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 (c) 1996, 2010, Oracle and/or its affiliates. All rights reserved. 24 */ 25 26 /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */ 27 /* All Rights Reserved */ 28 29 /* 30 * University Copyright- Copyright (c) 1982, 1986, 1988 31 * The Regents of the University of California 32 * All Rights Reserved 33 * 34 * University Acknowledgment- Portions of this document are derived from 35 * software developed by the University of California, Berkeley, and its 36 * contributors. 37 */ 38 39 #ifndef _SYS_SOCKETVAR_H 40 #define _SYS_SOCKETVAR_H 41 42 #include <sys/types.h> 43 #include <sys/stream.h> 44 #include <sys/t_lock.h> 45 #include <sys/cred.h> 46 #include <sys/vnode.h> 47 #include <sys/file.h> 48 #include <sys/param.h> 49 #include <sys/zone.h> 50 #include <sys/sdt.h> 51 #include <sys/modctl.h> 52 #include <sys/atomic.h> 53 #include <sys/socket.h> 54 #include <sys/ksocket.h> 55 #include <sys/kstat.h> 56 57 #ifdef _KERNEL 58 #include <sys/vfs_opreg.h> 59 #endif 60 61 #ifdef __cplusplus 62 extern "C" { 63 #endif 64 65 /* 66 * Internal representation of the address used to represent addresses 67 * in the loopback transport for AF_UNIX. While the sockaddr_un is used 68 * as the sockfs layer address for AF_UNIX the pathnames contained in 69 * these addresses are not unique (due to relative pathnames) thus can not 70 * be used in the transport. 71 * 72 * The transport level address consists of a magic number (used to separate the 73 * name space for specific and implicit binds). For a specific bind 74 * this is followed by a "vnode *" which ensures that all specific binds 75 * have a unique transport level address. For implicit binds the latter 76 * part of the address is a byte string (of the same length as a pointer) 77 * that is assigned by the loopback transport. 78 * 79 * The uniqueness assumes that the loopback transport has a separate namespace 80 * for sockets in order to avoid name conflicts with e.g. TLI use of the 81 * same transport. 82 */ 83 struct so_ux_addr { 84 void *soua_vp; /* vnode pointer or assigned by tl */ 85 uint_t soua_magic; /* See below */ 86 }; 87 88 #define SOU_MAGIC_EXPLICIT 0x75787670 /* "uxvp" */ 89 #define SOU_MAGIC_IMPLICIT 0x616e6f6e /* "anon" */ 90 91 struct sockaddr_ux { 92 sa_family_t sou_family; /* AF_UNIX */ 93 struct so_ux_addr sou_addr; 94 }; 95 96 #if defined(_KERNEL) || defined(_KMEMUSER) 97 98 #include <sys/socket_proto.h> 99 100 typedef struct sonodeops sonodeops_t; 101 typedef struct sonode sonode_t; 102 103 struct sodirect_s; 104 105 /* 106 * The sonode represents a socket. A sonode never exist in the file system 107 * name space and can not be opened using open() - only the socket, socketpair 108 * and accept calls create sonodes. 109 * 110 * The locking of sockfs uses the so_lock mutex plus the SOLOCKED and 111 * SOREADLOCKED flags in so_flag. The mutex protects all the state in the 112 * sonode. It is expected that the underlying transport protocol serializes 113 * socket operations, so sockfs will not normally not single-thread 114 * operations. However, certain sockets, including TPI based ones, can only 115 * handle one control operation at a time. The SOLOCKED flag is used to 116 * single-thread operations from sockfs users to prevent e.g. multiple bind() 117 * calls to operate on the same sonode concurrently. The SOREADLOCKED flag is 118 * used to ensure that only one thread sleeps in kstrgetmsg for a given 119 * sonode. This is needed to ensure atomic operation for things like 120 * MSG_WAITALL. 121 * 122 * The so_fallback_rwlock is used to ensure that for sockets that can 123 * fall back to TPI, the fallback is not initiated until all pending 124 * operations have completed. 125 * 126 * Note that so_lock is sometimes held across calls that might go to sleep 127 * (kmem_alloc and soallocproto*). This implies that no other lock in 128 * the system should be held when calling into sockfs; from the system call 129 * side or from strrput (in case of TPI based sockets). If locks are held 130 * while calling into sockfs the system might hang when running low on memory. 