ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/cvsroot/Async-Interrupt/Interrupt.pm
(Generate patch)

Comparing cvsroot/Async-Interrupt/Interrupt.pm (file contents):
Revision 1.2 by root, Thu Jul 2 15:13:03 2009 UTC vs.
Revision 1.18 by root, Tue Jul 28 12:50:16 2009 UTC

7 use Async::Interrupt; 7 use Async::Interrupt;
8 8
9=head1 DESCRIPTION 9=head1 DESCRIPTION
10 10
11This module implements a single feature only of interest to advanced perl 11This module implements a single feature only of interest to advanced perl
12modules, namely asynchronous interruptions (think "unix signals", which 12modules, namely asynchronous interruptions (think "UNIX signals", which
13are very similar). 13are very similar).
14 14
15Sometimes, modules wish to run code asynchronously (in another thread), 15Sometimes, modules wish to run code asynchronously (in another thread,
16and then signal the perl interpreter on certain events. One common way is 16or from a signal handler), and then signal the perl interpreter on
17to write some data to a pipe and use an event handling toolkit to watch 17certain events. One common way is to write some data to a pipe and use an
18for I/O events. Another way is to send a signal. Those methods are slow, 18event handling toolkit to watch for I/O events. Another way is to send
19and in the case of a pipe, also not asynchronous - it won't interrupt a 19a signal. Those methods are slow, and in the case of a pipe, also not
20running perl interpreter. 20asynchronous - it won't interrupt a running perl interpreter.
21 21
22This module implements asynchronous notifications that enable you to 22This module implements asynchronous notifications that enable you to
23signal running perl code form another thread, asynchronously, without 23signal running perl code from another thread, asynchronously, and
24issuing syscalls. 24sometimes even without using a single syscall.
25 25
26It works by creating an C<Async::Interrupt> object for each such use. This 26=head2 USAGE SCENARIOS
27object stores a perl and/or a C-level callback that is invoked when the 27
28C<Async::Interrupt> object gets signalled. It is executed at the next time 28=over 4
29the perl interpreter is running (i.e. it will interrupt a computation, but 29
30not an XS function or a syscall). 30=item Race-free signal handling
31
32There seems to be no way to do race-free signal handling in perl: to
33catch a signal, you have to execute Perl code, and between entering the
34interpreter C<select> function (or other blocking functions) and executing
35the select syscall is a small but relevant timespan during which signals
36will be queued, but perl signal handlers will not be executed and the
37blocking syscall will not be interrupted.
38
39You can use this module to bind a signal to a callback while at the same
40time activating an event pipe that you can C<select> on, fixing the race
41completely.
42
43This can be used to implement the signal hadling in event loops,
44e.g. L<AnyEvent>, L<POE>, L<IO::Async::Loop> and so on.
45
46=item Background threads want speedy reporting
47
48Assume you want very exact timing, and you can spare an extra cpu core
49for that. Then you can run an extra thread that signals your perl
50interpreter. This means you can get a very exact timing source while your
51perl code is number crunching, without even using a syscall to communicate
52between your threads.
53
54For example the deliantra game server uses a variant of this technique
55to interrupt background processes regularly to send map updates to game
56clients.
57
58Or L<EV::Loop::Async> uses an interrupt object to wake up perl when new
59events have arrived.
60
61L<IO::AIO> and L<BDB> could also use this to speed up result reporting.
62
63=item Speedy event loop invocation
64
65One could use this module e.g. in L<Coro> to interrupt a running coro-thread
66and cause it to enter the event loop.
67
68Or one could bind to C<SIGIO> and tell some important sockets to send this
69signal, causing the event loop to be entered to reduce network latency.
70
71=back
72
73=head2 HOW TO USE
74
75You can use this module by creating an C<Async::Interrupt> object for each
76such event source. This object stores a perl and/or a C-level callback
77that is invoked when the C<Async::Interrupt> object gets signalled. It is
78executed at the next time the perl interpreter is running (i.e. it will
79interrupt a computation, but not an XS function or a syscall).
31 80
32You can signal the C<Async::Interrupt> object either by calling it's C<< 81You can signal the C<Async::Interrupt> object either by calling it's C<<
33->signal >> method, or, more commonly, by calling a C function. 82->signal >> method, or, more commonly, by calling a C function. There is
83also the built-in (POSIX) signal source.
