ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent/lib/AnyEvent/Handle.pm
Revision: 1.241
Committed: Fri Sep 5 22:17:26 2014 UTC (9 years, 9 months ago) by root
Branch: MAIN
Changes since 1.240: +33 -10 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 =head1 NAME
2
3 AnyEvent::Handle - non-blocking I/O on streaming handles via AnyEvent
4
5 =head1 SYNOPSIS
6
7 use AnyEvent;
8 use AnyEvent::Handle;
9
10 my $cv = AnyEvent->condvar;
11
12 my $hdl; $hdl = new AnyEvent::Handle
13 fh => \*STDIN,
14 on_error => sub {
15 my ($hdl, $fatal, $msg) = @_;
16 AE::log error => $msg;
17 $hdl->destroy;
18 $cv->send;
19 };
20
21 # send some request line
22 $hdl->push_write ("getinfo\015\012");
23
24 # read the response line
25 $hdl->push_read (line => sub {
26 my ($hdl, $line) = @_;
27 say "got line <$line>";
28 $cv->send;
29 });
30
31 $cv->recv;
32
33 =head1 DESCRIPTION
34
35 This is a helper module to make it easier to do event-based I/O on
36 stream-based filehandles (sockets, pipes, and other stream things).
37
38 The L<AnyEvent::Intro> tutorial contains some well-documented
39 AnyEvent::Handle examples.
40
41 In the following, where the documentation refers to "bytes", it means
42 characters. As sysread and syswrite are used for all I/O, their
43 treatment of characters applies to this module as well.
44
45 At the very minimum, you should specify C<fh> or C<connect>, and the
46 C<on_error> callback.
47
48 All callbacks will be invoked with the handle object as their first
49 argument.
50
51 =cut
52
53 package AnyEvent::Handle;
54
55 use Scalar::Util ();
56 use List::Util ();
57 use Carp ();
58 use Errno qw(EAGAIN EINTR);
59
60 use AnyEvent (); BEGIN { AnyEvent::common_sense }
61 use AnyEvent::Util qw(WSAEWOULDBLOCK);
62
63 our $VERSION = $AnyEvent::VERSION;
64
65 sub _load_func($) {
66 my $func = $_[0];
67
68 unless (defined &$func) {
69 my $pkg = $func;
70 do {
71 $pkg =~ s/::[^:]+$//
72 or return;
73 eval "require $pkg";
74 } until defined &$func;
75 }
76
77 \&$func
78 }
79
80 sub MAX_READ_SIZE() { 131072 }
81
82 =head1 METHODS
83
84 =over 4
85
86 =item $handle = B<new> AnyEvent::Handle fh => $filehandle, key => value...
87
88 The constructor supports these arguments (all as C<< key => value >> pairs).
89
90 =over 4
91
92 =item fh => $filehandle [C<fh> or C<connect> MANDATORY]
93
94 The filehandle this L<AnyEvent::Handle> object will operate on.
95 NOTE: The filehandle will be set to non-blocking mode (using
96 C<AnyEvent::Util::fh_nonblocking>) by the constructor and needs to stay in
97 that mode.
98
99 =item connect => [$host, $service] [C<fh> or C<connect> MANDATORY]
100
101 Try to connect to the specified host and service (port), using
102 C<AnyEvent::Socket::tcp_connect>. The C<$host> additionally becomes the
103 default C<peername>.
104
105 You have to specify either this parameter, or C<fh>, above.
106
107 It is possible to push requests on the read and write queues, and modify
108 properties of the stream, even while AnyEvent::Handle is connecting.
109
110 When this parameter is specified, then the C<on_prepare>,
111 C<on_connect_error> and C<on_connect> callbacks will be called under the
112 appropriate circumstances:
113
114 =over 4
115
116 =item on_prepare => $cb->($handle)
117
118 This (rarely used) callback is called before a new connection is
119 attempted, but after the file handle has been created (you can access that
120 file handle via C<< $handle->{fh} >>). It could be used to prepare the
121 file handle with parameters required for the actual connect (as opposed to
122 settings that can be changed when the connection is already established).
123
124 The return value of this callback should be the connect timeout value in
125 seconds (or C<0>, or C<undef>, or the empty list, to indicate that the
126 default timeout is to be used).
127
128 =item on_connect => $cb->($handle, $host, $port, $retry->())
129
130 This callback is called when a connection has been successfully established.
131
132 The peer's numeric host and port (the socket peername) are passed as
133 parameters, together with a retry callback. At the time it is called the
134 read and write queues, EOF status, TLS status and similar properties of
135 the handle will have been reset.
136
137 It is not allowed to use the read or write queues while the handle object
138 is connecting.
139
140 If, for some reason, the handle is not acceptable, calling C<$retry> will
141 continue with the next connection target (in case of multi-homed hosts or
142 SRV records there can be multiple connection endpoints). The C<$retry>
143 callback can be invoked after the connect callback returns, i.e. one can
144 start a handshake and then decide to retry with the next host if the
145 handshake fails.
146
147 In most cases, you should ignore the C<$retry> parameter.
148
149 =item on_connect_error => $cb->($handle, $message)
150
151 This callback is called when the connection could not be
152 established. C<$!> will contain the relevant error code, and C<$message> a
153 message describing it (usually the same as C<"$!">).
154
155 If this callback isn't specified, then C<on_error> will be called with a
156 fatal error instead.
157
158 =back
159
160 =item on_error => $cb->($handle, $fatal, $message)
161
162 This is the error callback, which is called when, well, some error
163 occured, such as not being able to resolve the hostname, failure to
164 connect, or a read error.
165
166 Some errors are fatal (which is indicated by C<$fatal> being true). On
167 fatal errors the handle object will be destroyed (by a call to C<< ->
168 destroy >>) after invoking the error callback (which means you are free to
169 examine the handle object). Examples of fatal errors are an EOF condition
170 with active (but unsatisfiable) read watchers (C<EPIPE>) or I/O errors. In
171 cases where the other side can close the connection at will, it is
172 often easiest to not report C<EPIPE> errors in this callback.
173
174 AnyEvent::Handle tries to find an appropriate error code for you to check
175 against, but in some cases (TLS errors), this does not work well.
176
177 If you report the error to the user, it is recommended to always output
178 the C<$message> argument in human-readable error messages (you don't need
179 to report C<"$!"> if you report C<$message>).
180
181 If you want to react programmatically to the error, then looking at C<$!>
182 and comparing it against some of the documented C<Errno> values is usually
183 better than looking at the C<$message>.
184
185 Non-fatal errors can be retried by returning, but it is recommended
186 to simply ignore this parameter and instead abondon the handle object
187 when this callback is invoked. Examples of non-fatal errors are timeouts
188 C<ETIMEDOUT>) or badly-formatted data (C<EBADMSG>).
189
190 On entry to the callback, the value of C<$!> contains the operating
191 system error code (or C<ENOSPC>, C<EPIPE>, C<ETIMEDOUT>, C<EBADMSG> or
192 C<EPROTO>).
193
194 While not mandatory, it is I<highly> recommended to set this callback, as
195 you will not be notified of errors otherwise. The default just calls
196 C<croak>.
197
198 =item on_read => $cb->($handle)
199
200 This sets the default read callback, which is called when data arrives
201 and no read request is in the queue (unlike read queue callbacks, this
202 callback will only be called when at least one octet of data is in the
203 read buffer).
204
205 To access (and remove data from) the read buffer, use the C<< ->rbuf >>
206 method or access the C<< $handle->{rbuf} >> member directly. Note that you
207 must not enlarge or modify the read buffer, you can only remove data at
208 the beginning from it.
209
210 You can also call C<< ->push_read (...) >> or any other function that
211 modifies the read queue. Or do both. Or ...
212
213 When an EOF condition is detected, AnyEvent::Handle will first try to
214 feed all the remaining data to the queued callbacks and C<on_read> before
215 calling the C<on_eof> callback. If no progress can be made, then a fatal
216 error will be raised (with C<$!> set to C<EPIPE>).
217
218 Note that, unlike requests in the read queue, an C<on_read> callback
219 doesn't mean you I<require> some data: if there is an EOF and there
220 are outstanding read requests then an error will be flagged. With an
221 C<on_read> callback, the C<on_eof> callback will be invoked.
222
223 =item on_eof => $cb->($handle)
224
225 Set the callback to be called when an end-of-file condition is detected,
226 i.e. in the case of a socket, when the other side has closed the
227 connection cleanly, and there are no outstanding read requests in the
228 queue (if there are read requests, then an EOF counts as an unexpected
229 connection close and will be flagged as an error).
230
231 For sockets, this just means that the other side has stopped sending data,
232 you can still try to write data, and, in fact, one can return from the EOF
233 callback and continue writing data, as only the read part has been shut
234 down.
235
236 If an EOF condition has been detected but no C<on_eof> callback has been
237 set, then a fatal error will be raised with C<$!> set to <0>.
238
239 =item on_drain => $cb->($handle)
240
241 This sets the callback that is called once when the write buffer becomes
242 empty (and immediately when the handle object is created).
243
244 To append to the write buffer, use the C<< ->push_write >> method.
245
246 This callback is useful when you don't want to put all of your write data
247 into the queue at once, for example, when you want to write the contents
248 of some file to the socket you might not want to read the whole file into
249 memory and push it into the queue, but instead only read more data from
250 the file when the write queue becomes empty.
251
252 =item timeout => $fractional_seconds
253
254 =item rtimeout => $fractional_seconds
255
256 =item wtimeout => $fractional_seconds
257
258 If non-zero, then these enables an "inactivity" timeout: whenever this
259 many seconds pass without a successful read or write on the underlying
260 file handle (or a call to C<timeout_reset>), the C<on_timeout> callback
261 will be invoked (and if that one is missing, a non-fatal C<ETIMEDOUT>
262 error will be raised).
263
264 There are three variants of the timeouts that work independently of each
265 other, for both read and write (triggered when nothing was read I<OR>
266 written), just read (triggered when nothing was read), and just write:
267 C<timeout>, C<rtimeout> and C<wtimeout>, with corresponding callbacks
268 C<on_timeout>, C<on_rtimeout> and C<on_wtimeout>, and reset functions
269 C<timeout_reset>, C<rtimeout_reset>, and C<wtimeout_reset>.
270
271 Note that timeout processing is active even when you do not have any
272 outstanding read or write requests: If you plan to keep the connection
273 idle then you should disable the timeout temporarily or ignore the
274 timeout in the corresponding C<on_timeout> callback, in which case
275 AnyEvent::Handle will simply restart the timeout.
276
277 Zero (the default) disables the corresponding timeout.
278
279 =item on_timeout => $cb->($handle)
280
281 =item on_rtimeout => $cb->($handle)
282
283 =item on_wtimeout => $cb->($handle)
284
285 Called whenever the inactivity timeout passes. If you return from this
286 callback, then the timeout will be reset as if some activity had happened,
287 so this condition is not fatal in any way.
288
289 =item rbuf_max => <bytes>
290
291 If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>)
292 when the read buffer ever (strictly) exceeds this size. This is useful to
293 avoid some forms of denial-of-service attacks.
294
295 For example, a server accepting connections from untrusted sources should
296 be configured to accept only so-and-so much data that it cannot act on
297 (for example, when expecting a line, an attacker could send an unlimited
298 amount of data without a callback ever being called as long as the line
299 isn't finished).
300
301 =item wbuf_max => <bytes>
302
303 If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>)
304 when the write buffer ever (strictly) exceeds this size. This is useful to
305 avoid some forms of denial-of-service attacks.
306
307 Although the units of this parameter is bytes, this is the I<raw> number
308 of bytes not yet accepted by the kernel. This can make a difference when
309 you e.g. use TLS, as TLS typically makes your write data larger (but it
310 can also make it smaller due to compression).
311
312 As an example of when this limit is useful, take a chat server that sends
313 chat messages to a client. If the client does not read those in a timely
314 manner then the send buffer in the server would grow unbounded.
315
316 =item autocork => <boolean>
317
318 When disabled (the default), C<push_write> will try to immediately
319 write the data to the handle if possible. This avoids having to register
320 a write watcher and wait for the next event loop iteration, but can
321 be inefficient if you write multiple small chunks (on the wire, this
322 disadvantage is usually avoided by your kernel's nagle algorithm, see
323 C<no_delay>, but this option can save costly syscalls).
324
325 When enabled, writes will always be queued till the next event loop
326 iteration. This is efficient when you do many small writes per iteration,
327 but less efficient when you do a single write only per iteration (or when
328 the write buffer often is full). It also increases write latency.
329
330 =item no_delay => <boolean>
331
332 When doing small writes on sockets, your operating system kernel might
333 wait a bit for more data before actually sending it out. This is called
334 the Nagle algorithm, and usually it is beneficial.
335
336 In some situations you want as low a delay as possible, which can be
337 accomplishd by setting this option to a true value.
338
339 The default is your operating system's default behaviour (most likely
340 enabled). This option explicitly enables or disables it, if possible.
341
342 =item keepalive => <boolean>
343
344 Enables (default disable) the SO_KEEPALIVE option on the stream socket:
345 normally, TCP connections have no time-out once established, so TCP
346 connections, once established, can stay alive forever even when the other
347 side has long gone. TCP keepalives are a cheap way to take down long-lived
348 TCP connections when the other side becomes unreachable. While the default
349 is OS-dependent, TCP keepalives usually kick in after around two hours,
350 and, if the other side doesn't reply, take down the TCP connection some 10
351 to 15 minutes later.
352
353 It is harmless to specify this option for file handles that do not support
354 keepalives, and enabling it on connections that are potentially long-lived
355 is usually a good idea.
356
357 =item oobinline => <boolean>
358
359 BSD majorly fucked up the implementation of TCP urgent data. The result
360 is that almost no OS implements TCP according to the specs, and every OS
361 implements it slightly differently.
362
363 If you want to handle TCP urgent data, then setting this flag (the default
364 is enabled) gives you the most portable way of getting urgent data, by
365 putting it into the stream.
366
367 Since BSD emulation of OOB data on top of TCP's urgent data can have
368 security implications, AnyEvent::Handle sets this flag automatically
369 unless explicitly specified. Note that setting this flag after
370 establishing a connection I<may> be a bit too late (data loss could
371 already have occured on BSD systems), but at least it will protect you
372 from most attacks.
373
374 =item read_size => <bytes>
375
376 The initial read block size, the number of bytes this module will try
377 to read during each loop iteration. Each handle object will consume
378 at least this amount of memory for the read buffer as well, so when
379 handling many connections watch out for memory requirements). See also
380 C<max_read_size>. Default: C<2048>.
381
382 =item max_read_size => <bytes>
383
384 The maximum read buffer size used by the dynamic adjustment
385 algorithm: Each time AnyEvent::Handle can read C<read_size> bytes in
386 one go it will double C<read_size> up to the maximum given by this
387 option. Default: C<131072> or C<read_size>, whichever is higher.
388
389 =item low_water_mark => <bytes>
390
391 Sets the number of bytes (default: C<0>) that make up an "empty" write
392 buffer: If the buffer reaches this size or gets even samller it is
393 considered empty.
394
395 Sometimes it can be beneficial (for performance reasons) to add data to
396 the write buffer before it is fully drained, but this is a rare case, as
397 the operating system kernel usually buffers data as well, so the default
398 is good in almost all cases.
399
400 =item linger => <seconds>
401
402 If this is non-zero (default: C<3600>), the destructor of the
403 AnyEvent::Handle object will check whether there is still outstanding
404 write data and will install a watcher that will write this data to the
405 socket. No errors will be reported (this mostly matches how the operating
406 system treats outstanding data at socket close time).
407
408 This will not work for partial TLS data that could not be encoded
409 yet. This data will be lost. Calling the C<stoptls> method in time might
410 help.
411
412 =item peername => $string
413
414 A string used to identify the remote site - usually the DNS hostname
415 (I<not> IDN!) used to create the connection, rarely the IP address.
416
417 Apart from being useful in error messages, this string is also used in TLS
418 peername verification (see C<verify_peername> in L<AnyEvent::TLS>). This
419 verification will be skipped when C<peername> is not specified or is
420 C<undef>.
421
422 =item tls => "accept" | "connect" | Net::SSLeay::SSL object
423
424 When this parameter is given, it enables TLS (SSL) mode, that means
425 AnyEvent will start a TLS handshake as soon as the connection has been
426 established and will transparently encrypt/decrypt data afterwards.
427
428 All TLS protocol errors will be signalled as C<EPROTO>, with an
429 appropriate error message.
430
431 TLS mode requires Net::SSLeay to be installed (it will be loaded
432 automatically when you try to create a TLS handle): this module doesn't
433 have a dependency on that module, so if your module requires it, you have
434 to add the dependency yourself. If Net::SSLeay cannot be loaded or is too
435 old, you get an C<EPROTO> error.
436
437 Unlike TCP, TLS has a server and client side: for the TLS server side, use
438 C<accept>, and for the TLS client side of a connection, use C<connect>
439 mode.
440
441 You can also provide your own TLS connection object, but you have
442 to make sure that you call either C<Net::SSLeay::set_connect_state>
443 or C<Net::SSLeay::set_accept_state> on it before you pass it to
444 AnyEvent::Handle. Also, this module will take ownership of this connection
445 object.
446
447 At some future point, AnyEvent::Handle might switch to another TLS
448 implementation, then the option to use your own session object will go
449 away.
450
451 B<IMPORTANT:> since Net::SSLeay "objects" are really only integers,
452 passing in the wrong integer will lead to certain crash. This most often
453 happens when one uses a stylish C<< tls => 1 >> and is surprised about the
454 segmentation fault.
455
456 Use the C<< ->starttls >> method if you need to start TLS negotiation later.
457
458 =item tls_ctx => $anyevent_tls
459
460 Use the given C<AnyEvent::TLS> object to create the new TLS connection
461 (unless a connection object was specified directly). If this
462 parameter is missing (or C<undef>), then AnyEvent::Handle will use
463 C<AnyEvent::Handle::TLS_CTX>.
464
465 Instead of an object, you can also specify a hash reference with C<< key
466 => value >> pairs. Those will be passed to L<AnyEvent::TLS> to create a
467 new TLS context object.
468
469 =item on_starttls => $cb->($handle, $success[, $error_message])
470
471 This callback will be invoked when the TLS/SSL handshake has finished. If
472 C<$success> is true, then the TLS handshake succeeded, otherwise it failed
473 (C<on_stoptls> will not be called in this case).
474
475 The session in C<< $handle->{tls} >> can still be examined in this
476 callback, even when the handshake was not successful.
477
478 TLS handshake failures will not cause C<on_error> to be invoked when this
479 callback is in effect, instead, the error message will be passed to C<on_starttls>.
480
481 Without this callback, handshake failures lead to C<on_error> being
482 called as usual.
483
484 Note that you cannot just call C<starttls> again in this callback. If you
485 need to do that, start an zero-second timer instead whose callback can
486 then call C<< ->starttls >> again.
487
488 =item on_stoptls => $cb->($handle)
489
490 When a SSLv3/TLS shutdown/close notify/EOF is detected and this callback is
491 set, then it will be invoked after freeing the TLS session. If it is not,
492 then a TLS shutdown condition will be treated like a normal EOF condition
493 on the handle.
494
495 The session in C<< $handle->{tls} >> can still be examined in this
496 callback.
497
498 This callback will only be called on TLS shutdowns, not when the
499 underlying handle signals EOF.
500
501 =item json => L<JSON>, L<JSON::PP> or L<JSON::XS> object
502
503 This is the json coder object used by the C<json> read and write types.
504
505 If you don't supply it, then AnyEvent::Handle will create and use a
506 suitable one (on demand), which will write and expect UTF-8 encoded
507 JSON texts (either using L<JSON::XS> or L<JSON>). The written texts are
508 guaranteed not to contain any newline character.
509
510 For security reasons, this encoder will likely I<not> handle numbers and
511 strings, only arrays and objects/hashes. The reason is that originally
512 JSON was self-delimited, but Dougles Crockford thought it was a splendid
513 idea to redefine JSON incompatibly, so this is no longer true.
514
515 For protocols that used back-to-back JSON texts, this might lead to
516 run-ins, where two or more JSON texts will be interpreted as one JSON
517 text.
518
519 For this reason, if the default encoder uses L<JSON::XS>, it will default
520 to not allowing anything but arrays and objects/hashes, at least for the
521 forseeable future (it will change at some point). This might or might not
522 be true for the L<JSON> module, so this might cause a security issue.
523
524 If you depend on either behaviour, you should create your own json object
525 and pass it in explicitly.
526
527 =item cbor => L<CBOR::XS> object
528
529 This is the cbor coder object used by the C<cbor> read and write types.
530
531 If you don't supply it, then AnyEvent::Handle will create and use a
532 suitable one (on demand), which will write CBOR without using extensions,
533 if possible.
534
535 Note that you are responsible to depend on the L<CBOR::XS> module if you
536 want to use this functionality, as AnyEvent does not have a dependency on
537 it itself.
538
539 =back
540
541 =cut
542
543 sub new {
544 my $class = shift;
545 my $self = bless { @_ }, $class;
546
547 if ($self->{fh}) {
548 $self->_start;
549 return unless $self->{fh}; # could be gone by now
550
551 } elsif ($self->{connect}) {
552 require AnyEvent::Socket;
553
554 $self->{peername} = $self->{connect}[0]
555 unless exists $self->{peername};
556
557 $self->{_skip_drain_rbuf} = 1;
558
559 {
560 Scalar::Util::weaken (my $self = $self);
561
562 $self->{_connect} =
563 AnyEvent::Socket::tcp_connect (
564 $self->{connect}[0],
565 $self->{connect}[1],
566 sub {
567 my ($fh, $host, $port, $retry) = @_;
568
569 delete $self->{_connect}; # no longer needed
570
571 if ($fh) {
572 $self->{fh} = $fh;
573
574 delete $self->{_skip_drain_rbuf};
575 $self->_start;
576
577 $self->{on_connect}
578 and $self->{on_connect}($self, $host, $port, sub {
579 delete @$self{qw(fh _tw _rtw _wtw _ww _rw _eof _queue rbuf _wbuf tls _tls_rbuf _tls_wbuf)};
580 $self->{_skip_drain_rbuf} = 1;
581 &$retry;
582 });
583
584 } else {
585 if ($self->{on_connect_error}) {
586 $self->{on_connect_error}($self, "$!");
587 $self->destroy if $self;
588 } else {
589 $self->_error ($!, 1);
590 }
591 }
592 },
593 sub {
594 local $self->{fh} = $_[0];
595
596 $self->{on_prepare}
597 ? $self->{on_prepare}->($self)
598 : ()
599 }
600 );
601 }
602
603 } else {
604 Carp::croak "AnyEvent::Handle: either an existing fh or the connect parameter must be specified";
605 }
606
607 $self
608 }
609
610 sub _start {
611 my ($self) = @_;
612
613 # too many clueless people try to use udp and similar sockets
614 # with AnyEvent::Handle, do them a favour.
615 my $type = getsockopt $self->{fh}, Socket::SOL_SOCKET (), Socket::SO_TYPE ();
616 Carp::croak "AnyEvent::Handle: only stream sockets supported, anything else will NOT work!"
617 if Socket::SOCK_STREAM () != (unpack "I", $type) && defined $type;
618
619 AnyEvent::Util::fh_nonblocking $self->{fh}, 1;
620
621 $self->{_activity} =
622 $self->{_ractivity} =
623 $self->{_wactivity} = AE::now;
624
625 $self->{read_size} ||= 2048;
626 $self->{max_read_size} = $self->{read_size}
627 if $self->{read_size} > ($self->{max_read_size} || MAX_READ_SIZE);
628
629 $self->timeout (delete $self->{timeout} ) if $self->{timeout};
630 $self->rtimeout (delete $self->{rtimeout} ) if $self->{rtimeout};
631 $self->wtimeout (delete $self->{wtimeout} ) if $self->{wtimeout};
632
633 $self->no_delay (delete $self->{no_delay} ) if exists $self->{no_delay} && $self->{no_delay};
634 $self->keepalive (delete $self->{keepalive}) if exists $self->{keepalive} && $self->{keepalive};
635
636 $self->oobinline (exists $self->{oobinline} ? delete $self->{oobinline} : 1);
637
638 $self->starttls (delete $self->{tls}, delete $self->{tls_ctx})
639 if $self->{tls};
640
641 $self->on_drain (delete $self->{on_drain} ) if $self->{on_drain};
642
643 $self->start_read
644 if $self->{on_read} || @{ $self->{_queue} };
645
646 $self->_drain_wbuf;
647 }
648
649 sub _error {
650 my ($self, $errno, $fatal, $message) = @_;
651
652 $! = $errno;
653 $message ||= "$!";
654
655 if ($self->{on_error}) {
656 $self->{on_error}($self, $fatal, $message);
657 $self->destroy if $fatal;
658 } elsif ($self->{fh} || $self->{connect}) {
659 $self->destroy;
660 Carp::croak "AnyEvent::Handle uncaught error: $message";
661 }
662 }
663
664 =item $fh = $handle->fh
665
666 This method returns the file handle used to create the L<AnyEvent::Handle> object.
667
668 =cut
669
670 sub fh { $_[0]{fh} }
671
672 =item $handle->on_error ($cb)
673
674 Replace the current C<on_error> callback (see the C<on_error> constructor argument).
675
676 =cut
677
678 sub on_error {
679 $_[0]{on_error} = $_[1];
680 }
681
682 =item $handle->on_eof ($cb)
683
684 Replace the current C<on_eof> callback (see the C<on_eof> constructor argument).
685
686 =cut
687
688 sub on_eof {
689 $_[0]{on_eof} = $_[1];
690 }
691
692 =item $handle->on_timeout ($cb)
693
694 =item $handle->on_rtimeout ($cb)
695
696 =item $handle->on_wtimeout ($cb)
697
698 Replace the current C<on_timeout>, C<on_rtimeout> or C<on_wtimeout>
699 callback, or disables the callback (but not the timeout) if C<$cb> =
700 C<undef>. See the C<timeout> constructor argument and method.
701
702 =cut
703
704 # see below
705
706 =item $handle->autocork ($boolean)
707
708 Enables or disables the current autocork behaviour (see C<autocork>
709 constructor argument). Changes will only take effect on the next write.
710
711 =cut
712
713 sub autocork {
714 $_[0]{autocork} = $_[1];
715 }
716
717 =item $handle->no_delay ($boolean)
718
719 Enables or disables the C<no_delay> setting (see constructor argument of
720 the same name for details).
721
722 =cut
723
724 sub no_delay {
725 $_[0]{no_delay} = $_[1];
726
727 setsockopt $_[0]{fh}, Socket::IPPROTO_TCP (), Socket::TCP_NODELAY (), int $_[1]
728 if $_[0]{fh};
729 }
730
731 =item $handle->keepalive ($boolean)
732
733 Enables or disables the C<keepalive> setting (see constructor argument of
734 the same name for details).
735
736 =cut
737
738 sub keepalive {
739 $_[0]{keepalive} = $_[1];
740
741 eval {
742 local $SIG{__DIE__};
743 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
744 if $_[0]{fh};
745 };
746 }
747
748 =item $handle->oobinline ($boolean)
749
750 Enables or disables the C<oobinline> setting (see constructor argument of
751 the same name for details).
752
753 =cut
754
755 sub oobinline {
756 $_[0]{oobinline} = $_[1];
757
758 eval {
759 local $SIG{__DIE__};
760 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_OOBINLINE (), int $_[1]
761 if $_[0]{fh};
762 };
763 }
764
765 =item $handle->keepalive ($boolean)
766
767 Enables or disables the C<keepalive> setting (see constructor argument of
768 the same name for details).
769
770 =cut
771
772 sub keepalive {
773 $_[0]{keepalive} = $_[1];
774
775 eval {
776 local $SIG{__DIE__};
777 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
778 if $_[0]{fh};
779 };
780 }
781
782 =item $handle->on_starttls ($cb)
783
784 Replace the current C<on_starttls> callback (see the C<on_starttls> constructor argument).
785
786 =cut
787
788 sub on_starttls {
789 $_[0]{on_starttls} = $_[1];
790 }
791
792 =item $handle->on_stoptls ($cb)
793
794 Replace the current C<on_stoptls> callback (see the C<on_stoptls> constructor argument).
795
796 =cut
797
798 sub on_stoptls {
799 $_[0]{on_stoptls} = $_[1];
800 }
801
802 =item $handle->rbuf_max ($max_octets)
803
804 Configures the C<rbuf_max> setting (C<undef> disables it).
805
806 =item $handle->wbuf_max ($max_octets)
807
808 Configures the C<wbuf_max> setting (C<undef> disables it).
809
810 =cut
811
812 sub rbuf_max {
813 $_[0]{rbuf_max} = $_[1];
814 }
815
816 sub wbuf_max {
817 $_[0]{wbuf_max} = $_[1];
818 }
819
820 #############################################################################
821
822 =item $handle->timeout ($seconds)
823
824 =item $handle->rtimeout ($seconds)
825
826 =item $handle->wtimeout ($seconds)
827
828 Configures (or disables) the inactivity timeout.
829
830 The timeout will be checked instantly, so this method might destroy the
831 handle before it returns.
832
833 =item $handle->timeout_reset
834
835 =item $handle->rtimeout_reset
836
837 =item $handle->wtimeout_reset
838
839 Reset the activity timeout, as if data was received or sent.
840
841 These methods are cheap to call.
842
843 =cut
844
845 for my $dir ("", "r", "w") {
846 my $timeout = "${dir}timeout";
847 my $tw = "_${dir}tw";
848 my $on_timeout = "on_${dir}timeout";
849 my $activity = "_${dir}activity";
850 my $cb;
851
852 *$on_timeout = sub {
853 $_[0]{$on_timeout} = $_[1];
854 };
855
856 *$timeout = sub {
857 my ($self, $new_value) = @_;
858
859 $new_value >= 0
860 or Carp::croak "AnyEvent::Handle->$timeout called with negative timeout ($new_value), caught";
861
862 $self->{$timeout} = $new_value;
863 delete $self->{$tw}; &$cb;
864 };
865
866 *{"${dir}timeout_reset"} = sub {
867 $_[0]{$activity} = AE::now;
868 };
869
870 # main workhorse:
871 # reset the timeout watcher, as neccessary
872 # also check for time-outs
873 $cb = sub {
874 my ($self) = @_;
875
876 if ($self->{$timeout} && $self->{fh}) {
877 my $NOW = AE::now;
878
879 # when would the timeout trigger?
880 my $after = $self->{$activity} + $self->{$timeout} - $NOW;
881
882 # now or in the past already?
883 if ($after <= 0) {
884 $self->{$activity} = $NOW;
885
886 if ($self->{$on_timeout}) {
887 $self->{$on_timeout}($self);
888 } else {
889 $self->_error (Errno::ETIMEDOUT);
890 }
891
892 # callback could have changed timeout value, optimise
893 return unless $self->{$timeout};
894
895 # calculate new after
896 $after = $self->{$timeout};
897 }
898
899 Scalar::Util::weaken $self;
900 return unless $self; # ->error could have destroyed $self
901
902 $self->{$tw} ||= AE::timer $after, 0, sub {
903 delete $self->{$tw};
904 $cb->($self);
905 };
906 } else {
907 delete $self->{$tw};
908 }
909 }
910 }
911
912 #############################################################################
913
914 =back
915
916 =head2 WRITE QUEUE
917
918 AnyEvent::Handle manages two queues per handle, one for writing and one
919 for reading.
920
921 The write queue is very simple: you can add data to its end, and
922 AnyEvent::Handle will automatically try to get rid of it for you.
923
924 When data could be written and the write buffer is shorter then the low
925 water mark, the C<on_drain> callback will be invoked once.
926
927 =over 4
928
929 =item $handle->on_drain ($cb)
930
931 Sets the C<on_drain> callback or clears it (see the description of
932 C<on_drain> in the constructor).
933
934 This method may invoke callbacks (and therefore the handle might be
935 destroyed after it returns).
936
937 =cut
938
939 sub on_drain {
940 my ($self, $cb) = @_;
941
942 $self->{on_drain} = $cb;
943
944 $cb->($self)
945 if $cb && $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf});
946 }
947
948 =item $handle->push_write ($data)
949
950 Queues the given scalar to be written. You can push as much data as
951 you want (only limited by the available memory and C<wbuf_max>), as
952 C<AnyEvent::Handle> buffers it independently of the kernel.
953
954 This method may invoke callbacks (and therefore the handle might be
955 destroyed after it returns).
956
957 =cut
958
959 sub _drain_wbuf {
960 my ($self) = @_;
961
962 if (!$self->{_ww} && length $self->{wbuf}) {
963
964 Scalar::Util::weaken $self;
965
966 my $cb = sub {
967 my $len = syswrite $self->{fh}, $self->{wbuf};
968
969 if (defined $len) {
970 substr $self->{wbuf}, 0, $len, "";
971
972 $self->{_activity} = $self->{_wactivity} = AE::now;
973
974 $self->{on_drain}($self)
975 if $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf})
976 && $self->{on_drain};
977
978 delete $self->{_ww} unless length $self->{wbuf};
979 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
980 $self->_error ($!, 1);
981 }
982 };
983
984 # try to write data immediately
985 $cb->() unless $self->{autocork};
986
987 # if still data left in wbuf, we need to poll
988 $self->{_ww} = AE::io $self->{fh}, 1, $cb
989 if length $self->{wbuf};
990
991 if (
992 defined $self->{wbuf_max}
993 && $self->{wbuf_max} < length $self->{wbuf}
994 ) {
995 $self->_error (Errno::ENOSPC, 1), return;
996 }
997 };
998 }
999
1000 our %WH;
1001
1002 # deprecated
1003 sub register_write_type($$) {
1004 $WH{$_[0]} = $_[1];
1005 }
1006
1007 sub push_write {
1008 my $self = shift;
1009
1010 if (@_ > 1) {
1011 my $type = shift;
1012
1013 @_ = ($WH{$type} ||= _load_func "$type\::anyevent_write_type"
1014 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_write")
1015 ->($self, @_);
1016 }
1017
1018 # we downgrade here to avoid hard-to-track-down bugs,
1019 # and diagnose the problem earlier and better.
1020
1021 if ($self->{tls}) {
1022 utf8::downgrade $self->{_tls_wbuf} .= $_[0];
1023 &_dotls ($self) if $self->{fh};
1024 } else {
1025 utf8::downgrade $self->{wbuf} .= $_[0];
1026 $self->_drain_wbuf if $self->{fh};
1027 }
1028 }
1029
1030 =item $handle->push_write (type => @args)
1031
1032 Instead of formatting your data yourself, you can also let this module
1033 do the job by specifying a type and type-specific arguments. You
1034 can also specify the (fully qualified) name of a package, in which
1035 case AnyEvent tries to load the package and then expects to find the
1036 C<anyevent_write_type> function inside (see "custom write types", below).
1037
1038 Predefined types are (if you have ideas for additional types, feel free to
1039 drop by and tell us):
1040
1041 =over 4
1042
1043 =item netstring => $string
1044
1045 Formats the given value as netstring
1046 (http://cr.yp.to/proto/netstrings.txt, this is not a recommendation to use them).
1047
1048 =cut
1049
1050 register_write_type netstring => sub {
1051 my ($self, $string) = @_;
1052
1053 (length $string) . ":$string,"
1054 };
1055
1056 =item packstring => $format, $data
1057
1058 An octet string prefixed with an encoded length. The encoding C<$format>
1059 uses the same format as a Perl C<pack> format, but must specify a single
1060 integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
1061 optional C<!>, C<< < >> or C<< > >> modifier).
1062
1063 =cut
1064
1065 register_write_type packstring => sub {
1066 my ($self, $format, $string) = @_;
1067
1068 pack "$format/a*", $string
1069 };
1070
1071 =item json => $array_or_hashref
1072
1073 Encodes the given hash or array reference into a JSON object. Unless you
1074 provide your own JSON object, this means it will be encoded to JSON text
1075 in UTF-8.
1076
1077 The default encoder might or might not handle every type of JSON value -
1078 it might be limited to arrays and objects for security reasons. See the
1079 C<json> constructor attribute for more details.
1080
1081 JSON objects (and arrays) are self-delimiting, so if you only use arrays
1082 and hashes, you can write JSON at one end of a handle and read them at the
1083 other end without using any additional framing.
1084
1085 The JSON text generated by the default encoder is guaranteed not to
1086 contain any newlines: While this module doesn't need delimiters after or
1087 between JSON texts to be able to read them, many other languages depend on
1088 them.
1089
1090 A simple RPC protocol that interoperates easily with other languages is
1091 to send JSON arrays (or objects, although arrays are usually the better
1092 choice as they mimic how function argument passing works) and a newline
1093 after each JSON text:
1094
1095 $handle->push_write (json => ["method", "arg1", "arg2"]); # whatever
1096 $handle->push_write ("\012");
1097
1098 An AnyEvent::Handle receiver would simply use the C<json> read type and
1099 rely on the fact that the newline will be skipped as leading whitespace:
1100
1101 $handle->push_read (json => sub { my $array = $_[1]; ... });
1102
1103 Other languages could read single lines terminated by a newline and pass
1104 this line into their JSON decoder of choice.
1105
1106 =item cbor => $perl_scalar
1107
1108 Encodes the given scalar into a CBOR value. Unless you provide your own
1109 L<CBOR::XS> object, this means it will be encoded to a CBOR string not
1110 using any extensions, if possible.
1111
1112 CBOR values are self-delimiting, so you can write CBOR at one end of
1113 a handle and read them at the other end without using any additional
1114 framing.
1115
1116 A simple nd very very fast RPC protocol that interoperates with
1117 other languages is to send CBOR and receive CBOR values (arrays are
1118 recommended):
1119
1120 $handle->push_write (cbor => ["method", "arg1", "arg2"]); # whatever
1121
1122 An AnyEvent::Handle receiver would simply use the C<cbor> read type:
1123
1124 $handle->push_read (cbor => sub { my $array = $_[1]; ... });
1125
1126 =cut
1127
1128 sub json_coder() {
1129 eval { require JSON::XS; JSON::XS->new->utf8 }
1130 || do { require JSON::PP; JSON::PP->new->utf8 }
1131 }
1132
1133 register_write_type json => sub {
1134 my ($self, $ref) = @_;
1135
1136 ($self->{json} ||= json_coder)
1137 ->encode ($ref)
1138 };
1139
1140 sub cbor_coder() {
1141 require CBOR::XS;
1142 CBOR::XS->new
1143 }
1144
1145 register_write_type cbor => sub {
1146 my ($self, $scalar) = @_;
1147
1148 ($self->{cbor} ||= cbor_coder)
1149 ->encode ($scalar)
1150 };
1151
1152 =item storable => $reference
1153
1154 Freezes the given reference using L<Storable> and writes it to the
1155 handle. Uses the C<nfreeze> format.
1156
1157 =cut
1158
1159 register_write_type storable => sub {
1160 my ($self, $ref) = @_;
1161
1162 require Storable unless $Storable::VERSION;
1163
1164 pack "w/a*", Storable::nfreeze ($ref)
1165 };
1166
1167 =back
1168
1169 =item $handle->push_shutdown
1170
1171 Sometimes you know you want to close the socket after writing your data
1172 before it was actually written. One way to do that is to replace your
1173 C<on_drain> handler by a callback that shuts down the socket (and set
1174 C<low_water_mark> to C<0>). This method is a shorthand for just that, and
1175 replaces the C<on_drain> callback with:
1176
1177 sub { shutdown $_[0]{fh}, 1 }
1178
1179 This simply shuts down the write side and signals an EOF condition to the
1180 the peer.
1181
1182 You can rely on the normal read queue and C<on_eof> handling
1183 afterwards. This is the cleanest way to close a connection.
1184
1185 This method may invoke callbacks (and therefore the handle might be
1186 destroyed after it returns).
1187
1188 =cut
1189
1190 sub push_shutdown {
1191 my ($self) = @_;
1192
1193 delete $self->{low_water_mark};
1194 $self->on_drain (sub { shutdown $_[0]{fh}, 1 });
1195 }
1196
1197 =item custom write types - Package::anyevent_write_type $handle, @args
1198
1199 Instead of one of the predefined types, you can also specify the name of
1200 a package. AnyEvent will try to load the package and then expects to find
1201 a function named C<anyevent_write_type> inside. If it isn't found, it
1202 progressively tries to load the parent package until it either finds the
1203 function (good) or runs out of packages (bad).
1204
1205 Whenever the given C<type> is used, C<push_write> will the function with
1206 the handle object and the remaining arguments.
1207
1208 The function is supposed to return a single octet string that will be
1209 appended to the write buffer, so you can mentally treat this function as a
1210 "arguments to on-the-wire-format" converter.
1211
1212 Example: implement a custom write type C<join> that joins the remaining
1213 arguments using the first one.
1214
1215 $handle->push_write (My::Type => " ", 1,2,3);
1216
1217 # uses the following package, which can be defined in the "My::Type" or in
1218 # the "My" modules to be auto-loaded, or just about anywhere when the
1219 # My::Type::anyevent_write_type is defined before invoking it.
1220
1221 package My::Type;
1222
1223 sub anyevent_write_type {
1224 my ($handle, $delim, @args) = @_;
1225
1226 join $delim, @args
1227 }
1228
1229 =cut
1230
1231 #############################################################################
1232
1233 =back
1234
1235 =head2 READ QUEUE
1236
1237 AnyEvent::Handle manages two queues per handle, one for writing and one
1238 for reading.
1239
1240 The read queue is more complex than the write queue. It can be used in two
1241 ways, the "simple" way, using only C<on_read> and the "complex" way, using
1242 a queue.
1243
1244 In the simple case, you just install an C<on_read> callback and whenever
1245 new data arrives, it will be called. You can then remove some data (if
1246 enough is there) from the read buffer (C<< $handle->rbuf >>). Or you can
1247 leave the data there if you want to accumulate more (e.g. when only a
1248 partial message has been received so far), or change the read queue with
1249 e.g. C<push_read>.
1250
1251 In the more complex case, you want to queue multiple callbacks. In this
1252 case, AnyEvent::Handle will call the first queued callback each time new
1253 data arrives (also the first time it is queued) and remove it when it has
1254 done its job (see C<push_read>, below).
1255
1256 This way you can, for example, push three line-reads, followed by reading
1257 a chunk of data, and AnyEvent::Handle will execute them in order.
1258
1259 Example 1: EPP protocol parser. EPP sends 4 byte length info, followed by
1260 the specified number of bytes which give an XML datagram.
1261
1262 # in the default state, expect some header bytes
1263 $handle->on_read (sub {
1264 # some data is here, now queue the length-header-read (4 octets)
1265 shift->unshift_read (chunk => 4, sub {
1266 # header arrived, decode
1267 my $len = unpack "N", $_[1];
1268
1269 # now read the payload
1270 shift->unshift_read (chunk => $len, sub {
1271 my $xml = $_[1];
1272 # handle xml
1273 });
1274 });
1275 });
1276
1277 Example 2: Implement a client for a protocol that replies either with "OK"
1278 and another line or "ERROR" for the first request that is sent, and 64
1279 bytes for the second request. Due to the availability of a queue, we can
1280 just pipeline sending both requests and manipulate the queue as necessary
1281 in the callbacks.
1282
1283 When the first callback is called and sees an "OK" response, it will
1284 C<unshift> another line-read. This line-read will be queued I<before> the
1285 64-byte chunk callback.
1286
1287 # request one, returns either "OK + extra line" or "ERROR"
1288 $handle->push_write ("request 1\015\012");
1289
1290 # we expect "ERROR" or "OK" as response, so push a line read
1291 $handle->push_read (line => sub {
1292 # if we got an "OK", we have to _prepend_ another line,
1293 # so it will be read before the second request reads its 64 bytes
1294 # which are already in the queue when this callback is called
1295 # we don't do this in case we got an error
1296 if ($_[1] eq "OK") {
1297 $_[0]->unshift_read (line => sub {
1298 my $response = $_[1];
1299 ...
1300 });
1301 }
1302 });
1303
1304 # request two, simply returns 64 octets
1305 $handle->push_write ("request 2\015\012");
1306
1307 # simply read 64 bytes, always
1308 $handle->push_read (chunk => 64, sub {
1309 my $response = $_[1];
1310 ...
1311 });
1312
1313 =over 4
1314
1315 =cut
1316
1317 sub _drain_rbuf {
1318 my ($self) = @_;
1319
1320 # avoid recursion
1321 return if $self->{_skip_drain_rbuf};
1322 local $self->{_skip_drain_rbuf} = 1;
1323
1324 while () {
1325 # we need to use a separate tls read buffer, as we must not receive data while
1326 # we are draining the buffer, and this can only happen with TLS.
1327 $self->{rbuf} .= delete $self->{_tls_rbuf}
1328 if exists $self->{_tls_rbuf};
1329
1330 my $len = length $self->{rbuf};
1331
1332 if (my $cb = shift @{ $self->{_queue} }) {
1333 unless ($cb->($self)) {
1334 # no progress can be made
1335 # (not enough data and no data forthcoming)
1336 $self->_error (Errno::EPIPE, 1), return
1337 if $self->{_eof};
1338
1339 unshift @{ $self->{_queue} }, $cb;
1340 last;
1341 }
1342 } elsif ($self->{on_read}) {
1343 last unless $len;
1344
1345 $self->{on_read}($self);
1346
1347 if (
1348 $len == length $self->{rbuf} # if no data has been consumed
1349 && !@{ $self->{_queue} } # and the queue is still empty
1350 && $self->{on_read} # but we still have on_read
1351 ) {
1352 # no further data will arrive
1353 # so no progress can be made
1354 $self->_error (Errno::EPIPE, 1), return
1355 if $self->{_eof};
1356
1357 last; # more data might arrive
1358 }
1359 } else {
1360 # read side becomes idle
1361 delete $self->{_rw} unless $self->{tls};
1362 last;
1363 }
1364 }
1365
1366 if ($self->{_eof}) {
1367 $self->{on_eof}
1368 ? $self->{on_eof}($self)
1369 : $self->_error (0, 1, "Unexpected end-of-file");
1370
1371 return;
1372 }
1373
1374 if (
1375 defined $self->{rbuf_max}
1376 && $self->{rbuf_max} < length $self->{rbuf}
1377 ) {
1378 $self->_error (Errno::ENOSPC, 1), return;
1379 }
1380
1381 # may need to restart read watcher
1382 unless ($self->{_rw}) {
1383 $self->start_read
1384 if $self->{on_read} || @{ $self->{_queue} };
1385 }
1386 }
1387
1388 =item $handle->on_read ($cb)
1389
1390 This replaces the currently set C<on_read> callback, or clears it (when
1391 the new callback is C<undef>). See the description of C<on_read> in the
1392 constructor.
1393
1394 This method may invoke callbacks (and therefore the handle might be
1395 destroyed after it returns).
1396
1397 =cut
1398
1399 sub on_read {
1400 my ($self, $cb) = @_;
1401
1402 $self->{on_read} = $cb;
1403 $self->_drain_rbuf if $cb;
1404 }
1405
1406 =item $handle->rbuf
1407
1408 Returns the read buffer (as a modifiable lvalue). You can also access the
1409 read buffer directly as the C<< ->{rbuf} >> member, if you want (this is
1410 much faster, and no less clean).
1411
1412 The only operation allowed on the read buffer (apart from looking at it)
1413 is removing data from its beginning. Otherwise modifying or appending to
1414 it is not allowed and will lead to hard-to-track-down bugs.
1415
1416 NOTE: The read buffer should only be used or modified in the C<on_read>
1417 callback or when C<push_read> or C<unshift_read> are used with a single
1418 callback (i.e. untyped). Typed C<push_read> and C<unshift_read> methods
1419 will manage the read buffer on their own.
1420
1421 =cut
1422
1423 sub rbuf : lvalue {
1424 $_[0]{rbuf}
1425 }
1426
1427 =item $handle->push_read ($cb)
1428
1429 =item $handle->unshift_read ($cb)
1430
1431 Append the given callback to the end of the queue (C<push_read>) or
1432 prepend it (C<unshift_read>).
1433
1434 The callback is called each time some additional read data arrives.
1435
1436 It must check whether enough data is in the read buffer already.
1437
1438 If not enough data is available, it must return the empty list or a false
1439 value, in which case it will be called repeatedly until enough data is
1440 available (or an error condition is detected).
1441
1442 If enough data was available, then the callback must remove all data it is
1443 interested in (which can be none at all) and return a true value. After returning
1444 true, it will be removed from the queue.
1445
1446 These methods may invoke callbacks (and therefore the handle might be
1447 destroyed after it returns).
1448
1449 =cut
1450
1451 our %RH;
1452
1453 sub register_read_type($$) {
1454 $RH{$_[0]} = $_[1];
1455 }
1456
1457 sub push_read {
1458 my $self = shift;
1459 my $cb = pop;
1460
1461 if (@_) {
1462 my $type = shift;
1463
1464 $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
1465 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_read")
1466 ->($self, $cb, @_);
1467 }
1468
1469 push @{ $self->{_queue} }, $cb;
1470 $self->_drain_rbuf;
1471 }
1472
1473 sub unshift_read {
1474 my $self = shift;
1475 my $cb = pop;
1476
1477 if (@_) {
1478 my $type = shift;
1479
1480 $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
1481 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::unshift_read")
1482 ->($self, $cb, @_);
1483 }
1484
1485 unshift @{ $self->{_queue} }, $cb;
1486 $self->_drain_rbuf;
1487 }
1488
1489 =item $handle->push_read (type => @args, $cb)
1490
1491 =item $handle->unshift_read (type => @args, $cb)
1492
1493 Instead of providing a callback that parses the data itself you can chose
1494 between a number of predefined parsing formats, for chunks of data, lines
1495 etc. You can also specify the (fully qualified) name of a package, in
1496 which case AnyEvent tries to load the package and then expects to find the
1497 C<anyevent_read_type> function inside (see "custom read types", below).
1498
1499 Predefined types are (if you have ideas for additional types, feel free to
1500 drop by and tell us):
1501
1502 =over 4
1503
1504 =item chunk => $octets, $cb->($handle, $data)
1505
1506 Invoke the callback only once C<$octets> bytes have been read. Pass the
1507 data read to the callback. The callback will never be called with less
1508 data.
1509
1510 Example: read 2 bytes.
1511
1512 $handle->push_read (chunk => 2, sub {
1513 say "yay " . unpack "H*", $_[1];
1514 });
1515
1516 =cut
1517
1518 register_read_type chunk => sub {
1519 my ($self, $cb, $len) = @_;
1520
1521 sub {
1522 $len <= length $_[0]{rbuf} or return;
1523 $cb->($_[0], substr $_[0]{rbuf}, 0, $len, "");
1524 1
1525 }
1526 };
1527
1528 =item line => [$eol, ]$cb->($handle, $line, $eol)
1529
1530 The callback will be called only once a full line (including the end of
1531 line marker, C<$eol>) has been read. This line (excluding the end of line
1532 marker) will be passed to the callback as second argument (C<$line>), and
1533 the end of line marker as the third argument (C<$eol>).
1534
1535 The end of line marker, C<$eol>, can be either a string, in which case it
1536 will be interpreted as a fixed record end marker, or it can be a regex
1537 object (e.g. created by C<qr>), in which case it is interpreted as a
1538 regular expression.
1539
1540 The end of line marker argument C<$eol> is optional, if it is missing (NOT
1541 undef), then C<qr|\015?\012|> is used (which is good for most internet
1542 protocols).
1543
1544 Partial lines at the end of the stream will never be returned, as they are
1545 not marked by the end of line marker.
1546
1547 =cut
1548
1549 register_read_type line => sub {
1550 my ($self, $cb, $eol) = @_;
1551
1552 if (@_ < 3) {
1553 # this is faster then the generic code below
1554 sub {
1555 (my $pos = index $_[0]{rbuf}, "\012") >= 0
1556 or return;
1557
1558 (my $str = substr $_[0]{rbuf}, 0, $pos + 1, "") =~ s/(\015?\012)\Z// or die;
1559 $cb->($_[0], $str, "$1");
1560 1
1561 }
1562 } else {
1563 $eol = quotemeta $eol unless ref $eol;
1564 $eol = qr|^(.*?)($eol)|s;
1565
1566 sub {
1567 $_[0]{rbuf} =~ s/$eol// or return;
1568
1569 $cb->($_[0], "$1", "$2");
1570 1
1571 }
1572 }
1573 };
1574
1575 =item regex => $accept[, $reject[, $skip], $cb->($handle, $data)
1576
1577 Makes a regex match against the regex object C<$accept> and returns
1578 everything up to and including the match.
1579
1580 Example: read a single line terminated by '\n'.
1581
1582 $handle->push_read (regex => qr<\n>, sub { ... });
1583
1584 If C<$reject> is given and not undef, then it determines when the data is
1585 to be rejected: it is matched against the data when the C<$accept> regex
1586 does not match and generates an C<EBADMSG> error when it matches. This is
1587 useful to quickly reject wrong data (to avoid waiting for a timeout or a
1588 receive buffer overflow).
1589
1590 Example: expect a single decimal number followed by whitespace, reject
1591 anything else (not the use of an anchor).
1592
1593 $handle->push_read (regex => qr<^[0-9]+\s>, qr<[^0-9]>, sub { ... });
1594
1595 If C<$skip> is given and not C<undef>, then it will be matched against
1596 the receive buffer when neither C<$accept> nor C<$reject> match,
1597 and everything preceding and including the match will be accepted
1598 unconditionally. This is useful to skip large amounts of data that you
1599 know cannot be matched, so that the C<$accept> or C<$reject> regex do not
1600 have to start matching from the beginning. This is purely an optimisation
1601 and is usually worth it only when you expect more than a few kilobytes.
1602
1603 Example: expect a http header, which ends at C<\015\012\015\012>. Since we
1604 expect the header to be very large (it isn't in practice, but...), we use
1605 a skip regex to skip initial portions. The skip regex is tricky in that
1606 it only accepts something not ending in either \015 or \012, as these are
1607 required for the accept regex.
1608
1609 $handle->push_read (regex =>
1610 qr<\015\012\015\012>,
1611 undef, # no reject
1612 qr<^.*[^\015\012]>,
1613 sub { ... });
1614
1615 =cut
1616
1617 register_read_type regex => sub {
1618 my ($self, $cb, $accept, $reject, $skip) = @_;
1619
1620 my $data;
1621 my $rbuf = \$self->{rbuf};
1622
1623 sub {
1624 # accept
1625 if ($$rbuf =~ $accept) {
1626 $data .= substr $$rbuf, 0, $+[0], "";
1627 $cb->($_[0], $data);
1628 return 1;
1629 }
1630
1631 # reject
1632 if ($reject && $$rbuf =~ $reject) {
1633 $_[0]->_error (Errno::EBADMSG);
1634 }
1635
1636 # skip
1637 if ($skip && $$rbuf =~ $skip) {
1638 $data .= substr $$rbuf, 0, $+[0], "";
1639 }
1640
1641 ()
1642 }
1643 };
1644
1645 =item netstring => $cb->($handle, $string)
1646
1647 A netstring (http://cr.yp.to/proto/netstrings.txt, this is not an endorsement).
1648
1649 Throws an error with C<$!> set to EBADMSG on format violations.
1650
1651 =cut
1652
1653 register_read_type netstring => sub {
1654 my ($self, $cb) = @_;
1655
1656 sub {
1657 unless ($_[0]{rbuf} =~ s/^(0|[1-9][0-9]*)://) {
1658 if ($_[0]{rbuf} =~ /[^0-9]/) {
1659 $_[0]->_error (Errno::EBADMSG);
1660 }
1661 return;
1662 }
1663
1664 my $len = $1;
1665
1666 $_[0]->unshift_read (chunk => $len, sub {
1667 my $string = $_[1];
1668 $_[0]->unshift_read (chunk => 1, sub {
1669 if ($_[1] eq ",") {
1670 $cb->($_[0], $string);
1671 } else {
1672 $_[0]->_error (Errno::EBADMSG);
1673 }
1674 });
1675 });
1676
1677 1
1678 }
1679 };
1680
1681 =item packstring => $format, $cb->($handle, $string)
1682
1683 An octet string prefixed with an encoded length. The encoding C<$format>
1684 uses the same format as a Perl C<pack> format, but must specify a single
1685 integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
1686 optional C<!>, C<< < >> or C<< > >> modifier).
1687
1688 For example, DNS over TCP uses a prefix of C<n> (2 octet network order),
1689 EPP uses a prefix of C<N> (4 octtes).
1690
1691 Example: read a block of data prefixed by its length in BER-encoded
1692 format (very efficient).
1693
1694 $handle->push_read (packstring => "w", sub {
1695 my ($handle, $data) = @_;
1696 });
1697
1698 =cut
1699
1700 register_read_type packstring => sub {
1701 my ($self, $cb, $format) = @_;
1702
1703 sub {
1704 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1705 defined (my $len = eval { unpack $format, $_[0]{rbuf} })
1706 or return;
1707
1708 $format = length pack $format, $len;
1709
1710 # bypass unshift if we already have the remaining chunk
1711 if ($format + $len <= length $_[0]{rbuf}) {
1712 my $data = substr $_[0]{rbuf}, $format, $len;
1713 substr $_[0]{rbuf}, 0, $format + $len, "";
1714 $cb->($_[0], $data);
1715 } else {
1716 # remove prefix
1717 substr $_[0]{rbuf}, 0, $format, "";
1718
1719 # read remaining chunk
1720 $_[0]->unshift_read (chunk => $len, $cb);
1721 }
1722
1723 1
1724 }
1725 };
1726
1727 =item json => $cb->($handle, $hash_or_arrayref)
1728
1729 Reads a JSON object or array, decodes it and passes it to the
1730 callback. When a parse error occurs, an C<EBADMSG> error will be raised.
1731
1732 If a C<json> object was passed to the constructor, then that will be
1733 used for the final decode, otherwise it will create a L<JSON::XS> or
1734 L<JSON::PP> coder object expecting UTF-8.
1735
1736 This read type uses the incremental parser available with JSON version
1737 2.09 (and JSON::XS version 2.2) and above.
1738
1739 Since JSON texts are fully self-delimiting, the C<json> read and write
1740 types are an ideal simple RPC protocol: just exchange JSON datagrams. See
1741 the C<json> write type description, above, for an actual example.
1742
1743 =cut
1744
1745 register_read_type json => sub {
1746 my ($self, $cb) = @_;
1747
1748 my $json = $self->{json} ||= json_coder;
1749
1750 my $data;
1751
1752 sub {
1753 my $ref = eval { $json->incr_parse ($_[0]{rbuf}) };
1754
1755 if ($ref) {
1756 $_[0]{rbuf} = $json->incr_text;
1757 $json->incr_text = "";
1758 $cb->($_[0], $ref);
1759
1760 1
1761 } elsif ($@) {
1762 # error case
1763 $json->incr_skip;
1764
1765 $_[0]{rbuf} = $json->incr_text;
1766 $json->incr_text = "";
1767
1768 $_[0]->_error (Errno::EBADMSG);
1769
1770 ()
1771 } else {
1772 $_[0]{rbuf} = "";
1773
1774 ()
1775 }
1776 }
1777 };
1778
1779 =item cbor => $cb->($handle, $scalar)
1780
1781 Reads a CBOR value, decodes it and passes it to the callback. When a parse
1782 error occurs, an C<EBADMSG> error will be raised.
1783
1784 If a L<CBOR::XS> object was passed to the constructor, then that will be
1785 used for the final decode, otherwise it will create a CBOR coder without
1786 enabling any options.
1787
1788 You have to provide a dependency to L<CBOR::XS> on your own: this module
1789 will load the L<CBOR::XS> module, but AnyEvent does not depend on it
1790 itself.
1791
1792 Since CBOR values are fully self-delimiting, the C<cbor> read and write
1793 types are an ideal simple RPC protocol: just exchange CBOR datagrams. See
1794 the C<cbor> write type description, above, for an actual example.
1795
1796 =cut
1797
1798 register_read_type cbor => sub {
1799 my ($self, $cb) = @_;
1800
1801 my $cbor = $self->{cbor} ||= cbor_coder;
1802
1803 my $data;
1804
1805 sub {
1806 my (@value) = eval { $cbor->incr_parse ($_[0]{rbuf}) };
1807
1808 if (@value) {
1809 $cb->($_[0], @value);
1810
1811 1
1812 } elsif ($@) {
1813 # error case
1814 $cbor->incr_reset;
1815
1816 $_[0]->_error (Errno::EBADMSG);
1817
1818 ()
1819 } else {
1820 ()
1821 }
1822 }
1823 };
1824
1825 =item storable => $cb->($handle, $ref)
1826
1827 Deserialises a L<Storable> frozen representation as written by the
1828 C<storable> write type (BER-encoded length prefix followed by nfreeze'd
1829 data).
1830
1831 Raises C<EBADMSG> error if the data could not be decoded.
1832
1833 =cut
1834
1835 register_read_type storable => sub {
1836 my ($self, $cb) = @_;
1837
1838 require Storable unless $Storable::VERSION;
1839
1840 sub {
1841 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1842 defined (my $len = eval { unpack "w", $_[0]{rbuf} })
1843 or return;
1844
1845 my $format = length pack "w", $len;
1846
1847 # bypass unshift if we already have the remaining chunk
1848 if ($format + $len <= length $_[0]{rbuf}) {
1849 my $data = substr $_[0]{rbuf}, $format, $len;
1850 substr $_[0]{rbuf}, 0, $format + $len, "";
1851
1852 eval { $cb->($_[0], Storable::thaw ($data)); 1 }
1853 or return $_[0]->_error (Errno::EBADMSG);
1854 } else {
1855 # remove prefix
1856 substr $_[0]{rbuf}, 0, $format, "";
1857
1858 # read remaining chunk
1859 $_[0]->unshift_read (chunk => $len, sub {
1860 eval { $cb->($_[0], Storable::thaw ($_[1])); 1 }
1861 or $_[0]->_error (Errno::EBADMSG);
1862 });
1863 }
1864
1865 1
1866 }
1867 };
1868
1869 =item tls_detect => $cb->($handle, $detect, $major, $minor)
1870
1871 Checks the input stream for a valid SSL or TLS handshake TLSPaintext
1872 record without consuming anything. Only SSL version 3 or higher
1873 is handled, up to the fictituous protocol 4.x (but both SSL3+ and
1874 SSL2-compatible framing is supported).
1875
1876 If it detects that the input data is likely TLS, it calls the callback
1877 with a true value for C<$detect> and the (on-wire) TLS version as second
1878 and third argument (C<$major> is C<3>, and C<$minor> is 0..3 for SSL
1879 3.0, TLS 1.0, 1.1 and 1.2, respectively). If it detects the input to
1880 be definitely not TLS, it calls the callback with a false value for
1881 C<$detect>.
1882
1883 The callback could use this information to decide whether or not to start
1884 TLS negotiation.
1885
1886 In all cases the data read so far is passed to the following read
1887 handlers.
1888
1889 Usually you want to use the C<tls_autostart> read type instead.
1890
1891 If you want to design a protocol that works in the presence of TLS
1892 dtection, make sure that any non-TLS data doesn't start with the octet 22
1893 (ASCII SYN, 16 hex) or 128-255 (i.e. highest bit set). The checks this
1894 read type does are a bit more strict, but might losen in the future to
1895 accomodate protocol changes.
1896
1897 This read type does not rely on L<AnyEvent::TLS> (and thus, not on
1898 L<Net::SSLeay>).
1899
1900 =item tls_autostart => $tls[, $tls_ctx]
1901
1902 Tries to detect a valid SSL or TLS handshake. If one is detected, it tries
1903 to start tls by calling C<starttls> with the given arguments.
1904
1905 In practise, C<$tls> must be C<accept>, or a Net::SSLeay context that has
1906 been configured to accept, as servers do not normally send a handshake on
1907 their own and ths cannot be detected in this way.
1908
1909 See C<tls_detect> above for more details.
1910
1911 Example: give the client a chance to start TLS before accepting a text
1912 line.
1913
1914 $hdl->push_read (tls_detect => "accept");
1915 $hdl->push_read (line => sub {
1916 print "received ", ($_[0]{tls} ? "encrypted" : "cleartext"), " <$_[1]>\n";
1917 });
1918
1919 =cut
1920
1921 register_read_type tls_detect => sub {
1922 my ($self, $cb) = @_;
1923
1924 sub {
1925 # this regex matches a full or partial tls record
1926 if (
1927 # ssl3+: type(22=handshake) major(=3) minor(any) length_hi
1928 $self->{rbuf} =~ /^(?:\z| \x16 (\z| [\x03\x04] (?:\z| . (?:\z| [\x00-\x40] ))))/xs
1929 # ssl2 comapatible: len_hi len_lo type(1) major minor dummy(forlength)
1930 or $self->{rbuf} =~ /^(?:\z| [\x80-\xff] (?:\z| . (?:\z| \x01 (\z| [\x03\x04] (?:\z| . (?:\z| . ))))))/xs
1931 ) {
1932 return if 3 != length $1; # partial match, can't decide yet
1933
1934 # full match, valid TLS record
1935 my ($major, $minor) = unpack "CC", $1;
1936 $cb->($self, "accept", $major + $minor * 0.1);
1937 } else {
1938 # mismatch == guaranteed not TLS
1939 $cb->($self, undef);
1940 }
1941
1942 1
1943 }
1944 };
1945
1946 register_read_type tls_autostart => sub {
1947 my ($self, @tls) = @_;
1948
1949 $RH{tls_detect}($self, sub {
1950 return unless $_[1];
1951 $_[0]->starttls (@tls);
1952 })
1953 };
1954
1955 =back
1956
1957 =item custom read types - Package::anyevent_read_type $handle, $cb, @args
1958
1959 Instead of one of the predefined types, you can also specify the name
1960 of a package. AnyEvent will try to load the package and then expects to
1961 find a function named C<anyevent_read_type> inside. If it isn't found, it
1962 progressively tries to load the parent package until it either finds the
1963 function (good) or runs out of packages (bad).
1964
1965 Whenever this type is used, C<push_read> will invoke the function with the
1966 handle object, the original callback and the remaining arguments.
1967
1968 The function is supposed to return a callback (usually a closure) that
1969 works as a plain read callback (see C<< ->push_read ($cb) >>), so you can
1970 mentally treat the function as a "configurable read type to read callback"
1971 converter.
1972
1973 It should invoke the original callback when it is done reading (remember
1974 to pass C<$handle> as first argument as all other callbacks do that,
1975 although there is no strict requirement on this).
1976
1977 For examples, see the source of this module (F<perldoc -m
1978 AnyEvent::Handle>, search for C<register_read_type>)).
1979
1980 =item $handle->stop_read
1981
1982 =item $handle->start_read
1983
1984 In rare cases you actually do not want to read anything from the
1985 socket. In this case you can call C<stop_read>. Neither C<on_read> nor
1986 any queued callbacks will be executed then. To start reading again, call
1987 C<start_read>.
1988
1989 Note that AnyEvent::Handle will automatically C<start_read> for you when
1990 you change the C<on_read> callback or push/unshift a read callback, and it
1991 will automatically C<stop_read> for you when neither C<on_read> is set nor
1992 there are any read requests in the queue.
1993
1994 In older versions of this module (<= 5.3), these methods had no effect,
1995 as TLS does not support half-duplex connections. In current versions they
1996 work as expected, as this behaviour is required to avoid certain resource
1997 attacks, where the program would be forced to read (and buffer) arbitrary
1998 amounts of data before being able to send some data. The drawback is that
1999 some readings of the the SSL/TLS specifications basically require this
2000 attack to be working, as SSL/TLS implementations might stall sending data
2001 during a rehandshake.
2002
2003 As a guideline, during the initial handshake, you should not stop reading,
2004 and as a client, it might cause problems, depending on your application.
2005
2006 =cut
2007
2008 sub stop_read {
2009 my ($self) = @_;
2010
2011 delete $self->{_rw};
2012 }
2013
2014 sub start_read {
2015 my ($self) = @_;
2016
2017 unless ($self->{_rw} || $self->{_eof} || !$self->{fh}) {
2018 Scalar::Util::weaken $self;
2019
2020 $self->{_rw} = AE::io $self->{fh}, 0, sub {
2021 my $rbuf = \($self->{tls} ? my $buf : $self->{rbuf});
2022 my $len = sysread $self->{fh}, $$rbuf, $self->{read_size}, length $$rbuf;
2023
2024 if ($len > 0) {
2025 $self->{_activity} = $self->{_ractivity} = AE::now;
2026
2027 if ($self->{tls}) {
2028 Net::SSLeay::BIO_write ($self->{_rbio}, $$rbuf);
2029
2030 &_dotls ($self);
2031 } else {
2032 $self->_drain_rbuf;
2033 }
2034
2035 if ($len == $self->{read_size}) {
2036 $self->{read_size} *= 2;
2037 $self->{read_size} = $self->{max_read_size} || MAX_READ_SIZE
2038 if $self->{read_size} > ($self->{max_read_size} || MAX_READ_SIZE);
2039 }
2040
2041 } elsif (defined $len) {
2042 delete $self->{_rw};
2043 $self->{_eof} = 1;
2044 $self->_drain_rbuf;
2045
2046 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
2047 return $self->_error ($!, 1);
2048 }
2049 };
2050 }
2051 }
2052
2053 our $ERROR_SYSCALL;
2054 our $ERROR_WANT_READ;
2055
2056 sub _tls_error {
2057 my ($self, $err) = @_;
2058
2059 return $self->_error ($!, 1)
2060 if $err == Net::SSLeay::ERROR_SYSCALL ();
2061
2062 my $err = Net::SSLeay::ERR_error_string (Net::SSLeay::ERR_get_error ());
2063
2064 # reduce error string to look less scary
2065 $err =~ s/^error:[0-9a-fA-F]{8}:[^:]+:([^:]+):/\L$1: /;
2066
2067 if ($self->{_on_starttls}) {
2068 (delete $self->{_on_starttls})->($self, undef, $err);
2069 &_freetls;
2070 } else {
2071 &_freetls;
2072 $self->_error (Errno::EPROTO, 1, $err);
2073 }
2074 }
2075
2076 # poll the write BIO and send the data if applicable
2077 # also decode read data if possible
2078 # this is basiclaly our TLS state machine
2079 # more efficient implementations are possible with openssl,
2080 # but not with the buggy and incomplete Net::SSLeay.
2081 sub _dotls {
2082 my ($self) = @_;
2083
2084 my $tmp;
2085
2086 while (length $self->{_tls_wbuf}) {
2087 if (($tmp = Net::SSLeay::write ($self->{tls}, $self->{_tls_wbuf})) <= 0) {
2088 $tmp = Net::SSLeay::get_error ($self->{tls}, $tmp);
2089
2090 return $self->_tls_error ($tmp)
2091 if $tmp != $ERROR_WANT_READ
2092 && ($tmp != $ERROR_SYSCALL || $!);
2093
2094 last;
2095 }
2096
2097 substr $self->{_tls_wbuf}, 0, $tmp, "";
2098 }
2099
2100 while (defined ($tmp = Net::SSLeay::read ($self->{tls}))) {
2101 unless (length $tmp) {
2102 $self->{_on_starttls}
2103 and (delete $self->{_on_starttls})->($self, undef, "EOF during handshake"); # ???
2104 &_freetls;
2105
2106 if ($self->{on_stoptls}) {
2107 $self->{on_stoptls}($self);
2108 return;
2109 } else {
2110 # let's treat SSL-eof as we treat normal EOF
2111 delete $self->{_rw};
2112 $self->{_eof} = 1;
2113 }
2114 }
2115
2116 $self->{_tls_rbuf} .= $tmp;
2117 $self->_drain_rbuf;
2118 $self->{tls} or return; # tls session might have gone away in callback
2119 }
2120
2121 $tmp = Net::SSLeay::get_error ($self->{tls}, -1); # -1 is not neccessarily correct, but Net::SSLeay doesn't tell us
2122 return $self->_tls_error ($tmp)
2123 if $tmp != $ERROR_WANT_READ
2124 && ($tmp != $ERROR_SYSCALL || $!);
2125
2126 while (length ($tmp = Net::SSLeay::BIO_read ($self->{_wbio}))) {
2127 $self->{wbuf} .= $tmp;
2128 $self->_drain_wbuf;
2129 $self->{tls} or return; # tls session might have gone away in callback
2130 }
2131
2132 $self->{_on_starttls}
2133 and Net::SSLeay::state ($self->{tls}) == Net::SSLeay::ST_OK ()
2134 and (delete $self->{_on_starttls})->($self, 1, "TLS/SSL connection established");
2135 }
2136
2137 =item $handle->starttls ($tls[, $tls_ctx])
2138
2139 Instead of starting TLS negotiation immediately when the AnyEvent::Handle
2140 object is created, you can also do that at a later time by calling
2141 C<starttls>. See the C<tls> constructor argument for general info.
2142
2143 Starting TLS is currently an asynchronous operation - when you push some
2144 write data and then call C<< ->starttls >> then TLS negotiation will start
2145 immediately, after which the queued write data is then sent. This might
2146 change in future versions, so best make sure you have no outstanding write
2147 data when calling this method.
2148
2149 The first argument is the same as the C<tls> constructor argument (either
2150 C<"connect">, C<"accept"> or an existing Net::SSLeay object).
2151
2152 The second argument is the optional C<AnyEvent::TLS> object that is used
2153 when AnyEvent::Handle has to create its own TLS connection object, or
2154 a hash reference with C<< key => value >> pairs that will be used to
2155 construct a new context.
2156
2157 The TLS connection object will end up in C<< $handle->{tls} >>, the TLS
2158 context in C<< $handle->{tls_ctx} >> after this call and can be used or
2159 changed to your liking. Note that the handshake might have already started
2160 when this function returns.
2161
2162 Due to bugs in OpenSSL, it might or might not be possible to do multiple
2163 handshakes on the same stream. It is best to not attempt to use the
2164 stream after stopping TLS.
2165
2166 This method may invoke callbacks (and therefore the handle might be
2167 destroyed after it returns).
2168
2169 =cut
2170
2171 our %TLS_CACHE; #TODO not yet documented, should we?
2172
2173 sub starttls {
2174 my ($self, $tls, $ctx) = @_;
2175
2176 Carp::croak "It is an error to call starttls on an AnyEvent::Handle object while TLS is already active, caught"
2177 if $self->{tls};
2178
2179 unless (defined $AnyEvent::TLS::VERSION) {
2180 eval {
2181 require Net::SSLeay;
2182 require AnyEvent::TLS;
2183 1
2184 } or return $self->_error (Errno::EPROTO, 1, "TLS support not available on this system");
2185 }
2186
2187 $self->{tls} = $tls;
2188 $self->{tls_ctx} = $ctx if @_ > 2;
2189
2190 return unless $self->{fh};
2191
2192 $ERROR_SYSCALL = Net::SSLeay::ERROR_SYSCALL ();
2193 $ERROR_WANT_READ = Net::SSLeay::ERROR_WANT_READ ();
2194
2195 $tls = delete $self->{tls};
2196 $ctx = $self->{tls_ctx};
2197
2198 local $Carp::CarpLevel = 1; # skip ourselves when creating a new context or session
2199
2200 if ("HASH" eq ref $ctx) {
2201 if ($ctx->{cache}) {
2202 my $key = $ctx+0;
2203 $ctx = $TLS_CACHE{$key} ||= new AnyEvent::TLS %$ctx;
2204 } else {
2205 $ctx = new AnyEvent::TLS %$ctx;
2206 }
2207 }
2208
2209 $self->{tls_ctx} = $ctx || TLS_CTX ();
2210 $self->{tls} = $tls = $self->{tls_ctx}->_get_session ($tls, $self, $self->{peername});
2211
2212 # basically, this is deep magic (because SSL_read should have the same issues)
2213 # but the openssl maintainers basically said: "trust us, it just works".
2214 # (unfortunately, we have to hardcode constants because the abysmally misdesigned
2215 # and mismaintained ssleay-module doesn't even offer them).
2216 # http://www.mail-archive.com/openssl-dev@openssl.org/msg22420.html
2217 #
2218 # in short: this is a mess.
2219 #
2220 # note that we do not try to keep the length constant between writes as we are required to do.
2221 # we assume that most (but not all) of this insanity only applies to non-blocking cases,
2222 # and we drive openssl fully in blocking mode here. Or maybe we don't - openssl seems to
2223 # have identity issues in that area.
2224 # Net::SSLeay::CTX_set_mode ($ssl,
2225 # (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ENABLE_PARTIAL_WRITE () } || 1)
2226 # | (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ACCEPT_MOVING_WRITE_BUFFER () } || 2));
2227 Net::SSLeay::CTX_set_mode ($tls, 1|2);
2228
2229 $self->{_rbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
2230 $self->{_wbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
2231
2232 Net::SSLeay::BIO_write ($self->{_rbio}, $self->{rbuf});
2233 $self->{rbuf} = "";
2234
2235 Net::SSLeay::set_bio ($tls, $self->{_rbio}, $self->{_wbio});
2236
2237 $self->{_on_starttls} = sub { $_[0]{on_starttls}(@_) }
2238 if $self->{on_starttls};
2239
2240 &_dotls; # need to trigger the initial handshake
2241 $self->start_read; # make sure we actually do read
2242 }
2243
2244 =item $handle->stoptls
2245
2246 Shuts down the SSL connection - this makes a proper EOF handshake by
2247 sending a close notify to the other side, but since OpenSSL doesn't
2248 support non-blocking shut downs, it is not guaranteed that you can re-use
2249 the stream afterwards.
2250
2251 This method may invoke callbacks (and therefore the handle might be
2252 destroyed after it returns).
2253
2254 =cut
2255
2256 sub stoptls {
2257 my ($self) = @_;
2258
2259 if ($self->{tls} && $self->{fh}) {
2260 Net::SSLeay::shutdown ($self->{tls});
2261
2262 &_dotls;
2263
2264 # # we don't give a shit. no, we do, but we can't. no...#d#
2265 # # we, we... have to use openssl :/#d#
2266 # &_freetls;#d#
2267 }
2268 }
2269
2270 sub _freetls {
2271 my ($self) = @_;
2272
2273 return unless $self->{tls};
2274
2275 $self->{tls_ctx}->_put_session (delete $self->{tls})
2276 if $self->{tls} > 0;
2277
2278 delete @$self{qw(_rbio _wbio _tls_wbuf _on_starttls)};
2279 }
2280
2281 =item $handle->resettls
2282
2283 This rarely-used method simply resets and TLS state on the handle, usually
2284 causing data loss.
2285
2286 One case where it may be useful is when you want to skip over the data in
2287 the stream but you are not interested in interpreting it, so data loss is
2288 no concern.
2289
2290 =cut
2291
2292 *resettls = \&_freetls;
2293
2294 sub DESTROY {
2295 my ($self) = @_;
2296
2297 &_freetls;
2298
2299 my $linger = exists $self->{linger} ? $self->{linger} : 3600;
2300
2301 if ($linger && length $self->{wbuf} && $self->{fh}) {
2302 my $fh = delete $self->{fh};
2303 my $wbuf = delete $self->{wbuf};
2304
2305 my @linger;
2306
2307 push @linger, AE::io $fh, 1, sub {
2308 my $len = syswrite $fh, $wbuf, length $wbuf;
2309
2310 if ($len > 0) {
2311 substr $wbuf, 0, $len, "";
2312 } elsif (defined $len || ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK)) {
2313 @linger = (); # end
2314 }
2315 };
2316 push @linger, AE::timer $linger, 0, sub {
2317 @linger = ();
2318 };
2319 }
2320 }
2321
2322 =item $handle->destroy
2323
2324 Shuts down the handle object as much as possible - this call ensures that
2325 no further callbacks will be invoked and as many resources as possible
2326 will be freed. Any method you will call on the handle object after
2327 destroying it in this way will be silently ignored (and it will return the
2328 empty list).
2329
2330 Normally, you can just "forget" any references to an AnyEvent::Handle
2331 object and it will simply shut down. This works in fatal error and EOF
2332 callbacks, as well as code outside. It does I<NOT> work in a read or write
2333 callback, so when you want to destroy the AnyEvent::Handle object from
2334 within such an callback. You I<MUST> call C<< ->destroy >> explicitly in
2335 that case.
2336
2337 Destroying the handle object in this way has the advantage that callbacks
2338 will be removed as well, so if those are the only reference holders (as
2339 is common), then one doesn't need to do anything special to break any
2340 reference cycles.
2341
2342 The handle might still linger in the background and write out remaining
2343 data, as specified by the C<linger> option, however.
2344
2345 =cut
2346
2347 sub destroy {
2348 my ($self) = @_;
2349
2350 $self->DESTROY;
2351 %$self = ();
2352 bless $self, "AnyEvent::Handle::destroyed";
2353 }
2354
2355 sub AnyEvent::Handle::destroyed::AUTOLOAD {
2356 #nop
2357 }
2358
2359 =item $handle->destroyed
2360
2361 Returns false as long as the handle hasn't been destroyed by a call to C<<
2362 ->destroy >>, true otherwise.
2363
2364 Can be useful to decide whether the handle is still valid after some
2365 callback possibly destroyed the handle. For example, C<< ->push_write >>,
2366 C<< ->starttls >> and other methods can call user callbacks, which in turn
2367 can destroy the handle, so work can be avoided by checking sometimes:
2368
2369 $hdl->starttls ("accept");
2370 return if $hdl->destroyed;
2371 $hdl->push_write (...
2372
2373 Note that the call to C<push_write> will silently be ignored if the handle
2374 has been destroyed, so often you can just ignore the possibility of the
2375 handle being destroyed.
2376
2377 =cut
2378
2379 sub destroyed { 0 }
2380 sub AnyEvent::Handle::destroyed::destroyed { 1 }
2381
2382 =item AnyEvent::Handle::TLS_CTX
2383
2384 This function creates and returns the AnyEvent::TLS object used by default
2385 for TLS mode.
2386
2387 The context is created by calling L<AnyEvent::TLS> without any arguments.
2388
2389 =cut
2390
2391 our $TLS_CTX;
2392
2393 sub TLS_CTX() {
2394 $TLS_CTX ||= do {
2395 require AnyEvent::TLS;
2396
2397 new AnyEvent::TLS
2398 }
2399 }
2400
2401 =back
2402
2403
2404 =head1 NONFREQUENTLY ASKED QUESTIONS
2405
2406 =over 4
2407
2408 =item I C<undef> the AnyEvent::Handle reference inside my callback and
2409 still get further invocations!
2410
2411 That's because AnyEvent::Handle keeps a reference to itself when handling
2412 read or write callbacks.
2413
2414 It is only safe to "forget" the reference inside EOF or error callbacks,
2415 from within all other callbacks, you need to explicitly call the C<<
2416 ->destroy >> method.
2417
2418 =item Why is my C<on_eof> callback never called?
2419
2420 Probably because your C<on_error> callback is being called instead: When
2421 you have outstanding requests in your read queue, then an EOF is
2422 considered an error as you clearly expected some data.
2423
2424 To avoid this, make sure you have an empty read queue whenever your handle
2425 is supposed to be "idle" (i.e. connection closes are O.K.). You can set
2426 an C<on_read> handler that simply pushes the first read requests in the
2427 queue.
2428
2429 See also the next question, which explains this in a bit more detail.
2430
2431 =item How can I serve requests in a loop?
2432
2433 Most protocols consist of some setup phase (authentication for example)
2434 followed by a request handling phase, where the server waits for requests
2435 and handles them, in a loop.
2436
2437 There are two important variants: The first (traditional, better) variant
2438 handles requests until the server gets some QUIT command, causing it to
2439 close the connection first (highly desirable for a busy TCP server). A
2440 client dropping the connection is an error, which means this variant can
2441 detect an unexpected detection close.
2442
2443 To handle this case, always make sure you have a non-empty read queue, by
2444 pushing the "read request start" handler on it:
2445
2446 # we assume a request starts with a single line
2447 my @start_request; @start_request = (line => sub {
2448 my ($hdl, $line) = @_;
2449
2450 ... handle request
2451
2452 # push next request read, possibly from a nested callback
2453 $hdl->push_read (@start_request);
2454 });
2455
2456 # auth done, now go into request handling loop
2457 # now push the first @start_request
2458 $hdl->push_read (@start_request);
2459
2460 By always having an outstanding C<push_read>, the handle always expects
2461 some data and raises the C<EPIPE> error when the connction is dropped
2462 unexpectedly.
2463
2464 The second variant is a protocol where the client can drop the connection
2465 at any time. For TCP, this means that the server machine may run out of
2466 sockets easier, and in general, it means you cannot distinguish a protocl
2467 failure/client crash from a normal connection close. Nevertheless, these
2468 kinds of protocols are common (and sometimes even the best solution to the
2469 problem).
2470
2471 Having an outstanding read request at all times is possible if you ignore
2472 C<EPIPE> errors, but this doesn't help with when the client drops the
2473 connection during a request, which would still be an error.
2474
2475 A better solution is to push the initial request read in an C<on_read>
2476 callback. This avoids an error, as when the server doesn't expect data
2477 (i.e. is idly waiting for the next request, an EOF will not raise an
2478 error, but simply result in an C<on_eof> callback. It is also a bit slower
2479 and simpler:
2480
2481 # auth done, now go into request handling loop
2482 $hdl->on_read (sub {
2483 my ($hdl) = @_;
2484
2485 # called each time we receive data but the read queue is empty
2486 # simply start read the request
2487
2488 $hdl->push_read (line => sub {
2489 my ($hdl, $line) = @_;
2490
2491 ... handle request
2492
2493 # do nothing special when the request has been handled, just
2494 # let the request queue go empty.
2495 });
2496 });
2497
2498 =item I get different callback invocations in TLS mode/Why can't I pause
2499 reading?
2500
2501 Unlike, say, TCP, TLS connections do not consist of two independent
2502 communication channels, one for each direction. Or put differently, the
2503 read and write directions are not independent of each other: you cannot
2504 write data unless you are also prepared to read, and vice versa.
2505
2506 This means that, in TLS mode, you might get C<on_error> or C<on_eof>
2507 callback invocations when you are not expecting any read data - the reason
2508 is that AnyEvent::Handle always reads in TLS mode.
2509
2510 During the connection, you have to make sure that you always have a
2511 non-empty read-queue, or an C<on_read> watcher. At the end of the
2512 connection (or when you no longer want to use it) you can call the
2513 C<destroy> method.
2514
2515 =item How do I read data until the other side closes the connection?
2516
2517 If you just want to read your data into a perl scalar, the easiest way
2518 to achieve this is by setting an C<on_read> callback that does nothing,
2519 clearing the C<on_eof> callback and in the C<on_error> callback, the data
2520 will be in C<$_[0]{rbuf}>:
2521
2522 $handle->on_read (sub { });
2523 $handle->on_eof (undef);
2524 $handle->on_error (sub {
2525 my $data = delete $_[0]{rbuf};
2526 });
2527
2528 Note that this example removes the C<rbuf> member from the handle object,
2529 which is not normally allowed by the API. It is expressly permitted in
2530 this case only, as the handle object needs to be destroyed afterwards.
2531
2532 The reason to use C<on_error> is that TCP connections, due to latencies
2533 and packets loss, might get closed quite violently with an error, when in
2534 fact all data has been received.
2535
2536 It is usually better to use acknowledgements when transferring data,
2537 to make sure the other side hasn't just died and you got the data
2538 intact. This is also one reason why so many internet protocols have an
2539 explicit QUIT command.
2540
2541 =item I don't want to destroy the handle too early - how do I wait until
2542 all data has been written?
2543
2544 After writing your last bits of data, set the C<on_drain> callback
2545 and destroy the handle in there - with the default setting of
2546 C<low_water_mark> this will be called precisely when all data has been
2547 written to the socket:
2548
2549 $handle->push_write (...);
2550 $handle->on_drain (sub {
2551 AE::log debug => "All data submitted to the kernel.";
2552 undef $handle;
2553 });
2554
2555 If you just want to queue some data and then signal EOF to the other side,
2556 consider using C<< ->push_shutdown >> instead.
2557
2558 =item I want to contact a TLS/SSL server, I don't care about security.
2559
2560 If your TLS server is a pure TLS server (e.g. HTTPS) that only speaks TLS,
2561 connect to it and then create the AnyEvent::Handle with the C<tls>
2562 parameter:
2563
2564 tcp_connect $host, $port, sub {
2565 my ($fh) = @_;
2566
2567 my $handle = new AnyEvent::Handle
2568 fh => $fh,
2569 tls => "connect",
2570 on_error => sub { ... };
2571
2572 $handle->push_write (...);
2573 };
2574
2575 =item I want to contact a TLS/SSL server, I do care about security.
2576
2577 Then you should additionally enable certificate verification, including
2578 peername verification, if the protocol you use supports it (see
2579 L<AnyEvent::TLS>, C<verify_peername>).
2580
2581 E.g. for HTTPS:
2582
2583 tcp_connect $host, $port, sub {
2584 my ($fh) = @_;
2585
2586 my $handle = new AnyEvent::Handle
2587 fh => $fh,
2588 peername => $host,
2589 tls => "connect",
2590 tls_ctx => { verify => 1, verify_peername => "https" },
2591 ...
2592
2593 Note that you must specify the hostname you connected to (or whatever
2594 "peername" the protocol needs) as the C<peername> argument, otherwise no
2595 peername verification will be done.
2596
2597 The above will use the system-dependent default set of trusted CA
2598 certificates. If you want to check against a specific CA, add the
2599 C<ca_file> (or C<ca_cert>) arguments to C<tls_ctx>:
2600
2601 tls_ctx => {
2602 verify => 1,
2603 verify_peername => "https",
2604 ca_file => "my-ca-cert.pem",
2605 },
2606
2607 =item I want to create a TLS/SSL server, how do I do that?
2608
2609 Well, you first need to get a server certificate and key. You have
2610 three options: a) ask a CA (buy one, use cacert.org etc.) b) create a
2611 self-signed certificate (cheap. check the search engine of your choice,
2612 there are many tutorials on the net) or c) make your own CA (tinyca2 is a
2613 nice program for that purpose).
2614
2615 Then create a file with your private key (in PEM format, see
2616 L<AnyEvent::TLS>), followed by the certificate (also in PEM format). The
2617 file should then look like this:
2618
2619 -----BEGIN RSA PRIVATE KEY-----
2620 ...header data
2621 ... lots of base64'y-stuff
2622 -----END RSA PRIVATE KEY-----
2623
2624 -----BEGIN CERTIFICATE-----
2625 ... lots of base64'y-stuff
2626 -----END CERTIFICATE-----
2627
2628 The important bits are the "PRIVATE KEY" and "CERTIFICATE" parts. Then
2629 specify this file as C<cert_file>:
2630
2631 tcp_server undef, $port, sub {
2632 my ($fh) = @_;
2633
2634 my $handle = new AnyEvent::Handle
2635 fh => $fh,
2636 tls => "accept",
2637 tls_ctx => { cert_file => "my-server-keycert.pem" },
2638 ...
2639
2640 When you have intermediate CA certificates that your clients might not
2641 know about, just append them to the C<cert_file>.
2642
2643 =back
2644
2645 =head1 SUBCLASSING AnyEvent::Handle
2646
2647 In many cases, you might want to subclass AnyEvent::Handle.
2648
2649 To make this easier, a given version of AnyEvent::Handle uses these
2650 conventions:
2651
2652 =over 4
2653
2654 =item * all constructor arguments become object members.
2655
2656 At least initially, when you pass a C<tls>-argument to the constructor it
2657 will end up in C<< $handle->{tls} >>. Those members might be changed or
2658 mutated later on (for example C<tls> will hold the TLS connection object).
2659
2660 =item * other object member names are prefixed with an C<_>.
2661
2662 All object members not explicitly documented (internal use) are prefixed
2663 with an underscore character, so the remaining non-C<_>-namespace is free
2664 for use for subclasses.
2665
2666 =item * all members not documented here and not prefixed with an underscore
2667 are free to use in subclasses.
2668
2669 Of course, new versions of AnyEvent::Handle may introduce more "public"
2670 member variables, but that's just life. At least it is documented.
2671
2672 =back
2673
2674 =head1 AUTHOR
2675
2676 Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>.
2677
2678 =cut
2679
2680 1
2681