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 2008 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 clock_t so_sndtimeo; /* send timeout */ 178 clock_t so_rcvtimeo; /* recv timeout */ 179 180 mblk_t *so_oobmsg; /* outofline oob data */ 181 ssize_t so_oobmark; /* offset of the oob data */ 182 183 pid_t so_pgrp; /* pgrp for signals */ 184 185 cred_t *so_peercred; /* connected socket peer cred */ 186 pid_t so_cpid; /* connected socket peer cached pid */ 187 zoneid_t so_zoneid; /* opener's zoneid */ 188 189 struct pollhead so_poll_list; /* common pollhead */ 190 short so_pollev; /* events that should be generated */ 191 192 /* Receive */ 193 unsigned int so_rcv_queued; 194 mblk_t *so_rcv_q_head; 195 mblk_t *so_rcv_q_last_head; 196 mblk_t *so_rcv_head; /* 1st mblk in the list */ 197 mblk_t *so_rcv_last_head; /* last mblk in b_next chain */ 198 kcondvar_t so_rcv_cv; 199 uint_t so_rcv_wanted; /* # of bytes wanted by app */ 200 timeout_id_t so_rcv_timer_tid; 201 202 #define so_rcv_thresh so_proto_props.sopp_rcvthresh 203 #define so_rcv_timer_interval so_proto_props.sopp_rcvtimer 204 205 kcondvar_t so_snd_cv; 206 uint32_t 207 so_snd_qfull: 1, /* Transmit full */ 208 so_rcv_wakeup: 1, 209 so_snd_wakeup: 1, 210 so_not_str: 1, /* B_TRUE if not streams based socket */ 211 so_pad_to_bit_31: 28; 212 213 /* Communication channel with protocol */ 214 sock_lower_handle_t so_proto_handle; 215 sock_downcalls_t *so_downcalls; 216 217 struct sock_proto_props so_proto_props; /* protocol settings */ 218 boolean_t so_flowctrld; /* Flow controlled */ 219 uint_t so_copyflag; /* Copy related flag */ 220 kcondvar_t so_copy_cv; /* Copy cond variable */ 221 222 /* kernel sockets */ 223 ksocket_callbacks_t so_ksock_callbacks; 224 void *so_ksock_cb_arg; /* callback argument */ 225 kcondvar_t so_closing_cv; 226 227 /* != NULL for sodirect_t enabled socket */ 228 sodirect_t *so_direct; 229 }; 230 231 /* 232 * We do an initial check for events without holding locks. However, 233 * if there are no event available, then we redo the check for POLLIN 234 * events under the lock. 235 */ 236 #define SO_HAVE_DATA(so) \ 237 ((so)->so_rcv_timer_tid == 0 && (so->so_rcv_queued > 0)) || \ 238 ((so)->so_rcv_queued > (so)->so_rcv_thresh) || \ 239 ((so)->so_state & SS_CANTRCVMORE) 240 241 /* 242 * Events handled by the protocol (in case sd_poll is set) 243 */ 244 #define SO_PROTO_POLLEV (POLLIN|POLLRDNORM|POLLRDBAND) 245 246 247 #endif /* _KERNEL || _KMEMUSER */ 248 249 /* flags */ 250 #define SOMOD 0x0001 /* update socket modification time */ 251 #define SOACC 0x0002 /* update socket access time */ 252 253 #define SOLOCKED 0x0010 /* use to serialize open/closes */ 254 #define SOREADLOCKED 0x0020 /* serialize kstrgetmsg calls */ 255 #define SOWANT 0x0040 /* some process waiting on lock */ 256 #define SOCLONE 0x0080 /* child of clone driver */ 257 #define SOASYNC_UNBIND 0x0100 /* wait for ACK of async unbind */ 258 259 #define SOCK_IS_NONSTR(so) ((so)->so_not_str) 260 261 /* 262 * Socket state bits. 263 */ 264 #define SS_ISCONNECTED 0x00000001 /* socket connected to a peer */ 265 #define SS_ISCONNECTING 0x00000002 /* in process, connecting to peer */ 266 #define SS_ISDISCONNECTING 0x00000004 /* in process of disconnecting */ 267 #define SS_CANTSENDMORE 0x00000008 /* can't send more data to peer */ 268 269 #define SS_CANTRCVMORE 0x00000010 /* can't receive more data */ 270 #define SS_ISBOUND 0x00000020 /* socket is bound */ 271 #define SS_NDELAY 0x00000040 /* FNDELAY non-blocking */ 272 #define SS_NONBLOCK 0x00000080 /* O_NONBLOCK non-blocking */ 273 274 #define SS_ASYNC 0x00000100 /* async i/o notify */ 275 #define SS_ACCEPTCONN 0x00000200 /* listen done */ 276 /* unused 0x00000400 */ /* was SS_HASCONNIND */ 277 #define SS_SAVEDEOR 0x00000800 /* Saved MSG_EOR rcv side state */ 278 279 #define SS_RCVATMARK 0x00001000 /* at mark on input */ 280 #define SS_OOBPEND 0x00002000 /* OOB pending or present - poll */ 281 #define SS_HAVEOOBDATA 0x00004000 /* OOB data present */ 282 #define SS_HADOOBDATA 0x00008000 /* OOB data consumed */ 283 #define SS_CLOSING 0x00010000 /* in process of closing */ 284 285 /* unused 0x00020000 */ /* was SS_FADDR_NOXLATE */ 286 /* unused 0x00040000 */ /* was SS_HASDATA */ 287 /* unused 0x00080000 */ /* was SS_DONEREAD */ 288 /* unused 0x00100000 */ /* was SS_MOREDATA */ 289 /* unused 0x00200000 */ /* was SS_DIRECT */ 290 291 #define SS_SODIRECT 0x00400000 /* transport supports sodirect */ 292 293 /* unused 0x01000000 */ /* was SS_LADDR_VALID */ 294 /* unused 0x02000000 */ /* was SS_FADDR_VALID */ 295 296 #define SS_SENTLASTREADSIG 0x10000000 /* last rx signal has been sent */ 297 #define SS_SENTLASTWRITESIG 0x20000000 /* last tx signal has been sent */ 298 299 #define SS_FALLBACK_PENDING 0x40000000 300 #define SS_FALLBACK_COMP 0x80000000 301 302 303 /* Set of states when the socket can't be rebound */ 304 #define SS_CANTREBIND (SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING|\ 305 SS_CANTSENDMORE|SS_CANTRCVMORE|SS_ACCEPTCONN) 306 307 /* 308 * Sockets that can fall back to TPI must ensure that fall back is not 309 * initiated while a thread is using a socket. 310 */ 311 #define SO_BLOCK_FALLBACK(so, fn) { \ 312 ASSERT(MUTEX_NOT_HELD(&(so)->so_lock)); \ 313 rw_enter(&(so)->so_fallback_rwlock, RW_READER); \ 314 if ((so)->so_state & SS_FALLBACK_COMP) { \ 315 rw_exit(&(so)->so_fallback_rwlock); \ 316 return (fn); \ 317 } \ 318 } 319 320 #define SO_UNBLOCK_FALLBACK(so) { \ 321 rw_exit(&(so)->so_fallback_rwlock); \ 322 } 323 324 /* Poll events */ 325 #define SO_POLLEV_IN 0x1 /* POLLIN wakeup needed */ 326 #define SO_POLLEV_ALWAYS 0x2 /* wakeups */ 327 328 /* 329 * Characteristics of sockets. Not changed after the socket is created. 330 */ 331 #define SM_PRIV 0x001 /* privileged for broadcast, raw... */ 332 #define SM_ATOMIC 0x002 /* atomic data transmission */ 333 #define SM_ADDR 0x004 /* addresses given with messages */ 334 #define SM_CONNREQUIRED 0x008 /* connection required by protocol */ 335 336 #define SM_FDPASSING 0x010 /* passes file descriptors */ 337 #define SM_EXDATA 0x020 /* Can handle T_EXDATA_REQ */ 338 #define SM_OPTDATA 0x040 /* Can handle T_OPTDATA_REQ */ 339 #define SM_BYTESTREAM 0x080 /* Byte stream - can use M_DATA */ 340 341 #define SM_ACCEPTOR_ID 0x100 /* so_acceptor_id is valid */ 342 343 #define SM_KERNEL 0x200 /* kernel socket */ 344 345 /* The modes below are only for non-streams sockets */ 346 #define SM_ACCEPTSUPP 0x400 /* can handle accept() */ 347 #define SM_SENDFILESUPP 0x800 /* Private: proto supp sendfile */ 348 349 /* 350 * Socket versions. Used by the socket library when calling _so_socket(). 