1 @node Low-Level I/O, File System Interface, I/O on Streams, Top
2 @c %MENU% Low-level, less portable I/O
3 @chapter Low-Level Input/Output
5 This chapter describes functions for performing low-level input/output
6 operations on file descriptors. These functions include the primitives
7 for the higher-level I/O functions described in @ref{I/O on Streams}, as
8 well as functions for performing low-level control operations for which
9 there are no equivalents on streams.
11 Stream-level I/O is more flexible and usually more convenient;
12 therefore, programmers generally use the descriptor-level functions only
13 when necessary. These are some of the usual reasons:
17 For reading binary files in large chunks.
20 For reading an entire file into core before parsing it.
23 To perform operations other than data transfer, which can only be done
24 with a descriptor. (You can use @code{fileno} to get the descriptor
25 corresponding to a stream.)
28 To pass descriptors to a child process. (The child can create its own
29 stream to use a descriptor that it inherits, but cannot inherit a stream
34 * Opening and Closing Files:: How to open and close file
36 * Truncating Files:: Change the size of a file.
37 * I/O Primitives:: Reading and writing data.
38 * File Position Primitive:: Setting a descriptor's file
40 * Descriptors and Streams:: Converting descriptor to stream
42 * Stream/Descriptor Precautions:: Precautions needed if you use both
43 descriptors and streams.
44 * Scatter-Gather:: Fast I/O to discontinous buffers.
45 * Memory-mapped I/O:: Using files like memory.
46 * Waiting for I/O:: How to check for input or output
47 on multiple file descriptors.
48 * Synchronizing I/O:: Making sure all I/O actions completed.
49 * Asynchronous I/O:: Perform I/O in parallel.
50 * Control Operations:: Various other operations on file
52 * Duplicating Descriptors:: Fcntl commands for duplicating
54 * Descriptor Flags:: Fcntl commands for manipulating
55 flags associated with file
57 * File Status Flags:: Fcntl commands for manipulating
58 flags associated with open files.
59 * File Locks:: Fcntl commands for implementing
61 * Interrupt Input:: Getting an asynchronous signal when
63 * IOCTLs:: Generic I/O Control operations.
67 @node Opening and Closing Files
68 @section Opening and Closing Files
70 @cindex opening a file descriptor
71 @cindex closing a file descriptor
72 This section describes the primitives for opening and closing files
73 using file descriptors. The @code{open} and @code{creat} functions are
74 declared in the header file @file{fcntl.h}, while @code{close} is
75 declared in @file{unistd.h}.
81 @deftypefun int open (const char *@var{filename}, int @var{flags}[, mode_t @var{mode}])
82 The @code{open} function creates and returns a new file descriptor
83 for the file named by @var{filename}. Initially, the file position
84 indicator for the file is at the beginning of the file. The argument
85 @var{mode} is used only when a file is created, but it doesn't hurt
86 to supply the argument in any case.
88 The @var{flags} argument controls how the file is to be opened. This is
89 a bit mask; you create the value by the bitwise OR of the appropriate
90 parameters (using the @samp{|} operator in C).
91 @xref{File Status Flags}, for the parameters available.
93 The normal return value from @code{open} is a non-negative integer file
94 descriptor. In the case of an error, a value of @math{-1} is returned
95 instead. In addition to the usual file name errors (@pxref{File
96 Name Errors}), the following @code{errno} error conditions are defined
101 The file exists but is not readable/writable as requested by the @var{flags}
102 argument, the file does not exist and the directory is unwritable so
103 it cannot be created.
106 Both @code{O_CREAT} and @code{O_EXCL} are set, and the named file already
110 The @code{open} operation was interrupted by a signal.
111 @xref{Interrupted Primitives}.
114 The @var{flags} argument specified write access, and the file is a directory.
117 The process has too many files open.
118 The maximum number of file descriptors is controlled by the
119 @code{RLIMIT_NOFILE} resource limit; @pxref{Limits on Resources}.
122 The entire system, or perhaps the file system which contains the
123 directory, cannot support any additional open files at the moment.
124 (This problem cannot happen on the GNU system.)
127 The named file does not exist, and @code{O_CREAT} is not specified.
130 The directory or file system that would contain the new file cannot be
131 extended, because there is no disk space left.
134 @code{O_NONBLOCK} and @code{O_WRONLY} are both set in the @var{flags}
135 argument, the file named by @var{filename} is a FIFO (@pxref{Pipes and
136 FIFOs}), and no process has the file open for reading.
139 The file resides on a read-only file system and any of @w{@code{O_WRONLY}},
140 @code{O_RDWR}, and @code{O_TRUNC} are set in the @var{flags} argument,
141 or @code{O_CREAT} is set and the file does not already exist.
146 If on a 32 bits machine the sources are translated with
147 @code{_FILE_OFFSET_BITS == 64} the function @code{open} returns a file
148 descriptor opened in the large file mode which enables the file handling
149 functions to use files up to @math{2^63} bytes in size and offset from
150 @math{-2^63} to @math{2^63}. This happens transparently for the user
151 since all of the lowlevel file handling functions are equally replaced.
153 This function is a cancelation point in multi-threaded programs. This
154 is a problem if the thread allocates some resources (like memory, file
155 descriptors, semaphores or whatever) at the time @code{open} is
156 called. If the thread gets canceled these resources stay allocated
157 until the program ends. To avoid this calls to @code{open} should be
158 protected using cancelation handlers.
159 @c ref pthread_cleanup_push / pthread_cleanup_pop
161 The @code{open} function is the underlying primitive for the @code{fopen}
162 and @code{freopen} functions, that create streams.
167 @deftypefun int open64 (const char *@var{filename}, int @var{flags}[, mode_t @var{mode}])
168 This function is similar to @code{open}. It returns a file descriptor
169 which can be used to access the file named by @var{filename}. The only
170 the difference is that on 32 bits systems the file is opened in the
171 large file mode. I.e., file length and file offsets can exceed 31 bits.
173 When the sources are translated with @code{_FILE_OFFSET_BITS == 64} this
174 function is actually available under the name @code{open}. I.e., the
175 new, extended API using 64 bit file sizes and offsets transparently
176 replaces the old API.
181 @deftypefn {Obsolete function} int creat (const char *@var{filename}, mode_t @var{mode})
182 This function is obsolete. The call:
185 creat (@var{filename}, @var{mode})
192 open (@var{filename}, O_WRONLY | O_CREAT | O_TRUNC, @var{mode})
195 If on a 32 bits machine the sources are translated with
196 @code{_FILE_OFFSET_BITS == 64} the function @code{creat} returns a file
197 descriptor opened in the large file mode which enables the file handling
198 functions to use files up to @math{2^63} in size and offset from
199 @math{-2^63} to @math{2^63}. This happens transparently for the user
200 since all of the lowlevel file handling functions are equally replaced.
205 @deftypefn {Obsolete function} int creat64 (const char *@var{filename}, mode_t @var{mode})
206 This function is similar to @code{creat}. It returns a file descriptor
207 which can be used to access the file named by @var{filename}. The only
208 the difference is that on 32 bits systems the file is opened in the
209 large file mode. I.e., file length and file offsets can exceed 31 bits.
211 To use this file descriptor one must not use the normal operations but
212 instead the counterparts named @code{*64}, e.g., @code{read64}.
214 When the sources are translated with @code{_FILE_OFFSET_BITS == 64} this
215 function is actually available under the name @code{open}. I.e., the
216 new, extended API using 64 bit file sizes and offsets transparently
217 replaces the old API.
222 @deftypefun int close (int @var{filedes})
223 The function @code{close} closes the file descriptor @var{filedes}.
224 Closing a file has the following consequences:
228 The file descriptor is deallocated.
231 Any record locks owned by the process on the file are unlocked.
234 When all file descriptors associated with a pipe or FIFO have been closed,
235 any unread data is discarded.
238 This function is a cancelation point in multi-threaded programs. This
239 is a problem if the thread allocates some resources (like memory, file
240 descriptors, semaphores or whatever) at the time @code{close} is
241 called. If the thread gets canceled these resources stay allocated
242 until the program ends. To avoid this calls to @code{close} should be
243 protected using cancelation handlers.
244 @c ref pthread_cleanup_push / pthread_cleanup_pop
246 The normal return value from @code{close} is @math{0}; a value of @math{-1}
247 is returned in case of failure. The following @code{errno} error
248 conditions are defined for this function:
252 The @var{filedes} argument is not a valid file descriptor.
255 The @code{close} call was interrupted by a signal.
256 @xref{Interrupted Primitives}.
257 Here is an example of how to handle @code{EINTR} properly:
260 TEMP_FAILURE_RETRY (close (desc));
266 When the file is accessed by NFS, these errors from @code{write} can sometimes
267 not be detected until @code{close}. @xref{I/O Primitives}, for details
271 Please note that there is @emph{no} separate @code{close64} function.
272 This is not necessary since this function does not determine nor depend
273 on the mode of the file. The kernel which performs the @code{close}
274 operation knows for which mode the descriptor is used and can handle
278 To close a stream, call @code{fclose} (@pxref{Closing Streams}) instead
279 of trying to close its underlying file descriptor with @code{close}.
280 This flushes any buffered output and updates the stream object to
281 indicate that it is closed.
284 @section Input and Output Primitives
286 This section describes the functions for performing primitive input and
287 output operations on file descriptors: @code{read}, @code{write}, and
288 @code{lseek}. These functions are declared in the header file
294 @deftp {Data Type} ssize_t
295 This data type is used to represent the sizes of blocks that can be
296 read or written in a single operation. It is similar to @code{size_t},
297 but must be a signed type.
300 @cindex reading from a file descriptor
303 @deftypefun ssize_t read (int @var{filedes}, void *@var{buffer}, size_t @var{size})
304 The @code{read} function reads up to @var{size} bytes from the file
305 with descriptor @var{filedes}, storing the results in the @var{buffer}.
306 (This is not necessarily a character string and there is no terminating
307 null character added.)
309 @cindex end-of-file, on a file descriptor
310 The return value is the number of bytes actually read. This might be
311 less than @var{size}; for example, if there aren't that many bytes left
312 in the file or if there aren't that many bytes immediately available.
313 The exact behavior depends on what kind of file it is. Note that
314 reading less than @var{size} bytes is not an error.
316 A value of zero indicates end-of-file (except if the value of the
317 @var{size} argument is also zero). This is not considered an error.
318 If you keep calling @code{read} while at end-of-file, it will keep
319 returning zero and doing nothing else.
321 If @code{read} returns at least one character, there is no way you can
322 tell whether end-of-file was reached. But if you did reach the end, the
323 next read will return zero.
325 In case of an error, @code{read} returns @math{-1}. The following
326 @code{errno} error conditions are defined for this function:
330 Normally, when no input is immediately available, @code{read} waits for
331 some input. But if the @code{O_NONBLOCK} flag is set for the file
332 (@pxref{File Status Flags}), @code{read} returns immediately without
333 reading any data, and reports this error.
335 @strong{Compatibility Note:} Most versions of BSD Unix use a different
336 error code for this: @code{EWOULDBLOCK}. In the GNU library,
337 @code{EWOULDBLOCK} is an alias for @code{EAGAIN}, so it doesn't matter
340 On some systems, reading a large amount of data from a character special
341 file can also fail with @code{EAGAIN} if the kernel cannot find enough
342 physical memory to lock down the user's pages. This is limited to
343 devices that transfer with direct memory access into the user's memory,
344 which means it does not include terminals, since they always use
345 separate buffers inside the kernel. This problem never happens in the
348 Any condition that could result in @code{EAGAIN} can instead result in a
349 successful @code{read} which returns fewer bytes than requested.
350 Calling @code{read} again immediately would result in @code{EAGAIN}.
353 The @var{filedes} argument is not a valid file descriptor,
354 or is not open for reading.
357 @code{read} was interrupted by a signal while it was waiting for input.
358 @xref{Interrupted Primitives}. A signal will not necessary cause
359 @code{read} to return @code{EINTR}; it may instead result in a
360 successful @code{read} which returns fewer bytes than requested.
363 For many devices, and for disk files, this error code indicates
366 @code{EIO} also occurs when a background process tries to read from the
367 controlling terminal, and the normal action of stopping the process by
368 sending it a @code{SIGTTIN} signal isn't working. This might happen if
369 signal is being blocked or ignored, or because the process group is
370 orphaned. @xref{Job Control}, for more information about job control,
371 and @ref{Signal Handling}, for information about signals.
374 Please note that there is no function named @code{read64}. This is not
375 necessary since this function does not directly modify or handle the
376 possibly wide file offset. Since the kernel handles this state
377 internally the @code{read} function can be used for all cases.
379 This function is a cancelation point in multi-threaded programs. This
380 is a problem if the thread allocates some resources (like memory, file
381 descriptors, semaphores or whatever) at the time @code{read} is
382 called. If the thread gets canceled these resources stay allocated
383 until the program ends. To avoid this calls to @code{read} should be
384 protected using cancelation handlers.
385 @c ref pthread_cleanup_push / pthread_cleanup_pop
387 The @code{read} function is the underlying primitive for all of the
388 functions that read from streams, such as @code{fgetc}.
393 @deftypefun ssize_t pread (int @var{filedes}, void *@var{buffer}, size_t @var{size}, off_t @var{offset})
394 The @code{pread} function is similar to the @code{read} function. The
395 first three arguments are identical and also the return values and error
398 The difference is the fourth argument and its handling. The data block
399 is not read from the current position of the file descriptor
400 @code{filedes}. Instead the data is read from the file starting at
401 position @var{offset}. The position of the file descriptor itself is
402 not effected by the operation. The value is the same as before the call.
404 When the source file is compiled with @code{_FILE_OFFSET_BITS == 64} the
405 @code{pread} function is in fact @code{pread64} and the type
406 @code{off_t} has 64 bits which makes it possible to handle files up to
407 @math{2^63} bytes in length.
409 The return value of @code{pread} describes the number of bytes read.
410 In the error case it returns @math{-1} like @code{read} does and the
411 error codes are also the same. Only there are a few more error codes:
414 The value given for @var{offset} is negative and therefore illegal.
417 The file descriptor @var{filedes} is associate with a pipe or a FIFO and
418 this device does not allow positioning of the file pointer.
421 The function is an extension defined in the Unix Single Specification
427 @deftypefun ssize_t pread64 (int @var{filedes}, void *@var{buffer}, size_t @var{size}, off64_t @var{offset})
428 This function is similar to the @code{pread} function. The difference
429 is that the @var{offset} parameter is of type @code{off64_t} instead of
430 @code{off_t} which makes it possible on 32 bits machines to address
431 files larger than @math{2^31} bytes and up to @math{2^63} bytes. The
432 file descriptor @code{filedes} must be opened using @code{open64} since
433 otherwise the large offsets possible with @code{off64_t} will lead to
434 errors with a descriptor in small file mode.
436 When the source file is compiled with @code{_FILE_OFFSET_BITS == 64} on a
437 32 bits machine this function is actually available under the name
438 @code{pread} and so transparently replaces the 32 bits interface.
