13. Low-Level Input/Output

This chapter describes functions for performing low-level input/output operations on file descriptors. These functions include the primitives for the higher-level I/O functions described in Input/Output on Streams, as well as functions for performing low-level control operations for which there are no equivalents on streams.

Stream-level I/O is more flexible and usually more convenient; therefore, programmers generally use the descriptor-level functions only when necessary. These are some of the usual reasons:

• For reading binary files in large chunks.
• For reading an entire file into core before parsing it.
• To perform operations other than data transfer, which can only be done with a descriptor. (You can use fileno to get the descriptor corresponding to a stream.)
• To pass descriptors to a child process. (The child can create its own stream to use a descriptor that it inherits, but cannot inherit a stream directly.)

13.1 Opening and Closing Files

This section describes the primitives for opening and closing files using file descriptors. The open and creat functions are declared in the header file ‘fcntl.h’, while close is declared in ‘unistd.h’.

Function: int open (const char *filename, int flags[, mode_t mode])

The open function creates and returns a new file descriptor for the file named by filename. Initially, the file position indicator for the file is at the beginning of the file. The argument mode is used only when a file is created, but it doesn't hurt to supply the argument in any case.

The flags argument controls how the file is to be opened. This is a bit mask; you create the value by the bitwise OR of the appropriate parameters (using the ‘|’ operator in C). See section File Status Flags, for the parameters available.

The normal return value from open is a non-negative integer file descriptor. In the case of an error, a value of -1 is returned instead. In addition to the usual file name errors (see section File Name Errors), the following errno error conditions are defined for this function:

EACCES

The file exists but is not readable/writable as requested by the flags argument, the file does not exist and the directory is unwritable so it cannot be created.

EEXIST

Both O_CREAT and O_EXCL are set, and the named file already exists.

EINTR

The open operation was interrupted by a signal. See section Primitives Interrupted by Signals.

EISDIR

The flags argument specified write access, and the file is a directory.

EMFILE

The process has too many files open. The maximum number of file descriptors is controlled by the RLIMIT_NOFILE resource limit; see section Limiting Resource Usage.

ENFILE

The entire system, or perhaps the file system which contains the directory, cannot support any additional open files at the moment. (This problem cannot happen on the GNU system.)

ENOENT

The named file does not exist, and O_CREAT is not specified.

ENOSPC

The directory or file system that would contain the new file cannot be extended, because there is no disk space left.

ENXIO

O_NONBLOCK and O_WRONLY are both set in the flags argument, the file named by filename is a FIFO (see section Pipes and FIFOs), and no process has the file open for reading.

EROFS

The file resides on a read-only file system and any of O_WRONLY, O_RDWR, and O_TRUNC are set in the flags argument, or O_CREAT is set and the file does not already exist.

If on a 32 bit machine the sources are translated with _FILE_OFFSET_BITS == 64 the function open returns a file descriptor opened in the large file mode which enables the file handling functions to use files up to 2^63 bytes in size and offset from -2^63 to 2^63. This happens transparently for the user since all of the lowlevel file handling functions are equally replaced.

This function is a cancellation point in multi-threaded programs. This is a problem if the thread allocates some resources (like memory, file descriptors, semaphores or whatever) at the time open is called. If the thread gets canceled these resources stay allocated until the program ends. To avoid this calls to open should be protected using cancellation handlers.

The open function is the underlying primitive for the fopen and freopen functions, that create streams.

Function: int open64 (const char *filename, int flags[, mode_t mode])

This function is similar to open. It returns a file descriptor which can be used to access the file named by filename. The only difference is that on 32 bit systems the file is opened in the large file mode. I.e., file length and file offsets can exceed 31 bits.

When the sources are translated with _FILE_OFFSET_BITS == 64 this function is actually available under the name open. I.e., the new, extended API using 64 bit file sizes and offsets transparently replaces the old API.

Obsolete function: int creat (const char *filename, mode_t mode)

This function is obsolete. The call:

creat (filename, mode)

is equivalent to:

open (filename, O_WRONLY | O_CREAT | O_TRUNC, mode)

If on a 32 bit machine the sources are translated with _FILE_OFFSET_BITS == 64 the function creat returns a file descriptor opened in the large file mode which enables the file handling functions to use files up to 2^63 in size and offset from -2^63 to 2^63. This happens transparently for the user since all of the lowlevel file handling functions are equally replaced.

Obsolete function: int creat64 (const char *filename, mode_t mode)

This function is similar to creat. It returns a file descriptor which can be used to access the file named by filename. The only the difference is that on 32 bit systems the file is opened in the large file mode. I.e., file length and file offsets can exceed 31 bits.

To use this file descriptor one must not use the normal operations but instead the counterparts named *64, e.g., read64.

When the sources are translated with _FILE_OFFSET_BITS == 64 this function is actually available under the name open. I.e., the new, extended API using 64 bit file sizes and offsets transparently replaces the old API.

Function: int close (int filedes)

The function close closes the file descriptor filedes. Closing a file has the following consequences:

• The file descriptor is deallocated.
• Any record locks owned by the process on the file are unlocked.
• When all file descriptors associated with a pipe or FIFO have been closed, any unread data is discarded.

This function is a cancellation point in multi-threaded programs. This is a problem if the thread allocates some resources (like memory, file descriptors, semaphores or whatever) at the time close is called. If the thread gets canceled these resources stay allocated until the program ends. To avoid this, calls to close should be protected using cancellation handlers.

The normal return value from close is 0; a value of -1 is returned in case of failure. The following errno error conditions are defined for this function:

EBADF

The filedes argument is not a valid file descriptor.

EINTR

The close call was interrupted by a signal. See section Primitives Interrupted by Signals. Here is an example of how to handle EINTR properly:

TEMP_FAILURE_RETRY (close (desc));

ENOSPC

EIO

EDQUOT

When the file is accessed by NFS, these errors from write can sometimes not be detected until close. See section Input and Output Primitives, for details on their meaning.

Please note that there is no separate close64 function. This is not necessary since this function does not determine nor depend on the mode of the file. The kernel which performs the close operation knows which mode the descriptor is used for and can handle this situation.

To close a stream, call fclose (see section Closing Streams) instead of trying to close its underlying file descriptor with close. This flushes any buffered output and updates the stream object to indicate that it is closed.

13.2 Input and Output Primitives

This section describes the functions for performing primitive input and output operations on file descriptors: read, write, and lseek. These functions are declared in the header file ‘unistd.h’.

Data Type: ssize_t

This data type is used to represent the sizes of blocks that can be read or written in a single operation. It is similar to size_t, but must be a signed type.

Function: ssize_t read (int filedes, void *buffer, size_t size)

The read function reads up to size bytes from the file with descriptor filedes, storing the results in the buffer. (This is not necessarily a character string, and no terminating null character is added.)

The return value is the number of bytes actually read. This might be less than size; for example, if there aren't that many bytes left in the file or if there aren't that many bytes immediately available. The exact behavior depends on what kind of file it is. Note that reading less than size bytes is not an error.

A value of zero indicates end-of-file (except if the value of the size argument is also zero). This is not considered an error. If you keep calling read while at end-of-file, it will keep returning zero and doing nothing else.

If read returns at least one character, there is no way you can tell whether end-of-file was reached. But if you did reach the end, the next read will return zero.

In case of an error, read returns -1. The following errno error conditions are defined for this function:

EAGAIN

Normally, when no input is immediately available, read waits for some input. But if the O_NONBLOCK flag is set for the file (see section File Status Flags), read returns immediately without reading any data, and reports this error.

Compatibility Note: Most versions of BSD Unix use a different error code for this: EWOULDBLOCK. In the GNU library, EWOULDBLOCK is an alias for EAGAIN, so it doesn't matter which name you use.

On some systems, reading a large amount of data from a character special file can also fail with EAGAIN if the kernel cannot find enough physical memory to lock down the user's pages. This is limited to devices that transfer with direct memory access into the user's memory, which means it does not include terminals, since they always use separate buffers inside the kernel. This problem never happens in the GNU system.

Any condition that could result in EAGAIN can instead result in a successful read which returns fewer bytes than requested. Calling read again immediately would result in EAGAIN.

EBADF

The filedes argument is not a valid file descriptor, or is not open for reading.

EINTR

read was interrupted by a signal while it was waiting for input. See section Primitives Interrupted by Signals. A signal will not necessary cause read to return EINTR; it may instead result in a successful read which returns fewer bytes than requested.

EIO

For many devices, and for disk files, this error code indicates a hardware error.

EIO also occurs when a background process tries to read from the controlling terminal, and the normal action of stopping the process by sending it a SIGTTIN signal isn't working. This might happen if the signal is being blocked or ignored, or because the process group is orphaned. See section Job Control, for more information about job control, and Signal Handling, for information about signals.

EINVAL

In some systems, when reading from a character or block device, position and size offsets must be aligned to a particular block size. This error indicates that the offsets were not properly aligned.

Please note that there is no function named read64. This is not necessary since this function does not directly modify or handle the possibly wide file offset. Since the kernel handles this state internally, the read function can be used for all cases.

This function is a cancellation point in multi-threaded programs. This is a problem if the thread allocates some resources (like memory, file descriptors, semaphores or whatever) at the time read is called. If the thread gets canceled these resources stay allocated until the program ends. To avoid this, calls to read should be protected using cancellation handlers.

The read function is the underlying primitive for all of the functions that read from streams, such as fgetc.

Function: ssize_t pread (int filedes, void *buffer, size_t size, off_t offset)

The pread function is similar to the read function. The first three arguments are identical, and the return values and error codes also correspond.

The difference is the fourth argument and its handling. The data block is not read from the current position of the file descriptor filedes. Instead the data is read from the file starting at position offset. The position of the file descriptor itself is not affected by the operation. The value is the same as before the call.

When the source file is compiled with _FILE_OFFSET_BITS == 64 the pread function is in fact pread64 and the type off_t has 64 bits, which makes it possible to handle files up to 2^63 bytes in length.

The return value of pread describes the number of bytes read. In the error case it returns -1 like read does and the error codes are also the same, with these additions:

EINVAL

The value given for offset is negative and therefore illegal.

ESPIPE

The file descriptor filedes is associate with a pipe or a FIFO and this device does not allow positioning of the file pointer.

The function is an extension defined in the Unix Single Specification version 2.

Function: ssize_t pread64 (int filedes, void *buffer, size_t size, off64_t offset)

This function is similar to the pread function. The difference is that the offset parameter is of type off64_t instead of off_t which makes it possible on 32 bit machines to address files larger than 2^31 bytes and up to 2^63 bytes. The file descriptor filedes must be opened using open64 since otherwise the large offsets possible with off64_t will lead to errors with a descriptor in small file mode.

When the source file is compiled with _FILE_OFFSET_BITS == 64 on a 32 bit machine this function is actually available under the name pread and so transparently replaces the 32 bit interface.

Function: ssize_t write (int filedes, const void *buffer, size_t size)

The write function writes up to size bytes from buffer to the file with descriptor filedes. The data in buffer is not necessarily a character string and a null character is output like any other character.

The return value is the number of bytes actually written. This may be size, but can always be smaller. Your program should always call write in a loop, iterating until all the data is written.

Once write returns, the data is enqueued to be written and can be read back right away, but it is not necessarily written out to permanent storage immediately. You can use fsync when you need to be sure your data has been permanently stored before continuing. (It is more efficient for the system to batch up consecutive writes and do them all at once when convenient. Normally they will always be written to disk within a minute or less.) Modern systems provide another function fdatasync which guarantees integrity only for the file data and is therefore faster. You can use the O_FSYNC open mode to make write always store the data to disk before returning; see section I/O Operating Modes.