34 84
35The C<< ->signal_func >> returns the address of the C function that is to 85The C<< ->signal_func >> returns the address of the C function that is to
36be called (plus an argument to be used during the call). The signalling 86be called (plus an argument to be used during the call). The signalling
37function also takes an integer argument in the range SIG_ATOMIC_MIN to 87function also takes an integer argument in the range SIG_ATOMIC_MIN to
38SIG_ATOMIC_MAX (guaranteed to allow at least 0..127). 88SIG_ATOMIC_MAX (guaranteed to allow at least 0..127).
39 89
40Since this kind of interruption is fast, but can only interrupt a 90Since this kind of interruption is fast, but can only interrupt a
41I<running> interpreter, there is optional support for also signalling a 91I<running> interpreter, there is optional support for signalling a pipe
42pipe - that means you can also wait for the pipe to become readable while 92- that means you can also wait for the pipe to become readable (e.g. via
93L<EV> or L<AnyEvent>). This, of course, incurs the overhead of a C<read>
94and C<write> syscall.
95
96=head1 USAGE EXAMPLES
97
98=head2 Async::Interrupt to implement race-free signal handling
99
100This example uses a single event pipe for all signals, and one
101Async::Interrupt per signal.
102
103First, create the event pipe and hook it into the event loop (this code is
104actually what L<AnyEvent> uses itself):
105
106 $SIGPIPE = new Async::Interrupt::EventPipe;
107 $SIGPIPE_W = AnyEvent->io (
108 fh => $SIGPIPE->fileno,
109 poll => "r",
110 cb => \&_signal_check,
111 );
112
113Then, for each signal to hook, create an Async::Interrupt object. The
114callback just sets a global variable, as we are only interested in
115synchronous signals (i.e. when the event loop polls), which is why the
116pipe draining is not done automatically.
117
118 my $interrupt = new Async::Interrupt
119 cb => sub { undef $SIGNAL_RECEIVED{$signum} }
120 signal => $signal,
121 pipe => [$SIGPIPE_R->filenos],
122 pipe_autodrain => 0,
123 ;
124
125Finally, the I/O callback for the event pipe handles the signals:
126
127 sub _signal_check {
128 # drain the pipe first
129 $SIGPIPE->drain;
130
131 # two loops, just to be sure
132 while (%SIGNAL_RECEIVED) {
133 for (keys %SIGNAL_RECEIVED) {
134 delete $SIGNAL_RECEIVED{$_};
135 warn "signal $_ received\n";
136 }
137 }
138 }
139
140
141
142=head1 THE Async::Interrupt CLASS
43 143
44=over 4 144=over 4
45 145
46=cut 146=cut
47 147
48package Async::Interrupt; 148package Async::Interrupt;
49 149
50no warnings; 150use common::sense;
51 151
52BEGIN { 152BEGIN {
153 # the next line forces initialisation of internal
154 # signal handling # variables
155 $SIG{KILL} = sub { };
156
53 $VERSION = '0.02'; 157 our $VERSION = '0.6';
54 158
55 require XSLoader; 159 require XSLoader;
56 XSLoader::load Async::Interrupt::, $VERSION; 160 XSLoader::load ("Async::Interrupt", $VERSION);
57} 161}
58 162
59our $DIED = sub { warn "$@" }; 163our $DIED = sub { warn "$@" };
60 164
61=item $async = new Async::Interrupt key => value... 165=item $async = new Async::Interrupt key => value...
99C<$value> is the C<value> passed to some earlier call to either C<$signal> 203C<$value> is the C<value> passed to some earlier call to either C<$signal>
100or the C<signal_func> function. 204or the C<signal_func> function.
101 205
102Note that, because the callback can be invoked at almost any time, you 206Note that, because the callback can be invoked at almost any time, you
103have to be careful at saving and restoring global variables that Perl 207have to be careful at saving and restoring global variables that Perl
104might use (the excetpion is C<errno>, which is aved and restored by 208might use (the exception is C<errno>, which is saved and restored by
105Async::Interrupt). The callback itself runs as part of the perl context, 209Async::Interrupt). The callback itself runs as part of the perl context,
106so you can call any perl functions and modify any perl data structures (in 210so you can call any perl functions and modify any perl data structures (in
107which case the requireemnts set out for C<cb> apply as well). 211which case the requirements set out for C<cb> apply as well).
212
213=item var => $scalar_ref
214
215When specified, then the given argument must be a reference to a
216scalar. The scalar will be set to C<0> initially. Signalling the interrupt
217object will set it to the passed value, handling the interrupt will reset
218it to C<0> again.
