--- IO-AIO/AIO.pm 2005/08/28 11:05:50 1.39 +++ IO-AIO/AIO.pm 2006/10/24 03:40:38 1.70 @@ -17,64 +17,80 @@ $_[0] > 0 or die "read error: $!"; }; - # Event + # version 2+ has request and group objects + use IO::AIO 2; + + aioreq_pri 4; # give next request a very high priority + my $req = aio_unlink "/tmp/file", sub { }; + $req->cancel; # cancel request if still in queue + + my $grp = aio_group sub { print "all stats done\n" }; + add $grp aio_stat "..." for ...; + + # AnyEvent integration + open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!"; + my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb }); + + # Event integration Event->io (fd => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb); - # Glib/Gtk2 + # Glib/Gtk2 integration add_watch Glib::IO IO::AIO::poll_fileno, in => sub { IO::AIO::poll_cb; 1 }; - # Tk + # Tk integration Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", readable => \&IO::AIO::poll_cb); - # Danga::Socket + # Danga::Socket integration Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => \&IO::AIO::poll_cb); - =head1 DESCRIPTION This module implements asynchronous I/O using whatever means your operating system supports. Currently, a number of 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. In the future, this module might make use of the native -aio functions available on many operating systems. However, they are often -not well-supported (Linux doesn't allow them on normal files currently, -for example), and they would only support aio_read and aio_write, so the -remaining functionality would have to be implemented using threads anyway. - -Although the module will work with in the presence of other threads, it is -currently not reentrant, so use appropriate locking yourself, always call -C from within the same thread, or never call C (or other -C functions) recursively. +and signal their completion. You don't need thread support in perl, and +the threads created by this module will not be visible to perl. In the +future, this module might make use of the native aio functions available +on many operating systems. However, they are often not well-supported +(Linux doesn't allow them on normal files currently, for example), +and they would only support aio_read and aio_write, so the remaining +functionality would have to be implemented using threads anyway. + +Although the module will work with in the presence of other threads, +it is currently not reentrant in any way, so use appropriate locking +yourself, always call C from within the same thread, or never +call C (or other C functions) recursively. =cut package IO::AIO; no warnings; +use strict 'vars'; use base 'Exporter'; -use Fcntl (); - BEGIN { - $VERSION = 1.6; + our $VERSION = '2.0'; + + our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat + aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink + aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move + aio_group aio_nop); + our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice)); + our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush + min_parallel max_parallel max_outstanding nreqs); - @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat - aio_lstat aio_unlink aio_rmdir aio_readdir aio_symlink - aio_fsync aio_fdatasync aio_readahead); - @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel - max_outstanding nreqs); + @IO::AIO::GRP::ISA = 'IO::AIO::REQ'; require XSLoader; - XSLoader::load IO::AIO, $VERSION; + XSLoader::load ("IO::AIO", $VERSION); } =head1 FUNCTIONS @@ -92,6 +108,9 @@ All functions expecting a filehandle keep a copy of the filehandle internally until the request has finished. +All requests return objects of type L that allow further +manipulation of those requests while they are in-flight. + The pathnames you pass to these routines I be absolute and encoded in byte form. The reason for the former is that at the time the request is being executed, the current working directory could have @@ -107,7 +126,35 @@ =over 4 -=item aio_open $pathname, $flags, $mode, $callback +=item aioreq_pri $pri + +Sets the priority for the next aio request. The default priority +is C<0>, the minimum and maximum priorities are C<-4> and C<4>, +respectively. Requests with higher priority will be serviced first. + +The priority will be reset to C<0> after each call to one of the C +functions. + +Example: open a file with low priority, then read something from it with +higher priority so the read request is serviced before other low priority +open requests (potentially spamming the cache): + + aioreq_pri -3; + aio_open ..., sub { + return unless $_[0]; + + aioreq_pri -2; + aio_read $_[0], ..., sub { + ... + }; + }; + +=item aioreq_nice $pri_adjust + +Similar to C, but subtracts the given value from the current +priority, so