NAME
   Linux::AIO - linux-specific aio implemented using clone

SYNOPSIS
    use Linux::AIO;

    # This module has been mostly superseded by IO::AIO.

DESCRIPTION
   *This module has been mostly superseded by IO::AIO, which is API
   compatible.*

   This module implements asynchronous I/O using the means available to
   Linux - clone. It does not hook into the POSIX aio_* functions because
   Linux does not yet support these in the kernel (even as of 2.6.12, only
   O_DIRECT files are supported) and even if, it would only allow aio_read
   and write, not open, stat and so on.

   Instead, in this module a number of (non-posix) threads are started that
   execute your read/writes and signal their completion. You don't need
   thread support in your libc or perl, and the threads created by this
   module will not be visible to the pthreads library.

   NOTICE: the threads created by this module will automatically be killed
   when the thread calling min_parallel exits. Make sure you only ever call
   min_parallel from the same thread that loaded this module.

   Although the module will work with in the presence of other threads, it
   is not reentrant, so use appropriate locking yourself.

 API NOTES
   All the "aio_*" calls are more or less thin wrappers around the syscall
   with the same name (sans "aio_"). The arguments are similar or
   identical, and they all accept an additional $callback argument which
   must be a code reference. This code reference will get called with the
   syscall return code (e.g. most syscalls return -1 on error, unlike perl,
   which usually delivers "false") as it's sole argument when the given
   syscall has been executed asynchronously.

   All functions that expect a filehandle will also accept a file
   descriptor.

   The filenames you pass to these routines *must* be absolute. The reason
   is that at the time the request is being executed, the current working
   directory could have changed. Alternatively, you can make sure that you
   never change the current working directory.

   Linux::AIO::min_parallel $nthreads
       Set the minimum number of AIO threads to $nthreads. The default is
       1, which means a single asynchronous operation can be done at one
       time (the number of outstanding operations, however, is unlimited).

       It is recommended to keep the number of threads low, as some linux
       kernel versions will scale negatively with the number of threads
       (higher parallelity => MUCH higher latency).

       Under normal circumstances you don't need to call this function, as
       this module automatically starts a single async thread.

   Linux::AIO::max_parallel $nthreads
       Sets the maximum number of AIO threads to $nthreads. If more than
       the specified number of threads are currently running, kill them.
       This function blocks until the limit is reached.

       This module automatically runs "max_parallel 0" at program end, to
       ensure that all threads are killed and that there are no outstanding
       requests.

       Under normal circumstances you don't need to call this function.

   $fileno = Linux::AIO::poll_fileno
       Return the *request result pipe filehandle*. This filehandle must be
       polled for reading by some mechanism outside this module (e.g. Event
       or select, see below). If the pipe becomes readable you have to call
       "poll_cb" to check the results.

       See "poll_cb" for an example.

   Linux::AIO::poll_cb
       Process all outstanding events on the result pipe. You have to call
       this regularly. Returns the number of events processed. Returns
       immediately when no events are outstanding.

       You can use Event to multiplex, e.g.:

          Event->io (fd => Linux::AIO::poll_fileno,
                     poll => 'r', async => 1,
                     cb => \&Linux::AIO::poll_cb);

   Linux::AIO::poll_wait
       Wait till the result filehandle becomes ready for reading (simply
       does a select on the filehandle. This is useful if you want to
       synchronously wait for some requests to finish).

       See "nreqs" for an example.

   Linux::AIO::nreqs
       Returns the number of requests currently outstanding.

       Example: wait till there are no outstanding requests anymore:

          Linux::AIO::poll_wait, Linux::AIO::poll_cb
             while Linux::AIO::nreqs;

   aio_open $pathname, $flags, $mode, $callback
       Asynchronously open or create a file and call the callback with the
       filedescriptor (NOT a perl filehandle, sorry for that, but watch
       out, this might change in the future).

       The pathname passed to "aio_open" must be absolute. See API NOTES,
       above, for an explanation.

       The $mode argument is a bitmask. See the "Fcntl" module for a list.
       They are the same as used in "sysopen".

       Example:

          aio_open "/etc/passwd", O_RDONLY, 0, sub {
             if ($_[0] >= 0) {
                open my $fh, "<&=$_[0]";
                print "open successful, fh is $fh\n";
                ...
             } else {
                die "open failed: $!\n";
             }
          };

   aio_close $fh, $callback
       Asynchronously close a file and call the callback with the result
       code.

   aio_read $fh,$offset,$length, $data,$dataoffset,$callback
   aio_write $fh,$offset,$length, $data,$dataoffset,$callback
       Reads or writes "length" bytes from the specified "fh" and "offset"
       into the scalar given by "data" and offset "dataoffset" and calls
       the callback without the actual number of bytes read (or -1 on
       error, just like the syscall).

       Example: Read 15 bytes at offset 7 into scalar $buffer, strating at
       offset 0 within the scalar:

          aio_read $fh, 7, 15, $buffer, 0, sub {
             $_[0] >= 0 or die "read error: $!";
             print "read <$buffer>\n";
          };

   aio_readahead $fh,$offset,$length, $callback
       Asynchronously reads the specified byte range into the page cache,
       using the "readahead" syscall.

       readahead() populates the page cache with data from a file so that
       subsequent reads from that file will not block on disk I/O. The
       $offset argument specifies the starting point from which data is to
       be read and $length specifies the number of bytes to be read. I/O is
       performed in whole pages, so that offset is effectively rounded down
       to a page boundary and bytes are read up to the next page boundary
       greater than or equal to (off-set+length). aio_readahead() does not
       read beyond the end of the file. The current file offset of the file
       is left unchanged.

   aio_stat $fh_or_path, $callback
   aio_lstat $fh, $callback
       Works like perl's "stat" or "lstat" in void context. The callback
       will be called after the stat and the results will be available
       using "stat _" or "-s _" etc...

       The pathname passed to "aio_stat" must be absolute. See API NOTES,
       above, for an explanation.

       Currently, the stats are always 64-bit-stats, i.e. instead of
       returning an error when stat'ing a large file, the results will be
       silently truncated unless perl itself is compiled with large file
       support.

       Example: Print the length of /etc/passwd:

          aio_stat "/etc/passwd", sub {
             $_[0] and die "stat failed: $!";
             print "size is ", -s _, "\n";
          };

   aio_unlink $pathname, $callback
       Asynchronously unlink (delete) a file and call the callback with the
       result code.

   aio_fsync $fh, $callback
       Asynchronously call fsync on the given filehandle and call the
       callback with the fsync result code.

   aio_fdatasync $fh, $callback
       Asynchronously call fdatasync on the given filehandle and call the
       callback with the fdatasync result code.

BUGS
   This module has been extensively tested in a large and very busy
   webserver for many years now.

      - aio_open gives a fd, but all other functions expect a perl filehandle.

SEE ALSO
   Coro, IO::AIO.

AUTHOR
    Marc Lehmann <[email protected]>
    http://home.schmorp.de/

You are viewing proxied material from ftp.icm.edu.pl. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.