351 */ 352 #define SOV_STREAM 0 /* Not a socket - just a stream */ 353 #define SOV_DEFAULT 1 /* Select based on so_default_version */ 354 #define SOV_SOCKSTREAM 2 /* Socket plus streams operations */ 355 #define SOV_SOCKBSD 3 /* Socket with no streams operations */ 356 #define SOV_XPG4_2 4 /* Xnet socket */ 357 358 #if defined(_KERNEL) || defined(_KMEMUSER) 359 360 /* 361 * sonode create and destroy functions. 362 */ 363 typedef struct sonode *(*so_create_func_t)(struct sockparams *, 364 int, int, int, int, int, int *, cred_t *); 365 typedef void (*so_destroy_func_t)(struct sonode *); 366 367 /* STREAM device information */ 368 typedef struct sdev_info { 369 char *sd_devpath; 370 int sd_devpathlen; /* Is 0 if sp_devpath is a static string */ 371 vnode_t *sd_vnode; 372 } sdev_info_t; 373 374 #define SOCKMOD_VERSION 1 375 /* name of the TPI pseudo socket module */ 376 #define SOTPI_SMOD_NAME "socktpi" 377 378 typedef struct __smod_priv_s { 379 so_create_func_t smodp_sock_create_func; 380 so_destroy_func_t smodp_sock_destroy_func; 381 so_proto_fallback_func_t smodp_proto_fallback_func; 382 } __smod_priv_t; 383 384 /* 385 * Socket module register information 386 */ 387 typedef struct smod_reg_s { 388 int smod_version; 389 char *smod_name; 390 size_t smod_uc_version; 391 size_t smod_dc_version; 392 so_proto_create_func_t smod_proto_create_func; 393 394 /* __smod_priv_data must be NULL */ 395 __smod_priv_t *__smod_priv; 396 } smod_reg_t; 397 398 /* 399 * Socket module information 400 */ 401 typedef struct smod_info { 402 int smod_version; 403 char *smod_name; 404 uint_t smod_refcnt; /* # of entries */ 405 size_t smod_uc_version; /* upcall version */ 406 size_t smod_dc_version; /* down call version */ 407 so_proto_create_func_t smod_proto_create_func; 408 so_proto_fallback_func_t smod_proto_fallback_func; 409 so_create_func_t smod_sock_create_func; 410 so_destroy_func_t smod_sock_destroy_func; 411 list_node_t smod_node; 412 } smod_info_t; 413 414 /* 415 * sockparams 416 * 417 * Used for mapping family/type/protocol to module 418 */ 419 struct sockparams { 420 /* 421 * The family, type, protocol, sdev_info and smod_info are 422 * set when the entry is created, and they will never change 423 * thereafter. 424 */ 425 int sp_family; 426 int sp_type; 427 int sp_protocol; 428 429 sdev_info_t sp_sdev_info; /* STREAM device */ 430 char *sp_smod_name; /* socket module name */ 431 smod_info_t *sp_smod_info; /* socket module */ 432 433 kmutex_t sp_lock; /* lock for refcnt */ 434 uint64_t sp_refcnt; /* entry reference count */ 435 436 /* 437 * The entries below are only modified while holding 438 * splist_lock as a writer. 439 */ 440 int sp_flags; /* see below */ 441 list_node_t sp_node; 442 }; 443 444 445 /* 446 * sockparams flags 447 */ 448 #define SOCKPARAMS_EPHEMERAL 0x1 /* temp. entry, not on global list */ 449 450 extern void sockparams_init(void); 451 extern struct sockparams *sockparams_hold_ephemeral_bydev(int, int, int, 452 const char *, int, int *); 453 extern struct sockparams *sockparams_hold_ephemeral_bymod(int, int, int, 454 const char *, int, int *); 455 extern void sockparams_ephemeral_drop_last_ref(struct sockparams *); 456 457 extern void smod_init(void); 458 extern void smod_add(smod_info_t *); 459 extern int smod_register(const smod_reg_t *); 460 extern int smod_unregister(const char *); 461 extern smod_info_t *smod_lookup_byname(const char *); 462 463 #define SOCKPARAMS_HAS_DEVICE(sp) \ 464 ((sp)->sp_sdev_info.sd_devpath != NULL) 465 466 /* Increase the smod_info_t reference count */ 467 #define SMOD_INC_REF(smodp) { \ 468 ASSERT((smodp) != NULL); \ 469 DTRACE_PROBE1(smodinfo__inc__ref, struct smod_info *, (smodp)); \ 470 atomic_inc_uint(&(smodp)->smod_refcnt); \ 471 } 472 473 /* 474 * Decreace the socket module entry reference count. 