In the case of an error, write returns -1. The following errno error conditions are defined for this function:

EAGAIN

Normally, write blocks until the write operation is complete. But if the O_NONBLOCK flag is set for the file (see section Control Operations on Files), it returns immediately without writing any data and reports this error. An example of a situation that might cause the process to block on output is writing to a terminal device that supports flow control, where output has been suspended by receipt of a STOP character.

Compatibility Note: Most versions of BSD Unix use a different error code for this: EWOULDBLOCK. In the GNU library, EWOULDBLOCK is an alias for EAGAIN, so it doesn't matter which name you use.

On some systems, writing a large amount of data from a character special file can also fail with EAGAIN if the kernel cannot find enough physical memory to lock down the user's pages. This is limited to devices that transfer with direct memory access into the user's memory, which means it does not include terminals, since they always use separate buffers inside the kernel. This problem does not arise in the GNU system.

EBADF

The filedes argument is not a valid file descriptor, or is not open for writing.

EFBIG

The size of the file would become larger than the implementation can support.

EINTR

The write operation was interrupted by a signal while it was blocked waiting for completion. A signal will not necessarily cause write to return EINTR; it may instead result in a successful write which writes fewer bytes than requested. See section Primitives Interrupted by Signals.

EIO

For many devices, and for disk files, this error code indicates a hardware error.

ENOSPC

The device containing the file is full.

EPIPE

This error is returned when you try to write to a pipe or FIFO that isn't open for reading by any process. When this happens, a SIGPIPE signal is also sent to the process; see Signal Handling.

EINVAL

In some systems, when writing to a character or block device, position and size offsets must be aligned to a particular block size. This error indicates that the offsets were not properly aligned.

Unless you have arranged to prevent EINTR failures, you should check errno after each failing call to write, and if the error was EINTR, you should simply repeat the call. See section Primitives Interrupted by Signals. The easy way to do this is with the macro TEMP_FAILURE_RETRY, as follows:

nbytes = TEMP_FAILURE_RETRY (write (desc, buffer, count));

Please note that there is no function named write64. This is not necessary since this function does not directly modify or handle the possibly wide file offset. Since the kernel handles this state internally the write function can be used for all cases.

This function is a cancellation point in multi-threaded programs. This is a problem if the thread allocates some resources (like memory, file descriptors, semaphores or whatever) at the time write is called. If the thread gets canceled these resources stay allocated until the program ends. To avoid this, calls to write should be protected using cancellation handlers.

The write function is the underlying primitive for all of the functions that write to streams, such as fputc.

Function: ssize_t pwrite (int filedes, const void *buffer, size_t size, off_t offset)

The pwrite function is similar to the write function. The first three arguments are identical, and the return values and error codes also correspond.

The difference is the fourth argument and its handling. The data block is not written to the current position of the file descriptor filedes. Instead the data is written to the file starting at position offset. The position of the file descriptor itself is not affected by the operation. The value is the same as before the call.

When the source file is compiled with _FILE_OFFSET_BITS == 64 the pwrite function is in fact pwrite64 and the type off_t has 64 bits, which makes it possible to handle files up to 2^63 bytes in length.

The return value of pwrite describes the number of written bytes. In the error case it returns -1 like write does and the error codes are also the same, with these additions:

EINVAL

The value given for offset is negative and therefore illegal.

ESPIPE

The file descriptor filedes is associated with a pipe or a FIFO and this device does not allow positioning of the file pointer.

The function is an extension defined in the Unix Single Specification version 2.

Function: ssize_t pwrite64 (int filedes, const void *buffer, size_t size, off64_t offset)

This function is similar to the pwrite function. The difference is that the offset parameter is of type off64_t instead of off_t which makes it possible on 32 bit machines to address files larger than 2^31 bytes and up to 2^63 bytes. The file descriptor filedes must be opened using open64 since otherwise the large offsets possible with off64_t will lead to errors with a descriptor in small file mode.

When the source file is compiled using _FILE_OFFSET_BITS == 64 on a 32 bit machine this function is actually available under the name pwrite and so transparently replaces the 32 bit interface.

13.3 Setting the File Position of a Descriptor

Just as you can set the file position of a stream with fseek, you can set the file position of a descriptor with lseek. This specifies the position in the file for the next read or write operation. See section File Positioning, for more information on the file position and what it means.

To read the current file position value from a descriptor, use lseek (desc, 0, SEEK_CUR).

Function: off_t lseek (int filedes, off_t offset, int whence)

The lseek function is used to change the file position of the file with descriptor filedes.

The whence argument specifies how the offset should be interpreted, in the same way as for the fseek function, and it must be one of the symbolic constants SEEK_SET, SEEK_CUR, or SEEK_END.

SEEK_SET

Specifies that whence is a count of characters from the beginning of the file.

SEEK_CUR

Specifies that whence is a count of characters from the current file position. This count may be positive or negative.

SEEK_END

Specifies that whence is a count of characters from the end of the file. A negative count specifies a position within the current extent of the file; a positive count specifies a position past the current end. If you set the position past the current end, and actually write data, you will extend the file with zeros up to that position.

The return value from lseek is normally the resulting file position, measured in bytes from the beginning of the file. You can use this feature together with SEEK_CUR to read the current file position.

If you want to append to the file, setting the file position to the current end of file with SEEK_END is not sufficient. Another process may write more data after you seek but before you write, extending the file so the position you write onto clobbers their data. Instead, use the O_APPEND operating mode; see section I/O Operating Modes.

You can set the file position past the current end of the file. This does not by itself make the file longer; lseek never changes the file. But subsequent output at that position will extend the file. Characters between the previous end of file and the new position are filled with zeros. Extending the file in this way can create a “hole”: the blocks of zeros are not actually allocated on disk, so the file takes up less space than it appears to; it is then called a “sparse file”.

If the file position cannot be changed, or the operation is in some way invalid, lseek returns a value of -1. The following errno error conditions are defined for this function:

EBADF

The filedes is not a valid file descriptor.

EINVAL

The whence argument value is not valid, or the resulting file offset is not valid. A file offset is invalid.

ESPIPE

The filedes corresponds to an object that cannot be positioned, such as a pipe, FIFO or terminal device. (POSIX.1 specifies this error only for pipes and FIFOs, but in the GNU system, you always get ESPIPE if the object is not seekable.)

When the source file is compiled with _FILE_OFFSET_BITS == 64 the lseek function is in fact lseek64 and the type off_t has 64 bits which makes it possible to handle files up to 2^63 bytes in length.

This function is a cancellation point in multi-threaded programs. This is a problem if the thread allocates some resources (like memory, file descriptors, semaphores or whatever) at the time lseek is called. If the thread gets canceled these resources stay allocated until the program ends. To avoid this calls to lseek should be protected using cancellation handlers.

The lseek function is the underlying primitive for the fseek, fseeko, ftell, ftello and rewind functions, which operate on streams instead of file descriptors.

Function: off64_t lseek64 (int filedes, off64_t offset, int whence)

This function is similar to the lseek function. The difference is that the offset parameter is of type off64_t instead of off_t which makes it possible on 32 bit machines to address files larger than 2^31 bytes and up to 2^63 bytes. The file descriptor filedes must be opened using open64 since otherwise the large offsets possible with off64_t will lead to errors with a descriptor in small file mode.

When the source file is compiled with _FILE_OFFSET_BITS == 64 on a 32 bits machine this function is actually available under the name lseek and so transparently replaces the 32 bit interface.

You can have multiple descriptors for the same file if you open the file more than once, or if you duplicate a descriptor with dup. Descriptors that come from separate calls to open have independent file positions; using lseek on one descriptor has no effect on the other. For example,

{
  int d1, d2;
  char buf[4];
  d1 = open ("foo", O_RDONLY);
  d2 = open ("foo", O_RDONLY);
  lseek (d1, 1024, SEEK_SET);
  read (d2, buf, 4);
}

will read the first four characters of the file ‘foo’. (The error-checking code necessary for a real program has been omitted here for brevity.)

By contrast, descriptors made by duplication share a common file position with the original descriptor that was duplicated. Anything which alters the file position of one of the duplicates, including reading or writing data, affects all of them alike. Thus, for example,

{
  int d1, d2, d3;
  char buf1[4], buf2[4];
  d1 = open ("foo", O_RDONLY);
  d2 = dup (d1);
  d3 = dup (d2);
  lseek (d3, 1024, SEEK_SET);
  read (d1, buf1, 4);
  read (d2, buf2, 4);
}

will read four characters starting with the 1024'th character of ‘foo’, and then four more characters starting with the 1028'th character.

Data Type: off_t

This is an arithmetic data type used to represent file sizes. In the GNU system, this is equivalent to fpos_t or long int.

If the source is compiled with _FILE_OFFSET_BITS == 64 this type is transparently replaced by off64_t.

Data Type: off64_t

This type is used similar to off_t. The difference is that even on 32 bit machines, where the off_t type would have 32 bits, off64_t has 64 bits and so is able to address files up to 2^63 bytes in length.

When compiling with _FILE_OFFSET_BITS == 64 this type is available under the name off_t.

These aliases for the ‘SEEK_…’ constants exist for the sake of compatibility with older BSD systems. They are defined in two different header files: ‘fcntl.h’ and ‘sys/file.h’.

L_SET

An alias for SEEK_SET.

L_INCR

An alias for SEEK_CUR.

L_XTND

An alias for SEEK_END.

13.4 Descriptors and Streams

Given an open file descriptor, you can create a stream for it with the fdopen function. You can get the underlying file descriptor for an existing stream with the fileno function. These functions are declared in the header file ‘stdio.h’.

Function: FILE * fdopen (int filedes, const char *opentype)

The fdopen function returns a new stream for the file descriptor filedes.

The opentype argument is interpreted in the same way as for the fopen function (see section Opening Streams), except that the ‘b’ option is not permitted; this is because GNU makes no distinction between text and binary files. Also, "w" and "w+" do not cause truncation of the file; these have an effect only when opening a file, and in this case the file has already been opened. You must make sure that the opentype argument matches the actual mode of the open file descriptor.

The return value is the new stream. If the stream cannot be created (for example, if the modes for the file indicated by the file descriptor do not permit the access specified by the opentype argument), a null pointer is returned instead.

In some other systems, fdopen may fail to detect that the modes for file descriptor do not permit the access specified by opentype. The GNU C library always checks for this.

For an example showing the use of the fdopen function, see Creating a Pipe.

Function: int fileno (FILE *stream)

This function returns the file descriptor associated with the stream stream. If an error is detected (for example, if the stream is not valid) or if stream does not do I/O to a file, fileno returns -1.

Function: int fileno_unlocked (FILE *stream)

The fileno_unlocked function is equivalent to the fileno function except that it does not implicitly lock the stream if the state is FSETLOCKING_INTERNAL.

This function is a GNU extension.

There are also symbolic constants defined in ‘unistd.h’ for the file descriptors belonging to the standard streams stdin, stdout, and stderr; see Standard Streams.

STDIN_FILENO

This macro has value 0, which is the file descriptor for standard input.

STDOUT_FILENO

This macro has value 1, which is the file descriptor for standard output.

STDERR_FILENO

This macro has value 2, which is the file descriptor for standard error output.

13.5 Dangers of Mixing Streams and Descriptors

You can have multiple file descriptors and streams (let's call both streams and descriptors “channels” for short) connected to the same file, but you must take care to avoid confusion between channels. There are two cases to consider: linked channels that share a single file position value, and independent channels that have their own file positions.

It's best to use just one channel in your program for actual data transfer to any given file, except when all the access is for input. For example, if you open a pipe (something you can only do at the file descriptor level), either do all I/O with the descriptor, or construct a stream from the descriptor with fdopen and then do all I/O with the stream.

13.5.1 Linked Channels

Channels that come from a single opening share the same file position; we call them linked channels. Linked channels result when you make a stream from a descriptor using fdopen, when you get a descriptor from a stream with fileno, when you copy a descriptor with dup or dup2, and when descriptors are inherited during fork. For files that don't support random access, such as terminals and pipes, all channels are effectively linked. On random-access files, all append-type output streams are effectively linked to each other.

