk, getaddrs.assoc_id); |
| 3979 | if (!asoc) |
| 3980 | return -EINVAL; |
| 3981 | |
| 3982 | to = (void __user *)getaddrs.addrs; |
| 3983 | list_for_each(pos, &asoc->peer.transport_addr_list) { |
| 3984 | from = list_entry(pos, struct sctp_transport, transports); |
Al Viro | b3f5b3b | 2006-11-20 17:22:43 -0800 | [diff] [blame] | 3985 | memcpy(&temp, &from->ipaddr, sizeof(temp)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3986 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp); |
| 3987 | addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3988 | if (copy_to_user(to, &temp, addrlen)) |
| 3989 | return -EFAULT; |
| 3990 | to += addrlen ; |
| 3991 | cnt ++; |
| 3992 | if (cnt >= getaddrs.addr_num) break; |
| 3993 | } |
| 3994 | getaddrs.addr_num = cnt; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 3995 | if (put_user(len, optlen)) |
| 3996 | return -EFAULT; |
| 3997 | if (copy_to_user(optval, &getaddrs, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3998 | return -EFAULT; |
| 3999 | |
| 4000 | return 0; |
| 4001 | } |
| 4002 | |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4003 | static int sctp_getsockopt_peer_addrs(struct sock *sk, int len, |
| 4004 | char __user *optval, int __user *optlen) |
| 4005 | { |
| 4006 | struct sctp_association *asoc; |
| 4007 | struct list_head *pos; |
| 4008 | int cnt = 0; |
| 4009 | struct sctp_getaddrs getaddrs; |
| 4010 | struct sctp_transport *from; |
| 4011 | void __user *to; |
| 4012 | union sctp_addr temp; |
| 4013 | struct sctp_sock *sp = sctp_sk(sk); |
| 4014 | int addrlen; |
| 4015 | size_t space_left; |
| 4016 | int bytes_copied; |
| 4017 | |
| 4018 | if (len < sizeof(struct sctp_getaddrs)) |
| 4019 | return -EINVAL; |
| 4020 | |
| 4021 | if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs))) |
| 4022 | return -EFAULT; |
| 4023 | |
| 4024 | /* For UDP-style sockets, id specifies the association to query. */ |
| 4025 | asoc = sctp_id2assoc(sk, getaddrs.assoc_id); |
| 4026 | if (!asoc) |
| 4027 | return -EINVAL; |
| 4028 | |
| 4029 | to = optval + offsetof(struct sctp_getaddrs,addrs); |
Neil Horman | 186e234 | 2007-06-18 19:59:16 -0400 | [diff] [blame] | 4030 | space_left = len - offsetof(struct sctp_getaddrs,addrs); |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4031 | |
| 4032 | list_for_each(pos, &asoc->peer.transport_addr_list) { |
| 4033 | from = list_entry(pos, struct sctp_transport, transports); |
Al Viro | b3f5b3b | 2006-11-20 17:22:43 -0800 | [diff] [blame] | 4034 | memcpy(&temp, &from->ipaddr, sizeof(temp)); |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4035 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp); |
| 4036 | addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4037 | if (space_left < addrlen) |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4038 | return -ENOMEM; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4039 | if (copy_to_user(to, &temp, addrlen)) |
| 4040 | return -EFAULT; |
| 4041 | to += addrlen; |
| 4042 | cnt++; |
| 4043 | space_left -= addrlen; |
| 4044 | } |
| 4045 | |
| 4046 | if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) |
| 4047 | return -EFAULT; |
| 4048 | bytes_copied = ((char __user *)to) - optval; |
| 4049 | if (put_user(bytes_copied, optlen)) |
| 4050 | return -EFAULT; |
| 4051 | |
| 4052 | return 0; |
| 4053 | } |
| 4054 | |
| 4055 | static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len, |
| 4056 | char __user *optval, |
| 4057 | int __user *optlen) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4058 | { |
| 4059 | sctp_assoc_t id; |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4060 | struct list_head *pos; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4061 | struct sctp_bind_addr *bp; |
| 4062 | struct sctp_association *asoc; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4063 | struct sctp_sockaddr_entry *addr; |
| 4064 | rwlock_t *addr_lock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4065 | int cnt = 0; |
| 4066 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4067 | if (len < sizeof(sctp_assoc_t)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4068 | return -EINVAL; |
| 4069 | |
| 4070 | if (copy_from_user(&id, optval, sizeof(sctp_assoc_t))) |
| 4071 | return -EFAULT; |
| 4072 | |
| 4073 | /* |
| 4074 | * For UDP-style sockets, id specifies the association to query. |
| 4075 | * If the id field is set to the value '0' then the locally bound |
| 4076 | * addresses are returned without regard to any particular |
| 4077 | * association. |
| 4078 | */ |
| 4079 | if (0 == id) { |
| 4080 | bp = &sctp_sk(sk)->ep->base.bind_addr; |
| 4081 | addr_lock = &sctp_sk(sk)->ep->base.addr_lock; |
| 4082 | } else { |
| 4083 | asoc = sctp_id2assoc(sk, id); |
| 4084 | if (!asoc) |
| 4085 | return -EINVAL; |
| 4086 | bp = &asoc->base.bind_addr; |
| 4087 | addr_lock = &asoc->base.addr_lock; |
| 4088 | } |
| 4089 | |
| 4090 | sctp_read_lock(addr_lock); |
| 4091 | |
| 4092 | /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid |
| 4093 | * addresses from the global local address list. |
| 4094 | */ |
| 4095 | if (sctp_list_single_entry(&bp->address_list)) { |
| 4096 | addr = list_entry(bp->address_list.next, |
| 4097 | struct sctp_sockaddr_entry, list); |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4098 | if (sctp_is_any(&addr->a)) { |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4099 | rcu_read_lock(); |
| 4100 | list_for_each_entry_rcu(addr, |
| 4101 | &sctp_local_addr_list, list) { |
| 4102 | if (!addr->valid) |
| 4103 | continue; |
| 4104 | |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 4105 | if ((PF_INET == sk->sk_family) && |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4106 | (AF_INET6 == addr->a.sa.sa_family)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4107 | continue; |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4108 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4109 | cnt++; |
| 4110 | } |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4111 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4112 | } else { |
| 4113 | cnt = 1; |
| 4114 | } |
| 4115 | goto done; |
| 4116 | } |
| 4117 | |
| 4118 | list_for_each(pos, &bp->address_list) { |
| 4119 | cnt ++; |
| 4120 | } |
| 4121 | |
| 4122 | done: |
| 4123 | sctp_read_unlock(addr_lock); |
| 4124 | return cnt; |
| 4125 | } |
| 4126 | |
| 4127 | /* Helper function that copies local addresses to user and returns the number |
| 4128 | * of addresses copied. |
| 4129 | */ |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4130 | static int sctp_copy_laddrs_old(struct sock *sk, __u16 port, |
| 4131 | int max_addrs, void *to, |
| 4132 | int *bytes_copied) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4133 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4134 | struct sctp_sockaddr_entry *addr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4135 | union sctp_addr temp; |
| 4136 | int cnt = 0; |
| 4137 | int addrlen; |
| 4138 | |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4139 | rcu_read_lock(); |
| 4140 | list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) { |
| 4141 | if (!addr->valid) |
| 4142 | continue; |
| 4143 | |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 4144 | if ((PF_INET == sk->sk_family) && |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4145 | (AF_INET6 == addr->a.sa.sa_family)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4146 | continue; |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4147 | memcpy(&temp, &addr->a, sizeof(temp)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4148 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk), |
| 4149 | &temp); |
| 4150 | addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4151 | memcpy(to, &temp, addrlen); |
Sridhar Samudrala | 29c7cf9 | 2006-12-13 16:26:26 -0800 | [diff] [blame] | 4152 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4153 | to += addrlen; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4154 | *bytes_copied += addrlen; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4155 | cnt ++; |
| 4156 | if (cnt >= max_addrs) break; |
| 4157 | } |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4158 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4159 | |
| 4160 | return cnt; |
| 4161 | } |
| 4162 | |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4163 | static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to, |
| 4164 | size_t space_left, int *bytes_copied) |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4165 | { |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4166 | struct sctp_sockaddr_entry *addr; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4167 | union sctp_addr temp; |
| 4168 | int cnt = 0; |
| 4169 | int addrlen; |
| 4170 | |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4171 | rcu_read_lock(); |
| 4172 | list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) { |
| 4173 | if (!addr->valid) |
| 4174 | continue; |
| 4175 | |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 4176 | if ((PF_INET == sk->sk_family) && |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4177 | (AF_INET6 == addr->a.sa.sa_family)) |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4178 | continue; |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4179 | memcpy(&temp, &addr->a, sizeof(temp)); |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4180 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk), |
| 4181 | &temp); |
| 4182 | addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len; |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4183 | if (space_left < addrlen) { |
| 4184 | cnt = -ENOMEM; |
| 4185 | break; |
| 4186 | } |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4187 | memcpy(to, &temp, addrlen); |
Sridhar Samudrala | 29c7cf9 | 2006-12-13 16:26:26 -0800 | [diff] [blame] | 4188 | |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4189 | to += addrlen; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4190 | cnt ++; |
| 4191 | space_left -= addrlen; |
Vlad Yasevich | 3663c30 | 2007-07-03 12:43:12 -0400 | [diff] [blame] | 4192 | *bytes_copied += addrlen; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4193 | } |
Vlad Yasevich | 2930354 | 2007-09-16 16:02:12 -0700 | [diff] [blame^] | 4194 | rcu_read_unlock(); |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4195 | |
| 4196 | return cnt; |
| 4197 | } |
| 4198 | |
| 4199 | /* Old API for getting list of local addresses. Does not work for 32-bit |
| 4200 | * programs running on a 64-bit kernel |
| 4201 | */ |
| 4202 | static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len, |
| 4203 | char __user *optval, int __user *optlen) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4204 | { |
| 4205 | struct sctp_bind_addr *bp; |
| 4206 | struct sctp_association *asoc; |
| 4207 | struct list_head *pos; |
| 4208 | int cnt = 0; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4209 | struct sctp_getaddrs_old getaddrs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4210 | struct sctp_sockaddr_entry *addr; |
| 4211 | void __user *to; |
| 4212 | union sctp_addr temp; |
| 4213 | struct sctp_sock *sp = sctp_sk(sk); |
| 4214 | int addrlen; |
| 4215 | rwlock_t *addr_lock; |
| 4216 | int err = 0; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4217 | void *addrs; |
Vlad Yasevich | 70b57b8 | 2007-05-09 13:51:31 -0700 | [diff] [blame] | 4218 | void *buf; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4219 | int bytes_copied = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4220 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4221 | if (len < sizeof(struct sctp_getaddrs_old)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4222 | return -EINVAL; |
| 4223 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4224 | len = sizeof(struct sctp_getaddrs_old); |
| 4225 | if (copy_from_user(&getaddrs, optval, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4226 | return -EFAULT; |
| 4227 | |