475 * When no one mapping to the entry, we try to unload the module from the 476 * kernel. If the module can't unload, just leave the module entry with 477 * a zero refcnt. 478 */ 479 #define SMOD_DEC_REF(sp, smodp) { \ 480 ASSERT((smodp) != NULL); \ 481 ASSERT((smodp)->smod_refcnt != 0); \ 482 atomic_dec_uint(&(smodp)->smod_refcnt); \ 483 /* \ 484 * No need to atomically check the return value because the \ 485 * socket module framework will verify that no one is using \ 486 * the module before unloading. Worst thing that can happen \ 487 * here is multiple calls to mod_remove_by_name(), which is OK. \ 488 */ \ 489 if ((smodp)->smod_refcnt == 0) \ 490 (void) mod_remove_by_name((sp)->sp_smod_name); \ 491 } 492 493 /* Increase the reference count */ 494 #define SOCKPARAMS_INC_REF(sp) { \ 495 ASSERT((sp) != NULL); \ 496 DTRACE_PROBE1(sockparams__inc__ref, struct sockparams *, (sp)); \ 497 mutex_enter(&(sp)->sp_lock); \ 498 (sp)->sp_refcnt++; \ 499 ASSERT((sp)->sp_refcnt != 0); \ 500 mutex_exit(&(sp)->sp_lock); \ 501 } 502 503 /* 504 * Decrease the reference count. 505 * 506 * If the sockparams is ephemeral, then the thread dropping the last ref 507 * count will destroy the entry. 508 */ 509 #define SOCKPARAMS_DEC_REF(sp) { \ 510 ASSERT((sp) != NULL); \ 511 DTRACE_PROBE1(sockparams__dec__ref, struct sockparams *, (sp)); \ 512 mutex_enter(&(sp)->sp_lock); \ 513 ASSERT((sp)->sp_refcnt > 0); \ 514 if ((sp)->sp_refcnt == 1) { \ 515 if ((sp)->sp_flags & SOCKPARAMS_EPHEMERAL) { \ 516 mutex_exit(&(sp)->sp_lock); \ 517 sockparams_ephemeral_drop_last_ref((sp)); \ 518 } else { \ 519 (sp)->sp_refcnt--; \ 520 if ((sp)->sp_smod_info != NULL) \ 521 SMOD_DEC_REF(sp, (sp)->sp_smod_info); \ 522 (sp)->sp_smod_info = NULL; \ 523 mutex_exit(&(sp)->sp_lock); \ 524 } \ 525 } else { \ 526 (sp)->sp_refcnt--; \ 527 mutex_exit(&(sp)->sp_lock); \ 528 } \ 529 } 530 531 /* 532 * Used to traverse the list of AF_UNIX sockets to construct the kstat 533 * for netstat(1m). 534 */ 535 struct socklist { 536 kmutex_t sl_lock; 537 struct sonode *sl_list; 538 }; 539 540 extern struct socklist socklist; 541 /* 542 * ss_full_waits is the number of times the reader thread 543 * waits when the queue is full and ss_empty_waits is the number 544 * of times the consumer thread waits when the queue is empty. 545 * No locks for these as they are just indicators of whether 546 * disk or network or both is slow or fast. 547 */ 548 struct sendfile_stats { 549 uint32_t ss_file_cached; 550 uint32_t ss_file_not_cached; 551 uint32_t ss_full_waits; 552 uint32_t ss_empty_waits; 553 uint32_t ss_file_segmap; 554 }; 555 556 /* 557 * A single sendfile request is represented by snf_req. 558 */ 559 typedef struct snf_req { 560 struct snf_req *sr_next; 561 mblk_t *sr_mp_head; 562 mblk_t *sr_mp_tail; 563 kmutex_t sr_lock; 564 kcondvar_t sr_cv; 565 uint_t sr_qlen; 566 int sr_hiwat; 567 int sr_lowat; 