If you have been using a stream for I/O (or have just opened the stream), and you want to do I/O using another channel (either a stream or a descriptor) that is linked to it, you must first clean up the stream that you have been using. See section Cleaning Streams.

Terminating a process, or executing a new program in the process, destroys all the streams in the process. If descriptors linked to these streams persist in other processes, their file positions become undefined as a result. To prevent this, you must clean up the streams before destroying them.

13.5.2 Independent Channels

When you open channels (streams or descriptors) separately on a seekable file, each channel has its own file position. These are called independent channels.

The system handles each channel independently. Most of the time, this is quite predictable and natural (especially for input): each channel can read or write sequentially at its own place in the file. However, if some of the channels are streams, you must take these precautions:

• You should clean an output stream after use, before doing anything else that might read or write from the same part of the file.
• You should clean an input stream before reading data that may have been modified using an independent channel. Otherwise, you might read obsolete data that had been in the stream's buffer.

If you do output to one channel at the end of the file, this will certainly leave the other independent channels positioned somewhere before the new end. You cannot reliably set their file positions to the new end of file before writing, because the file can always be extended by another process between when you set the file position and when you write the data. Instead, use an append-type descriptor or stream; they always output at the current end of the file. In order to make the end-of-file position accurate, you must clean the output channel you were using, if it is a stream.

It's impossible for two channels to have separate file pointers for a file that doesn't support random access. Thus, channels for reading or writing such files are always linked, never independent. Append-type channels are also always linked. For these channels, follow the rules for linked channels; see Linked Channels.

13.5.3 Cleaning Streams

On the GNU system, you can clean up any stream with fclean:

Function: int fclean (FILE *stream)

Clean up the stream stream so that its buffer is empty. If stream is doing output, force it out. If stream is doing input, give the data in the buffer back to the system, arranging to reread it.

On other systems, you can use fflush to clean a stream in most cases.

You can skip the fclean or fflush if you know the stream is already clean. A stream is clean whenever its buffer is empty. For example, an unbuffered stream is always clean. An input stream that is at end-of-file is clean. A line-buffered stream is clean when the last character output was a newline. However, a just-opened input stream might not be clean, as its input buffer might not be empty.

There is one case in which cleaning a stream is impossible on most systems. This is when the stream is doing input from a file that is not random-access. Such streams typically read ahead, and when the file is not random access, there is no way to give back the excess data already read. When an input stream reads from a random-access file, fflush does clean the stream, but leaves the file pointer at an unpredictable place; you must set the file pointer before doing any further I/O. On the GNU system, using fclean avoids both of these problems.

Closing an output-only stream also does fflush, so this is a valid way of cleaning an output stream. On the GNU system, closing an input stream does fclean.

You need not clean a stream before using its descriptor for control operations such as setting terminal modes; these operations don't affect the file position and are not affected by it. You can use any descriptor for these operations, and all channels are affected simultaneously. However, text already “output” to a stream but still buffered by the stream will be subject to the new terminal modes when subsequently flushed. To make sure “past” output is covered by the terminal settings that were in effect at the time, flush the output streams for that terminal before setting the modes. See section Terminal Modes.

13.6 Fast Scatter-Gather I/O

Some applications may need to read or write data to multiple buffers, which are separated in memory. Although this can be done easily enough with multiple calls to read and write, it is inefficient because there is overhead associated with each kernel call.

Instead, many platforms provide special high-speed primitives to perform these scatter-gather operations in a single kernel call. The GNU C library will provide an emulation on any system that lacks these primitives, so they are not a portability threat. They are defined in sys/uio.h.

These functions are controlled with arrays of iovec structures, which describe the location and size of each buffer.

Data Type: struct iovec

The iovec structure describes a buffer. It contains two fields:

void *iov_base

Contains the address of a buffer.

size_t iov_len

Contains the length of the buffer.

Function: ssize_t readv (int filedes, const struct iovec *vector, int count)

The readv function reads data from filedes and scatters it into the buffers described in vector, which is taken to be count structures long. As each buffer is filled, data is sent to the next.

Note that readv is not guaranteed to fill all the buffers. It may stop at any point, for the same reasons read would.

The return value is a count of bytes (not buffers) read, 0 indicating end-of-file, or -1 indicating an error. The possible errors are the same as in read.

Function: ssize_t writev (int filedes, const struct iovec *vector, int count)

The writev function gathers data from the buffers described in vector, which is taken to be count structures long, and writes them to filedes. As each buffer is written, it moves on to the next.

Like readv, writev may stop midstream under the same conditions write would.

The return value is a count of bytes written, or -1 indicating an error. The possible errors are the same as in write.

Note that if the buffers are small (under about 1kB), high-level streams may be easier to use than these functions. However, readv and writev are more efficient when the individual buffers themselves (as opposed to the total output), are large. In that case, a high-level stream would not be able to cache the data effectively.

13.7 Memory-mapped I/O

On modern operating systems, it is possible to mmap (pronounced “em-map”) a file to a region of memory. When this is done, the file can be accessed just like an array in the program.

This is more efficient than read or write, as only the regions of the file that a program actually accesses are loaded. Accesses to not-yet-loaded parts of the mmapped region are handled in the same way as swapped out pages.

Since mmapped pages can be stored back to their file when physical memory is low, it is possible to mmap files orders of magnitude larger than both the physical memory and swap space. The only limit is address space. The theoretical limit is 4GB on a 32-bit machine - however, the actual limit will be smaller since some areas will be reserved for other purposes. If the LFS interface is used the file size on 32-bit systems is not limited to 2GB (offsets are signed which reduces the addressable area of 4GB by half); the full 64-bit are available.

Memory mapping only works on entire pages of memory. Thus, addresses for mapping must be page-aligned, and length values will be rounded up. To determine the size of a page the machine uses one should use

size_t page_size = (size_t) sysconf (_SC_PAGESIZE);

These functions are declared in ‘sys/mman.h’.

Function: void * mmap (void *address, size_t length,int protect, int flags, int filedes, off_t offset)

The mmap function creates a new mapping, connected to bytes (offset) to (offset + length - 1) in the file open on filedes. A new reference for the file specified by filedes is created, which is not removed by closing the file.

address gives a preferred starting address for the mapping. NULL expresses no preference. Any previous mapping at that address is automatically removed. The address you give may still be changed, unless you use the MAP_FIXED flag.

protect contains flags that control what kind of access is permitted. They include PROT_READ, PROT_WRITE, and PROT_EXEC, which permit reading, writing, and execution, respectively. Inappropriate access will cause a segfault (see section Program Error Signals).

Note that most hardware designs cannot support write permission without read permission, and many do not distinguish read and execute permission. Thus, you may receive wider permissions than you ask for, and mappings of write-only files may be denied even if you do not use PROT_READ.

flags contains flags that control the nature of the map. One of MAP_SHARED or MAP_PRIVATE must be specified.

They include:

MAP_PRIVATE

This specifies that writes to the region should never be written back to the attached file. Instead, a copy is made for the process, and the region will be swapped normally if memory runs low. No other process will see the changes.

Since private mappings effectively revert to ordinary memory when written to, you must have enough virtual memory for a copy of the entire mmapped region if you use this mode with PROT_WRITE.

MAP_SHARED

This specifies that writes to the region will be written back to the file. Changes made will be shared immediately with other processes mmaping the same file.

Note that actual writing may take place at any time. You need to use msync, described below, if it is important that other processes using conventional I/O get a consistent view of the file.

MAP_FIXED

This forces the system to use the exact mapping address specified in address and fail if it can't.

MAP_ANONYMOUS

MAP_ANON

This flag tells the system to create an anonymous mapping, not connected to a file. filedes and off are ignored, and the region is initialized with zeros.

Anonymous maps are used as the basic primitive to extend the heap on some systems. They are also useful to share data between multiple tasks without creating a file.

On some systems using private anonymous mmaps is more efficient than using malloc for large blocks. This is not an issue with the GNU C library, as the included malloc automatically uses mmap where appropriate.

mmap returns the address of the new mapping, or -1 for an error.

Possible errors include:

EINVAL

Either address was unusable, or inconsistent flags were given.

EACCES

filedes was not open for the type of access specified in protect.

ENOMEM

Either there is not enough memory for the operation, or the process is out of address space.

ENODEV

This file is of a type that doesn't support mapping.

ENOEXEC

The file is on a filesystem that doesn't support mapping.

Function: void * mmap64 (void *address, size_t length,int protect, int flags, int filedes, off64_t offset)

The mmap64 function is equivalent to the mmap function but the offset parameter is of type off64_t. On 32-bit systems this allows the file associated with the filedes descriptor to be larger than 2GB. filedes must be a descriptor returned from a call to open64 or fopen64 and freopen64 where the descriptor is retrieved with fileno.

When the sources are translated with _FILE_OFFSET_BITS == 64 this function is actually available under the name mmap. I.e., the new, extended API using 64 bit file sizes and offsets transparently replaces the old API.

Function: int munmap (void *addr, size_t length)

munmap removes any memory maps from (addr) to (addr + length). length should be the length of the mapping.

It is safe to unmap multiple mappings in one command, or include unmapped space in the range. It is also possible to unmap only part of an existing mapping. However, only entire pages can be removed. If length is not an even number of pages, it will be rounded up.

It returns 0 for success and -1 for an error.

One error is possible:

EINVAL

The memory range given was outside the user mmap range or wasn't page aligned.

Function: int msync (void *address, size_t length, int flags)

When using shared mappings, the kernel can write the file at any time before the mapping is removed. To be certain data has actually been written to the file and will be accessible to non-memory-mapped I/O, it is necessary to use this function.

It operates on the region address to (address + length). It may be used on part of a mapping or multiple mappings, however the region given should not contain any unmapped space.

flags can contain some options:

MS_SYNC

This flag makes sure the data is actually written to disk. Normally msync only makes sure that accesses to a file with conventional I/O reflect the recent changes.

MS_ASYNC

This tells msync to begin the synchronization, but not to wait for it to complete.

msync returns 0 for success and -1 for error. Errors include:

EINVAL

An invalid region was given, or the flags were invalid.

EFAULT

There is no existing mapping in at least part of the given region.

Function: void * mremap (void *address, size_t length, size_t new_length, int flag)

This function can be used to change the size of an existing memory area. address and length must cover a region entirely mapped in the same mmap statement. A new mapping with the same characteristics will be returned with the length new_length.

One option is possible, MREMAP_MAYMOVE. If it is given in flags, the system may remove the existing mapping and create a new one of the desired length in another location.

The address of the resulting mapping is returned, or -1. Possible error codes include:

EFAULT

There is no existing mapping in at least part of the original region, or the region covers two or more distinct mappings.

EINVAL

The address given is misaligned or inappropriate.

EAGAIN

The region has pages locked, and if extended it would exceed the process's resource limit for locked pages. See section Limiting Resource Usage.

ENOMEM

The region is private writable, and insufficient virtual memory is available to extend it. Also, this error will occur if MREMAP_MAYMOVE is not given and the extension would collide with another mapped region.

This function is only available on a few systems. Except for performing optional optimizations one should not rely on this function.

Not all file descriptors may be mapped. Sockets, pipes, and most devices only allow sequential access and do not fit into the mapping abstraction. In addition, some regular files may not be mmapable, and older kernels may not support mapping at all. Thus, programs using mmap should have a fallback method to use should it fail. See (standards)Mmap section `Mmap' in GNU Coding Standards.

Function: int madvise (void *addr, size_t length, int advice)

This function can be used to provide the system with advice about the intended usage patterns of the memory region starting at addr and extending length bytes.

The valid BSD values for advice are:

MADV_NORMAL

The region should receive no further special treatment.

MADV_RANDOM

The region will be accessed via random page references. The kernel should page-in the minimal number of pages for each page fault.