| 4228 | if (getaddrs.addr_num <= 0) return -EINVAL; |
| 4229 | /* |
| 4230 | * For UDP-style sockets, id specifies the association to query. |
| 4231 | * If the id field is set to the value '0' then the locally bound |
| 4232 | * addresses are returned without regard to any particular |
| 4233 | * association. |
| 4234 | */ |
| 4235 | if (0 == getaddrs.assoc_id) { |
| 4236 | bp = &sctp_sk(sk)->ep->base.bind_addr; |
| 4237 | addr_lock = &sctp_sk(sk)->ep->base.addr_lock; |
| 4238 | } else { |
| 4239 | asoc = sctp_id2assoc(sk, getaddrs.assoc_id); |
| 4240 | if (!asoc) |
| 4241 | return -EINVAL; |
| 4242 | bp = &asoc->base.bind_addr; |
| 4243 | addr_lock = &asoc->base.addr_lock; |
| 4244 | } |
| 4245 | |
| 4246 | to = getaddrs.addrs; |
| 4247 | |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4248 | /* Allocate space for a local instance of packed array to hold all |
| 4249 | * the data. We store addresses here first and then put write them |
| 4250 | * to the user in one shot. |
| 4251 | */ |
| 4252 | addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num, |
| 4253 | GFP_KERNEL); |
| 4254 | if (!addrs) |
| 4255 | return -ENOMEM; |
| 4256 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4257 | sctp_read_lock(addr_lock); |
| 4258 | |
| 4259 | /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid |
| 4260 | * addresses from the global local address list. |
| 4261 | */ |
| 4262 | if (sctp_list_single_entry(&bp->address_list)) { |
| 4263 | addr = list_entry(bp->address_list.next, |
| 4264 | struct sctp_sockaddr_entry, list); |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4265 | if (sctp_is_any(&addr->a)) { |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4266 | cnt = sctp_copy_laddrs_old(sk, bp->port, |
| 4267 | getaddrs.addr_num, |
| 4268 | addrs, &bytes_copied); |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 4269 | goto copy_getaddrs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4270 | } |
| 4271 | } |
| 4272 | |
Vlad Yasevich | 70b57b8 | 2007-05-09 13:51:31 -0700 | [diff] [blame] | 4273 | buf = addrs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4274 | list_for_each(pos, &bp->address_list) { |
| 4275 | addr = list_entry(pos, struct sctp_sockaddr_entry, list); |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4276 | memcpy(&temp, &addr->a, sizeof(temp)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4277 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp); |
| 4278 | addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len; |
Vlad Yasevich | 70b57b8 | 2007-05-09 13:51:31 -0700 | [diff] [blame] | 4279 | memcpy(buf, &temp, addrlen); |
| 4280 | buf += addrlen; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4281 | bytes_copied += addrlen; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4282 | cnt ++; |
| 4283 | if (cnt >= getaddrs.addr_num) break; |
| 4284 | } |
| 4285 | |
| 4286 | copy_getaddrs: |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4287 | sctp_read_unlock(addr_lock); |
| 4288 | |
| 4289 | /* copy the entire address list into the user provided space */ |
| 4290 | if (copy_to_user(to, addrs, bytes_copied)) { |
| 4291 | err = -EFAULT; |
| 4292 | goto error; |
| 4293 | } |
| 4294 | |
| 4295 | /* copy the leading structure back to user */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4296 | getaddrs.addr_num = cnt; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4297 | if (copy_to_user(optval, &getaddrs, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4298 | err = -EFAULT; |
| 4299 | |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4300 | error: |
| 4301 | kfree(addrs); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4302 | return err; |
| 4303 | } |
| 4304 | |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4305 | static int sctp_getsockopt_local_addrs(struct sock *sk, int len, |
| 4306 | char __user *optval, int __user *optlen) |
| 4307 | { |
| 4308 | struct sctp_bind_addr *bp; |
| 4309 | struct sctp_association *asoc; |
| 4310 | struct list_head *pos; |
| 4311 | int cnt = 0; |
| 4312 | struct sctp_getaddrs getaddrs; |
| 4313 | struct sctp_sockaddr_entry *addr; |
| 4314 | void __user *to; |
| 4315 | union sctp_addr temp; |
| 4316 | struct sctp_sock *sp = sctp_sk(sk); |
| 4317 | int addrlen; |
| 4318 | rwlock_t *addr_lock; |
| 4319 | int err = 0; |
| 4320 | size_t space_left; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4321 | int bytes_copied = 0; |
| 4322 | void *addrs; |
Vlad Yasevich | 70b57b8 | 2007-05-09 13:51:31 -0700 | [diff] [blame] | 4323 | void *buf; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4324 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4325 | if (len < sizeof(struct sctp_getaddrs)) |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4326 | return -EINVAL; |
| 4327 | |
| 4328 | if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs))) |
| 4329 | return -EFAULT; |
| 4330 | |
| 4331 | /* |
| 4332 | * For UDP-style sockets, id specifies the association to query. |
| 4333 | * If the id field is set to the value '0' then the locally bound |
| 4334 | * addresses are returned without regard to any particular |
| 4335 | * association. |
| 4336 | */ |
| 4337 | if (0 == getaddrs.assoc_id) { |
| 4338 | bp = &sctp_sk(sk)->ep->base.bind_addr; |
| 4339 | addr_lock = &sctp_sk(sk)->ep->base.addr_lock; |
| 4340 | } else { |
| 4341 | asoc = sctp_id2assoc(sk, getaddrs.assoc_id); |
| 4342 | if (!asoc) |
| 4343 | return -EINVAL; |
| 4344 | bp = &asoc->base.bind_addr; |
| 4345 | addr_lock = &asoc->base.addr_lock; |
| 4346 | } |
| 4347 | |
| 4348 | to = optval + offsetof(struct sctp_getaddrs,addrs); |
Neil Horman | 186e234 | 2007-06-18 19:59:16 -0400 | [diff] [blame] | 4349 | space_left = len - offsetof(struct sctp_getaddrs,addrs); |
| 4350 | |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4351 | addrs = kmalloc(space_left, GFP_KERNEL); |
| 4352 | if (!addrs) |
| 4353 | return -ENOMEM; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4354 | |
| 4355 | sctp_read_lock(addr_lock); |
| 4356 | |
| 4357 | /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid |
| 4358 | * addresses from the global local address list. |
| 4359 | */ |
| 4360 | if (sctp_list_single_entry(&bp->address_list)) { |
| 4361 | addr = list_entry(bp->address_list.next, |
| 4362 | struct sctp_sockaddr_entry, list); |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4363 | if (sctp_is_any(&addr->a)) { |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4364 | cnt = sctp_copy_laddrs(sk, bp->port, addrs, |
| 4365 | space_left, &bytes_copied); |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4366 | if (cnt < 0) { |
| 4367 | err = cnt; |
Sebastian Siewior | d6f9fda | 2007-07-27 22:55:59 +0200 | [diff] [blame] | 4368 | goto error_lock; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4369 | } |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 4370 | goto copy_getaddrs; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4371 | } |
| 4372 | } |
| 4373 | |
Vlad Yasevich | 70b57b8 | 2007-05-09 13:51:31 -0700 | [diff] [blame] | 4374 | buf = addrs; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4375 | list_for_each(pos, &bp->address_list) { |
| 4376 | addr = list_entry(pos, struct sctp_sockaddr_entry, list); |
Al Viro | 6244be4 | 2006-11-20 17:21:44 -0800 | [diff] [blame] | 4377 | memcpy(&temp, &addr->a, sizeof(temp)); |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4378 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp); |
| 4379 | addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4380 | if (space_left < addrlen) { |
| 4381 | err = -ENOMEM; /*fixme: right error?*/ |
Sebastian Siewior | d6f9fda | 2007-07-27 22:55:59 +0200 | [diff] [blame] | 4382 | goto error_lock; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4383 | } |
Vlad Yasevich | 70b57b8 | 2007-05-09 13:51:31 -0700 | [diff] [blame] | 4384 | memcpy(buf, &temp, addrlen); |
| 4385 | buf += addrlen; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4386 | bytes_copied += addrlen; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4387 | cnt ++; |
| 4388 | space_left -= addrlen; |
| 4389 | } |
| 4390 | |
| 4391 | copy_getaddrs: |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4392 | sctp_read_unlock(addr_lock); |
| 4393 | |
| 4394 | if (copy_to_user(to, addrs, bytes_copied)) { |
| 4395 | err = -EFAULT; |
Sebastian Siewior | d6f9fda | 2007-07-27 22:55:59 +0200 | [diff] [blame] | 4396 | goto out; |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4397 | } |
Vlad Yasevich | fe979ac | 2007-05-23 11:11:37 -0400 | [diff] [blame] | 4398 | if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) { |
| 4399 | err = -EFAULT; |
Sebastian Siewior | d6f9fda | 2007-07-27 22:55:59 +0200 | [diff] [blame] | 4400 | goto out; |
Vlad Yasevich | fe979ac | 2007-05-23 11:11:37 -0400 | [diff] [blame] | 4401 | } |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4402 | if (put_user(bytes_copied, optlen)) |
Vlad Yasevich | fe979ac | 2007-05-23 11:11:37 -0400 | [diff] [blame] | 4403 | err = -EFAULT; |
Sebastian Siewior | d6f9fda | 2007-07-27 22:55:59 +0200 | [diff] [blame] | 4404 | |
| 4405 | goto out; |
| 4406 | |
| 4407 | error_lock: |
| 4408 | sctp_read_unlock(addr_lock); |
| 4409 | |
| 4410 | out: |
Vlad Yasevich | aad97f3 | 2007-04-28 21:09:04 -0700 | [diff] [blame] | 4411 | kfree(addrs); |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4412 | return err; |
| 4413 | } |
| 4414 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4415 | /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR) |
| 4416 | * |
| 4417 | * Requests that the local SCTP stack use the enclosed peer address as |
| 4418 | * the association primary. The enclosed address must be one of the |
| 4419 | * association peer's addresses. |
| 4420 | */ |
| 4421 | static int sctp_getsockopt_primary_addr(struct sock *sk, int len, |
| 4422 | char __user *optval, int __user *optlen) |
| 4423 | { |
| 4424 | struct sctp_prim prim; |
| 4425 | struct sctp_association *asoc; |
| 4426 | struct sctp_sock *sp = sctp_sk(sk); |
| 4427 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4428 | if (len < sizeof(struct sctp_prim)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4429 | return -EINVAL; |
| 4430 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4431 | len = sizeof(struct sctp_prim); |
| 4432 | |
| 4433 | if (copy_from_user(&prim, optval, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4434 | return -EFAULT; |
| 4435 | |
| 4436 | asoc = sctp_id2assoc(sk, prim.ssp_assoc_id); |
| 4437 | if (!asoc) |
| 4438 | return -EINVAL; |
| 4439 | |
| 4440 | if (!asoc->peer.primary_path) |
| 4441 | return -ENOTCONN; |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 4442 | |
Al Viro | 8cec6b8 | 2006-11-20 17:23:01 -0800 | [diff] [blame] | 4443 | memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr, |
| 4444 | asoc->peer.primary_path->af_specific->sockaddr_len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4445 | |
| 4446 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, |
| 4447 | (union sctp_addr *)&prim.ssp_addr); |
| 4448 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4449 | if (put_user(len, optlen)) |
| 4450 | return -EFAULT; |
| 4451 | if (copy_to_user(optval, &prim, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4452 | return -EFAULT; |
| 4453 | |
| 4454 | return 0; |
| 4455 | } |
| 4456 | |
| 4457 | /* |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 4458 | * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4459 | * |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 4460 | * Requests that the local endpoint set the specified Adaptation Layer |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4461 | * Indication parameter for all future INIT and INIT-ACK exchanges. |
| 4462 | */ |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 4463 | static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4464 | char __user *optval, int __user *optlen) |
