bsd/man/man2/kqueue.2

   1 .\"
   2 .\" Copyright (c) 2008-2015 Apple Inc.  All rights reserved.
   3 .\"
   4 .\" @APPLE_LICENSE_HEADER_START@
   5 .\"
   6 .\" This file contains Original Code and/or Modifications of Original Code
   7 .\" as defined in and that are subject to the Apple Public Source License
   8 .\" Version 2.0 (the 'License'). You may not use this file except in
   9 .\" compliance with the License. Please obtain a copy of the License at
  10 .\" http://www.opensource.apple.com/apsl/ and read it before using this
  11 .\" file.
  12 .\"
  13 .\" The Original Code and all software distributed under the License are
  14 .\" distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  15 .\" EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  16 .\" INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  17 .\" FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  18 .\" Please see the License for the specific language governing rights and
  19 .\" limitations under the License.
  20 .\"
  21 .\" @APPLE_LICENSE_HEADER_END@
  22 .\"
  23 .\"
  24 .\" Copyright (c) 2000 Jonathan Lemon
  25 .\" All rights reserved.
  26 .\"
  27 .\" Redistribution and use in source and binary forms, with or without
  28 .\" modification, are permitted provided that the following conditions
  29 .\" are met:
  30 .\" 1. Redistributions of source code must retain the above copyright
  31 .\"    notice, this list of conditions and the following disclaimer.
  32 .\" 2. Redistributions in binary form must reproduce the above copyright
  33 .\"    notice, this list of conditions and the following disclaimer in the
  34 .\"    documentation and/or other materials provided with the distribution.
  35 .\"
  36 .\" THIS SOFTWARE IS PROVIDED ``AS IS'' AND
  37 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  38 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  39 .\" ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  40 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  41 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  42 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  43 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  44 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  45 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  46 .\" SUCH DAMAGE.
  47 .\"
  48 .\" $FreeBSD: src/lib/libc/sys/kqueue.2,v 1.32 2002/12/19 09:40:25 ru Exp $
  49 .\"
  50 .Dd October 21, 2008
  51 .Dt KQUEUE 2
  52 .Os
  53 .Sh NAME
  54 .Nm kqueue ,
  55 .Nm kevent ,
  56 and
  57 .Nm kevent64
  58 .Nd kernel event notification mechanism
  59 .Sh LIBRARY
  60 .Lb libc
  61 .Sh SYNOPSIS
  62 .In sys/types.h
  63 .In sys/event.h
  64 .In sys/time.h
  65 .Ft int
  66 .Fn kqueue "void"
  67 .Ft int
  68 .Fn kevent "int kq" "const struct kevent *changelist" "int nchanges" "struct kevent *eventlist" "int nevents" "const struct timespec *timeout"
  69 .Ft int
  70 .Fn kevent64 "int kq" "const struct kevent64_s *changelist" "int nchanges" "struct kevent64_s *eventlist" "int nevents" "unsigned int flags" "const struct timespec *timeout"
  71 .Fn EV_SET "&kev" ident filter flags fflags data udata
  72 .Fn EV_SET64 "&kev" ident filter flags fflags data udata "ext[0]" "ext[1]"
  73 .Sh DESCRIPTION
  74 The
  75 .Fn kqueue
  76 system call
  77 provides a generic method of notifying the user when a kernel
  78 event (kevent) happens or a condition holds, based on the results
  79 of small pieces of kernel code termed filters.
  80 A kevent is identified by an (ident, filter) pair and specifies
  81 the interesting conditions to be notified about for that pair.
  82 An (ident, filter) pair can only appear once in a given kqueue.
  83 Subsequent attempts to register the same pair for a given kqueue
  84 will result in the replacement of the conditions being watched,
  85 not an addition.
  86 .Pp
  87 The filter identified in a kevent is executed upon the initial
  88 registration of that event in order to detect whether a preexisting
  89 condition is present, and is also executed whenever an event is
  90 passed to the filter for evaluation.