568 int sr_operation; 569 struct vnode *sr_vp; 570 file_t *sr_fp; 571 ssize_t sr_maxpsz; 572 u_offset_t sr_file_off; 573 u_offset_t sr_file_size; 574 #define SR_READ_DONE 0x80000000 575 int sr_read_error; 576 int sr_write_error; 577 } snf_req_t; 578 579 /* A queue of sendfile requests */ 580 struct sendfile_queue { 581 snf_req_t *snfq_req_head; 582 snf_req_t *snfq_req_tail; 583 kmutex_t snfq_lock; 584 kcondvar_t snfq_cv; 585 int snfq_svc_threads; /* # of service threads */ 586 int snfq_idle_cnt; /* # of idling threads */ 587 int snfq_max_threads; 588 int snfq_req_cnt; /* Number of requests */ 589 }; 590 591 #define READ_OP 1 592 #define SNFQ_TIMEOUT (60 * 5 * hz) /* 5 minutes */ 593 594 /* Socket network operations switch */ 595 struct sonodeops { 596 int (*sop_init)(struct sonode *, struct sonode *, cred_t *, 597 int); 598 int (*sop_accept)(struct sonode *, int, cred_t *, struct sonode **); 599 int (*sop_bind)(struct sonode *, struct sockaddr *, socklen_t, 600 int, cred_t *); 601 int (*sop_listen)(struct sonode *, int, cred_t *); 602 int (*sop_connect)(struct sonode *, const struct sockaddr *, 603 socklen_t, int, int, cred_t *); 604 int (*sop_recvmsg)(struct sonode *, struct msghdr *, 605 struct uio *, cred_t *); 606 int (*sop_sendmsg)(struct sonode *, struct msghdr *, 607 struct uio *, cred_t *); 608 int (*sop_sendmblk)(struct sonode *, struct msghdr *, int, 609 cred_t *, mblk_t **); 610 int (*sop_getpeername)(struct sonode *, struct sockaddr *, 611 socklen_t *, boolean_t, cred_t *); 612 int (*sop_getsockname)(struct sonode *, struct sockaddr *, 613 socklen_t *, cred_t *); 614 int (*sop_shutdown)(struct sonode *, int, cred_t *); 615 int (*sop_getsockopt)(struct sonode *, int, int, void *, 616 socklen_t *, int, cred_t *); 617 int (*sop_setsockopt)(struct sonode *, int, int, const void *, 618 socklen_t, cred_t *); 619 int (*sop_ioctl)(struct sonode *, int, intptr_t, int, 620 cred_t *, int32_t *); 621 int (*sop_poll)(struct sonode *, short, int, short *, 622 struct pollhead **); 623 int (*sop_close)(struct sonode *, int, cred_t *); 624 }; 625 626 #define SOP_INIT(so, flag, cr, flags) \ 627 ((so)->so_ops->sop_init((so), (flag), (cr), (flags))) 628 #define SOP_ACCEPT(so, fflag, cr, nsop) \ 629 ((so)->so_ops->sop_accept((so), (fflag), (cr), (nsop))) 630 #define SOP_BIND(so, name, namelen, flags, cr) \ 631 ((so)->so_ops->sop_bind((so), (name), (namelen), (flags), (cr))) 632 #define SOP_LISTEN(so, backlog, cr) \ 633 ((so)->so_ops->sop_listen((so), (backlog), (cr))) 634 #define SOP_CONNECT(so, name, namelen, fflag, flags, cr) \ 635 ((so)->so_ops->sop_connect((so), (name), (namelen), (fflag), (flags), \ 636 (cr))) 637 #define SOP_RECVMSG(so, msg, uiop, cr) \ 638 ((so)->so_ops->sop_recvmsg((so), (msg), (uiop), (cr))) 639 #define SOP_SENDMSG(so, msg, uiop, cr) \ 640 ((so)->so_ops->sop_sendmsg((so), (msg), (uiop), (cr))) 641 #define SOP_SENDMBLK(so, msg, size, cr, mpp) \ 642 ((so)->so_ops->sop_sendmblk((so), (msg), (size), (cr), (mpp))) 643 #define SOP_GETPEERNAME(so, addr, addrlen, accept, cr) \ 644 ((so)->so_ops->sop_getpeername((so), (addr), (addrlen), (accept), (cr))) 645 #define SOP_GETSOCKNAME(so, addr, addrlen, cr) \ 646 ((so)->so_ops->sop_getsockname((so), (addr), (addrlen), (cr))) 647 #define SOP_SHUTDOWN(so, how, cr) \ 648 ((so)->so_ops->sop_shutdown((so), (how), (cr))) 649 #define SOP_GETSOCKOPT(so, level, optionname, optval, optlenp, flags, cr) \ 650 ((so)->so_ops->sop_getsockopt((so), (level), (optionname), \ 651 (optval), (optlenp), (flags), (cr))) 652 #define SOP_SETSOCKOPT(so, level, optionname, optval, optlen, cr) \ 653 ((so)->so_ops->sop_setsockopt((so), (level), (optionname), \ 654 (optval), (optlen), (cr))) 655 #define SOP_IOCTL(so, cmd, arg, mode, cr, rvalp) \ 656 ((so)->so_ops->sop_ioctl((so), (cmd), (arg), (mode), (cr), (rvalp))) 657 #define SOP_POLL(so, events, anyyet, reventsp, phpp) \ 658 ((so)->so_ops->sop_poll((so), (events), (anyyet), (reventsp), (phpp))) 659 #define SOP_CLOSE(so, flag, cr) \ 660 ((so)->so_ops->sop_close((so), (flag), (cr))) 661 662 #endif /* defined(_KERNEL) || defined(_KMEMUSER) */ 663 664 #ifdef _KERNEL 665 666 #define ISALIGNED_cmsghdr(addr) \ 667 (((uintptr_t)(addr) & (_CMSG_HDR_ALIGNMENT - 1)) == 0) 668 669 #define ROUNDUP_cmsglen(len) \ 670 (((len) + _CMSG_HDR_ALIGNMENT - 1) & ~(_CMSG_HDR_ALIGNMENT - 1)) 671 672 #define IS_NON_STREAM_SOCK(vp) \ 673 ((vp)->v_type == VSOCK && (vp)->v_stream == NULL) 674 /* 675 * Macros that operate on struct cmsghdr. 676 * Used in parsing msg_control. 677 * The CMSG_VALID macro does not assume that the last option buffer is padded. 678 */ 679 #define CMSG_NEXT(cmsg) \ 680 (struct cmsghdr *)((uintptr_t)(cmsg) + \ 681 ROUNDUP_cmsglen((cmsg)->cmsg_len)) 682 #define CMSG_CONTENT(cmsg) (&((cmsg)[1])) 683 #define CMSG_CONTENTLEN(cmsg) ((cmsg)->cmsg_len - sizeof (struct cmsghdr)) 684 #define CMSG_VALID(cmsg, start, end) \ 685 (ISALIGNED_cmsghdr(cmsg) && \ 686 ((uintptr_t)(cmsg) >= (uintptr_t)(start)) && \ 687 ((uintptr_t)(cmsg) < (uintptr_t)(end)) && \ 688 ((ssize_t)(cmsg)->cmsg_len >= sizeof (struct cmsghdr)) && \ 689 ((uintptr_t)(cmsg) + (cmsg)->cmsg_len <= (uintptr_t)(end))) 690 691 /* 692 * Maximum size of any argument that is copied in (addresses, options, 693 * access rights). MUST be at least MAXPATHLEN + 3. 694 * BSD and SunOS 4.X limited this to MLEN or MCLBYTES. 695 */ 696 #define SO_MAXARGSIZE 8192 697 698 /* 699 * Convert between vnode and sonode 700 */ 701 #define VTOSO(vp) ((struct sonode *)((vp)->v_data)) 702 #define SOTOV(sp) ((sp)->so_vnode) 703 704 /* 705 * Internal flags for sobind() 706 */ 707 #define _SOBIND_REBIND 0x01 /* Bind to existing local address */ 708 #define _SOBIND_UNSPEC 0x02 /* Bind to unspecified address */ 709 #define _SOBIND_LOCK_HELD 0x04 /* so_excl_lock held by caller */ 710 #define _SOBIND_NOXLATE 0x08 /* No addr translation for AF_UNIX */ 711 #define _SOBIND_XPG4_2 0x10 /* xpg4.2 semantics */ 712 #define _SOBIND_SOCKBSD 0x20 /* BSD semantics */ 713 #define _SOBIND_LISTEN 0x40 /* Make into SS_ACCEPTCONN */ 714 #define _SOBIND_SOCKETPAIR 0x80 /* Internal flag for so_socketpair() */ 715 /* to enable listen with backlog = 1 */ 716 717 /* 718 * Internal flags for sounbind() 719 */ 720 #define _SOUNBIND_REBIND 0x01 /* Don't clear fields - will rebind */ 721 722 /* 723 * Internal flags for soconnect() 724 */ 725 #define _SOCONNECT_NOXLATE 0x01 /* No addr translation for AF_UNIX */ 726 #define _SOCONNECT_DID_BIND 0x02 /* Unbind when connect fails */ 727 #define _SOCONNECT_XPG4_2 0x04 /* xpg4.2 semantics */ 728 729 /* 730 * Internal flags for sodisconnect() 731 */ 732 #define _SODISCONNECT_LOCK_HELD 0x01 /* so_excl_lock held by caller */ 733 734 /* 735 * Internal flags for sotpi_getsockopt(). 