131 */ 132 struct sonode { 133 struct vnode *so_vnode; /* vnode associated with this sonode */ 134 135 sonodeops_t *so_ops; /* operations vector for this sonode */ 136 void *so_priv; /* sonode private data */ 137 138 krwlock_t so_fallback_rwlock; 139 kmutex_t so_lock; /* protects sonode fields */ 140 141 kcondvar_t so_state_cv; /* synchronize state changes */ 142 kcondvar_t so_single_cv; /* wait due to SOLOCKED */ 143 kcondvar_t so_read_cv; /* wait due to SOREADLOCKED */ 144 145 /* These fields are protected by so_lock */ 146 147 uint_t so_state; /* internal state flags SS_*, below */ 148 uint_t so_mode; /* characteristics on socket. SM_* */ 149 ushort_t so_flag; /* flags, see below */ 150 int so_count; /* count of opened references */ 151 152 sock_connid_t so_proto_connid; /* protocol generation number */ 153 154 ushort_t so_error; /* error affecting connection */ 155 156 struct sockparams *so_sockparams; /* vnode or socket module */ 157 /* Needed to recreate the same socket for accept */ 158 short so_family; 159 short so_type; 160 short so_protocol; 161 short so_version; /* From so_socket call */ 162 163 /* Accept queue */ 164 kmutex_t so_acceptq_lock; /* protects accept queue */ 165 struct sonode *so_acceptq_next; /* acceptq list node */ 166 struct sonode *so_acceptq_head; 167 struct sonode **so_acceptq_tail; 168 unsigned int so_acceptq_len; 169 unsigned int so_backlog; /* Listen backlog */ 170 kcondvar_t so_acceptq_cv; /* wait for new conn. */ 171 172 /* Options */ 173 short so_options; /* From socket call, see socket.h */ 174 struct linger so_linger; /* SO_LINGER value */ 175 #define so_sndbuf so_proto_props.sopp_txhiwat /* SO_SNDBUF value */ 176 #define so_sndlowat so_proto_props.sopp_txlowat /* tx low water mark */ 177 #define so_rcvbuf so_proto_props.sopp_rxhiwat /* SO_RCVBUF value */ 178 #define so_rcvlowat so_proto_props.sopp_rxlowat /* rx low water mark */ 179 #define so_max_addr_len so_proto_props.sopp_maxaddrlen 180 #define so_minpsz so_proto_props.sopp_minpsz 181 #define so_maxpsz so_proto_props.sopp_maxpsz 182 183 int so_xpg_rcvbuf; /* SO_RCVBUF value for XPG4 socket */ 184 clock_t so_sndtimeo; /* send timeout */ 185 clock_t so_rcvtimeo; /* recv timeout */ 186 187 mblk_t *so_oobmsg; /* outofline oob data */ 188 ssize_t so_oobmark; /* offset of the oob data */ 189 190 pid_t so_pgrp; /* pgrp for signals */ 191 192 cred_t *so_peercred; /* connected socket peer cred */ 193 pid_t so_cpid; /* connected socket peer cached pid */ 194 zoneid_t so_zoneid; /* opener's zoneid */ 195 196 struct pollhead so_poll_list; /* common pollhead */ 197 short so_pollev; /* events that should be generated */ 198 199 /* Receive */ 200 unsigned int so_rcv_queued; /* # bytes on both rcv lists */ 201 mblk_t *so_rcv_q_head; /* processing/copyout rcv queue */ 202 mblk_t *so_rcv_q_last_head; 203 mblk_t *so_rcv_head; /* protocol prequeue */ 204 mblk_t *so_rcv_last_head; /* last mblk in b_next chain */ 205 kcondvar_t so_rcv_cv; /* wait for data */ 206 uint_t so_rcv_wanted; /* # of bytes wanted by app */ 207 timeout_id_t so_rcv_timer_tid; 208 209 #define so_rcv_thresh so_proto_props.sopp_rcvthresh 210 #define so_rcv_timer_interval so_proto_props.sopp_rcvtimer 211 212 kcondvar_t so_snd_cv; /* wait for snd buffers */ 213 uint32_t 214 so_snd_qfull: 1, /* Transmit full */ 215 so_rcv_wakeup: 1, 216 so_snd_wakeup: 1, 217 so_not_str: 1, /* B_TRUE if not streams based socket */ 218 so_pad_to_bit_31: 28; 219 220 /* Communication channel with protocol */ 221 sock_lower_handle_t so_proto_handle; 222 sock_downcalls_t *so_downcalls; 223 224 struct sock_proto_props so_proto_props; /* protocol settings */ 225 boolean_t so_flowctrld; /* Flow controlled */ 226 uint_t so_copyflag; /* Copy related flag */ 227 kcondvar_t so_copy_cv; /* Copy cond variable */ 228 229 /* kernel sockets */ 230 ksocket_callbacks_t so_ksock_callbacks; 231 void *so_ksock_cb_arg; /* callback argument */ 232 kcondvar_t so_closing_cv; 233 234 /* != NULL for sodirect enabled socket */ 235 struct sodirect_s *so_direct; 236 }; 237 238 #define SO_HAVE_DATA(so) \ 239 /* \ 240 * For the (tid == 0) case we must check so_rcv_{q_,}head \ 241 * rather than (so_rcv_queued > 0), since the latter does not \ 242 * take into account mblks with only control/name information. \ 243 */ \ 244 ((so)->so_rcv_timer_tid == 0 && ((so)->so_rcv_head != NULL || \ 245 (so)->so_rcv_q_head != NULL)) || \ 246 ((so)->so_state & SS_CANTRCVMORE) 247 248 /* 249 * Events handled by the protocol (in case sd_poll is set) 250 */ 251 #define SO_PROTO_POLLEV (POLLIN|POLLRDNORM|POLLRDBAND) 252 253 254 #endif /* _KERNEL || _KMEMUSER */ 255 256 /* flags */ 257 #define SOMOD 0x0001 /* update socket modification time */ 258 #define SOACC 0x0002 /* update socket access time */ 259 260 #define SOLOCKED 0x0010 /* use to serialize open/closes */ 261 #define SOREADLOCKED 0x0020 /* serialize kstrgetmsg calls */ 262 #define SOCLONE 0x0040 /* child of clone driver */ 263 #define SOASYNC_UNBIND 0x0080 /* wait for ACK of async unbind */ 264 265 #define SOCK_IS_NONSTR(so) ((so)->so_not_str) 266 267 /* 268 * Socket state bits. 269 */ 270 #define SS_ISCONNECTED 0x00000001 /* socket connected to a peer */ 271 #define SS_ISCONNECTING 0x00000002 /* in process, connecting to peer */ 272 #define SS_ISDISCONNECTING 0x00000004 /* in process of disconnecting */ 273 #define SS_CANTSENDMORE 0x00000008 /* can't send more data to peer */ 274 275 #define SS_CANTRCVMORE 0x00000010 /* can't receive more data */ 276 #define SS_ISBOUND 0x00000020 /* socket is bound */ 277 #define SS_NDELAY 0x00000040 /* FNDELAY non-blocking */ 278 #define SS_NONBLOCK 0x00000080 /* O_NONBLOCK non-blocking */ 279 280 #define SS_ASYNC 0x00000100 /* async i/o notify */ 281 #define SS_ACCEPTCONN 0x00000200 /* listen done */ 282 /* unused 0x00000400 */ /* was SS_HASCONNIND */ 283 #define SS_SAVEDEOR 0x00000800 /* Saved MSG_EOR rcv side state */ 284 285 #define SS_RCVATMARK 0x00001000 /* at mark on input */ 286 #define SS_OOBPEND 0x00002000 /* OOB pending or present - poll */ 287 #define SS_HAVEOOBDATA 0x00004000 /* OOB data present */ 288 #define SS_HADOOBDATA 0x00008000 /* OOB data consumed */ 289 #define SS_CLOSING 0x00010000 /* in process of closing */ 290 291 /* unused 0x00020000 */ /* was SS_FADDR_NOXLATE */ 292 /* unused 0x00040000 */ /* was SS_HASDATA */ 293 /* unused 0x00080000 */ /* was SS_DONEREAD */ 294 /* unused 0x00100000 */ /* was SS_MOREDATA */ 295 /* unused 0x00200000 */ /* was SS_DIRECT */ 296 297 #define SS_SODIRECT 0x00400000 /* transport supports sodirect */ 298 299 #define SS_SENTLASTREADSIG 0x01000000 /* last rx signal has been sent */ 300 #define SS_SENTLASTWRITESIG 0x02000000 /* last tx signal has been sent */ 301 302 #define SS_FALLBACK_DRAIN 0x20000000 /* data was/is being drained */ 303 #define SS_FALLBACK_PENDING 0x40000000 /* fallback is pending */ 304 #define SS_FALLBACK_COMP 0x80000000 /* fallback has completed */ 305 306 307 /* Set of states when the socket can't be rebound */ 308 #define SS_CANTREBIND (SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING|\ 309 SS_CANTSENDMORE|SS_CANTRCVMORE|SS_ACCEPTCONN) 310 311 /* 312 * Sockets that can fall back to TPI must ensure that fall back is not 313 * initiated while a thread is using a socket. 314 */ 315 #define SO_BLOCK_FALLBACK(so, fn) { \ 316 ASSERT(MUTEX_NOT_HELD(&(so)->so_lock)); \ 317 rw_enter(&(so)->so_fallback_rwlock, RW_READER); \ 318 if ((so)->so_state & SS_FALLBACK_COMP) { \ 319 rw_exit(&(so)->so_fallback_rwlock); \ 320 return (fn); \ 321 } \ 322 } 323 324 #define SO_UNBLOCK_FALLBACK(so) { \ 325 rw_exit(&(so)->so_fallback_rwlock); \ 326 } 327 328 /* Poll events */ 329 #define SO_POLLEV_IN 0x1 /* POLLIN wakeup needed */ 330 #define SO_POLLEV_ALWAYS 0x2 /* wakeups */ 331 332 /* 333 * Characteristics of sockets. Not changed after the socket is created. 334 */ 335 #define SM_PRIV 0x001 /* privileged for broadcast, raw... */ 336 #define SM_ATOMIC 0x002 /* atomic data transmission */ 337 #define SM_ADDR 0x004 /* addresses given with messages */ 338 #define SM_CONNREQUIRED 0x008 /* connection required by protocol */ 339 340 #define SM_FDPASSING 0x010 /* passes file descriptors */ 341 #define SM_EXDATA 0x020 /* Can handle T_EXDATA_REQ */ 342 #define SM_OPTDATA 0x040 /* Can handle T_OPTDATA_REQ */ 343 #define SM_BYTESTREAM 0x080 /* Byte stream - can use M_DATA */ 344 345 #define SM_ACCEPTOR_ID 0x100 /* so_acceptor_id is valid */ 346 347 #define SM_KERNEL 0x200 /* kernel socket */ 348 349 /* The modes below are only for non-streams sockets */ 350 #define SM_ACCEPTSUPP 0x400 /* can handle accept() */ 351 #define SM_SENDFILESUPP 0x800 /* Private: proto supp sendfile */ 352 353 /* 354 * Socket versions. Used by the socket library when calling _so_socket(). 