  91 If the filter determines that the condition should be reported,
  92 then the kevent is placed on the kqueue for the user to retrieve.
  93 .Pp
  94 The filter is also run when the user attempts to retrieve the kevent
  95 from the kqueue.
  96 If the filter indicates that the condition that triggered
  97 the event no longer holds, the kevent is removed from the kqueue and
  98 is not returned.
  99 .Pp
 100 Multiple events which trigger the filter do not result in multiple
 101 kevents being placed on the kqueue; instead, the filter will aggregate
 102 the events into a single struct kevent.
 103 Calling
 104 .Fn close
 105 on a file descriptor will remove any kevents that reference the descriptor.
 106 .Pp
 107 The
 108 .Fn kqueue
 109 system call
 110 creates a new kernel event queue and returns a descriptor.
 111 The queue is not inherited by a child created with
 112 .Xr fork 2 .
 113 .Pp
 114 The
 115 .Fn kevent
 116 and
 117 .Fn kevent64
 118 system calls
 119 are used to register events with the queue, and return any pending
 120 events to the user.
 121 The
 122 .Fa changelist
 123 argument
 124 is a pointer to an array of
 125 .Va kevent
 126 or
 127 .Va kevent64_s
 128 structures, as defined in
 129 .Aq Pa sys/event.h .
 130 All changes contained in the
 131 .Fa changelist
 132 are applied before any pending events are read from the queue.
 133 The
 134 .Fa nchanges
 135 argument
 136 gives the size of
 137 .Fa changelist .
 138 The
 139 .Fa eventlist
 140 argument
 141 is a pointer to an array of
 142 .Va kevent
 143 or
 144 .Va kevent64_s
 145 structures.
 146 The
 147 .Fa nevents
 148 argument
 149 determines the size of
 150 .Fa eventlist .
 151 If
 152 .Fa timeout
 153 is a non-NULL pointer, it specifies a maximum interval to wait
 154 for an event, which will be interpreted as a struct timespec.  If
 155 .Fa timeout
 156 is a NULL pointer, both
 157 .Fn kevent
 158 and
 159 .Fn kevent64
 160 wait indefinitely.  To effect a poll, the
 161 .Fa timeout
 162 argument should be non-NULL, pointing to a zero-valued
 163 .Va timespec
 164 structure.  The same array may be used for the
 165 .Fa changelist
 166 and
 167 .Fa eventlist .
 168 .Pp
 169 The
 170 .Fn EV_SET
 171 macro is provided for ease of initializing a
 172 .Va kevent
 173 structure. Similarly,
 174 .Fn EV_SET64
 175 initializes a
 176 .Va kevent64_s
 177 structure.
 178 .Pp
 179 The
 180 .Va kevent
 181 and
 182 .Va kevent64_s
 183 structures are defined as:
 184 .Bd -literal
 185 struct kevent {
 186         uintptr_t       ident;          /* identifier for this event */
 187         int16_t         filter;         /* filter for event */
 188         uint16_t        flags;          /* general flags */
 189         uint32_t        fflags;         /* filter-specific flags */
 190         intptr_t        data;           /* filter-specific data */
 191         void            *udata;         /* opaque user data identifier */
 192 };
 193
 194
 195 struct kevent64_s {
 196         uint64_t        ident;          /* identifier for this event */
 197         int16_t         filter;         /* filter for event */
 198         uint16_t        flags;          /* general flags */
 199         uint32_t        fflags;         /* filter-specific flags */
 200         int64_t         data;           /* filter-specific data */
 201         uint64_t        udata;          /* opaque user data identifier */
 202         uint64_t        ext[2];         /* filter-specific extensions */
 203 };
 204 .Ed
 205 .Pp
 206 ----
 207 .Pp
 208 The fields of
 209 .Fa struct kevent
 210 and
 211 .Fa struct kevent64_s
 212 are:
 213 .Bl -tag -width XXXfilter
 214 .It ident
 215 Value used to identify this event.
 216 The exact interpretation is determined by the attached filter,
 217 but often is a file descriptor.
