1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */ 28 /* All Rights Reserved */ 29 30 /* 31 * University Copyright- Copyright (c) 1982, 1986, 1988 32 * The Regents of the University of California 33 * All Rights Reserved 34 * 35 * University Acknowledgment- Portions of this document are derived from 36 * software developed by the University of California, Berkeley, and its 37 * contributors. 38 */ 39 40 #ifndef _SYS_SOCKETVAR_H 41 #define _SYS_SOCKETVAR_H 42 43 #include <sys/types.h> 44 #include <sys/stream.h> 45 #include <sys/t_lock.h> 46 #include <sys/cred.h> 47 #include <sys/vnode.h> 48 #include <sys/file.h> 49 #include <sys/param.h> 50 #include <sys/zone.h> 51 #include <sys/sdt.h> 52 #include <sys/modctl.h> 53 #include <sys/atomic.h> 54 #include <sys/socket.h> 55 #include <sys/ksocket.h> 56 #include <sys/kstat.h> 57 58 #ifdef __cplusplus 59 extern "C" { 60 #endif 61 62 /* 63 * Internal representation of the address used to represent addresses 64 * in the loopback transport for AF_UNIX. While the sockaddr_un is used 65 * as the sockfs layer address for AF_UNIX the pathnames contained in 66 * these addresses are not unique (due to relative pathnames) thus can not 67 * be used in the transport. 68 * 69 * The transport level address consists of a magic number (used to separate the 70 * name space for specific and implicit binds). For a specific bind 71 * this is followed by a "vnode *" which ensures that all specific binds 72 * have a unique transport level address. For implicit binds the latter 73 * part of the address is a byte string (of the same length as a pointer) 74 * that is assigned by the loopback transport. 75 * 76 * The uniqueness assumes that the loopback transport has a separate namespace 77 * for sockets in order to avoid name conflicts with e.g. TLI use of the 78 * same transport. 79 */ 80 struct so_ux_addr { 81 void *soua_vp; /* vnode pointer or assigned by tl */ 82 uint_t soua_magic; /* See below */ 83 }; 84 85 #define SOU_MAGIC_EXPLICIT 0x75787670 /* "uxvp" */ 86 #define SOU_MAGIC_IMPLICIT 0x616e6f6e /* "anon" */ 87 88 struct sockaddr_ux { 89 sa_family_t sou_family; /* AF_UNIX */ 90 struct so_ux_addr sou_addr; 91 }; 92 93 #if defined(_KERNEL) || defined(_KMEMUSER) 94 95 #include <sys/socket_proto.h> 96 97 typedef struct sonodeops sonodeops_t; 98 typedef struct sonode sonode_t; 99 100 struct sodirect_s; 101 102 /* 103 * The sonode represents a socket. A sonode never exist in the file system 104 * name space and can not be opened using open() - only the socket, socketpair 105 * and accept calls create sonodes. 106 * 107 * The locking of sockfs uses the so_lock mutex plus the SOLOCKED and 108 * SOREADLOCKED flags in so_flag. The mutex protects all the state in the 109 * sonode. It is expected that the underlying transport protocol serializes 110 * socket operations, so sockfs will not normally not single-thread 111 * operations. However, certain sockets, including TPI based ones, can only 112 * handle one control operation at a time. The SOLOCKED flag is used to 113 * single-thread operations from sockfs users to prevent e.g. multiple bind() 114 * calls to operate on the same sonode concurrently. The SOREADLOCKED flag is 115 * used to ensure that only one thread sleeps in kstrgetmsg for a given 116 * sonode. This is needed to ensure atomic operation for things like 117 * MSG_WAITALL. 118 * 119 * The so_fallback_rwlock is used to ensure that for sockets that can 120 * fall back to TPI, the fallback is not initiated until all pending 121 * operations have completed. 122 * 123 * Note that so_lock is sometimes held across calls that might go to sleep 124 * (kmem_alloc and soallocproto*). This implies that no other lock in 125 * the system should be held when calling into sockfs; from the system call 126 * side or from strrput (in case of TPI based sockets). If locks are held 127 * while calling into sockfs the system might hang when running low on memory. 