441 @cindex writing to a file descriptor
444 @deftypefun ssize_t write (int @var{filedes}, const void *@var{buffer}, size_t @var{size})
445 The @code{write} function writes up to @var{size} bytes from
446 @var{buffer} to the file with descriptor @var{filedes}. The data in
447 @var{buffer} is not necessarily a character string and a null character is
448 output like any other character.
450 The return value is the number of bytes actually written. This may be
451 @var{size}, but can always be smaller. Your program should always call
452 @code{write} in a loop, iterating until all the data is written.
454 Once @code{write} returns, the data is enqueued to be written and can be
455 read back right away, but it is not necessarily written out to permanent
456 storage immediately. You can use @code{fsync} when you need to be sure
457 your data has been permanently stored before continuing. (It is more
458 efficient for the system to batch up consecutive writes and do them all
459 at once when convenient. Normally they will always be written to disk
460 within a minute or less.) Modern systems provide another function
461 @code{fdatasync} which guarantees integrity only for the file data and
463 @c !!! xref fsync, fdatasync
464 You can use the @code{O_FSYNC} open mode to make @code{write} always
465 store the data to disk before returning; @pxref{Operating Modes}.
467 In the case of an error, @code{write} returns @math{-1}. The following
468 @code{errno} error conditions are defined for this function:
472 Normally, @code{write} blocks until the write operation is complete.
473 But if the @code{O_NONBLOCK} flag is set for the file (@pxref{Control
474 Operations}), it returns immediately without writing any data, and
475 reports this error. An example of a situation that might cause the
476 process to block on output is writing to a terminal device that supports
477 flow control, where output has been suspended by receipt of a STOP
480 @strong{Compatibility Note:} Most versions of BSD Unix use a different
481 error code for this: @code{EWOULDBLOCK}. In the GNU library,
482 @code{EWOULDBLOCK} is an alias for @code{EAGAIN}, so it doesn't matter
485 On some systems, writing a large amount of data from a character special
486 file can also fail with @code{EAGAIN} if the kernel cannot find enough
487 physical memory to lock down the user's pages. This is limited to
488 devices that transfer with direct memory access into the user's memory,
489 which means it does not include terminals, since they always use
490 separate buffers inside the kernel. This problem does not arise in the
494 The @var{filedes} argument is not a valid file descriptor,
495 or is not open for writing.
498 The size of the file would become larger than the implementation can support.
501 The @code{write} operation was interrupted by a signal while it was
502 blocked waiting for completion. A signal will not necessary cause
503 @code{write} to return @code{EINTR}; it may instead result in a
504 successful @code{write} which writes fewer bytes than requested.
505 @xref{Interrupted Primitives}.
508 For many devices, and for disk files, this error code indicates
512 The device containing the file is full.
515 This error is returned when you try to write to a pipe or FIFO that
516 isn't open for reading by any process. When this happens, a @code{SIGPIPE}
517 signal is also sent to the process; see @ref{Signal Handling}.
520 Unless you have arranged to prevent @code{EINTR} failures, you should
521 check @code{errno} after each failing call to @code{write}, and if the
522 error was @code{EINTR}, you should simply repeat the call.
523 @xref{Interrupted Primitives}. The easy way to do this is with the
524 macro @code{TEMP_FAILURE_RETRY}, as follows:
527 nbytes = TEMP_FAILURE_RETRY (write (desc, buffer, count));
530 Please note that there is no function named @code{write64}. This is not
531 necessary since this function does not directly modify or handle the
532 possibly wide file offset. Since the kernel handles this state
533 internally the @code{write} function can be used for all cases.
535 This function is a cancelation point in multi-threaded programs. This
536 is a problem if the thread allocates some resources (like memory, file
537 descriptors, semaphores or whatever) at the time @code{write} is
538 called. If the thread gets canceled these resources stay allocated
539 until the program ends. To avoid this calls to @code{write} should be
540 protected using cancelation handlers.
541 @c ref pthread_cleanup_push / pthread_cleanup_pop
543 The @code{write} function is the underlying primitive for all of the
544 functions that write to streams, such as @code{fputc}.
549 @deftypefun ssize_t pwrite (int @var{filedes}, const void *@var{buffer}, size_t @var{size}, off_t @var{offset})
550 The @code{pwrite} function is similar to the @code{write} function. The
551 first three arguments are identical and also the return values and error
554 The difference is the fourth argument and its handling. The data block
555 is not written to the current position of the file descriptor
556 @code{filedes}. Instead the data is written to the file starting at
557 position @var{offset}. The position of the file descriptor itself is
558 not effected by the operation. The value is the same as before the call.
560 When the source file is compiled with @code{_FILE_OFFSET_BITS == 64} the
561 @code{pwrite} function is in fact @code{pwrite64} and the type
562 @code{off_t} has 64 bits which makes it possible to handle files up to
563 @math{2^63} bytes in length.
565 The return value of @code{pwrite} describes the number of written bytes.
566 In the error case it returns @math{-1} like @code{write} does and the
567 error codes are also the same. Only there are a few more error codes:
570 The value given for @var{offset} is negative and therefore illegal.
573 The file descriptor @var{filedes} is associate with a pipe or a FIFO and
574 this device does not allow positioning of the file pointer.
577 The function is an extension defined in the Unix Single Specification
583 @deftypefun ssize_t pwrite64 (int @var{filedes}, const void *@var{buffer}, size_t @var{size}, off64_t @var{offset})
584 This function is similar to the @code{pwrite} function. The difference
585 is that the @var{offset} parameter is of type @code{off64_t} instead of
586 @code{off_t} which makes it possible on 32 bits machines to address
587 files larger than @math{2^31} bytes and up to @math{2^63} bytes. The
588 file descriptor @code{filedes} must be opened using @code{open64} since
589 otherwise the large offsets possible with @code{off64_t} will lead to
590 errors with a descriptor in small file mode.
592 When the source file is compiled using @code{_FILE_OFFSET_BITS == 64} on a
593 32 bits machine this function is actually available under the name
594 @code{pwrite} and so transparently replaces the 32 bits interface.
598 @node File Position Primitive
599 @section Setting the File Position of a Descriptor
601 Just as you can set the file position of a stream with @code{fseek}, you
602 can set the file position of a descriptor with @code{lseek}. This
603 specifies the position in the file for the next @code{read} or
604 @code{write} operation. @xref{File Positioning}, for more information
605 on the file position and what it means.
607 To read the current file position value from a descriptor, use
608 @code{lseek (@var{desc}, 0, SEEK_CUR)}.
610 @cindex file positioning on a file descriptor
611 @cindex positioning a file descriptor
612 @cindex seeking on a file descriptor
615 @deftypefun off_t lseek (int @var{filedes}, off_t @var{offset}, int @var{whence})
616 The @code{lseek} function is used to change the file position of the
617 file with descriptor @var{filedes}.
619 The @var{whence} argument specifies how the @var{offset} should be
620 interpreted in the same way as for the @code{fseek} function, and must be
621 one of the symbolic constants @code{SEEK_SET}, @code{SEEK_CUR}, or
626 Specifies that @var{whence} is a count of characters from the beginning
630 Specifies that @var{whence} is a count of characters from the current
631 file position. This count may be positive or negative.
634 Specifies that @var{whence} is a count of characters from the end of
635 the file. A negative count specifies a position within the current
636 extent of the file; a positive count specifies a position past the
637 current end. If you set the position past the current end, and
638 actually write data, you will extend the file with zeros up to that
642 The return value from @code{lseek} is normally the resulting file
643 position, measured in bytes from the beginning of the file.
644 You can use this feature together with @code{SEEK_CUR} to read the
645 current file position.
647 If you want to append to the file, setting the file position to the
648 current end of file with @code{SEEK_END} is not sufficient. Another
649 process may write more data after you seek but before you write,
650 extending the file so the position you write onto clobbers their data.
651 Instead, use the @code{O_APPEND} operating mode; @pxref{Operating Modes}.
653 You can set the file position past the current end of the file. This
654 does not by itself make the file longer; @code{lseek} never changes the
655 file. But subsequent output at that position will extend the file.
656 Characters between the previous end of file and the new position are
657 filled with zeros. Extending the file in this way can create a
658 ``hole'': the blocks of zeros are not actually allocated on disk, so the
659 file takes up less space than it appears so; it is then called a
662 @cindex holes in files
664 If the file position cannot be changed, or the operation is in some way
665 invalid, @code{lseek} returns a value of @math{-1}. The following
666 @code{errno} error conditions are defined for this function:
670 The @var{filedes} is not a valid file descriptor.
673 The @var{whence} argument value is not valid, or the resulting
674 file offset is not valid. A file offset is invalid.
677 The @var{filedes} corresponds to an object that cannot be positioned,
678 such as a pipe, FIFO or terminal device. (POSIX.1 specifies this error
679 only for pipes and FIFOs, but in the GNU system, you always get
680 @code{ESPIPE} if the object is not seekable.)
683 When the source file is compiled with @code{_FILE_OFFSET_BITS == 64} the
684 @code{lseek} function is in fact @code{lseek64} and the type
685 @code{off_t} has 64 bits which makes it possible to handle files up to
686 @math{2^63} bytes in length.
688 This function is a cancelation point in multi-threaded programs. This
689 is a problem if the thread allocates some resources (like memory, file
690 descriptors, semaphores or whatever) at the time @code{lseek} is
691 called. If the thread gets canceled these resources stay allocated
692 until the program ends. To avoid this calls to @code{lseek} should be
693 protected using cancelation handlers.
694 @c ref pthread_cleanup_push / pthread_cleanup_pop
696 The @code{lseek} function is the underlying primitive for the
697 @code{fseek}, @code{fseeko}, @code{ftell}, @code{ftello} and
698 @code{rewind} functions, which operate on streams instead of file
704 @deftypefun off64_t lseek64 (int @var{filedes}, off64_t @var{offset}, int @var{whence})
705 This function is similar to the @code{lseek} function. The difference
706 is that the @var{offset} parameter is of type @code{off64_t} instead of
707 @code{off_t} which makes it possible on 32 bits machines to address
708 files larger than @math{2^31} bytes and up to @math{2^63} bytes. The
709 file descriptor @code{filedes} must be opened using @code{open64} since
710 otherwise the large offsets possible with @code{off64_t} will lead to
711 errors with a descriptor in small file mode.
713 When the source file is compiled with @code{_FILE_OFFSET_BITS == 64} on a
714 32 bits machine this function is actually available under the name
715 @code{lseek} and so transparently replaces the 32 bits interface.
718 You can have multiple descriptors for the same file if you open the file
719 more than once, or if you duplicate a descriptor with @code{dup}.
720 Descriptors that come from separate calls to @code{open} have independent
721 file positions; using @code{lseek} on one descriptor has no effect on the
729 d1 = open ("foo", O_RDONLY);
730 d2 = open ("foo", O_RDONLY);
731 lseek (d1, 1024, SEEK_SET);
738 will read the first four characters of the file @file{foo}. (The
739 error-checking code necessary for a real program has been omitted here
742 By contrast, descriptors made by duplication share a common file
743 position with the original descriptor that was duplicated. Anything
744 which alters the file position of one of the duplicates, including
745 reading or writing data, affects all of them alike. Thus, for example,
750 char buf1[4], buf2[4];
751 d1 = open ("foo", O_RDONLY);
754 lseek (d3, 1024, SEEK_SET);
761 will read four characters starting with the 1024'th character of
762 @file{foo}, and then four more characters starting with the 1028'th
767 @deftp {Data Type} off_t
768 This is an arithmetic data type used to represent file sizes.
769 In the GNU system, this is equivalent to @code{fpos_t} or @code{long int}.
771 If the source is compiled with @code{_FILE_OFFSET_BITS == 64} this type
772 is transparently replaced by @code{off64_t}.
777 @deftp {Data Type} off64_t
778 This type is used similar to @code{off_t}. The difference is that even
779 on 32 bits machines, where the @code{off_t} type would have 32 bits,
780 @code{off64_t} has 64 bits and so is able to address files up to
781 @math{2^63} bytes in length.
783 When compiling with @code{_FILE_OFFSET_BITS == 64} this type is
784 available under the name @code{off_t}.
787 These aliases for the @samp{SEEK_@dots{}} constants exist for the sake
788 of compatibility with older BSD systems. They are defined in two
789 different header files: @file{fcntl.h} and @file{sys/file.h}.
793 An alias for @code{SEEK_SET}.
796 An alias for @code{SEEK_CUR}.
799 An alias for @code{SEEK_END}.
802 @node Descriptors and Streams
803 @section Descriptors and Streams
804 @cindex streams, and file descriptors
805 @cindex converting file descriptor to stream
806 @cindex extracting file descriptor from stream
808 Given an open file descriptor, you can create a stream for it with the
809 @code{fdopen} function. You can get the underlying file descriptor for
810 an existing stream with the @code{fileno} function. These functions are
811 declared in the header file @file{stdio.h}.
816 @deftypefun {FILE *} fdopen (int @var{filedes}, const char *@var{opentype})
817 The @code{fdopen} function returns a new stream for the file descriptor
820 The @var{opentype} argument is interpreted in the same way as for the
821 @code{fopen} function (@pxref{Opening Streams}), except that
822 the @samp{b} option is not permitted; this is because GNU makes no
823 distinction between text and binary files. Also, @code{"w"} and
824 @code{"w+"} do not cause truncation of the file; these have affect only
825 when opening a file, and in this case the file has already been opened.
826 You must make sure that the @var{opentype} argument matches the actual
827 mode of the open file descriptor.
829 The return value is the new stream. If the stream cannot be created
830 (for example, if the modes for the file indicated by the file descriptor
831 do not permit the access specified by the @var{opentype} argument), a
832 null pointer is returned instead.
834 In some other systems, @code{fdopen} may fail to detect that the modes
835 for file descriptor do not permit the access specified by
836 @code{opentype}. The GNU C library always checks for this.
839 For an example showing the use of the @code{fdopen} function,
840 see @ref{Creating a Pipe}.
844 @deftypefun int fileno (FILE *@var{stream})
845 This function returns the file descriptor associated with the stream
846 @var{stream}. If an error is detected (for example, if the @var{stream}
847 is not valid) or if @var{stream} does not do I/O to a file,
848 @code{fileno} returns @math{-1}.
851 @cindex standard file descriptors
852 @cindex file descriptors, standard
853 There are also symbolic constants defined in @file{unistd.h} for the
854 file descriptors belonging to the standard streams @code{stdin},
855 @code{stdout}, and @code{stderr}; see @ref{Standard Streams}.
863 This macro has value @code{0}, which is the file descriptor for
865 @cindex standard input file descriptor
870 @vindex STDOUT_FILENO
871 This macro has value @code{1}, which is the file descriptor for
873 @cindex standard output file descriptor
878 @vindex STDERR_FILENO
879 This macro has value @code{2}, which is the file descriptor for
880 standard error output.
882 @cindex standard error file descriptor
884 @node Stream/Descriptor Precautions
885 @section Dangers of Mixing Streams and Descriptors
887 @cindex streams and descriptors
888 @cindex descriptors and streams
889 @cindex mixing descriptors and streams
891 You can have multiple file descriptors and streams (let's call both
892 streams and descriptors ``channels'' for short) connected to the same
893 file, but you must take care to avoid confusion between channels. There
894 are two cases to consider: @dfn{linked} channels that share a single
895 file position value, and @dfn{independent} channels that have their own
898 It's best to use just one channel in your program for actual data
899 transfer to any given file, except when all the access is for input.