MADV_SEQUENTIAL

The region will be accessed via sequential page references. This may cause the kernel to aggressively read-ahead, expecting further sequential references after any page fault within this region.

MADV_WILLNEED

The region will be needed. The pages within this region may be pre-faulted in by the kernel.

MADV_DONTNEED

The region is no longer needed. The kernel may free these pages, causing any changes to the pages to be lost, as well as swapped out pages to be discarded.

The POSIX names are slightly different, but with the same meanings:

POSIX_MADV_NORMAL

This corresponds with BSD's MADV_NORMAL.

POSIX_MADV_RANDOM

This corresponds with BSD's MADV_RANDOM.

POSIX_MADV_SEQUENTIAL

This corresponds with BSD's MADV_SEQUENTIAL.

POSIX_MADV_WILLNEED

This corresponds with BSD's MADV_WILLNEED.

POSIX_MADV_DONTNEED

This corresponds with BSD's MADV_DONTNEED.

msync returns 0 for success and -1 for error. Errors include:

EINVAL

An invalid region was given, or the advice was invalid.

EFAULT

There is no existing mapping in at least part of the given region.

13.8 Waiting for Input or Output

Sometimes a program needs to accept input on multiple input channels whenever input arrives. For example, some workstations may have devices such as a digitizing tablet, function button box, or dial box that are connected via normal asynchronous serial interfaces; good user interface style requires responding immediately to input on any device. Another example is a program that acts as a server to several other processes via pipes or sockets.

You cannot normally use read for this purpose, because this blocks the program until input is available on one particular file descriptor; input on other channels won't wake it up. You could set nonblocking mode and poll each file descriptor in turn, but this is very inefficient.

A better solution is to use the select function. This blocks the program until input or output is ready on a specified set of file descriptors, or until a timer expires, whichever comes first. This facility is declared in the header file ‘sys/types.h’.

In the case of a server socket (see section Listening for Connections), we say that “input” is available when there are pending connections that could be accepted (see section Accepting Connections). accept for server sockets blocks and interacts with select just as read does for normal input.

The file descriptor sets for the select function are specified as fd_set objects. Here is the description of the data type and some macros for manipulating these objects.

Data Type: fd_set

The fd_set data type represents file descriptor sets for the select function. It is actually a bit array.

Macro: int FD_SETSIZE

The value of this macro is the maximum number of file descriptors that a fd_set object can hold information about. On systems with a fixed maximum number, FD_SETSIZE is at least that number. On some systems, including GNU, there is no absolute limit on the number of descriptors open, but this macro still has a constant value which controls the number of bits in an fd_set; if you get a file descriptor with a value as high as FD_SETSIZE, you cannot put that descriptor into an fd_set.

Macro: void FD_ZERO (fd_set *set)

This macro initializes the file descriptor set set to be the empty set.

Macro: void FD_SET (int filedes, fd_set *set)

This macro adds filedes to the file descriptor set set.

The filedes parameter must not have side effects since it is evaluated more than once.

Macro: void FD_CLR (int filedes, fd_set *set)

This macro removes filedes from the file descriptor set set.

The filedes parameter must not have side effects since it is evaluated more than once.

Macro: int FD_ISSET (int filedes, const fd_set *set)

This macro returns a nonzero value (true) if filedes is a member of the file descriptor set set, and zero (false) otherwise.

The filedes parameter must not have side effects since it is evaluated more than once.

Next, here is the description of the select function itself.

Function: int select (int nfds, fd_set *read-fds, fd_set *write-fds, fd_set *except-fds, struct timeval *timeout)

The select function blocks the calling process until there is activity on any of the specified sets of file descriptors, or until the timeout period has expired.

The file descriptors specified by the read-fds argument are checked to see if they are ready for reading; the write-fds file descriptors are checked to see if they are ready for writing; and the except-fds file descriptors are checked for exceptional conditions. You can pass a null pointer for any of these arguments if you are not interested in checking for that kind of condition.

A file descriptor is considered ready for reading if a read call will not block. This usually includes the read offset being at the end of the file or there is an error to report. A server socket is considered ready for reading if there is a pending connection which can be accepted with accept; see section Accepting Connections. A client socket is ready for writing when its connection is fully established; see section Making a Connection.

“Exceptional conditions” does not mean errors—errors are reported immediately when an erroneous system call is executed, and do not constitute a state of the descriptor. Rather, they include conditions such as the presence of an urgent message on a socket. (See section Sockets, for information on urgent messages.)

The select function checks only the first nfds file descriptors. The usual thing is to pass FD_SETSIZE as the value of this argument.

The timeout specifies the maximum time to wait. If you pass a null pointer for this argument, it means to block indefinitely until one of the file descriptors is ready. Otherwise, you should provide the time in struct timeval format; see High-Resolution Calendar. Specify zero as the time (a struct timeval containing all zeros) if you want to find out which descriptors are ready without waiting if none are ready.

The normal return value from select is the total number of ready file descriptors in all of the sets. Each of the argument sets is overwritten with information about the descriptors that are ready for the corresponding operation. Thus, to see if a particular descriptor desc has input, use FD_ISSET (desc, read-fds) after select returns.

If select returns because the timeout period expires, it returns a value of zero.

Any signal will cause select to return immediately. So if your program uses signals, you can't rely on select to keep waiting for the full time specified. If you want to be sure of waiting for a particular amount of time, you must check for EINTR and repeat the select with a newly calculated timeout based on the current time. See the example below. See also Primitives Interrupted by Signals.

If an error occurs, select returns -1 and does not modify the argument file descriptor sets. The following errno error conditions are defined for this function:

EBADF

One of the file descriptor sets specified an invalid file descriptor.

EINTR

The operation was interrupted by a signal. See section Primitives Interrupted by Signals.

EINVAL

The timeout argument is invalid; one of the components is negative or too large.

Portability Note: The select function is a BSD Unix feature.

Here is an example showing how you can use select to establish a timeout period for reading from a file descriptor. The input_timeout function blocks the calling process until input is available on the file descriptor, or until the timeout period expires.

#include <errno.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/time.h>

int
input_timeout (int filedes, unsigned int seconds)
{
  fd_set set;
  struct timeval timeout;

  /* Initialize the file descriptor set. */
  FD_ZERO (&set);
  FD_SET (filedes, &set);

  /* Initialize the timeout data structure. */
  timeout.tv_sec = seconds;
  timeout.tv_usec = 0;

  /* select returns 0 if timeout, 1 if input available, -1 if error. */
  return TEMP_FAILURE_RETRY (select (FD_SETSIZE,
                                     &set, NULL, NULL,
                                     &timeout));
}

int
main (void)
{
  fprintf (stderr, "select returned %d.\n",
           input_timeout (STDIN_FILENO, 5));
  return 0;
}

There is another example showing the use of select to multiplex input from multiple sockets in Byte Stream Connection Server Example.

13.9 Synchronizing I/O operations

In most modern operating systems, the normal I/O operations are not executed synchronously. I.e., even if a write system call returns, this does not mean the data is actually written to the media, e.g., the disk.

In situations where synchronization points are necessary, you can use special functions which ensure that all operations finish before they return.

Function: int sync (void)

A call to this function will not return as long as there is data which has not been written to the device. All dirty buffers in the kernel will be written and so an overall consistent system can be achieved (if no other process in parallel writes data).

A prototype for sync can be found in ‘unistd.h’.

The return value is zero to indicate no error.

Programs more often want to ensure that data written to a given file is committed, rather than all data in the system. For this, sync is overkill.

Function: int fsync (int fildes)

The fsync function can be used to make sure all data associated with the open file fildes is written to the device associated with the descriptor. The function call does not return unless all actions have finished.

A prototype for fsync can be found in ‘unistd.h’.

This function is a cancellation point in multi-threaded programs. This is a problem if the thread allocates some resources (like memory, file descriptors, semaphores or whatever) at the time fsync is called. If the thread gets canceled these resources stay allocated until the program ends. To avoid this, calls to fsync should be protected using cancellation handlers.

The return value of the function is zero if no error occurred. Otherwise it is -1 and the global variable errno is set to the following values:

EBADF

The descriptor fildes is not valid.

EINVAL

No synchronization is possible since the system does not implement this.

Sometimes it is not even necessary to write all data associated with a file descriptor. E.g., in database files which do not change in size it is enough to write all the file content data to the device. Meta-information, like the modification time etc., are not that important and leaving such information uncommitted does not prevent a successful recovering of the file in case of a problem.

Function: int fdatasync (int fildes)

When a call to the fdatasync function returns, it is ensured that all of the file data is written to the device. For all pending I/O operations, the parts guaranteeing data integrity finished.

Not all systems implement the fdatasync operation. On systems missing this functionality fdatasync is emulated by a call to fsync since the performed actions are a superset of those required by fdatasync.

The prototype for fdatasync is in ‘unistd.h’.

The return value of the function is zero if no error occurred. Otherwise it is -1 and the global variable errno is set to the following values:

EBADF

The descriptor fildes is not valid.

EINVAL

No synchronization is possible since the system does not implement this.

13.10 Perform I/O Operations in Parallel

The POSIX.1b standard defines a new set of I/O operations which can significantly reduce the time an application spends waiting at I/O. The new functions allow a program to initiate one or more I/O operations and then immediately resume normal work while the I/O operations are executed in parallel. This functionality is available if the ‘unistd.h’ file defines the symbol _POSIX_ASYNCHRONOUS_IO.

These functions are part of the library with realtime functions named ‘librt’. They are not actually part of the ‘libc’ binary. The implementation of these functions can be done using support in the kernel (if available) or using an implementation based on threads at userlevel. In the latter case it might be necessary to link applications with the thread library ‘libpthread’ in addition to ‘librt’.

All AIO operations operate on files which were opened previously. There might be arbitrarily many operations running for one file. The asynchronous I/O operations are controlled using a data structure named struct aiocb (AIO control block). It is defined in ‘aio.h’ as follows.

Data Type: struct aiocb

The POSIX.1b standard mandates that the struct aiocb structure contains at least the members described in the following table. There might be more elements which are used by the implementation, but depending upon these elements is not portable and is highly deprecated.

int aio_fildes

This element specifies the file descriptor to be used for the operation. It must be a legal descriptor, otherwise the operation will fail.

The device on which the file is opened must allow the seek operation. I.e., it is not possible to use any of the AIO operations on devices like terminals where an lseek call would lead to an error.

off_t aio_offset

This element specifies the offset in the file at which the operation (input or output) is performed. Since the operations are carried out in arbitrary order and more than one operation for one file descriptor can be started, one cannot expect a current read/write position of the file descriptor.

volatile void *aio_buf

This is a pointer to the buffer with the data to be written or the place where the read data is stored.

size_t aio_nbytes

This element specifies the length of the buffer pointed to by aio_buf.

int aio_reqprio

If the platform has defined _POSIX_PRIORITIZED_IO and _POSIX_PRIORITY_SCHEDULING, the AIO requests are processed based on the current scheduling priority. The aio_reqprio element can then be used to lower the priority of the AIO operation.

struct sigevent aio_sigevent

This element specifies how the calling process is notified once the operation terminates. If the sigev_notify element is SIGEV_NONE, no notification is sent. If it is SIGEV_SIGNAL, the signal determined by sigev_signo is sent. Otherwise, sigev_notify must be SIGEV_THREAD. In this case, a thread is created which starts executing the function pointed to by sigev_notify_function.

int aio_lio_opcode

This element is only used by the lio_listio and lio_listio64 functions. Since these functions allow an arbitrary number of operations to start at once, and each operation can be input or output (or nothing), the information must be stored in the control block. The possible values are:

LIO_READ

Start a read operation. Read from the file at position aio_offset and store the next aio_nbytes bytes in the buffer pointed to by aio_buf.