128 */ 129 struct sonode { 130 struct vnode *so_vnode; /* vnode associated with this sonode */ 131 132 sonodeops_t *so_ops; /* operations vector for this sonode */ 133 void *so_priv; /* sonode private data */ 134 135 krwlock_t so_fallback_rwlock; 136 kmutex_t so_lock; /* protects sonode fields */ 137 138 kcondvar_t so_state_cv; /* synchronize state changes */ 139 kcondvar_t so_want_cv; /* wait due to SOLOCKED */ 140 141 /* These fields are protected by so_lock */ 142 143 uint_t so_state; /* internal state flags SS_*, below */ 144 uint_t so_mode; /* characteristics on socket. SM_* */ 145 ushort_t so_flag; /* flags, see below */ 146 int so_count; /* count of opened references */ 147 148 sock_connid_t so_proto_connid; /* protocol generation number */ 149 150 ushort_t so_error; /* error affecting connection */ 151 152 struct sockparams *so_sockparams; /* vnode or socket module */ 153 /* Needed to recreate the same socket for accept */ 154 short so_family; 155 short so_type; 156 short so_protocol; 157 short so_version; /* From so_socket call */ 158 159 /* Accept queue */ 160 kmutex_t so_acceptq_lock; /* protects accept queue */ 161 struct sonode *so_acceptq_next; /* acceptq list node */ 162 struct sonode *so_acceptq_head; 163 struct sonode **so_acceptq_tail; 164 unsigned int so_acceptq_len; 165 unsigned int so_backlog; /* Listen backlog */ 166 kcondvar_t so_acceptq_cv; /* wait for new conn. */ 167 168 /* Options */ 169 short so_options; /* From socket call, see socket.h */ 170 struct linger so_linger; /* SO_LINGER value */ 171 #define so_sndbuf so_proto_props.sopp_txhiwat /* SO_SNDBUF value */ 172 #define so_sndlowat so_proto_props.sopp_txlowat /* tx low water mark */ 173 #define so_rcvbuf so_proto_props.sopp_rxhiwat /* SO_RCVBUF value */ 174 #define so_rcvlowat so_proto_props.sopp_rxlowat /* rx low water mark */ 175 #define so_max_addr_len so_proto_props.sopp_maxaddrlen 176 #define so_minpsz so_proto_props.sopp_minpsz 177 #define so_maxpsz so_proto_props.sopp_maxpsz 178 179 int so_xpg_rcvbuf; /* SO_RCVBUF value for XPG4 socket */ 180 clock_t so_sndtimeo; /* send timeout */ 181 clock_t so_rcvtimeo; /* recv timeout */ 182 183 mblk_t *so_oobmsg; /* outofline oob data */ 184 ssize_t so_oobmark; /* offset of the oob data */ 185 186 pid_t so_pgrp; /* pgrp for signals */ 187 188 cred_t *so_peercred; /* connected socket peer cred */ 189 pid_t so_cpid; /* connected socket peer cached pid */ 190 zoneid_t so_zoneid; /* opener's zoneid */ 191 192 struct pollhead so_poll_list; /* common pollhead */ 193 short so_pollev; /* events that should be generated */ 194 195 /* Receive */ 196 unsigned int so_rcv_queued; /* # bytes on both rcv lists */ 197 mblk_t *so_rcv_q_head; /* processing/copyout rcv queue */ 198 mblk_t *so_rcv_q_last_head; 199 mblk_t *so_rcv_head; /* protocol prequeue */ 200 mblk_t *so_rcv_last_head; /* last mblk in b_next chain */ 201 kcondvar_t so_rcv_cv; /* wait for data */ 202 uint_t so_rcv_wanted; /* # of bytes wanted by app */ 203 timeout_id_t so_rcv_timer_tid; 204 205 #define so_rcv_thresh so_proto_props.sopp_rcvthresh 206 #define so_rcv_timer_interval so_proto_props.sopp_rcvtimer 207 208 kcondvar_t so_snd_cv; /* wait for snd buffers */ 209 uint32_t 210 so_snd_qfull: 1, /* Transmit full */ 211 so_rcv_wakeup: 1, 212 so_snd_wakeup: 1, 213 so_not_str: 1, /* B_TRUE if not streams based socket */ 214 so_pad_to_bit_31: 28; 215 216 /* Communication channel with protocol */ 217 sock_lower_handle_t so_proto_handle; 218 sock_downcalls_t *so_downcalls; 219 220 struct sock_proto_props so_proto_props; /* protocol settings */ 221 boolean_t so_flowctrld; /* Flow controlled */ 222 uint_t so_copyflag; /* Copy related flag */ 223 kcondvar_t so_copy_cv; /* Copy cond variable */ 224 225 /* kernel sockets */ 226 ksocket_callbacks_t so_ksock_callbacks; 227 void *so_ksock_cb_arg; /* callback argument */ 228 kcondvar_t so_closing_cv; 229 230 /* != NULL for sodirect enabled socket */ 231 struct sodirect_s *so_direct; 232 }; 233 234 #define SO_HAVE_DATA(so) \ 235 /* \ 236 * For the (tid == 0) case we must check so_rcv_{q_,}head \ 237 * rather than (so_rcv_queued > 0), since the latter does not \ 238 * take into account mblks with only control/name information. \ 239 */ \ 240 ((so)->so_rcv_timer_tid == 0 && ((so)->so_rcv_head != NULL || \ 241 (so)->so_rcv_q_head != NULL)) || \ 242 ((so)->so_state & SS_CANTRCVMORE) 243 244 /* 245 * Events handled by the protocol (in case sd_poll is set) 246 */ 247 #define SO_PROTO_POLLEV (POLLIN|POLLRDNORM|POLLRDBAND) 248 249 250 #endif /* _KERNEL || _KMEMUSER */ 251 252 /* flags */ 253 #define SOMOD 0x0001 /* update socket modification time */ 254 #define SOACC 0x0002 /* update socket access time */ 255 256 #define SOLOCKED 0x0010 /* use to serialize open/closes */ 257 #define SOREADLOCKED 0x0020 /* serialize kstrgetmsg calls */ 258 #define SOWANT 0x0040 /* some process waiting on lock */ 259 #define SOCLONE 0x0080 /* child of clone driver */ 260 #define SOASYNC_UNBIND 0x0100 /* wait for ACK of async unbind */ 261 262 #define SOCK_IS_NONSTR(so) ((so)->so_not_str) 263 264 /* 265 * Socket state bits. 266 */ 267 #define SS_ISCONNECTED 0x00000001 /* socket connected to a peer */ 268 #define SS_ISCONNECTING 0x00000002 /* in process, connecting to peer */ 269 #define SS_ISDISCONNECTING 0x00000004 /* in process of disconnecting */ 270 #define SS_CANTSENDMORE 0x00000008 /* can't send more data to peer */ 271 272 #define SS_CANTRCVMORE 0x00000010 /* can't receive more data */ 273 #define SS_ISBOUND 0x00000020 /* socket is bound */ 274 #define SS_NDELAY 0x00000040 /* FNDELAY non-blocking */ 275 #define SS_NONBLOCK 0x00000080 /* O_NONBLOCK non-blocking */ 276 277 #define SS_ASYNC 0x00000100 /* async i/o notify */ 278 #define SS_ACCEPTCONN 0x00000200 /* listen done */ 279 /* unused 0x00000400 */ /* was SS_HASCONNIND */ 280 #define SS_SAVEDEOR 0x00000800 /* Saved MSG_EOR rcv side state */ 281 282 #define SS_RCVATMARK 0x00001000 /* at mark on input */ 283 #define SS_OOBPEND 0x00002000 /* OOB pending or present - poll */ 284 #define SS_HAVEOOBDATA 0x00004000 /* OOB data present */ 285 #define SS_HADOOBDATA 0x00008000 /* OOB data consumed */ 286 #define SS_CLOSING 0x00010000 /* in process of closing */ 287 288 /* unused 0x00020000 */ /* was SS_FADDR_NOXLATE */ 289 /* unused 0x00040000 */ /* was SS_HASDATA */ 290 /* unused 0x00080000 */ /* was SS_DONEREAD */ 291 /* unused 0x00100000 */ /* was SS_MOREDATA */ 292 /* unused 0x00200000 */ /* was SS_DIRECT */ 293 294 #define SS_SODIRECT 0x00400000 /* transport supports sodirect */ 295 296 #define SS_SENTLASTREADSIG 0x01000000 /* last rx signal has been sent */ 297 #define SS_SENTLASTWRITESIG 0x02000000 /* last tx signal has been sent */ 298 299 #define SS_FALLBACK_DRAIN 0x20000000 /* data was/is being drained */ 300 #define SS_FALLBACK_PENDING 0x40000000 /* fallback is pending */ 301 #define SS_FALLBACK_COMP 0x80000000 /* fallback has completed */ 302 303 304 /* Set of states when the socket can't be rebound */ 305 #define SS_CANTREBIND (SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING|\ 306 SS_CANTSENDMORE|SS_CANTRCVMORE|SS_ACCEPTCONN) 307 308 /* 309 * Sockets that can fall back to TPI must ensure that fall back is not 310 * initiated while a thread is using a socket. 311 */ 312 #define SO_BLOCK_FALLBACK(so, fn) { \ 313 ASSERT(MUTEX_NOT_HELD(&(so)->so_lock)); \ 314 rw_enter(&(so)->so_fallback_rwlock, RW_READER); \ 315 if ((so)->so_state & SS_FALLBACK_COMP) { \ 316 rw_exit(&(so)->so_fallback_rwlock); \ 317 return (fn); \ 318 } \ 319 } 320 321 #define SO_UNBLOCK_FALLBACK(so) { \ 322 rw_exit(&(so)->so_fallback_rwlock); \ 323 } 324 325 /* Poll events */ 326 #define SO_POLLEV_IN 0x1 /* POLLIN wakeup needed */ 327 #define SO_POLLEV_ALWAYS 0x2 /* wakeups */ 328 329 /* 330 * Characteristics of sockets. Not changed after the socket is created. 331 */ 332 #define SM_PRIV 0x001 /* privileged for broadcast, raw... */ 333 #define SM_ATOMIC 0x002 /* atomic data transmission */ 334 #define SM_ADDR 0x004 /* addresses given with messages */ 335 #define SM_CONNREQUIRED 0x008 /* connection required by protocol */ 336 337 #define SM_FDPASSING 0x010 /* passes file descriptors */ 338 #define SM_EXDATA 0x020 /* Can handle T_EXDATA_REQ */ 339 #define SM_OPTDATA 0x040 /* Can handle T_OPTDATA_REQ */ 340 #define SM_BYTESTREAM 0x080 /* Byte stream - can use M_DATA */ 341 342 #define SM_ACCEPTOR_ID 0x100 /* so_acceptor_id is valid */ 343 344 #define SM_KERNEL 0x200 /* kernel socket */ 345 346 /* The modes below are only for non-streams sockets */ 347 #define SM_ACCEPTSUPP 0x400 /* can handle accept() */ 348 #define SM_SENDFILESUPP 0x800 /* Private: proto supp sendfile */ 349 350 /* 351 * Socket versions. Used by the socket library when calling _so_socket(). 