355 */ 356 #define SOV_STREAM 0 /* Not a socket - just a stream */ 357 #define SOV_DEFAULT 1 /* Select based on so_default_version */ 358 #define SOV_SOCKSTREAM 2 /* Socket plus streams operations */ 359 #define SOV_SOCKBSD 3 /* Socket with no streams operations */ 360 #define SOV_XPG4_2 4 /* Xnet socket */ 361 362 #if defined(_KERNEL) || defined(_KMEMUSER) 363 364 /* 365 * sonode create and destroy functions. 366 */ 367 typedef struct sonode *(*so_create_func_t)(struct sockparams *, 368 int, int, int, int, int, int *, cred_t *); 369 typedef void (*so_destroy_func_t)(struct sonode *); 370 371 /* STREAM device information */ 372 typedef struct sdev_info { 373 char *sd_devpath; 374 int sd_devpathlen; /* Is 0 if sp_devpath is a static string */ 375 vnode_t *sd_vnode; 376 } sdev_info_t; 377 378 #define SOCKMOD_VERSION 1 379 /* name of the TPI pseudo socket module */ 380 #define SOTPI_SMOD_NAME "socktpi" 381 382 typedef struct __smod_priv_s { 383 so_create_func_t smodp_sock_create_func; 384 so_destroy_func_t smodp_sock_destroy_func; 385 so_proto_fallback_func_t smodp_proto_fallback_func; 386 } __smod_priv_t; 387 388 /* 389 * Socket module register information 390 */ 391 typedef struct smod_reg_s { 392 int smod_version; 393 char *smod_name; 394 size_t smod_uc_version; 395 size_t smod_dc_version; 396 so_proto_create_func_t smod_proto_create_func; 397 398 /* __smod_priv_data must be NULL */ 399 __smod_priv_t *__smod_priv; 400 } smod_reg_t; 401 402 /* 403 * Socket module information 404 */ 405 typedef struct smod_info { 406 int smod_version; 407 char *smod_name; 408 uint_t smod_refcnt; /* # of entries */ 409 size_t smod_uc_version; /* upcall version */ 410 size_t smod_dc_version; /* down call version */ 411 so_proto_create_func_t smod_proto_create_func; 412 so_proto_fallback_func_t smod_proto_fallback_func; 413 so_create_func_t smod_sock_create_func; 414 so_destroy_func_t smod_sock_destroy_func; 415 list_node_t smod_node; 416 } smod_info_t; 417 418 typedef struct sockparams_stats { 419 kstat_named_t sps_nfallback; /* # of fallbacks to TPI */ 420 kstat_named_t sps_nactive; /* # of active sockets */ 421 kstat_named_t sps_ncreate; /* total # of created sockets */ 422 } sockparams_stats_t; 423 424 /* 425 * sockparams 426 * 427 * Used for mapping family/type/protocol to a socket module or STREAMS device 428 */ 429 struct sockparams { 430 /* 431 * The family, type, protocol, sdev_info and smod_name are 432 * set when the entry is created, and they will never change 433 * thereafter. 434 */ 435 int sp_family; 436 int sp_type; 437 int sp_protocol; 438 439 sdev_info_t sp_sdev_info; /* STREAM device */ 440 char *sp_smod_name; /* socket module name */ 441 442 kmutex_t sp_lock; /* lock for refcnt and smod_info */ 443 uint64_t sp_refcnt; /* entry reference count */ 444 smod_info_t *sp_smod_info; /* socket module */ 445 446 sockparams_stats_t sp_stats; 447 kstat_t *sp_kstat; 448 449 /* 450 * The entries below are only modified while holding 451 * splist_lock as a writer. 452 */ 453 int sp_flags; /* see below */ 454 list_node_t sp_node; 455 }; 456 457 458 /* 459 * sockparams flags 460 */ 461 #define SOCKPARAMS_EPHEMERAL 0x1 /* temp. entry, not on global list */ 462 463 extern void sockparams_init(void); 464 extern struct sockparams *sockparams_hold_ephemeral_bydev(int, int, int, 465 const char *, int, int *); 466 extern struct sockparams *sockparams_hold_ephemeral_bymod(int, int, int, 467 const char *, int, int *); 468 extern void sockparams_ephemeral_drop_last_ref(struct sockparams *); 469 470 extern void smod_init(void); 471 extern void smod_add(smod_info_t *); 472 extern int smod_register(const smod_reg_t *); 473 extern int smod_unregister(const char *); 474 extern smod_info_t *smod_lookup_byname(const char *); 475 476 #define SOCKPARAMS_HAS_DEVICE(sp) \ 477 ((sp)->sp_sdev_info.sd_devpath != NULL) 478 479 /* Increase the smod_info_t reference count */ 480 #define SMOD_INC_REF(smodp) { \ 481 ASSERT((smodp) != NULL); \ 482 DTRACE_PROBE1(smodinfo__inc__ref, struct smod_info *, (smodp)); \ 483 atomic_inc_uint(&(smodp)->smod_refcnt); \ 484 } 485 486 /* 487 * Decreace the socket module entry reference count. 