effects are cumulative. + +=item aio_open $pathname, $flags, $mode, $callback->($fh) Asynchronously open or create a file and call the callback with a newly created filehandle for the file. @@ -134,7 +181,7 @@ } }; -=item aio_close $fh, $callback +=item aio_close $fh, $callback->($status) Asynchronously close a file and call the callback with the result code. I although accepted, you should not pass in a perl @@ -145,9 +192,9 @@ This is supposed to be a bug in the API, so that might change. It's therefore best to avoid this function. -=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback +=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) -=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback +=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) Reads or writes C bytes from the specified C and C into the scalar given by C and offset C and calls the @@ -166,7 +213,77 @@ print "read $_[0] bytes: <$buffer>\n"; }; -=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback +=item aio_move $srcpath, $dstpath, $callback->($status) + +[EXPERIMENTAL due to internal aio_group use] + +Try to move the I (directories not supported as either source or +destination) from C<$srcpath> to C<$dstpath> and call the callback with +the C<0> (error) or C<-1> ok. + +This is a composite request that tries to rename(2) the file first. If +rename files with C, it creates the destination file with mode 0200 +and copies the contents of the source file into it using C, +followed by restoring atime, mtime, access mode and uid/gid, in that +order, and unlinking the C<$srcpath>. + +If an error occurs, the partial destination file will be unlinked, if +possible, except when setting atime, mtime, access mode and uid/gid, where +errors are being ignored. + +=cut + +sub aio_move($$$) { + my ($src, $dst, $cb) = @_; + + my $grp = aio_group $cb; + + add $grp aio_rename $src, $dst, sub { + if ($_[0] && $! == EXDEV) { + add $grp aio_open $src, O_RDONLY, 0, sub { + if (my $src_fh = $_[0]) { + my @stat = stat $src_fh; + + add $grp aio_open $dst, O_WRONLY, 0200, sub { + if (my $dst_fh = $_[0]) { + add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub { + close $src_fh; + + if ($_[0] == $stat[7]) { + utime $stat[8], $stat[9], $dst; + chmod $stat[2] & 07777, $dst_fh; + chown $stat[4], $stat[5], $dst_fh; + close $dst_fh; + + add $grp aio_unlink $src, sub { + $grp->result ($_[0]); + }; + } else { + my $errno = $!; + add $grp aio_unlink $dst, sub { + $! = $errno; + $grp->result (-1); + }; + } + }; + } else { + $grp->result (-1); + } + }, + + } else { + $grp->result (-1); + } + }; + } else { + $grp->result ($_[0]); + } + }; + + $grp +} + +=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts reading at byte offset C<$in_offset>, and starts writing at the current @@ -189,7 +306,7 @@ value equals C<$length> one can assume that C<$length> bytes have been read. -=item aio_readahead $fh,$offset,$length, $callback +=item aio_readahead $fh,$offset,$length, $callback->($retval) C populates the page cache with data from a file so that subsequent reads from that file will not block on disk I/O. The C<$offset> @@ -203,9 +320,9 @@ If that syscall doesn't exist (likely if your OS isn't Linux) it will be emulated by simply reading the data, which would have a similar effect. -=item aio_stat $fh_or_path, $callback +=item aio_stat $fh_or_path, $callback->($status) -=item aio_lstat $fh, $callback +=item aio_lstat $fh, $callback->($status) Works like perl's C or C in void context. The callback will be called after the stat and the results will be available using C @@ -225,17 +342,32 @@ print "size is ", -s _, "\n"; }; -=item aio_unlink $pathname, $callback +=item aio_unlink $pathname, $callback->($status) Asynchronously unlink (delete) a file and call the callback with the result code. -=item aio_rmdir $pathname, $callback +=item aio_link $srcpath, $dstpath, $callback->($status) + +Asynchronously create a new link to the existing object at C<$srcpath> at +the path C<$dstpath> and call the callback with the result code. + +=item aio_symlink $srcpath, $dstpath, $callback->($status) + +Asynchronously create a new symbolic link to the existing object at C<$srcpath> at +the path C<$dstpath> and call the callback with the result code. + +=item aio_rename $srcpath, $dstpath, $callback->($status) + +Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as +rename(2) and call the callback with the result code. + +=item aio_rmdir $pathname, $callback->($status) Asynchronously rmdir (delete) a directory and call the callback with the result code. -=item aio_readdir $pathname $callback +=item aio_readdir $pathname, $callback->($entries) Unlike the POSIX call of the same name, C reads an entire directory (i.e. opendir + readdir + closedir). The entries will not be @@ -244,12 +376,167 @@ The callback a single argument which is either C or an array-ref with the filenames. -=item aio_fsync $fh, $callback +=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs) + +[EXPERIMENTAL due to internal aio_group use] + +Scans a directory (similar to C) but additionally tries to +separate the entries of directory C<$path> into two sets of names, ones +you can recurse into (directories or links to them), and ones you cannot +recurse into (everything else). + +C is a composite request that creates of many sub requests_ +C<$maxreq> specifies the maximum number of outstanding aio requests that +this function generates. If it is C<< <= 0 >>, then a suitable default +will be chosen (currently 6). + +On error, the callback is called without arguments, otherwise it receives +two array-refs with path-relative entry names. + +Example: + + aio_scandir $dir, 0, sub { + my ($dirs, $nondirs) = @_; + print "real directories: @$dirs\n"; + print "everything else: @$nondirs\n"; + }; + +Implementation notes. + +The C cannot be avoided, but C'ing every entry can. + +After reading the directory, the modification time, size etc. of the +directory before and after the readdir is checked, and if they match (and +isn't the current time), the link count will be used to decide how many +entries are directories (if >= 2). Otherwise, no knowledge of the number +of subdirectories will be assumed. + +Then entries will be sorted into likely directories (everything without +a non-initial dot currently) and likely non-directories (everything +else). Then every entry plus an appended C will be C'ed, +likely directories first. If that succeeds, it assumes that the entry +is a directory or a symlink to directory (which will be checked +seperately). This is often faster than stat'ing the entry itself because +filesystems might detect the type of the entry without reading the inode +data (e.g. ext2fs filetype feature). + +If the known number of directories (link count - 2) has been reached, the +rest of the entries is assumed to be non-directories. + +This only works with certainty on POSIX (= UNIX) filesystems, which +fortunately are the vast majority of filesystems around. + +It will also likely work on non-POSIX filesystems with reduced efficiency +as those tend to return 0 or 1 as link counts, which disables the +directory counting heuristic. + +=cut + +sub aio_scandir($$$) { + my ($path, $maxreq, $cb) = @_; + + my $grp = aio_group $cb; + + $maxreq = 6 if $maxreq <= 0; + + # stat once + add $grp aio_stat $path, sub { + return $grp->result () if $_[0]; + my $now = time; + my $hash1 = join ":", (stat _)[0,1,3,7,9]; + + # read the directory entries + add $grp aio_readdir $path, sub { + my $entries = shift + or return $grp->result (); + + # stat the dir another time + add $grp aio_stat $path, sub { + my $hash2 = join ":", (stat _)[0,1,3,7,9]; + + my $ndirs; + + # take the slow route if anything looks fishy + if ($hash1 ne $hash2 or (stat _)[9] == $now) { + $ndirs = -1; + } else { + # if nlink == 2, we are finished + # on non-posix-fs's, we rely on nlink < 2 + $ndirs = (stat _)[3] - 2 + or return $grp->result ([], $entries); + } + + # sort into likely dirs and likely nondirs + # dirs == files without ".", short entries first + $entries = [map $_->[0], + sort { $b->[1] cmp $a->[1] } + map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length], + @$entries]; + + my (@dirs, @nondirs); + + my ($statcb, $schedcb); + my $nreq = 0; + + my $statgrp = add $grp aio_group; + + $schedcb = sub { + if (@$entries) { + if ($nreq < $maxreq) { + my $ent = pop @$entries; + $nreq++; + add $statgrp aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) }; + } + } elsif (!$nreq) { + # finished + $statgrp->cancel; + undef $statcb; + undef $schedcb; + $grp->result (\@dirs, \@nondirs); + } + }; + $statcb = sub { + my ($status, $entry) = @_; + + if ($status < 0) { + $nreq--; + push @nondirs, $entry; + &$schedcb; + } else { + # need to check for real directory + add $grp aio_lstat "$path/$entry", sub { + $nreq--; + + if (-d _) { + push @dirs, $entry; + + if (!--$ndirs) { + push @nondirs, @$entries; + $entries = []; + } + } else { + push @nondirs, $entry; + } + + &$schedcb; + } + } + }; + + &$schedcb while @$entries && $nreq < $maxreq; + }; + }; + }; + + $grp +} + +=item aio_fsync $fh, $callback->($status) Asynchronously call fsync on the given filehandle and call the callback with the fsync result code. -=item aio_fdatasync $fh, $callback +=item aio_fdatasync $fh, $callback->($status) Asynchronously call fdatasync on the given filehandle and call the callback with the fdatasync result code. @@ -257,6 +544,206 @@ If this call isn't available because your OS lacks it or it couldn't be detected, it will be emulated by calling C instead. +=item aio_group $callback->(...) + +[EXPERIMENTAL] + +This is a very special aio request: Instead of doing something, it is a +container for other aio requests, which is useful if you want to bundle +many requests into a single, composite, request. + +Returns an object of class L. See its documentation below +for more info. + +Example: + + my $grp = aio_group sub { + print "all stats done\n"; + }; + + add $grp + (aio_stat ...), + (aio_stat ...), + ...; + +=item aio_nop $callback->() + +This is a special request - it does nothing in itself and is only used for +side effects, such as when you want to add a dummy request to a group so +that finishing the requests in the group depends on executing the given +code. + +While this request does nothing, it still goes through the execution +phase and still requires a worker thread. Thus, the callback will not +be executed immediately but only after other requests in the queue have +entered their execution phase. This can be used to measure request +latency. + +=item IO::AIO::aio_sleep $fractional_seconds, $callback->() *NOT EXPORTED* + +Mainly used for debugging and benchmarking, this aio request puts one of +the request workers to sleep for the given time. + +While it is theoretically handy to have simple I/O scheduling requests +like sleep and file handle readable/writable, the overhead this creates +is immense, so do not use this function except to put your application +under artificial I/O pressure. + +=back + +=head2 IO::AIO::REQ CLASS + +All non-aggregate C functions return an object of this class when +called in non-void context. + +A request always moves through the following five states in its lifetime, +in order: B (request has been created, but has not been executed +yet), B (request is currently being executed), B +(request has been executed but callback has not been called yet), +B (results are being processed synchronously, includes calling the +callback) and B (request has reached the end of its lifetime and +holds no resources anymore). + +=over 4 + +=item cancel $req + +Cancels the request, if possible. Has the effect of skipping execution +when entering the B state and skipping calling the callback when +entering the the B state, but will leave the request otherwise +untouched. That means that requests that currently execute will not be +stopped and resources held by the request will not be freed prematurely. + +=item cb $req $callback->(...) + +Replace (or simply set) the callback registered to the request. + +=back + +=head2 IO::AIO::GRP CLASS + +This class is a subclass of L, so all its methods apply to +objects of this class, too. + +A IO::AIO::GRP object is a special request that can contain multiple other +aio requests. + +You create one by calling the C constructing function with a +callback that will be called when all contained requests have entered the +C state: + + my $grp = aio_group sub { + print "all requests are done\n"; + }; + +You add requests by calling the C method with one or more +C objects: + + $grp->add (aio_unlink "..."); + + add $grp aio_stat "...", sub { + $_[0] or return $grp->result ("error"); + + # add another request dynamically, if first succeeded + add $grp aio_open "...", sub { + $grp->result ("ok"); + }; + }; + +This makes it very easy to create composite requests (see the source of +C for an application) that work and feel like simple requests. + +=over 4 + +=item * The IO::AIO::GRP objects will be cleaned up during calls to +C, just like any other request. + +=item * They can be canceled like any other request. Canceling will cancel not +only the request itself, but also all requests it contains. + +=item * They can also can also be added to other IO::AIO::GRP objects. + +=item * You must not add requests to a group from within the group callback (or +any later time). + +=item * This does not harmonise well with C, so best do +not combine C with it. Groups and feeders are recommended for +this kind of concurrency-limiting. + +=back + +Their lifetime, simplified, looks like this: when they are empty, they +will finish very quickly. If they contain only requests that are in the +C state, they will also finish. Otherwise they will continue to +exist. + +That means after creating a group you have some time to add requests. And +in the callbacks of those requests, you can add further requests to the +group. And only when all those requests have finished will the the group +itself finish. + +=over 4 + +=item add $grp ... + +=item $grp->add (...) + +Add one or more requests to the group. Any type of L can +be added, including other groups, as long as you do not create circular +dependencies. + +Returns all its arguments. + +=item $grp->result (...) + +Set the result value(s) that will