LIO_WRITE

Start a write operation. Write aio_nbytes bytes starting at aio_buf into the file starting at position aio_offset.

LIO_NOP

Do nothing for this control block. This value is useful sometimes when an array of struct aiocb values contains holes, i.e., some of the values must not be handled although the whole array is presented to the lio_listio function.

When the sources are compiled using _FILE_OFFSET_BITS == 64 on a 32 bit machine, this type is in fact struct aiocb64, since the LFS interface transparently replaces the struct aiocb definition.

For use with the AIO functions defined in the LFS, there is a similar type defined which replaces the types of the appropriate members with larger types but otherwise is equivalent to struct aiocb. Particularly, all member names are the same.

Data Type: struct aiocb64

int aio_fildes

This element specifies the file descriptor which is used for the operation. It must be a legal descriptor since otherwise the operation fails for obvious reasons.

The device on which the file is opened must allow the seek operation. I.e., it is not possible to use any of the AIO operations on devices like terminals where an lseek call would lead to an error.

off64_t aio_offset

This element specifies at which offset in the file the operation (input or output) is performed. Since the operation are carried in arbitrary order and more than one operation for one file descriptor can be started, one cannot expect a current read/write position of the file descriptor.

volatile void *aio_buf

This is a pointer to the buffer with the data to be written or the place where the read data is stored.

size_t aio_nbytes

This element specifies the length of the buffer pointed to by aio_buf.

int aio_reqprio

If for the platform _POSIX_PRIORITIZED_IO and _POSIX_PRIORITY_SCHEDULING are defined the AIO requests are processed based on the current scheduling priority. The aio_reqprio element can then be used to lower the priority of the AIO operation.

struct sigevent aio_sigevent

This element specifies how the calling process is notified once the operation terminates. If the sigev_notify, element is SIGEV_NONE no notification is sent. If it is SIGEV_SIGNAL, the signal determined by sigev_signo is sent. Otherwise, sigev_notify must be SIGEV_THREAD in which case a thread which starts executing the function pointed to by sigev_notify_function.

int aio_lio_opcode

This element is only used by the lio_listio and [lio_listio64 functions. Since these functions allow an arbitrary number of operations to start at once, and since each operation can be input or output (or nothing), the information must be stored in the control block. See the description of struct aiocb for a description of the possible values.

When the sources are compiled using _FILE_OFFSET_BITS == 64 on a 32 bit machine, this type is available under the name struct aiocb64, since the LFS transparently replaces the old interface.

13.10.1 Asynchronous Read and Write Operations

Function: int aio_read (struct aiocb *aiocbp)

This function initiates an asynchronous read operation. It immediately returns after the operation was enqueued or when an error was encountered.

The first aiocbp->aio_nbytes bytes of the file for which aiocbp->aio_fildes is a descriptor are written to the buffer starting at aiocbp->aio_buf. Reading starts at the absolute position aiocbp->aio_offset in the file.

If prioritized I/O is supported by the platform the aiocbp->aio_reqprio value is used to adjust the priority before the request is actually enqueued.

The calling process is notified about the termination of the read request according to the aiocbp->aio_sigevent value.

When aio_read returns, the return value is zero if no error occurred that can be found before the process is enqueued. If such an early error is found, the function returns -1 and sets errno to one of the following values:

EAGAIN

The request was not enqueued due to (temporarily) exceeded resource limitations.

ENOSYS

The aio_read function is not implemented.

EBADF

The aiocbp->aio_fildes descriptor is not valid. This condition need not be recognized before enqueueing the request and so this error might also be signaled asynchronously.

EINVAL

The aiocbp->aio_offset or aiocbp->aio_reqpiro value is invalid. This condition need not be recognized before enqueueing the request and so this error might also be signaled asynchronously.

If aio_read returns zero, the current status of the request can be queried using aio_error and aio_return functions. As long as the value returned by aio_error is EINPROGRESS the operation has not yet completed. If aio_error returns zero, the operation successfully terminated, otherwise the value is to be interpreted as an error code. If the function terminated, the result of the operation can be obtained using a call to aio_return. The returned value is the same as an equivalent call to read would have returned. Possible error codes returned by aio_error are:

EBADF

The aiocbp->aio_fildes descriptor is not valid.

ECANCELED

The operation was canceled before the operation was finished (see section Cancellation of AIO Operations)

EINVAL

The aiocbp->aio_offset value is invalid.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is in fact aio_read64 since the LFS interface transparently replaces the normal implementation.

Function: int aio_read64 (struct aiocb *aiocbp)

This function is similar to the aio_read function. The only difference is that on 32 bit machines, the file descriptor should be opened in the large file mode. Internally, aio_read64 uses functionality equivalent to lseek64 (see section Setting the File Position of a Descriptor) to position the file descriptor correctly for the reading, as opposed to lseek functionality used in aio_read.

When the sources are compiled with _FILE_OFFSET_BITS == 64, this function is available under the name aio_read and so transparently replaces the interface for small files on 32 bit machines.

To write data asynchronously to a file, there exists an equivalent pair of functions with a very similar interface.

Function: int aio_write (struct aiocb *aiocbp)

This function initiates an asynchronous write operation. The function call immediately returns after the operation was enqueued or if before this happens an error was encountered.

The first aiocbp->aio_nbytes bytes from the buffer starting at aiocbp->aio_buf are written to the file for which aiocbp->aio_fildes is an descriptor, starting at the absolute position aiocbp->aio_offset in the file.

If prioritized I/O is supported by the platform, the aiocbp->aio_reqprio value is used to adjust the priority before the request is actually enqueued.

The calling process is notified about the termination of the read request according to the aiocbp->aio_sigevent value.

When aio_write returns, the return value is zero if no error occurred that can be found before the process is enqueued. If such an early error is found the function returns -1 and sets errno to one of the following values.

EAGAIN

The request was not enqueued due to (temporarily) exceeded resource limitations.

ENOSYS

The aio_write function is not implemented.

EBADF

The aiocbp->aio_fildes descriptor is not valid. This condition may not be recognized before enqueueing the request, and so this error might also be signaled asynchronously.

EINVAL

The aiocbp->aio_offset or aiocbp->aio_reqprio value is invalid. This condition may not be recognized before enqueueing the request and so this error might also be signaled asynchronously.

In the case aio_write returns zero, the current status of the request can be queried using aio_error and aio_return functions. As long as the value returned by aio_error is EINPROGRESS the operation has not yet completed. If aio_error returns zero, the operation successfully terminated, otherwise the value is to be interpreted as an error code. If the function terminated, the result of the operation can be get using a call to aio_return. The returned value is the same as an equivalent call to read would have returned. Possible error codes returned by aio_error are:

EBADF

The aiocbp->aio_fildes descriptor is not valid.

ECANCELED

The operation was canceled before the operation was finished. (see section Cancellation of AIO Operations)

EINVAL

The aiocbp->aio_offset value is invalid.

When the sources are compiled with _FILE_OFFSET_BITS == 64, this function is in fact aio_write64 since the LFS interface transparently replaces the normal implementation.

Function: int aio_write64 (struct aiocb *aiocbp)

This function is similar to the aio_write function. The only difference is that on 32 bit machines the file descriptor should be opened in the large file mode. Internally aio_write64 uses functionality equivalent to lseek64 (see section Setting the File Position of a Descriptor) to position the file descriptor correctly for the writing, as opposed to lseek functionality used in aio_write.

When the sources are compiled with _FILE_OFFSET_BITS == 64, this function is available under the name aio_write and so transparently replaces the interface for small files on 32 bit machines.

Besides these functions with the more or less traditional interface, POSIX.1b also defines a function which can initiate more than one operation at a time, and which can handle freely mixed read and write operations. It is therefore similar to a combination of readv and writev.

Function: int lio_listio (int mode, struct aiocb *const list[], int nent, struct sigevent *sig)

The lio_listio function can be used to enqueue an arbitrary number of read and write requests at one time. The requests can all be meant for the same file, all for different files or every solution in between.

lio_listio gets the nent requests from the array pointed to by list. The operation to be performed is determined by the aio_lio_opcode member in each element of list. If this field is LIO_READ a read operation is enqueued, similar to a call of aio_read for this element of the array (except that the way the termination is signalled is different, as we will see below). If the aio_lio_opcode member is LIO_WRITE a write operation is enqueued. Otherwise the aio_lio_opcode must be LIO_NOP in which case this element of list is simply ignored. This “operation” is useful in situations where one has a fixed array of struct aiocb elements from which only a few need to be handled at a time. Another situation is where the lio_listio call was canceled before all requests are processed (see section Cancellation of AIO Operations) and the remaining requests have to be reissued.

The other members of each element of the array pointed to by list must have values suitable for the operation as described in the documentation for aio_read and aio_write above.

The mode argument determines how lio_listio behaves after having enqueued all the requests. If mode is LIO_WAIT it waits until all requests terminated. Otherwise mode must be LIO_NOWAIT and in this case the function returns immediately after having enqueued all the requests. In this case the caller gets a notification of the termination of all requests according to the sig parameter. If sig is NULL no notification is send. Otherwise a signal is sent or a thread is started, just as described in the description for aio_read or aio_write.

If mode is LIO_WAIT, the return value of lio_listio is 0 when all requests completed successfully. Otherwise the function return -1 and errno is set accordingly. To find out which request or requests failed one has to use the aio_error function on all the elements of the array list.

In case mode is LIO_NOWAIT, the function returns 0 if all requests were enqueued correctly. The current state of the requests can be found using aio_error and aio_return as described above. If lio_listio returns -1 in this mode, the global variable errno is set accordingly. If a request did not yet terminate, a call to aio_error returns EINPROGRESS. If the value is different, the request is finished and the error value (or 0) is returned and the result of the operation can be retrieved using aio_return.

Possible values for errno are:

EAGAIN

The resources necessary to queue all the requests are not available at the moment. The error status for each element of list must be checked to determine which request failed.

Another reason could be that the system wide limit of AIO requests is exceeded. This cannot be the case for the implementation on GNU systems since no arbitrary limits exist.

EINVAL

The mode parameter is invalid or nent is larger than AIO_LISTIO_MAX.

EIO

One or more of the request's I/O operations failed. The error status of each request should be checked to determine which one failed.

ENOSYS

The lio_listio function is not supported.

If the mode parameter is LIO_NOWAIT and the caller cancels a request, the error status for this request returned by aio_error is ECANCELED.

When the sources are compiled with _FILE_OFFSET_BITS == 64, this function is in fact lio_listio64 since the LFS interface transparently replaces the normal implementation.

Function: int lio_listio64 (int mode, struct aiocb *const list, int nent, struct sigevent *sig)

This function is similar to the lio_listio function. The only difference is that on 32 bit machines, the file descriptor should be opened in the large file mode. Internally, lio_listio64 uses functionality equivalent to lseek64 (see section Setting the File Position of a Descriptor) to position the file descriptor correctly for the reading or writing, as opposed to lseek functionality used in lio_listio.

When the sources are compiled with _FILE_OFFSET_BITS == 64, this function is available under the name lio_listio and so transparently replaces the interface for small files on 32 bit machines.

13.10.2 Getting the Status of AIO Operations

As already described in the documentation of the functions in the last section, it must be possible to get information about the status of an I/O request. When the operation is performed truly asynchronously (as with aio_read and aio_write and with lio_listio when the mode is LIO_NOWAIT), one sometimes needs to know whether a specific request already terminated and if so, what the result was. The following two functions allow you to get this kind of information.

Function: int aio_error (const struct aiocb *aiocbp)

This function determines the error state of the request described by the struct aiocb variable pointed to by aiocbp. If the request has not yet terminated the value returned is always EINPROGRESS. Once the request has terminated the value aio_error returns is either 0 if the request completed successfully or it returns the value which would be stored in the errno variable if the request would have been done using read, write, or fsync.