736 */ 737 #define _SOGETSOCKOPT_XPG4_2 0x01 /* xpg4.2 semantics */ 738 739 /* 740 * Internal flags for soallocproto*() 741 */ 742 #define _ALLOC_NOSLEEP 0 /* Don't sleep for memory */ 743 #define _ALLOC_INTR 1 /* Sleep until interrupt */ 744 #define _ALLOC_SLEEP 2 /* Sleep forever */ 745 746 /* 747 * Internal structure for handling AF_UNIX file descriptor passing 748 */ 749 struct fdbuf { 750 int fd_size; /* In bytes, for kmem_free */ 751 int fd_numfd; /* Number of elements below */ 752 char *fd_ebuf; /* Extra buffer to free */ 753 int fd_ebuflen; 754 frtn_t fd_frtn; 755 struct file *fd_fds[1]; /* One or more */ 756 }; 757 #define FDBUF_HDRSIZE (sizeof (struct fdbuf) - sizeof (struct file *)) 758 759 /* 760 * Variable that can be patched to set what version of socket socket() 761 * will create. 762 */ 763 extern int so_default_version; 764 765 #ifdef DEBUG 766 /* Turn on extra testing capabilities */ 767 #define SOCK_TEST 768 #endif /* DEBUG */ 769 770 #ifdef DEBUG 771 char *pr_state(uint_t, uint_t); 772 char *pr_addr(int, struct sockaddr *, t_uscalar_t); 773 int so_verify_oobstate(struct sonode *); 774 #endif /* DEBUG */ 775 776 /* 777 * DEBUG macros 778 */ 779 #if defined(DEBUG) 780 #define SOCK_DEBUG 781 782 extern int sockdebug; 783 extern int sockprinterr; 784 785 #define eprint(args) printf args 786 #define eprintso(so, args) \ 787 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) printf args; } 788 #define eprintline(error) \ 789 { \ 790 if (error != EINTR && (sockprinterr || sockdebug > 0)) \ 791 printf("socket error %d: line %d file %s\n", \ 792 (error), __LINE__, __FILE__); \ 793 } 794 795 #define eprintsoline(so, error) \ 796 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) \ 797 printf("socket(%p) error %d: line %d file %s\n", \ 798 (void *)(so), (error), __LINE__, __FILE__); \ 799 } 800 #define dprint(level, args) { if (sockdebug > (level)) printf args; } 801 #define dprintso(so, level, args) \ 802 { if (sockdebug > (level) && ((so)->so_options & SO_DEBUG)) printf args; } 803 804 #else /* define(DEBUG) */ 805 806 #define eprint(args) {} 807 #define eprintso(so, args) {} 808 #define eprintline(error) {} 809 #define eprintsoline(so, error) {} 810 #define dprint(level, args) {} 811 #define dprintso(so, level, args) {} 812 813 #endif /* defined(DEBUG) */ 814 815 extern struct vfsops sock_vfsops; 816 extern struct vnodeops *socket_vnodeops; 817 extern const struct fs_operation_def socket_vnodeops_template[]; 818 819 extern dev_t sockdev; 820 821 /* 822 * sockfs functions 823 */ 824 extern int sock_getmsg(vnode_t *, struct strbuf *, struct strbuf *, 825 uchar_t *, int *, int, rval_t *); 826 extern int sock_putmsg(vnode_t *, struct strbuf *, struct strbuf *, 827 uchar_t, int, int); 828 extern int sogetvp(char *, vnode_t **, int); 829 extern int sockinit(int, char *); 830 extern int soconfig(int, int, int, char *, int, char *); 831 extern int solookup(int, int, int, struct sockparams **); 832 extern void so_lock_single(struct sonode *); 833 extern void so_unlock_single(struct sonode *, int); 834 extern int so_lock_read(struct sonode *, int); 835 extern int so_lock_read_intr(struct sonode *, int); 836 extern void so_unlock_read(struct sonode *); 837 extern void *sogetoff(mblk_t *, t_uscalar_t, t_uscalar_t, uint_t); 