352 */ 353 #define SOV_STREAM 0 /* Not a socket - just a stream */ 354 #define SOV_DEFAULT 1 /* Select based on so_default_version */ 355 #define SOV_SOCKSTREAM 2 /* Socket plus streams operations */ 356 #define SOV_SOCKBSD 3 /* Socket with no streams operations */ 357 #define SOV_XPG4_2 4 /* Xnet socket */ 358 359 #if defined(_KERNEL) || defined(_KMEMUSER) 360 361 /* 362 * sonode create and destroy functions. 363 */ 364 typedef struct sonode *(*so_create_func_t)(struct sockparams *, 365 int, int, int, int, int, int *, cred_t *); 366 typedef void (*so_destroy_func_t)(struct sonode *); 367 368 /* STREAM device information */ 369 typedef struct sdev_info { 370 char *sd_devpath; 371 int sd_devpathlen; /* Is 0 if sp_devpath is a static string */ 372 vnode_t *sd_vnode; 373 } sdev_info_t; 374 375 #define SOCKMOD_VERSION 1 376 /* name of the TPI pseudo socket module */ 377 #define SOTPI_SMOD_NAME "socktpi" 378 379 typedef struct __smod_priv_s { 380 so_create_func_t smodp_sock_create_func; 381 so_destroy_func_t smodp_sock_destroy_func; 382 so_proto_fallback_func_t smodp_proto_fallback_func; 383 } __smod_priv_t; 384 385 /* 386 * Socket module register information 387 */ 388 typedef struct smod_reg_s { 389 int smod_version; 390 char *smod_name; 391 size_t smod_uc_version; 392 size_t smod_dc_version; 393 so_proto_create_func_t smod_proto_create_func; 394 395 /* __smod_priv_data must be NULL */ 396 __smod_priv_t *__smod_priv; 397 } smod_reg_t; 398 399 /* 400 * Socket module information 401 */ 402 typedef struct smod_info { 403 int smod_version; 404 char *smod_name; 405 uint_t smod_refcnt; /* # of entries */ 406 size_t smod_uc_version; /* upcall version */ 407 size_t smod_dc_version; /* down call version */ 408 so_proto_create_func_t smod_proto_create_func; 409 so_proto_fallback_func_t smod_proto_fallback_func; 410 so_create_func_t smod_sock_create_func; 411 so_destroy_func_t smod_sock_destroy_func; 412 list_node_t smod_node; 413 } smod_info_t; 414 415 typedef struct sockparams_stats { 416 kstat_named_t sps_nfallback; /* # of fallbacks to TPI */ 417 kstat_named_t sps_nactive; /* # of active sockets */ 418 kstat_named_t sps_ncreate; /* total # of created sockets */ 419 } sockparams_stats_t; 420 421 /* 422 * sockparams 423 * 424 * Used for mapping family/type/protocol to module 425 */ 426 struct sockparams { 427 /* 428 * The family, type, protocol, sdev_info and smod_info are 429 * set when the entry is created, and they will never change 430 * thereafter. 431 */ 432 int sp_family; 433 int sp_type; 434 int sp_protocol; 435 436 sdev_info_t sp_sdev_info; /* STREAM device */ 437 char *sp_smod_name; /* socket module name */ 438 smod_info_t *sp_smod_info; /* socket module */ 439 440 kmutex_t sp_lock; /* lock for refcnt */ 441 uint64_t sp_refcnt; /* entry reference count */ 442 sockparams_stats_t sp_stats; 443 kstat_t *sp_kstat; 444 445 /* 446 * The entries below are only modified while holding 447 * splist_lock as a writer. 448 */ 449 int sp_flags; /* see below */ 450 list_node_t sp_node; 451 }; 452 453 454 /* 455 * sockparams flags 456 */ 457 #define SOCKPARAMS_EPHEMERAL 0x1 /* temp. entry, not on global list */ 458 459 extern void sockparams_init(void); 460 extern struct sockparams *sockparams_hold_ephemeral_bydev(int, int, int, 461 const char *, int, int *); 462 extern struct sockparams *sockparams_hold_ephemeral_bymod(int, int, int, 463 const char *, int, int *); 464 extern void sockparams_ephemeral_drop_last_ref(struct sockparams *); 465 466 extern void smod_init(void); 467 extern void smod_add(smod_info_t *); 468 extern int smod_register(const smod_reg_t *); 469 extern int smod_unregister(const char *); 470 extern smod_info_t *smod_lookup_byname(const char *); 471 472 #define SOCKPARAMS_HAS_DEVICE(sp) \ 473 ((sp)->sp_sdev_info.sd_devpath != NULL) 474 475 /* Increase the smod_info_t reference count */ 476 #define SMOD_INC_REF(smodp) { \ 477 ASSERT((smodp) != NULL); \ 478 DTRACE_PROBE1(smodinfo__inc__ref, struct smod_info *, (smodp)); \ 479 atomic_inc_uint(&(smodp)->smod_refcnt); \ 480 } 481 482 /* 483 * Decreace the socket module entry reference count. 