900 For example, if you open a pipe (something you can only do at the file
901 descriptor level), either do all I/O with the descriptor, or construct a
902 stream from the descriptor with @code{fdopen} and then do all I/O with
906 * Linked Channels:: Dealing with channels sharing a file position.
907 * Independent Channels:: Dealing with separately opened, unlinked channels.
908 * Cleaning Streams:: Cleaning a stream makes it safe to use
912 @node Linked Channels
913 @subsection Linked Channels
914 @cindex linked channels
916 Channels that come from a single opening share the same file position;
917 we call them @dfn{linked} channels. Linked channels result when you
918 make a stream from a descriptor using @code{fdopen}, when you get a
919 descriptor from a stream with @code{fileno}, when you copy a descriptor
920 with @code{dup} or @code{dup2}, and when descriptors are inherited
921 during @code{fork}. For files that don't support random access, such as
922 terminals and pipes, @emph{all} channels are effectively linked. On
923 random-access files, all append-type output streams are effectively
924 linked to each other.
926 @cindex cleaning up a stream
927 If you have been using a stream for I/O, and you want to do I/O using
928 another channel (either a stream or a descriptor) that is linked to it,
929 you must first @dfn{clean up} the stream that you have been using.
930 @xref{Cleaning Streams}.
932 Terminating a process, or executing a new program in the process,
933 destroys all the streams in the process. If descriptors linked to these
934 streams persist in other processes, their file positions become
935 undefined as a result. To prevent this, you must clean up the streams
936 before destroying them.
938 @node Independent Channels
939 @subsection Independent Channels
940 @cindex independent channels
942 When you open channels (streams or descriptors) separately on a seekable
943 file, each channel has its own file position. These are called
944 @dfn{independent channels}.
946 The system handles each channel independently. Most of the time, this
947 is quite predictable and natural (especially for input): each channel
948 can read or write sequentially at its own place in the file. However,
949 if some of the channels are streams, you must take these precautions:
953 You should clean an output stream after use, before doing anything else
954 that might read or write from the same part of the file.
957 You should clean an input stream before reading data that may have been
958 modified using an independent channel. Otherwise, you might read
959 obsolete data that had been in the stream's buffer.
962 If you do output to one channel at the end of the file, this will
963 certainly leave the other independent channels positioned somewhere
964 before the new end. You cannot reliably set their file positions to the
965 new end of file before writing, because the file can always be extended
966 by another process between when you set the file position and when you
967 write the data. Instead, use an append-type descriptor or stream; they
968 always output at the current end of the file. In order to make the
969 end-of-file position accurate, you must clean the output channel you
970 were using, if it is a stream.
972 It's impossible for two channels to have separate file pointers for a
973 file that doesn't support random access. Thus, channels for reading or
974 writing such files are always linked, never independent. Append-type
975 channels are also always linked. For these channels, follow the rules
976 for linked channels; see @ref{Linked Channels}.
978 @node Cleaning Streams
979 @subsection Cleaning Streams
981 On the GNU system, you can clean up any stream with @code{fclean}:
985 @deftypefun int fclean (FILE *@var{stream})
986 Clean up the stream @var{stream} so that its buffer is empty. If
987 @var{stream} is doing output, force it out. If @var{stream} is doing
988 input, give the data in the buffer back to the system, arranging to
992 On other systems, you can use @code{fflush} to clean a stream in most
995 You can skip the @code{fclean} or @code{fflush} if you know the stream
996 is already clean. A stream is clean whenever its buffer is empty. For
997 example, an unbuffered stream is always clean. An input stream that is
998 at end-of-file is clean. A line-buffered stream is clean when the last
999 character output was a newline.
1001 There is one case in which cleaning a stream is impossible on most
1002 systems. This is when the stream is doing input from a file that is not
1003 random-access. Such streams typically read ahead, and when the file is
1004 not random access, there is no way to give back the excess data already
1005 read. When an input stream reads from a random-access file,
1006 @code{fflush} does clean the stream, but leaves the file pointer at an
1007 unpredictable place; you must set the file pointer before doing any
1008 further I/O. On the GNU system, using @code{fclean} avoids both of
1011 Closing an output-only stream also does @code{fflush}, so this is a
1012 valid way of cleaning an output stream. On the GNU system, closing an
1013 input stream does @code{fclean}.
1015 You need not clean a stream before using its descriptor for control
1016 operations such as setting terminal modes; these operations don't affect
1017 the file position and are not affected by it. You can use any
1018 descriptor for these operations, and all channels are affected
1019 simultaneously. However, text already ``output'' to a stream but still
1020 buffered by the stream will be subject to the new terminal modes when
1021 subsequently flushed. To make sure ``past'' output is covered by the
1022 terminal settings that were in effect at the time, flush the output
1023 streams for that terminal before setting the modes. @xref{Terminal
1026 @node Scatter-Gather
1027 @section Fast Scatter-Gather I/O
1028 @cindex scatter-gather
1030 Some applications may need to read or write data to multiple buffers,
1031 which are seperated in memory. Although this can be done easily enough
1032 with multiple calls to @code{read} and @code{write}, it is inefficent
1033 because there is overhead associated with each kernel call.
1035 Instead, many platforms provide special high-speed primitives to perform
1036 these @dfn{scatter-gather} operations in a single kernel call. The GNU C
1037 library will provide an emulation on any system that lacks these
1038 primitives, so they are not a portability threat. They are defined in
1041 These functions are controlled with arrays of @code{iovec} structures,
1042 which describe the location and size of each buffer.
1044 @deftp {Data Type} {struct iovec}
1046 The @code{iovec} structure describes a buffer. It contains two fields:
1050 @item void *iov_base
1051 Contains the address of a buffer.
1053 @item size_t iov_len
1054 Contains the length of the buffer.
1059 @deftypefun ssize_t readv (int @var{filedes}, const struct iovec *@var{vector}, int @var{count})
1061 The @code{readv} function reads data from @var{filedes} and scatters it
1062 into the buffers described in @var{vector}, which is taken to be
1063 @var{count} structures long. As each buffer is filled, data is sent to the
1066 Note that @code{readv} is not guaranteed to fill all the buffers.
1067 It may stop at any point, for the same reasons @code{read} would.
1069 The return value is a count of bytes (@emph{not} buffers) read, @math{0}
1070 indicating end-of-file, or @math{-1} indicating an error. The possible
1071 errors are the same as in @code{read}.
1075 @deftypefun ssize_t writev (int @var{filedes}, const struct iovec *@var{vector}, int @var{count})
1077 The @code{writev} function gathers data from the buffers described in
1078 @var{vector}, which is taken to be @var{count} structures long, and writes
1079 them to @code{filedes}. As each buffer is written, it moves on to the
1082 Like @code{readv}, @code{writev} may stop midstream under the same
1083 conditions @code{write} would.
1085 The return value is a count of bytes written, or @math{-1} indicating an
1086 error. The possible errors are the same as in @code{write}.
1090 @c Note - I haven't read this anywhere. I surmised it from my knowledge
1091 @c of computer science. Thus, there could be subtleties I'm missing.
1093 Note that if the buffers are small (under about 1kB), high-level streams
1094 may be easier to use than these functions. However, @code{readv} and
1095 @code{writev} are more efficient when the individual buffers themselves
1096 (as opposed to the total output), are large. In that case, a high-level
1097 stream would not be able to cache the data effectively.
1099 @node Memory-mapped I/O
1100 @section Memory-mapped I/O
1102 On modern operating systems, it is possible to @dfn{mmap} (pronounced
1103 ``em-map'') a file to a region of memory. When this is done, the file can
1104 be accessed just like an array in the program.
1106 This is more efficent than @code{read} or @code{write}, as only regions
1107 of the file a program actually accesses are loaded. Accesses to
1108 not-yet-loaded parts of the mmapped region are handled in the same way as
1111 Since mmapped pages can be stored back to their file when physical memory
1112 is low, it is possible to mmap files orders of magnitude larger than both
1113 the physical memory @emph{and} swap space. The only limit is address
1114 space. The theoretical limit is 4GB on a 32-bit machine - however, the
1115 actual limit will be smaller since some areas will be reserved for other
1118 Memory mapping only works on entire pages of memory. Thus, addresses
1119 for mapping must be page-aligned, and length values will be rounded up.
1120 To determine the size of a page the machine uses one should use
1123 size_t page_size = (size_t) sysconf (_SC_PAGESIZE);
1126 These functions are declared in @file{sys/mman.h}.
1128 @deftypefun {void *} mmap (void *@var{address}, size_t @var{length},int @var{protect}, int @var{flags}, int @var{filedes}, off_t @var{offset})
1130 The @code{mmap} function creates a new mapping, connected to bytes
1131 (@var{offset}) to (@var{offset} + @var{length}) in the file open on
1134 @var{address} gives a preferred starting address for the mapping.
1135 @code{NULL} expresses no preference. Any previous mapping at that
1136 address is automatically removed. The address you give may still be
1137 changed, unless you use the @code{MAP_FIXED} flag.
1142 @var{protect} contains flags that control what kind of access is
1143 permitted. They include @code{PROT_READ}, @code{PROT_WRITE}, and
1144 @code{PROT_EXEC}, which permit reading, writing, and execution,
1145 respectively. Inappropriate access will cause a segfault (@pxref{Program
1148 Note that most hardware designs cannot support write permission without
1149 read permission, and many do not distinguish read and execute permission.
1150 Thus, you may recieve wider permissions than you ask for, and mappings of
1151 write-only files may be denied even if you do not use @code{PROT_READ}.
1153 @var{flags} contains flags that control the nature of the map.
1154 One of @code{MAP_SHARED} or @code{MAP_PRIVATE} must be specified.
1160 This specifies that writes to the region should never be written back
1161 to the attached file. Instead, a copy is made for the process, and the
1162 region will be swapped normally if memory runs low. No other process will
1165 Since private mappings effectively revert to ordinary memory
1166 when written to, you must have enough virtual memory for a copy of
1167 the entire mmapped region if you use this mode with @code{PROT_WRITE}.
1170 This specifies that writes to the region will be written back to the
1171 file. Changes made will be shared immediately with other processes
1172 mmaping the same file.
1174 Note that actual writing may take place at any time. You need to use
1175 @code{msync}, described below, if it is important that other processes
1176 using conventional I/O get a consistent view of the file.
1179 This forces the system to use the exact mapping address specified in
1180 @var{address} and fail if it can't.
1182 @c One of these is official - the other is obviously an obsolete synonym
1186 This flag tells the system to create an anonymous mapping, not connected
1187 to a file. @var{filedes} and @var{off} are ignored, and the region is
1188 initialized with zeros.
1190 Anonymous maps are used as the basic primitive to extend the heap on some
1191 systems. They are also useful to share data between multiple tasks
1192 without creating a file.
1194 On some systems using private anonymous mmaps is more efficent than using
1195 @code{malloc} for large blocks. This is not an issue with the GNU C library,
1196 as the included @code{malloc} automatically uses @code{mmap} where appropriate.
1198 @c Linux has some other MAP_ options, which I have not discussed here.
1199 @c MAP_DENYWRITE, MAP_EXECUTABLE and MAP_GROWSDOWN don't seem applicable to
1200 @c user programs (and I don't understand the last two). MAP_LOCKED does
1201 @c not appear to be implemented.
1205 @code{mmap} returns the address of the new mapping, or @math{-1} for an
1208 Possible errors include:
1214 Either @var{address} was unusable, or inconsistent @var{flags} were
1219 @var{filedes} was not open for the type of access specified in @var{protect}.
1223 Either there is not enough memory for the operation, or the process is
1224 out of address space.
1228 This file is of a type that doesn't support mapping.
1232 The file is on a filesystem that doesn't support mapping.
1234 @c On Linux, EAGAIN will appear if the file has a conflicting mandatory lock.
1235 @c However mandatory locks are not discussed in this manual.
1237 @c Similarly, ETXTBSY will occur if the MAP_DENYWRITE flag (not documented
1238 @c here) is used and the file is already open for writing.
1244 @deftypefun int munmap (void *@var{addr}, size_t @var{length})
1246 @code{munmap} removes any memory maps from (@var{addr}) to (@var{addr} +
1247 @var{length}). @var{length} should be the length of the mapping.
1249 It is safe to un-map multiple mappings in one command, or include unmapped
1250 space in the range. It is also possible to unmap only part of an existing
1251 mapping, however only entire pages can be removed. If @var{length} is not
1252 an even number of pages, it will be rounded up.
1254 It returns @math{0} for success and @math{-1} for an error.
1256 One error is possible:
1261 The memory range given was outside the user mmap range, or wasn't page
1268 @deftypefun int msync (void *@var{address}, size_t @var{length}, int @var{flags})
1270 When using shared mappings, the kernel can write the file at any time
1271 before the mapping is removed. To be certain data has actually been
1272 written to the file and will be accessable to non-memory-mapped I/O, it
1273 is neccessary to use this function.
1275 It operates on the region @var{address} to (@var{address} + @var{length}).
1276 It may be used on part of a mapping or multiple mappings, however the
1277 region given should not contain any unmapped space.
1279 @var{flags} can contain some options:
1285 This flag makes sure the data is actually written @emph{to disk}.
1286 Normally @code{msync} only makes sure that accesses to a file with
1287 conventional I/O reflect the recent changes.
1291 This tells @code{msync} to begin the synchronization, but not to wait for
1294 @c Linux also has MS_INVALIDATE, which I don't understand.
1298 @code{msync} returns @math{0} for success and @math{-1} for
1299 error. Errors include:
1304 An invalid region was given, or the @var{flags} were invalid.
1307 There is no existing mapping in at least part of the given region.
1313 @deftypefun {void *} mremap (void *@var{address}, size_t @var{length}, size_t @var{new_length}, int @var{flag})
1315 This function can be used to change the size of an existing memory
1316 area. @var{address} and @var{length} must cover a region entirely mapped
1317 in the same @code{mmap} statement. A new mapping with the same
1318 characteristics will be returned, but a with the length @var{new_length}
1321 One option is possible, @code{MREMAP_MAYMOVE}. If it is given in
1322 @var{flags}, the system may remove the existing mapping and create a new
1323 one of the desired length in another location.
1325 The address of the resulting mapping is returned, or @math{-1}. Possible
1326 error codes include:
1328 This function is only available on a few systems. Except for performing
1329 optional optimizations one should not rely on this function.
1333 There is no existing mapping in at least part of the original region, or
1334 the region covers two or more distinct mappings.
1337 The address given is misaligned or inappropriate.
1340 The region has pages locked, and if extended it would exceed the
1341 process's resource limit for locked pages. @xref{Limits on Resources}.
1344 The region is private writable, and insufficent virtual memory is
1345 available to extend it. Also, this error will occur if
1346 @code{MREMAP_MAYMOVE} is not given and the extension would collide with
1347 another mapped region.
1352 Not all file descriptors may be mapped. Sockets, pipes, and most devices
1353 only allow sequential access and do not fit into the mapping abstraction.