 218 .It filter
 219 Identifies the kernel filter used to process this event.  The pre-defined
 220 system filters are described below.
 221 .It flags
 222 Actions to perform on the event.
 223 .It fflags
 224 Filter-specific flags.
 225 .It data
 226 Filter-specific data value.
 227 .It udata
 228 Opaque user-defined value passed through the kernel unchanged.
 229 .El
 230 .Pp
 231 In addition,
 232 .Fa struct kevent64_s
 233 contains:
 234 .Bl -tag -width XXXfilter
 235 .It ext[2]
 236 This field stores extensions for the event's filter. What type of extension depends on
 237 what type of filter is being used.
 238 .El
 239 .Pp
 240 ----
 241 .Pp
 242 The
 243 .Va flags
 244 field can contain the following values:
 245 .Bl -tag -width XXXEV_ONESHOT
 246 .It EV_ADD
 247 Adds the event to the kqueue.  Re-adding an existing event
 248 will modify the parameters of the original event, and not result
 249 in a duplicate entry.  Adding an event automatically enables it,
 250 unless overridden by the EV_DISABLE flag.
 251 .It EV_ENABLE
 252 Permit
 253 .Fn kevent
 254 and
 255 .Fn kevent64
 256 to return the event if it is triggered.
 257 .It EV_DISABLE
 258 Disable the event so
 259 .Fn kevent
 260 and
 261 .Fn kevent64
 262 will not return it.  The filter itself is not disabled.
 263 .It EV_DELETE
 264 Removes the event from the kqueue.  Events which are attached to
 265 file descriptors are automatically deleted on the last close of
 266 the descriptor.
 267 .It EV_RECEIPT
 268 This flag is useful for making bulk changes to a kqueue without draining any
 269 pending events. When passed as input, it forces EV_ERROR to always be returned.
 270 When a filter is successfully added, the
 271 .Va data
 272 field will be zero.
 273 .It EV_ONESHOT
 274 Causes the event to return only the first occurrence of the filter
 275 being triggered.  After the user retrieves the event from the kqueue,
 276 it is deleted.
 277 .It EV_CLEAR
 278 After the event is retrieved by the user, its state is reset.
 279 This is useful for filters which report state transitions
 280 instead of the current state.  Note that some filters may automatically
 281 set this flag internally.
 282 .It EV_EOF
 283 Filters may set this flag to indicate filter-specific EOF condition.
 284 .It EV_OOBAND
 285 Read filter on socket may set this flag to indicate the presence of out of
 286 band data on the descriptor.
 287 .It EV_ERROR
 288 See
 289 .Sx RETURN VALUES
 290 below.
 291 .El
 292 .Pp
 293 ----
 294 .Pp
 295 The predefined system filters are listed below.
 296 Arguments may be passed to and from the filter via the
 297 .Va fflags
 298 and
 299 .Va data
 300 fields in the
 301 .Va kevent
 302 or
 303 .Va kevent64_s
 304 structure.
 305 .Bl -tag -width EVFILT_MACHPORT
 306 .It EVFILT_READ
 307 Takes a file descriptor as the identifier, and returns whenever
 308 there is data available to read.
 309 The behavior of the filter is slightly different depending
 310 on the descriptor type.
 311 .Pp
 312 .Bl -tag -width 2n
 313 .It Sockets
 314 Sockets which have previously been passed to
 315 .Fn listen
 316 return when there is an incoming connection pending.
 317 .Va data
 318 contains the size of the listen backlog.
 319 .Pp
 320 Other socket descriptors return when there is data to be read,
 321 subject to the
 322 .Dv SO_RCVLOWAT
 323 value of the socket buffer.
 324 This may be overridden with a per-filter low water mark at the
 325 time the filter is added by setting the
 326 NOTE_LOWAT
 327 flag in
 328 .Va fflags ,
 329 and specifying the new low water mark in
 330 .Va data .