The function can return ENOSYS if it is not implemented. It could also return EINVAL if the aiocbp parameter does not refer to an asynchronous operation whose return status is not yet known.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is in fact aio_error64 since the LFS interface transparently replaces the normal implementation.

Function: int aio_error64 (const struct aiocb64 *aiocbp)

This function is similar to aio_error with the only difference that the argument is a reference to a variable of type struct aiocb64.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is available under the name aio_error and so transparently replaces the interface for small files on 32 bit machines.

Function: ssize_t aio_return (const struct aiocb *aiocbp)

This function can be used to retrieve the return status of the operation carried out by the request described in the variable pointed to by aiocbp. As long as the error status of this request as returned by aio_error is EINPROGRESS the return of this function is undefined.

Once the request is finished this function can be used exactly once to retrieve the return value. Following calls might lead to undefined behavior. The return value itself is the value which would have been returned by the read, write, or fsync call.

The function can return ENOSYS if it is not implemented. It could also return EINVAL if the aiocbp parameter does not refer to an asynchronous operation whose return status is not yet known.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is in fact aio_return64 since the LFS interface transparently replaces the normal implementation.

Function: int aio_return64 (const struct aiocb64 *aiocbp)

This function is similar to aio_return with the only difference that the argument is a reference to a variable of type struct aiocb64.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is available under the name aio_return and so transparently replaces the interface for small files on 32 bit machines.

13.10.3 Getting into a Consistent State

When dealing with asynchronous operations it is sometimes necessary to get into a consistent state. This would mean for AIO that one wants to know whether a certain request or a group of request were processed. This could be done by waiting for the notification sent by the system after the operation terminated, but this sometimes would mean wasting resources (mainly computation time). Instead POSIX.1b defines two functions which will help with most kinds of consistency.

The aio_fsync and aio_fsync64 functions are only available if the symbol _POSIX_SYNCHRONIZED_IO is defined in ‘unistd.h’.

Function: int aio_fsync (int op, struct aiocb *aiocbp)

Calling this function forces all I/O operations operating queued at the time of the function call operating on the file descriptor aiocbp->aio_fildes into the synchronized I/O completion state (see section Synchronizing I/O operations). The aio_fsync function returns immediately but the notification through the method described in aiocbp->aio_sigevent will happen only after all requests for this file descriptor have terminated and the file is synchronized. This also means that requests for this very same file descriptor which are queued after the synchronization request are not affected.

If op is O_DSYNC the synchronization happens as with a call to fdatasync. Otherwise op should be O_SYNC and the synchronization happens as with fsync.

As long as the synchronization has not happened, a call to aio_error with the reference to the object pointed to by aiocbp returns EINPROGRESS. Once the synchronization is done aio_error return 0 if the synchronization was not successful. Otherwise the value returned is the value to which the fsync or fdatasync function would have set the errno variable. In this case nothing can be assumed about the consistency for the data written to this file descriptor.

The return value of this function is 0 if the request was successfully enqueued. Otherwise the return value is -1 and errno is set to one of the following values:

EAGAIN

The request could not be enqueued due to temporary lack of resources.

EBADF

The file descriptor aiocbp->aio_fildes is not valid or not open for writing.

EINVAL

The implementation does not support I/O synchronization or the op parameter is other than O_DSYNC and O_SYNC.

ENOSYS

This function is not implemented.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is in fact aio_fsync64 since the LFS interface transparently replaces the normal implementation.

Function: int aio_fsync64 (int op, struct aiocb64 *aiocbp)

This function is similar to aio_fsync with the only difference that the argument is a reference to a variable of type struct aiocb64.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is available under the name aio_fsync and so transparently replaces the interface for small files on 32 bit machines.

Another method of synchronization is to wait until one or more requests of a specific set terminated. This could be achieved by the aio_* functions to notify the initiating process about the termination but in some situations this is not the ideal solution. In a program which constantly updates clients somehow connected to the server it is not always the best solution to go round robin since some connections might be slow. On the other hand letting the aio_* function notify the caller might also be not the best solution since whenever the process works on preparing data for on client it makes no sense to be interrupted by a notification since the new client will not be handled before the current client is served. For situations like this aio_suspend should be used.

Function: int aio_suspend (const struct aiocb *const list[], int nent, const struct timespec *timeout)

When calling this function, the calling thread is suspended until at least one of the requests pointed to by the nent elements of the array list has completed. If any of the requests has already completed at the time aio_suspend is called, the function returns immediately. Whether a request has terminated or not is determined by comparing the error status of the request with EINPROGRESS. If an element of list is NULL, the entry is simply ignored.

If no request has finished, the calling process is suspended. If timeout is NULL, the process is not woken until a request has finished. If timeout is not NULL, the process remains suspended at least as long as specified in timeout. In this case, aio_suspend returns with an error.

The return value of the function is 0 if one or more requests from the list have terminated. Otherwise the function returns -1 and errno is set to one of the following values:

EAGAIN

None of the requests from the list completed in the time specified by timeout.

EINTR

A signal interrupted the aio_suspend function. This signal might also be sent by the AIO implementation while signalling the termination of one of the requests.

ENOSYS

The aio_suspend function is not implemented.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is in fact aio_suspend64 since the LFS interface transparently replaces the normal implementation.

Function: int aio_suspend64 (const struct aiocb64 *const list[], int nent, const struct timespec *timeout)

This function is similar to aio_suspend with the only difference that the argument is a reference to a variable of type struct aiocb64.

When the sources are compiled with _FILE_OFFSET_BITS == 64 this function is available under the name aio_suspend and so transparently replaces the interface for small files on 32 bit machines.

13.10.4 Cancellation of AIO Operations

When one or more requests are asynchronously processed, it might be useful in some situations to cancel a selected operation, e.g., if it becomes obvious that the written data is no longer accurate and would have to be overwritten soon. As an example, assume an application, which writes data in files in a situation where new incoming data would have to be written in a file which will be updated by an enqueued request. The POSIX AIO implementation provides such a function, but this function is not capable of forcing the cancellation of the request. It is up to the implementation to decide whether it is possible to cancel the operation or not. Therefore using this function is merely a hint.

Function: int aio_cancel (int fildes, struct aiocb *aiocbp)

The aio_cancel function can be used to cancel one or more outstanding requests. If the aiocbp parameter is NULL, the function tries to cancel all of the outstanding requests which would process the file descriptor fildes (i.e., whose aio_fildes member is fildes). If aiocbp is not NULL, aio_cancel attempts to cancel the specific request pointed to by aiocbp.

For requests which were successfully canceled, the normal notification about the termination of the request should take place. I.e., depending on the struct sigevent object which controls this, nothing happens, a signal is sent or a thread is started. If the request cannot be canceled, it terminates the usual way after performing the operation.

After a request is successfully canceled, a call to aio_error with a reference to this request as the parameter will return ECANCELED and a call to aio_return will return -1. If the request wasn't canceled and is still running the error status is still EINPROGRESS.

The return value of the function is AIO_CANCELED if there were requests which haven't terminated and which were successfully canceled. If there is one or more requests left which couldn't be canceled, the return value is AIO_NOTCANCELED. In this case aio_error must be used to find out which of the, perhaps multiple, requests (in aiocbp is NULL) weren't successfully canceled. If all requests already terminated at the time aio_cancel is called the return value is AIO_ALLDONE.

If an error occurred during the execution of aio_cancel the function returns -1 and sets errno to one of the following values.

EBADF

The file descriptor fildes is not valid.

ENOSYS

aio_cancel is not implemented.

When the sources are compiled with _FILE_OFFSET_BITS == 64, this function is in fact aio_cancel64 since the LFS interface transparently replaces the normal implementation.

Function: int aio_cancel64 (int fildes, struct aiocb64 *aiocbp)

This function is similar to aio_cancel with the only difference that the argument is a reference to a variable of type struct aiocb64.

When the sources are compiled with _FILE_OFFSET_BITS == 64, this function is available under the name aio_cancel and so transparently replaces the interface for small files on 32 bit machines.

13.10.5 How to optimize the AIO implementation

The POSIX standard does not specify how the AIO functions are implemented. They could be system calls, but it is also possible to emulate them at userlevel.

At the point of this writing, the available implementation is a userlevel implementation which uses threads for handling the enqueued requests. While this implementation requires making some decisions about limitations, hard limitations are something which is best avoided in the GNU C library. Therefore, the GNU C library provides a means for tuning the AIO implementation according to the individual use.

Data Type: struct aioinit

This data type is used to pass the configuration or tunable parameters to the implementation. The program has to initialize the members of this struct and pass it to the implementation using the aio_init function.

int aio_threads

This member specifies the maximal number of threads which may be used at any one time.

int aio_num

This number provides an estimate on the maximal number of simultaneously enqueued requests.

int aio_locks

Unused.

int aio_usedba

Unused.

int aio_debug

Unused.

int aio_numusers

Unused.

int aio_reserved[2]

Unused.

Function: void aio_init (const struct aioinit *init)

This function must be called before any other AIO function. Calling it is completely voluntary, as it is only meant to help the AIO implementation perform better.

Before calling the aio_init, function the members of a variable of type struct aioinit must be initialized. Then a reference to this variable is passed as the parameter to aio_init which itself may or may not pay attention to the hints.

The function has no return value and no error cases are defined. It is a extension which follows a proposal from the SGI implementation in Irix 6. It is not covered by POSIX.1b or Unix98.

13.11 Control Operations on Files

This section describes how you can perform various other operations on file descriptors, such as inquiring about or setting flags describing the status of the file descriptor, manipulating record locks, and the like. All of these operations are performed by the function fcntl.

The second argument to the fcntl function is a command that specifies which operation to perform. The function and macros that name various flags that are used with it are declared in the header file ‘fcntl.h’. Many of these flags are also used by the open function; see Opening and Closing Files.

Function: int fcntl (int filedes, int command, …)

The fcntl function performs the operation specified by command on the file descriptor filedes. Some commands require additional arguments to be supplied. These additional arguments and the return value and error conditions are given in the detailed descriptions of the individual commands.

Briefly, here is a list of what the various commands are.

F_DUPFD

Duplicate the file descriptor (return another file descriptor pointing to the same open file). See section Duplicating Descriptors.

F_GETFD

Get flags associated with the file descriptor. See section File Descriptor Flags.

F_SETFD

Set flags associated with the file descriptor. See section File Descriptor Flags.

F_GETFL

Get flags associated with the open file. See section File Status Flags.

F_SETFL

Set flags associated with the open file. See section File Status Flags.

F_GETLK

Get a file lock. See section File Locks.

F_SETLK

Set or clear a file lock. See section File Locks.

F_SETLKW

Like F_SETLK, but wait for completion. See section File Locks.

F_GETOWN

Get process or process group ID to receive SIGIO signals. See section Interrupt-Driven Input.

F_SETOWN

Set process or process group ID to receive SIGIO signals. See section Interrupt-Driven Input.

This function is a cancellation point in multi-threaded programs. This is a problem if the thread allocates some resources (like memory, file descriptors, semaphores or whatever) at the time fcntl is called. If the thread gets canceled these resources stay allocated until the program ends. To avoid this calls to fcntl should be protected using cancellation handlers.

13.12 Duplicating Descriptors

You can duplicate a file descriptor, or allocate another file descriptor that refers to the same open file as the original. Duplicate descriptors share one file position and one set of file status flags (see section File Status Flags), but each has its own set of file descriptor flags (see section File Descriptor Flags).

The major use of duplicating a file descriptor is to implement redirection of input or output: that is, to change the file or pipe that a particular file descriptor corresponds to.

You can perform this operation using the fcntl function with the F_DUPFD command, but there are also convenient functions dup and dup2 for duplicating descriptors.

The fcntl function and flags are declared in ‘fcntl.h’, while prototypes for dup and dup2 are in the header file ‘unistd.h’.

Function: int dup (int old)

This function copies descriptor old to the first available descriptor number (the first number not currently open). It is equivalent to fcntl (old, F_DUPFD, 0).

Function: int dup2 (int old, int new)

This function copies the descriptor old to descriptor number new.

If old is an invalid descriptor, then dup2 does nothing; it does not close new. Otherwise, the new duplicate of old replaces any previous meaning of descriptor new, as if new were closed first.

If old and new are different numbers, and old is a valid descriptor number, then dup2 is equivalent to:

close (new); fcntl (old, F_DUPFD, new)

However, dup2 does this atomically; there is no instant in the middle of calling dup2 at which new is closed and not yet a duplicate of old.

Macro: int F_DUPFD

This macro is used as the command argument to fcntl, to copy the file descriptor given as the first argument.

The form of the call in this case is:

fcntl (old, F_DUPFD, next-filedes)

The next-filedes argument is of type int and specifies that the file descriptor returned should be the next available one greater than or equal to this value.

The return value from fcntl with this command is normally the value of the new file descriptor. A return value of -1 indicates an error. The following errno error conditions are defined for this command:

EBADF

The old argument is invalid.

EINVAL

The next-filedes argument is invalid.

EMFILE

There are no more file descriptors available—your program is already using the maximum. In BSD and GNU, the maximum is controlled by a resource limit that can be changed; see section Limiting Resource Usage, for more information about the RLIMIT_NOFILE limit.

ENFILE is not a possible error code for dup2 because dup2 does not create a new opening of a file; duplicate descriptors do not count toward the limit which ENFILE indicates. EMFILE is possible because it refers to the limit on distinct descriptor numbers in use in one process.

Here is an example showing how to use dup2 to do redirection. Typically, redirection of the standard streams (like stdin) is done by a shell or shell-like program before calling one of the exec functions (see section Executing a File) to execute a new program in a child process. When the new program is executed, it creates and initializes the standard streams to point to the corresponding file descriptors, before its main function is invoked.

So, to redirect standard input to a file, the shell could do something like:

pid = fork ();
if (pid == 0)
  {
    char *filename;
    char *program;
    int file;
    …
    file = TEMP_FAILURE_RETRY (open (filename, O_RDONLY));
    dup2 (file, STDIN_FILENO);
    TEMP_FAILURE_RETRY (close (file));
    execv (program, NULL);
  }

There is also a more detailed example showing how to implement redirection in the context of a pipeline of processes in Launching Jobs.

13.13 File Descriptor Flags

File descriptor flags are miscellaneous attributes of a file descriptor. These flags are associated with particular file descriptors, so that if you have created duplicate file descriptors from a single opening of a file, each descriptor has its own set of flags.

Currently there is just one file descriptor flag: FD_CLOEXEC, which causes the descriptor to be closed if you use any of the exec… functions (see section Executing a File).

The symbols in this section are defined in the header file ‘fcntl.h’.

Macro: int F_GETFD

This macro is used as the command argument to fcntl, to specify that it should return the file descriptor flags associated with the filedes argument.

The normal return value from fcntl with this command is a nonnegative number which can be interpreted as the bitwise OR of the individual flags (except that currently there is only one flag to use).

In case of an error, fcntl returns -1. The following errno error conditions are defined for this command:

EBADF

The filedes argument is invalid.

Macro: int F_SETFD

This macro is used as the command argument to fcntl, to specify that it should set the file descriptor flags associated with the filedes argument. This requires a third int argument to specify the new flags, so the form of the call is:

fcntl (filedes, F_SETFD, new-flags)

The normal return value from fcntl with this command is an unspecified value other than -1, which indicates an error. The flags and error conditions are the same as for the F_GETFD command.

The following macro is defined for use as a file descriptor flag with the fcntl function. The value is an integer constant usable as a bit mask value.

Macro: int FD_CLOEXEC

This flag specifies that the file descriptor should be closed when an exec function is invoked; see Executing a File. When a file descriptor is allocated (as with open or dup), this bit is initially cleared on the new file descriptor, meaning that descriptor will survive into the new program after exec.

If you want to modify the file descriptor flags, you should get the current flags with F_GETFD and modify the value. Don't assume that the flags listed here are the only ones that are implemented; your program may be run years from now and more flags may exist then. For example, here is a function to set or clear the flag FD_CLOEXEC without altering any other flags:

/* Set the FD_CLOEXEC flag of desc if value is nonzero,
   or clear the flag if value is 0.
   Return 0 on success, or -1 on error with errno set. */

int
set_cloexec_flag (int desc, int value)
{
  int oldflags = fcntl (desc, F_GETFD, 0);
  /* If reading the flags failed, return error indication now. */
  if (oldflags < 0)
    return oldflags;
  /* Set just the flag we want to set. */
  if (value != 0)
    oldflags |= FD_CLOEXEC;
  else
    oldflags &= ~FD_CLOEXEC;
  /* Store modified flag word in the descriptor. */
  return fcntl (desc, F_SETFD, oldflags);
}

13.14 File Status Flags

File status flags are used to specify attributes of the opening of a file. Unlike the file descriptor flags discussed in File Descriptor Flags, the file status flags are shared by duplicated file descriptors resulting from a single opening of the file. The file status flags are specified with the flags argument to open; see section Opening and Closing Files.

File status flags fall into three categories, which are described in the following sections.

• File Access Modes, specify what type of access is allowed to the file: reading, writing, or both. They are set by open and are returned by fcntl, but cannot be changed.
• Open-time Flags, control details of what open will do. These flags are not preserved after the open call.
• I/O Operating Modes, affect how operations such as read and write are done. They are set by open, and can be fetched or changed with fcntl.

The symbols in this section are defined in the header file ‘fcntl.h’.

13.14.1 File Access Modes

The file access modes allow a file descriptor to be used for reading, writing, or both. (In the GNU system, they can also allow none of these, and allow execution of the file as a program.) The access modes are chosen when the file is opened, and never change.

Macro: int O_RDONLY

Open the file for read access.

Macro: int O_WRONLY

Open the file for write access.

Macro: int O_RDWR

Open the file for both reading and writing.

In the GNU system (and not in other systems), O_RDONLY and O_WRONLY are independent bits that can be bitwise-ORed together, and it is valid for either bit to be set or clear. This means that O_RDWR is the same as O_RDONLY|O_WRONLY. A file access mode of zero is permissible; it allows no operations that do input or output to the file, but does allow other operations such as fchmod. On the GNU system, since “read-only” or “write-only” is a misnomer, ‘fcntl.h’ defines additional names for the file access modes. These names are preferred when writing GNU-specific code. But most programs will want to be portable to other POSIX.1 systems and should use the POSIX.1 names above instead.

Macro: int O_READ

Open the file for reading. Same as O_RDONLY; only defined on GNU.

Macro: int O_WRITE

Open the file for writing. Same as O_WRONLY; only defined on GNU.

Macro: int O_EXEC

Open the file for executing. Only defined on GNU.

To determine the file access mode with fcntl, you must extract the access mode bits from the retrieved file status flags. In the GNU system, you can just test the O_READ and O_WRITE bits in the flags word. But in other POSIX.1 systems, reading and writing access modes are not stored as distinct bit flags. The portable way to extract the file access mode bits is with O_ACCMODE.

Macro: int O_ACCMODE

This macro stands for a mask that can be bitwise-ANDed with the file status flag value to produce a value representing the file access mode. The mode will be O_RDONLY, O_WRONLY, or O_RDWR. (In the GNU system it could also be zero, and it never includes the O_EXEC bit.)

13.14.2 Open-time Flags

The open-time flags specify options affecting how open will behave. These options are not preserved once the file is open. The exception to this is O_NONBLOCK, which is also an I/O operating mode and so it is saved. See section Opening and Closing Files, for how to call open.

There are two sorts of options specified by open-time flags.

• File name translation flags affect how open looks up the file name to locate the file, and whether the file can be created.
• Open-time action flags specify extra operations that open will perform on the file once it is open.

Here are the file name translation flags.

Macro: int O_CREAT

If set, the file will be created if it doesn't already exist.

Macro: int O_EXCL

If both O_CREAT and O_EXCL are set, then open fails if the specified file already exists. This is guaranteed to never clobber an existing file.

Macro: int O_NONBLOCK

This prevents open from blocking for a “long time” to open the file. This is only meaningful for some kinds of files, usually devices such as serial ports; when it is not meaningful, it is harmless and ignored. Often opening a port to a modem blocks until the modem reports carrier detection; if O_NONBLOCK is specified, open will return immediately without a carrier.

Note that the O_NONBLOCK flag is overloaded as both an I/O operating mode and a file name translation flag. This means that specifying O_NONBLOCK in open also sets nonblocking I/O mode; see section I/O Operating Modes. To open the file without blocking but do normal I/O that blocks, you must call open with O_NONBLOCK set and then call fcntl to turn the bit off.

Macro: int O_NOCTTY

If the named file is a terminal device, don't make it the controlling terminal for the process. See section Job Control, for information about what it means to be the controlling terminal.

In the GNU system and 4.4 BSD, opening a file never makes it the controlling terminal and O_NOCTTY is zero. However, other systems may use a nonzero value for O_NOCTTY and set the controlling terminal when you open a file that is a terminal device; so to be portable, use O_NOCTTY when it is important to avoid this.

The following three file name translation flags exist only in the GNU system.

Macro: int O_IGNORE_CTTY

Do not recognize the named file as the controlling terminal, even if it refers to the process's existing controlling terminal device. Operations on the new file descriptor will never induce job control signals. See section Job Control.

Macro: int O_NOLINK

If the named file is a symbolic link, open the link itself instead of the file it refers to. (fstat on the new file descriptor will return the information returned by lstat on the link's name.)

Macro: int O_NOTRANS

If the named file is specially translated, do not invoke the translator. Open the bare file the translator itself sees.

The open-time action flags tell open to do additional operations which are not really related to opening the file. The reason to do them as part of open instead of in separate calls is that open can do them atomically.

Macro: int O_TRUNC

Truncate the file to zero length. This option is only useful for regular files, not special files such as directories or FIFOs. POSIX.1 requires that you open the file for writing to use O_TRUNC. In BSD and GNU you must have permission to write the file to truncate it, but you need not open for write access.

This is the only open-time action flag specified by POSIX.1. There is no good reason for truncation to be done by open, instead of by calling ftruncate afterwards. The O_TRUNC flag existed in Unix before ftruncate was invented, and is retained for backward compatibility.

The remaining operating modes are BSD extensions. They exist only on some systems. On other systems, these macros are not defined.

Macro: int O_SHLOCK

Acquire a shared lock on the file, as with flock. See section File Locks.

If O_CREAT is specified, the locking is done atomically when creating the file. You are guaranteed that no other process will get the lock on the new file first.

Macro: int O_EXLOCK

Acquire an exclusive lock on the file, as with flock. See section File Locks. This is atomic like O_SHLOCK.

13.14.3 I/O Operating Modes

The operating modes affect how input and output operations using a file descriptor work. These flags are set by open and can be fetched and changed with fcntl.

Macro: int O_APPEND

The bit that enables append mode for the file. If set, then all write operations write the data at the end of the file, extending it, regardless of the current file position. This is the only reliable way to append to a file. In append mode, you are guaranteed that the data you write will always go to the current end of the file, regardless of other processes writing to the file. Conversely, if you simply set the file position to the end of file and write, then another process can extend the file after you set the file position but before you write, resulting in your data appearing someplace before the real end of file.

Macro: int O_NONBLOCK

The bit that enables nonblocking mode for the file. If this bit is set, read requests on the file can return immediately with a failure status if there is no input immediately available, instead of blocking. Likewise, write requests can also return immediately with a failure status if the output can't be written immediately.

Note that the O_NONBLOCK flag is overloaded as both an I/O operating mode and a file name translation flag; see section Open-time Flags.