488 * When no one mapping to the entry, we try to unload the module from the 489 * kernel. If the module can't unload, just leave the module entry with 490 * a zero refcnt. 491 */ 492 #define SMOD_DEC_REF(smodp, modname) { \ 493 ASSERT((smodp) != NULL); \ 494 ASSERT((smodp)->smod_refcnt != 0); \ 495 atomic_dec_uint(&(smodp)->smod_refcnt); \ 496 /* \ 497 * No need to atomically check the return value because the \ 498 * socket module framework will verify that no one is using \ 499 * the module before unloading. Worst thing that can happen \ 500 * here is multiple calls to mod_remove_by_name(), which is OK. \ 501 */ \ 502 if ((smodp)->smod_refcnt == 0) \ 503 (void) mod_remove_by_name(modname); \ 504 } 505 506 /* Increase the reference count */ 507 #define SOCKPARAMS_INC_REF(sp) { \ 508 ASSERT((sp) != NULL); \ 509 DTRACE_PROBE1(sockparams__inc__ref, struct sockparams *, (sp)); \ 510 mutex_enter(&(sp)->sp_lock); \ 511 (sp)->sp_refcnt++; \ 512 ASSERT((sp)->sp_refcnt != 0); \ 513 mutex_exit(&(sp)->sp_lock); \ 514 } 515 516 /* 517 * Decrease the reference count. 518 * 519 * If the sockparams is ephemeral, then the thread dropping the last ref 520 * count will destroy the entry. 521 */ 522 #define SOCKPARAMS_DEC_REF(sp) { \ 523 ASSERT((sp) != NULL); \ 524 DTRACE_PROBE1(sockparams__dec__ref, struct sockparams *, (sp)); \ 525 mutex_enter(&(sp)->sp_lock); \ 526 ASSERT((sp)->sp_refcnt > 0); \ 527 if ((sp)->sp_refcnt == 1) { \ 528 if ((sp)->sp_flags & SOCKPARAMS_EPHEMERAL) { \ 529 mutex_exit(&(sp)->sp_lock); \ 530 sockparams_ephemeral_drop_last_ref((sp)); \ 531 } else { \ 532 (sp)->sp_refcnt--; \ 533 if ((sp)->sp_smod_info != NULL) { \ 534 SMOD_DEC_REF((sp)->sp_smod_info, \ 535 (sp)->sp_smod_name); \ 536 } \ 537 (sp)->sp_smod_info = NULL; \ 538 mutex_exit(&(sp)->sp_lock); \ 539 } \ 540 } else { \ 541 (sp)->sp_refcnt--; \ 542 mutex_exit(&(sp)->sp_lock); \ 543 } \ 544 } 545 546 /* 547 * Used to traverse the list of AF_UNIX sockets to construct the kstat 548 * for netstat(1m). 549 */ 550 struct socklist { 551 kmutex_t sl_lock; 552 struct sonode *sl_list; 553 }; 554 555 extern struct socklist socklist; 556 /* 557 * ss_full_waits is the number of times the reader thread 558 * waits when the queue is full and ss_empty_waits is the number 559 * of times the consumer thread waits when the queue is empty. 560 * No locks for these as they are just indicators of whether 561 * disk or network or both is slow or fast. 562 */ 563 struct sendfile_stats { 564 uint32_t ss_file_cached; 565 uint32_t ss_file_not_cached; 566 uint32_t ss_full_waits; 567 uint32_t ss_empty_waits; 568 uint32_t ss_file_segmap; 569 }; 570 571 /* 572 * A single sendfile request is represented by snf_req. 573 */ 574 typedef struct snf_req { 575 struct snf_req *sr_next; 576 mblk_t *sr_mp_head; 577 mblk_t *sr_mp_tail; 578 kmutex_t sr_lock; 579 kcondvar_t sr_cv; 580 uint_t sr_qlen; 581 int sr_hiwat; 582 int sr_lowat; 583 int sr_operation; 584 struct vnode *sr_vp; 585 file_t *sr_fp; 586 ssize_t sr_maxpsz; 587 u_offset_t sr_file_off; 588 u_offset_t sr_file_size; 589 #define SR_READ_DONE 0x80000000 590 int sr_read_error; 591 int sr_write_error; 592 } snf_req_t; 593 594 /* A queue of sendfile requests */ 595 struct sendfile_queue { 596 snf_req_t *snfq_req_head; 597 snf_req_t *snfq_req_tail; 598 kmutex_t snfq_lock; 599 kcondvar_t snfq_cv; 600 int snfq_svc_threads; /* # of service threads */ 601 int snfq_idle_cnt; /* # of idling threads */ 602 int snfq_max_threads; 603 int snfq_req_cnt; /* Number of requests */ 604 }; 605 606 #define READ_OP 1 607 #define SNFQ_TIMEOUT (60 * 5 * hz) /* 5 minutes */ 608 609 /* Socket network operations switch */ 610 struct sonodeops { 611 int (*sop_init)(struct sonode *, struct sonode *, cred_t *, 612 int); 613 int (*sop_accept)(struct sonode *, int, cred_t *, struct sonode **); 614 int (*sop_bind)(struct sonode *, struct sockaddr *, socklen_t, 615 int, cred_t *); 616 int (*sop_listen)(struct sonode *, int, cred_t *); 617 int (*sop_connect)(struct sonode *, const struct sockaddr *, 618 socklen_t, int, int, cred_t *); 619 int (*sop_recvmsg)(struct sonode *, struct msghdr *, 620 struct uio *, cred_t *); 621 int (*sop_sendmsg)(struct sonode *, struct msghdr *, 622 struct uio *, cred_t *); 623 int (*sop_sendmblk)(struct sonode *, struct msghdr *, int, 624 cred_t *, mblk_t **); 625 int (*sop_getpeername)(struct sonode *, struct sockaddr *, 626 socklen_t *, boolean_t, cred_t *); 627 int (*sop_getsockname)(struct sonode *, struct sockaddr *, 628 socklen_t *, cred_t *); 629 int (*sop_shutdown)(struct sonode *, int, cred_t *); 630 int (*sop_getsockopt)(struct sonode *, int, int, void *, 631 socklen_t *, int, cred_t *); 632 int (*sop_setsockopt)(struct sonode *, int, int, const void *, 633 socklen_t, cred_t *); 634 int (*sop_ioctl)(struct sonode *, int, intptr_t, int, 635 cred_t *, int32_t *); 636 int (*sop_poll)(struct sonode *, short, int, short *, 637 struct pollhead **); 638 int (*sop_close)(struct sonode *, int, cred_t *); 639 }; 640 641 #define SOP_INIT(so, flag, cr, flags) \ 642 ((so)->so_ops->sop_init((so), (flag), (cr), (flags))) 643 #define SOP_ACCEPT(so, fflag, cr, nsop) \ 644 ((so)->so_ops->sop_accept((so), (fflag), (cr), (nsop))) 645 #define SOP_BIND(so, name, namelen, flags, cr) \ 646 ((so)->so_ops->sop_bind((so), (name), (namelen), (flags), (cr))) 647 #define SOP_LISTEN(so, backlog, cr) \ 648 ((so)->so_ops->sop_listen((so), (backlog), (cr))) 649 #define SOP_CONNECT(so, name, namelen, fflag, flags, cr) \ 650 ((so)->so_ops->sop_connect((so), (name), (namelen), (fflag), (flags), \ 651 (cr))) 652 #define SOP_RECVMSG(so, msg, uiop, cr) \ 653 ((so)->so_ops->sop_recvmsg((so), (msg), (uiop), (cr))) 654 #define SOP_SENDMSG(so, msg, uiop, cr) \ 655 ((so)->so_ops->sop_sendmsg((so), (msg), (uiop), (cr))) 656 #define SOP_SENDMBLK(so, msg, size, cr, mpp) \ 657 ((so)->so_ops->sop_sendmblk((so), (msg), (size), (cr), (mpp))) 658 #define SOP_GETPEERNAME(so, addr, addrlen, accept, cr) \ 659 ((so)->so_ops->sop_getpeername((so), (addr), (addrlen), (accept), (cr))) 660 #define SOP_GETSOCKNAME(so, addr, addrlen, cr) \ 661 ((so)->so_ops->sop_getsockname((so), (addr), (addrlen), (cr))) 662 #define SOP_SHUTDOWN(so, how, cr) \ 663 ((so)->so_ops->sop_shutdown((so), (how), (cr))) 664 #define SOP_GETSOCKOPT(so, level, optionname, optval, optlenp, flags, cr) \ 665 ((so)->so_ops->sop_getsockopt((so), (level), (optionname), \ 666 (optval), (optlenp), (flags), (cr))) 667 #define SOP_SETSOCKOPT(so, level, optionname, optval, optlen, cr) \ 668 ((so)->so_ops->sop_setsockopt((so), (level), (optionname), \ 669 (optval), (optlen), (cr))) 670 #define SOP_IOCTL(so, cmd, arg, mode, cr, rvalp) \ 671 ((so)->so_ops->sop_ioctl((so), (cmd), (arg), (mode), (cr), (rvalp))) 672 #define SOP_POLL(so, events, anyyet, reventsp, phpp) \ 673 ((so)->so_ops->sop_poll((so), (events), (anyyet), (reventsp), (phpp))) 674 #define SOP_CLOSE(so, flag, cr) \ 675 ((so)->so_ops->sop_close((so), (flag), (cr))) 676 677 #endif /* defined(_KERNEL) || defined(_KMEMUSER) */ 678 679 #ifdef _KERNEL 680 681 #define ISALIGNED_cmsghdr(addr) \ 682 (((uintptr_t)(addr) & (_CMSG_HDR_ALIGNMENT - 1)) == 0) 683 684 #define ROUNDUP_cmsglen(len) \ 685 (((len) + _CMSG_HDR_ALIGNMENT - 1) & ~(_CMSG_HDR_ALIGNMENT - 1)) 686 687 #define IS_NON_STREAM_SOCK(vp) \ 688 ((vp)->v_type == VSOCK && (vp)->v_stream == NULL) 689 /* 690 * Macros that operate on struct cmsghdr. 691 * Used in parsing msg_control. 692 * The CMSG_VALID macro does not assume that the last option buffer is padded. 693 */ 694 #define CMSG_NEXT(cmsg) \ 695 (struct cmsghdr *)((uintptr_t)(cmsg) + \ 696 ROUNDUP_cmsglen((cmsg)->cmsg_len)) 697 #define CMSG_CONTENT(cmsg) (&((cmsg)[1])) 698 #define CMSG_CONTENTLEN(cmsg) ((cmsg)->cmsg_len - sizeof (struct cmsghdr)) 699 #define CMSG_VALID(cmsg, start, end) \ 700 (ISALIGNED_cmsghdr(cmsg) && \ 701 ((uintptr_t)(cmsg) >= (uintptr_t)(start)) && \ 702 ((uintptr_t)(cmsg) < (uintptr_t)(end)) && \ 703 ((ssize_t)(cmsg)->cmsg_len >= sizeof (struct cmsghdr)) && \ 704 ((uintptr_t)(cmsg) + (cmsg)->cmsg_len <= (uintptr_t)(end))) 705 706 /* 707 * Maximum size of any argument that is copied in (addresses, options, 708 * access rights). MUST be at least MAXPATHLEN + 3. 