| 4465 | { |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 4466 | struct sctp_setadaptation adaptation; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4467 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4468 | if (len < sizeof(struct sctp_setadaptation)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4469 | return -EINVAL; |
| 4470 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4471 | len = sizeof(struct sctp_setadaptation); |
| 4472 | |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 4473 | adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4474 | |
| 4475 | if (put_user(len, optlen)) |
| 4476 | return -EFAULT; |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 4477 | if (copy_to_user(optval, &adaptation, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4478 | return -EFAULT; |
Ivan Skytte Jorgensen | a1ab358 | 2005-10-28 15:33:24 -0700 | [diff] [blame] | 4479 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4480 | return 0; |
| 4481 | } |
| 4482 | |
| 4483 | /* |
| 4484 | * |
| 4485 | * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM) |
| 4486 | * |
| 4487 | * Applications that wish to use the sendto() system call may wish to |
| 4488 | * specify a default set of parameters that would normally be supplied |
| 4489 | * through the inclusion of ancillary data. This socket option allows |
| 4490 | * such an application to set the default sctp_sndrcvinfo structure. |
| 4491 | |
| 4492 | |
| 4493 | * The application that wishes to use this socket option simply passes |
| 4494 | * in to this call the sctp_sndrcvinfo structure defined in Section |
| 4495 | * 5.2.2) The input parameters accepted by this call include |
| 4496 | * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context, |
| 4497 | * sinfo_timetolive. The user must provide the sinfo_assoc_id field in |
| 4498 | * to this call if the caller is using the UDP model. |
| 4499 | * |
| 4500 | * For getsockopt, it get the default sctp_sndrcvinfo structure. |
| 4501 | */ |
| 4502 | static int sctp_getsockopt_default_send_param(struct sock *sk, |
| 4503 | int len, char __user *optval, |
| 4504 | int __user *optlen) |
| 4505 | { |
| 4506 | struct sctp_sndrcvinfo info; |
| 4507 | struct sctp_association *asoc; |
| 4508 | struct sctp_sock *sp = sctp_sk(sk); |
| 4509 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4510 | if (len < sizeof(struct sctp_sndrcvinfo)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4511 | return -EINVAL; |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4512 | |
| 4513 | len = sizeof(struct sctp_sndrcvinfo); |
| 4514 | |
| 4515 | if (copy_from_user(&info, optval, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4516 | return -EFAULT; |
| 4517 | |
| 4518 | asoc = sctp_id2assoc(sk, info.sinfo_assoc_id); |
| 4519 | if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP)) |
| 4520 | return -EINVAL; |
| 4521 | |
| 4522 | if (asoc) { |
| 4523 | info.sinfo_stream = asoc->default_stream; |
| 4524 | info.sinfo_flags = asoc->default_flags; |
| 4525 | info.sinfo_ppid = asoc->default_ppid; |
| 4526 | info.sinfo_context = asoc->default_context; |
| 4527 | info.sinfo_timetolive = asoc->default_timetolive; |
| 4528 | } else { |
| 4529 | info.sinfo_stream = sp->default_stream; |
| 4530 | info.sinfo_flags = sp->default_flags; |
| 4531 | info.sinfo_ppid = sp->default_ppid; |
| 4532 | info.sinfo_context = sp->default_context; |
| 4533 | info.sinfo_timetolive = sp->default_timetolive; |
| 4534 | } |
| 4535 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4536 | if (put_user(len, optlen)) |
| 4537 | return -EFAULT; |
| 4538 | if (copy_to_user(optval, &info, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4539 | return -EFAULT; |
| 4540 | |
| 4541 | return 0; |
| 4542 | } |
| 4543 | |
| 4544 | /* |
| 4545 | * |
| 4546 | * 7.1.5 SCTP_NODELAY |
| 4547 | * |
| 4548 | * Turn on/off any Nagle-like algorithm. This means that packets are |
| 4549 | * generally sent as soon as possible and no unnecessary delays are |
| 4550 | * introduced, at the cost of more packets in the network. Expects an |
| 4551 | * integer boolean flag. |
| 4552 | */ |
| 4553 | |
| 4554 | static int sctp_getsockopt_nodelay(struct sock *sk, int len, |
| 4555 | char __user *optval, int __user *optlen) |
| 4556 | { |
| 4557 | int val; |
| 4558 | |
| 4559 | if (len < sizeof(int)) |
| 4560 | return -EINVAL; |
| 4561 | |
| 4562 | len = sizeof(int); |
| 4563 | val = (sctp_sk(sk)->nodelay == 1); |
| 4564 | if (put_user(len, optlen)) |
| 4565 | return -EFAULT; |
| 4566 | if (copy_to_user(optval, &val, len)) |
| 4567 | return -EFAULT; |
| 4568 | return 0; |
| 4569 | } |
| 4570 | |
| 4571 | /* |
| 4572 | * |
| 4573 | * 7.1.1 SCTP_RTOINFO |
| 4574 | * |
| 4575 | * The protocol parameters used to initialize and bound retransmission |
| 4576 | * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access |
| 4577 | * and modify these parameters. |
| 4578 | * All parameters are time values, in milliseconds. A value of 0, when |
| 4579 | * modifying the parameters, indicates that the current value should not |
| 4580 | * be changed. |
| 4581 | * |
| 4582 | */ |
| 4583 | static int sctp_getsockopt_rtoinfo(struct sock *sk, int len, |
| 4584 | char __user *optval, |
| 4585 | int __user *optlen) { |
| 4586 | struct sctp_rtoinfo rtoinfo; |
| 4587 | struct sctp_association *asoc; |
| 4588 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4589 | if (len < sizeof (struct sctp_rtoinfo)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4590 | return -EINVAL; |
| 4591 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4592 | len = sizeof(struct sctp_rtoinfo); |
| 4593 | |
| 4594 | if (copy_from_user(&rtoinfo, optval, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4595 | return -EFAULT; |
| 4596 | |
| 4597 | asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id); |
| 4598 | |
| 4599 | if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP)) |
| 4600 | return -EINVAL; |
| 4601 | |
| 4602 | /* Values corresponding to the specific association. */ |
| 4603 | if (asoc) { |
| 4604 | rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial); |
| 4605 | rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max); |
| 4606 | rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min); |
| 4607 | } else { |
| 4608 | /* Values corresponding to the endpoint. */ |
| 4609 | struct sctp_sock *sp = sctp_sk(sk); |
| 4610 | |
| 4611 | rtoinfo.srto_initial = sp->rtoinfo.srto_initial; |
| 4612 | rtoinfo.srto_max = sp->rtoinfo.srto_max; |
| 4613 | rtoinfo.srto_min = sp->rtoinfo.srto_min; |
| 4614 | } |
| 4615 | |
| 4616 | if (put_user(len, optlen)) |
| 4617 | return -EFAULT; |
| 4618 | |
| 4619 | if (copy_to_user(optval, &rtoinfo, len)) |
| 4620 | return -EFAULT; |
| 4621 | |
| 4622 | return 0; |
| 4623 | } |
| 4624 | |
| 4625 | /* |
| 4626 | * |
| 4627 | * 7.1.2 SCTP_ASSOCINFO |
| 4628 | * |
Michael Opdenacker | 59c5159 | 2007-05-09 08:57:56 +0200 | [diff] [blame] | 4629 | * This option is used to tune the maximum retransmission attempts |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4630 | * of the association. |
| 4631 | * Returns an error if the new association retransmission value is |
| 4632 | * greater than the sum of the retransmission value of the peer. |
| 4633 | * See [SCTP] for more information. |
| 4634 | * |
| 4635 | */ |
| 4636 | static int sctp_getsockopt_associnfo(struct sock *sk, int len, |
| 4637 | char __user *optval, |
| 4638 | int __user *optlen) |
| 4639 | { |
| 4640 | |
| 4641 | struct sctp_assocparams assocparams; |
| 4642 | struct sctp_association *asoc; |
| 4643 | struct list_head *pos; |
| 4644 | int cnt = 0; |
| 4645 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4646 | if (len < sizeof (struct sctp_assocparams)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4647 | return -EINVAL; |
| 4648 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4649 | len = sizeof(struct sctp_assocparams); |
| 4650 | |
| 4651 | if (copy_from_user(&assocparams, optval, len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4652 | return -EFAULT; |
| 4653 | |
| 4654 | asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id); |
| 4655 | |
| 4656 | if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP)) |
| 4657 | return -EINVAL; |
| 4658 | |
| 4659 | /* Values correspoinding to the specific association */ |
Vladislav Yasevich | 1733721 | 2005-04-28 11:57:54 -0700 | [diff] [blame] | 4660 | if (asoc) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4661 | assocparams.sasoc_asocmaxrxt = asoc->max_retrans; |
| 4662 | assocparams.sasoc_peer_rwnd = asoc->peer.rwnd; |
| 4663 | assocparams.sasoc_local_rwnd = asoc->a_rwnd; |
| 4664 | assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec |
| 4665 | * 1000) + |
| 4666 | (asoc->cookie_life.tv_usec |
| 4667 | / 1000); |
| 4668 | |
| 4669 | list_for_each(pos, &asoc->peer.transport_addr_list) { |
| 4670 | cnt ++; |
| 4671 | } |
| 4672 | |
| 4673 | assocparams.sasoc_number_peer_destinations = cnt; |
| 4674 | } else { |
| 4675 | /* Values corresponding to the endpoint */ |
| 4676 | struct sctp_sock *sp = sctp_sk(sk); |
| 4677 | |
| 4678 | assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt; |
| 4679 | assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd; |
| 4680 | assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd; |
| 4681 | assocparams.sasoc_cookie_life = |
| 4682 | sp->assocparams.sasoc_cookie_life; |
| 4683 | assocparams.sasoc_number_peer_destinations = |
| 4684 | sp->assocparams. |
| 4685 | sasoc_number_peer_destinations; |
| 4686 | } |
| 4687 | |
| 4688 | if (put_user(len, optlen)) |
| 4689 | return -EFAULT; |
| 4690 | |
| 4691 | if (copy_to_user(optval, &assocparams, len)) |
| 4692 | return -EFAULT; |
| 4693 | |
| 4694 | return 0; |
| 4695 | } |
| 4696 | |
| 4697 | /* |
| 4698 | * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR) |
| 4699 | * |
| 4700 | * This socket option is a boolean flag which turns on or off mapped V4 |
| 4701 | * addresses. If this option is turned on and the socket is type |
| 4702 | * PF_INET6, then IPv4 addresses will be mapped to V6 representation. |
| 4703 | * If this option is turned off, then no mapping will be done of V4 |
| 4704 | * addresses and a user will receive both PF_INET6 and PF_INET type |
| 4705 | * addresses on the socket. |
| 4706 | */ |
| 4707 | static int sctp_getsockopt_mappedv4(struct sock *sk, int len, |
| 4708 | char __user *optval, int __user *optlen) |
| 4709 | { |
| 4710 | int val; |
| 4711 | struct sctp_sock *sp = sctp_sk(sk); |
| 4712 | |
| 4713 | if (len < sizeof(int)) |
| 4714 | return -EINVAL; |
| 4715 | |
| 4716 | len = sizeof(int); |
| 4717 | val = sp->v4mapped; |
| 4718 | if (put_user(len, optlen)) |
| 4719 | return -EFAULT; |
| 4720 | if (copy_to_user(optval, &val, len)) |
| 4721 | return -EFAULT; |
| 4722 | |
| 4723 | return 0; |
| 4724 | } |
| 4725 | |
| 4726 | /* |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 4727 | * 7.1.29. Set or Get the default context (SCTP_CONTEXT) |
| 4728 | * (chapter and verse is quoted at sctp_setsockopt_context()) |
| 4729 | */ |
| 4730 | static int sctp_getsockopt_context(struct sock *sk, int len, |
| 4731 | char __user *optval, int __user *optlen) |
| 4732 | { |
| 4733 | struct sctp_assoc_value params; |
| 4734 | struct sctp_sock *sp; |
| 4735 | struct sctp_association *asoc; |
| 4736 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4737 | if (len < sizeof(struct sctp_assoc_value)) |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 4738 | return -EINVAL; |
| 4739 | |
Neil Horman | 408f22e | 2007-06-16 14:03:45 -0400 | [diff] [blame] | 4740 | len = sizeof(struct sctp_assoc_value); |
| 4741 | |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 4742 | if (copy_from_user(¶ms, optval, len)) |
| 4743 | return -EFAULT; |
| 4744 | |
| 4745 | sp = sctp_sk(sk); |
| 4746 | |
| 4747 | if (params.assoc_id != 0) { |
| 4748 | asoc = sctp_id2assoc(sk, params.assoc_id); |
| 4749 | if (!asoc) |
| 4750 | return -EINVAL; |
| 4751 | params.assoc_value = asoc->default_rcv_context; |