Macro: int O_NDELAY

This is an obsolete name for O_NONBLOCK, provided for compatibility with BSD. It is not defined by the POSIX.1 standard.

The remaining operating modes are BSD and GNU extensions. They exist only on some systems. On other systems, these macros are not defined.

Macro: int O_ASYNC

The bit that enables asynchronous input mode. If set, then SIGIO signals will be generated when input is available. See section Interrupt-Driven Input.

Asynchronous input mode is a BSD feature.

Macro: int O_FSYNC

The bit that enables synchronous writing for the file. If set, each write call will make sure the data is reliably stored on disk before returning.

Synchronous writing is a BSD feature.

Macro: int O_SYNC

This is another name for O_FSYNC. They have the same value.

Macro: int O_NOATIME

If this bit is set, read will not update the access time of the file. See section File Times. This is used by programs that do backups, so that backing a file up does not count as reading it. Only the owner of the file or the superuser may use this bit.

This is a GNU extension.

13.14.4 Getting and Setting File Status Flags

The fcntl function can fetch or change file status flags.

Macro: int F_GETFL

This macro is used as the command argument to fcntl, to read the file status flags for the open file with descriptor filedes.

The normal return value from fcntl with this command is a nonnegative number which can be interpreted as the bitwise OR of the individual flags. Since the file access modes are not single-bit values, you can mask off other bits in the returned flags with O_ACCMODE to compare them.

In case of an error, fcntl returns -1. The following errno error conditions are defined for this command:

EBADF

The filedes argument is invalid.

Macro: int F_SETFL

This macro is used as the command argument to fcntl, to set the file status flags for the open file corresponding to the filedes argument. This command requires a third int argument to specify the new flags, so the call looks like this:

fcntl (filedes, F_SETFL, new-flags)

You can't change the access mode for the file in this way; that is, whether the file descriptor was opened for reading or writing.

The normal return value from fcntl with this command is an unspecified value other than -1, which indicates an error. The error conditions are the same as for the F_GETFL command.

If you want to modify the file status flags, you should get the current flags with F_GETFL and modify the value. Don't assume that the flags listed here are the only ones that are implemented; your program may be run years from now and more flags may exist then. For example, here is a function to set or clear the flag O_NONBLOCK without altering any other flags:

/* Set the O_NONBLOCK flag of desc if value is nonzero,
   or clear the flag if value is 0.
   Return 0 on success, or -1 on error with errno set. */

int
set_nonblock_flag (int desc, int value)
{
  int oldflags = fcntl (desc, F_GETFL, 0);
  /* If reading the flags failed, return error indication now. */
  if (oldflags == -1)
    return -1;
  /* Set just the flag we want to set. */
  if (value != 0)
    oldflags |= O_NONBLOCK;
  else
    oldflags &= ~O_NONBLOCK;
  /* Store modified flag word in the descriptor. */
  return fcntl (desc, F_SETFL, oldflags);
}

13.15 File Locks

The remaining fcntl commands are used to support record locking, which permits multiple cooperating programs to prevent each other from simultaneously accessing parts of a file in error-prone ways.

An exclusive or write lock gives a process exclusive access for writing to the specified part of the file. While a write lock is in place, no other process can lock that part of the file.

A shared or read lock prohibits any other process from requesting a write lock on the specified part of the file. However, other processes can request read locks.

The read and write functions do not actually check to see whether there are any locks in place. If you want to implement a locking protocol for a file shared by multiple processes, your application must do explicit fcntl calls to request and clear locks at the appropriate points.

Locks are associated with processes. A process can only have one kind of lock set for each byte of a given file. When any file descriptor for that file is closed by the process, all of the locks that process holds on that file are released, even if the locks were made using other descriptors that remain open. Likewise, locks are released when a process exits, and are not inherited by child processes created using fork (see section Creating a Process).

When making a lock, use a struct flock to specify what kind of lock and where. This data type and the associated macros for the fcntl function are declared in the header file ‘fcntl.h’.

Data Type: struct flock

This structure is used with the fcntl function to describe a file lock. It has these members:

short int l_type

Specifies the type of the lock; one of F_RDLCK, F_WRLCK, or F_UNLCK.

short int l_whence

This corresponds to the whence argument to fseek or lseek, and specifies what the offset is relative to. Its value can be one of SEEK_SET, SEEK_CUR, or SEEK_END.

off_t l_start

This specifies the offset of the start of the region to which the lock applies, and is given in bytes relative to the point specified by l_whence member.

off_t l_len

This specifies the length of the region to be locked. A value of 0 is treated specially; it means the region extends to the end of the file.

pid_t l_pid

This field is the process ID (see section Process Creation Concepts) of the process holding the lock. It is filled in by calling fcntl with the F_GETLK command, but is ignored when making a lock.

Macro: int F_GETLK

This macro is used as the command argument to fcntl, to specify that it should get information about a lock. This command requires a third argument of type struct flock * to be passed to fcntl, so that the form of the call is:

fcntl (filedes, F_GETLK, lockp)

If there is a lock already in place that would block the lock described by the lockp argument, information about that lock overwrites *lockp. Existing locks are not reported if they are compatible with making a new lock as specified. Thus, you should specify a lock type of F_WRLCK if you want to find out about both read and write locks, or F_RDLCK if you want to find out about write locks only.

There might be more than one lock affecting the region specified by the lockp argument, but fcntl only returns information about one of them. The l_whence member of the lockp structure is set to SEEK_SET and the l_start and l_len fields set to identify the locked region.

If no lock applies, the only change to the lockp structure is to update the l_type to a value of F_UNLCK.

The normal return value from fcntl with this command is an unspecified value other than -1, which is reserved to indicate an error. The following errno error conditions are defined for this command:

EBADF

The filedes argument is invalid.

EINVAL

Either the lockp argument doesn't specify valid lock information, or the file associated with filedes doesn't support locks.

Macro: int F_SETLK

This macro is used as the command argument to fcntl, to specify that it should set or clear a lock. This command requires a third argument of type struct flock * to be passed to fcntl, so that the form of the call is:

fcntl (filedes, F_SETLK, lockp)

If the process already has a lock on any part of the region, the old lock on that part is replaced with the new lock. You can remove a lock by specifying a lock type of F_UNLCK.

If the lock cannot be set, fcntl returns immediately with a value of -1. This function does not block waiting for other processes to release locks. If fcntl succeeds, it return a value other than -1.

The following errno error conditions are defined for this function:

EAGAIN

EACCES

The lock cannot be set because it is blocked by an existing lock on the file. Some systems use EAGAIN in this case, and other systems use EACCES; your program should treat them alike, after F_SETLK. (The GNU system always uses EAGAIN.)

EBADF

Either: the filedes argument is invalid; you requested a read lock but the filedes is not open for read access; or, you requested a write lock but the filedes is not open for write access.

EINVAL

Either the lockp argument doesn't specify valid lock information, or the file associated with filedes doesn't support locks.

ENOLCK

The system has run out of file lock resources; there are already too many file locks in place.

Well-designed file systems never report this error, because they have no limitation on the number of locks. However, you must still take account of the possibility of this error, as it could result from network access to a file system on another machine.

Macro: int F_SETLKW

This macro is used as the command argument to fcntl, to specify that it should set or clear a lock. It is just like the F_SETLK command, but causes the process to block (or wait) until the request can be specified.

This command requires a third argument of type struct flock *, as for the F_SETLK command.

The fcntl return values and errors are the same as for the F_SETLK command, but these additional errno error conditions are defined for this command:

EINTR

The function was interrupted by a signal while it was waiting. See section Primitives Interrupted by Signals.

EDEADLK

The specified region is being locked by another process. But that process is waiting to lock a region which the current process has locked, so waiting for the lock would result in deadlock. The system does not guarantee that it will detect all such conditions, but it lets you know if it notices one.

The following macros are defined for use as values for the l_type member of the flock structure. The values are integer constants.

F_RDLCK

This macro is used to specify a read (or shared) lock.

F_WRLCK

This macro is used to specify a write (or exclusive) lock.

F_UNLCK

This macro is used to specify that the region is unlocked.

As an example of a situation where file locking is useful, consider a program that can be run simultaneously by several different users, that logs status information to a common file. One example of such a program might be a game that uses a file to keep track of high scores. Another example might be a program that records usage or accounting information for billing purposes.

Having multiple copies of the program simultaneously writing to the file could cause the contents of the file to become mixed up. But you can prevent this kind of problem by setting a write lock on the file before actually writing to the file.

If the program also needs to read the file and wants to make sure that the contents of the file are in a consistent state, then it can also use a read lock. While the read lock is set, no other process can lock that part of the file for writing.

Remember that file locks are only a voluntary protocol for controlling access to a file. There is still potential for access to the file by programs that don't use the lock protocol.

13.16 Interrupt-Driven Input

If you set the O_ASYNC status flag on a file descriptor (see section File Status Flags), a SIGIO signal is sent whenever input or output becomes possible on that file descriptor. The process or process group to receive the signal can be selected by using the F_SETOWN command to the fcntl function. If the file descriptor is a socket, this also selects the recipient of SIGURG signals that are delivered when out-of-band data arrives on that socket; see Out-of-Band Data. (SIGURG is sent in any situation where select would report the socket as having an “exceptional condition”. See section Waiting for Input or Output.)

If the file descriptor corresponds to a terminal device, then SIGIO signals are sent to the foreground process group of the terminal. See section Job Control.

The symbols in this section are defined in the header file ‘fcntl.h’.

Macro: int F_GETOWN

This macro is used as the command argument to fcntl, to specify that it should get information about the process or process group to which SIGIO signals are sent. (For a terminal, this is actually the foreground process group ID, which you can get using tcgetpgrp; see Functions for Controlling Terminal Access.)

The return value is interpreted as a process ID; if negative, its absolute value is the process group ID.

The following errno error condition is defined for this command:

EBADF

The filedes argument is invalid.

Macro: int F_SETOWN

This macro is used as the command argument to fcntl, to specify that it should set the process or process group to which SIGIO signals are sent. This command requires a third argument of type pid_t to be passed to fcntl, so that the form of the call is:

fcntl (filedes, F_SETOWN, pid)

The pid argument should be a process ID. You can also pass a negative number whose absolute value is a process group ID.

The return value from fcntl with this command is -1 in case of error and some other value if successful. The following errno error conditions are defined for this command:

EBADF

The filedes argument is invalid.

ESRCH

There is no process or process group corresponding to pid.

13.17 Generic I/O Control operations

The GNU system can handle most input/output operations on many different devices and objects in terms of a few file primitives - read, write and lseek. However, most devices also have a few peculiar operations which do not fit into this model. Such as:

• Changing the character font used on a terminal.
• Telling a magnetic tape system to rewind or fast forward. (Since they cannot move in byte increments, lseek is inapplicable).
• Ejecting a disk from a drive.
• Playing an audio track from a CD-ROM drive.
• Maintaining routing tables for a network.

Although some such objects such as sockets and terminals (2) have special functions of their own, it would not be practical to create functions for all these cases.

Instead these minor operations, known as IOCTLs, are assigned code numbers and multiplexed through the ioctl function, defined in sys/ioctl.h. The code numbers themselves are defined in many different headers.

Function: int ioctl (int filedes, int command, …)

The ioctl function performs the generic I/O operation command on filedes.

A third argument is usually present, either a single number or a pointer to a structure. The meaning of this argument, the returned value, and any error codes depends upon the command used. Often -1 is returned for a failure.

On some systems, IOCTLs used by different devices share the same numbers. Thus, although use of an inappropriate IOCTL usually only produces an error, you should not attempt to use device-specific IOCTLs on an unknown device.

Most IOCTLs are OS-specific and/or only used in special system utilities, and are thus beyond the scope of this document. For an example of the use of an IOCTL, see Out-of-Band Data.

This document was generated by root on January, 9 2009 using texi2html 1.78.