484 * When no one mapping to the entry, we try to unload the module from the 485 * kernel. If the module can't unload, just leave the module entry with 486 * a zero refcnt. 487 */ 488 #define SMOD_DEC_REF(sp, smodp) { \ 489 ASSERT((smodp) != NULL); \ 490 ASSERT((smodp)->smod_refcnt != 0); \ 491 atomic_dec_uint(&(smodp)->smod_refcnt); \ 492 /* \ 493 * No need to atomically check the return value because the \ 494 * socket module framework will verify that no one is using \ 495 * the module before unloading. Worst thing that can happen \ 496 * here is multiple calls to mod_remove_by_name(), which is OK. \ 497 */ \ 498 if ((smodp)->smod_refcnt == 0) \ 499 (void) mod_remove_by_name((sp)->sp_smod_name); \ 500 } 501 502 /* Increase the reference count */ 503 #define SOCKPARAMS_INC_REF(sp) { \ 504 ASSERT((sp) != NULL); \ 505 DTRACE_PROBE1(sockparams__inc__ref, struct sockparams *, (sp)); \ 506 mutex_enter(&(sp)->sp_lock); \ 507 (sp)->sp_refcnt++; \ 508 ASSERT((sp)->sp_refcnt != 0); \ 509 mutex_exit(&(sp)->sp_lock); \ 510 } 511 512 /* 513 * Decrease the reference count. 514 * 515 * If the sockparams is ephemeral, then the thread dropping the last ref 516 * count will destroy the entry. 517 */ 518 #define SOCKPARAMS_DEC_REF(sp) { \ 519 ASSERT((sp) != NULL); \ 520 DTRACE_PROBE1(sockparams__dec__ref, struct sockparams *, (sp)); \ 521 mutex_enter(&(sp)->sp_lock); \ 522 ASSERT((sp)->sp_refcnt > 0); \ 523 if ((sp)->sp_refcnt == 1) { \ 524 if ((sp)->sp_flags & SOCKPARAMS_EPHEMERAL) { \ 525 mutex_exit(&(sp)->sp_lock); \ 526 sockparams_ephemeral_drop_last_ref((sp)); \ 527 } else { \ 528 (sp)->sp_refcnt--; \ 529 if ((sp)->sp_smod_info != NULL) \ 530 SMOD_DEC_REF(sp, (sp)->sp_smod_info); \ 531 (sp)->sp_smod_info = NULL; \ 532 mutex_exit(&(sp)->sp_lock); \ 533 } \ 534 } else { \ 535 (sp)->sp_refcnt--; \ 536 mutex_exit(&(sp)->sp_lock); \ 537 } \ 538 } 539 540 /* 541 * Used to traverse the list of AF_UNIX sockets to construct the kstat 542 * for netstat(1m). 543 */ 544 struct socklist { 545 kmutex_t sl_lock; 546 struct sonode *sl_list; 547 }; 548 549 extern struct socklist socklist; 550 /* 551 * ss_full_waits is the number of times the reader thread 552 * waits when the queue is full and ss_empty_waits is the number 553 * of times the consumer thread waits when the queue is empty. 554 * No locks for these as they are just indicators of whether 555 * disk or network or both is slow or fast. 556 */ 557 struct sendfile_stats { 558 uint32_t ss_file_cached; 559 uint32_t ss_file_not_cached; 560 uint32_t ss_full_waits; 561 uint32_t ss_empty_waits; 562 uint32_t ss_file_segmap; 563 }; 564 565 /* 566 * A single sendfile request is represented by snf_req. 567 */ 568 typedef struct snf_req { 569 struct snf_req *sr_next; 570 mblk_t *sr_mp_head; 571 mblk_t *sr_mp_tail; 572 kmutex_t sr_lock; 573 kcondvar_t sr_cv; 574 uint_t sr_qlen; 575 int sr_hiwat; 576 int sr_lowat; 577 int sr_operation; 578 struct vnode *sr_vp; 579 file_t *sr_fp; 580 ssize_t sr_maxpsz; 581 u_offset_t sr_file_off; 582 u_offset_t sr_file_size; 583 #define SR_READ_DONE 0x80000000 584 int sr_read_error; 585 int sr_write_error; 586 } snf_req_t; 587 588 /* A queue of sendfile requests */ 589 struct sendfile_queue { 590 snf_req_t *snfq_req_head; 591 snf_req_t *snfq_req_tail; 592 kmutex_t snfq_lock; 593 kcondvar_t snfq_cv; 594 int snfq_svc_threads; /* # of service threads */ 595 int snfq_idle_cnt; /* # of idling threads */ 596 int snfq_max_threads; 597 int snfq_req_cnt; /* Number of requests */ 598 }; 599 600 #define READ_OP 1 601 #define SNFQ_TIMEOUT (60 * 5 * hz) /* 5 minutes */ 602 603 /* Socket network operations switch */ 604 struct sonodeops { 605 int (*sop_init)(struct sonode *, struct sonode *, cred_t *, 606 int); 607 int (*sop_accept)(struct sonode *, int, cred_t *, struct sonode **); 608 int (*sop_bind)(struct sonode *, struct