 331 On return,
 332 .Va data
 333 contains the number of bytes of protocol data available to read.
 334 .Pp
 335 The presence of EV_OOBAND in
 336 .Va flags ,
 337 indicates the presence of out of band data on the socket
 338 .Va data
 339 equal to the potential number of OOB bytes availble to read.
 340 .Pp
 341 If the read direction of the socket has shutdown, then the filter
 342 also sets EV_EOF in
 343 .Va flags ,
 344 and returns the socket error (if any) in
 345 .Va fflags .
 346 It is possible for EOF to be returned (indicating the connection is gone)
 347 while there is still data pending in the socket buffer.
 348 .It Vnodes
 349 Returns when the file pointer is not at the end of file.
 350 .Va data
 351 contains the offset from current position to end of file,
 352 and may be negative.
 353 .It "Fifos, Pipes"
 354 Returns when the there is data to read;
 355 .Va data
 356 contains the number of bytes available.
 357 .Pp
 358 When the last writer disconnects, the filter will set EV_EOF in
 359 .Va flags .
 360 This may be cleared by passing in EV_CLEAR, at which point the
 361 filter will resume waiting for data to become available before
 362 returning.
 363 .El
 364 .It EVFILT_WRITE
 365 Takes a file descriptor as the identifier, and returns whenever
 366 it is possible to write to the descriptor.  For sockets, pipes
 367 and fifos,
 368 .Va data
 369 will contain the amount of space remaining in the write buffer.
 370 The filter will set EV_EOF when the reader disconnects, and for
 371 the fifo case, this may be cleared by use of EV_CLEAR.
 372 Note that this filter is not supported for vnodes.
 373 .Pp
 374 For sockets, the low water mark and socket error handling is
 375 identical to the EVFILT_READ case.
 376 .It EVFILT_AIO
 377 This filter is currently unsupported.
 378 .\"The sigevent portion of the AIO request is filled in, with
 379 .\".Va sigev_notify_kqueue
 380 .\"containing the descriptor of the kqueue that the event should
 381 .\"be attached to,
 382 .\".Va sigev_value
 383 .\"containing the udata value, and
 384 .\".Va sigev_notify
 385 .\"set to SIGEV_KEVENT.
 386 .\"When the
 387 .\".Fn aio_*
 388 .\"system call is made, the event will be registered
 389 .\"with the specified kqueue, and the
 390 .\".Va ident
 391 .\"argument set to the
 392 .\".Fa struct aiocb
 393 .\"returned by the
 394 .\".Fn aio_*
 395 .\"system call.
 396 .\"The filter returns under the same conditions as aio_error.
 397 .\".Pp
 398 .\"Alternatively, a kevent structure may be initialized, with
 399 .\".Va ident
 400 .\"containing the descriptor of the kqueue, and the
 401 .\"address of the kevent structure placed in the
 402 .\".Va aio_lio_opcode
 403 .\"field of the AIO request.  However, this approach will not work on
 404 .\"architectures with 64-bit pointers, and should be considered deprecated.
 405 .It EVFILT_VNODE
 406 Takes a file descriptor as the identifier and the events to watch for in
 407 .Va fflags ,
 408 and returns when one or more of the requested events occurs on the descriptor.
 409 The events to monitor are:
 410 .Bl -tag -width XXNOTE_RENAME
 411 .It NOTE_DELETE
 412 The
 413 .Fn unlink
 414 system call
 415 was called on the file referenced by the descriptor.
 416 .It NOTE_WRITE
 417 A write occurred on the file referenced by the descriptor.
 418 .It NOTE_EXTEND
 419 The file referenced by the descriptor was extended.
 420 .It NOTE_ATTRIB
 421 The file referenced by the descriptor had its attributes changed.
 422 .It NOTE_LINK
 423 The link count on the file changed.
 424 .It NOTE_RENAME
 425 The file referenced by the descriptor was renamed.
 426 .It NOTE_REVOKE
 427 Access to the file was revoked via
 428 .Xr revoke 2
 429 or the underlying fileystem was unmounted.
