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