sockaddr *, socklen_t, 609 int, cred_t *); 610 int (*sop_listen)(struct sonode *, int, cred_t *); 611 int (*sop_connect)(struct sonode *, const struct sockaddr *, 612 socklen_t, int, int, cred_t *); 613 int (*sop_recvmsg)(struct sonode *, struct msghdr *, 614 struct uio *, cred_t *); 615 int (*sop_sendmsg)(struct sonode *, struct msghdr *, 616 struct uio *, cred_t *); 617 int (*sop_sendmblk)(struct sonode *, struct msghdr *, int, 618 cred_t *, mblk_t **); 619 int (*sop_getpeername)(struct sonode *, struct sockaddr *, 620 socklen_t *, boolean_t, cred_t *); 621 int (*sop_getsockname)(struct sonode *, struct sockaddr *, 622 socklen_t *, cred_t *); 623 int (*sop_shutdown)(struct sonode *, int, cred_t *); 624 int (*sop_getsockopt)(struct sonode *, int, int, void *, 625 socklen_t *, int, cred_t *); 626 int (*sop_setsockopt)(struct sonode *, int, int, const void *, 627 socklen_t, cred_t *); 628 int (*sop_ioctl)(struct sonode *, int, intptr_t, int, 629 cred_t *, int32_t *); 630 int (*sop_poll)(struct sonode *, short, int, short *, 631 struct pollhead **); 632 int (*sop_close)(struct sonode *, int, cred_t *); 633 }; 634 635 #define SOP_INIT(so, flag, cr, flags) \ 636 ((so)->so_ops->sop_init((so), (flag), (cr), (flags))) 637 #define SOP_ACCEPT(so, fflag, cr, nsop) \ 638 ((so)->so_ops->sop_accept((so), (fflag), (cr), (nsop))) 639 #define SOP_BIND(so, name, namelen, flags, cr) \ 640 ((so)->so_ops->sop_bind((so), (name), (namelen), (flags), (cr))) 641 #define SOP_LISTEN(so, backlog, cr) \ 642 ((so)->so_ops->sop_listen((so), (backlog), (cr))) 643 #define SOP_CONNECT(so, name, namelen, fflag, flags, cr) \ 644 ((so)->so_ops->sop_connect((so), (name), (namelen), (fflag), (flags), \ 645 (cr))) 646 #define SOP_RECVMSG(so, msg, uiop, cr) \ 647 ((so)->so_ops->sop_recvmsg((so), (msg), (uiop), (cr))) 648 #define SOP_SENDMSG(so, msg, uiop, cr) \ 649 ((so)->so_ops->sop_sendmsg((so), (msg), (uiop), (cr))) 650 #define SOP_SENDMBLK(so, msg, size, cr, mpp) \ 651 ((so)->so_ops->sop_sendmblk((so), (msg), (size), (cr), (mpp))) 652 #define SOP_GETPEERNAME(so, addr, addrlen, accept, cr) \ 653 ((so)->so_ops->sop_getpeername((so), (addr), (addrlen), (accept), (cr))) 654 #define SOP_GETSOCKNAME(so, addr, addrlen, cr) \ 655 ((so)->so_ops->sop_getsockname((so), (addr), (addrlen), (cr))) 656 #define SOP_SHUTDOWN(so, how, cr) \ 657 ((so)->so_ops->sop_shutdown((so), (how), (cr))) 658 #define SOP_GETSOCKOPT(so, level, optionname, optval, optlenp, flags, cr) \ 659 ((so)->so_ops->sop_getsockopt((so), (level), (optionname), \ 660 (optval), (optlenp), (flags), (cr))) 661 #define SOP_SETSOCKOPT(so, level, optionname, optval, optlen, cr) \ 662 ((so)->so_ops->sop_setsockopt((so), (level), (optionname), \ 663 (optval), (optlen), (cr))) 664 #define SOP_IOCTL(so, cmd, arg, mode, cr, rvalp) \ 665 ((so)->so_ops->sop_ioctl((so), (cmd), (arg), (mode), (cr), (rvalp))) 666 #define SOP_POLL(so, events, anyyet, reventsp, phpp) \ 667 ((so)->so_ops->sop_poll((so), (events), (anyyet), (reventsp), (phpp))) 668 #define SOP_CLOSE(so, flag, cr) \ 669 ((so)->so_ops->sop_close((so), (flag), (cr))) 670 671 #endif /* defined(_KERNEL) || defined(_KMEMUSER) */ 672 673 #ifdef _KERNEL 674 675 #define ISALIGNED_cmsghdr(addr) \ 676 (((uintptr_t)(addr) & (_CMSG_HDR_ALIGNMENT - 1)) == 0) 677 678 #define ROUNDUP_cmsglen(len) \ 679 (((len) + _CMSG_HDR_ALIGNMENT - 1) & ~(_CMSG_HDR_ALIGNMENT - 1)) 680 681 #define IS_NON_STREAM_SOCK(vp) \ 682 ((vp)->v_type == VSOCK && (vp)->v_stream == NULL) 683 /* 684 * Macros that operate on struct cmsghdr. 685 * Used in parsing msg_control. 686 * The CMSG_VALID macro does not assume that the last option buffer is padded. 687 */ 688 #define CMSG_NEXT(cmsg) \ 689 (struct cmsghdr *)((uintptr_t)(cmsg) + \ 690 ROUNDUP_cmsglen((cmsg)->cmsg_len)) 691 #define CMSG_CONTENT(cmsg) (&((cmsg)[1])) 692 #define CMSG_CONTENTLEN(cmsg) ((cmsg)->cmsg_len - sizeof (struct cmsghdr)) 693 #define CMSG_VALID(cmsg, start, end) \ 694 (ISALIGNED_cmsghdr(cmsg) && \ 695 ((uintptr_t)(cmsg) >= (uintptr_t)(start)) && \ 696 ((uintptr_t)(cmsg) < (uintptr_t)(end)) && \ 697 ((ssize_t)(cmsg)->cmsg_len >= sizeof (struct cmsghdr)) && \ 698 ((uintptr_t)(cmsg) + (cmsg)->cmsg_len <= (uintptr_t)(end))) 699 700 /* 701 * Maximum size of any argument that is copied in (addresses, options, 702 * access rights). MUST be at least MAXPATHLEN + 3. 