 430 .El
 431 .Pp
 432 On return,
 433 .Va fflags
 434 contains the events which triggered the filter.
 435 .It EVFILT_PROC
 436 Takes the process ID to monitor as the identifier and the events to watch for
 437 in
 438 .Va fflags ,
 439 and returns when the process performs one or more of the requested events.
 440 If a process can normally see another process, it can attach an event to it.
 441 The events to monitor are:
 442 .Bl -tag -width NOTE_SIGNAL
 443 .It NOTE_EXIT
 444 The process has exited.
 445 .It NOTE_EXITSTATUS
 446 The process has exited and its exit status is in filter specific data. Valid only on child processes and to be used along with NOTE_EXIT.
 447 .It NOTE_FORK
 448 The process created a child process via
 449 .Xr fork 2
 450 or similar call.
 451 .It NOTE_EXEC
 452 The process executed a new process via
 453 .Xr execve 2
 454 or similar call.
 455 .It NOTE_SIGNAL
 456 The process was sent a signal. Status can be checked via
 457 .Xr waitpid 2
 458 or similar call.
 459 .It NOTE_REAP
 460 The process was reaped by the parent via
 461 .Xr wait 2
 462 or similar call. Deprecated, use NOTE_EXIT.
 463 .El
 464 .Pp
 465 On return,
 466 .Va fflags
 467 contains the events which triggered the filter.
 468 .It EVFILT_SIGNAL
 469 Takes the signal number to monitor as the identifier and returns
 470 when the given signal is generated for the process.
 471 This coexists with the
 472 .Fn signal
 473 and
 474 .Fn sigaction
 475 facilities, and has a lower precedence.  Only signals sent to the process,
 476 not to a particular thread, will trigger the filter. The filter will record
 477 all attempts to deliver a signal to a process, even if the signal has
 478 been marked as SIG_IGN.  Event notification happens before normal
 479 signal delivery processing.
 480 .Va data
 481 returns the number of times the signal has been generated since the last call to
 482 .Fn kevent .
 483 This filter automatically sets the EV_CLEAR flag internally.
 484 .It EVFILT_MACHPORT
 485 Takes the name of a mach port, or port set, in
 486 .Va ident
 487 and waits until a message is received on the port or port set. When a message
 488 is recieved, the size of the message is returned in
 489 .Va data
 490 and if
 491 .Va fflags
 492 is set to MACH_RCV_MSG, a pointer to the message is returned in ext[0].
 493 .It EVFILT_TIMER
 494 Establishes an interval timer with the data
 495 timer identified by
 496 .Va ident .
 497 When adding a timer,
 498 .Va data
 499 specifies the timeout period and
 500 .Va fflags
 501 can be set to one of the following:
 502 .Bl -tag -width NOTE_ABSOLUTE
 503 .It NOTE_SECONDS
 504 data is in seconds
 505 .It NOTE_USECONDS
 506 data is in microseconds
 507 .It NOTE_NSECONDS
 508 data is in nanoseconds
 509 .It NOTE_ABSOLUTE
 510 data is an absolute timeout
 511 .It NOTE_CRITICAL
 512 system makes a best effort to fire this timer as scheduled.
 513 .It NOTE_BACKGROUND
 514 system has extra leeway to coalesce this timer.
 515 .It NOTE_LEEWAY
 516 ext[1] holds user-supplied slop in deadline for timer coalescing.
 517 .El
 518 .Pp
 519 If fflags is not set, the default is milliseconds. The timer will be periodic unless EV_ONESHOT is specified.
 520 On return,
 521 .Va data
 522 contains the number of times the timeout has expired since the last call to
 523 .Fn kevent
 524 or
 525 .Fn kevent64 .