be passed to the group callback when all +subrequests have finished. By default, no argument will be passed. + +=item feed $grp $callback->($grp) + +[VERY EXPERIMENTAL] + +Sets a feeder/generator on this group: every group can have an attached +generator that generates requests if idle. The idea behind this is that, +although you could just queue as many requests as you want in a group, +this might starve other requests for a potentially long time. For +example, C might generate hundreds of thousands C +requests, delaying any later requests for a long time. + +To avoid this, and allow incremental generation of requests, you can +instead a group and set a feeder on it that generates those requests. The +feed callback will be called whenever there are few enough (see C, +below) requests active in the group itself and is expected to queue more +requests. + +The feed callback can queue as many requests as it likes (i.e. C does +not impose any limits). + +If the feed does not queue more requests when called, it will be +automatically removed from the group. + +If the feed limit is C<0>, it will be set to C<2> automatically. + +Example: + + # stat all files in @files, but only ever use four aio requests concurrently: + + my $grp = aio_group sub { print "finished\n" }; + limit $grp 4; + feed $grp sub { + my $file = pop @files + or return; + + add $grp aio_stat $file, sub { ... }; + }; + +=item limit $grp $num + +Sets the feeder limit for the group: The feeder will be called whenever +the group contains less than this many requests. + +Setting the limit to C<0> will pause the feeding process. + =back =head2 SUPPORT FUNCTIONS @@ -323,17 +810,18 @@ =item IO::AIO::min_parallel $nthreads -Set the minimum number of AIO threads to C<$nthreads>. The current default -is C<4>, which means four asynchronous operations can be done at one time -(the number of outstanding operations, however, is unlimited). +Set the minimum number of AIO threads to C<$nthreads>. The current +default is C<8>, which means eight asynchronous operations can execute +concurrently at any one time (the number of outstanding requests, +however, is unlimited). IO::AIO starts threads only on demand, when an AIO request is queued and no free thread exists. -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). With current Linux 2.6 versions, 4-32 -threads should be fine. +It is recommended to keep the number of threads relatively low, as some +Linux kernel versions will scale negatively with the number of threads +(higher parallelity => MUCH higher latency). With current Linux 2.6 +versions, 4-32 threads should be fine. Under most circumstances you don't need to call this function, as the module selects a default that is suitable for low to moderate load. @@ -354,6 +842,8 @@ =item $oldnreqs = IO::AIO::max_outstanding $nreqs +[DEPRECATED] + Sets the maximum number of outstanding requests to C<$nreqs>. If you try to queue up more than this number of requests, the caller will block until some requests have been handled. @@ -362,6 +852,10 @@ queue up many requests in a loop it often improves speed if you set this to a relatively low number, such as C<100>. +This function does not work well together with C's, and their +feeder interface is better suited to limiting concurrency, so do not use +this function. + Under normal circumstances you don't need to call this function. =back @@ -384,7 +878,7 @@ *$sym } -min_parallel 4; +min_parallel 8; END { max_parallel 0; @@ -394,17 +888,36 @@ =head2 FORK BEHAVIOUR +This module should do "the right thing" when the process using it forks: + Before the fork, IO::AIO enters a quiescent state where no requests can be added in other threads and no results will be processed. After the fork the parent simply leaves the quiescent state and continues request/result processing, while the child clears the request/result queue (so the requests started before the fork will only be handled in -the parent). Threats will be started on demand until the limit ste in the +the parent). Threads will be started on demand until the limit ste in the parent process has been reached again. +In short: the parent will, after a short pause, continue as if fork had +not been called, while the child will act as if IO::AIO has not been used +yet. + +=head2 MEMORY USAGE + +Each aio request uses - depending on your architecture - around 128 bytes +of memory. In addition, stat requests need a stat buffer (possibly a few +hundred bytes). Perl scalars and other data passed into aio requests will +also be locked. + +This is now awfully much, so queuing lots of requests is not usually a +problem. + +Each thread needs a stack area which is usually around 16k, sometimes much +larger, depending on the OS. + =head1 SEE ALSO -L, L. +L. =head1 AUTHOR