| 4752 | } else { |
| 4753 | params.assoc_value = sp->default_rcv_context; |
| 4754 | } |
| 4755 | |
| 4756 | if (put_user(len, optlen)) |
| 4757 | return -EFAULT; |
| 4758 | if (copy_to_user(optval, ¶ms, len)) |
| 4759 | return -EFAULT; |
| 4760 | |
| 4761 | return 0; |
| 4762 | } |
| 4763 | |
| 4764 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4765 | * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG) |
| 4766 | * |
| 4767 | * This socket option specifies the maximum size to put in any outgoing |
| 4768 | * SCTP chunk. If a message is larger than this size it will be |
| 4769 | * fragmented by SCTP into the specified size. Note that the underlying |
| 4770 | * SCTP implementation may fragment into smaller sized chunks when the |
| 4771 | * PMTU of the underlying association is smaller than the value set by |
| 4772 | * the user. |
| 4773 | */ |
| 4774 | static int sctp_getsockopt_maxseg(struct sock *sk, int len, |
| 4775 | char __user *optval, int __user *optlen) |
| 4776 | { |
| 4777 | int val; |
| 4778 | |
| 4779 | if (len < sizeof(int)) |
| 4780 | return -EINVAL; |
| 4781 | |
| 4782 | len = sizeof(int); |
| 4783 | |
| 4784 | val = sctp_sk(sk)->user_frag; |
| 4785 | if (put_user(len, optlen)) |
| 4786 | return -EFAULT; |
| 4787 | if (copy_to_user(optval, &val, len)) |
| 4788 | return -EFAULT; |
| 4789 | |
| 4790 | return 0; |
| 4791 | } |
| 4792 | |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 4793 | /* |
| 4794 | * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE) |
| 4795 | * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave()) |
| 4796 | */ |
| 4797 | static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len, |
| 4798 | char __user *optval, int __user *optlen) |
| 4799 | { |
| 4800 | int val; |
| 4801 | |
| 4802 | if (len < sizeof(int)) |
| 4803 | return -EINVAL; |
| 4804 | |
| 4805 | len = sizeof(int); |
| 4806 | |
| 4807 | val = sctp_sk(sk)->frag_interleave; |
| 4808 | if (put_user(len, optlen)) |
| 4809 | return -EFAULT; |
| 4810 | if (copy_to_user(optval, &val, len)) |
| 4811 | return -EFAULT; |
| 4812 | |
| 4813 | return 0; |
| 4814 | } |
| 4815 | |
Vlad Yasevich | d49d91d | 2007-03-23 11:32:00 -0700 | [diff] [blame] | 4816 | /* |
| 4817 | * 7.1.25. Set or Get the sctp partial delivery point |
| 4818 | * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point()) |
| 4819 | */ |
| 4820 | static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len, |
| 4821 | char __user *optval, |
| 4822 | int __user *optlen) |
| 4823 | { |
YOSHIFUJI Hideaki | 9cbcbf4 | 2007-07-19 10:44:50 +0900 | [diff] [blame] | 4824 | u32 val; |
Vlad Yasevich | d49d91d | 2007-03-23 11:32:00 -0700 | [diff] [blame] | 4825 | |
| 4826 | if (len < sizeof(u32)) |
| 4827 | return -EINVAL; |
| 4828 | |
| 4829 | len = sizeof(u32); |
| 4830 | |
| 4831 | val = sctp_sk(sk)->pd_point; |
| 4832 | if (put_user(len, optlen)) |
| 4833 | return -EFAULT; |
| 4834 | if (copy_to_user(optval, &val, len)) |
| 4835 | return -EFAULT; |
| 4836 | |
| 4837 | return -ENOTSUPP; |
| 4838 | } |
| 4839 | |
Vlad Yasevich | 7033157 | 2007-03-23 11:34:36 -0700 | [diff] [blame] | 4840 | /* |
| 4841 | * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST) |
| 4842 | * (chapter and verse is quoted at sctp_setsockopt_maxburst()) |
| 4843 | */ |
| 4844 | static int sctp_getsockopt_maxburst(struct sock *sk, int len, |
| 4845 | char __user *optval, |
| 4846 | int __user *optlen) |
| 4847 | { |
YOSHIFUJI Hideaki | 9cbcbf4 | 2007-07-19 10:44:50 +0900 | [diff] [blame] | 4848 | int val; |
Vlad Yasevich | 7033157 | 2007-03-23 11:34:36 -0700 | [diff] [blame] | 4849 | |
| 4850 | if (len < sizeof(int)) |
| 4851 | return -EINVAL; |
| 4852 | |
| 4853 | len = sizeof(int); |
| 4854 | |
| 4855 | val = sctp_sk(sk)->max_burst; |
| 4856 | if (put_user(len, optlen)) |
| 4857 | return -EFAULT; |
| 4858 | if (copy_to_user(optval, &val, len)) |
| 4859 | return -EFAULT; |
| 4860 | |
| 4861 | return -ENOTSUPP; |
| 4862 | } |
| 4863 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4864 | SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname, |
| 4865 | char __user *optval, int __user *optlen) |
| 4866 | { |
| 4867 | int retval = 0; |
| 4868 | int len; |
| 4869 | |
Frank Filz | 3f7a87d | 2005-06-20 13:14:57 -0700 | [diff] [blame] | 4870 | SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n", |
| 4871 | sk, optname); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4872 | |
| 4873 | /* I can hardly begin to describe how wrong this is. This is |
| 4874 | * so broken as to be worse than useless. The API draft |
| 4875 | * REALLY is NOT helpful here... I am not convinced that the |
| 4876 | * semantics of getsockopt() with a level OTHER THAN SOL_SCTP |
| 4877 | * are at all well-founded. |
| 4878 | */ |
| 4879 | if (level != SOL_SCTP) { |
| 4880 | struct sctp_af *af = sctp_sk(sk)->pf->af; |
| 4881 | |
| 4882 | retval = af->getsockopt(sk, level, optname, optval, optlen); |
| 4883 | return retval; |
| 4884 | } |
| 4885 | |
| 4886 | if (get_user(len, optlen)) |
| 4887 | return -EFAULT; |
| 4888 | |
| 4889 | sctp_lock_sock(sk); |
| 4890 | |
| 4891 | switch (optname) { |
| 4892 | case SCTP_STATUS: |
| 4893 | retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen); |
| 4894 | break; |
| 4895 | case SCTP_DISABLE_FRAGMENTS: |
| 4896 | retval = sctp_getsockopt_disable_fragments(sk, len, optval, |
| 4897 | optlen); |
| 4898 | break; |
| 4899 | case SCTP_EVENTS: |
| 4900 | retval = sctp_getsockopt_events(sk, len, optval, optlen); |
| 4901 | break; |
| 4902 | case SCTP_AUTOCLOSE: |
| 4903 | retval = sctp_getsockopt_autoclose(sk, len, optval, optlen); |
| 4904 | break; |
| 4905 | case SCTP_SOCKOPT_PEELOFF: |
| 4906 | retval = sctp_getsockopt_peeloff(sk, len, optval, optlen); |
| 4907 | break; |
| 4908 | case SCTP_PEER_ADDR_PARAMS: |
| 4909 | retval = sctp_getsockopt_peer_addr_params(sk, len, optval, |
| 4910 | optlen); |
| 4911 | break; |
Frank Filz | 7708610 | 2005-12-22 11:37:30 -0800 | [diff] [blame] | 4912 | case SCTP_DELAYED_ACK_TIME: |
| 4913 | retval = sctp_getsockopt_delayed_ack_time(sk, len, optval, |
| 4914 | optlen); |
| 4915 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4916 | case SCTP_INITMSG: |
| 4917 | retval = sctp_getsockopt_initmsg(sk, len, optval, optlen); |
| 4918 | break; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4919 | case SCTP_GET_PEER_ADDRS_NUM_OLD: |
| 4920 | retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval, |
| 4921 | optlen); |
| 4922 | break; |
| 4923 | case SCTP_GET_LOCAL_ADDRS_NUM_OLD: |
| 4924 | retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval, |
| 4925 | optlen); |
| 4926 | break; |
| 4927 | case SCTP_GET_PEER_ADDRS_OLD: |
| 4928 | retval = sctp_getsockopt_peer_addrs_old(sk, len, optval, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4929 | optlen); |
| 4930 | break; |
Ivan Skytte Jørgensen | 5fe467e | 2005-10-06 21:36:17 -0700 | [diff] [blame] | 4931 | case SCTP_GET_LOCAL_ADDRS_OLD: |
| 4932 | retval = sctp_getsockopt_local_addrs_old(sk, len, optval, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4933 | optlen); |
| 4934 | break; |
| 4935 | case SCTP_GET_PEER_ADDRS: |
| 4936 | retval = sctp_getsockopt_peer_addrs(sk, len, optval, |
| 4937 | optlen); |
| 4938 | break; |
| 4939 | case SCTP_GET_LOCAL_ADDRS: |
| 4940 | retval = sctp_getsockopt_local_addrs(sk, len, optval, |
| 4941 | optlen); |
| 4942 | break; |
| 4943 | case SCTP_DEFAULT_SEND_PARAM: |
| 4944 | retval = sctp_getsockopt_default_send_param(sk, len, |
| 4945 | optval, optlen); |
| 4946 | break; |
| 4947 | case SCTP_PRIMARY_ADDR: |
| 4948 | retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen); |
| 4949 | break; |
| 4950 | case SCTP_NODELAY: |
| 4951 | retval = sctp_getsockopt_nodelay(sk, len, optval, optlen); |
| 4952 | break; |
| 4953 | case SCTP_RTOINFO: |
| 4954 | retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen); |
| 4955 | break; |
| 4956 | case SCTP_ASSOCINFO: |
| 4957 | retval = sctp_getsockopt_associnfo(sk, len, optval, optlen); |
| 4958 | break; |
| 4959 | case SCTP_I_WANT_MAPPED_V4_ADDR: |
| 4960 | retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen); |
| 4961 | break; |
| 4962 | case SCTP_MAXSEG: |
| 4963 | retval = sctp_getsockopt_maxseg(sk, len, optval, optlen); |
| 4964 | break; |
| 4965 | case SCTP_GET_PEER_ADDR_INFO: |
| 4966 | retval = sctp_getsockopt_peer_addr_info(sk, len, optval, |
| 4967 | optlen); |
| 4968 | break; |
Ivan Skytte Jorgensen | 0f3fffd | 2006-12-20 16:07:04 -0800 | [diff] [blame] | 4969 | case SCTP_ADAPTATION_LAYER: |
| 4970 | retval = sctp_getsockopt_adaptation_layer(sk, len, optval, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4971 | optlen); |
| 4972 | break; |
Ivan Skytte Jorgensen | 6ab792f | 2006-12-13 16:34:22 -0800 | [diff] [blame] | 4973 | case SCTP_CONTEXT: |
| 4974 | retval = sctp_getsockopt_context(sk, len, optval, optlen); |
| 4975 | break; |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 4976 | case SCTP_FRAGMENT_INTERLEAVE: |
| 4977 | retval = sctp_getsockopt_fragment_interleave(sk, len, optval, |
| 4978 | optlen); |
| 4979 | break; |
Vlad Yasevich | d49d91d | 2007-03-23 11:32:00 -0700 | [diff] [blame] | 4980 | case SCTP_PARTIAL_DELIVERY_POINT: |
| 4981 | retval = sctp_getsockopt_partial_delivery_point(sk, len, optval, |
| 4982 | optlen); |
| 4983 | break; |
Vlad Yasevich | 7033157 | 2007-03-23 11:34:36 -0700 | [diff] [blame] | 4984 | case SCTP_MAX_BURST: |
| 4985 | retval = sctp_getsockopt_maxburst(sk, len, optval, optlen); |
| 4986 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4987 | default: |
| 4988 | retval = -ENOPROTOOPT; |
| 4989 | break; |
Stephen Hemminger | 3ff50b7 | 2007-04-20 17:09:22 -0700 | [diff] [blame] | 4990 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4991 | |
| 4992 | sctp_release_sock(sk); |
| 4993 | return retval; |
| 4994 | } |
| 4995 | |
| 4996 | static void sctp_hash(struct sock *sk) |
| 4997 | { |
| 4998 | /* STUB */ |
| 4999 | } |
| 5000 | |
| 5001 | static void sctp_unhash(struct sock *sk) |
| 5002 | { |
| 5003 | /* STUB */ |
| 5004 | } |
| 5005 | |
| 5006 | /* Check if port is acceptable. Possibly find first available port. |
| 5007 | * |
| 5008 | * The port hash table (contained in the 'global' SCTP protocol storage |
| 5009 | * returned by struct sctp_protocol *sctp_get_protocol()). The hash |
| 5010 | * table is an array of 4096 lists (sctp_bind_hashbucket). Each |
| 5011 | * list (the list number is the port number hashed out, so as you |
| 5012 | * would expect from a hash function, all the ports in a given list have |
| 5013 | * such a number that hashes out to the same list number; you were |
| 5014 | * expecting that, right?); so each list has a set of ports, with a |
| 5015 | * link to the socket (struct sock) that uses it, the port number and |
| 5016 | * a fastreuse flag (FIXME: NPI ipg). |
| 5017 | */ |
| 5018 | static struct sctp_bind_bucket *sctp_bucket_create( |
| 5019 | struct sctp_bind_hashbucket *head, unsigned short snum); |
| 5020 | |
| 5021 | static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr) |
| 5022 | { |
| 5023 | struct sctp_bind_hashbucket *head; /* hash list */ |
| 5024 | struct sctp_bind_bucket *pp; /* hash list port iterator */ |
| 5025 | unsigned short snum; |
| 5026 | int ret; |
| 5027 | |
Al Viro | 04afd8b | 2006-11-20 17:02:01 -0800 | [diff] [blame] | 5028 | snum = ntohs(addr->v4.sin_port); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5029 | |
| 5030 | SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum); |
| 5031 | sctp_local_bh_disable(); |
| 5032 | |
| 5033 | if (snum == 0) { |
| 5034 | /* Search for an available port. |
| 5035 | * |
| 5036 | * 'sctp_port_rover' was the last port assigned, so |
| 5037 | * we start to search from 'sctp_port_rover + |
| 5038 | * 1'. What we do is first check if port 'rover' is |
| 5039 | * already in the hash table; if not, we use that; if |
| 5040 | * it is, we try next. |
| 5041 | */ |
| 5042 | int low = sysctl_local_port_range[0]; |
| 5043 | int high = sysctl_local_port_range[1]; |
| 5044 | int remaining = (high - low) + 1; |
| 5045 | int rover; |
| 5046 | int index; |
| 5047 | |
| 5048 | sctp_spin_lock(&sctp_port_alloc_lock); |
| 5049 | rover = sctp_port_rover; |
| 5050 | do { |
| 5051 | rover++; |
| 5052 | if ((rover < low) || (rover > high)) |
| 5053 | rover = low; |
| 5054 | index = sctp_phashfn(rover); |
| 5055 | head = &sctp_port_hashtable[index]; |
| 5056 | sctp_spin_lock(&head->lock); |
| 5057 | for (pp = head->chain; pp; pp = pp->next) |
| 5058 | if (pp->port == rover) |
| 5059 | goto next; |
| 5060 | break; |
| 5061 | next: |
| 5062 | sctp_spin_unlock(&head->lock); |
| 5063 | } while (--remaining > 0); |
| 5064 | sctp_port_rover = rover; |
| 5065 | sctp_spin_unlock(&sctp_port_alloc_lock); |
| 5066 | |
| 5067 | /* Exhausted local port range during search? */ |
| 5068 | ret = 1; |
| 5069 | if (remaining <= 0) |
| 5070 | goto fail; |
| 5071 | |
| 5072 | /* OK, here is the one we will use. HEAD (the port |
| 5073 | * hash table list entry) is non-NULL and we hold it's |
| 5074 | * mutex. |
| 5075 | */ |
| 5076 | snum = rover; |
| 5077 | } else { |
| 5078 | /* We are given an specific port number; we verify |
| 5079 | * that it is not being used. If it is used, we will |
| 5080 | * exahust the search in the hash list corresponding |
| 5081 | * to the port number (snum) - we detect that with the |
| 5082 | * port iterator, pp being NULL. |
| 5083 | */ |
| 5084 | head = &sctp_port_hashtable[sctp_phashfn(snum)]; |
| 5085 | sctp_spin_lock(&head->lock); |
| 5086 | for (pp = head->chain; pp; pp = pp->next) { |
| 5087 | if (pp->port == snum) |
| 5088 | goto pp_found; |
| 5089 | } |
| 5090 | } |
| 5091 | pp = NULL; |
| 5092 | goto pp_not_found; |
| 5093 | pp_found: |
| 5094 | if (!hlist_empty(&pp->owner)) { |
| 5095 | /* We had a port hash table hit - there is an |
| 5096 | * available port (pp != NULL) and it is being |
| 5097 | * used by other socket (pp->owner not empty); that other |
| 5098 | * socket is going to be sk2. |
| 5099 | */ |
| 5100 | int reuse = sk->sk_reuse; |
| 5101 | struct sock *sk2; |
| 5102 | struct hlist_node *node; |
| 5103 | |
| 5104 | SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n"); |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5105 | if (pp->fastreuse && sk->sk_reuse && |
| 5106 | sk->sk_state != SCTP_SS_LISTENING) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5107 | goto success; |
| 5108 | |
| 5109 | /* Run through the list of sockets bound to the port |
| 5110 | * (pp->port) [via the pointers bind_next and |
| 5111 | * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one, |
| 5112 | * we get the endpoint they describe and run through |
| 5113 | * the endpoint's list of IP (v4 or v6) addresses, |
| 5114 | * comparing each of the addresses with the address of |
| 5115 | * the socket sk. If we find a match, then that means |
| 5116 | * that this port/socket (sk) combination are already |
| 5117 | * in an endpoint. |
| 5118 | */ |
| 5119 | sk_for_each_bound(sk2, node, &pp->owner) { |
| 5120 | struct sctp_endpoint *ep2; |
| 5121 | ep2 = sctp_sk(sk2)->ep; |
| 5122 | |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5123 | if (reuse && sk2->sk_reuse && |
| 5124 | sk2->sk_state != SCTP_SS_LISTENING) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5125 | continue; |
| 5126 | |
Al Viro | 7e1e4a2 | 2006-11-20 17:05:43 -0800 | [diff] [blame] | 5127 | if (sctp_bind_addr_match(&ep2->base.bind_addr, addr, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5128 | sctp_sk(sk))) { |
| 5129 | ret = (long)sk2; |
| 5130 | goto fail_unlock; |
| 5131 | } |
| 5132 | } |
| 5133 | SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n"); |
| 5134 | } |
| 5135 | pp_not_found: |
| 5136 | /* If there was a hash table miss, create a new port. */ |
| 5137 | ret = 1; |
| 5138 | if (!pp && !(pp = sctp_bucket_create(head, snum))) |
| 5139 | goto fail_unlock; |
| 5140 | |
| 5141 | /* In either case (hit or miss), make sure fastreuse is 1 only |
| 5142 | * if sk->sk_reuse is too (that is, if the caller requested |
| 5143 | * SO_REUSEADDR on this socket -sk-). |
| 5144 | */ |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5145 | if (hlist_empty(&pp->owner)) { |
| 5146 | if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING) |
| 5147 | pp->fastreuse = 1; |
| 5148 | else |
| 5149 | pp->fastreuse = 0; |
| 5150 | } else if (pp->fastreuse && |
| 5151 | (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5152 | pp->fastreuse = 0; |
| 5153 | |
| 5154 | /* We are set, so fill up all the data in the hash table |
| 5155 | * entry, tie the socket list information with the rest of the |
| 5156 | * sockets FIXME: Blurry, NPI (ipg). |
| 5157 | */ |
| 5158 | success: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5159 | if (!sctp_sk(sk)->bind_hash) { |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5160 | inet_sk(sk)->num = snum; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5161 | sk_add_bind_node(sk, &pp->owner); |
| 5162 | sctp_sk(sk)->bind_hash = pp; |
| 5163 | } |
| 5164 | ret = 0; |
| 5165 | |
| 5166 | fail_unlock: |
| 5167 | sctp_spin_unlock(&head->lock); |
| 5168 | |
| 5169 | fail: |
| 5170 | sctp_local_bh_enable(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5171 | return ret; |
| 5172 | } |
| 5173 | |
| 5174 | /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral |
| 5175 | * port is requested. |
| 5176 | */ |
| 5177 | static int sctp_get_port(struct sock *sk, unsigned short snum) |
| 5178 | { |
| 5179 | long ret; |
| 5180 | union sctp_addr addr; |
| 5181 | struct sctp_af *af = sctp_sk(sk)->pf->af; |
| 5182 | |
| 5183 | /* Set up a dummy address struct from the sk. */ |
| 5184 | af->from_sk(&addr, sk); |
| 5185 | addr.v4.sin_port = htons(snum); |
| 5186 | |
| 5187 | /* Note: sk->sk_num gets filled in if ephemeral port request. */ |
| 5188 | ret = sctp_get_port_local(sk, &addr); |
| 5189 | |
| 5190 | return (ret ? 1 : 0); |
| 5191 | } |
| 5192 | |
| 5193 | /* |
| 5194 | * 3.1.3 listen() - UDP Style Syntax |
| 5195 | * |
| 5196 | * By default, new associations are not accepted for UDP style sockets. |
| 5197 | * An application uses listen() to mark a socket as being able to |
| 5198 | * accept new associations. |
| 5199 | */ |
| 5200 | SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog) |
| 5201 | { |
| 5202 | struct sctp_sock *sp = sctp_sk(sk); |
| 5203 | struct sctp_endpoint *ep = sp->ep; |
| 5204 | |
| 5205 | /* Only UDP style sockets that are not peeled off are allowed to |
| 5206 | * listen(). |
| 5207 | */ |
| 5208 | if (!sctp_style(sk, UDP)) |
| 5209 | return -EINVAL; |
| 5210 | |
| 5211 | /* If backlog is zero, disable listening. */ |
| 5212 | if (!backlog) { |
| 5213 | if (sctp_sstate(sk, CLOSED)) |
| 5214 | return 0; |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 5215 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5216 | sctp_unhash_endpoint(ep); |
| 5217 | sk->sk_state = SCTP_SS_CLOSED; |
Vlad Yasevich | 498d630 | 2007-08-30 14:03:58 -0400 | [diff] [blame] | 5218 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5219 | } |
| 5220 | |
| 5221 | /* Return if we are already listening. */ |
| 5222 | if (sctp_sstate(sk, LISTENING)) |
| 5223 | return 0; |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 5224 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5225 | /* |
| 5226 | * If a bind() or sctp_bindx() is not called prior to a listen() |
| 5227 | * call that allows new associations to be accepted, the system |
| 5228 | * picks an ephemeral port and will choose an address set equivalent |
| 5229 | * to binding with a wildcard address. |
| 5230 | * |
| 5231 | * This is not currently spelled out in the SCTP sockets |
| 5232 | * extensions draft, but follows the practice as seen in TCP |
| 5233 | * sockets. |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5234 | * |
| 5235 | * Additionally, turn off fastreuse flag since we are not listening |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5236 | */ |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5237 | sk->sk_state = SCTP_SS_LISTENING; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5238 | if (!ep->base.bind_addr.port) { |
| 5239 | if (sctp_autobind(sk)) |
| 5240 | return -EAGAIN; |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5241 | } else |
| 5242 | sctp_sk(sk)->bind_hash->fastreuse = 0; |
| 5243 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5244 | sctp_hash_endpoint(ep); |
| 5245 | return 0; |
| 5246 | } |
| 5247 | |
| 5248 | /* |
| 5249 | * 4.1.3 listen() - TCP Style Syntax |
| 5250 | * |
| 5251 | * Applications uses listen() to ready the SCTP endpoint for accepting |
| 5252 | * inbound associations. |
| 5253 | */ |
| 5254 | SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog) |
| 5255 | { |
| 5256 | struct sctp_sock *sp = sctp_sk(sk); |
| 5257 | struct sctp_endpoint *ep = sp->ep; |
| 5258 | |
| 5259 | /* If backlog is zero, disable listening. */ |
| 5260 | if (!backlog) { |
| 5261 | if (sctp_sstate(sk, CLOSED)) |
| 5262 | return 0; |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 5263 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5264 | sctp_unhash_endpoint(ep); |
| 5265 | sk->sk_state = SCTP_SS_CLOSED; |
Vlad Yasevich | 498d630 | 2007-08-30 14:03:58 -0400 | [diff] [blame] | 5266 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5267 | } |
| 5268 | |
| 5269 | if (sctp_sstate(sk, LISTENING)) |
| 5270 | return 0; |
| 5271 | |
| 5272 | /* |
| 5273 | * If a bind() or sctp_bindx() is not called prior to a listen() |
| 5274 | * call that allows new associations to be accepted, the system |
| 5275 | * picks an ephemeral port and will choose an address set equivalent |
| 5276 | * to binding with a wildcard address. |
| 5277 | * |
| 5278 | * This is not currently spelled out in the SCTP sockets |
| 5279 | * extensions draft, but follows the practice as seen in TCP |
| 5280 | * sockets. |
| 5281 | */ |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5282 | sk->sk_state = SCTP_SS_LISTENING; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5283 | if (!ep->base.bind_addr.port) { |
| 5284 | if (sctp_autobind(sk)) |
| 5285 | return -EAGAIN; |
Vlad Yasevich | ce5325c | 2007-05-04 13:34:49 -0700 | [diff] [blame] | 5286 | } else |
| 5287 | sctp_sk(sk)->bind_hash->fastreuse = 0; |
| 5288 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5289 | sk->sk_max_ack_backlog = backlog; |
| 5290 | sctp_hash_endpoint(ep); |
| 5291 | return 0; |
| 5292 | } |
| 5293 | |
| 5294 | /* |
| 5295 | * Move a socket to LISTENING state. |
| 5296 | */ |
| 5297 | int sctp_inet_listen(struct socket *sock, int backlog) |
| 5298 | { |
| 5299 | struct sock *sk = sock->sk; |
Herbert Xu | 1b489e1 | 2006-08-20 15:07:14 +1000 | [diff] [blame] | 5300 | struct crypto_hash *tfm = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5301 | int err = -EINVAL; |
| 5302 | |
| 5303 | if (unlikely(backlog < 0)) |
| 5304 | goto out; |
| 5305 | |
| 5306 | sctp_lock_sock(sk); |
| 5307 | |
| 5308 | if (sock->state != SS_UNCONNECTED) |
| 5309 | goto out; |
| 5310 | |
| 5311 | /* Allocate HMAC for generating cookie. */ |
| 5312 | if (sctp_hmac_alg) { |
Herbert Xu | 1b489e1 | 2006-08-20 15:07:14 +1000 | [diff] [blame] | 5313 | tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC); |
Vlad Yasevich | 8dc4984 | 2007-05-09 13:50:20 -0700 | [diff] [blame] | 5314 | if (IS_ERR(tfm)) { |
| 5315 | if (net_ratelimit()) { |
| 5316 | printk(KERN_INFO |
| 5317 | "SCTP: failed to load transform for %s: %ld\n", |
| 5318 | sctp_hmac_alg, PTR_ERR(tfm)); |
| 5319 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5320 | err = -ENOSYS; |
| 5321 | goto out; |
| 5322 | } |
| 5323 | } |
| 5324 | |
| 5325 | switch (sock->type) { |
| 5326 | case SOCK_SEQPACKET: |
| 5327 | err = sctp_seqpacket_listen(sk, backlog); |
| 5328 | break; |
| 5329 | case SOCK_STREAM: |
| 5330 | err = sctp_stream_listen(sk, backlog); |
| 5331 | break; |
| 5332 | default: |
| 5333 | break; |
Stephen Hemminger | 3ff50b7 | 2007-04-20 17:09:22 -0700 | [diff] [blame] | 5334 | } |
| 5335 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5336 | if (err) |
| 5337 | goto cleanup; |
| 5338 | |
| 5339 | /* Store away the transform reference. */ |
| 5340 | sctp_sk(sk)->hmac = tfm; |
| 5341 | out: |
| 5342 | sctp_release_sock(sk); |
| 5343 | return err; |
| 5344 | cleanup: |
Herbert Xu | 1b489e1 | 2006-08-20 15:07:14 +1000 | [diff] [blame] | 5345 | crypto_free_hash(tfm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5346 | goto out; |
| 5347 | } |
| 5348 | |
| 5349 | /* |
| 5350 | * This function is done by modeling the current datagram_poll() and the |
| 5351 | * tcp_poll(). Note that, based on these implementations, we don't |
| 5352 | * lock the socket in this function, even though it seems that, |
| 5353 | * ideally, locking or some other mechanisms can be used to ensure |
Neil Horman | 9bffc4a | 2005-12-19 14:24:40 -0800 | [diff] [blame] | 5354 | * the integrity of the counters (sndbuf and wmem_alloc) used |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5355 | * in this place. We assume that we don't need locks either until proven |
| 5356 | * otherwise. |
| 5357 | * |
| 5358 | * Another thing to note is that we include the Async I/O support |
| 5359 | * here, again, by modeling the current TCP/UDP code. We don't have |
| 5360 | * a good way to test with it yet. |
| 5361 | */ |
| 5362 | unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait) |
| 5363 | { |
| 5364 | struct sock *sk = sock->sk; |
| 5365 | struct sctp_sock *sp = sctp_sk(sk); |
| 5366 | unsigned int mask; |
| 5367 | |
| 5368 | poll_wait(file, sk->sk_sleep, wait); |
| 5369 | |
| 5370 | /* A TCP-style listening socket becomes readable when the accept queue |
| 5371 | * is not empty. |
| 5372 | */ |
| 5373 | if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) |
| 5374 | return (!list_empty(&sp->ep->asocs)) ? |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 5375 | (POLLIN | POLLRDNORM) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5376 | |
| 5377 | mask = 0; |
| 5378 | |
| 5379 | /* Is there any exceptional events? */ |
| 5380 | if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) |
| 5381 | mask |= POLLERR; |
Davide Libenzi | f348d70 | 2006-03-25 03:07:39 -0800 | [diff] [blame] | 5382 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
| 5383 | mask |= POLLRDHUP; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5384 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
| 5385 | mask |= POLLHUP; |
| 5386 | |
| 5387 | /* Is it readable? Reconsider this code with TCP-style support. */ |
| 5388 | if (!skb_queue_empty(&sk->sk_receive_queue) || |
| 5389 | (sk->sk_shutdown & RCV_SHUTDOWN)) |
| 5390 | mask |= POLLIN | POLLRDNORM; |
| 5391 | |
| 5392 | /* The association is either gone or not ready. */ |
| 5393 | if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED)) |
| 5394 | return mask; |
| 5395 | |
| 5396 | /* Is it writable? */ |
| 5397 | if (sctp_writeable(sk)) { |
| 5398 | mask |= POLLOUT | POLLWRNORM; |
| 5399 | } else { |
| 5400 | set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); |
| 5401 | /* |
| 5402 | * Since the socket is not locked, the buffer |
| 5403 | * might be made available after the writeable check and |
| 5404 | * before the bit is set. This could cause a lost I/O |
| 5405 | * signal. tcp_poll() has a race breaker for this race |
| 5406 | * condition. Based on their implementation, we put |
| 5407 | * in the following code to cover it as well. |
| 5408 | */ |
| 5409 | if (sctp_writeable(sk)) |
| 5410 | mask |= POLLOUT | POLLWRNORM; |
| 5411 | } |
| 5412 | return mask; |
| 5413 | } |
| 5414 | |
| 5415 | /******************************************************************** |
| 5416 | * 2nd Level Abstractions |
| 5417 | ********************************************************************/ |
| 5418 | |
| 5419 | static struct sctp_bind_bucket *sctp_bucket_create( |
| 5420 | struct sctp_bind_hashbucket *head, unsigned short snum) |
| 5421 | { |
| 5422 | struct sctp_bind_bucket *pp; |
| 5423 | |
Christoph Lameter | 54e6ecb | 2006-12-06 20:33:16 -0800 | [diff] [blame] | 5424 | pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5425 | SCTP_DBG_OBJCNT_INC(bind_bucket); |
| 5426 | if (pp) { |
| 5427 | pp->port = snum; |
| 5428 | pp->fastreuse = 0; |
| 5429 | INIT_HLIST_HEAD(&pp->owner); |
| 5430 | if ((pp->next = head->chain) != NULL) |
| 5431 | pp->next->pprev = &pp->next; |
| 5432 | head->chain = pp; |
| 5433 | pp->pprev = &head->chain; |
| 5434 | } |
| 5435 | return pp; |
| 5436 | } |
| 5437 | |
| 5438 | /* Caller must hold hashbucket lock for this tb with local BH disabled */ |
| 5439 | static void sctp_bucket_destroy(struct sctp_bind_bucket *pp) |
| 5440 | { |
Sridhar Samudrala | 37fa687 | 2006-07-21 14:45:47 -0700 | [diff] [blame] | 5441 | if (pp && hlist_empty(&pp->owner)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5442 | if (pp->next) |
| 5443 | pp->next->pprev = pp->pprev; |
| 5444 | *(pp->pprev) = pp->next; |
| 5445 | kmem_cache_free(sctp_bucket_cachep, pp); |
| 5446 | SCTP_DBG_OBJCNT_DEC(bind_bucket); |
| 5447 | } |
| 5448 | } |
| 5449 | |
| 5450 | /* Release this socket's reference to a local port. */ |
| 5451 | static inline void __sctp_put_port(struct sock *sk) |
| 5452 | { |
| 5453 | struct sctp_bind_hashbucket *head = |
| 5454 | &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)]; |
| 5455 | struct sctp_bind_bucket *pp; |
| 5456 | |
| 5457 | sctp_spin_lock(&head->lock); |
| 5458 | pp = sctp_sk(sk)->bind_hash; |
| 5459 | __sk_del_bind_node(sk); |
| 5460 | sctp_sk(sk)->bind_hash = NULL; |
| 5461 | inet_sk(sk)->num = 0; |
| 5462 | sctp_bucket_destroy(pp); |
| 5463 | sctp_spin_unlock(&head->lock); |
| 5464 | } |
| 5465 | |
| 5466 | void sctp_put_port(struct sock *sk) |
| 5467 | { |
| 5468 | sctp_local_bh_disable(); |
| 5469 | __sctp_put_port(sk); |
| 5470 | sctp_local_bh_enable(); |
| 5471 | } |
| 5472 | |
| 5473 | /* |
| 5474 | * The system picks an ephemeral port and choose an address set equivalent |
| 5475 | * to binding with a wildcard address. |
| 5476 | * One of those addresses will be the primary address for the association. |
| 5477 | * This automatically enables the multihoming capability of SCTP. |
| 5478 | */ |
| 5479 | static int sctp_autobind(struct sock *sk) |
| 5480 | { |
| 5481 | union sctp_addr autoaddr; |
| 5482 | struct sctp_af *af; |
Al Viro | 6fbfa9f | 2006-11-20 17:24:53 -0800 | [diff] [blame] | 5483 | __be16 port; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5484 | |
| 5485 | /* Initialize a local sockaddr structure to INADDR_ANY. */ |
| 5486 | af = sctp_sk(sk)->pf->af; |
| 5487 | |
| 5488 | port = htons(inet_sk(sk)->num); |
| 5489 | af->inaddr_any(&autoaddr, port); |
| 5490 | |
| 5491 | return sctp_do_bind(sk, &autoaddr, af->sockaddr_len); |
| 5492 | } |
| 5493 | |
| 5494 | /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation. |
| 5495 | * |
| 5496 | * From RFC 2292 |
| 5497 | * 4.2 The cmsghdr Structure * |
| 5498 | * |
| 5499 | * When ancillary data is sent or received, any number of ancillary data |
| 5500 | * objects can be specified by the msg_control and msg_controllen members of |
| 5501 | * the msghdr structure, because each object is preceded by |
| 5502 | * a cmsghdr structure defining the object's length (the cmsg_len member). |
| 5503 | * Historically Berkeley-derived implementations have passed only one object |
| 5504 | * at a time, but this API allows multiple objects to be |
| 5505 | * passed in a single call to sendmsg() or recvmsg(). The following example |
| 5506 | * shows two ancillary data objects in a control buffer. |
| 5507 | * |
| 5508 | * |<--------------------------- msg_controllen -------------------------->| |
| 5509 | * | | |
| 5510 | * |
| 5511 | * |<----- ancillary data object ----->|<----- ancillary data object ----->| |
| 5512 | * |
| 5513 | * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->| |
| 5514 | * | | | |
| 5515 | * |
| 5516 | * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| | |
| 5517 | * |
| 5518 | * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| | |
| 5519 | * | | | | | |
| 5520 | * |
| 5521 | * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+ |
| 5522 | * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX| |
| 5523 | * |
| 5524 | * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX| |
| 5525 | * |
| 5526 | * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+ |
| 5527 | * ^ |
| 5528 | * | |
| 5529 | * |
| 5530 | * msg_control |
| 5531 | * points here |
| 5532 | */ |
| 5533 | SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg, |
| 5534 | sctp_cmsgs_t *cmsgs) |
| 5535 | { |
| 5536 | struct cmsghdr *cmsg; |
| 5537 | |
| 5538 | for (cmsg = CMSG_FIRSTHDR(msg); |
| 5539 | cmsg != NULL; |
| 5540 | cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) { |
| 5541 | if (!CMSG_OK(msg, cmsg)) |
| 5542 | return -EINVAL; |
| 5543 | |
| 5544 | /* Should we parse this header or ignore? */ |
| 5545 | if (cmsg->cmsg_level != IPPROTO_SCTP) |
| 5546 | continue; |
| 5547 | |
| 5548 | /* Strictly check lengths following example in SCM code. */ |
| 5549 | switch (cmsg->cmsg_type) { |
| 5550 | case SCTP_INIT: |
| 5551 | /* SCTP Socket API Extension |
| 5552 | * 5.2.1 SCTP Initiation Structure (SCTP_INIT) |
| 5553 | * |
| 5554 | * This cmsghdr structure provides information for |
| 5555 | * initializing new SCTP associations with sendmsg(). |
| 5556 | * The SCTP_INITMSG socket option uses this same data |
| 5557 | * structure. This structure is not used for |
| 5558 | * recvmsg(). |
| 5559 | * |
| 5560 | * cmsg_level cmsg_type cmsg_data[] |
| 5561 | * ------------ ------------ ---------------------- |
| 5562 | * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg |
| 5563 | */ |
| 5564 | if (cmsg->cmsg_len != |
| 5565 | CMSG_LEN(sizeof(struct sctp_initmsg))) |
| 5566 | return -EINVAL; |
| 5567 | cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg); |
| 5568 | break; |
| 5569 | |
| 5570 | case SCTP_SNDRCV: |
| 5571 | /* SCTP Socket API Extension |
| 5572 | * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV) |
| 5573 | * |
| 5574 | * This cmsghdr structure specifies SCTP options for |
| 5575 | * sendmsg() and describes SCTP header information |
| 5576 | * about a received message through recvmsg(). |
| 5577 | * |
| 5578 | * cmsg_level cmsg_type cmsg_data[] |
| 5579 | * ------------ ------------ ---------------------- |
| 5580 | * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo |
| 5581 | */ |
| 5582 | if (cmsg->cmsg_len != |
| 5583 | CMSG_LEN(sizeof(struct sctp_sndrcvinfo))) |
| 5584 | return -EINVAL; |
| 5585 | |
| 5586 | cmsgs->info = |
| 5587 | (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg); |
| 5588 | |
| 5589 | /* Minimally, validate the sinfo_flags. */ |
| 5590 | if (cmsgs->info->sinfo_flags & |
Ivan Skytte Jorgensen | eaa5c54 | 2005-10-28 15:10:00 -0700 | [diff] [blame] | 5591 | ~(SCTP_UNORDERED | SCTP_ADDR_OVER | |
| 5592 | SCTP_ABORT | SCTP_EOF)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5593 | return -EINVAL; |
| 5594 | break; |
| 5595 | |
| 5596 | default: |
| 5597 | return -EINVAL; |
Stephen Hemminger | 3ff50b7 | 2007-04-20 17:09:22 -0700 | [diff] [blame] | 5598 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5599 | } |
| 5600 | return 0; |
| 5601 | } |
| 5602 | |
| 5603 | /* |
| 5604 | * Wait for a packet.. |
| 5605 | * Note: This function is the same function as in core/datagram.c |
| 5606 | * with a few modifications to make lksctp work. |
| 5607 | */ |
| 5608 | static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p) |
| 5609 | { |
| 5610 | int error; |
| 5611 | DEFINE_WAIT(wait); |
| 5612 | |
| 5613 | prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE); |
| 5614 | |
| 5615 | /* Socket errors? */ |
| 5616 | error = sock_error(sk); |
| 5617 | if (error) |
| 5618 | goto out; |
| 5619 | |
| 5620 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
| 5621 | goto ready; |
| 5622 | |