703 * BSD and SunOS 4.X limited this to MLEN or MCLBYTES. 704 */ 705 #define SO_MAXARGSIZE 8192 706 707 /* 708 * Convert between vnode and sonode 709 */ 710 #define VTOSO(vp) ((struct sonode *)((vp)->v_data)) 711 #define SOTOV(sp) ((sp)->so_vnode) 712 713 /* 714 * Internal flags for sobind() 715 */ 716 #define _SOBIND_REBIND 0x01 /* Bind to existing local address */ 717 #define _SOBIND_UNSPEC 0x02 /* Bind to unspecified address */ 718 #define _SOBIND_LOCK_HELD 0x04 /* so_excl_lock held by caller */ 719 #define _SOBIND_NOXLATE 0x08 /* No addr translation for AF_UNIX */ 720 #define _SOBIND_XPG4_2 0x10 /* xpg4.2 semantics */ 721 #define _SOBIND_SOCKBSD 0x20 /* BSD semantics */ 722 #define _SOBIND_LISTEN 0x40 /* Make into SS_ACCEPTCONN */ 723 #define _SOBIND_SOCKETPAIR 0x80 /* Internal flag for so_socketpair() */ 724 /* to enable listen with backlog = 1 */ 725 726 /* 727 * Internal flags for sounbind() 728 */ 729 #define _SOUNBIND_REBIND 0x01 /* Don't clear fields - will rebind */ 730 731 /* 732 * Internal flags for soconnect() 733 */ 734 #define _SOCONNECT_NOXLATE 0x01 /* No addr translation for AF_UNIX */ 735 #define _SOCONNECT_DID_BIND 0x02 /* Unbind when connect fails */ 736 #define _SOCONNECT_XPG4_2 0x04 /* xpg4.2 semantics */ 737 738 /* 739 * Internal flags for sodisconnect() 740 */ 741 #define _SODISCONNECT_LOCK_HELD 0x01 /* so_excl_lock held by caller */ 742 743 /* 744 * Internal flags for sotpi_getsockopt(). 745 */ 746 #define _SOGETSOCKOPT_XPG4_2 0x01 /* xpg4.2 semantics */ 747 748 /* 749 * Internal flags for soallocproto*() 750 */ 751 #define _ALLOC_NOSLEEP 0 /* Don't sleep for memory */ 752 #define _ALLOC_INTR 1 /* Sleep until interrupt */ 753 #define _ALLOC_SLEEP 2 /* Sleep forever */ 754 755 /* 756 * Internal structure for handling AF_UNIX file descriptor passing 757 */ 758 struct fdbuf { 759 int fd_size; /* In bytes, for kmem_free */ 760 int fd_numfd; /* Number of elements below */ 761 char *fd_ebuf; /* Extra buffer to free */ 762 int fd_ebuflen; 763 frtn_t fd_frtn; 764 struct file *fd_fds[1]; /* One or more */ 765 }; 766 #define FDBUF_HDRSIZE (sizeof (struct fdbuf) - sizeof (struct file *)) 767 768 /* 769 * Variable that can be patched to set what version of socket socket() 770 * will create. 771 */ 772 extern int so_default_version; 773 774 #ifdef DEBUG 775 /* Turn on extra testing capabilities */ 776 #define SOCK_TEST 777 #endif /* DEBUG */ 778 779 #ifdef DEBUG 780 char *pr_state(uint_t, uint_t); 781 char *pr_addr(int, struct sockaddr *, t_uscalar_t); 782 int so_verify_oobstate(struct sonode *); 783 #endif /* DEBUG */ 784 785 /* 786 * DEBUG macros 787 */ 788 #if defined(DEBUG) 789 #define SOCK_DEBUG 790 791 extern int sockdebug; 792 extern int sockprinterr; 793 794 #define eprint(args) printf args 795 #define eprintso(so, args) \ 796 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) printf args; } 797 #define eprintline(error) \ 798 { \ 799 if (error != EINTR && (sockprinterr || sockdebug > 0)) \ 800 printf("socket error %d: line %d file %s\n", \ 801 (error), __LINE__, __FILE__); \ 802 } 803 804 #define eprintsoline(so, error) \ 805 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) \ 806 printf("socket(%p) error %d: line %d file %s\n", \ 807 (void *)(so), (error), __LINE__, __FILE__); \ 808 } 809 #define dprint(level, args) { if (sockdebug > (level)) printf args; } 810 #define dprintso(so, level, args) \ 811 { if (sockdebug > (level) && ((so)->so_options & SO_DEBUG)) printf args; } 812 813 #else /* define(DEBUG) */ 814 815 #define eprint(args) {} 816 #define eprintso(so, args) {} 817 #define eprintline(error) {} 818 #define eprintsoline(so, error) {} 819 #define dprint(level, args) {} 820 #define dprintso(so, level, args) {} 821 822 #endif /* defined(DEBUG) */ 823 824 extern struct vfsops sock_vfsops; 825 extern struct vnodeops *socket_vnodeops; 826 extern const struct fs_operation_def socket_vnodeops_template[]; 827 828 