1354 In addition, some regular files may not be mmapable, and older kernels may
1355 not support mapping at all. Thus, programs using @code{mmap} should
1356 have a fallback method to use should it fail. @xref{Mmap,,,standards,GNU
1359 @c XXX madvice documentation missing
1361 @node Waiting for I/O
1362 @section Waiting for Input or Output
1363 @cindex waiting for input or output
1364 @cindex multiplexing input
1365 @cindex input from multiple files
1367 Sometimes a program needs to accept input on multiple input channels
1368 whenever input arrives. For example, some workstations may have devices
1369 such as a digitizing tablet, function button box, or dial box that are
1370 connected via normal asynchronous serial interfaces; good user interface
1371 style requires responding immediately to input on any device. Another
1372 example is a program that acts as a server to several other processes
1373 via pipes or sockets.
1375 You cannot normally use @code{read} for this purpose, because this
1376 blocks the program until input is available on one particular file
1377 descriptor; input on other channels won't wake it up. You could set
1378 nonblocking mode and poll each file descriptor in turn, but this is very
1381 A better solution is to use the @code{select} function. This blocks the
1382 program until input or output is ready on a specified set of file
1383 descriptors, or until a timer expires, whichever comes first. This
1384 facility is declared in the header file @file{sys/types.h}.
1387 In the case of a server socket (@pxref{Listening}), we say that
1388 ``input'' is available when there are pending connections that could be
1389 accepted (@pxref{Accepting Connections}). @code{accept} for server
1390 sockets blocks and interacts with @code{select} just as @code{read} does
1393 @cindex file descriptor sets, for @code{select}
1394 The file descriptor sets for the @code{select} function are specified
1395 as @code{fd_set} objects. Here is the description of the data type
1396 and some macros for manipulating these objects.
1398 @comment sys/types.h
1400 @deftp {Data Type} fd_set
1401 The @code{fd_set} data type represents file descriptor sets for the
1402 @code{select} function. It is actually a bit array.
1405 @comment sys/types.h
1407 @deftypevr Macro int FD_SETSIZE
1408 The value of this macro is the maximum number of file descriptors that a
1409 @code{fd_set} object can hold information about. On systems with a
1410 fixed maximum number, @code{FD_SETSIZE} is at least that number. On
1411 some systems, including GNU, there is no absolute limit on the number of
1412 descriptors open, but this macro still has a constant value which
1413 controls the number of bits in an @code{fd_set}; if you get a file
1414 descriptor with a value as high as @code{FD_SETSIZE}, you cannot put
1415 that descriptor into an @code{fd_set}.
1418 @comment sys/types.h
1420 @deftypefn Macro void FD_ZERO (fd_set *@var{set})
1421 This macro initializes the file descriptor set @var{set} to be the
1425 @comment sys/types.h
1427 @deftypefn Macro void FD_SET (int @var{filedes}, fd_set *@var{set})
1428 This macro adds @var{filedes} to the file descriptor set @var{set}.
1431 @comment sys/types.h
1433 @deftypefn Macro void FD_CLR (int @var{filedes}, fd_set *@var{set})
1434 This macro removes @var{filedes} from the file descriptor set @var{set}.
1437 @comment sys/types.h
1439 @deftypefn Macro int FD_ISSET (int @var{filedes}, fd_set *@var{set})
1440 This macro returns a nonzero value (true) if @var{filedes} is a member
1441 of the file descriptor set @var{set}, and zero (false) otherwise.
1444 Next, here is the description of the @code{select} function itself.
1446 @comment sys/types.h
1448 @deftypefun int select (int @var{nfds}, fd_set *@var{read-fds}, fd_set *@var{write-fds}, fd_set *@var{except-fds}, struct timeval *@var{timeout})
1449 The @code{select} function blocks the calling process until there is
1450 activity on any of the specified sets of file descriptors, or until the
1451 timeout period has expired.
1453 The file descriptors specified by the @var{read-fds} argument are
1454 checked to see if they are ready for reading; the @var{write-fds} file
1455 descriptors are checked to see if they are ready for writing; and the
1456 @var{except-fds} file descriptors are checked for exceptional
1457 conditions. You can pass a null pointer for any of these arguments if
1458 you are not interested in checking for that kind of condition.
1460 A file descriptor is considered ready for reading if it is not at end of
1461 file. A server socket is considered ready for reading if there is a
1462 pending connection which can be accepted with @code{accept};
1463 @pxref{Accepting Connections}. A client socket is ready for writing when
1464 its connection is fully established; @pxref{Connecting}.
1466 ``Exceptional conditions'' does not mean errors---errors are reported
1467 immediately when an erroneous system call is executed, and do not
1468 constitute a state of the descriptor. Rather, they include conditions
1469 such as the presence of an urgent message on a socket. (@xref{Sockets},
1470 for information on urgent messages.)
1472 The @code{select} function checks only the first @var{nfds} file
1473 descriptors. The usual thing is to pass @code{FD_SETSIZE} as the value
1476 The @var{timeout} specifies the maximum time to wait. If you pass a
1477 null pointer for this argument, it means to block indefinitely until one
1478 of the file descriptors is ready. Otherwise, you should provide the
1479 time in @code{struct timeval} format; see @ref{High-Resolution
1480 Calendar}. Specify zero as the time (a @code{struct timeval} containing
1481 all zeros) if you want to find out which descriptors are ready without
1482 waiting if none are ready.
1484 The normal return value from @code{select} is the total number of ready file
1485 descriptors in all of the sets. Each of the argument sets is overwritten
1486 with information about the descriptors that are ready for the corresponding
1487 operation. Thus, to see if a particular descriptor @var{desc} has input,
1488 use @code{FD_ISSET (@var{desc}, @var{read-fds})} after @code{select} returns.
1490 If @code{select} returns because the timeout period expires, it returns
1493 Any signal will cause @code{select} to return immediately. So if your
1494 program uses signals, you can't rely on @code{select} to keep waiting
1495 for the full time specified. If you want to be sure of waiting for a
1496 particular amount of time, you must check for @code{EINTR} and repeat
1497 the @code{select} with a newly calculated timeout based on the current
1498 time. See the example below. See also @ref{Interrupted Primitives}.
1500 If an error occurs, @code{select} returns @code{-1} and does not modify
1501 the argument file descriptor sets. The following @code{errno} error
1502 conditions are defined for this function:
1506 One of the file descriptor sets specified an invalid file descriptor.
1509 The operation was interrupted by a signal. @xref{Interrupted Primitives}.
1512 The @var{timeout} argument is invalid; one of the components is negative
1517 @strong{Portability Note:} The @code{select} function is a BSD Unix
1520 Here is an example showing how you can use @code{select} to establish a
1521 timeout period for reading from a file descriptor. The @code{input_timeout}
1522 function blocks the calling process until input is available on the
1523 file descriptor, or until the timeout period expires.
1526 @include select.c.texi
1529 There is another example showing the use of @code{select} to multiplex
1530 input from multiple sockets in @ref{Server Example}.
1533 @node Synchronizing I/O
1534 @section Synchronizing I/O operations
1536 @cindex synchronizing
1537 In most modern operation systems the normal I/O operations are not
1538 executed synchronously. I.e., even if a @code{write} system call
1539 returns this does not mean the data is actually written to the media,
1542 In situations where synchronization points are necessary the user can
1543 use special functions which ensure that all operations finished before
1548 @deftypefun int sync (void)
1549 A call to this function will not return as long as there is data which
1550 that is not written to the device. All dirty buffers in the kernel will
1551 be written and so an overall consistent system can be achieved (if no
1552 other process in parallel writes data).
1554 A prototype for @code{sync} can be found in @file{unistd.h}.
1556 The return value is zero to indicate no error.
1559 More often it is wanted that not all data in the system is committed.
1560 Programs want to ensure that data written to a given file are all
1561 committed and in this situation @code{sync} is overkill.
1565 @deftypefun int fsync (int @var{fildes})
1566 The @code{fsync} can be used to make sure all data associated with the
1567 open file @var{fildes} is written to the device associated with the
1568 descriptor. The function call does not return unless all actions have
1571 A prototype for @code{fsync} can be found in @file{unistd.h}.
1573 This function is a cancelation point in multi-threaded programs. This
1574 is a problem if the thread allocates some resources (like memory, file
1575 descriptors, semaphores or whatever) at the time @code{fsync} is
1576 called. If the thread gets canceled these resources stay allocated
1577 until the program ends. To avoid this calls to @code{fsync} should be
1578 protected using cancelation handlers.
1579 @c ref pthread_cleanup_push / pthread_cleanup_pop
1581 The return value of the function is zero if no error occured. Otherwise
1582 it is @math{-1} and the global variable @var{errno} is set to the
1586 The descriptor @var{fildes} is not valid.
1589 No synchronization is possible since the system does not implement this.
1593 Sometimes it is not even necessary to write all data associated with a
1594 file descriptor. E.g., in database files which do not change in size it
1595 is enough to write all the file content data to the device.
1596 Meta-information like the modification time etc. are not that important
1597 and leaving such information uncommitted does not prevent a successful
1598 recovering of the file in case of a problem.
1602 @deftypefun int fdatasync (int @var{fildes})
1603 When a call to the @code{fdatasync} function returns it is made sure
1604 that all of the file data is written to the device. For all pending I/O
1605 operations the parts guaranteeing data integrity finished.
1607 Not all systems implement the @code{fdatasync} operation. On systems
1608 missing this functionality @code{fdatasync} is emulated by a call to
1609 @code{fsync} since the performed actions are a superset of those
1610 required by @code{fdatasyn}.
1612 The prototype for @code{fdatasync} is in @file{unistd.h}.
1614 The return value of the function is zero if no error occured. Otherwise
1615 it is @math{-1} and the global variable @var{errno} is set to the
1619 The descriptor @var{fildes} is not valid.
1622 No synchronization is possible since the system does not implement this.
1627 @node Asynchronous I/O
1628 @section Perform I/O Operations in Parallel
1630 The POSIX.1b standard defines a new set of I/O operations which can
1631 reduce the time an application spends waiting at I/O significantly. The
1632 new functions allow a program to initiate one or more I/O operations and
1633 then immediately resume the normal work while the I/O operations are
1634 executed in parallel. The functionality is available if the
1635 @file{unistd.h} file defines the symbol @code{_POSIX_ASYNCHRONOUS_IO}.
1637 These functions are part of the library with realtime functions named
1638 @file{librt}. They are not actually part of the @file{libc} binary.
1639 The implementation of these functions can be done using support in the
1640 kernel (if available) or using an implementation based on threads at
1641 userlevel. In the latter case it might be necessary to link applications
1642 with the thread library @file{libpthread} in addition to @file{librt}.
1644 All AIO operations operate on files which were opened previously. There
1645 might be arbitrary many operations for one file running. The
1646 asynchronous I/O operations are controlled using a data structure named
1647 @code{struct aiocb} (@dfn{AIO control block}). It is defined in
1648 @file{aio.h} as follows.
1652 @deftp {Data Type} {struct aiocb}
1653 The POSIX.1b standard mandates that the @code{struct aiocb} structure
1654 contains at least the members described in the following table. There
1655 might be more elements which are used by the implementation but
1656 depending on these elements is not portable and is highly deprecated.
1659 @item int aio_fildes
1660 This element specifies the file descriptor which is used for the
1661 operation. It must be a legal descriptor since otherwise the operation
1662 fails for obvious reasons.
1664 The device on which the file is opened must allow the seek operation.
1665 I.e., it is not possible to use any of the AIO operations on devices
1666 like terminals where an @code{lseek} call would lead to an error.
1668 @item off_t aio_offset
1669 This element specifies at which offset in the file the operation (input
1670 or output) is performed. Since the operations are carried out in arbitrary
1671 order and more than one operation for one file descriptor can be
1672 started, one cannot expect a current read/write position of the file
1675 @item volatile void *aio_buf
1676 This is a pointer to the buffer with the data to be written or the place
1677 where the read data is stored.
1679 @item size_t aio_nbytes
1680 This element specifies the length of the buffer pointed to by @code{aio_buf}.
1682 @item int aio_reqprio
1683 If the platform has defined @code{_POSIX_PRIORITIZED_IO} and
1684 @code{_POSIX_PRIORITY_SCHEDULING} the AIO requests are
1685 processed based on the current scheduling priority. The
1686 @code{aio_reqprio} element can then be used to lower the priority of the
1689 @item struct sigevent aio_sigevent
1690 This element specifies how the calling process is notified once the
1691 operation terminates. If the @code{sigev_notify} element is
1692 @code{SIGEV_NONE} no notification is send. If it is @code{SIGEV_SIGNAL}
1693 the signal determined by @code{sigev_signo} is send. Otherwise
1694 @code{sigev_notify} must be @code{SIGEV_THREAD}. In this case a thread
1695 is created which starts executing the function pointed to by
1696 @code{sigev_notify_function}.
1698 @item int aio_lio_opcode
1699 This element is only used by the @code{lio_listio} and
1700 @code{lio_listio64} functions. Since these functions allow to start an
1701 arbitrary number of operations at once and since each operation can be
1702 input or output (or nothing) the information must be stored in the
1703 control block. The possible values are:
1707 Start a read operation. Read from the file at position
1708 @code{aio_offset} and store the next @code{aio_nbytes} bytes in the
1709 buffer pointed to by @code{aio_buf}.
1712 Start a write operation. Write @code{aio_nbytes} bytes starting at
1713 @code{aio_buf} into the file starting at position @code{aio_offset}.
1716 Do nothing for this control block. This value is useful sometimes when
1717 an array of @code{struct aiocb} values contains holes, i.e., some of the
1718 values must not be handled although the whole array is presented to the
1719 @code{lio_listio} function.
1723 When the sources are compiled using @code{_FILE_OFFSET_BITS == 64} on a
1724 32 bits machine this type is in fact @code{struct aiocb64} since the LFS
1725 interface transparently replaces the @code{struct aiocb} definition.
1728 For use with the AIO functions defined in the LFS there is a similar type
1729 defined which replaces the types of the appropriate members with larger
1730 types but otherwise is equivalent to @code{struct aiocb}. Especially
1731 all member names are the same.
1735 @deftp {Data Type} {struct aiocb64}
1737 @item int aio_fildes
1738 This element specifies the file descriptor which is used for the
1739 operation. It must be a legal descriptor since otherwise the operation
1740 fails for obvious reasons.
1742 The device on which the file is opened must allow the seek operation.
1743 I.e., it is not possible to use any of the AIO operations on devices
1744 like terminals where an @code{lseek} call would lead to an error.
1746 @item off64_t aio_offset
1747 This element specified at which offset in the file the operation (input
1748 or output) is performed. Since the operation are carried in arbitrary
1749 order and more than one operation for one file descriptor can be
1750 started, one cannot expect a current read/write position of the file
1753 @item volatile void *aio_buf
1754 This is a pointer to the buffer with the data to be written or the place
1755 where the ead data is stored.