| 5623 | /* Socket shut down? */ |
| 5624 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
| 5625 | goto out; |
| 5626 | |
| 5627 | /* Sequenced packets can come disconnected. If so we report the |
| 5628 | * problem. |
| 5629 | */ |
| 5630 | error = -ENOTCONN; |
| 5631 | |
| 5632 | /* Is there a good reason to think that we may receive some data? */ |
| 5633 | if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING)) |
| 5634 | goto out; |
| 5635 | |
| 5636 | /* Handle signals. */ |
| 5637 | if (signal_pending(current)) |
| 5638 | goto interrupted; |
| 5639 | |
| 5640 | /* Let another process have a go. Since we are going to sleep |
| 5641 | * anyway. Note: This may cause odd behaviors if the message |
| 5642 | * does not fit in the user's buffer, but this seems to be the |
| 5643 | * only way to honor MSG_DONTWAIT realistically. |
| 5644 | */ |
| 5645 | sctp_release_sock(sk); |
| 5646 | *timeo_p = schedule_timeout(*timeo_p); |
| 5647 | sctp_lock_sock(sk); |
| 5648 | |
| 5649 | ready: |
| 5650 | finish_wait(sk->sk_sleep, &wait); |
| 5651 | return 0; |
| 5652 | |
| 5653 | interrupted: |
| 5654 | error = sock_intr_errno(*timeo_p); |
| 5655 | |
| 5656 | out: |
| 5657 | finish_wait(sk->sk_sleep, &wait); |
| 5658 | *err = error; |
| 5659 | return error; |
| 5660 | } |
| 5661 | |
| 5662 | /* Receive a datagram. |
| 5663 | * Note: This is pretty much the same routine as in core/datagram.c |
| 5664 | * with a few changes to make lksctp work. |
| 5665 | */ |
| 5666 | static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags, |
| 5667 | int noblock, int *err) |
| 5668 | { |
| 5669 | int error; |
| 5670 | struct sk_buff *skb; |
| 5671 | long timeo; |
| 5672 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5673 | timeo = sock_rcvtimeo(sk, noblock); |
| 5674 | |
| 5675 | SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n", |
| 5676 | timeo, MAX_SCHEDULE_TIMEOUT); |
| 5677 | |
| 5678 | do { |
| 5679 | /* Again only user level code calls this function, |
| 5680 | * so nothing interrupt level |
| 5681 | * will suddenly eat the receive_queue. |
| 5682 | * |
| 5683 | * Look at current nfs client by the way... |
| 5684 | * However, this function was corrent in any case. 8) |
| 5685 | */ |
| 5686 | if (flags & MSG_PEEK) { |
Herbert Xu | 1e061ab | 2005-06-18 22:56:42 -0700 | [diff] [blame] | 5687 | spin_lock_bh(&sk->sk_receive_queue.lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5688 | skb = skb_peek(&sk->sk_receive_queue); |
| 5689 | if (skb) |
| 5690 | atomic_inc(&skb->users); |
Herbert Xu | 1e061ab | 2005-06-18 22:56:42 -0700 | [diff] [blame] | 5691 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5692 | } else { |
| 5693 | skb = skb_dequeue(&sk->sk_receive_queue); |
| 5694 | } |
| 5695 | |
| 5696 | if (skb) |
| 5697 | return skb; |
| 5698 | |
Neil Horman | 6736dc3 | 2005-12-02 20:30:06 -0800 | [diff] [blame] | 5699 | /* Caller is allowed not to check sk->sk_err before calling. */ |
| 5700 | error = sock_error(sk); |
| 5701 | if (error) |
| 5702 | goto no_packet; |
| 5703 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5704 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
| 5705 | break; |
| 5706 | |
| 5707 | /* User doesn't want to wait. */ |
| 5708 | error = -EAGAIN; |
| 5709 | if (!timeo) |
| 5710 | goto no_packet; |
| 5711 | } while (sctp_wait_for_packet(sk, err, &timeo) == 0); |
| 5712 | |
| 5713 | return NULL; |
| 5714 | |
| 5715 | no_packet: |
| 5716 | *err = error; |
| 5717 | return NULL; |
| 5718 | } |
| 5719 | |
| 5720 | /* If sndbuf has changed, wake up per association sndbuf waiters. */ |
| 5721 | static void __sctp_write_space(struct sctp_association *asoc) |
| 5722 | { |
| 5723 | struct sock *sk = asoc->base.sk; |
| 5724 | struct socket *sock = sk->sk_socket; |
| 5725 | |
| 5726 | if ((sctp_wspace(asoc) > 0) && sock) { |
| 5727 | if (waitqueue_active(&asoc->wait)) |
| 5728 | wake_up_interruptible(&asoc->wait); |
| 5729 | |
| 5730 | if (sctp_writeable(sk)) { |
| 5731 | if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) |
| 5732 | wake_up_interruptible(sk->sk_sleep); |
| 5733 | |
| 5734 | /* Note that we try to include the Async I/O support |
| 5735 | * here by modeling from the current TCP/UDP code. |
| 5736 | * We have not tested with it yet. |
| 5737 | */ |
| 5738 | if (sock->fasync_list && |
| 5739 | !(sk->sk_shutdown & SEND_SHUTDOWN)) |
| 5740 | sock_wake_async(sock, 2, POLL_OUT); |
| 5741 | } |
| 5742 | } |
| 5743 | } |
| 5744 | |
| 5745 | /* Do accounting for the sndbuf space. |
| 5746 | * Decrement the used sndbuf space of the corresponding association by the |
| 5747 | * data size which was just transmitted(freed). |
| 5748 | */ |
| 5749 | static void sctp_wfree(struct sk_buff *skb) |
| 5750 | { |
| 5751 | struct sctp_association *asoc; |
| 5752 | struct sctp_chunk *chunk; |
| 5753 | struct sock *sk; |
| 5754 | |
| 5755 | /* Get the saved chunk pointer. */ |
| 5756 | chunk = *((struct sctp_chunk **)(skb->cb)); |
| 5757 | asoc = chunk->asoc; |
| 5758 | sk = asoc->base.sk; |
Neil Horman | 4eb701d | 2005-04-28 12:02:04 -0700 | [diff] [blame] | 5759 | asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) + |
| 5760 | sizeof(struct sk_buff) + |
| 5761 | sizeof(struct sctp_chunk); |
| 5762 | |
Neil Horman | 4eb701d | 2005-04-28 12:02:04 -0700 | [diff] [blame] | 5763 | atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc); |
| 5764 | |
| 5765 | sock_wfree(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5766 | __sctp_write_space(asoc); |
| 5767 | |
| 5768 | sctp_association_put(asoc); |
| 5769 | } |
| 5770 | |
Vlad Yasevich | 331c4ee | 2006-10-09 21:34:04 -0700 | [diff] [blame] | 5771 | /* Do accounting for the receive space on the socket. |
| 5772 | * Accounting for the association is done in ulpevent.c |
| 5773 | * We set this as a destructor for the cloned data skbs so that |
| 5774 | * accounting is done at the correct time. |
| 5775 | */ |
| 5776 | void sctp_sock_rfree(struct sk_buff *skb) |
| 5777 | { |
| 5778 | struct sock *sk = skb->sk; |
| 5779 | struct sctp_ulpevent *event = sctp_skb2event(skb); |
| 5780 | |
| 5781 | atomic_sub(event->rmem_len, &sk->sk_rmem_alloc); |
| 5782 | } |
| 5783 | |
| 5784 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5785 | /* Helper function to wait for space in the sndbuf. */ |
| 5786 | static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p, |
| 5787 | size_t msg_len) |
| 5788 | { |
| 5789 | struct sock *sk = asoc->base.sk; |
| 5790 | int err = 0; |
| 5791 | long current_timeo = *timeo_p; |
| 5792 | DEFINE_WAIT(wait); |
| 5793 | |
| 5794 | SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n", |
YOSHIFUJI Hideaki | d808ad9 | 2007-02-09 23:25:18 +0900 | [diff] [blame] | 5795 | asoc, (long)(*timeo_p), msg_len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5796 | |
| 5797 | /* Increment the association's refcnt. */ |
| 5798 | sctp_association_hold(asoc); |
| 5799 | |
| 5800 | /* Wait on the association specific sndbuf space. */ |
| 5801 | for (;;) { |
| 5802 | prepare_to_wait_exclusive(&asoc->wait, &wait, |
| 5803 | TASK_INTERRUPTIBLE); |
| 5804 | if (!*timeo_p) |
| 5805 | goto do_nonblock; |
| 5806 | if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING || |
| 5807 | asoc->base.dead) |
| 5808 | goto do_error; |
| 5809 | if (signal_pending(current)) |
| 5810 | goto do_interrupted; |
| 5811 | if (msg_len <= sctp_wspace(asoc)) |
| 5812 | break; |
| 5813 | |
| 5814 | /* Let another process have a go. Since we are going |
| 5815 | * to sleep anyway. |
| 5816 | */ |
| 5817 | sctp_release_sock(sk); |
| 5818 | current_timeo = schedule_timeout(current_timeo); |
Vladislav Yasevich | 61c9fed | 2006-05-19 11:01:18 -0700 | [diff] [blame] | 5819 | BUG_ON(sk != asoc->base.sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5820 | sctp_lock_sock(sk); |
| 5821 | |
| 5822 | *timeo_p = current_timeo; |
| 5823 | } |
| 5824 | |
| 5825 | out: |
| 5826 | finish_wait(&asoc->wait, &wait); |
| 5827 | |
| 5828 | /* Release the association's refcnt. */ |
| 5829 | sctp_association_put(asoc); |
| 5830 | |
| 5831 | return err; |
| 5832 | |
| 5833 | do_error: |
| 5834 | err = -EPIPE; |
| 5835 | goto out; |
| 5836 | |
| 5837 | do_interrupted: |
| 5838 | err = sock_intr_errno(*timeo_p); |
| 5839 | goto out; |
| 5840 | |
| 5841 | do_nonblock: |
| 5842 | err = -EAGAIN; |
| 5843 | goto out; |
| 5844 | } |
| 5845 | |
| 5846 | /* If socket sndbuf has changed, wake up all per association waiters. */ |
| 5847 | void sctp_write_space(struct sock *sk) |
| 5848 | { |
| 5849 | struct sctp_association *asoc; |
| 5850 | struct list_head *pos; |
| 5851 | |
| 5852 | /* Wake up the tasks in each wait queue. */ |
| 5853 | list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) { |
| 5854 | asoc = list_entry(pos, struct sctp_association, asocs); |
| 5855 | __sctp_write_space(asoc); |
| 5856 | } |
| 5857 | } |
| 5858 | |
| 5859 | /* Is there any sndbuf space available on the socket? |
| 5860 | * |
Neil Horman | 9bffc4a | 2005-12-19 14:24:40 -0800 | [diff] [blame] | 5861 | * Note that sk_wmem_alloc is the sum of the send buffers on all of the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5862 | * associations on the same socket. For a UDP-style socket with |
| 5863 | * multiple associations, it is possible for it to be "unwriteable" |
| 5864 | * prematurely. I assume that this is acceptable because |
| 5865 | * a premature "unwriteable" is better than an accidental "writeable" which |
| 5866 | * would cause an unwanted block under certain circumstances. For the 1-1 |
| 5867 | * UDP-style sockets or TCP-style sockets, this code should work. |
| 5868 | * - Daisy |
| 5869 | */ |
| 5870 | static int sctp_writeable(struct sock *sk) |
| 5871 | { |
| 5872 | int amt = 0; |
| 5873 | |
Neil Horman | 9bffc4a | 2005-12-19 14:24:40 -0800 | [diff] [blame] | 5874 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5875 | if (amt < 0) |
| 5876 | amt = 0; |
| 5877 | return amt; |
| 5878 | } |
| 5879 | |
| 5880 | /* Wait for an association to go into ESTABLISHED state. If timeout is 0, |
| 5881 | * returns immediately with EINPROGRESS. |
| 5882 | */ |
| 5883 | static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p) |
| 5884 | { |
| 5885 | struct sock *sk = asoc->base.sk; |
| 5886 | int err = 0; |
| 5887 | long current_timeo = *timeo_p; |
| 5888 | DEFINE_WAIT(wait); |
| 5889 | |
| 5890 | SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc, |
| 5891 | (long)(*timeo_p)); |
| 5892 | |
| 5893 | /* Increment the association's refcnt. */ |
| 5894 | sctp_association_hold(asoc); |
| 5895 | |
| 5896 | for (;;) { |
| 5897 | prepare_to_wait_exclusive(&asoc->wait, &wait, |
| 5898 | TASK_INTERRUPTIBLE); |
| 5899 | if (!*timeo_p) |
| 5900 | goto do_nonblock; |
| 5901 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
| 5902 | break; |
| 5903 | if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING || |
| 5904 | asoc->base.dead) |
| 5905 | goto do_error; |
| 5906 | if (signal_pending(current)) |
| 5907 | goto do_interrupted; |
| 5908 | |
| 5909 | if (sctp_state(asoc, ESTABLISHED)) |
| 5910 | break; |
| 5911 | |
| 5912 | /* Let another process have a go. Since we are going |
| 5913 | * to sleep anyway. |
| 5914 | */ |
| 5915 | sctp_release_sock(sk); |
| 5916 | current_timeo = schedule_timeout(current_timeo); |
| 5917 | sctp_lock_sock(sk); |
| 5918 | |
| 5919 | *timeo_p = current_timeo; |
| 5920 | } |
| 5921 | |
| 5922 | out: |
| 5923 | finish_wait(&asoc->wait, &wait); |
| 5924 | |
| 5925 | /* Release the association's refcnt. */ |
| 5926 | sctp_association_put(asoc); |
| 5927 | |
| 5928 | return err; |
| 5929 | |
| 5930 | do_error: |
Vlad Yasevich | 81845c2 | 2006-01-30 15:59:54 -0800 | [diff] [blame] | 5931 | if (asoc->init_err_counter + 1 > asoc->max_init_attempts) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5932 | err = -ETIMEDOUT; |
| 5933 | else |
| 5934 | err = -ECONNREFUSED; |
| 5935 | goto out; |
| 5936 | |
| 5937 | do_interrupted: |
| 5938 | err = sock_intr_errno(*timeo_p); |
| 5939 | goto out; |
| 5940 | |
| 5941 | do_nonblock: |