709 * BSD and SunOS 4.X limited this to MLEN or MCLBYTES. 710 */ 711 #define SO_MAXARGSIZE 8192 712 713 /* 714 * Convert between vnode and sonode 715 */ 716 #define VTOSO(vp) ((struct sonode *)((vp)->v_data)) 717 #define SOTOV(sp) ((sp)->so_vnode) 718 719 /* 720 * Internal flags for sobind() 721 */ 722 #define _SOBIND_REBIND 0x01 /* Bind to existing local address */ 723 #define _SOBIND_UNSPEC 0x02 /* Bind to unspecified address */ 724 #define _SOBIND_LOCK_HELD 0x04 /* so_excl_lock held by caller */ 725 #define _SOBIND_NOXLATE 0x08 /* No addr translation for AF_UNIX */ 726 #define _SOBIND_XPG4_2 0x10 /* xpg4.2 semantics */ 727 #define _SOBIND_SOCKBSD 0x20 /* BSD semantics */ 728 #define _SOBIND_LISTEN 0x40 /* Make into SS_ACCEPTCONN */ 729 #define _SOBIND_SOCKETPAIR 0x80 /* Internal flag for so_socketpair() */ 730 /* to enable listen with backlog = 1 */ 731 732 /* 733 * Internal flags for sounbind() 734 */ 735 #define _SOUNBIND_REBIND 0x01 /* Don't clear fields - will rebind */ 736 737 /* 738 * Internal flags for soconnect() 739 */ 740 #define _SOCONNECT_NOXLATE 0x01 /* No addr translation for AF_UNIX */ 741 #define _SOCONNECT_DID_BIND 0x02 /* Unbind when connect fails */ 742 #define _SOCONNECT_XPG4_2 0x04 /* xpg4.2 semantics */ 743 744 /* 745 * Internal flags for sodisconnect() 746 */ 747 #define _SODISCONNECT_LOCK_HELD 0x01 /* so_excl_lock held by caller */ 748 749 /* 750 * Internal flags for sotpi_getsockopt(). 751 */ 752 #define _SOGETSOCKOPT_XPG4_2 0x01 /* xpg4.2 semantics */ 753 754 /* 755 * Internal flags for soallocproto*() 756 */ 757 #define _ALLOC_NOSLEEP 0 /* Don't sleep for memory */ 758 #define _ALLOC_INTR 1 /* Sleep until interrupt */ 759 #define _ALLOC_SLEEP 2 /* Sleep forever */ 760 761 /* 762 * Internal structure for handling AF_UNIX file descriptor passing 763 */ 764 struct fdbuf { 765 int fd_size; /* In bytes, for kmem_free */ 766 int fd_numfd; /* Number of elements below */ 767 char *fd_ebuf; /* Extra buffer to free */ 768 int fd_ebuflen; 769 frtn_t fd_frtn; 770 struct file *fd_fds[1]; /* One or more */ 771 }; 772 #define FDBUF_HDRSIZE (sizeof (struct fdbuf) - sizeof (struct file *)) 773 774 /* 775 * Variable that can be patched to set what version of socket socket() 776 * will create. 777 */ 778 extern int so_default_version; 779 780 #ifdef DEBUG 781 /* Turn on extra testing capabilities */ 782 #define SOCK_TEST 783 #endif /* DEBUG */ 784 785 #ifdef DEBUG 786 char *pr_state(uint_t, uint_t); 787 char *pr_addr(int, struct sockaddr *, t_uscalar_t); 788 int so_verify_oobstate(struct sonode *); 789 #endif /* DEBUG */ 790 791 /* 792 * DEBUG macros 793 */ 794 #if defined(DEBUG) 795 #define SOCK_DEBUG 796 797 extern int sockdebug; 798 extern int sockprinterr; 799 800 #define eprint(args) printf args 801 #define eprintso(so, args) \ 802 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) printf args; } 803 #define eprintline(error) \ 804 { \ 805 if (error != EINTR && (sockprinterr || sockdebug > 0)) \ 806 printf("socket error %d: line %d file %s\n", \ 807 (error), __LINE__, __FILE__); \ 808 } 809 810 #define eprintsoline(so, error) \ 811 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) \ 812 printf("socket(%p) error %d: line %d file %s\n", \ 813 (void *)(so), (error), __LINE__, __FILE__); \ 814 } 815 #define dprint(level, args) { if (sockdebug > (level)) printf args; } 816 #define dprintso(so, level, args) \ 817 { if (sockdebug > (level) && ((so)->so_options & SO_DEBUG)) printf args; } 818 819 #else /* define(DEBUG) */ 820 821 #define eprint(args) {} 822 #define eprintso(so, args) {} 823 #define eprintline(error) {} 824 #define eprintsoline(so, error) {} 825 #define dprint(level, args) {} 826 #define dprintso(so, level, args) {} 827 828 #endif /* defined(DEBUG) */ 829 830 extern struct vfsops sock_vfsops; 831 extern struct vnodeops *socket_vnodeops; 832 extern const struct fs_operation_def