838 extern void so_getopt_srcaddr(void *, t_uscalar_t, 839 void **, t_uscalar_t *); 840 extern int so_getopt_unix_close(void *, t_uscalar_t); 841 extern void fdbuf_free(struct fdbuf *); 842 extern mblk_t *fdbuf_allocmsg(int, struct fdbuf *); 843 extern int fdbuf_create(void *, int, struct fdbuf **); 844 extern void so_closefds(void *, t_uscalar_t, int, int); 845 extern int so_getfdopt(void *, t_uscalar_t, int, void **, int *); 846 t_uscalar_t so_optlen(void *, t_uscalar_t, int); 847 extern void so_cmsg2opt(void *, t_uscalar_t, int, mblk_t *); 848 extern t_uscalar_t 849 so_cmsglen(mblk_t *, void *, t_uscalar_t, int); 850 extern int so_opt2cmsg(mblk_t *, void *, t_uscalar_t, int, 851 void *, t_uscalar_t); 852 extern void soisconnecting(struct sonode *); 853 extern void soisconnected(struct sonode *); 854 extern void soisdisconnected(struct sonode *, int); 855 extern void socantsendmore(struct sonode *); 856 extern void socantrcvmore(struct sonode *); 857 extern void soseterror(struct sonode *, int); 858 extern int sogeterr(struct sonode *, boolean_t); 859 extern int sowaitconnected(struct sonode *, int, int); 860 861 extern ssize_t soreadfile(file_t *, uchar_t *, u_offset_t, int *, size_t); 862 extern void *sock_kstat_init(zoneid_t); 863 extern void sock_kstat_fini(zoneid_t, void *); 864 extern struct sonode *getsonode(int, int *, file_t **); 865 /* 866 * Function wrappers (mostly around the sonode switch) for 867 * backward compatibility. 868 */ 869 extern int soaccept(struct sonode *, int, struct sonode **); 870 extern int sobind(struct sonode *, struct sockaddr *, socklen_t, 871 int, int); 872 extern int solisten(struct sonode *, int); 873 extern int soconnect(struct sonode *, const struct sockaddr *, socklen_t, 874 int, int); 875 extern int sorecvmsg(struct sonode *, struct nmsghdr *, struct uio *); 876 extern int sosendmsg(struct sonode *, struct nmsghdr *, struct uio *); 877 extern int soshutdown(struct sonode *, int); 878 extern int sogetsockopt(struct sonode *, int, int, void *, socklen_t *, 879 int); 880 extern int sosetsockopt(struct sonode *, int, int, const void *, 881 t_uscalar_t); 882 883 extern struct sonode *socreate(struct sockparams *, int, int, int, int, 884 int *); 885 886 extern int so_copyin(const void *, void *, size_t, int); 887 extern int so_copyout(const void *, void *, size_t, int); 888 889 #endif 890 891 /* 892 * Internal structure for obtaining sonode information from the socklist. 893 * These types match those corresponding in the sonode structure. 894 * This is not a published interface, and may change at any time. 895 */ 896 struct sockinfo { 897 uint_t si_size; /* real length of this struct */ 898 short si_family; 899 short si_type; 900 ushort_t si_flag; 901 uint_t si_state; 902 uint_t si_ux_laddr_sou_magic; 903 uint_t si_ux_faddr_sou_magic; 904 t_scalar_t si_serv_type; 905 t_uscalar_t si_laddr_soa_len; 906 t_uscalar_t si_faddr_soa_len; 907 uint16_t si_laddr_family; 908 uint16_t si_faddr_family; 909 char si_laddr_sun_path[MAXPATHLEN + 1]; /* NULL terminated */ 910 char si_faddr_sun_path[MAXPATHLEN + 1]; 911 boolean_t si_faddr_noxlate; 912 zoneid_t si_szoneid; 913 }; 914 915 #define SOCKMOD_PATH "socketmod" /* dir where sockmods are stored */ 916 917 #ifdef __cplusplus 918 } 919 #endif 920 921 #endif /* _SYS_SOCKETVAR_H */ 922