1757 @item size_t aio_nbytes
1758 This element specifies the length of the buffer pointed to by @code{aio_buf}.
1760 @item int aio_reqprio
1761 If for the platform @code{_POSIX_PRIORITIZED_IO} and
1762 @code{_POSIX_PRIORITY_SCHEDULING} is defined the AIO requests are
1763 processed based on the current scheduling priority. The
1764 @code{aio_reqprio} element can then be used to lower the priority of the
1767 @item struct sigevent aio_sigevent
1768 This element specifies how the calling process is notified once the
1769 operation terminates. If the @code{sigev_notify} element is
1770 @code{SIGEV_NONE} no notification is send. If it is @code{SIGEV_SIGNAL}
1771 the signal determined by @code{sigev_signo} is send. Otherwise
1772 @code{sigev_notify} must be @code{SIGEV_THREAD} in which case a thread
1773 which starts executing the function pointeed to by
1774 @code{sigev_notify_function}.
1776 @item int aio_lio_opcode
1777 This element is only used by the @code{lio_listio} and
1778 @code{[lio_listio64} functions. Since these functions allow to start an
1779 arbitrary number of operations at once and since each operation can be
1780 input or output (or nothing) the information must be stored in the
1781 control block. See the description of @code{struct aiocb} for a description
1782 of the possible values.
1785 When the sources are compiled using @code{_FILE_OFFSET_BITS == 64} on a
1786 32 bits machine this type is available under the name @code{struct
1787 aiocb64} since the LFS replaces transparently the old interface.
1791 * Asynchronous Reads/Writes:: Asynchronous Read and Write Operations.
1792 * Status of AIO Operations:: Getting the Status of AIO Operations.
1793 * Synchronizing AIO Operations:: Getting into a consistent state.
1794 * Cancel AIO Operations:: Cancelation of AIO Operations.
1795 * Configuration of AIO:: How to optimize the AIO implementation.
1798 @node Asynchronous Reads/Writes
1799 @subsection Asynchronous Read and Write Operations
1803 @deftypefun int aio_read (struct aiocb *@var{aiocbp})
1804 This function initiates an asynchronous read operation. The function
1805 call immediately returns after the operation was enqueued or when an
1806 error was encountered.
1808 The first @code{aiocbp->aio_nbytes} bytes of the file for which
1809 @code{aiocbp->aio_fildes} is a descriptor are written to the buffer
1810 starting at @code{aiocbp->aio_buf}. Reading starts at the absolute
1811 position @code{aiocbp->aio_offset} in the file.
1813 If prioritized I/O is supported by the platform the
1814 @code{aiocbp->aio_reqprio} value is used to adjust the priority before
1815 the request is actually enqueued.
1817 The calling process is notified about the termination of the read
1818 request according to the @code{aiocbp->aio_sigevent} value.
1820 When @code{aio_read} returns the return value is zero if no error
1821 occurred that can be found before the process is enqueued. If such an
1822 early error is found the function returns @math{-1} and sets
1823 @code{errno} to one of the following values.
1827 The request was not enqueued due to (temporarily) exceeded resource
1830 The @code{aio_read} function is not implemented.
1832 The @code{aiocbp->aio_fildes} descriptor is not valid. This condition
1833 needs not be recognized before enqueueing the request and so this error
1834 might also be signaled asynchronously.
1836 The @code{aiocbp->aio_offset} or @code{aiocbp->aio_reqpiro} value is
1837 invalid. This condition need not be recognized before enqueueing the
1838 request and so this error might also be signaled asynchrously.
1841 In the case @code{aio_read} returns zero the current status of the
1842 request can be queried using @code{aio_error} and @code{aio_return}
1843 functions. As long as the value returned by @code{aio_error} is
1844 @code{EINPROGRESS} the operation has not yet completed. If
1845 @code{aio_error} returns zero the operation successfully terminated,
1846 otherwise the value is to be interpreted as an error code. If the
1847 function terminated the result of the operation can be get using a call
1848 to @code{aio_return}. The returned value is the same as an equivalent
1849 call to @code{read} would have returned. Possible error codes returned
1850 by @code{aio_error} are:
1854 The @code{aiocbp->aio_fildes} descriptor is not valid.
1856 The operation was canceled before the operation was finished
1857 (@pxref{Cancel AIO Operations})
1859 The @code{aiocbp->aio_offset} value is invalid.
1862 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
1863 function is in fact @code{aio_read64} since the LFS interface transparently
1864 replaces the normal implementation.
1869 @deftypefun int aio_read64 (struct aiocb *@var{aiocbp})
1870 This function is similar to the @code{aio_read} function. The only
1871 difference is that on @w{32 bits} machines the file descriptor should
1872 be opened in the large file mode. Internally @code{aio_read64} uses
1873 functionality equivalent to @code{lseek64} (@pxref{File Position
1874 Primitive}) to position the file descriptor correctly for the reading,
1875 as opposed to @code{lseek} functionality used in @code{aio_read}.
1877 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
1878 function is available under the name @code{aio_read} and so transparently
1879 replaces the interface for small files on 32 bits machines.
1882 To write data asynchronously to a file there exists an equivalent pair
1883 of functions with a very similar interface.
1887 @deftypefun int aio_write (struct aiocb *@var{aiocbp})
1888 This function initiates an asynchronous write operation. The function
1889 call immediately returns after the operation was enqueued or if before
1890 this happens an error was encountered.
1892 The first @code{aiocbp->aio_nbytes} bytes from the buffer starting at
1893 @code{aiocbp->aio_buf} are written to the file for which
1894 @code{aiocbp->aio_fildes} is an descriptor, starting at the absolute
1895 position @code{aiocbp->aio_offset} in the file.
1897 If prioritized I/O is supported by the platform the
1898 @code{aiocbp->aio_reqprio} value is used to adjust the priority before
1899 the request is actually enqueued.
1901 The calling process is notified about the termination of the read
1902 request according to the @code{aiocbp->aio_sigevent} value.
1904 When @code{aio_write} returns the return value is zero if no error
1905 occurred that can be found before the process is enqueued. If such an
1906 early error is found the function returns @math{-1} and sets
1907 @code{errno} to one of the following values.
1911 The request was not enqueued due to (temporarily) exceeded resource
1914 The @code{aio_write} function is not implemented.
1916 The @code{aiocbp->aio_fildes} descriptor is not valid. This condition
1917 needs not be recognized before enqueueing the request and so this error
1918 might also be signaled asynchronously.
1920 The @code{aiocbp->aio_offset} or @code{aiocbp->aio_reqpiro} value is
1921 invalid. This condition needs not be recognized before enqueueing the
1922 request and so this error might also be signaled asynchronously.
1925 In the case @code{aio_write} returns zero the current status of the
1926 request can be queried using @code{aio_error} and @code{aio_return}
1927 functions. As long as the value returned by @code{aio_error} is
1928 @code{EINPROGRESS} the operation has not yet completed. If
1929 @code{aio_error} returns zero the operation successfully terminated,
1930 otherwise the value is to be interpreted as an error code. If the
1931 function terminated the result of the operation can be get using a call
1932 to @code{aio_return}. The returned value is the same as an equivalent
1933 call to @code{read} would have returned. Possible error code returned
1934 by @code{aio_error} are:
1938 The @code{aiocbp->aio_fildes} descriptor is not valid.
1940 The operation was canceled before the operation was finished
1941 (@pxref{Cancel AIO Operations})
1943 The @code{aiocbp->aio_offset} value is invalid.
1946 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
1947 function is in fact @code{aio_write64} since the LFS interface transparently
1948 replaces the normal implementation.
1953 @deftypefun int aio_write64 (struct aiocb *@var{aiocbp})
1954 This function is similar to the @code{aio_write} function. The only
1955 difference is that on @w{32 bits} machines the file descriptor should
1956 be opened in the large file mode. Internally @code{aio_write64} uses
1957 functionality equivalent to @code{lseek64} (@pxref{File Position
1958 Primitive}) to position the file descriptor correctly for the writing,
1959 as opposed to @code{lseek} functionality used in @code{aio_write}.
1961 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
1962 function is available under the name @code{aio_write} and so transparently
1963 replaces the interface for small files on 32 bits machines.
1966 Beside these functions with the more or less traditional interface
1967 POSIX.1b also defines a function with can initiate more than one
1968 operation at once and which can handled freely mixed read and write
1969 operation. It is therefore similar to a combination of @code{readv} and
1974 @deftypefun int lio_listio (int @var{mode}, struct aiocb *const @var{list}[], int @var{nent}, struct sigevent *@var{sig})
1975 The @code{lio_listio} function can be used to enqueue an arbitrary
1976 number of read and write requests at one time. The requests can all be
1977 meant for the same file, all for different files or every solution in
1980 @code{lio_listio} gets the @var{nent} requests from the array pointed to
1981 by @var{list}. What operation has to be performed is determined by the
1982 @code{aio_lio_opcode} member in each element of @var{list}. If this
1983 field is @code{LIO_READ} an read operation is queued, similar to a call
1984 of @code{aio_read} for this element of the array (except that the way
1985 the termination is signalled is different, as we will see below). If
1986 the @code{aio_lio_opcode} member is @code{LIO_WRITE} an write operation
1987 is enqueued. Otherwise the @code{aio_lio_opcode} must be @code{LIO_NOP}
1988 in which case this element of @var{list} is simply ignored. This
1989 ``operation'' is useful in situations where one has a fixed array of
1990 @code{struct aiocb} elements from which only a few need to be handled at
1991 a time. Another situation is where the @code{lio_listio} call was
1992 cancelled before all requests are processed (@pxref{Cancel AIO
1993 Operations}) and the remaining requests have to be reissued.
1995 The other members of each element of the array pointed to by
1996 @code{list} must have values suitable for the operation as described in
1997 the documentation for @code{aio_read} and @code{aio_write} above.
1999 The @var{mode} argument determines how @code{lio_listio} behaves after
2000 having enqueued all the requests. If @var{mode} is @code{LIO_WAIT} it
2001 waits until all requests terminated. Otherwise @var{mode} must be
2002 @code{LIO_NOWAIT} and in this case the function returns immediately after
2003 having enqueued all the requests. In this case the caller gets a
2004 notification of the termination of all requests according to the
2005 @var{sig} parameter. If @var{sig} is @code{NULL} no notification is
2006 send. Otherwise a signal is sent or a thread is started, just as
2007 described in the description for @code{aio_read} or @code{aio_write}.
2009 If @var{mode} is @code{LIO_WAIT} the return value of @code{lio_listio}
2010 is @math{0} when all requests completed successfully. Otherwise the
2011 function return @math{-1} and @code{errno} is set accordingly. To find
2012 out which request or requests failed one has to use the @code{aio_error}
2013 function on all the elements of the array @var{list}.
2015 In case @var{mode} is @code{LIO_NOWAIT} the function return @math{0} if
2016 all requests were enqueued correctly. The current state of the requests
2017 can be found using @code{aio_error} and @code{aio_return} as described
2018 above. In case @code{lio_listio} returns @math{-1} in this mode the
2019 global variable @code{errno} is set accordingly. If a request did not
2020 yet terminate a call to @code{aio_error} returns @code{EINPROGRESS}. If
2021 the value is different the request is finished and the error value (or
2022 @math{0}) is returned and the result of the operation can be retrieved
2023 using @code{aio_return}.
2025 Possible values for @code{errno} are:
2029 The resources necessary to queue all the requests are not available in
2030 the moment. The error status for each element of @var{list} must be
2031 checked which request failed.
2033 Another reason could be that the system wide limit of AIO requests is
2034 exceeded. This cannot be the case for the implementation on GNU systems
2035 since no arbitrary limits exist.
2037 The @var{mode} parameter is invalid or @var{nent} is larger than
2038 @code{AIO_LISTIO_MAX}.
2040 One or more of the request's I/O operations failed. The error status of
2041 each request should be checked for which one failed.
2043 The @code{lio_listio} function is not supported.
2046 If the @var{mode} parameter is @code{LIO_NOWAIT} and the caller cancels
2047 an request the error status for this request returned by
2048 @code{aio_error} is @code{ECANCELED}.
2050 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2051 function is in fact @code{lio_listio64} since the LFS interface
2052 transparently replaces the normal implementation.
2057 @deftypefun int lio_listio64 (int @var{mode}, struct aiocb *const @var{list}, int @var{nent}, struct sigevent *@var{sig})
2058 This function is similar to the @code{aio_listio} function. The only
2059 difference is that only @w{32 bits} machines the file descriptor should
2060 be opened in the large file mode. Internally @code{lio_listio64} uses
2061 functionality equivalent to @code{lseek64} (@pxref{File Position
2062 Primitive}) to position the file descriptor correctly for the reading or
2063 writing, as opposed to @code{lseek} functionality used in
2066 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2067 function is available under the name @code{lio_listio} and so
2068 transparently replaces the interface for small files on 32 bits
2072 @node Status of AIO Operations
2073 @subsection Getting the Status of AIO Operations
2075 As already described in the documentation of the functions in the last
2076 section it must be possible to get information about the status of a I/O
2077 request. When the operation is performed really asynchronous (as with
2078 @code{aio_read} and @code{aio_write} and with @code{aio_listio} when the
2079 mode is @code{LIO_NOWAIT}) one sometimes needs to know whether a
2080 specific request already terminated and if yes, what the result was..
2081 The following two function allow to get this kind of information.
2085 @deftypefun int aio_error (const struct aiocb *@var{aiocbp})
2086 This function determines the error state of the request described by the
2087 @code{struct aiocb} variable pointed to by @var{aiocbp}. If the
2088 request has not yet terminated the value returned is always
2089 @code{EINPROGRESS}. Once the request has terminated the value
2090 @code{aio_error} returns is either @math{0} if the request completed
2091 successfully or it returns the value which would be stored in the
2092 @code{errno} variable if the request would have been done using
2093 @code{read}, @code{write}, or @code{fsync}.
2095 The function can return @code{ENOSYS} if it is not implemented. It
2096 could also return @code{EINVAL} if the @var{aiocbp} parameter does not
2097 refer to an asynchronous operation whose return status is not yet known.
2099 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2100 function is in fact @code{aio_error64} since the LFS interface
2101 transparently replaces the normal implementation.
2106 @deftypefun int aio_error64 (const struct aiocb64 *@var{aiocbp})
2107 This function is similar to @code{aio_error} with the only difference
2108 that the argument is a reference to a variable of type @code{struct
2111 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2112 function is available under the name @code{aio_error} and so
2113 transparently replaces the interface for small files on 32 bits
2119 @deftypefun ssize_t aio_return (const struct aiocb *@var{aiocbp})
2120 This function can be used to retrieve the return status of the operation
2121 carried out by the request described in the variable pointed to by
2122 @var{aiocbp}. As long as the error status of this request as returned
2123 by @code{aio_error} is @code{EINPROGRESS} the return of this function is
2126 Once the request is finished this function can be used exactly once to
2127 retrieve the return value. Following calls might lead to undefined
2128 behaviour. The return value itself is the value which would have been
2129 returned by the @code{read}, @code{write}, or @code{fsync} call.
2131 The function can return @code{ENOSYS} if it is not implemented. It
2132 could also return @code{EINVAL} if the @var{aiocbp} parameter does not
2133 refer to an asynchronous operation whose return status is not yet known.
2135 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2136 function is in fact @code{aio_return64} since the LFS interface
2137 transparently replaces the normal implementation.