| 5942 | err = -EINPROGRESS; |
| 5943 | goto out; |
| 5944 | } |
| 5945 | |
| 5946 | static int sctp_wait_for_accept(struct sock *sk, long timeo) |
| 5947 | { |
| 5948 | struct sctp_endpoint *ep; |
| 5949 | int err = 0; |
| 5950 | DEFINE_WAIT(wait); |
| 5951 | |
| 5952 | ep = sctp_sk(sk)->ep; |
| 5953 | |
| 5954 | |
| 5955 | for (;;) { |
| 5956 | prepare_to_wait_exclusive(sk->sk_sleep, &wait, |
| 5957 | TASK_INTERRUPTIBLE); |
| 5958 | |
| 5959 | if (list_empty(&ep->asocs)) { |
| 5960 | sctp_release_sock(sk); |
| 5961 | timeo = schedule_timeout(timeo); |
| 5962 | sctp_lock_sock(sk); |
| 5963 | } |
| 5964 | |
| 5965 | err = -EINVAL; |
| 5966 | if (!sctp_sstate(sk, LISTENING)) |
| 5967 | break; |
| 5968 | |
| 5969 | err = 0; |
| 5970 | if (!list_empty(&ep->asocs)) |
| 5971 | break; |
| 5972 | |
| 5973 | err = sock_intr_errno(timeo); |
| 5974 | if (signal_pending(current)) |
| 5975 | break; |
| 5976 | |
| 5977 | err = -EAGAIN; |
| 5978 | if (!timeo) |
| 5979 | break; |
| 5980 | } |
| 5981 | |
| 5982 | finish_wait(sk->sk_sleep, &wait); |
| 5983 | |
| 5984 | return err; |
| 5985 | } |
| 5986 | |
sebastian@breakpoint.cc | 0467521 | 2007-07-26 23:21:31 +0200 | [diff] [blame] | 5987 | static void sctp_wait_for_close(struct sock *sk, long timeout) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5988 | { |
| 5989 | DEFINE_WAIT(wait); |
| 5990 | |
| 5991 | do { |
| 5992 | prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE); |
| 5993 | if (list_empty(&sctp_sk(sk)->ep->asocs)) |
| 5994 | break; |
| 5995 | sctp_release_sock(sk); |
| 5996 | timeout = schedule_timeout(timeout); |
| 5997 | sctp_lock_sock(sk); |
| 5998 | } while (!signal_pending(current) && timeout); |
| 5999 | |
| 6000 | finish_wait(sk->sk_sleep, &wait); |
| 6001 | } |
| 6002 | |
Tsutomu Fujii | ea2bc48 | 2007-04-17 12:49:53 -0700 | [diff] [blame] | 6003 | static void sctp_sock_rfree_frag(struct sk_buff *skb) |
| 6004 | { |
| 6005 | struct sk_buff *frag; |
| 6006 | |
| 6007 | if (!skb->data_len) |
| 6008 | goto done; |
| 6009 | |
| 6010 | /* Don't forget the fragments. */ |
| 6011 | for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) |
| 6012 | sctp_sock_rfree_frag(frag); |
| 6013 | |
| 6014 | done: |
| 6015 | sctp_sock_rfree(skb); |
| 6016 | } |
| 6017 | |
| 6018 | static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk) |
| 6019 | { |
| 6020 | struct sk_buff *frag; |
| 6021 | |
| 6022 | if (!skb->data_len) |
| 6023 | goto done; |
| 6024 | |
| 6025 | /* Don't forget the fragments. */ |
| 6026 | for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) |
| 6027 | sctp_skb_set_owner_r_frag(frag, sk); |
| 6028 | |
| 6029 | done: |
| 6030 | sctp_skb_set_owner_r(skb, sk); |
| 6031 | } |
| 6032 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6033 | /* Populate the fields of the newsk from the oldsk and migrate the assoc |
| 6034 | * and its messages to the newsk. |
| 6035 | */ |
| 6036 | static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk, |
| 6037 | struct sctp_association *assoc, |
| 6038 | sctp_socket_type_t type) |
| 6039 | { |
| 6040 | struct sctp_sock *oldsp = sctp_sk(oldsk); |
| 6041 | struct sctp_sock *newsp = sctp_sk(newsk); |
| 6042 | struct sctp_bind_bucket *pp; /* hash list port iterator */ |
| 6043 | struct sctp_endpoint *newep = newsp->ep; |
| 6044 | struct sk_buff *skb, *tmp; |
| 6045 | struct sctp_ulpevent *event; |
Vladislav Yasevich | 4243cac | 2005-06-13 15:10:49 -0700 | [diff] [blame] | 6046 | int flags = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6047 | |
| 6048 | /* Migrate socket buffer sizes and all the socket level options to the |
| 6049 | * new socket. |
| 6050 | */ |
| 6051 | newsk->sk_sndbuf = oldsk->sk_sndbuf; |
| 6052 | newsk->sk_rcvbuf = oldsk->sk_rcvbuf; |
| 6053 | /* Brute force copy old sctp opt. */ |
| 6054 | inet_sk_copy_descendant(newsk, oldsk); |
| 6055 | |
| 6056 | /* Restore the ep value that was overwritten with the above structure |
| 6057 | * copy. |
| 6058 | */ |
| 6059 | newsp->ep = newep; |
| 6060 | newsp->hmac = NULL; |
| 6061 | |
| 6062 | /* Hook this new socket in to the bind_hash list. */ |
| 6063 | pp = sctp_sk(oldsk)->bind_hash; |
| 6064 | sk_add_bind_node(newsk, &pp->owner); |
| 6065 | sctp_sk(newsk)->bind_hash = pp; |
| 6066 | inet_sk(newsk)->num = inet_sk(oldsk)->num; |
| 6067 | |
Vladislav Yasevich | 4243cac | 2005-06-13 15:10:49 -0700 | [diff] [blame] | 6068 | /* Copy the bind_addr list from the original endpoint to the new |
| 6069 | * endpoint so that we can handle restarts properly |
| 6070 | */ |
Vladislav Yasevich | eb5fa39 | 2006-08-22 00:23:13 -0700 | [diff] [blame] | 6071 | if (PF_INET6 == assoc->base.sk->sk_family) |
| 6072 | flags = SCTP_ADDR6_ALLOWED; |
Vladislav Yasevich | 4243cac | 2005-06-13 15:10:49 -0700 | [diff] [blame] | 6073 | if (assoc->peer.ipv4_address) |
| 6074 | flags |= SCTP_ADDR4_PEERSUPP; |
| 6075 | if (assoc->peer.ipv6_address) |
| 6076 | flags |= SCTP_ADDR6_PEERSUPP; |
| 6077 | sctp_bind_addr_copy(&newsp->ep->base.bind_addr, |
| 6078 | &oldsp->ep->base.bind_addr, |
| 6079 | SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags); |
| 6080 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6081 | /* Move any messages in the old socket's receive queue that are for the |
| 6082 | * peeled off association to the new socket's receive queue. |
| 6083 | */ |
| 6084 | sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) { |
| 6085 | event = sctp_skb2event(skb); |
| 6086 | if (event->asoc == assoc) { |
Tsutomu Fujii | ea2bc48 | 2007-04-17 12:49:53 -0700 | [diff] [blame] | 6087 | sctp_sock_rfree_frag(skb); |
David S. Miller | 8728b83 | 2005-08-09 19:25:21 -0700 | [diff] [blame] | 6088 | __skb_unlink(skb, &oldsk->sk_receive_queue); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6089 | __skb_queue_tail(&newsk->sk_receive_queue, skb); |
Tsutomu Fujii | ea2bc48 | 2007-04-17 12:49:53 -0700 | [diff] [blame] | 6090 | sctp_skb_set_owner_r_frag(skb, newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6091 | } |
| 6092 | } |
| 6093 | |
| 6094 | /* Clean up any messages pending delivery due to partial |
| 6095 | * delivery. Three cases: |
| 6096 | * 1) No partial deliver; no work. |
| 6097 | * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby. |
| 6098 | * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue. |
| 6099 | */ |
| 6100 | skb_queue_head_init(&newsp->pd_lobby); |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 6101 | atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6102 | |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 6103 | if (atomic_read(&sctp_sk(oldsk)->pd_mode)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6104 | struct sk_buff_head *queue; |
| 6105 | |
| 6106 | /* Decide which queue to move pd_lobby skbs to. */ |
| 6107 | if (assoc->ulpq.pd_mode) { |
| 6108 | queue = &newsp->pd_lobby; |
| 6109 | } else |
| 6110 | queue = &newsk->sk_receive_queue; |
| 6111 | |
| 6112 | /* Walk through the pd_lobby, looking for skbs that |
| 6113 | * need moved to the new socket. |
| 6114 | */ |
| 6115 | sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) { |
| 6116 | event = sctp_skb2event(skb); |
| 6117 | if (event->asoc == assoc) { |
Tsutomu Fujii | ea2bc48 | 2007-04-17 12:49:53 -0700 | [diff] [blame] | 6118 | sctp_sock_rfree_frag(skb); |
David S. Miller | 8728b83 | 2005-08-09 19:25:21 -0700 | [diff] [blame] | 6119 | __skb_unlink(skb, &oldsp->pd_lobby); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6120 | __skb_queue_tail(queue, skb); |
Tsutomu Fujii | ea2bc48 | 2007-04-17 12:49:53 -0700 | [diff] [blame] | 6121 | sctp_skb_set_owner_r_frag(skb, newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6122 | } |
| 6123 | } |
| 6124 | |
| 6125 | /* Clear up any skbs waiting for the partial |
| 6126 | * delivery to finish. |
| 6127 | */ |
| 6128 | if (assoc->ulpq.pd_mode) |
Vlad Yasevich | b6e1331 | 2007-04-20 12:23:15 -0700 | [diff] [blame] | 6129 | sctp_clear_pd(oldsk, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6130 | |
| 6131 | } |
| 6132 | |
Tsutomu Fujii | ea2bc48 | 2007-04-17 12:49:53 -0700 | [diff] [blame] | 6133 | sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) { |
| 6134 | sctp_sock_rfree_frag(skb); |
| 6135 | sctp_skb_set_owner_r_frag(skb, newsk); |
| 6136 | } |
| 6137 | |
| 6138 | sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) { |
| 6139 | sctp_sock_rfree_frag(skb); |
| 6140 | sctp_skb_set_owner_r_frag(skb, newsk); |
| 6141 | } |
| 6142 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6143 | /* Set the type of socket to indicate that it is peeled off from the |
| 6144 | * original UDP-style socket or created with the accept() call on a |
| 6145 | * TCP-style socket.. |
| 6146 | */ |
| 6147 | newsp->type = type; |
| 6148 | |
Vladislav Yasevich | 61c9fed | 2006-05-19 11:01:18 -0700 | [diff] [blame] | 6149 | /* Mark the new socket "in-use" by the user so that any packets |
| 6150 | * that may arrive on the association after we've moved it are |
| 6151 | * queued to the backlog. This prevents a potential race between |
| 6152 | * backlog processing on the old socket and new-packet processing |
| 6153 | * on the new socket. |
Zach Brown | 5131a18 | 2007-06-22 15:14:46 -0700 | [diff] [blame] | 6154 | * |
| 6155 | * The caller has just allocated newsk so we can guarantee that other |
| 6156 | * paths won't try to lock it and then oldsk. |
Vladislav Yasevich | 61c9fed | 2006-05-19 11:01:18 -0700 | [diff] [blame] | 6157 | */ |
Zach Brown | 5131a18 | 2007-06-22 15:14:46 -0700 | [diff] [blame] | 6158 | lock_sock_nested(newsk, SINGLE_DEPTH_NESTING); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6159 | sctp_assoc_migrate(assoc, newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6160 | |
| 6161 | /* If the association on the newsk is already closed before accept() |
| 6162 | * is called, set RCV_SHUTDOWN flag. |
| 6163 | */ |
| 6164 | if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) |
| 6165 | newsk->sk_shutdown |= RCV_SHUTDOWN; |
| 6166 | |
| 6167 | newsk->sk_state = SCTP_SS_ESTABLISHED; |
Vladislav Yasevich | 61c9fed | 2006-05-19 11:01:18 -0700 | [diff] [blame] | 6168 | sctp_release_sock(newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6169 | } |
| 6170 | |
| 6171 | /* This proto struct describes the ULP interface for SCTP. */ |
| 6172 | struct proto sctp_prot = { |
| 6173 | .name = "SCTP", |
| 6174 | .owner = THIS_MODULE, |
| 6175 | .close = sctp_close, |
| 6176 | .connect = sctp_connect, |
| 6177 | .disconnect = sctp_disconnect, |
| 6178 | .accept = sctp_accept, |
| 6179 | .ioctl = sctp_ioctl, |
| 6180 | .init = sctp_init_sock, |
| 6181 | .destroy = sctp_destroy_sock, |
| 6182 | .shutdown = sctp_shutdown, |
| 6183 | .setsockopt = sctp_setsockopt, |
| 6184 | .getsockopt = sctp_getsockopt, |
| 6185 | .sendmsg = sctp_sendmsg, |
| 6186 | .recvmsg = sctp_recvmsg, |
| 6187 | .bind = sctp_bind, |
| 6188 | .backlog_rcv = sctp_backlog_rcv, |
| 6189 | .hash = sctp_hash, |
| 6190 | .unhash = sctp_unhash, |
| 6191 | .get_port = sctp_get_port, |
| 6192 | .obj_size = sizeof(struct sctp_sock), |
| 6193 | }; |
| 6194 | |
| 6195 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) |
| 6196 | struct proto sctpv6_prot = { |
| 6197 | .name = "SCTPv6", |
| 6198 | .owner = THIS_MODULE, |
| 6199 | .close = sctp_close, |
| 6200 | .connect = sctp_connect, |
| 6201 | .disconnect = sctp_disconnect, |
| 6202 | .accept = sctp_accept, |
| 6203 | .ioctl = sctp_ioctl, |
| 6204 | .init = sctp_init_sock, |
| 6205 | .destroy = sctp_destroy_sock, |
| 6206 | .shutdown = sctp_shutdown, |
| 6207 | .setsockopt = sctp_setsockopt, |
| 6208 | .getsockopt = sctp_getsockopt, |
| 6209 | .sendmsg = sctp_sendmsg, |
| 6210 | .recvmsg = sctp_recvmsg, |
| 6211 | .bind = sctp_bind, |
| 6212 | .backlog_rcv = sctp_backlog_rcv, |
| 6213 | .hash = sctp_hash, |
| 6214 | .unhash = sctp_unhash, |
| 6215 | .get_port = sctp_get_port, |
| 6216 | .obj_size = sizeof(struct sctp6_sock), |
| 6217 | }; |
| 6218 | #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */ |