cvsroot/Async-Interrupt/Interrupt.pm

=head1 NAME

Async::Interrupt - allow C/XS libraries to interrupt perl asynchronously

=head1 SYNOPSIS

 use Async::Interrupt;

=head1 DESCRIPTION

This module implements a single feature only of interest to advanced perl
modules, namely asynchronous interruptions (think "UNIX signals", which
are very similar).

Sometimes, modules wish to run code asynchronously (in another thread,
or from a signal handler), and then signal the perl interpreter on
certain events. One common way is to write some data to a pipe and use an
event handling toolkit to watch for I/O events. Another way is to send
a signal. Those methods are slow, and in the case of a pipe, also not
asynchronous - it won't interrupt a running perl interpreter.

This module implements asynchronous notifications that enable you to
signal running perl code from another thread, asynchronously, and
sometimes even without using a single syscall.

=head2 USAGE SCENARIOS

=over 4

=item Race-free signal handling

There seems to be no way to do race-free signal handling in perl: to
catch a signal, you have to execute Perl code, and between entering the
interpreter C<select> function (or other blocking functions) and executing
the select syscall is a small but relevant timespan during which signals
will be queued, but perl signal handlers will not be executed and the
blocking syscall will not be interrupted.

You can use this module to bind a signal to a callback while at the same
time activating an event pipe that you can C<select> on, fixing the race
completely.

This can be used to implement the signal hadling in event loops,
e.g. L<AnyEvent>, L<POE>, L<IO::Async::Loop> and so on.

=item Background threads want speedy reporting

Assume you want very exact timing, and you can spare an extra cpu core
for that. Then you can run an extra thread that signals your perl
interpreter. This means you can get a very exact timing source while your
perl code is number crunching, without even using a syscall to communicate
between your threads.

For example the deliantra game server uses a variant of this technique
to interrupt background processes regularly to send map updates to game
clients.

L<IO::AIO> and L<BDB> could also use this to speed up result reporting.

=item Speedy event loop invocation

One could use this module e.g. in L<Coro> to interrupt a running coro-thread
and cause it to enter the event loop.

Or one could bind to C<SIGIO> and tell some important sockets to send this
signal, causing the event loop to be entered to reduce network latency.

=back

=head2 HOW TO USE

You can use this module by creating an C<Async::Interrupt> object for each
such event source. This object stores a perl and/or a C-level callback
that is invoked when the C<Async::Interrupt> object gets signalled. It is
executed at the next time the perl interpreter is running (i.e. it will
interrupt a computation, but not an XS function or a syscall).

You can signal the C<Async::Interrupt> object either by calling it's C<<
->signal >> method, or, more commonly, by calling a C function. There is
also the built-in (POSIX) signal source.

The C<< ->signal_func >> returns the address of the C function that is to
be called (plus an argument to be used during the call). The signalling
function also takes an integer argument in the range SIG_ATOMIC_MIN to
SIG_ATOMIC_MAX (guaranteed to allow at least 0..127).

Since this kind of interruption is fast, but can only interrupt a
I<running> interpreter, there is optional support for signalling a pipe
- that means you can also wait for the pipe to become readable (e.g. via
L<EV> or L<AnyEvent>). This, of course, incurs the overhead of a C<read>
and C<write> syscall.

=over 4

=cut

package Async::Interrupt;

use common::sense;

BEGIN {
   # the next line forces initialisation of internal
   # signal handling # variables
   $SIG{KILL} = sub { };

   our $VERSION = '0.041';

   require XSLoader;
   XSLoader::load ("Async::Interrupt", $VERSION);
}

our $DIED = sub { warn "$@" };

=item $async = new Async::Interrupt key => value...

Creates a new Async::Interrupt object. You may only use async
notifications on this object while it exists, so you need to keep a
reference to it at all times while it is used.

Optional constructor arguments include (normally you would specify at
least one of C<cb> or C<c_cb>).

=over 4

=item cb => $coderef->($value)

Registers a perl callback to be invoked whenever the async interrupt is
signalled.

Note that, since this callback can be invoked at basically any time, it
must not modify any well-known global variables such as C<$/> without
restoring them again before returning.

The exceptions are C<$!> and C<$@>, which are saved and restored by
Async::Interrupt.

If the callback should throw an exception, then it will be caught,
and C<$Async::Interrupt::DIED> will be called with C<$@> containing
the exception.  The default will simply C<warn> about the message and
continue.

=item c_cb => [$c_func, $c_arg]

Registers a C callback the be invoked whenever the async interrupt is
signalled.

The C callback must have the following prototype:

   void c_func (pTHX_ void *c_arg, int value);

Both C<$c_func> and C<$c_arg> must be specified as integers/IVs, and
C<$value> is the C<value> passed to some earlier call to either C<$signal>
or the C<signal_func> function.

Note that, because the callback can be invoked at almost any time, you
have to be careful at saving and restoring global variables that Perl
might use (the exception is C<errno>, which is saved and restored by
Async::Interrupt). The callback itself runs as part of the perl context,
so you can call any perl functions and modify any perl data structures (in
which case the requirements set out for C<cb> apply as well).

=item signal => $signame_or_value

When this parameter is specified, then the Async::Interrupt will hook the
given signal, that is, it will effectively call C<< ->signal (0) >> each time
the given signal is caught by the process.

Only one async can hook a given signal, and the signal will be restored to
defaults when the Async::Interrupt object gets destroyed.

=item pipe => [$fileno_or_fh_for_reading, $fileno_or_fh_for_writing]

Specifies two file descriptors (or file handles) that should be signalled
whenever the async interrupt is signalled. This means a single octet will
be written to it, and before the callback is being invoked, it will be
read again. Due to races, it is unlikely but possible that multiple octets
are written. It is required that the file handles are both in nonblocking
mode.

You can get a portable pipe and set non-blocking mode portably by using
e.g. L<AnyEvent::Util> from the L<AnyEvent> distribution.

It is also possible to pass in a linux eventfd as both read and write
handle (which is faster than a pipe).

The object will keep a reference to the file handles.

This can be used to ensure that async notifications will interrupt event
frameworks as well.

=back

=cut

sub new {
   my ($class, %arg) = @_;

   bless \(_alloc $arg{cb}, @{$arg{c_cb}}[0,1], @{$arg{pipe}}[0,1], $arg{signal}), $class
}

=item ($signal_func, $signal_arg) = $async->signal_func

Returns the address of a function to call asynchronously. The function has
the following prototype and needs to be passed the specified C<$c_arg>,
which is a C<void *> cast to C<IV>:

   void (*signal_func) (void *signal_arg, int value)

An example call would look like:

   signal_func (signal_arg, 0);

The function is safe to call from within signal and thread contexts, at
any time. The specified C<value> is passed to both C and Perl callback.

C<$value> must be in the valid range for a C<sig_atomic_t> (0..127 is
portable).

If the function is called while the Async::Interrupt object is already
signaled but before the callbacks are being executed, then the stored
C<value> is either the old or the new one. Due to the asynchronous
nature of the code, the C<value> can even be passed to two consecutive
invocations of the callback.

=item $async->signal ($value=0)

This signals the given async object from Perl code. Semi-obviously, this
will instantly trigger the callback invocation.

C<$value> must be in the valid range for a C<sig_atomic_t> (0..127 is
portable).

=item $async->block

=item $async->unblock

Sometimes you need a "critical section" of code that will not be
interrupted by an Async::Interrupt. This can be implemented by calling C<<
$async->block >> before the critical section, and C<< $async->unblock >>
afterwards.

Note that there must be exactly one call of C<unblock> for every previous
call to C<block> (i.e. calls can nest).

Since ensuring this in the presence of exceptions and threads is
usually more difficult than you imagine, I recommend using C<<
$async->scoped_block >> instead.

=item $async->scope_block

This call C<< $async->block >> and installs a handler that is called when
the current scope is exited (via an exception, by canceling the Coro
thread, by calling last/goto etc.).

This is the recommended (and fastest) way to implement critical sections.

=item $async->pipe_enable

=item $async->pipe_disable

Enable/disable signalling the pipe when the interrupt occurs (default is
enabled). Writing to a pipe is relatively expensive, so it can be disabled
when you know you are not waiting for it (for example, with L<EV> you
could disable the pipe in a check watcher, and enable it in a prepare
watcher).

Note that when C<fd_disable> is in effect, no attempt to read from the
pipe will be done.

=cut

1;

=back

=head1 EXAMPLE

There really should be a complete C/XS example. Bug me about it. Better
yet, create one.

=head1 IMPLEMENTATION DETAILS AND LIMITATIONS

This module works by "hijacking" SIGKILL, which is guaranteed to always
exist, but also cannot be caught, so is always available.

Basically, this module fakes the occurance of a SIGKILL signal and
then intercepts the interpreter handling it. This makes normal signal
handling slower (probably unmeasurably, though), but has the advantage
of not requiring a special runops function, nor slowing down normal perl
execution a bit.

It assumes that C<sig_atomic_t> and C<int> are both async-safe to modify
(C<sig_atomic_> is used by this module, and perl itself uses C<int>, so we
can assume that this is quite portable, at least w.r.t. signals).

=head1 AUTHOR

 Marc Lehmann <schmorp@schmorp.de>
 http://home.schmorp.de/

=cut

Revision:	1.11
Committed:	Tue Jul 14 19:29:26 2009 UTC (14 years, 11 months ago) by root
Branch:	MAIN
Changes since 1.10:	+5 -1 lines
Log Message:	* empty log message *
#	User	Rev	Content
1	root	1.1	=head1 NAME
2
3			Async::Interrupt - allow C/XS libraries to interrupt perl asynchronously
4
5			=head1 SYNOPSIS
6
7			use Async::Interrupt;
8
9			=head1 DESCRIPTION
10
11			This module implements a single feature only of interest to advanced perl
12	root	1.4	modules, namely asynchronous interruptions (think "UNIX signals", which
13	root	1.1	are very similar).
14
15	root	1.8	Sometimes, modules wish to run code asynchronously (in another thread,
16			or from a signal handler), and then signal the perl interpreter on
17			certain events. One common way is to write some data to a pipe and use an
18			event handling toolkit to watch for I/O events. Another way is to send
19			a signal. Those methods are slow, and in the case of a pipe, also not
20			asynchronous - it won't interrupt a running perl interpreter.
21	root	1.1
22			This module implements asynchronous notifications that enable you to
23	root	1.8	signal running perl code from another thread, asynchronously, and
24			sometimes even without using a single syscall.
25	root	1.1
26	root	1.8	=head2 USAGE SCENARIOS
27
28			=over 4
29
30			=item Race-free signal handling
31
32			There seems to be no way to do race-free signal handling in perl: to
33			catch a signal, you have to execute Perl code, and between entering the
34			interpreter C<select> function (or other blocking functions) and executing
35			the select syscall is a small but relevant timespan during which signals
36			will be queued, but perl signal handlers will not be executed and the
37			blocking syscall will not be interrupted.
38
39			You can use this module to bind a signal to a callback while at the same
40			time activating an event pipe that you can C<select> on, fixing the race
41			completely.
42
43			This can be used to implement the signal hadling in event loops,
44			e.g. L<AnyEvent>, L<POE>, L<IO::Async::Loop> and so on.
45
46			=item Background threads want speedy reporting
47
48			Assume you want very exact timing, and you can spare an extra cpu core
49			for that. Then you can run an extra thread that signals your perl
50			interpreter. This means you can get a very exact timing source while your
51			perl code is number crunching, without even using a syscall to communicate
52			between your threads.
53
54			For example the deliantra game server uses a variant of this technique
55			to interrupt background processes regularly to send map updates to game
56			clients.
57
58			L<IO::AIO> and L<BDB> could also use this to speed up result reporting.
59
60			=item Speedy event loop invocation
61
62			One could use this module e.g. in L<Coro> to interrupt a running coro-thread
63			and cause it to enter the event loop.
64
65			Or one could bind to C<SIGIO> and tell some important sockets to send this
66			signal, causing the event loop to be entered to reduce network latency.
67
68			=back
69
70			=head2 HOW TO USE
71
72			You can use this module by creating an C<Async::Interrupt> object for each
73			such event source. This object stores a perl and/or a C-level callback
74			that is invoked when the C<Async::Interrupt> object gets signalled. It is
75			executed at the next time the perl interpreter is running (i.e. it will
76			interrupt a computation, but not an XS function or a syscall).
77	root	1.2
78			You can signal the C<Async::Interrupt> object either by calling it's C<<
79	root	1.8	->signal >> method, or, more commonly, by calling a C function. There is
80			also the built-in (POSIX) signal source.
81	root	1.2
82			The C<< ->signal_func >> returns the address of the C function that is to
83			be called (plus an argument to be used during the call). The signalling
84			function also takes an integer argument in the range SIG_ATOMIC_MIN to
85			SIG_ATOMIC_MAX (guaranteed to allow at least 0..127).
86
87			Since this kind of interruption is fast, but can only interrupt a
88	root	1.8	I<running> interpreter, there is optional support for signalling a pipe
89			- that means you can also wait for the pipe to become readable (e.g. via
90			L<EV> or L<AnyEvent>). This, of course, incurs the overhead of a C<read>
91			and C<write> syscall.
92	root	1.2
93	root	1.1	=over 4
94
95			=cut
96
97			package Async::Interrupt;
98
99	root	1.10	use common::sense;
100	root	1.2
101	root	1.1	BEGIN {
102	root	1.11	# the next line forces initialisation of internal
103			# signal handling # variables
104			$SIG{KILL} = sub { };
105
106	root	1.10	our $VERSION = '0.041';
107	root	1.1
108			require XSLoader;
109	root	1.11	XSLoader::load ("Async::Interrupt", $VERSION);
110	root	1.1	}
111
112	root	1.2	our $DIED = sub { warn "$@" };
113
114	root	1.1	=item $async = new Async::Interrupt key => value...
115
116			Creates a new Async::Interrupt object. You may only use async
117			notifications on this object while it exists, so you need to keep a
118			reference to it at all times while it is used.
119
120			Optional constructor arguments include (normally you would specify at
121			least one of C<cb> or C<c_cb>).
122
123			=over 4
124
125			=item cb => $coderef->($value)
126
127			Registers a perl callback to be invoked whenever the async interrupt is
128			signalled.
129
130			Note that, since this callback can be invoked at basically any time, it
131	root	1.2	must not modify any well-known global variables such as C<$/> without
132			restoring them again before returning.
133
134			The exceptions are C<$!> and C<$@>, which are saved and restored by
135			Async::Interrupt.
136	root	1.1
137	root	1.2	If the callback should throw an exception, then it will be caught,
138			and C<$Async::Interrupt::DIED> will be called with C<$@> containing
139			the exception. The default will simply C<warn> about the message and
140			continue.
141
142			=item c_cb => [$c_func, $c_arg]
143	root	1.1
144			Registers a C callback the be invoked whenever the async interrupt is
145			signalled.
146
147			The C callback must have the following prototype:
148
149	root	1.2	void c_func (pTHX_ void *c_arg, int value);
150	root	1.1
151	root	1.2	Both C<$c_func> and C<$c_arg> must be specified as integers/IVs, and
152			C<$value> is the C<value> passed to some earlier call to either C<$signal>
153			or the C<signal_func> function.
154	root	1.1
155			Note that, because the callback can be invoked at almost any time, you
156			have to be careful at saving and restoring global variables that Perl
157	root	1.4	might use (the exception is C<errno>, which is saved and restored by
158	root	1.2	Async::Interrupt). The callback itself runs as part of the perl context,
159			so you can call any perl functions and modify any perl data structures (in
160	root	1.4	which case the requirements set out for C<cb> apply as well).
161	root	1.1
162	root	1.6	=item signal => $signame_or_value
163
164			When this parameter is specified, then the Async::Interrupt will hook the
165			given signal, that is, it will effectively call C<< ->signal (0) >> each time
166			the given signal is caught by the process.
167
168			Only one async can hook a given signal, and the signal will be restored to
169			defaults when the Async::Interrupt object gets destroyed.
170
171	root	1.2	=item pipe => [$fileno_or_fh_for_reading, $fileno_or_fh_for_writing]
172	root	1.1
173	root	1.2	Specifies two file descriptors (or file handles) that should be signalled
174	root	1.1	whenever the async interrupt is signalled. This means a single octet will
175			be written to it, and before the callback is being invoked, it will be
176			read again. Due to races, it is unlikely but possible that multiple octets
177	root	1.2	are written. It is required that the file handles are both in nonblocking
178			mode.
179	root	1.1
180	root	1.6	You can get a portable pipe and set non-blocking mode portably by using
181			e.g. L<AnyEvent::Util> from the L<AnyEvent> distribution.
182
183			It is also possible to pass in a linux eventfd as both read and write
184			handle (which is faster than a pipe).
185	root	1.1
186	root	1.2	The object will keep a reference to the file handles.
187	root	1.1
188			This can be used to ensure that async notifications will interrupt event
189			frameworks as well.
190
191			=back
192
193			=cut
194
195			sub new {
196			my ($class, %arg) = @_;
197
198	root	1.6	bless \(_alloc $arg{cb}, @{$arg{c_cb}}[0,1], @{$arg{pipe}}[0,1], $arg{signal}), $class
199	root	1.1	}
200
201	root	1.2	=item ($signal_func, $signal_arg) = $async->signal_func
202	root	1.1
203			Returns the address of a function to call asynchronously. The function has
204			the following prototype and needs to be passed the specified C<$c_arg>,
205			which is a C<void *> cast to C<IV>:
206
207			void (signal_func) (void signal_arg, int value)
208
209			An example call would look like:
210
211			signal_func (signal_arg, 0);
212
213	root	1.2	The function is safe to call from within signal and thread contexts, at
214	root	1.1	any time. The specified C<value> is passed to both C and Perl callback.
215
216	root	1.2	C<$value> must be in the valid range for a C<sig_atomic_t> (0..127 is
217			portable).
218
219	root	1.1	If the function is called while the Async::Interrupt object is already
220			signaled but before the callbacks are being executed, then the stored
221	root	1.2	C<value> is either the old or the new one. Due to the asynchronous
222			nature of the code, the C<value> can even be passed to two consecutive
223			invocations of the callback.
224	root	1.1
225			=item $async->signal ($value=0)
226
227			This signals the given async object from Perl code. Semi-obviously, this
228			will instantly trigger the callback invocation.
229
230	root	1.2	C<$value> must be in the valid range for a C<sig_atomic_t> (0..127 is
231			portable).
232
233			=item $async->block
234
235	root	1.3	=item $async->unblock
236
237			Sometimes you need a "critical section" of code that will not be
238			interrupted by an Async::Interrupt. This can be implemented by calling C<<
239			$async->block >> before the critical section, and C<< $async->unblock >>
240			afterwards.
241
242	root	1.4	Note that there must be exactly one call of C<unblock> for every previous
243	root	1.3	call to C<block> (i.e. calls can nest).
244
245	root	1.4	Since ensuring this in the presence of exceptions and threads is
246	root	1.3	usually more difficult than you imagine, I recommend using C<<
247			$async->scoped_block >> instead.
248	root	1.2
249	root	1.3	=item $async->scope_block
250
251			This call C<< $async->block >> and installs a handler that is called when
252			the current scope is exited (via an exception, by canceling the Coro
253			thread, by calling last/goto etc.).
254
255			This is the recommended (and fastest) way to implement critical sections.
256	root	1.2
257	root	1.6	=item $async->pipe_enable
258
259			=item $async->pipe_disable
260
261			Enable/disable signalling the pipe when the interrupt occurs (default is
262			enabled). Writing to a pipe is relatively expensive, so it can be disabled
263			when you know you are not waiting for it (for example, with L<EV> you
264			could disable the pipe in a check watcher, and enable it in a prepare
265			watcher).
266
267			Note that when C<fd_disable> is in effect, no attempt to read from the
268			pipe will be done.
269
270	root	1.1	=cut
271
272			1;
273
274			=back
275
276	root	1.2	=head1 EXAMPLE
277
278	root	1.8	There really should be a complete C/XS example. Bug me about it. Better
279			yet, create one.
280	root	1.2
281			=head1 IMPLEMENTATION DETAILS AND LIMITATIONS
282
283	root	1.8	This module works by "hijacking" SIGKILL, which is guaranteed to always
284			exist, but also cannot be caught, so is always available.
285	root	1.2
286	root	1.8	Basically, this module fakes the occurance of a SIGKILL signal and
287			then intercepts the interpreter handling it. This makes normal signal
288			handling slower (probably unmeasurably, though), but has the advantage
289			of not requiring a special runops function, nor slowing down normal perl
290			execution a bit.
291
292			It assumes that C<sig_atomic_t> and C<int> are both async-safe to modify
293			(C<sig_atomic_> is used by this module, and perl itself uses C<int>, so we
294			can assume that this is quite portable, at least w.r.t. signals).
295	root	1.2
296	root	1.1	=head1 AUTHOR
297
298			Marc Lehmann <schmorp@schmorp.de>
299			http://home.schmorp.de/
300
301			=cut
302