2142 @deftypefun int aio_return64 (const struct aiocb64 *@var{aiocbp})
2143 This function is similar to @code{aio_return} with the only difference
2144 that the argument is a reference to a variable of type @code{struct
2147 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2148 function is available under the name @code{aio_return} and so
2149 transparently replaces the interface for small files on 32 bits
2153 @node Synchronizing AIO Operations
2154 @subsection Getting into a Consistent State
2156 When dealing with asynchronous operations it is sometimes necessary to
2157 get into a consistent state. This would mean for AIO that one wants to
2158 know whether a certain request or a group of request were processed.
2159 This could be done by waiting for the notification sent by the system
2160 after the operation terminated but this sometimes would mean wasting
2161 resources (mainly computation time). Instead POSIX.1b defines two
2162 functions which will help with most kinds of consistency.
2164 The @code{aio_fsync} and @code{aio_fsync64} functions are only available
2165 if in @file{unistd.h} the symbol @code{_POSIX_SYNCHRONIZED_IO} is
2168 @cindex synchronizing
2171 @deftypefun int aio_fsync (int @var{op}, struct aiocb *@var{aiocbp})
2172 Calling this function forces all I/O operations operating queued at the
2173 time of the function call operating on the file descriptor
2174 @code{aiocbp->aio_fildes} into the synchronized I/O completion state
2175 (@pxref{Synchronizing I/O}). The @code{aio_fsync} function return
2176 immediately but the notification through the method described in
2177 @code{aiocbp->aio_sigevent} will happen only after all requests for this
2178 file descriptor terminated and the file is synchronized. This also
2179 means that requests for this very same file descriptor which are queued
2180 after the synchronization request are not effected.
2182 If @var{op} is @code{O_DSYNC} the synchronization happens as with a call
2183 to @code{fdatasync}. Otherwise @var{op} should be @code{O_SYNC} and
2184 the synchronization happens as with @code{fsync}.
2186 As long as the synchronization has not happened a call to
2187 @code{aio_error} with the reference to the object pointed to by
2188 @var{aiocbp} returns @code{EINPROGRESS}. Once the synchronization is
2189 done @code{aio_error} return @math{0} if the synchronization was not
2190 successful. Otherwise the value returned is the value to which the
2191 @code{fsync} or @code{fdatasync} function would have set the
2192 @code{errno} variable. In this case nothing can be assumed about the
2193 consistency for the data written to this file descriptor.
2195 The return value of this function is @math{0} if the request was
2196 successfully filed. Otherwise the return value is @math{-1} and
2197 @code{errno} is set to one of the following values:
2201 The request could not be enqueued due to temporary lack of resources.
2203 The file descriptor @code{aiocbp->aio_fildes} is not valid or not open
2206 The implementation does not support I/O synchronization or the @var{op}
2207 parameter is other than @code{O_DSYNC} and @code{O_SYNC}.
2209 This function is not implemented.
2212 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2213 function is in fact @code{aio_return64} since the LFS interface
2214 transparently replaces the normal implementation.
2219 @deftypefun int aio_fsync64 (int @var{op}, struct aiocb64 *@var{aiocbp})
2220 This function is similar to @code{aio_fsync} with the only difference
2221 that the argument is a reference to a variable of type @code{struct
2224 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2225 function is available under the name @code{aio_fsync} and so
2226 transparently replaces the interface for small files on 32 bits
2230 Another method of synchronization is to wait until one or more requests of a
2231 specific set terminated. This could be achieved by the @code{aio_*}
2232 functions to notify the initiating process about the termination but in
2233 some situations this is not the ideal solution. In a program which
2234 constantly updates clients somehow connected to the server it is not
2235 always the best solution to go round robin since some connections might
2236 be slow. On the other hand letting the @code{aio_*} function notify the
2237 caller might also be not the best solution since whenever the process
2238 works on preparing data for on client it makes no sense to be
2239 interrupted by a notification since the new client will not be handled
2240 before the current client is served. For situations like this
2241 @code{aio_suspend} should be used.
2245 @deftypefun int aio_suspend (const struct aiocb *const @var{list}[], int @var{nent}, const struct timespec *@var{timeout})
2246 When calling this function the calling thread is suspended until at
2247 least one of the requests pointed to by the @var{nent} elements of the
2248 array @var{list} has completed. If any of the requests already has
2249 completed at the time @code{aio_suspend} is called the function returns
2250 immediately. Whether a request has terminated or not is done by
2251 comparing the error status of the request with @code{EINPROGRESS}. If
2252 an element of @var{list} is @code{NULL} the entry is simply ignored.
2254 If no request has finished the calling process is suspended. If
2255 @var{timeout} is @code{NULL} the process is not waked until a request
2256 finished. If @var{timeout} is not @code{NULL} the process remains
2257 suspended at as long as specified in @var{timeout}. In this case
2258 @code{aio_suspend} returns with an error.
2260 The return value of the function is @math{0} if one or more requests
2261 from the @var{list} have terminated. Otherwise the function returns
2262 @math{-1} and @code{errno} is set to one of the following values:
2266 None of the requests from the @var{list} completed in the time specified
2269 A signal interrupted the @code{aio_suspend} function. This signal might
2270 also be sent by the AIO implementation while signalling the termination
2271 of one of the requests.
2273 The @code{aio_suspend} function is not implemented.
2276 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2277 function is in fact @code{aio_suspend64} since the LFS interface
2278 transparently replaces the normal implementation.
2283 @deftypefun int aio_suspend64 (const struct aiocb64 *const @var{list}[], int @var{nent}, const struct timespec *@var{timeout})
2284 This function is similar to @code{aio_suspend} with the only difference
2285 that the argument is a reference to a variable of type @code{struct
2288 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2289 function is available under the name @code{aio_suspend} and so
2290 transparently replaces the interface for small files on 32 bits
2294 @node Cancel AIO Operations
2295 @subsection Cancelation of AIO Operations
2297 When one or more requests are asynchronously processed it might be
2298 useful in some situations to cancel a selected operation, e.g., if it
2299 becomes obvious that the written data is not anymore accurate and would
2300 have to be overwritten soon. As an example assume an application, which
2301 writes data in files in a situation where new incoming data would have
2302 to be written in a file which will be updated by an enqueued request.
2303 The POSIX AIO implementation provides such a function but this function
2304 is not capable to force the cancelation of the request. It is up to the
2305 implementation to decide whether it is possible to cancel the operation
2306 or not. Therefore using this function is merely a hint.
2310 @deftypefun int aio_cancel (int @var{fildes}, struct aiocb *@var{aiocbp})
2311 The @code{aio_cancel} function can be used to cancel one or more
2312 outstanding requests. If the @var{aiocbp} parameter is @code{NULL} the
2313 function tries to cancel all outstanding requests which would process
2314 the file descriptor @var{fildes} (i.e.,, whose @code{aio_fildes} member
2315 is @var{fildes}). If @var{aiocbp} is not @code{NULL} the very specific
2316 request pointed to by @var{aiocbp} is tried to be canceled.
2318 For requests which were successfully canceled the normal notification
2319 about the termination of the request should take place. I.e., depending
2320 on the @code{struct sigevent} object which controls this, nothing
2321 happens, a signal is sent or a thread is started. If the request cannot
2322 be canceled it terminates the usual way after performing te operation.
2324 After a request is successfully canceled a call to @code{aio_error} with
2325 a reference to this request as the parameter will return
2326 @code{ECANCELED} and a call to @code{aio_return} will return @math{-1}.
2327 If the request wasn't canceled and is still running the error status is
2328 still @code{EINPROGRESS}.
2330 The return value of the function is @code{AIO_CANCELED} if there were
2331 requests which haven't terminated and which successfully were canceled.
2332 If there is one or more request left which couldn't be canceled the
2333 return value is @code{AIO_NOTCANCELED}. In this case @code{aio_error}
2334 must be used to find out which of the perhaps multiple requests (in
2335 @var{aiocbp} is @code{NULL}) wasn't successfully canceled. If all
2336 requests already terminated at the time @code{aio_cancel} is called the
2337 return value is @code{AIO_ALLDONE}.
2339 If an error occurred during the execution of @code{aio_cancel} the
2340 function returns @math{-1} and sets @code{errno} to one of the following
2345 The file descriptor @var{fildes} is not valid.
2347 @code{aio_cancel} is not implemented.
2350 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2351 function is in fact @code{aio_cancel64} since the LFS interface
2352 transparently replaces the normal implementation.
2357 @deftypefun int aio_cancel64 (int @var{fildes}, struct aiocb *@var{aiocbp})
2358 This function is similar to @code{aio_cancel} with the only difference
2359 that the argument is a reference to a variable of type @code{struct
2362 When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} this
2363 function is available under the name @code{aio_cancel} and so
2364 transparently replaces the interface for small files on 32 bits
2368 @node Configuration of AIO
2369 @subsection How to optimize the AIO implementation
2371 The POSIX standard does not specify how the AIO functions are
2372 implemented. They could be system calls but it is also possible to
2373 emulate them at userlevel.
2375 At least the available implementation at the point of this writing is a
2376 userlevel implementation which uses threads for handling the enqueued
2377 requests. This implementation requires to make some decisions about
2378 limitations but hard limitations are something which better should be
2379 avoided the GNU C library implementation provides a mean to tune the AIO
2380 implementation individually for each use.
2384 @deftp {Data Type} {struct aioinit}
2385 This data type is used to pass the configuration or tunable parameters
2386 to the implementation. The program has to initialize the members of
2387 this struct and pass it to the implementation using the @code{aio_init}
2391 @item int aio_threads
2392 This member specifies the maximal number of threads which must be used
2395 This number provides an estimate on the maximal number of simultaneously
2399 @item int aio_usedba
2403 @item int aio_numusers
2405 @item int aio_reserved[2]
2412 @deftypefun void aio_init (const struct aioinit *@var{init})
2413 This function must be called before any other AIO function. Calling it
2414 is completely voluntarily since it only is meant to help the AIO
2415 implementation to perform better.
2417 Before calling the @code{aio_init} function the members of a variable of
2418 type @code{struct aioinit} must be initialized. Then a reference to
2419 this variable is passed as the parameter to @code{aio_init} which itself
2420 may or may not pay attention to the hints.
2422 The function has no return value and no error cases are defined. It is
2423 a extension which follows a proposal from the SGI implementation in
2424 @w{Irix 6}. It is not covered by POSIX.1b or Unix98.
2427 @node Control Operations
2428 @section Control Operations on Files
2430 @cindex control operations on files
2431 @cindex @code{fcntl} function
2432 This section describes how you can perform various other operations on
2433 file descriptors, such as inquiring about or setting flags describing
2434 the status of the file descriptor, manipulating record locks, and the
2435 like. All of these operations are performed by the function @code{fcntl}.
2437 The second argument to the @code{fcntl} function is a command that
2438 specifies which operation to perform. The function and macros that name
2439 various flags that are used with it are declared in the header file
2440 @file{fcntl.h}. Many of these flags are also used by the @code{open}
2441 function; see @ref{Opening and Closing Files}.
2446 @deftypefun int fcntl (int @var{filedes}, int @var{command}, @dots{})
2447 The @code{fcntl} function performs the operation specified by
2448 @var{command} on the file descriptor @var{filedes}. Some commands
2449 require additional arguments to be supplied. These additional arguments
2450 and the return value and error conditions are given in the detailed
2451 descriptions of the individual commands.
2453 Briefly, here is a list of what the various commands are.
2457 Duplicate the file descriptor (return another file descriptor pointing
2458 to the same open file). @xref{Duplicating Descriptors}.
2461 Get flags associated with the file descriptor. @xref{Descriptor Flags}.
2464 Set flags associated with the file descriptor. @xref{Descriptor Flags}.
2467 Get flags associated with the open file. @xref{File Status Flags}.
2470 Set flags associated with the open file. @xref{File Status Flags}.
2473 Get a file lock. @xref{File Locks}.
2476 Set or clear a file lock. @xref{File Locks}.
2479 Like @code{F_SETLK}, but wait for completion. @xref{File Locks}.
2482 Get process or process group ID to receive @code{SIGIO} signals.
2483 @xref{Interrupt Input}.
2486 Set process or process group ID to receive @code{SIGIO} signals.
2487 @xref{Interrupt Input}.
2490 This function is a cancelation point in multi-threaded programs. This
2491 is a problem if the thread allocates some resources (like memory, file
2492 descriptors, semaphores or whatever) at the time @code{fcntl} is
2493 called. If the thread gets canceled these resources stay allocated
2494 until the program ends. To avoid this calls to @code{fcntl} should be
2495 protected using cancelation handlers.
2496 @c ref pthread_cleanup_push / pthread_cleanup_pop
2500 @node Duplicating Descriptors
2501 @section Duplicating Descriptors
2503 @cindex duplicating file descriptors
2504 @cindex redirecting input and output
2506 You can @dfn{duplicate} a file descriptor, or allocate another file
2507 descriptor that refers to the same open file as the original. Duplicate
2508 descriptors share one file position and one set of file status flags
2509 (@pxref{File Status Flags}), but each has its own set of file descriptor
2510 flags (@pxref{Descriptor Flags}).
2512 The major use of duplicating a file descriptor is to implement
2513 @dfn{redirection} of input or output: that is, to change the
2514 file or pipe that a particular file descriptor corresponds to.
2516 You can perform this operation using the @code{fcntl} function with the
2517 @code{F_DUPFD} command, but there are also convenient functions
2518 @code{dup} and @code{dup2} for duplicating descriptors.
2522 The @code{fcntl} function and flags are declared in @file{fcntl.h},
2523 while prototypes for @code{dup} and @code{dup2} are in the header file
2528 @deftypefun int dup (int @var{old})
2529 This function copies descriptor @var{old} to the first available
2530 descriptor number (the first number not currently open). It is
2531 equivalent to @code{fcntl (@var{old}, F_DUPFD, 0)}.
2536 @deftypefun int dup2 (int @var{old}, int @var{new})
2537 This function copies the descriptor @var{old} to descriptor number
2540 If @var{old} is an invalid descriptor, then @code{dup2} does nothing; it
2541 does not close @var{new}. Otherwise, the new duplicate of @var{old}
2542 replaces any previous meaning of descriptor @var{new}, as if @var{new}
2545 If @var{old} and @var{new} are different numbers, and @var{old} is a
2546 valid descriptor number, then @code{dup2} is equivalent to:
2550 fcntl (@var{old}, F_DUPFD, @var{new})
2553 However, @code{dup2} does this atomically; there is no instant in the
2554 middle of calling @code{dup2} at which @var{new} is closed and not yet a
2555 duplicate of @var{old}.
2560 @deftypevr Macro int F_DUPFD
2561 This macro is used as the @var{command} argument to @code{fcntl}, to
2562 copy the file descriptor given as the first argument.
2564 The form of the call in this case is:
2567 fcntl (@var{old}, F_DUPFD, @var{next-filedes})
2570 The @var{next-filedes} argument is of type @code{int} and specifies that
2571 the file descriptor returned should be the next available one greater
2572 than or equal to this value.