extern dev_t sockdev; 829 830 /* 831 * sockfs functions 832 */ 833 extern int sock_getmsg(vnode_t *, struct strbuf *, struct strbuf *, 834 uchar_t *, int *, int, rval_t *); 835 extern int sock_putmsg(vnode_t *, struct strbuf *, struct strbuf *, 836 uchar_t, int, int); 837 extern int sogetvp(char *, vnode_t **, int); 838 extern int sockinit(int, char *); 839 extern int soconfig(int, int, int, char *, int, char *); 840 extern int solookup(int, int, int, struct sockparams **); 841 extern void so_lock_single(struct sonode *); 842 extern void so_unlock_single(struct sonode *, int); 843 extern int so_lock_read(struct sonode *, int); 844 extern int so_lock_read_intr(struct sonode *, int); 845 extern void so_unlock_read(struct sonode *); 846 extern void *sogetoff(mblk_t *, t_uscalar_t, t_uscalar_t, uint_t); 847 extern void so_getopt_srcaddr(void *, t_uscalar_t, 848 void **, t_uscalar_t *); 849 extern int so_getopt_unix_close(void *, t_uscalar_t); 850 extern void fdbuf_free(struct fdbuf *); 851 extern mblk_t *fdbuf_allocmsg(int, struct fdbuf *); 852 extern int fdbuf_create(void *, int, struct fdbuf **); 853 extern void so_closefds(void *, t_uscalar_t, int, int); 854 extern int so_getfdopt(void *, t_uscalar_t, int, void **, int *); 855 t_uscalar_t so_optlen(void *, t_uscalar_t, int); 856 extern void so_cmsg2opt(void *, t_uscalar_t, int, mblk_t *); 857 extern t_uscalar_t 858 so_cmsglen(mblk_t *, void *, t_uscalar_t, int); 859 extern int so_opt2cmsg(mblk_t *, void *, t_uscalar_t, int, 860 void *, t_uscalar_t); 861 extern void soisconnecting(struct sonode *); 862 extern void soisconnected(struct sonode *); 863 extern void soisdisconnected(struct sonode *, int); 864 extern void socantsendmore(struct sonode *); 865 extern void socantrcvmore(struct sonode *); 866 extern void soseterror(struct sonode *, int); 867 extern int sogeterr(struct sonode *, boolean_t); 868 extern int sowaitconnected(struct sonode *, int, int); 869 870 extern ssize_t soreadfile(file_t *, uchar_t *, u_offset_t, int *, size_t); 871 extern void *sock_kstat_init(zoneid_t); 872 extern void sock_kstat_fini(zoneid_t, void *); 873 extern struct sonode *getsonode(int, int *, file_t **); 874 /* 875 * Function wrappers (mostly around the sonode switch) for 876 * backward compatibility. 877 */ 878 extern int soaccept(struct sonode *, int, struct sonode **); 879 extern int sobind(struct sonode *, struct sockaddr *, socklen_t, 880 int, int); 881 extern int solisten(struct sonode *, int); 882 extern int soconnect(struct sonode *, const struct sockaddr *, socklen_t, 883 int, int); 884 extern int sorecvmsg(struct sonode *, struct nmsghdr *, struct uio *); 885 extern int sosendmsg(struct sonode *, struct nmsghdr *, struct uio *); 886 extern int soshutdown(struct sonode *, int); 887 extern int sogetsockopt(struct sonode *, int, int, void *, socklen_t *, 888 int); 889 extern int sosetsockopt(struct sonode *, int, int, const void *, 890 t_uscalar_t); 891 892 extern struct sonode *socreate(struct sockparams *, int, int, int, int, 893 int *); 894 895 extern int so_copyin(const void *, void *, size_t, int); 896 extern int so_copyout(const void *, void *, size_t, int); 897 898 #endif 899 900 /* 901 * Internal structure for obtaining sonode information from the socklist. 902 * These types match those corresponding in the sonode structure. 903 * This is not a published interface, and may change at any time. 904 */ 905 struct sockinfo { 906 uint_t si_size; /* real length of this struct */ 907 short si_family; 908 short si_type; 909 ushort_t si_flag; 910 uint_t si_state; 911 uint_t si_ux_laddr_sou_magic; 912 uint_t si_ux_faddr_sou_magic; 913 t_scalar_t si_serv_type; 914 t_uscalar_t si_laddr_soa_len; 915 t_uscalar_t si_faddr_soa_len; 916 uint16_t si_laddr_family; 917 uint16_t si_faddr_family; 918 char si_laddr_sun_path[MAXPATHLEN + 1]; /* NULL terminated */ 919 char si_faddr_sun_path[MAXPATHLEN + 1]; 920 boolean_t si_faddr_noxlate; 921 zoneid_t si_szoneid; 922 }; 923 924 #define SOCKMOD_PATH "socketmod" /* dir where sockmods are stored */ 925 926 #ifdef __cplusplus 927 } 928 #endif 929 930 #endif /* _SYS_SOCKETVAR_H */ 931