socket_vnodeops_template[]; 833 834 extern dev_t sockdev; 835 836 /* 837 * sockfs functions 838 */ 839 extern int sock_getmsg(vnode_t *, struct strbuf *, struct strbuf *, 840 uchar_t *, int *, int, rval_t *); 841 extern int sock_putmsg(vnode_t *, struct strbuf *, struct strbuf *, 842 uchar_t, int, int); 843 extern int sogetvp(char *, vnode_t **, int); 844 extern int sockinit(int, char *); 845 extern int soconfig(int, int, int, char *, int, char *); 846 extern int solookup(int, int, int, struct sockparams **); 847 extern void so_lock_single(struct sonode *); 848 extern void so_unlock_single(struct sonode *, int); 849 extern int so_lock_read(struct sonode *, int); 850 extern int so_lock_read_intr(struct sonode *, int); 851 extern void so_unlock_read(struct sonode *); 852 extern void *sogetoff(mblk_t *, t_uscalar_t, t_uscalar_t, uint_t); 853 extern void so_getopt_srcaddr(void *, t_uscalar_t, 854 void **, t_uscalar_t *); 855 extern int so_getopt_unix_close(void *, t_uscalar_t); 856 extern void fdbuf_free(struct fdbuf *); 857 extern mblk_t *fdbuf_allocmsg(int, struct fdbuf *); 858 extern int fdbuf_create(void *, int, struct fdbuf **); 859 extern void so_closefds(void *, t_uscalar_t, int, int); 860 extern int so_getfdopt(void *, t_uscalar_t, int, void **, int *); 861 t_uscalar_t so_optlen(void *, t_uscalar_t, int); 862 extern void so_cmsg2opt(void *, t_uscalar_t, int, mblk_t *); 863 extern t_uscalar_t 864 so_cmsglen(mblk_t *, void *, t_uscalar_t, int); 865 extern int so_opt2cmsg(mblk_t *, void *, t_uscalar_t, int, 866 void *, t_uscalar_t); 867 extern void soisconnecting(struct sonode *); 868 extern void soisconnected(struct sonode *); 869 extern void soisdisconnected(struct sonode *, int); 870 extern void socantsendmore(struct sonode *); 871 extern void socantrcvmore(struct sonode *); 872 extern void soseterror(struct sonode *, int); 873 extern int sogeterr(struct sonode *, boolean_t); 874 extern int sowaitconnected(struct sonode *, int, int); 875 876 extern ssize_t soreadfile(file_t *, uchar_t *, u_offset_t, int *, size_t); 877 extern void *sock_kstat_init(zoneid_t); 878 extern void sock_kstat_fini(zoneid_t, void *); 879 extern struct sonode *getsonode(int, int *, file_t **); 880 /* 881 * Function wrappers (mostly around the sonode switch) for 882 * backward compatibility. 883 */ 884 extern int soaccept(struct sonode *, int, struct sonode **); 885 extern int sobind(struct sonode *, struct sockaddr *, socklen_t, 886 int, int); 887 extern int solisten(struct sonode *, int); 888 extern int soconnect(struct sonode *, const struct sockaddr *, socklen_t, 889 int, int); 890 extern int sorecvmsg(struct sonode *, struct nmsghdr *, struct uio *); 891 extern int sosendmsg(struct sonode *, struct nmsghdr *, struct uio *); 892 extern int soshutdown(struct sonode *, int); 893 extern int sogetsockopt(struct sonode *, int, int, void *, socklen_t *, 894 int); 895 extern int sosetsockopt(struct sonode *, int, int, const void *, 896 t_uscalar_t); 897 898 extern struct sonode *socreate(struct sockparams *, int, int, int, int, 899 int *); 900 901 extern int so_copyin(const void *, void *, size_t, int); 902 extern int so_copyout(const void *, void *, size_t, int); 903 904 #endif 905 906 /* 907 * Internal structure for obtaining sonode information from the socklist. 908 * These types match those corresponding in the sonode structure. 909 * This is not a published interface, and may change at any time. 910 */ 911 struct sockinfo { 912 uint_t si_size; /* real length of this struct */ 913 short si_family; 914 short si_type; 915 ushort_t si_flag; 916 uint_t si_state; 917 uint_t si_ux_laddr_sou_magic; 918 uint_t si_ux_faddr_sou_magic; 919 t_scalar_t si_serv_type; 920 t_uscalar_t si_laddr_soa_len; 921 t_uscalar_t si_faddr_soa_len; 922 uint16_t si_laddr_family; 923 uint16_t si_faddr_family; 924 char si_laddr_sun_path[MAXPATHLEN + 1]; /* NULL terminated */ 925 char si_faddr_sun_path[MAXPATHLEN + 1]; 926 boolean_t si_faddr_noxlate; 927 zoneid_t si_szoneid; 928 }; 929 930 #define SOCKMOD_PATH "socketmod" /* dir where sockmods are stored */ 931 932 #ifdef __cplusplus 933 } 934 #endif 935 936 #endif /* _SYS_SOCKETVAR_H */ 937