2574 The return value from @code{fcntl} with this command is normally the value
2575 of the new file descriptor. A return value of @math{-1} indicates an
2576 error. The following @code{errno} error conditions are defined for
2581 The @var{old} argument is invalid.
2584 The @var{next-filedes} argument is invalid.
2587 There are no more file descriptors available---your program is already
2588 using the maximum. In BSD and GNU, the maximum is controlled by a
2589 resource limit that can be changed; @pxref{Limits on Resources}, for
2590 more information about the @code{RLIMIT_NOFILE} limit.
2593 @code{ENFILE} is not a possible error code for @code{dup2} because
2594 @code{dup2} does not create a new opening of a file; duplicate
2595 descriptors do not count toward the limit which @code{ENFILE}
2596 indicates. @code{EMFILE} is possible because it refers to the limit on
2597 distinct descriptor numbers in use in one process.
2600 Here is an example showing how to use @code{dup2} to do redirection.
2601 Typically, redirection of the standard streams (like @code{stdin}) is
2602 done by a shell or shell-like program before calling one of the
2603 @code{exec} functions (@pxref{Executing a File}) to execute a new
2604 program in a child process. When the new program is executed, it
2605 creates and initializes the standard streams to point to the
2606 corresponding file descriptors, before its @code{main} function is
2609 So, to redirect standard input to a file, the shell could do something
2620 file = TEMP_FAILURE_RETRY (open (filename, O_RDONLY));
2621 dup2 (file, STDIN_FILENO);
2622 TEMP_FAILURE_RETRY (close (file));
2623 execv (program, NULL);
2627 There is also a more detailed example showing how to implement redirection
2628 in the context of a pipeline of processes in @ref{Launching Jobs}.
2631 @node Descriptor Flags
2632 @section File Descriptor Flags
2633 @cindex file descriptor flags
2635 @dfn{File descriptor flags} are miscellaneous attributes of a file
2636 descriptor. These flags are associated with particular file
2637 descriptors, so that if you have created duplicate file descriptors
2638 from a single opening of a file, each descriptor has its own set of flags.
2640 Currently there is just one file descriptor flag: @code{FD_CLOEXEC},
2641 which causes the descriptor to be closed if you use any of the
2642 @code{exec@dots{}} functions (@pxref{Executing a File}).
2644 The symbols in this section are defined in the header file
2650 @deftypevr Macro int F_GETFD
2651 This macro is used as the @var{command} argument to @code{fcntl}, to
2652 specify that it should return the file descriptor flags associated
2653 with the @var{filedes} argument.
2655 The normal return value from @code{fcntl} with this command is a
2656 nonnegative number which can be interpreted as the bitwise OR of the
2657 individual flags (except that currently there is only one flag to use).
2659 In case of an error, @code{fcntl} returns @math{-1}. The following
2660 @code{errno} error conditions are defined for this command:
2664 The @var{filedes} argument is invalid.
2671 @deftypevr Macro int F_SETFD
2672 This macro is used as the @var{command} argument to @code{fcntl}, to
2673 specify that it should set the file descriptor flags associated with the
2674 @var{filedes} argument. This requires a third @code{int} argument to
2675 specify the new flags, so the form of the call is:
2678 fcntl (@var{filedes}, F_SETFD, @var{new-flags})
2681 The normal return value from @code{fcntl} with this command is an
2682 unspecified value other than @math{-1}, which indicates an error.
2683 The flags and error conditions are the same as for the @code{F_GETFD}
2687 The following macro is defined for use as a file descriptor flag with
2688 the @code{fcntl} function. The value is an integer constant usable
2689 as a bit mask value.
2693 @deftypevr Macro int FD_CLOEXEC
2694 @cindex close-on-exec (file descriptor flag)
2695 This flag specifies that the file descriptor should be closed when
2696 an @code{exec} function is invoked; see @ref{Executing a File}. When
2697 a file descriptor is allocated (as with @code{open} or @code{dup}),
2698 this bit is initially cleared on the new file descriptor, meaning that
2699 descriptor will survive into the new program after @code{exec}.
2702 If you want to modify the file descriptor flags, you should get the
2703 current flags with @code{F_GETFD} and modify the value. Don't assume
2704 that the flags listed here are the only ones that are implemented; your
2705 program may be run years from now and more flags may exist then. For
2706 example, here is a function to set or clear the flag @code{FD_CLOEXEC}
2707 without altering any other flags:
2710 /* @r{Set the @code{FD_CLOEXEC} flag of @var{desc} if @var{value} is nonzero,}
2711 @r{or clear the flag if @var{value} is 0.}
2712 @r{Return 0 on success, or -1 on error with @code{errno} set.} */
2715 set_cloexec_flag (int desc, int value)
2717 int oldflags = fcntl (desc, F_GETFD, 0);
2718 /* @r{If reading the flags failed, return error indication now.}
2721 /* @r{Set just the flag we want to set.} */
2723 oldflags |= FD_CLOEXEC;
2725 oldflags &= ~FD_CLOEXEC;
2726 /* @r{Store modified flag word in the descriptor.} */
2727 return fcntl (desc, F_SETFD, oldflags);
2731 @node File Status Flags
2732 @section File Status Flags
2733 @cindex file status flags
2735 @dfn{File status flags} are used to specify attributes of the opening of a
2736 file. Unlike the file descriptor flags discussed in @ref{Descriptor
2737 Flags}, the file status flags are shared by duplicated file descriptors
2738 resulting from a single opening of the file. The file status flags are
2739 specified with the @var{flags} argument to @code{open};
2740 @pxref{Opening and Closing Files}.
2742 File status flags fall into three categories, which are described in the
2747 @ref{Access Modes}, specify what type of access is allowed to the
2748 file: reading, writing, or both. They are set by @code{open} and are
2749 returned by @code{fcntl}, but cannot be changed.
2752 @ref{Open-time Flags}, control details of what @code{open} will do.
2753 These flags are not preserved after the @code{open} call.
2756 @ref{Operating Modes}, affect how operations such as @code{read} and
2757 @code{write} are done. They are set by @code{open}, and can be fetched or
2758 changed with @code{fcntl}.
2761 The symbols in this section are defined in the header file
2766 * Access Modes:: Whether the descriptor can read or write.
2767 * Open-time Flags:: Details of @code{open}.
2768 * Operating Modes:: Special modes to control I/O operations.
2769 * Getting File Status Flags:: Fetching and changing these flags.
2773 @subsection File Access Modes
2775 The file access modes allow a file descriptor to be used for reading,
2776 writing, or both. (In the GNU system, they can also allow none of these,
2777 and allow execution of the file as a program.) The access modes are chosen
2778 when the file is opened, and never change.
2782 @deftypevr Macro int O_RDONLY
2783 Open the file for read access.
2788 @deftypevr Macro int O_WRONLY
2789 Open the file for write access.
2794 @deftypevr Macro int O_RDWR
2795 Open the file for both reading and writing.
2798 In the GNU system (and not in other systems), @code{O_RDONLY} and
2799 @code{O_WRONLY} are independent bits that can be bitwise-ORed together,
2800 and it is valid for either bit to be set or clear. This means that
2801 @code{O_RDWR} is the same as @code{O_RDONLY|O_WRONLY}. A file access
2802 mode of zero is permissible; it allows no operations that do input or
2803 output to the file, but does allow other operations such as
2804 @code{fchmod}. On the GNU system, since ``read-only'' or ``write-only''
2805 is a misnomer, @file{fcntl.h} defines additional names for the file
2806 access modes. These names are preferred when writing GNU-specific code.
2807 But most programs will want to be portable to other POSIX.1 systems and
2808 should use the POSIX.1 names above instead.
2812 @deftypevr Macro int O_READ
2813 Open the file for reading. Same as @code{O_RDWR}; only defined on GNU.
2818 @deftypevr Macro int O_WRITE
2819 Open the file for reading. Same as @code{O_WRONLY}; only defined on GNU.
2824 @deftypevr Macro int O_EXEC
2825 Open the file for executing. Only defined on GNU.
2828 To determine the file access mode with @code{fcntl}, you must extract
2829 the access mode bits from the retrieved file status flags. In the GNU
2830 system, you can just test the @code{O_READ} and @code{O_WRITE} bits in
2831 the flags word. But in other POSIX.1 systems, reading and writing
2832 access modes are not stored as distinct bit flags. The portable way to
2833 extract the file access mode bits is with @code{O_ACCMODE}.
2837 @deftypevr Macro int O_ACCMODE
2838 This macro stands for a mask that can be bitwise-ANDed with the file
2839 status flag value to produce a value representing the file access mode.
2840 The mode will be @code{O_RDONLY}, @code{O_WRONLY}, or @code{O_RDWR}.
2841 (In the GNU system it could also be zero, and it never includes the
2845 @node Open-time Flags
2846 @subsection Open-time Flags
2848 The open-time flags specify options affecting how @code{open} will behave.
2849 These options are not preserved once the file is open. The exception to
2850 this is @code{O_NONBLOCK}, which is also an I/O operating mode and so it
2851 @emph{is} saved. @xref{Opening and Closing Files}, for how to call
2854 There are two sorts of options specified by open-time flags.
2858 @dfn{File name translation flags} affect how @code{open} looks up the
2859 file name to locate the file, and whether the file can be created.
2860 @cindex file name translation flags
2861 @cindex flags, file name translation
2864 @dfn{Open-time action flags} specify extra operations that @code{open} will
2865 perform on the file once it is open.
2866 @cindex open-time action flags
2867 @cindex flags, open-time action
2870 Here are the file name translation flags.
2874 @deftypevr Macro int O_CREAT
2875 If set, the file will be created if it doesn't already exist.
2876 @c !!! mode arg, umask
2877 @cindex create on open (file status flag)
2882 @deftypevr Macro int O_EXCL
2883 If both @code{O_CREAT} and @code{O_EXCL} are set, then @code{open} fails
2884 if the specified file already exists. This is guaranteed to never
2885 clobber an existing file.
2890 @deftypevr Macro int O_NONBLOCK
2891 @cindex non-blocking open
2892 This prevents @code{open} from blocking for a ``long time'' to open the
2893 file. This is only meaningful for some kinds of files, usually devices
2894 such as serial ports; when it is not meaningful, it is harmless and
2895 ignored. Often opening a port to a modem blocks until the modem reports
2896 carrier detection; if @code{O_NONBLOCK} is specified, @code{open} will
2897 return immediately without a carrier.
2899 Note that the @code{O_NONBLOCK} flag is overloaded as both an I/O operating
2900 mode and a file name translation flag. This means that specifying
2901 @code{O_NONBLOCK} in @code{open} also sets nonblocking I/O mode;
2902 @pxref{Operating Modes}. To open the file without blocking but do normal
2903 I/O that blocks, you must call @code{open} with @code{O_NONBLOCK} set and
2904 then call @code{fcntl} to turn the bit off.
2909 @deftypevr Macro int O_NOCTTY
2910 If the named file is a terminal device, don't make it the controlling
2911 terminal for the process. @xref{Job Control}, for information about
2912 what it means to be the controlling terminal.
2914 In the GNU system and 4.4 BSD, opening a file never makes it the
2915 controlling terminal and @code{O_NOCTTY} is zero. However, other
2916 systems may use a nonzero value for @code{O_NOCTTY} and set the
2917 controlling terminal when you open a file that is a terminal device; so
2918 to be portable, use @code{O_NOCTTY} when it is important to avoid this.
2919 @cindex controlling terminal, setting
2922 The following three file name translation flags exist only in the GNU system.
2926 @deftypevr Macro int O_IGNORE_CTTY
2927 Do not recognize the named file as the controlling terminal, even if it
2928 refers to the process's existing controlling terminal device. Operations
2929 on the new file descriptor will never induce job control signals.
2935 @deftypevr Macro int O_NOLINK
2936 If the named file is a symbolic link, open the link itself instead of
2937 the file it refers to. (@code{fstat} on the new file descriptor will
2938 return the information returned by @code{lstat} on the link's name.)
2939 @cindex symbolic link, opening
2944 @deftypevr Macro int O_NOTRANS
2945 If the named file is specially translated, do not invoke the translator.
2946 Open the bare file the translator itself sees.
2950 The open-time action flags tell @code{open} to do additional operations
2951 which are not really related to opening the file. The reason to do them
2952 as part of @code{open} instead of in separate calls is that @code{open}
2953 can do them @i{atomically}.
2957 @deftypevr Macro int O_TRUNC
2958 Truncate the file to zero length. This option is only useful for
2959 regular files, not special files such as directories or FIFOs. POSIX.1
2960 requires that you open the file for writing to use @code{O_TRUNC}. In
2961 BSD and GNU you must have permission to write the file to truncate it,
2962 but you need not open for write access.
2964 This is the only open-time action flag specified by POSIX.1. There is
2965 no good reason for truncation to be done by @code{open}, instead of by
2966 calling @code{ftruncate} afterwards. The @code{O_TRUNC} flag existed in
2967 Unix before @code{ftruncate} was invented, and is retained for backward
2971 The remaining operating modes are BSD extensions. They exist only
2972 on some systems. On other systems, these macros are not defined.
2976 @deftypevr Macro int O_SHLOCK
2977 Acquire a shared lock on the file, as with @code{flock}.
2980 If @code{O_CREAT} is specified, the locking is done atomically when
2981 creating the file. You are guaranteed that no other process will get
2982 the lock on the new file first.
2987 @deftypevr Macro int O_EXLOCK
2988 Acquire an exclusive lock on the file, as with @code{flock}.
2989 @xref{File Locks}. This is atomic like @code{O_SHLOCK}.
2992 @node Operating Modes
2993 @subsection I/O Operating Modes
2995 The operating modes affect how input and output operations using a file
2996 descriptor work. These flags are set by @code{open} and can be fetched
2997 and changed with @code{fcntl}.
3001 @deftypevr Macro int O_APPEND
3002 The bit that enables append mode for the file. If set, then all
3003 @code{write} operations write the data at the end of the file, extending
3004 it, regardless of the current file position. This is the only reliable
3005 way to append to a file. In append mode, you are guaranteed that the
3006 data you write will always go to the current end of the file, regardless
3007 of other processes writing to the file. Conversely, if you simply set
3008 the file position to the end of file and write, then another process can
3009 extend the file after you set the file position but before you write,
3010 resulting in your data appearing someplace before the real end of file.
3015 @deftypevr Macro int O_NONBLOCK
3016 The bit that enables nonblocking mode for the file. If this bit is set,
3017 @code{read} requests on the file can return immediately with a failure
3018 status if there is no input immediately available, instead of blocking.
3019 Likewise, @code{write} requests can also return immediately with a
3020 failure status if the output can't be written immediately.
3022 Note that the @code{O_NONBLOCK} flag is overloaded as both an I/O
3023 operating mode and a file name translation flag; @pxref{Open-time Flags}.
3028 @deftypevr Macro int O_NDELAY
3029 This is an obsolete name for @code{O_NONBLOCK}, provided for
3030 compatibility with BSD. It is not defined by the POSIX.1 standard.
3033 The remaining operating modes are BSD and GNU extensions. They exist only
3034 on some systems. On other systems, these macros are not defined.
3038 @deftypevr Macro int O_ASYNC
3039 The bit that enables asynchronous input mode. If set, then @code{SIGIO}
3040 signals will be generated when input is available. @xref{Interrupt Input}.
3042 Asynchronous input mode is a BSD feature.
3047 @deftypevr Macro int O_FSYNC
3048 The bit that enables synchronous writing for the file. If set, each
3049 @code{write} call will make sure the data is reliably stored on disk before
3050 returning. @c !!! xref fsync
3052 Synchronous writing is a BSD feature.
3057 @deftypevr Macro int O_SYNC
3058 This is another name for @code{O_FSYNC}. They have the same value.
3063 @deftypevr Macro int O_NOATIME
3064 If this bit is set, @code{read} will not update the access time of the
3065 file. @xref{File Times}. This is used by programs that do backups, so
3066 that backing a file up does not count as reading it.
3067 Only the owner of the file or the superuser may use this bit.
3069 This is a GNU extension.
3072 @node Getting File Status Flags
3073 @subsection Getting and Setting File Status Flags
3075 The @code{fcntl} function can fetch or change file status flags.
3079 @deftypevr Macro int F_GETFL
3080 This macro is used as the @var{command} argument to @code{fcntl}, to
3081 read the file status flags for the open file with descriptor
3084 The normal return value from @code{fcntl} with this command is a
3085 nonnegative number which can be interpreted as the bitwise OR of the
3086 individual flags. Since the file access modes are not single-bit values,
3087 you can mask off other bits in the returned flags with @code{O_ACCMODE}
3090 In case of an error, @code{fcntl} returns @math{-1}. The following
3091 @code{errno} error conditions are defined for this command:
3095 The @var{filedes} argument is invalid.
3101 @deftypevr Macro int F_SETFL
3102 This macro is used as the @var{command} argument to @code{fcntl}, to set
3103 the file status flags for the open file corresponding to the
3104 @var{filedes} argument. This command requires a third @code{int}
3105 argument to specify the new flags, so the call looks like this:
3108 fcntl (@var{filedes}, F_SETFL, @var{new-flags})
3111 You can't change the access mode for the file in this way; that is,
3112 whether the file descriptor was opened for reading or writing.
3114 The normal return value from @code{fcntl} with this command is an
3115 unspecified value other than @math{-1}, which indicates an error. The
3116 error conditions are the same as for the @code{F_GETFL} command.
3119 If you want to modify the file status flags, you should get the current
3120 flags with @code{F_GETFL} and modify the value. Don't assume that the
3121 flags listed here are the only ones that are implemented; your program
3122 may be run years from now and more flags may exist then. For example,
3123 here is a function to set or clear the flag @code{O_NONBLOCK} without
3124 altering any other flags:
3128 /* @r{Set the @code{O_NONBLOCK} flag of @var{desc} if @var{value} is nonzero,}
3129 @r{or clear the flag if @var{value} is 0.}
3130 @r{Return 0 on success, or -1 on error with @code{errno} set.} */
3133 set_nonblock_flag (int desc, int value)
3135 int oldflags = fcntl (desc, F_GETFL, 0);
3136 /* @r{If reading the flags failed, return error indication now.} */
3139 /* @r{Set just the flag we want to set.} */
3141 oldflags |= O_NONBLOCK;
3143 oldflags &= ~O_NONBLOCK;
3144 /* @r{Store modified flag word in the descriptor.} */
3145 return fcntl (desc, F_SETFL, oldflags);
3154 @cindex record locking
3155 The remaining @code{fcntl} commands are used to support @dfn{record
3156 locking}, which permits multiple cooperating programs to prevent each
3157 other from simultaneously accessing parts of a file in error-prone
3160 @cindex exclusive lock
3162 An @dfn{exclusive} or @dfn{write} lock gives a process exclusive access
3163 for writing to the specified part of the file. While a write lock is in
3164 place, no other process can lock that part of the file.
3168 A @dfn{shared} or @dfn{read} lock prohibits any other process from
3169 requesting a write lock on the specified part of the file. However,
3170 other processes can request read locks.
3172 The @code{read} and @code{write} functions do not actually check to see
3173 whether there are any locks in place. If you want to implement a
3174 locking protocol for a file shared by multiple processes, your application
3175 must do explicit @code{fcntl} calls to request and clear locks at the
3178 Locks are associated with processes. A process can only have one kind
3179 of lock set for each byte of a given file. When any file descriptor for
3180 that file is closed by the process, all of the locks that process holds
3181 on that file are released, even if the locks were made using other
3182 descriptors that remain open. Likewise, locks are released when a
3183 process exits, and are not inherited by child processes created using
3184 @code{fork} (@pxref{Creating a Process}).
3186 When making a lock, use a @code{struct flock} to specify what kind of
3187 lock and where. This data type and the associated macros for the
3188 @code{fcntl} function are declared in the header file @file{fcntl.h}.
3193 @deftp {Data Type} {struct flock}
3194 This structure is used with the @code{fcntl} function to describe a file
3195 lock. It has these members:
3198 @item short int l_type
3199 Specifies the type of the lock; one of @code{F_RDLCK}, @code{F_WRLCK}, or
3202 @item short int l_whence
3203 This corresponds to the @var{whence} argument to @code{fseek} or
3204 @code{lseek}, and specifies what the offset is relative to. Its value
3205 can be one of @code{SEEK_SET}, @code{SEEK_CUR}, or @code{SEEK_END}.
3208 This specifies the offset of the start of the region to which the lock
3209 applies, and is given in bytes relative to the point specified by
3210 @code{l_whence} member.
3213 This specifies the length of the region to be locked. A value of
3214 @code{0} is treated specially; it means the region extends to the end of
3218 This field is the process ID (@pxref{Process Creation Concepts}) of the
3219 process holding the lock. It is filled in by calling @code{fcntl} with
3220 the @code{F_GETLK} command, but is ignored when making a lock.
3226 @deftypevr Macro int F_GETLK
3227 This macro is used as the @var{command} argument to @code{fcntl}, to
3228 specify that it should get information about a lock. This command
3229 requires a third argument of type @w{@code{struct flock *}} to be passed
3230 to @code{fcntl}, so that the form of the call is:
3233 fcntl (@var{filedes}, F_GETLK, @var{lockp})
3236 If there is a lock already in place that would block the lock described
3237 by the @var{lockp} argument, information about that lock overwrites
3238 @code{*@var{lockp}}. Existing locks are not reported if they are
3239 compatible with making a new lock as specified. Thus, you should
3240 specify a lock type of @code{F_WRLCK} if you want to find out about both
3241 read and write locks, or @code{F_RDLCK} if you want to find out about
3244 There might be more than one lock affecting the region specified by the
3245 @var{lockp} argument, but @code{fcntl} only returns information about
3246 one of them. The @code{l_whence} member of the @var{lockp} structure is
3247 set to @code{SEEK_SET} and the @code{l_start} and @code{l_len} fields
3248 set to identify the locked region.
3250 If no lock applies, the only change to the @var{lockp} structure is to
3251 update the @code{l_type} to a value of @code{F_UNLCK}.
3253 The normal return value from @code{fcntl} with this command is an
3254 unspecified value other than @math{-1}, which is reserved to indicate an
3255 error. The following @code{errno} error conditions are defined for
3260 The @var{filedes} argument is invalid.
3263 Either the @var{lockp} argument doesn't specify valid lock information,
3264 or the file associated with @var{filedes} doesn't support locks.
3270 @deftypevr Macro int F_SETLK
3271 This macro is used as the @var{command} argument to @code{fcntl}, to
3272 specify that it should set or clear a lock. This command requires a
3273 third argument of type @w{@code{struct flock *}} to be passed to
3274 @code{fcntl}, so that the form of the call is:
3277 fcntl (@var{filedes}, F_SETLK, @var{lockp})
3280 If the process already has a lock on any part of the region, the old lock
3281 on that part is replaced with the new lock. You can remove a lock
3282 by specifying a lock type of @code{F_UNLCK}.
3284 If the lock cannot be set, @code{fcntl} returns immediately with a value
3285 of @math{-1}. This function does not block waiting for other processes
3286 to release locks. If @code{fcntl} succeeds, it return a value other
3289 The following @code{errno} error conditions are defined for this
3295 The lock cannot be set because it is blocked by an existing lock on the
3296 file. Some systems use @code{EAGAIN} in this case, and other systems
3297 use @code{EACCES}; your program should treat them alike, after
3298 @code{F_SETLK}. (The GNU system always uses @code{EAGAIN}.)
3301 Either: the @var{filedes} argument is invalid; you requested a read lock
3302 but the @var{filedes} is not open for read access; or, you requested a
3303 write lock but the @var{filedes} is not open for write access.
3306 Either the @var{lockp} argument doesn't specify valid lock information,
3307 or the file associated with @var{filedes} doesn't support locks.
3310 The system has run out of file lock resources; there are already too
3311 many file locks in place.
3313 Well-designed file systems never report this error, because they have no
3314 limitation on the number of locks. However, you must still take account
3315 of the possibility of this error, as it could result from network access
3316 to a file system on another machine.
3322 @deftypevr Macro int F_SETLKW
3323 This macro is used as the @var{command} argument to @code{fcntl}, to
3324 specify that it should set or clear a lock. It is just like the
3325 @code{F_SETLK} command, but causes the process to block (or wait)
3326 until the request can be specified.
3328 This command requires a third argument of type @code{struct flock *}, as
3329 for the @code{F_SETLK} command.
3331 The @code{fcntl} return values and errors are the same as for the
3332 @code{F_SETLK} command, but these additional @code{errno} error conditions
3333 are defined for this command:
3337 The function was interrupted by a signal while it was waiting.
3338 @xref{Interrupted Primitives}.
3341 The specified region is being locked by another process. But that
3342 process is waiting to lock a region which the current process has
3343 locked, so waiting for the lock would result in deadlock. The system
3344 does not guarantee that it will detect all such conditions, but it lets
3345 you know if it notices one.
3350 The following macros are defined for use as values for the @code{l_type}
3351 member of the @code{flock} structure. The values are integer constants.
3358 This macro is used to specify a read (or shared) lock.
3364 This macro is used to specify a write (or exclusive) lock.
3370 This macro is used to specify that the region is unlocked.
3373 As an example of a situation where file locking is useful, consider a
3374 program that can be run simultaneously by several different users, that
3375 logs status information to a common file. One example of such a program
3376 might be a game that uses a file to keep track of high scores. Another
3377 example might be a program that records usage or accounting information
3378 for billing purposes.
3380 Having multiple copies of the program simultaneously writing to the
3381 file could cause the contents of the file to become mixed up. But
3382 you can prevent this kind of problem by setting a write lock on the
3383 file before actually writing to the file.
3385 If the program also needs to read the file and wants to make sure that
3386 the contents of the file are in a consistent state, then it can also use
3387 a read lock. While the read lock is set, no other process can lock
3388 that part of the file for writing.
3390 @c ??? This section could use an example program.
3392 Remember that file locks are only a @emph{voluntary} protocol for
3393 controlling access to a file. There is still potential for access to
3394 the file by programs that don't use the lock protocol.
3396 @node Interrupt Input
3397 @section Interrupt-Driven Input
3399 @cindex interrupt-driven input
3400 If you set the @code{O_ASYNC} status flag on a file descriptor
3401 (@pxref{File Status Flags}), a @code{SIGIO} signal is sent whenever
3402 input or output becomes possible on that file descriptor. The process
3403 or process group to receive the signal can be selected by using the
3404 @code{F_SETOWN} command to the @code{fcntl} function. If the file
3405 descriptor is a socket, this also selects the recipient of @code{SIGURG}
3406 signals that are delivered when out-of-band data arrives on that socket;
3407 see @ref{Out-of-Band Data}. (@code{SIGURG} is sent in any situation
3408 where @code{select} would report the socket as having an ``exceptional
3409 condition''. @xref{Waiting for I/O}.)
3411 If the file descriptor corresponds to a terminal device, then @code{SIGIO}
3412 signals are sent to the foreground process group of the terminal.
3416 The symbols in this section are defined in the header file
3421 @deftypevr Macro int F_GETOWN
3422 This macro is used as the @var{command} argument to @code{fcntl}, to
3423 specify that it should get information about the process or process
3424 group to which @code{SIGIO} signals are sent. (For a terminal, this is
3425 actually the foreground process group ID, which you can get using
3426 @code{tcgetpgrp}; see @ref{Terminal Access Functions}.)
3428 The return value is interpreted as a process ID; if negative, its
3429 absolute value is the process group ID.
3431 The following @code{errno} error condition is defined for this command:
3435 The @var{filedes} argument is invalid.
3441 @deftypevr Macro int F_SETOWN
3442 This macro is used as the @var{command} argument to @code{fcntl}, to
3443 specify that it should set the process or process group to which
3444 @code{SIGIO} signals are sent. This command requires a third argument
3445 of type @code{pid_t} to be passed to @code{fcntl}, so that the form of
3449 fcntl (@var{filedes}, F_SETOWN, @var{pid})
3452 The @var{pid} argument should be a process ID. You can also pass a
3453 negative number whose absolute value is a process group ID.
3455 The return value from @code{fcntl} with this command is @math{-1}
3456 in case of error and some other value if successful. The following
3457 @code{errno} error conditions are defined for this command:
3461 The @var{filedes} argument is invalid.
3464 There is no process or process group corresponding to @var{pid}.
3468 @c ??? This section could use an example program.
3471 @section Generic I/O Control operations
3472 @cindex generic i/o control operations
3475 The GNU system can handle most input/output operations on many different
3476 devices and objects in terms of a few file primitives - @code{read},
3477 @code{write} and @code{lseek}. However, most devices also have a few
3478 peculiar operations which do not fit into this model. Such as:
3483 Changing the character font used on a terminal.
3486 Telling a magnetic tape system to rewind or fast forward. (Since they
3487 cannot move in byte increments, @code{lseek} is inapplicable).
3490 Ejecting a disk from a drive.
3493 Playing an audio track from a CD-ROM drive.
3496 Maintaining routing tables for a network.
3500 Although some such objects such as sockets and terminals
3501 @footnote{Actually, the terminal-specific functions are implemented with
3502 IOCTLs on many platforms.} have special functions of their own, it would
3503 not be practical to create functions for all these cases.
3505 Instead these minor operations, known as @dfn{IOCTL}s, are assigned code
3506 numbers and multiplexed through the @code{ioctl} function, defined in
3507 @code{sys/ioctl.h}. The code numbers themselves are defined in many
3510 @deftypefun int ioctl (int @var{filedes}, int @var{command}, @dots{})
3512 The @code{ioctl} function performs the generic I/O operation
3513 @var{command} on @var{filedes}.
3515 A third argument is usually present, either a single number or a pointer
3516 to a structure. The meaning of this argument, the returned value, and
3517 any error codes depends upon the command used. Often @math{-1} is
3518 returned for a failure.
3522 On some systems, IOCTLs used by different devices share the same numbers.
3523 Thus, although use of an inappropriate IOCTL @emph{usually} only produces
3524 an error, you should not attempt to use device-specific IOCTLs on an
3527 Most IOCTLs are OS-specific and/or only used in special system utilities,
3528 and are thus beyond the scope of this document. For an example of the use
3529 of an IOCTL, see @ref{Out-of-Band Data}.