 526 This filter automatically sets the EV_CLEAR flag internally.
 527 .El
 528 .Pp
 529 ----
 530 .Pp
 531 In the
 532 .Va ext[2]
 533 field of the
 534 .Va kevent64_s
 535 struture,
 536 .Va ext[0]
 537 is only used with the EVFILT_MACHPORT filter.
 538 With other filters,
 539 .Va ext[0]
 540 is passed through
 541 .Fn kevent64
 542 much like
 543 .Va udata .
 544 .Va ext[1]
 545 can always be used like
 546 .Va udata .
 547 For the use of ext[0], see the EVFILT_MACHPORT filter above.
 548 .Sh RETURN VALUES
 549 The
 550 .Fn kqueue
 551 system call
 552 creates a new kernel event queue and returns a file descriptor.
 553 If there was an error creating the kernel event queue, a value of -1 is
 554 returned and errno set.
 555 .Pp
 556 The
 557 .Fn kevent
 558 and
 559 .Fn kevent64
 560 system calls
 561 return the number of events placed in the
 562 .Fa eventlist ,
 563 up to the value given by
 564 .Fa nevents .
 565 If an error occurs while processing an element of the
 566 .Fa changelist
 567 and there is enough room in the
 568 .Fa eventlist ,
 569 then the event will be placed in the
 570 .Fa eventlist
 571 with
 572 .Dv EV_ERROR
 573 set in
 574 .Va flags
 575 and the system error in
 576 .Va data .
 577 Otherwise,
 578 .Dv -1
 579 will be returned, and
 580 .Dv errno
 581 will be set to indicate the error condition.
 582 If the time limit expires, then
 583 .Fn kevent
 584 and
 585 .Fn kevent64
 586 return 0.
 587 .Sh ERRORS
 588 The
 589 .Fn kqueue
 590 system call fails if:
 591 .Bl -tag -width Er
 592 .It Bq Er ENOMEM
 593 The kernel failed to allocate enough memory for the kernel queue.
 594 .It Bq Er EMFILE
 595 The per-process descriptor table is full.
 596 .It Bq Er ENFILE
 597 The system file table is full.
 598 .El
 599 .Pp
 600 The
 601 .Fn kevent
 602 and
 603 .Fn kevent64
 604 system calls fail if:
 605 .Bl -tag -width Er
 606 .It Bq Er EACCES
 607 The process does not have permission to register a filter.
 608 .It Bq Er EFAULT
 609 There was an error reading or writing the
 610 .Va kevent
 611 or
 612 .Va kevent64_s
 613 structure.
 614 .It Bq Er EBADF
 615 The specified descriptor is invalid.
 616 .It Bq Er EINTR
 617 A signal was delivered before the timeout expired and before any
 618 events were placed on the kqueue for return.
 619 .It Bq Er EINVAL
 620 The specified time limit or filter is invalid.
 621 .It Bq Er ENOENT
 622 The event could not be found to be modified or deleted.
 623 .It Bq Er ENOMEM
 624 No memory was available to register the event.
 625 .It Bq Er ESRCH
 626 The specified process to attach to does not exist.
 627 .El
 628 .Sh SEE ALSO
 629 .Xr aio_error 2 ,
 630 .Xr aio_read 2 ,
 631 .Xr aio_return 2 ,
 632 .Xr read 2 ,
 633 .Xr select 2 ,
 634 .Xr sigaction 2 ,
 635 .Xr write 2 ,
 636 .Xr signal 3
 637 .Sh HISTORY
 638 The
 639 .Fn kqueue
 640 and
 641 .Fn kevent
 642 system calls first appeared in
 643 .Fx 4.1 .
 644 .Sh AUTHORS
 645 The
 646 .Fn kqueue
 647 system and this manual page were written by
 648 .An Jonathan Lemon Aq jlemon@FreeBSD.org .
 649 .Sh BUGS
 650 Not all filesystem types support kqueue-style notifications.
 651 And even some that do, like some remote filesystems, may only
 652 support a subset of the notification semantics described
 653 here.