1.\" Copyright (c) 2000 Jonathan Lemon 2.\" All rights reserved. 3.\" 4.\" Redistribution and use in source and binary forms, with or without 5.\" modification, are permitted provided that the following conditions 6.\" are met: 7.\" 1. Redistributions of source code must retain the above copyright 8.\" notice, this list of conditions and the following disclaimer. 9.\" 2. Redistributions in binary form must reproduce the above copyright 10.\" notice, this list of conditions and the following disclaimer in the 11.\" documentation and/or other materials provided with the distribution. 12.\" 13.\" THIS SOFTWARE IS PROVIDED ``AS IS'' AND 14.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23.\" SUCH DAMAGE. 24.\" 25.\" $FreeBSD: src/lib/libc/sys/kqueue.2,v 1.1.2.16 2002/07/02 21:05:08 mp Exp $ 26.\" 27.Dd February 26, 2013 28.Dt KQUEUE 2 29.Os 30.Sh NAME 31.Nm kqueue , 32.Nm kevent 33.Nd kernel event notification mechanism 34.Sh LIBRARY 35.Lb libc 36.Sh SYNOPSIS 37.In sys/types.h 38.In sys/event.h 39.In sys/time.h 40.Ft int 41.Fn kqueue "void" 42.Ft int 43.Fn kevent "int kq" "const struct kevent *changelist" "int nchanges" "struct kevent *eventlist" "int nevents" "const struct timespec *timeout" 44.Fn EV_SET "&kev" ident filter flags fflags data udata 45.Sh DESCRIPTION 46.Fn kqueue 47provides a generic method of notifying the user when an event 48happens or a condition holds, based on the results of small 49pieces of kernel code termed filters. 50A kevent is identified by the (ident, filter) pair; there may only 51be one unique kevent per kqueue. 52.Pp 53The filter is executed upon the initial registration of a kevent 54in order to detect whether a preexisting condition is present, and is also 55executed whenever an event is passed to the filter for evaluation. 56If the filter determines that the condition should be reported, 57then the kevent is placed on the kqueue for the user to retrieve. 58.Pp 59The filter is also run when the user attempts to retrieve the kevent 60from the kqueue. 61If the filter indicates that the condition that triggered 62the event no longer holds, the kevent is removed from the kqueue and 63is not returned. 64.Pp 65Multiple events which trigger the filter do not result in multiple 66kevents being placed on the kqueue; instead, the filter will aggregate 67the events into a single struct kevent. 68Calling 69.Fn close 70on a file descriptor will remove any kevents that reference the descriptor. 71.Pp 72.Fn kqueue 73creates a new kernel event queue and returns a descriptor. 74The queue is not inherited by a child created with 75.Xr fork 2 . 76However, if 77.Xr rfork 2 78is called without the 79.Dv RFFDG 80flag, then the descriptor table is shared, 81which will allow sharing of the kqueue between two processes. 82.Pp 83.Fn kevent 84is used to register events with the queue, and return any pending 85events to the user. 86.Fa changelist 87is a pointer to an array of 88.Va kevent 89structures, as defined in 90.In sys/event.h . 91All changes contained in the 92.Fa changelist 93are applied before any pending events are read from the queue. 94.Fa nchanges 95gives the size of 96.Fa changelist . 97.Fa eventlist 98is a pointer to an array of kevent structures. 99.Fa nevents 100determines the size of 101.Fa eventlist . 102If 103.Fa timeout 104is a non-NULL pointer, it specifies a maximum interval to wait 105for an event, which will be interpreted as a struct timespec. 106If 107.Fa timeout 108is a NULL pointer, 109.Fn kevent 110waits indefinitely. 111To effect a poll, the 112.Fa timeout 113argument should be non-NULL, pointing to a zero-valued 114.Va timespec 115structure. 116The same array may be used for the 117.Fa changelist 118and 119.Fa eventlist . 120.Pp 121.Fn EV_SET 122is a macro which is provided for ease of initializing a 123kevent structure. 124.Pp 125The 126.Va kevent 127structure is defined as: 128.Bd -literal 129struct kevent { 130 uintptr_t ident; /* identifier for this event */ 131 short filter; /* filter for event */ 132 u_short flags; /* action flags for kqueue */ 133 u_int fflags; /* filter flag value */ 134 intptr_t data; /* filter data value */ 135 void *udata; /* opaque user data identifier */ 136}; 137.Ed 138.Pp 139The fields of 140.Fa struct kevent 141are: 142.Bl -tag -width XXXfilter 143.It ident 144Value used to identify this event. 145The exact interpretation is determined by the attached filter, 146but often is a file descriptor. 147.It filter 148Identifies the kernel filter used to process this event. 149The pre-defined system filters are described below. 150.It flags 151Actions to perform on the event. 152.It fflags 153Filter-specific flags. 154.It data 155Filter-specific data value. 156.It udata 157Opaque user-defined value passed through the kernel unchanged. 158.El 159.Pp 160The 161.Va flags 162field can contain the following values: 163.Bl -tag -width ".Dv EV_ONESHOT" 164.It Dv EV_ADD 165Adds the event to the kqueue. 166Re-adding an existing event will modify the parameters of the original 167event, and not result in a duplicate entry. 168Adding an event automatically enables it, unless overridden by the 169.Dv EV_DISABLE 170flag. 171.It Dv EV_ENABLE 172Permit 173.Fn kevent 174to return the event if it is triggered. 175.It Dv EV_DISABLE 176Disable the event so 177.Fn kevent 178will not return it. 179The filter itself is not disabled. 180.It Dv EV_DELETE 181Removes the event from the kqueue. 182Events which are attached to file descriptors are automatically 183deleted on the last close of the descriptor. 184.It Dv EV_ONESHOT 185Causes the event to return only the first occurrence of the filter 186being triggered. 187After the user retrieves the event from the kqueue, it is deleted. 188.It Dv EV_CLEAR 189After the event is retrieved by the user, its state is reset. 190This is useful for filters which report state transitions 191instead of the current state. 192Note that some filters may automatically set this flag internally. 193.It Dv EV_EOF 194Filters may set this flag to indicate filter-specific EOF condition. 195.It Dv EV_ERROR 196See 197.Sx RETURN VALUES 198below. 199.El 200.Pp 201The predefined system filters are listed below. 202Arguments may be passed to and from the filter via the 203.Va fflags 204and 205.Va data 206fields in the kevent structure. 207.Bl -tag -width ".Dv EVFILT_SIGNAL" 208.It Dv EVFILT_READ 209Takes a descriptor as the identifier, and returns whenever 210there is data available to read. 211The behavior of the filter is slightly different depending 212on the descriptor type. 213.Bl -tag -width 2n 214.It Sockets 215Sockets which have previously been passed to 216.Fn listen 217return when there is an incoming connection pending. 218.Va data 219contains the size of the listen backlog. 220.Pp 221Other socket descriptors return when there is data to be read, 222subject to the 223.Dv SO_RCVLOWAT 224value of the socket buffer. 225This may be overridden with a per-filter low water mark at the 226time the filter is added by setting the 227.Dv NOTE_LOWAT 228flag in 229.Va fflags , 230and specifying the new low water mark in 231.Va data . 232On return, 233.Va data 234contains the number of bytes in the socket buffer. 235.Pp 236If the read direction of the socket has shutdown, then the filter also sets 237.Dv EV_EOF 238in 239.Va flags , 240and returns the socket error (if any) in 241.Va fflags . 242It is possible for EOF to be returned (indicating the connection is gone) 243while there is still data pending in the socket buffer. 244.It Vnodes 245Returns when the file pointer is not at the end of file. 246.Va data 247contains the offset from current position to end of file, 248and may be negative. 249.It "Fifos, Pipes" 250Returns when the there is data to read; 251.Va data 252contains the number of bytes available. 253.Pp 254When the last writer disconnects, the filter will set 255.Dv EV_EOF 256in 257.Va flags . 258This may be cleared by passing in 259.Dv EV_CLEAR , 260at which point the filter will resume waiting for data to become 261available before returning. 262.El 263.It Dv EVFILT_WRITE 264Takes a descriptor as the identifier, and returns whenever 265it is possible to write to the descriptor. 266For sockets, pipes and fifos, 267.Va data 268will contain the amount of space remaining in the write buffer. 269The filter will set 270.Dv EV_EOF 271when the reader disconnects, and for the fifo case, this may be cleared 272by use of 273.Dv EV_CLEAR . 274Note that this filter is not supported for vnodes. 275.Pp 276For sockets, the low water mark and socket error handling is 277identical to the 278.Dv EVFILT_READ 279case. 280.It Dv EVFILT_EXCEPT 281Takes a descriptor as the identifier, and returns whenever one of the 282specified exceptional conditions has occurred on the descriptor. Conditions 283are specified in 284.Va fflags . 285Currently, a filter can monitor the reception of out-of-band data with 286.Dv NOTE_OOB . 287.It Dv EVFILT_AIO 288The sigevent portion of the AIO request is filled in, with 289.Va sigev_notify_kqueue 290containing the descriptor of the kqueue that the event should 291be attached to, 292.Va sigev_value 293containing the udata value, and 294.Va sigev_notify 295set to 296.Dv SIGEV_KEVENT . 297When the aio_* function is called, the event will be registered 298with the specified kqueue, and the 299.Va ident 300argument set to the 301.Fa struct aiocb 302returned by the aio_* function. 303The filter returns under the same conditions as aio_error. 304.Pp 305Alternatively, a kevent structure may be initialized, with 306.Va ident 307containing the descriptor of the kqueue, and the 308address of the kevent structure placed in the 309.Va aio_lio_opcode 310field of the AIO request. 311However, this approach will not work on architectures with 64-bit 312pointers, and should be considered deprecated. 313.It Dv EVFILT_VNODE 314Takes a file descriptor as the identifier and the events to watch for in 315.Va fflags , 316and returns when one or more of the requested events occurs on the descriptor. 317The events to monitor are: 318.Bl -tag -width ".Dv NOTE_RENAME" 319.It Dv NOTE_DELETE 320.Fn unlink 321was called on the file referenced by the descriptor. 322.It Dv NOTE_WRITE 323A write occurred on the file referenced by the descriptor. 324.It Dv NOTE_EXTEND 325The file referenced by the descriptor was extended. 326.It Dv NOTE_ATTRIB 327The file referenced by the descriptor had its attributes changed. 328.It Dv NOTE_LINK 329The link count on the file changed. 330.It Dv NOTE_RENAME 331The file referenced by the descriptor was renamed. 332.It Dv NOTE_REVOKE 333Access to the file was revoked via 334.Xr revoke 2 335or the underlying fileystem was unmounted. 336.El 337.Pp 338On return, 339.Va fflags 340contains the events which triggered the filter. 341.It Dv EVFILT_PROC 342Takes the process ID to monitor as the identifier and the events to watch for 343in 344.Va fflags , 345and returns when the process performs one or more of the requested events. 346If a process can normally see another process, it can attach an event to it. 347The events to monitor are: 348.Bl -tag -width ".Dv NOTE_TRACKERR" 349.It Dv NOTE_EXIT 350The process has exited. 351.It Dv NOTE_FORK 352The process has called 353.Fn fork . 354.It Dv NOTE_EXEC 355The process has executed a new process via 356.Xr execve 2 357or similar call. 358.It Dv NOTE_TRACK 359Follow a process across 360.Fn fork 361calls. 362The parent process will return with 363.Dv NOTE_TRACK 364set in the 365.Va fflags 366field, while the child process will return with 367.Dv NOTE_CHILD 368set in 369.Va fflags 370and the parent PID in 371.Va data . 372.It Dv NOTE_TRACKERR 373This flag is returned if the system was unable to attach an event to 374the child process, usually due to resource limitations. 375.El 376.Pp 377On return, 378.Va fflags 379contains the events which triggered the filter. 380.It Dv EVFILT_SIGNAL 381Takes the signal number to monitor as the identifier and returns 382when the given signal is delivered to the process. 383This coexists with the 384.Fn signal 385and 386.Fn sigaction 387facilities, and has a lower precedence. 388The filter will record all attempts to deliver a signal to a process, 389even if the signal has been marked as 390.Dv SIG_IGN . 391Event notification happens after normal signal delivery processing. 392.Va data 393returns the number of times the signal has occurred since the last call to 394.Fn kevent . 395This filter automatically sets the 396.Dv EV_CLEAR 397flag internally. 398.It Dv EVFILT_TIMER 399Establishes an arbitrary timer identified by 400.Va ident . 401When adding a timer, 402.Va data 403specifies the timeout period in milliseconds. 404The timer will be periodic unless 405.Dv EV_ONESHOT 406is specified. 407On return, 408.Va data 409contains the number of times the timeout has expired since the last call to 410.Fn kevent . 411This filter automatically sets the 412.Dv EV_CLEAR 413flag internally. 414.El 415.Sh RETURN VALUES 416.Fn kqueue 417creates a new kernel event queue and returns a file descriptor. 418If there was an error creating the kernel event queue, a value of -1 is 419returned and 420.Va errno 421set. 422.Pp 423.Fn kevent 424returns the number of events placed in the 425.Fa eventlist , 426up to the value given by 427.Fa nevents . 428If an error occurs while processing an element of the 429.Fa changelist 430and there is enough room in the 431.Fa eventlist , 432then the event will be placed in the 433.Fa eventlist 434with 435.Dv EV_ERROR 436set in 437.Va flags 438and the system error in 439.Va data . 440Otherwise, 441.Dv -1 442will be returned, and 443.Va errno 444will be set to indicate the error condition. 445If the time limit expires, then 446.Fn kevent 447returns 0. 448.Sh ERRORS 449The 450.Fn kqueue 451function fails if: 452.Bl -tag -width Er 453.It Bq Er ENOMEM 454The kernel failed to allocate enough memory for the kernel queue. 455.It Bq Er EMFILE 456The per-process descriptor table is full. 457.It Bq Er ENFILE 458The system file table is full. 459.El 460.Pp 461The 462.Fn kevent 463function fails if: 464.Bl -tag -width Er 465.It Bq Er EACCES 466The process does not have permission to register a filter. 467.It Bq Er EFAULT 468There was an error reading or writing the 469.Va kevent 470structure. 471.It Bq Er EBADF 472The specified descriptor is invalid. 473.It Bq Er EINTR 474A signal was delivered before the timeout expired and before any 475events were placed on the kqueue for return. 476.It Bq Er EINVAL 477The specified time limit or filter is invalid. 478.It Bq Er ENOENT 479The event could not be found to be modified or deleted. 480.It Bq Er ENOMEM 481No memory was available to register the event. 482.It Bq Er ESRCH 483The specified process to attach to does not exist. 484.El 485.Sh SEE ALSO 486.Xr poll 2 , 487.Xr read 2 , 488.Xr select 2 , 489.Xr sigaction 2 , 490.Xr write 2 , 491.Xr aio_error 3 , 492.Xr aio_read 3 , 493.Xr aio_return 3 , 494.Xr signal 3 495.Sh HISTORY 496The 497.Fn kqueue 498and 499.Fn kevent 500functions first appeared in 501.Fx 4.1 . 502.Sh AUTHORS 503The 504.Fn kqueue 505system and this manual page were written by 506.An Jonathan Lemon Aq Mt jlemon@FreeBSD.org . 507.Sh BUGS 508Currently it is only possible to watch a 509.Xr vnode 9 510on 511.Xr HAMMER 5 , 512.Xr tmpfs 5 513and 514.Xr UFS 5 515file systems. 516