219
220Note that the only thing you are legally allowed to do is to is to check
221the variable in a boolean or integer context (e.g. comparing it with a
222string, or printing it, will I<destroy> it and might cause your program to
223crash or worse).
224
225=item signal => $signame_or_value
226
227When this parameter is specified, then the Async::Interrupt will hook the
228given signal, that is, it will effectively call C<< ->signal (0) >> each time
229the given signal is caught by the process.
230
231Only one async can hook a given signal, and the signal will be restored to
232defaults when the Async::Interrupt object gets destroyed.
108 233
109=item pipe => [$fileno_or_fh_for_reading, $fileno_or_fh_for_writing] 234=item pipe => [$fileno_or_fh_for_reading, $fileno_or_fh_for_writing]
110 235
111Specifies two file descriptors (or file handles) that should be signalled 236Specifies two file descriptors (or file handles) that should be signalled
112whenever the async interrupt is signalled. This means a single octet will 237whenever the async interrupt is signalled. This means a single octet will
113be written to it, and before the callback is being invoked, it will be 238be written to it, and before the callback is being invoked, it will be
114read again. Due to races, it is unlikely but possible that multiple octets 239read again. Due to races, it is unlikely but possible that multiple octets
115are written. It is required that the file handles are both in nonblocking 240are written. It is required that the file handles are both in nonblocking
116mode. 241mode.
117 242
118(You can get a portable pipe and set non-blocking mode portably by using
119e.g. L<AnyEvent::Util> from the L<AnyEvent> distro).
120
121The object will keep a reference to the file handles. 243The object will keep a reference to the file handles.
122 244
123This can be used to ensure that async notifications will interrupt event 245This can be used to ensure that async notifications will interrupt event
124frameworks as well. 246frameworks as well.
125 247
248Note that C<Async::Interrupt> will create a suitable signal fd
249automatically when your program requests one, so you don't have to specify
250this argument when all you want is an extra file descriptor to watch.
251
252If you want to share a single event pipe between multiple Async::Interrupt
253objects, you can use the C<Async::Interrupt::EventPipe> class to manage
254those.
255
126=back 256=back
127 257
128=cut 258=cut
129 259
130sub new { 260sub new {
131 my ($class, %arg) = @_; 261 my ($class, %arg) = @_;
132 262
133 bless \(_alloc $arg{cb}, @{$arg{c_cb}}[0,1], @{$arg{pipe}}[0,1]), $class 263 bless \(_alloc $arg{cb}, @{$arg{c_cb}}[0,1], @{$arg{pipe}}[0,1], $arg{signal}, $arg{var}), $class
134} 264}
135 265
136=item ($signal_func, $signal_arg) = $async->signal_func 266=item ($signal_func, $signal_arg) = $async->signal_func
137 267
138Returns the address of a function to call asynchronously. The function has 268Returns the address of a function to call asynchronously. The function
139the following prototype and needs to be passed the specified C<$c_arg>, 269has the following prototype and needs to be passed the specified
140which is a C<void *> cast to C<IV>: 270C<$signal_arg>, which is a C<void *> cast to C<IV>:
141 271
142 void (*signal_func) (void *signal_arg, int value) 272 void (*signal_func) (void *signal_arg, int value)
143 273
144An example call would look like: 274An example call would look like:
145 275
146 signal_func (signal_arg, 0); 276 signal_func (signal_arg, 0);
147 277
148The function is safe to call from within signal and thread contexts, at 278The function is safe to call from within signal and thread contexts, at
149any time. The specified C<value> is passed to both C and Perl callback. 279any time. The specified C<value> is passed to both C and Perl callback.
150 280
151C<$value> must be in the valid range for a C<sig_atomic_t> (0..127 is 281C<$value> must be in the valid range for a C<sig_atomic_t>, except C<0>
152portable). 282(1..127 is portable).
153 283
154If the function is called while the Async::Interrupt object is already 284If the function is called while the Async::Interrupt object is already
155signaled but before the callbacks are being executed, then the stored 285signaled but before the callbacks are being executed, then the stored
156C<value> is either the old or the new one. Due to the asynchronous 286C<value> is either the old or the new one. Due to the asynchronous
157nature of the code, the C<value> can even be passed to two consecutive 287nature of the code, the C<value> can even be passed to two consecutive
158invocations of the callback. 288invocations of the callback.
159 289
290=item $address = $async->c_var
291
292Returns the address (cast to IV) of an C<IV> variable. The variable is set
293to C<0> initially and gets set to the passed value whenever the object
294gets signalled, and reset to C<0> once the interrupt has been handled.
295
296Note that it is often beneficial to just call C<PERL_ASYNC_CHECK ()> to
297handle any interrupts.
298
299Example: call some XS function to store the address, then show C code
300waiting for it.
301
302 my_xs_func $async->c_var;
303
304 static IV *valuep;
305
306 void
307 my_xs_func (void *addr)
308 CODE:
309 valuep = (IV *)addr;
310
311 // code in a loop, waiting
312 while (!*valuep)
313 ; // do something
314
160=item $async->signal ($value=0) 315=item $async->signal ($value=1)
161 316
162This signals the given async object from Perl code. Semi-obviously, this 317This signals the given async object from Perl code. Semi-obviously, this
163will instantly trigger the callback invocation. 318will instantly trigger the callback invocation (it does not, as the name
319might imply, do anything with POSIX signals).
164 320
165C<$value> must be in the valid range for a C<sig_atomic_t> (0..127 is 321C<$value> must be in the valid range for a C<sig_atomic_t>, except C<0>
166portable). 322(1..127 is portable).
323
324=item $async->signal_hysteresis ($enable)
325
326Enables or disables signal hysteresis (default: disabled). If a POSIX
327signal is used as a signal source for the interrupt object, then enabling
328signal hysteresis causes Async::Interrupt to reset the signal action to
329C<SIG_IGN> in the signal handler and restore it just before handling the
330interruption.
331
332When you expect a lot of signals (e.g. when using SIGIO), then enabling
333signal hysteresis can reduce the number of handler invocations
334considerably, at the cost of two extra syscalls.
335
336Note that setting the signal to C<SIG_IGN> can have unintended side
337effects when you fork and exec other programs, as often they do nto expect
338signals to be ignored by default.
167 339
168=item $async->block 340=item $async->block
169 341
170Sometimes you need a "critical section" of code where
171
172=item $async->unblock 342=item $async->unblock
173 343
344Sometimes you need a "critical section" of code that will not be
345interrupted by an Async::Interrupt. This can be implemented by calling C<<
346$async->block >> before the critical section, and C<< $async->unblock >>
347afterwards.
348
349Note that there must be exactly one call of C<unblock> for every previous
350call to C<block> (i.e. calls can nest).
351
352Since ensuring this in the presence of exceptions and threads is
353usually more difficult than you imagine, I recommend using C<<
354$async->scoped_block >> instead.
355
356=item $async->scope_block
357
358This call C<< $async->block >> and installs a handler that is called when
359the current scope is exited (via an exception, by canceling the Coro
360thread, by calling last/goto etc.).
361
362This is the recommended (and fastest) way to implement critical sections.
363
364=item ($block_func, $block_arg) = $async->scope_block_func
365
366Returns the address of a function that implements the C<scope_block>
367functionality.
368
369It has the following prototype and needs to be passed the specified
370C<$block_arg>, which is a C<void *> cast to C<IV>:
371
372 void (*block_func) (void *block_arg)
373
374An example call would look like:
375
376 block_func (block_arg);
377
378The function is safe to call only from within the toplevel of a perl XS
379function and will call C<LEAVE> and C<ENTER> (in this order!).
380
381=item $async->pipe_enable
382
383=item $async->pipe_disable
384
385Enable/disable signalling the pipe when the interrupt occurs (default is
386enabled). Writing to a pipe is relatively expensive, so it can be disabled
387when you know you are not waiting for it (for example, with L<EV> you
388could disable the pipe in a check watcher, and enable it in a prepare
389watcher).
390
391Note that currently, while C<pipe_disable> is in effect, no attempt to
392read from the pipe will be done when handling events. This might change as
393soon as I realize why this is a mistake.
394
395=item $fileno = $async->pipe_fileno
396
397Returns the reading side of the signalling pipe. If no signalling pipe is
398currently attached to the object, it will dynamically create one.
399
400Note that the only valid oepration on this file descriptor is to wait
401until it is readable. The fd might belong currently to a pipe, a tcp
402socket, or an eventfd, depending on the platform, and is guaranteed to be
403C<select>able.
404
405=item $async->pipe_autodrain ($enable)
406
407Enables (C<1>) or disables (C<0>) automatic draining of the pipe (default:
408enabled). When automatic draining is enabled, then Async::Interrupt will
409automatically clear the pipe. Otherwise the user is responsible for this
410draining.
411
412This is useful when you want to share one pipe among many Async::Interrupt
413objects.
414
415=item $async->post_fork
416
417The object will not normally be usable after a fork (as the pipe fd is
418shared between processes). Calling this method after a fork in the child
419ensures that the object will work as expected again. It only needs to be
420called when the async object is used in the child.
421
422This only works when the pipe was created by Async::Interrupt.
423
424Async::Interrupt ensures that the reading file descriptor does not change
425it's value.
426
427=item $signum = Async::Interrupt::sig2num $signame_or_number
428
429=item $signame = Async::Interrupt::sig2name $signame_or_number
430
431These two convenience functions simply convert a signal name or number to
432the corresponding name or number. They are not used by this module and
433exist just because perl doesn't have a nice way to do this on its own.
434
435They will return C<undef> on illegal names or numbers.
436
437=back
438
439=head1 THE Async::Interrupt::EventPipe CLASS
440
441Pipes are the predominent utility to make asynchronous signals
442synchronous. However, pipes are hard to come by: they don't exist on the
443broken windows platform, and on GNU/Linux systems, you might want to use
444an C<eventfd> instead.
445
446This class creates selectable event pipes in a portable fashion: on
447windows, it will try to create a tcp socket pair, on GNU/Linux, it will
448try to create an eventfd and everywhere else it will try to use a normal
449pipe.
450
451=over 4
452
453=item $epipe = new Async::Interrupt::EventPipe
454
455This creates and returns an eventpipe object. This object is simply a
456blessed array reference:
457
458=item ($r_fd, $w_fd) = $epipe->filenos
459
460Returns the read-side file descriptor and the write-side file descriptor.
461
462Example: pass an eventpipe object as pipe to the Async::Interrupt
463constructor, and create an AnyEvent watcher for the read side.
464
465 my $epipe = new Async::Interrupt::EventPipe;
466 my $asy = new Async::Interrupt pipe => [$epipe->filenos];
467 my $iow = AnyEvent->io (fh => $epipe->fileno, poll => 'r', cb => sub { });
468
469=item $r_fd = $epipe->fileno
470
471Return only the reading/listening side.
472
473=item $epipe->signal
474
475Write something to the pipe, in a portable fashion.
476
477=item $epipe->drain
478
479Drain (empty) the pipe.
480
481=item $epipe->renew
482
483Recreates the pipe (useful after a fork). The reading side will not change
484it's file descriptor number, but the writing side might.
485
486=back
487
174=cut 488=cut
175 489
1761; 4901;
177 491
178=back
179
180=head1 EXAMPLE 492=head1 EXAMPLE
181 493
182#TODO 494There really should be a complete C/XS example. Bug me about it. Better
495yet, create one.
183 496
184=head1 IMPLEMENTATION DETAILS AND LIMITATIONS 497=head1 IMPLEMENTATION DETAILS AND LIMITATIONS
185 498
186This module works by "hijacking" SIGKILL, which is guarenteed to be always 499This module works by "hijacking" SIGKILL, which is guaranteed to always
187available in perl, but also cannot be caught, so is always available. 500exist, but also cannot be caught, so is always available.
188 501
189Basically, this module fakes the receive of a SIGKILL signal and 502Basically, this module fakes the occurance of a SIGKILL signal and
190then catches it. This makes normal signal handling slower (probably 503then intercepts the interpreter handling it. This makes normal signal
191unmeasurably), but has the advantage of not requiring a special runops nor 504handling slower (probably unmeasurably, though), but has the advantage
192slowing down normal perl execution a bit. 505of not requiring a special runops function, nor slowing down normal perl
506execution a bit.
193 507
194It assumes that C<sig_atomic_t> and C<int> are both exception-safe to 508It assumes that C<sig_atomic_t>, C<int> and C<IV> are all async-safe to
195modify (C<sig_atomic_> is used by this module, and perl itself uses 509modify.
196C<int>, so we can assume that this is quite portbale, at least w.r.t.
197signals).
198 510
199=head1 AUTHOR 511=head1 AUTHOR
200 512
201 Marc Lehmann <schmorp@schmorp.de> 513 Marc Lehmann <schmorp@schmorp.de>
202 http://home.schmorp.de/ 514 http://home.schmorp.de/

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines