ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent/lib/AnyEvent/Handle.pm
(Generate patch)

Comparing AnyEvent/lib/AnyEvent/Handle.pm (file contents):
Revision 1.26 by root, Sat May 24 15:20:46 2008 UTC vs.
Revision 1.183 by root, Thu Sep 3 12:45:35 2009 UTC

1package AnyEvent::Handle;
2
3no warnings;
4use strict;
5
6use AnyEvent ();
7use AnyEvent::Util ();
8use Scalar::Util ();
9use Carp ();
10use Fcntl ();
11use Errno qw/EAGAIN EINTR/;
12
13=head1 NAME 1=head1 NAME
14 2
15AnyEvent::Handle - non-blocking I/O on file handles via AnyEvent 3AnyEvent::Handle - non-blocking I/O on file handles via AnyEvent
16
17This module is experimental.
18
19=cut
20
21our $VERSION = '0.04';
22 4
23=head1 SYNOPSIS 5=head1 SYNOPSIS
24 6
25 use AnyEvent; 7 use AnyEvent;
26 use AnyEvent::Handle; 8 use AnyEvent::Handle;
27 9
28 my $cv = AnyEvent->condvar; 10 my $cv = AnyEvent->condvar;
29 11
30 my $ae_fh = AnyEvent::Handle->new (fh => \*STDIN); 12 my $hdl; $hdl = new AnyEvent::Handle
31
32 #TODO
33
34 # or use the constructor to pass the callback:
35
36 my $ae_fh2 =
37 AnyEvent::Handle->new (
38 fh => \*STDIN, 13 fh => \*STDIN,
39 on_eof => sub { 14 on_error => sub {
40 $cv->broadcast; 15 my ($hdl, $fatal, $msg) = @_;
41 }, 16 warn "got error $msg\n";
42 #TODO 17 $hdl->destroy;
18 $cv->send;
43 ); 19 );
44 20
45 $cv->wait; 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, $line) = @_;
27 warn "got line <$line>\n";
28 $cv->send;
29 });
30
31 $cv->recv;
46 32
47=head1 DESCRIPTION 33=head1 DESCRIPTION
48 34
49This module is a helper module to make it easier to do event-based I/O on 35This module is a helper module to make it easier to do event-based I/O on
50filehandles. For utility functions for doing non-blocking connects and accepts 36filehandles.
51on sockets see L<AnyEvent::Util>. 37
38The L<AnyEvent::Intro> tutorial contains some well-documented
39AnyEvent::Handle examples.
52 40
53In the following, when the documentation refers to of "bytes" then this 41In the following, when the documentation refers to of "bytes" then this
54means characters. As sysread and syswrite are used for all I/O, their 42means characters. As sysread and syswrite are used for all I/O, their
55treatment of characters applies to this module as well. 43treatment of characters applies to this module as well.
56 44
45At the very minimum, you should specify C<fh> or C<connect>, and the
46C<on_error> callback.
47
57All callbacks will be invoked with the handle object as their first 48All callbacks will be invoked with the handle object as their first
58argument. 49argument.
59 50
51=cut
52
53package AnyEvent::Handle;
54
55use Scalar::Util ();
56use List::Util ();
57use Carp ();
58use Errno qw(EAGAIN EINTR);
59
60use AnyEvent (); BEGIN { AnyEvent::common_sense }
61use AnyEvent::Util qw(WSAEWOULDBLOCK);
62
63our $VERSION = $AnyEvent::VERSION;
64
60=head1 METHODS 65=head1 METHODS
61 66
62=over 4 67=over 4
63 68
64=item B<new (%args)> 69=item $handle = B<new> AnyEvent::TLS fh => $filehandle, key => value...
65 70
66The constructor supports these arguments (all as key => value pairs). 71The constructor supports these arguments (all as C<< key => value >> pairs).
67 72
68=over 4 73=over 4
69 74
70=item fh => $filehandle [MANDATORY] 75=item fh => $filehandle [C<fh> or C<connect> MANDATORY]
71 76
72The filehandle this L<AnyEvent::Handle> object will operate on. 77The filehandle this L<AnyEvent::Handle> object will operate on.
73
74NOTE: The filehandle will be set to non-blocking (using 78NOTE: The filehandle will be set to non-blocking mode (using
75AnyEvent::Util::fh_nonblocking). 79C<AnyEvent::Util::fh_nonblocking>) by the constructor and needs to stay in
80that mode.
76 81
77=item on_eof => $cb->($self) 82=item connect => [$host, $service] [C<fh> or C<connect> MANDATORY]
78 83
79Set the callback to be called on EOF. 84Try to connect to the specified host and service (port), using
85C<AnyEvent::Socket::tcp_connect>. The C<$host> additionally becomes the
86default C<peername>.
80 87
81While not mandatory, it is highly recommended to set an eof callback, 88You have to specify either this parameter, or C<fh>, above.
82otherwise you might end up with a closed socket while you are still
83waiting for data.
84 89
90It is possible to push requests on the read and write queues, and modify
91properties of the stream, even while AnyEvent::Handle is connecting.
92
93When this parameter is specified, then the C<on_prepare>,
94C<on_connect_error> and C<on_connect> callbacks will be called under the
95appropriate circumstances:
96
97=over 4
98
85=item on_error => $cb->($self) 99=item on_prepare => $cb->($handle)
86 100
101This (rarely used) callback is called before a new connection is
102attempted, but after the file handle has been created. It could be used to
103prepare the file handle with parameters required for the actual connect
104(as opposed to settings that can be changed when the connection is already
105established).
106
107The return value of this callback should be the connect timeout value in
108seconds (or C<0>, or C<undef>, or the empty list, to indicate the default
109timeout is to be used).
110
111=item on_connect => $cb->($handle, $host, $port, $retry->())
112
113This callback is called when a connection has been successfully established.
114
115The actual numeric host and port (the socket peername) are passed as
116parameters, together with a retry callback.
117
118When, for some reason, the handle is not acceptable, then calling
119C<$retry> will continue with the next conenction target (in case of
120multi-homed hosts or SRV records there can be multiple connection
121endpoints). When it is called then the read and write queues, eof status,
122tls status and similar properties of the handle are being reset.
123
124In most cases, ignoring the C<$retry> parameter is the way to go.
125
126=item on_connect_error => $cb->($handle, $message)
127
128This callback is called when the conenction could not be
129established. C<$!> will contain the relevant error code, and C<$message> a
130message describing it (usually the same as C<"$!">).
131
132If this callback isn't specified, then C<on_error> will be called with a
133fatal error instead.
134
135=back
136
137=item on_error => $cb->($handle, $fatal, $message)
138
87This is the fatal error callback, that is called when, well, a fatal error 139This is the error callback, which is called when, well, some error
88occurs, such as not being able to resolve the hostname, failure to connect 140occured, such as not being able to resolve the hostname, failure to
89or a read error. 141connect or a read error.
90 142
91The object will not be in a usable state when this callback has been 143Some errors are fatal (which is indicated by C<$fatal> being true). On
92called. 144fatal errors the handle object will be destroyed (by a call to C<< ->
145destroy >>) after invoking the error callback (which means you are free to
146examine the handle object). Examples of fatal errors are an EOF condition
147with active (but unsatisifable) read watchers (C<EPIPE>) or I/O errors. In
148cases where the other side can close the connection at their will it is
149often easiest to not report C<EPIPE> errors in this callback.
150
151AnyEvent::Handle tries to find an appropriate error code for you to check
152against, but in some cases (TLS errors), this does not work well. It is
153recommended to always output the C<$message> argument in human-readable
154error messages (it's usually the same as C<"$!">).
155
156Non-fatal errors can be retried by simply returning, but it is recommended
157to simply ignore this parameter and instead abondon the handle object
158when this callback is invoked. Examples of non-fatal errors are timeouts
159C<ETIMEDOUT>) or badly-formatted data (C<EBADMSG>).
93 160
94On callback entrance, the value of C<$!> contains the operating system 161On callback entrance, the value of C<$!> contains the operating system
95error (or C<ENOSPC> or C<EPIPE>). 162error code (or C<ENOSPC>, C<EPIPE>, C<ETIMEDOUT>, C<EBADMSG> or
163C<EPROTO>).
96 164
97While not mandatory, it is I<highly> recommended to set this callback, as 165While not mandatory, it is I<highly> recommended to set this callback, as
98you will not be notified of errors otherwise. The default simply calls 166you will not be notified of errors otherwise. The default simply calls
99die. 167C<croak>.
100 168
101=item on_read => $cb->($self) 169=item on_read => $cb->($handle)
102 170
103This sets the default read callback, which is called when data arrives 171This sets the default read callback, which is called when data arrives
104and no read request is in the queue. 172and no read request is in the queue (unlike read queue callbacks, this
173callback will only be called when at least one octet of data is in the
174read buffer).
105 175
106To access (and remove data from) the read buffer, use the C<< ->rbuf >> 176To access (and remove data from) the read buffer, use the C<< ->rbuf >>
107method or access the C<$self->{rbuf}> member directly. 177method or access the C<< $handle->{rbuf} >> member directly. Note that you
178must not enlarge or modify the read buffer, you can only remove data at
179the beginning from it.
108 180
109When an EOF condition is detected then AnyEvent::Handle will first try to 181When an EOF condition is detected then AnyEvent::Handle will first try to
110feed all the remaining data to the queued callbacks and C<on_read> before 182feed all the remaining data to the queued callbacks and C<on_read> before
111calling the C<on_eof> callback. If no progress can be made, then a fatal 183calling the C<on_eof> callback. If no progress can be made, then a fatal
112error will be raised (with C<$!> set to C<EPIPE>). 184error will be raised (with C<$!> set to C<EPIPE>).
113 185
186Note that, unlike requests in the read queue, an C<on_read> callback
187doesn't mean you I<require> some data: if there is an EOF and there
188are outstanding read requests then an error will be flagged. With an
189C<on_read> callback, the C<on_eof> callback will be invoked.
190
191=item on_eof => $cb->($handle)
192
193Set the callback to be called when an end-of-file condition is detected,
194i.e. in the case of a socket, when the other side has closed the
195connection cleanly, and there are no outstanding read requests in the
196queue (if there are read requests, then an EOF counts as an unexpected
197connection close and will be flagged as an error).
198
199For sockets, this just means that the other side has stopped sending data,
200you can still try to write data, and, in fact, one can return from the EOF
201callback and continue writing data, as only the read part has been shut
202down.
203
204If an EOF condition has been detected but no C<on_eof> callback has been
205set, then a fatal error will be raised with C<$!> set to <0>.
206
114=item on_drain => $cb->() 207=item on_drain => $cb->($handle)
115 208
116This sets the callback that is called when the write buffer becomes empty 209This sets the callback that is called when the write buffer becomes empty
117(or when the callback is set and the buffer is empty already). 210(or when the callback is set and the buffer is empty already).
118 211
119To append to the write buffer, use the C<< ->push_write >> method. 212To append to the write buffer, use the C<< ->push_write >> method.
120 213
214This callback is useful when you don't want to put all of your write data
215into the queue at once, for example, when you want to write the contents
216of some file to the socket you might not want to read the whole file into
217memory and push it into the queue, but instead only read more data from
218the file when the write queue becomes empty.
219
220=item timeout => $fractional_seconds
221
222=item rtimeout => $fractional_seconds
223
224=item wtimeout => $fractional_seconds
225
226If non-zero, then these enables an "inactivity" timeout: whenever this
227many seconds pass without a successful read or write on the underlying
228file handle (or a call to C<timeout_reset>), the C<on_timeout> callback
229will be invoked (and if that one is missing, a non-fatal C<ETIMEDOUT>
230error will be raised).
231
232There are three variants of the timeouts that work fully independent
233of each other, for both read and write, just read, and just write:
234C<timeout>, C<rtimeout> and C<wtimeout>, with corresponding callbacks
235C<on_timeout>, C<on_rtimeout> and C<on_wtimeout>, and reset functions
236C<timeout_reset>, C<rtimeout_reset>, and C<wtimeout_reset>.
237
238Note that timeout processing is also active when you currently do not have
239any outstanding read or write requests: If you plan to keep the connection
240idle then you should disable the timout temporarily or ignore the timeout
241in the C<on_timeout> callback, in which case AnyEvent::Handle will simply
242restart the timeout.
243
244Zero (the default) disables this timeout.
245
246=item on_timeout => $cb->($handle)
247
248Called whenever the inactivity timeout passes. If you return from this
249callback, then the timeout will be reset as if some activity had happened,
250so this condition is not fatal in any way.
251
121=item rbuf_max => <bytes> 252=item rbuf_max => <bytes>
122 253
123If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>) 254If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>)
124when the read buffer ever (strictly) exceeds this size. This is useful to 255when the read buffer ever (strictly) exceeds this size. This is useful to
125avoid denial-of-service attacks. 256avoid some forms of denial-of-service attacks.
126 257
127For example, a server accepting connections from untrusted sources should 258For example, a server accepting connections from untrusted sources should
128be configured to accept only so-and-so much data that it cannot act on 259be configured to accept only so-and-so much data that it cannot act on
129(for example, when expecting a line, an attacker could send an unlimited 260(for example, when expecting a line, an attacker could send an unlimited
130amount of data without a callback ever being called as long as the line 261amount of data without a callback ever being called as long as the line
131isn't finished). 262isn't finished).
132 263
264=item autocork => <boolean>
265
266When disabled (the default), then C<push_write> will try to immediately
267write the data to the handle, if possible. This avoids having to register
268a write watcher and wait for the next event loop iteration, but can
269be inefficient if you write multiple small chunks (on the wire, this
270disadvantage is usually avoided by your kernel's nagle algorithm, see
271C<no_delay>, but this option can save costly syscalls).
272
273When enabled, then writes will always be queued till the next event loop
274iteration. This is efficient when you do many small writes per iteration,
275but less efficient when you do a single write only per iteration (or when
276the write buffer often is full). It also increases write latency.
277
278=item no_delay => <boolean>
279
280When doing small writes on sockets, your operating system kernel might
281wait a bit for more data before actually sending it out. This is called
282the Nagle algorithm, and usually it is beneficial.
283
284In some situations you want as low a delay as possible, which can be
285accomplishd by setting this option to a true value.
286
287The default is your opertaing system's default behaviour (most likely
288enabled), this option explicitly enables or disables it, if possible.
289
290=item keepalive => <boolean>
291
292Enables (default disable) the SO_KEEPALIVE option on the stream socket:
293normally, TCP connections have no time-out once established, so TCP
294conenctions, once established, can stay alive forever even when the other
295side has long gone. TCP keepalives are a cheap way to take down long-lived
296TCP connections whent he other side becomes unreachable. While the default
297is OS-dependent, TCP keepalives usually kick in after around two hours,
298and, if the other side doesn't reply, take down the TCP connection some 10
299to 15 minutes later.
300
301It is harmless to specify this option for file handles that do not support
302keepalives, and enabling it on connections that are potentially long-lived
303is usually a good idea.
304
305=item oobinline => <boolean>
306
307BSD majorly fucked up the implementation of TCP urgent data. The result
308is that almost no OS implements TCP according to the specs, and every OS
309implements it slightly differently.
310
311If you want to handle TCP urgent data, then setting this flag (the default
312is enabled) gives you the most portable way of getting urgent data, by
313putting it into the stream.
314
315Since BSD emulation of OOB data on top of TCP's urgent data can have
316security implications, AnyEvent::Handle sets this flag automatically
317unless explicitly specified.
318
133=item read_size => <bytes> 319=item read_size => <bytes>
134 320
135The default read block size (the amount of bytes this module will try to read 321The default read block size (the amount of bytes this module will
136on each [loop iteration). Default: C<4096>. 322try to read during each loop iteration, which affects memory
323requirements). Default: C<8192>.
137 324
138=item low_water_mark => <bytes> 325=item low_water_mark => <bytes>
139 326
140Sets the amount of bytes (default: C<0>) that make up an "empty" write 327Sets the amount of bytes (default: C<0>) that make up an "empty" write
141buffer: If the write reaches this size or gets even samller it is 328buffer: If the write reaches this size or gets even samller it is
142considered empty. 329considered empty.
143 330
331Sometimes it can be beneficial (for performance reasons) to add data to
332the write buffer before it is fully drained, but this is a rare case, as
333the operating system kernel usually buffers data as well, so the default
334is good in almost all cases.
335
336=item linger => <seconds>
337
338If non-zero (default: C<3600>), then the destructor of the
339AnyEvent::Handle object will check whether there is still outstanding
340write data and will install a watcher that will write this data to the
341socket. No errors will be reported (this mostly matches how the operating
342system treats outstanding data at socket close time).
343
344This will not work for partial TLS data that could not be encoded
345yet. This data will be lost. Calling the C<stoptls> method in time might
346help.
347
348=item peername => $string
349
350A string used to identify the remote site - usually the DNS hostname
351(I<not> IDN!) used to create the connection, rarely the IP address.
352
353Apart from being useful in error messages, this string is also used in TLS
354peername verification (see C<verify_peername> in L<AnyEvent::TLS>). This
355verification will be skipped when C<peername> is not specified or
356C<undef>.
357
144=item tls => "accept" | "connect" | Net::SSLeay::SSL object 358=item tls => "accept" | "connect" | Net::SSLeay::SSL object
145 359
146When this parameter is given, it enables TLS (SSL) mode, that means it 360When this parameter is given, it enables TLS (SSL) mode, that means
147will start making tls handshake and will transparently encrypt/decrypt 361AnyEvent will start a TLS handshake as soon as the conenction has been
148data. 362established and will transparently encrypt/decrypt data afterwards.
363
364All TLS protocol errors will be signalled as C<EPROTO>, with an
365appropriate error message.
149 366
150TLS mode requires Net::SSLeay to be installed (it will be loaded 367TLS mode requires Net::SSLeay to be installed (it will be loaded
151automatically when you try to create a TLS handle). 368automatically when you try to create a TLS handle): this module doesn't
369have a dependency on that module, so if your module requires it, you have
370to add the dependency yourself.
152 371
153For the TLS server side, use C<accept>, and for the TLS client side of a 372Unlike TCP, TLS has a server and client side: for the TLS server side, use
154connection, use C<connect> mode. 373C<accept>, and for the TLS client side of a connection, use C<connect>
374mode.
155 375
156You can also provide your own TLS connection object, but you have 376You can also provide your own TLS connection object, but you have
157to make sure that you call either C<Net::SSLeay::set_connect_state> 377to make sure that you call either C<Net::SSLeay::set_connect_state>
158or C<Net::SSLeay::set_accept_state> on it before you pass it to 378or C<Net::SSLeay::set_accept_state> on it before you pass it to
159AnyEvent::Handle. 379AnyEvent::Handle. Also, this module will take ownership of this connection
380object.
160 381
382At some future point, AnyEvent::Handle might switch to another TLS
383implementation, then the option to use your own session object will go
384away.
385
386B<IMPORTANT:> since Net::SSLeay "objects" are really only integers,
387passing in the wrong integer will lead to certain crash. This most often
388happens when one uses a stylish C<< tls => 1 >> and is surprised about the
389segmentation fault.
390
161See the C<starttls> method if you need to start TLs negotiation later. 391See the C<< ->starttls >> method for when need to start TLS negotiation later.
162 392
163=item tls_ctx => $ssl_ctx 393=item tls_ctx => $anyevent_tls
164 394
165Use the given Net::SSLeay::CTX object to create the new TLS connection 395Use the given C<AnyEvent::TLS> object to create the new TLS connection
166(unless a connection object was specified directly). If this parameter is 396(unless a connection object was specified directly). If this parameter is
167missing, then AnyEvent::Handle will use C<AnyEvent::Handle::TLS_CTX>. 397missing, then AnyEvent::Handle will use C<AnyEvent::Handle::TLS_CTX>.
168 398
399Instead of an object, you can also specify a hash reference with C<< key
400=> value >> pairs. Those will be passed to L<AnyEvent::TLS> to create a
401new TLS context object.
402
403=item on_starttls => $cb->($handle, $success[, $error_message])
404
405This callback will be invoked when the TLS/SSL handshake has finished. If
406C<$success> is true, then the TLS handshake succeeded, otherwise it failed
407(C<on_stoptls> will not be called in this case).
408
409The session in C<< $handle->{tls} >> can still be examined in this
410callback, even when the handshake was not successful.
411
412TLS handshake failures will not cause C<on_error> to be invoked when this
413callback is in effect, instead, the error message will be passed to C<on_starttls>.
414
415Without this callback, handshake failures lead to C<on_error> being
416called, as normal.
417
418Note that you cannot call C<starttls> right again in this callback. If you
419need to do that, start an zero-second timer instead whose callback can
420then call C<< ->starttls >> again.
421
422=item on_stoptls => $cb->($handle)
423
424When a SSLv3/TLS shutdown/close notify/EOF is detected and this callback is
425set, then it will be invoked after freeing the TLS session. If it is not,
426then a TLS shutdown condition will be treated like a normal EOF condition
427on the handle.
428
429The session in C<< $handle->{tls} >> can still be examined in this
430callback.
431
432This callback will only be called on TLS shutdowns, not when the
433underlying handle signals EOF.
434
435=item json => JSON or JSON::XS object
436
437This is the json coder object used by the C<json> read and write types.
438
439If you don't supply it, then AnyEvent::Handle will create and use a
440suitable one (on demand), which will write and expect UTF-8 encoded JSON
441texts.
442
443Note that you are responsible to depend on the JSON module if you want to
444use this functionality, as AnyEvent does not have a dependency itself.
445
169=back 446=back
170 447
171=cut 448=cut
172 449
173sub new { 450sub new {
174 my $class = shift; 451 my $class = shift;
175
176 my $self = bless { @_ }, $class; 452 my $self = bless { @_ }, $class;
177 453
178 $self->{fh} or Carp::croak "mandatory argument fh is missing"; 454 if ($self->{fh}) {
455 $self->_start;
456 return unless $self->{fh}; # could be gone by now
457
458 } elsif ($self->{connect}) {
459 require AnyEvent::Socket;
460
461 $self->{peername} = $self->{connect}[0]
462 unless exists $self->{peername};
463
464 $self->{_skip_drain_rbuf} = 1;
465
466 {
467 Scalar::Util::weaken (my $self = $self);
468
469 $self->{_connect} =
470 AnyEvent::Socket::tcp_connect (
471 $self->{connect}[0],
472 $self->{connect}[1],
473 sub {
474 my ($fh, $host, $port, $retry) = @_;
475
476 if ($fh) {
477 $self->{fh} = $fh;
478
479 delete $self->{_skip_drain_rbuf};
480 $self->_start;
481
482 $self->{on_connect}
483 and $self->{on_connect}($self, $host, $port, sub {
484 delete @$self{qw(fh _tw _rtw _wtw _ww _rw _eof _queue rbuf _wbuf tls _tls_rbuf _tls_wbuf)};
485 $self->{_skip_drain_rbuf} = 1;
486 &$retry;
487 });
488
489 } else {
490 if ($self->{on_connect_error}) {
491 $self->{on_connect_error}($self, "$!");
492 $self->destroy;
493 } else {
494 $self->_error ($!, 1);
495 }
496 }
497 },
498 sub {
499 local $self->{fh} = $_[0];
500
501 $self->{on_prepare}
502 ? $self->{on_prepare}->($self)
503 : ()
504 }
505 );
506 }
507
508 } else {
509 Carp::croak "AnyEvent::Handle: either an existing fh or the connect parameter must be specified";
510 }
511
512 $self
513}
514
515sub _start {
516 my ($self) = @_;
179 517
180 AnyEvent::Util::fh_nonblocking $self->{fh}, 1; 518 AnyEvent::Util::fh_nonblocking $self->{fh}, 1;
181 519
182 if ($self->{tls}) { 520 $self->{_activity} =
183 require Net::SSLeay; 521 $self->{_ractivity} =
522 $self->{_wactivity} = AE::now;
523
524 $self->timeout (delete $self->{timeout} ) if $self->{timeout};
525 $self->rtimeout (delete $self->{rtimeout} ) if $self->{rtimeout};
526 $self->wtimeout (delete $self->{wtimeout} ) if $self->{wtimeout};
527
528 $self->no_delay (delete $self->{no_delay} ) if exists $self->{no_delay} && $self->{no_delay};
529 $self->keepalive (delete $self->{keepalive}) if exists $self->{keepalive} && $self->{keepalive};
530
531 $self->oobinline (exists $self->{oobinline} ? delete $self->{oobinline} : 1);
532
184 $self->starttls (delete $self->{tls}, delete $self->{tls_ctx}); 533 $self->starttls (delete $self->{tls}, delete $self->{tls_ctx})
185 } 534 if $self->{tls};
186 535
187 $self->on_eof (delete $self->{on_eof} ) if $self->{on_eof};
188 $self->on_error (delete $self->{on_error}) if $self->{on_error};
189 $self->on_drain (delete $self->{on_drain}) if $self->{on_drain}; 536 $self->on_drain (delete $self->{on_drain}) if $self->{on_drain};
190 $self->on_read (delete $self->{on_read} ) if $self->{on_read};
191 537
192 $self->start_read; 538 $self->start_read
539 if $self->{on_read} || @{ $self->{_queue} };
193 540
194 $self 541 $self->_drain_wbuf;
195} 542}
196 543
197sub _shutdown {
198 my ($self) = @_;
199
200 delete $self->{rw};
201 delete $self->{ww};
202 delete $self->{fh};
203}
204
205sub error { 544sub _error {
206 my ($self) = @_; 545 my ($self, $errno, $fatal, $message) = @_;
207 546
208 { 547 $! = $errno;
209 local $!; 548 $message ||= "$!";
210 $self->_shutdown;
211 }
212 549
213 if ($self->{on_error}) { 550 if ($self->{on_error}) {
214 $self->{on_error}($self); 551 $self->{on_error}($self, $fatal, $message);
215 } else { 552 $self->destroy if $fatal;
553 } elsif ($self->{fh}) {
554 $self->destroy;
216 die "AnyEvent::Handle uncaught fatal error: $!"; 555 Carp::croak "AnyEvent::Handle uncaught error: $message";
217 } 556 }
218} 557}
219 558
220=item $fh = $handle->fh 559=item $fh = $handle->fh
221 560
222This method returns the file handle of the L<AnyEvent::Handle> object. 561This method returns the file handle used to create the L<AnyEvent::Handle> object.
223 562
224=cut 563=cut
225 564
226sub fh { $_[0]->{fh} } 565sub fh { $_[0]{fh} }
227 566
228=item $handle->on_error ($cb) 567=item $handle->on_error ($cb)
229 568
230Replace the current C<on_error> callback (see the C<on_error> constructor argument). 569Replace the current C<on_error> callback (see the C<on_error> constructor argument).
231 570
243 582
244sub on_eof { 583sub on_eof {
245 $_[0]{on_eof} = $_[1]; 584 $_[0]{on_eof} = $_[1];
246} 585}
247 586
587=item $handle->on_timeout ($cb)
588
589=item $handle->on_rtimeout ($cb)
590
591=item $handle->on_wtimeout ($cb)
592
593Replace the current C<on_timeout>, C<on_rtimeout> or C<on_wtimeout>
594callback, or disables the callback (but not the timeout) if C<$cb> =
595C<undef>. See the C<timeout> constructor argument and method.
596
597=cut
598
599# see below
600
601=item $handle->autocork ($boolean)
602
603Enables or disables the current autocork behaviour (see C<autocork>
604constructor argument). Changes will only take effect on the next write.
605
606=cut
607
608sub autocork {
609 $_[0]{autocork} = $_[1];
610}
611
612=item $handle->no_delay ($boolean)
613
614Enables or disables the C<no_delay> setting (see constructor argument of
615the same name for details).
616
617=cut
618
619sub no_delay {
620 $_[0]{no_delay} = $_[1];
621
622 eval {
623 local $SIG{__DIE__};
624 setsockopt $_[0]{fh}, Socket::IPPROTO_TCP (), Socket::TCP_NODELAY (), int $_[1]
625 if $_[0]{fh};
626 };
627}
628
629=item $handle->keepalive ($boolean)
630
631Enables or disables the C<keepalive> setting (see constructor argument of
632the same name for details).
633
634=cut
635
636sub keepalive {
637 $_[0]{keepalive} = $_[1];
638
639 eval {
640 local $SIG{__DIE__};
641 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
642 if $_[0]{fh};
643 };
644}
645
646=item $handle->oobinline ($boolean)
647
648Enables or disables the C<oobinline> setting (see constructor argument of
649the same name for details).
650
651=cut
652
653sub oobinline {
654 $_[0]{oobinline} = $_[1];
655
656 eval {
657 local $SIG{__DIE__};
658 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_OOBINLINE (), int $_[1]
659 if $_[0]{fh};
660 };
661}
662
663=item $handle->keepalive ($boolean)
664
665Enables or disables the C<keepalive> setting (see constructor argument of
666the same name for details).
667
668=cut
669
670sub keepalive {
671 $_[0]{keepalive} = $_[1];
672
673 eval {
674 local $SIG{__DIE__};
675 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
676 if $_[0]{fh};
677 };
678}
679
680=item $handle->on_starttls ($cb)
681
682Replace the current C<on_starttls> callback (see the C<on_starttls> constructor argument).
683
684=cut
685
686sub on_starttls {
687 $_[0]{on_starttls} = $_[1];
688}
689
690=item $handle->on_stoptls ($cb)
691
692Replace the current C<on_stoptls> callback (see the C<on_stoptls> constructor argument).
693
694=cut
695
696sub on_starttls {
697 $_[0]{on_stoptls} = $_[1];
698}
699
700=item $handle->rbuf_max ($max_octets)
701
702Configures the C<rbuf_max> setting (C<undef> disables it).
703
704=cut
705
706sub rbuf_max {
707 $_[0]{rbuf_max} = $_[1];
708}
709
710#############################################################################
711
712=item $handle->timeout ($seconds)
713
714=item $handle->rtimeout ($seconds)
715
716=item $handle->wtimeout ($seconds)
717
718Configures (or disables) the inactivity timeout.
719
720=item $handle->timeout_reset
721
722=item $handle->rtimeout_reset
723
724=item $handle->wtimeout_reset
725
726Reset the activity timeout, as if data was received or sent.
727
728These methods are cheap to call.
729
730=cut
731
732for my $dir ("", "r", "w") {
733 my $timeout = "${dir}timeout";
734 my $tw = "_${dir}tw";
735 my $on_timeout = "on_${dir}timeout";
736 my $activity = "_${dir}activity";
737 my $cb;
738
739 *$on_timeout = sub {
740 $_[0]{$on_timeout} = $_[1];
741 };
742
743 *$timeout = sub {
744 my ($self, $new_value) = @_;
745
746 $self->{$timeout} = $new_value;
747 delete $self->{$tw}; &$cb;
748 };
749
750 *{"${dir}timeout_reset"} = sub {
751 $_[0]{$activity} = AE::now;
752 };
753
754 # main workhorse:
755 # reset the timeout watcher, as neccessary
756 # also check for time-outs
757 $cb = sub {
758 my ($self) = @_;
759
760 if ($self->{$timeout} && $self->{fh}) {
761 my $NOW = AE::now;
762
763 # when would the timeout trigger?
764 my $after = $self->{$activity} + $self->{$timeout} - $NOW;
765
766 # now or in the past already?
767 if ($after <= 0) {
768 $self->{$activity} = $NOW;
769
770 if ($self->{$on_timeout}) {
771 $self->{$on_timeout}($self);
772 } else {
773 $self->_error (Errno::ETIMEDOUT);
774 }
775
776 # callback could have changed timeout value, optimise
777 return unless $self->{$timeout};
778
779 # calculate new after
780 $after = $self->{$timeout};
781 }
782
783 Scalar::Util::weaken $self;
784 return unless $self; # ->error could have destroyed $self
785
786 $self->{$tw} ||= AE::timer $after, 0, sub {
787 delete $self->{$tw};
788 $cb->($self);
789 };
790 } else {
791 delete $self->{$tw};
792 }
793 }
794}
795
248############################################################################# 796#############################################################################
249 797
250=back 798=back
251 799
252=head2 WRITE QUEUE 800=head2 WRITE QUEUE
273 my ($self, $cb) = @_; 821 my ($self, $cb) = @_;
274 822
275 $self->{on_drain} = $cb; 823 $self->{on_drain} = $cb;
276 824
277 $cb->($self) 825 $cb->($self)
278 if $cb && $self->{low_water_mark} >= length $self->{wbuf}; 826 if $cb && $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf});
279} 827}
280 828
281=item $handle->push_write ($data) 829=item $handle->push_write ($data)
282 830
283Queues the given scalar to be written. You can push as much data as you 831Queues the given scalar to be written. You can push as much data as you
287=cut 835=cut
288 836
289sub _drain_wbuf { 837sub _drain_wbuf {
290 my ($self) = @_; 838 my ($self) = @_;
291 839
292 unless ($self->{ww}) { 840 if (!$self->{_ww} && length $self->{wbuf}) {
841
293 Scalar::Util::weaken $self; 842 Scalar::Util::weaken $self;
843
294 my $cb = sub { 844 my $cb = sub {
295 my $len = syswrite $self->{fh}, $self->{wbuf}; 845 my $len = syswrite $self->{fh}, $self->{wbuf};
296 846
297 if ($len > 0) { 847 if (defined $len) {
298 substr $self->{wbuf}, 0, $len, ""; 848 substr $self->{wbuf}, 0, $len, "";
299 849
850 $self->{_activity} = $self->{_wactivity} = AE::now;
851
300 $self->{on_drain}($self) 852 $self->{on_drain}($self)
301 if $self->{low_water_mark} >= length $self->{wbuf} 853 if $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf})
302 && $self->{on_drain}; 854 && $self->{on_drain};
303 855
304 delete $self->{ww} unless length $self->{wbuf}; 856 delete $self->{_ww} unless length $self->{wbuf};
305 } elsif ($! != EAGAIN && $! != EINTR) { 857 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
306 $self->error; 858 $self->_error ($!, 1);
307 } 859 }
308 }; 860 };
309 861
310 $self->{ww} = AnyEvent->io (fh => $self->{fh}, poll => "w", cb => $cb); 862 # try to write data immediately
863 $cb->() unless $self->{autocork};
311 864
312 $cb->($self); 865 # if still data left in wbuf, we need to poll
866 $self->{_ww} = AE::io $self->{fh}, 1, $cb
867 if length $self->{wbuf};
313 }; 868 };
869}
870
871our %WH;
872
873sub register_write_type($$) {
874 $WH{$_[0]} = $_[1];
314} 875}
315 876
316sub push_write { 877sub push_write {
317 my $self = shift; 878 my $self = shift;
318 879
880 if (@_ > 1) {
881 my $type = shift;
882
883 @_ = ($WH{$type} or Carp::croak "unsupported type passed to AnyEvent::Handle::push_write")
884 ->($self, @_);
885 }
886
319 if ($self->{filter_w}) { 887 if ($self->{tls}) {
320 $self->{filter_w}->($self, \$_[0]); 888 $self->{_tls_wbuf} .= $_[0];
889 &_dotls ($self) if $self->{fh};
321 } else { 890 } else {
322 $self->{wbuf} .= $_[0]; 891 $self->{wbuf} .= $_[0];
323 $self->_drain_wbuf; 892 $self->_drain_wbuf if $self->{fh};
324 } 893 }
325} 894}
895
896=item $handle->push_write (type => @args)
897
898Instead of formatting your data yourself, you can also let this module do
899the job by specifying a type and type-specific arguments.
900
901Predefined types are (if you have ideas for additional types, feel free to
902drop by and tell us):
903
904=over 4
905
906=item netstring => $string
907
908Formats the given value as netstring
909(http://cr.yp.to/proto/netstrings.txt, this is not a recommendation to use them).
910
911=cut
912
913register_write_type netstring => sub {
914 my ($self, $string) = @_;
915
916 (length $string) . ":$string,"
917};
918
919=item packstring => $format, $data
920
921An octet string prefixed with an encoded length. The encoding C<$format>
922uses the same format as a Perl C<pack> format, but must specify a single
923integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
924optional C<!>, C<< < >> or C<< > >> modifier).
925
926=cut
927
928register_write_type packstring => sub {
929 my ($self, $format, $string) = @_;
930
931 pack "$format/a*", $string
932};
933
934=item json => $array_or_hashref
935
936Encodes the given hash or array reference into a JSON object. Unless you
937provide your own JSON object, this means it will be encoded to JSON text
938in UTF-8.
939
940JSON objects (and arrays) are self-delimiting, so you can write JSON at
941one end of a handle and read them at the other end without using any
942additional framing.
943
944The generated JSON text is guaranteed not to contain any newlines: While
945this module doesn't need delimiters after or between JSON texts to be
946able to read them, many other languages depend on that.
947
948A simple RPC protocol that interoperates easily with others is to send
949JSON arrays (or objects, although arrays are usually the better choice as
950they mimic how function argument passing works) and a newline after each
951JSON text:
952
953 $handle->push_write (json => ["method", "arg1", "arg2"]); # whatever
954 $handle->push_write ("\012");
955
956An AnyEvent::Handle receiver would simply use the C<json> read type and
957rely on the fact that the newline will be skipped as leading whitespace:
958
959 $handle->push_read (json => sub { my $array = $_[1]; ... });
960
961Other languages could read single lines terminated by a newline and pass
962this line into their JSON decoder of choice.
963
964=cut
965
966sub json_coder() {
967 eval { require JSON::XS; JSON::XS->new->utf8 }
968 || do { require JSON; JSON->new->utf8 }
969}
970
971register_write_type json => sub {
972 my ($self, $ref) = @_;
973
974 my $json = $self->{json} ||= json_coder;
975
976 $json->encode ($ref)
977};
978
979=item storable => $reference
980
981Freezes the given reference using L<Storable> and writes it to the
982handle. Uses the C<nfreeze> format.
983
984=cut
985
986register_write_type storable => sub {
987 my ($self, $ref) = @_;
988
989 require Storable;
990
991 pack "w/a*", Storable::nfreeze ($ref)
992};
993
994=back
995
996=item $handle->push_shutdown
997
998Sometimes you know you want to close the socket after writing your data
999before it was actually written. One way to do that is to replace your
1000C<on_drain> handler by a callback that shuts down the socket (and set
1001C<low_water_mark> to C<0>). This method is a shorthand for just that, and
1002replaces the C<on_drain> callback with:
1003
1004 sub { shutdown $_[0]{fh}, 1 } # for push_shutdown
1005
1006This simply shuts down the write side and signals an EOF condition to the
1007the peer.
1008
1009You can rely on the normal read queue and C<on_eof> handling
1010afterwards. This is the cleanest way to close a connection.
1011
1012=cut
1013
1014sub push_shutdown {
1015 my ($self) = @_;
1016
1017 delete $self->{low_water_mark};
1018 $self->on_drain (sub { shutdown $_[0]{fh}, 1 });
1019}
1020
1021=item AnyEvent::Handle::register_write_type type => $coderef->($handle, @args)
1022
1023This function (not method) lets you add your own types to C<push_write>.
1024Whenever the given C<type> is used, C<push_write> will invoke the code
1025reference with the handle object and the remaining arguments.
1026
1027The code reference is supposed to return a single octet string that will
1028be appended to the write buffer.
1029
1030Note that this is a function, and all types registered this way will be
1031global, so try to use unique names.
1032
1033=cut
326 1034
327############################################################################# 1035#############################################################################
328 1036
329=back 1037=back
330 1038
337ways, the "simple" way, using only C<on_read> and the "complex" way, using 1045ways, the "simple" way, using only C<on_read> and the "complex" way, using
338a queue. 1046a queue.
339 1047
340In the simple case, you just install an C<on_read> callback and whenever 1048In the simple case, you just install an C<on_read> callback and whenever
341new data arrives, it will be called. You can then remove some data (if 1049new data arrives, it will be called. You can then remove some data (if
342enough is there) from the read buffer (C<< $handle->rbuf >>) if you want 1050enough is there) from the read buffer (C<< $handle->rbuf >>). Or you cna
343or not. 1051leave the data there if you want to accumulate more (e.g. when only a
1052partial message has been received so far).
344 1053
345In the more complex case, you want to queue multiple callbacks. In this 1054In the more complex case, you want to queue multiple callbacks. In this
346case, AnyEvent::Handle will call the first queued callback each time new 1055case, AnyEvent::Handle will call the first queued callback each time new
347data arrives and removes it when it has done its job (see C<push_read>, 1056data arrives (also the first time it is queued) and removes it when it has
348below). 1057done its job (see C<push_read>, below).
349 1058
350This way you can, for example, push three line-reads, followed by reading 1059This way you can, for example, push three line-reads, followed by reading
351a chunk of data, and AnyEvent::Handle will execute them in order. 1060a chunk of data, and AnyEvent::Handle will execute them in order.
352 1061
353Example 1: EPP protocol parser. EPP sends 4 byte length info, followed by 1062Example 1: EPP protocol parser. EPP sends 4 byte length info, followed by
354the specified number of bytes which give an XML datagram. 1063the specified number of bytes which give an XML datagram.
355 1064
356 # in the default state, expect some header bytes 1065 # in the default state, expect some header bytes
357 $handle->on_read (sub { 1066 $handle->on_read (sub {
358 # some data is here, now queue the length-header-read (4 octets) 1067 # some data is here, now queue the length-header-read (4 octets)
359 shift->unshift_read_chunk (4, sub { 1068 shift->unshift_read (chunk => 4, sub {
360 # header arrived, decode 1069 # header arrived, decode
361 my $len = unpack "N", $_[1]; 1070 my $len = unpack "N", $_[1];
362 1071
363 # now read the payload 1072 # now read the payload
364 shift->unshift_read_chunk ($len, sub { 1073 shift->unshift_read (chunk => $len, sub {
365 my $xml = $_[1]; 1074 my $xml = $_[1];
366 # handle xml 1075 # handle xml
367 }); 1076 });
368 }); 1077 });
369 }); 1078 });
370 1079
371Example 2: Implement a client for a protocol that replies either with 1080Example 2: Implement a client for a protocol that replies either with "OK"
372"OK" and another line or "ERROR" for one request, and 64 bytes for the 1081and another line or "ERROR" for the first request that is sent, and 64
373second request. Due tot he availability of a full queue, we can just 1082bytes for the second request. Due to the availability of a queue, we can
374pipeline sending both requests and manipulate the queue as necessary in 1083just pipeline sending both requests and manipulate the queue as necessary
375the callbacks: 1084in the callbacks.
376 1085
377 # request one 1086When the first callback is called and sees an "OK" response, it will
1087C<unshift> another line-read. This line-read will be queued I<before> the
108864-byte chunk callback.
1089
1090 # request one, returns either "OK + extra line" or "ERROR"
378 $handle->push_write ("request 1\015\012"); 1091 $handle->push_write ("request 1\015\012");
379 1092
380 # we expect "ERROR" or "OK" as response, so push a line read 1093 # we expect "ERROR" or "OK" as response, so push a line read
381 $handle->push_read_line (sub { 1094 $handle->push_read (line => sub {
382 # if we got an "OK", we have to _prepend_ another line, 1095 # if we got an "OK", we have to _prepend_ another line,
383 # so it will be read before the second request reads its 64 bytes 1096 # so it will be read before the second request reads its 64 bytes
384 # which are already in the queue when this callback is called 1097 # which are already in the queue when this callback is called
385 # we don't do this in case we got an error 1098 # we don't do this in case we got an error
386 if ($_[1] eq "OK") { 1099 if ($_[1] eq "OK") {
387 $_[0]->unshift_read_line (sub { 1100 $_[0]->unshift_read (line => sub {
388 my $response = $_[1]; 1101 my $response = $_[1];
389 ... 1102 ...
390 }); 1103 });
391 } 1104 }
392 }); 1105 });
393 1106
394 # request two 1107 # request two, simply returns 64 octets
395 $handle->push_write ("request 2\015\012"); 1108 $handle->push_write ("request 2\015\012");
396 1109
397 # simply read 64 bytes, always 1110 # simply read 64 bytes, always
398 $handle->push_read_chunk (64, sub { 1111 $handle->push_read (chunk => 64, sub {
399 my $response = $_[1]; 1112 my $response = $_[1];
400 ... 1113 ...
401 }); 1114 });
402 1115
403=over 4 1116=over 4
404 1117
405=cut 1118=cut
406 1119
407sub _drain_rbuf { 1120sub _drain_rbuf {
408 my ($self) = @_; 1121 my ($self) = @_;
1122
1123 # avoid recursion
1124 return if $self->{_skip_drain_rbuf};
1125 local $self->{_skip_drain_rbuf} = 1;
1126
1127 while () {
1128 # we need to use a separate tls read buffer, as we must not receive data while
1129 # we are draining the buffer, and this can only happen with TLS.
1130 $self->{rbuf} .= delete $self->{_tls_rbuf}
1131 if exists $self->{_tls_rbuf};
1132
1133 my $len = length $self->{rbuf};
1134
1135 if (my $cb = shift @{ $self->{_queue} }) {
1136 unless ($cb->($self)) {
1137 # no progress can be made
1138 # (not enough data and no data forthcoming)
1139 $self->_error (Errno::EPIPE, 1), return
1140 if $self->{_eof};
1141
1142 unshift @{ $self->{_queue} }, $cb;
1143 last;
1144 }
1145 } elsif ($self->{on_read}) {
1146 last unless $len;
1147
1148 $self->{on_read}($self);
1149
1150 if (
1151 $len == length $self->{rbuf} # if no data has been consumed
1152 && !@{ $self->{_queue} } # and the queue is still empty
1153 && $self->{on_read} # but we still have on_read
1154 ) {
1155 # no further data will arrive
1156 # so no progress can be made
1157 $self->_error (Errno::EPIPE, 1), return
1158 if $self->{_eof};
1159
1160 last; # more data might arrive
1161 }
1162 } else {
1163 # read side becomes idle
1164 delete $self->{_rw} unless $self->{tls};
1165 last;
1166 }
1167 }
1168
1169 if ($self->{_eof}) {
1170 $self->{on_eof}
1171 ? $self->{on_eof}($self)
1172 : $self->_error (0, 1, "Unexpected end-of-file");
1173
1174 return;
1175 }
409 1176
410 if ( 1177 if (
411 defined $self->{rbuf_max} 1178 defined $self->{rbuf_max}
412 && $self->{rbuf_max} < length $self->{rbuf} 1179 && $self->{rbuf_max} < length $self->{rbuf}
413 ) { 1180 ) {
414 $! = &Errno::ENOSPC; return $self->error; 1181 $self->_error (Errno::ENOSPC, 1), return;
415 } 1182 }
416 1183
417 return if $self->{in_drain}; 1184 # may need to restart read watcher
418 local $self->{in_drain} = 1; 1185 unless ($self->{_rw}) {
419 1186 $self->start_read
420 while (my $len = length $self->{rbuf}) { 1187 if $self->{on_read} || @{ $self->{_queue} };
421 no strict 'refs';
422 if (my $cb = shift @{ $self->{queue} }) {
423 if (!$cb->($self)) {
424 if ($self->{eof}) {
425 # no progress can be made (not enough data and no data forthcoming)
426 $! = &Errno::EPIPE; return $self->error;
427 }
428
429 unshift @{ $self->{queue} }, $cb;
430 return;
431 }
432 } elsif ($self->{on_read}) {
433 $self->{on_read}($self);
434
435 if (
436 $self->{eof} # if no further data will arrive
437 && $len == length $self->{rbuf} # and no data has been consumed
438 && !@{ $self->{queue} } # and the queue is still empty
439 && $self->{on_read} # and we still want to read data
440 ) {
441 # then no progress can be made
442 $! = &Errno::EPIPE; return $self->error;
443 }
444 } else {
445 # read side becomes idle
446 delete $self->{rw};
447 return;
448 }
449 }
450
451 if ($self->{eof}) {
452 $self->_shutdown;
453 $self->{on_eof}($self)
454 if $self->{on_eof};
455 } 1188 }
456} 1189}
457 1190
458=item $handle->on_read ($cb) 1191=item $handle->on_read ($cb)
459 1192
465 1198
466sub on_read { 1199sub on_read {
467 my ($self, $cb) = @_; 1200 my ($self, $cb) = @_;
468 1201
469 $self->{on_read} = $cb; 1202 $self->{on_read} = $cb;
1203 $self->_drain_rbuf if $cb;
470} 1204}
471 1205
472=item $handle->rbuf 1206=item $handle->rbuf
473 1207
474Returns the read buffer (as a modifiable lvalue). 1208Returns the read buffer (as a modifiable lvalue).
475 1209
476You can access the read buffer directly as the C<< ->{rbuf} >> member, if 1210You can access the read buffer directly as the C<< ->{rbuf} >>
477you want. 1211member, if you want. However, the only operation allowed on the
1212read buffer (apart from looking at it) is removing data from its
1213beginning. Otherwise modifying or appending to it is not allowed and will
1214lead to hard-to-track-down bugs.
478 1215
479NOTE: The read buffer should only be used or modified if the C<on_read>, 1216NOTE: The read buffer should only be used or modified if the C<on_read>,
480C<push_read> or C<unshift_read> methods are used. The other read methods 1217C<push_read> or C<unshift_read> methods are used. The other read methods
481automatically manage the read buffer. 1218automatically manage the read buffer.
482 1219
505interested in (which can be none at all) and return a true value. After returning 1242interested in (which can be none at all) and return a true value. After returning
506true, it will be removed from the queue. 1243true, it will be removed from the queue.
507 1244
508=cut 1245=cut
509 1246
1247our %RH;
1248
1249sub register_read_type($$) {
1250 $RH{$_[0]} = $_[1];
1251}
1252
510sub push_read { 1253sub push_read {
511 my ($self, $cb) = @_; 1254 my $self = shift;
1255 my $cb = pop;
512 1256
1257 if (@_) {
1258 my $type = shift;
1259
1260 $cb = ($RH{$type} or Carp::croak "unsupported type passed to AnyEvent::Handle::push_read")
1261 ->($self, $cb, @_);
1262 }
1263
513 push @{ $self->{queue} }, $cb; 1264 push @{ $self->{_queue} }, $cb;
514 $self->_drain_rbuf; 1265 $self->_drain_rbuf;
515} 1266}
516 1267
517sub unshift_read { 1268sub unshift_read {
518 my ($self, $cb) = @_; 1269 my $self = shift;
1270 my $cb = pop;
519 1271
1272 if (@_) {
1273 my $type = shift;
1274
1275 $cb = ($RH{$type} or Carp::croak "unsupported type passed to AnyEvent::Handle::unshift_read")
1276 ->($self, $cb, @_);
1277 }
1278
520 push @{ $self->{queue} }, $cb; 1279 unshift @{ $self->{_queue} }, $cb;
521 $self->_drain_rbuf; 1280 $self->_drain_rbuf;
522} 1281}
523 1282
524=item $handle->push_read_chunk ($len, $cb->($self, $data)) 1283=item $handle->push_read (type => @args, $cb)
525 1284
526=item $handle->unshift_read_chunk ($len, $cb->($self, $data)) 1285=item $handle->unshift_read (type => @args, $cb)
527 1286
528Append the given callback to the end of the queue (C<push_read_chunk>) or 1287Instead of providing a callback that parses the data itself you can chose
529prepend it (C<unshift_read_chunk>). 1288between a number of predefined parsing formats, for chunks of data, lines
1289etc.
530 1290
531The callback will be called only once C<$len> bytes have been read, and 1291Predefined types are (if you have ideas for additional types, feel free to
532these C<$len> bytes will be passed to the callback. 1292drop by and tell us):
533 1293
534=cut 1294=over 4
535 1295
536sub _read_chunk($$) { 1296=item chunk => $octets, $cb->($handle, $data)
1297
1298Invoke the callback only once C<$octets> bytes have been read. Pass the
1299data read to the callback. The callback will never be called with less
1300data.
1301
1302Example: read 2 bytes.
1303
1304 $handle->push_read (chunk => 2, sub {
1305 warn "yay ", unpack "H*", $_[1];
1306 });
1307
1308=cut
1309
1310register_read_type chunk => sub {
537 my ($self, $len, $cb) = @_; 1311 my ($self, $cb, $len) = @_;
538 1312
539 sub { 1313 sub {
540 $len <= length $_[0]{rbuf} or return; 1314 $len <= length $_[0]{rbuf} or return;
541 $cb->($_[0], substr $_[0]{rbuf}, 0, $len, ""); 1315 $cb->($_[0], substr $_[0]{rbuf}, 0, $len, "");
542 1 1316 1
543 } 1317 }
544} 1318};
545 1319
546sub push_read_chunk { 1320=item line => [$eol, ]$cb->($handle, $line, $eol)
547 $_[0]->push_read (&_read_chunk);
548}
549
550
551sub unshift_read_chunk {
552 $_[0]->unshift_read (&_read_chunk);
553}
554
555=item $handle->push_read_line ([$eol, ]$cb->($self, $line, $eol))
556
557=item $handle->unshift_read_line ([$eol, ]$cb->($self, $line, $eol))
558
559Append the given callback to the end of the queue (C<push_read_line>) or
560prepend it (C<unshift_read_line>).
561 1321
562The callback will be called only once a full line (including the end of 1322The callback will be called only once a full line (including the end of
563line marker, C<$eol>) has been read. This line (excluding the end of line 1323line marker, C<$eol>) has been read. This line (excluding the end of line
564marker) will be passed to the callback as second argument (C<$line>), and 1324marker) will be passed to the callback as second argument (C<$line>), and
565the end of line marker as the third argument (C<$eol>). 1325the end of line marker as the third argument (C<$eol>).
576Partial lines at the end of the stream will never be returned, as they are 1336Partial lines at the end of the stream will never be returned, as they are
577not marked by the end of line marker. 1337not marked by the end of line marker.
578 1338
579=cut 1339=cut
580 1340
581sub _read_line($$) { 1341register_read_type line => sub {
582 my $self = shift; 1342 my ($self, $cb, $eol) = @_;
583 my $cb = pop;
584 my $eol = @_ ? shift : qr|(\015?\012)|;
585 my $pos;
586 1343
1344 if (@_ < 3) {
1345 # this is more than twice as fast as the generic code below
1346 sub {
1347 $_[0]{rbuf} =~ s/^([^\015\012]*)(\015?\012)// or return;
1348
1349 $cb->($_[0], $1, $2);
1350 1
1351 }
1352 } else {
587 $eol = quotemeta $eol unless ref $eol; 1353 $eol = quotemeta $eol unless ref $eol;
588 $eol = qr|^(.*?)($eol)|s; 1354 $eol = qr|^(.*?)($eol)|s;
1355
1356 sub {
1357 $_[0]{rbuf} =~ s/$eol// or return;
1358
1359 $cb->($_[0], $1, $2);
1360 1
1361 }
1362 }
1363};
1364
1365=item regex => $accept[, $reject[, $skip], $cb->($handle, $data)
1366
1367Makes a regex match against the regex object C<$accept> and returns
1368everything up to and including the match.
1369
1370Example: read a single line terminated by '\n'.
1371
1372 $handle->push_read (regex => qr<\n>, sub { ... });
1373
1374If C<$reject> is given and not undef, then it determines when the data is
1375to be rejected: it is matched against the data when the C<$accept> regex
1376does not match and generates an C<EBADMSG> error when it matches. This is
1377useful to quickly reject wrong data (to avoid waiting for a timeout or a
1378receive buffer overflow).
1379
1380Example: expect a single decimal number followed by whitespace, reject
1381anything else (not the use of an anchor).
1382
1383 $handle->push_read (regex => qr<^[0-9]+\s>, qr<[^0-9]>, sub { ... });
1384
1385If C<$skip> is given and not C<undef>, then it will be matched against
1386the receive buffer when neither C<$accept> nor C<$reject> match,
1387and everything preceding and including the match will be accepted
1388unconditionally. This is useful to skip large amounts of data that you
1389know cannot be matched, so that the C<$accept> or C<$reject> regex do not
1390have to start matching from the beginning. This is purely an optimisation
1391and is usually worth only when you expect more than a few kilobytes.
1392
1393Example: expect a http header, which ends at C<\015\012\015\012>. Since we
1394expect the header to be very large (it isn't in practise, but...), we use
1395a skip regex to skip initial portions. The skip regex is tricky in that
1396it only accepts something not ending in either \015 or \012, as these are
1397required for the accept regex.
1398
1399 $handle->push_read (regex =>
1400 qr<\015\012\015\012>,
1401 undef, # no reject
1402 qr<^.*[^\015\012]>,
1403 sub { ... });
1404
1405=cut
1406
1407register_read_type regex => sub {
1408 my ($self, $cb, $accept, $reject, $skip) = @_;
1409
1410 my $data;
1411 my $rbuf = \$self->{rbuf};
589 1412
590 sub { 1413 sub {
591 $_[0]{rbuf} =~ s/$eol// or return; 1414 # accept
1415 if ($$rbuf =~ $accept) {
1416 $data .= substr $$rbuf, 0, $+[0], "";
1417 $cb->($self, $data);
1418 return 1;
1419 }
1420
1421 # reject
1422 if ($reject && $$rbuf =~ $reject) {
1423 $self->_error (Errno::EBADMSG);
1424 }
592 1425
593 $cb->($_[0], $1, $2); 1426 # skip
1427 if ($skip && $$rbuf =~ $skip) {
1428 $data .= substr $$rbuf, 0, $+[0], "";
1429 }
1430
1431 ()
1432 }
1433};
1434
1435=item netstring => $cb->($handle, $string)
1436
1437A netstring (http://cr.yp.to/proto/netstrings.txt, this is not an endorsement).
1438
1439Throws an error with C<$!> set to EBADMSG on format violations.
1440
1441=cut
1442
1443register_read_type netstring => sub {
1444 my ($self, $cb) = @_;
1445
1446 sub {
1447 unless ($_[0]{rbuf} =~ s/^(0|[1-9][0-9]*)://) {
1448 if ($_[0]{rbuf} =~ /[^0-9]/) {
1449 $self->_error (Errno::EBADMSG);
1450 }
1451 return;
1452 }
1453
1454 my $len = $1;
1455
1456 $self->unshift_read (chunk => $len, sub {
1457 my $string = $_[1];
1458 $_[0]->unshift_read (chunk => 1, sub {
1459 if ($_[1] eq ",") {
1460 $cb->($_[0], $string);
1461 } else {
1462 $self->_error (Errno::EBADMSG);
1463 }
1464 });
1465 });
1466
594 1 1467 1
595 } 1468 }
596} 1469};
597 1470
598sub push_read_line { 1471=item packstring => $format, $cb->($handle, $string)
599 $_[0]->push_read (&_read_line);
600}
601 1472
602sub unshift_read_line { 1473An octet string prefixed with an encoded length. The encoding C<$format>
603 $_[0]->unshift_read (&_read_line); 1474uses the same format as a Perl C<pack> format, but must specify a single
604} 1475integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
1476optional C<!>, C<< < >> or C<< > >> modifier).
1477
1478For example, DNS over TCP uses a prefix of C<n> (2 octet network order),
1479EPP uses a prefix of C<N> (4 octtes).
1480
1481Example: read a block of data prefixed by its length in BER-encoded
1482format (very efficient).
1483
1484 $handle->push_read (packstring => "w", sub {
1485 my ($handle, $data) = @_;
1486 });
1487
1488=cut
1489
1490register_read_type packstring => sub {
1491 my ($self, $cb, $format) = @_;
1492
1493 sub {
1494 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1495 defined (my $len = eval { unpack $format, $_[0]{rbuf} })
1496 or return;
1497
1498 $format = length pack $format, $len;
1499
1500 # bypass unshift if we already have the remaining chunk
1501 if ($format + $len <= length $_[0]{rbuf}) {
1502 my $data = substr $_[0]{rbuf}, $format, $len;
1503 substr $_[0]{rbuf}, 0, $format + $len, "";
1504 $cb->($_[0], $data);
1505 } else {
1506 # remove prefix
1507 substr $_[0]{rbuf}, 0, $format, "";
1508
1509 # read remaining chunk
1510 $_[0]->unshift_read (chunk => $len, $cb);
1511 }
1512
1513 1
1514 }
1515};
1516
1517=item json => $cb->($handle, $hash_or_arrayref)
1518
1519Reads a JSON object or array, decodes it and passes it to the
1520callback. When a parse error occurs, an C<EBADMSG> error will be raised.
1521
1522If a C<json> object was passed to the constructor, then that will be used
1523for the final decode, otherwise it will create a JSON coder expecting UTF-8.
1524
1525This read type uses the incremental parser available with JSON version
15262.09 (and JSON::XS version 2.2) and above. You have to provide a
1527dependency on your own: this module will load the JSON module, but
1528AnyEvent does not depend on it itself.
1529
1530Since JSON texts are fully self-delimiting, the C<json> read and write
1531types are an ideal simple RPC protocol: just exchange JSON datagrams. See
1532the C<json> write type description, above, for an actual example.
1533
1534=cut
1535
1536register_read_type json => sub {
1537 my ($self, $cb) = @_;
1538
1539 my $json = $self->{json} ||= json_coder;
1540
1541 my $data;
1542 my $rbuf = \$self->{rbuf};
1543
1544 sub {
1545 my $ref = eval { $json->incr_parse ($self->{rbuf}) };
1546
1547 if ($ref) {
1548 $self->{rbuf} = $json->incr_text;
1549 $json->incr_text = "";
1550 $cb->($self, $ref);
1551
1552 1
1553 } elsif ($@) {
1554 # error case
1555 $json->incr_skip;
1556
1557 $self->{rbuf} = $json->incr_text;
1558 $json->incr_text = "";
1559
1560 $self->_error (Errno::EBADMSG);
1561
1562 ()
1563 } else {
1564 $self->{rbuf} = "";
1565
1566 ()
1567 }
1568 }
1569};
1570
1571=item storable => $cb->($handle, $ref)
1572
1573Deserialises a L<Storable> frozen representation as written by the
1574C<storable> write type (BER-encoded length prefix followed by nfreeze'd
1575data).
1576
1577Raises C<EBADMSG> error if the data could not be decoded.
1578
1579=cut
1580
1581register_read_type storable => sub {
1582 my ($self, $cb) = @_;
1583
1584 require Storable;
1585
1586 sub {
1587 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1588 defined (my $len = eval { unpack "w", $_[0]{rbuf} })
1589 or return;
1590
1591 my $format = length pack "w", $len;
1592
1593 # bypass unshift if we already have the remaining chunk
1594 if ($format + $len <= length $_[0]{rbuf}) {
1595 my $data = substr $_[0]{rbuf}, $format, $len;
1596 substr $_[0]{rbuf}, 0, $format + $len, "";
1597 $cb->($_[0], Storable::thaw ($data));
1598 } else {
1599 # remove prefix
1600 substr $_[0]{rbuf}, 0, $format, "";
1601
1602 # read remaining chunk
1603 $_[0]->unshift_read (chunk => $len, sub {
1604 if (my $ref = eval { Storable::thaw ($_[1]) }) {
1605 $cb->($_[0], $ref);
1606 } else {
1607 $self->_error (Errno::EBADMSG);
1608 }
1609 });
1610 }
1611
1612 1
1613 }
1614};
1615
1616=back
1617
1618=item AnyEvent::Handle::register_read_type type => $coderef->($handle, $cb, @args)
1619
1620This function (not method) lets you add your own types to C<push_read>.
1621
1622Whenever the given C<type> is used, C<push_read> will invoke the code
1623reference with the handle object, the callback and the remaining
1624arguments.
1625
1626The code reference is supposed to return a callback (usually a closure)
1627that works as a plain read callback (see C<< ->push_read ($cb) >>).
1628
1629It should invoke the passed callback when it is done reading (remember to
1630pass C<$handle> as first argument as all other callbacks do that).
1631
1632Note that this is a function, and all types registered this way will be
1633global, so try to use unique names.
1634
1635For examples, see the source of this module (F<perldoc -m AnyEvent::Handle>,
1636search for C<register_read_type>)).
605 1637
606=item $handle->stop_read 1638=item $handle->stop_read
607 1639
608=item $handle->start_read 1640=item $handle->start_read
609 1641
610In rare cases you actually do not want to read anything from the 1642In rare cases you actually do not want to read anything from the
611socket. In this case you can call C<stop_read>. Neither C<on_read> no 1643socket. In this case you can call C<stop_read>. Neither C<on_read> nor
612any queued callbacks will be executed then. To start reading again, call 1644any queued callbacks will be executed then. To start reading again, call
613C<start_read>. 1645C<start_read>.
614 1646
1647Note that AnyEvent::Handle will automatically C<start_read> for you when
1648you change the C<on_read> callback or push/unshift a read callback, and it
1649will automatically C<stop_read> for you when neither C<on_read> is set nor
1650there are any read requests in the queue.
1651
1652These methods will have no effect when in TLS mode (as TLS doesn't support
1653half-duplex connections).
1654
615=cut 1655=cut
616 1656
617sub stop_read { 1657sub stop_read {
618 my ($self) = @_; 1658 my ($self) = @_;
619 1659
620 delete $self->{rw}; 1660 delete $self->{_rw} unless $self->{tls};
621} 1661}
622 1662
623sub start_read { 1663sub start_read {
624 my ($self) = @_; 1664 my ($self) = @_;
625 1665
626 unless ($self->{rw} || $self->{eof}) { 1666 unless ($self->{_rw} || $self->{_eof}) {
627 Scalar::Util::weaken $self; 1667 Scalar::Util::weaken $self;
628 1668
629 $self->{rw} = AnyEvent->io (fh => $self->{fh}, poll => "r", cb => sub { 1669 $self->{_rw} = AE::io $self->{fh}, 0, sub {
630 my $rbuf = $self->{filter_r} ? \my $buf : \$self->{rbuf}; 1670 my $rbuf = \($self->{tls} ? my $buf : $self->{rbuf});
631 my $len = sysread $self->{fh}, $$rbuf, $self->{read_size} || 8192, length $$rbuf; 1671 my $len = sysread $self->{fh}, $$rbuf, $self->{read_size} || 8192, length $$rbuf;
632 1672
633 if ($len > 0) { 1673 if ($len > 0) {
634 $self->{filter_r} 1674 $self->{_activity} = $self->{_ractivity} = AE::now;
635 ? $self->{filter_r}->($self, $rbuf) 1675
1676 if ($self->{tls}) {
1677 Net::SSLeay::BIO_write ($self->{_rbio}, $$rbuf);
1678
1679 &_dotls ($self);
1680 } else {
636 : $self->_drain_rbuf; 1681 $self->_drain_rbuf;
1682 }
637 1683
638 } elsif (defined $len) { 1684 } elsif (defined $len) {
639 delete $self->{rw}; 1685 delete $self->{_rw};
640 $self->{eof} = 1; 1686 $self->{_eof} = 1;
641 $self->_drain_rbuf; 1687 $self->_drain_rbuf;
642 1688
643 } elsif ($! != EAGAIN && $! != EINTR) { 1689 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
644 return $self->error; 1690 return $self->_error ($!, 1);
645 } 1691 }
646 }); 1692 };
647 } 1693 }
648} 1694}
649 1695
1696our $ERROR_SYSCALL;
1697our $ERROR_WANT_READ;
1698
1699sub _tls_error {
1700 my ($self, $err) = @_;
1701
1702 return $self->_error ($!, 1)
1703 if $err == Net::SSLeay::ERROR_SYSCALL ();
1704
1705 my $err =Net::SSLeay::ERR_error_string (Net::SSLeay::ERR_get_error ());
1706
1707 # reduce error string to look less scary
1708 $err =~ s/^error:[0-9a-fA-F]{8}:[^:]+:([^:]+):/\L$1: /;
1709
1710 if ($self->{_on_starttls}) {
1711 (delete $self->{_on_starttls})->($self, undef, $err);
1712 &_freetls;
1713 } else {
1714 &_freetls;
1715 $self->_error (Errno::EPROTO, 1, $err);
1716 }
1717}
1718
1719# poll the write BIO and send the data if applicable
1720# also decode read data if possible
1721# this is basiclaly our TLS state machine
1722# more efficient implementations are possible with openssl,
1723# but not with the buggy and incomplete Net::SSLeay.
650sub _dotls { 1724sub _dotls {
651 my ($self) = @_; 1725 my ($self) = @_;
652 1726
1727 my $tmp;
1728
653 if (length $self->{tls_wbuf}) { 1729 if (length $self->{_tls_wbuf}) {
654 while ((my $len = Net::SSLeay::write ($self->{tls}, $self->{tls_wbuf})) > 0) { 1730 while (($tmp = Net::SSLeay::write ($self->{tls}, $self->{_tls_wbuf})) > 0) {
655 substr $self->{tls_wbuf}, 0, $len, ""; 1731 substr $self->{_tls_wbuf}, 0, $tmp, "";
656 } 1732 }
657 }
658 1733
1734 $tmp = Net::SSLeay::get_error ($self->{tls}, $tmp);
1735 return $self->_tls_error ($tmp)
1736 if $tmp != $ERROR_WANT_READ
1737 && ($tmp != $ERROR_SYSCALL || $!);
1738 }
1739
659 if (defined (my $buf = Net::SSLeay::BIO_read ($self->{tls_wbio}))) { 1740 while (defined ($tmp = Net::SSLeay::read ($self->{tls}))) {
1741 unless (length $tmp) {
1742 $self->{_on_starttls}
1743 and (delete $self->{_on_starttls})->($self, undef, "EOF during handshake"); # ???
1744 &_freetls;
1745
1746 if ($self->{on_stoptls}) {
1747 $self->{on_stoptls}($self);
1748 return;
1749 } else {
1750 # let's treat SSL-eof as we treat normal EOF
1751 delete $self->{_rw};
1752 $self->{_eof} = 1;
1753 }
1754 }
1755
1756 $self->{_tls_rbuf} .= $tmp;
1757 $self->_drain_rbuf;
1758 $self->{tls} or return; # tls session might have gone away in callback
1759 }
1760
1761 $tmp = Net::SSLeay::get_error ($self->{tls}, -1);
1762 return $self->_tls_error ($tmp)
1763 if $tmp != $ERROR_WANT_READ
1764 && ($tmp != $ERROR_SYSCALL || $!);
1765
1766 while (length ($tmp = Net::SSLeay::BIO_read ($self->{_wbio}))) {
660 $self->{wbuf} .= $buf; 1767 $self->{wbuf} .= $tmp;
661 $self->_drain_wbuf; 1768 $self->_drain_wbuf;
662 } 1769 }
663 1770
664 while (defined (my $buf = Net::SSLeay::read ($self->{tls}))) { 1771 $self->{_on_starttls}
665 $self->{rbuf} .= $buf; 1772 and Net::SSLeay::state ($self->{tls}) == Net::SSLeay::ST_OK ()
666 $self->_drain_rbuf; 1773 and (delete $self->{_on_starttls})->($self, 1, "TLS/SSL connection established");
667 }
668
669 my $err = Net::SSLeay::get_error ($self->{tls}, -1);
670
671 if ($err!= Net::SSLeay::ERROR_WANT_READ ()) {
672 if ($err == Net::SSLeay::ERROR_SYSCALL ()) {
673 $self->error;
674 } elsif ($err == Net::SSLeay::ERROR_SSL ()) {
675 $! = &Errno::EIO;
676 $self->error;
677 }
678
679 # all others are fine for our purposes
680 }
681} 1774}
682 1775
683=item $handle->starttls ($tls[, $tls_ctx]) 1776=item $handle->starttls ($tls[, $tls_ctx])
684 1777
685Instead of starting TLS negotiation immediately when the AnyEvent::Handle 1778Instead of starting TLS negotiation immediately when the AnyEvent::Handle
686object is created, you can also do that at a later time by calling 1779object is created, you can also do that at a later time by calling
687C<starttls>. 1780C<starttls>.
688 1781
1782Starting TLS is currently an asynchronous operation - when you push some
1783write data and then call C<< ->starttls >> then TLS negotiation will start
1784immediately, after which the queued write data is then sent.
1785
689The first argument is the same as the C<tls> constructor argument (either 1786The first argument is the same as the C<tls> constructor argument (either
690C<"connect">, C<"accept"> or an existing Net::SSLeay object). 1787C<"connect">, C<"accept"> or an existing Net::SSLeay object).
691 1788
692The second argument is the optional C<Net::SSLeay::CTX> object that is 1789The second argument is the optional C<AnyEvent::TLS> object that is used
693used when AnyEvent::Handle has to create its own TLS connection object. 1790when AnyEvent::Handle has to create its own TLS connection object, or
1791a hash reference with C<< key => value >> pairs that will be used to
1792construct a new context.
694 1793
695=cut 1794The TLS connection object will end up in C<< $handle->{tls} >>, the TLS
1795context in C<< $handle->{tls_ctx} >> after this call and can be used or
1796changed to your liking. Note that the handshake might have already started
1797when this function returns.
696 1798
697# TODO: maybe document... 1799Due to bugs in OpenSSL, it might or might not be possible to do multiple
1800handshakes on the same stream. Best do not attempt to use the stream after
1801stopping TLS.
1802
1803=cut
1804
1805our %TLS_CACHE; #TODO not yet documented, should we?
1806
698sub starttls { 1807sub starttls {
699 my ($self, $ssl, $ctx) = @_; 1808 my ($self, $tls, $ctx) = @_;
700 1809
701 $self->stoptls; 1810 Carp::croak "It is an error to call starttls on an AnyEvent::Handle object while TLS is already active, caught"
1811 if $self->{tls};
702 1812
703 if ($ssl eq "accept") { 1813 $self->{tls} = $tls;
704 $ssl = Net::SSLeay::new ($ctx || TLS_CTX ()); 1814 $self->{tls_ctx} = $ctx if @_ > 2;
705 Net::SSLeay::set_accept_state ($ssl); 1815
706 } elsif ($ssl eq "connect") { 1816 return unless $self->{fh};
707 $ssl = Net::SSLeay::new ($ctx || TLS_CTX ()); 1817
708 Net::SSLeay::set_connect_state ($ssl); 1818 require Net::SSLeay;
1819
1820 $ERROR_SYSCALL = Net::SSLeay::ERROR_SYSCALL ();
1821 $ERROR_WANT_READ = Net::SSLeay::ERROR_WANT_READ ();
1822
1823 $tls = delete $self->{tls};
1824 $ctx = $self->{tls_ctx};
1825
1826 local $Carp::CarpLevel = 1; # skip ourselves when creating a new context or session
1827
1828 if ("HASH" eq ref $ctx) {
1829 require AnyEvent::TLS;
1830
1831 if ($ctx->{cache}) {
1832 my $key = $ctx+0;
1833 $ctx = $TLS_CACHE{$key} ||= new AnyEvent::TLS %$ctx;
1834 } else {
1835 $ctx = new AnyEvent::TLS %$ctx;
1836 }
1837 }
709 } 1838
710 1839 $self->{tls_ctx} = $ctx || TLS_CTX ();
711 $self->{tls} = $ssl; 1840 $self->{tls} = $tls = $self->{tls_ctx}->_get_session ($tls, $self, $self->{peername});
712 1841
713 # basically, this is deep magic (because SSL_read should have the same issues) 1842 # basically, this is deep magic (because SSL_read should have the same issues)
714 # but the openssl maintainers basically said: "trust us, it just works". 1843 # but the openssl maintainers basically said: "trust us, it just works".
715 # (unfortunately, we have to hardcode constants because the abysmally misdesigned 1844 # (unfortunately, we have to hardcode constants because the abysmally misdesigned
716 # and mismaintained ssleay-module doesn't even offer them). 1845 # and mismaintained ssleay-module doesn't even offer them).
1846 # http://www.mail-archive.com/openssl-dev@openssl.org/msg22420.html
1847 #
1848 # in short: this is a mess.
1849 #
1850 # note that we do not try to keep the length constant between writes as we are required to do.
1851 # we assume that most (but not all) of this insanity only applies to non-blocking cases,
1852 # and we drive openssl fully in blocking mode here. Or maybe we don't - openssl seems to
1853 # have identity issues in that area.
717 Net::SSLeay::CTX_set_mode ($self->{tls}, 1854# Net::SSLeay::CTX_set_mode ($ssl,
718 (eval { Net::SSLeay::MODE_ENABLE_PARTIAL_WRITE () } || 1) 1855# (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ENABLE_PARTIAL_WRITE () } || 1)
719 | (eval { Net::SSLeay::MODE_ACCEPT_MOVING_WRITE_BUFFER () } || 2)); 1856# | (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ACCEPT_MOVING_WRITE_BUFFER () } || 2));
1857 Net::SSLeay::CTX_set_mode ($tls, 1|2);
720 1858
721 $self->{tls_rbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ()); 1859 $self->{_rbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
722 $self->{tls_wbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ()); 1860 $self->{_wbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
723 1861
1862 Net::SSLeay::BIO_write ($self->{_rbio}, delete $self->{rbuf});
1863
724 Net::SSLeay::set_bio ($ssl, $self->{tls_rbio}, $self->{tls_wbio}); 1864 Net::SSLeay::set_bio ($tls, $self->{_rbio}, $self->{_wbio});
725 1865
726 $self->{filter_w} = sub { 1866 $self->{_on_starttls} = sub { $_[0]{on_starttls}(@_) }
727 $_[0]{tls_wbuf} .= ${$_[1]}; 1867 if $self->{on_starttls};
728 &_dotls; 1868
729 }; 1869 &_dotls; # need to trigger the initial handshake
730 $self->{filter_r} = sub { 1870 $self->start_read; # make sure we actually do read
731 Net::SSLeay::BIO_write ($_[0]{tls_rbio}, ${$_[1]});
732 &_dotls;
733 };
734} 1871}
735 1872
736=item $handle->stoptls 1873=item $handle->stoptls
737 1874
738Destroys the SSL connection, if any. Partial read or write data will be 1875Shuts down the SSL connection - this makes a proper EOF handshake by
739lost. 1876sending a close notify to the other side, but since OpenSSL doesn't
1877support non-blocking shut downs, it is not guarenteed that you can re-use
1878the stream afterwards.
740 1879
741=cut 1880=cut
742 1881
743sub stoptls { 1882sub stoptls {
744 my ($self) = @_; 1883 my ($self) = @_;
745 1884
746 Net::SSLeay::free (delete $self->{tls}) if $self->{tls}; 1885 if ($self->{tls}) {
747 delete $self->{tls_rbio}; 1886 Net::SSLeay::shutdown ($self->{tls});
748 delete $self->{tls_wbio}; 1887
749 delete $self->{tls_wbuf}; 1888 &_dotls;
750 delete $self->{filter_r}; 1889
751 delete $self->{filter_w}; 1890# # we don't give a shit. no, we do, but we can't. no...#d#
1891# # we, we... have to use openssl :/#d#
1892# &_freetls;#d#
1893 }
1894}
1895
1896sub _freetls {
1897 my ($self) = @_;
1898
1899 return unless $self->{tls};
1900
1901 $self->{tls_ctx}->_put_session (delete $self->{tls})
1902 if $self->{tls} > 0;
1903
1904 delete @$self{qw(_rbio _wbio _tls_wbuf _on_starttls)};
752} 1905}
753 1906
754sub DESTROY { 1907sub DESTROY {
755 my $self = shift; 1908 my ($self) = @_;
756 1909
757 $self->stoptls; 1910 &_freetls;
1911
1912 my $linger = exists $self->{linger} ? $self->{linger} : 3600;
1913
1914 if ($linger && length $self->{wbuf} && $self->{fh}) {
1915 my $fh = delete $self->{fh};
1916 my $wbuf = delete $self->{wbuf};
1917
1918 my @linger;
1919
1920 push @linger, AE::io $fh, 1, sub {
1921 my $len = syswrite $fh, $wbuf, length $wbuf;
1922
1923 if ($len > 0) {
1924 substr $wbuf, 0, $len, "";
1925 } else {
1926 @linger = (); # end
1927 }
1928 };
1929 push @linger, AE::timer $linger, 0, sub {
1930 @linger = ();
1931 };
1932 }
1933}
1934
1935=item $handle->destroy
1936
1937Shuts down the handle object as much as possible - this call ensures that
1938no further callbacks will be invoked and as many resources as possible
1939will be freed. Any method you will call on the handle object after
1940destroying it in this way will be silently ignored (and it will return the
1941empty list).
1942
1943Normally, you can just "forget" any references to an AnyEvent::Handle
1944object and it will simply shut down. This works in fatal error and EOF
1945callbacks, as well as code outside. It does I<NOT> work in a read or write
1946callback, so when you want to destroy the AnyEvent::Handle object from
1947within such an callback. You I<MUST> call C<< ->destroy >> explicitly in
1948that case.
1949
1950Destroying the handle object in this way has the advantage that callbacks
1951will be removed as well, so if those are the only reference holders (as
1952is common), then one doesn't need to do anything special to break any
1953reference cycles.
1954
1955The handle might still linger in the background and write out remaining
1956data, as specified by the C<linger> option, however.
1957
1958=cut
1959
1960sub destroy {
1961 my ($self) = @_;
1962
1963 $self->DESTROY;
1964 %$self = ();
1965 bless $self, "AnyEvent::Handle::destroyed";
1966}
1967
1968sub AnyEvent::Handle::destroyed::AUTOLOAD {
1969 #nop
758} 1970}
759 1971
760=item AnyEvent::Handle::TLS_CTX 1972=item AnyEvent::Handle::TLS_CTX
761 1973
762This function creates and returns the Net::SSLeay::CTX object used by 1974This function creates and returns the AnyEvent::TLS object used by default
763default for TLS mode. 1975for TLS mode.
764 1976
765The context is created like this: 1977The context is created by calling L<AnyEvent::TLS> without any arguments.
766
767 Net::SSLeay::load_error_strings;
768 Net::SSLeay::SSLeay_add_ssl_algorithms;
769 Net::SSLeay::randomize;
770
771 my $CTX = Net::SSLeay::CTX_new;
772
773 Net::SSLeay::CTX_set_options $CTX, Net::SSLeay::OP_ALL
774 1978
775=cut 1979=cut
776 1980
777our $TLS_CTX; 1981our $TLS_CTX;
778 1982
779sub TLS_CTX() { 1983sub TLS_CTX() {
780 $TLS_CTX || do { 1984 $TLS_CTX ||= do {
781 require Net::SSLeay; 1985 require AnyEvent::TLS;
782 1986
783 Net::SSLeay::load_error_strings (); 1987 new AnyEvent::TLS
784 Net::SSLeay::SSLeay_add_ssl_algorithms ();
785 Net::SSLeay::randomize ();
786
787 $TLS_CTX = Net::SSLeay::CTX_new ();
788
789 Net::SSLeay::CTX_set_options ($TLS_CTX, Net::SSLeay::OP_ALL ());
790
791 $TLS_CTX
792 } 1988 }
793} 1989}
794 1990
795=back 1991=back
796 1992
1993
1994=head1 NONFREQUENTLY ASKED QUESTIONS
1995
1996=over 4
1997
1998=item I C<undef> the AnyEvent::Handle reference inside my callback and
1999still get further invocations!
2000
2001That's because AnyEvent::Handle keeps a reference to itself when handling
2002read or write callbacks.
2003
2004It is only safe to "forget" the reference inside EOF or error callbacks,
2005from within all other callbacks, you need to explicitly call the C<<
2006->destroy >> method.
2007
2008=item I get different callback invocations in TLS mode/Why can't I pause
2009reading?
2010
2011Unlike, say, TCP, TLS connections do not consist of two independent
2012communication channels, one for each direction. Or put differently. The
2013read and write directions are not independent of each other: you cannot
2014write data unless you are also prepared to read, and vice versa.
2015
2016This can mean than, in TLS mode, you might get C<on_error> or C<on_eof>
2017callback invocations when you are not expecting any read data - the reason
2018is that AnyEvent::Handle always reads in TLS mode.
2019
2020During the connection, you have to make sure that you always have a
2021non-empty read-queue, or an C<on_read> watcher. At the end of the
2022connection (or when you no longer want to use it) you can call the
2023C<destroy> method.
2024
2025=item How do I read data until the other side closes the connection?
2026
2027If you just want to read your data into a perl scalar, the easiest way
2028to achieve this is by setting an C<on_read> callback that does nothing,
2029clearing the C<on_eof> callback and in the C<on_error> callback, the data
2030will be in C<$_[0]{rbuf}>:
2031
2032 $handle->on_read (sub { });
2033 $handle->on_eof (undef);
2034 $handle->on_error (sub {
2035 my $data = delete $_[0]{rbuf};
2036 });
2037
2038The reason to use C<on_error> is that TCP connections, due to latencies
2039and packets loss, might get closed quite violently with an error, when in
2040fact, all data has been received.
2041
2042It is usually better to use acknowledgements when transferring data,
2043to make sure the other side hasn't just died and you got the data
2044intact. This is also one reason why so many internet protocols have an
2045explicit QUIT command.
2046
2047=item I don't want to destroy the handle too early - how do I wait until
2048all data has been written?
2049
2050After writing your last bits of data, set the C<on_drain> callback
2051and destroy the handle in there - with the default setting of
2052C<low_water_mark> this will be called precisely when all data has been
2053written to the socket:
2054
2055 $handle->push_write (...);
2056 $handle->on_drain (sub {
2057 warn "all data submitted to the kernel\n";
2058 undef $handle;
2059 });
2060
2061If you just want to queue some data and then signal EOF to the other side,
2062consider using C<< ->push_shutdown >> instead.
2063
2064=item I want to contact a TLS/SSL server, I don't care about security.
2065
2066If your TLS server is a pure TLS server (e.g. HTTPS) that only speaks TLS,
2067simply connect to it and then create the AnyEvent::Handle with the C<tls>
2068parameter:
2069
2070 tcp_connect $host, $port, sub {
2071 my ($fh) = @_;
2072
2073 my $handle = new AnyEvent::Handle
2074 fh => $fh,
2075 tls => "connect",
2076 on_error => sub { ... };
2077
2078 $handle->push_write (...);
2079 };
2080
2081=item I want to contact a TLS/SSL server, I do care about security.
2082
2083Then you should additionally enable certificate verification, including
2084peername verification, if the protocol you use supports it (see
2085L<AnyEvent::TLS>, C<verify_peername>).
2086
2087E.g. for HTTPS:
2088
2089 tcp_connect $host, $port, sub {
2090 my ($fh) = @_;
2091
2092 my $handle = new AnyEvent::Handle
2093 fh => $fh,
2094 peername => $host,
2095 tls => "connect",
2096 tls_ctx => { verify => 1, verify_peername => "https" },
2097 ...
2098
2099Note that you must specify the hostname you connected to (or whatever
2100"peername" the protocol needs) as the C<peername> argument, otherwise no
2101peername verification will be done.
2102
2103The above will use the system-dependent default set of trusted CA
2104certificates. If you want to check against a specific CA, add the
2105C<ca_file> (or C<ca_cert>) arguments to C<tls_ctx>:
2106
2107 tls_ctx => {
2108 verify => 1,
2109 verify_peername => "https",
2110 ca_file => "my-ca-cert.pem",
2111 },
2112
2113=item I want to create a TLS/SSL server, how do I do that?
2114
2115Well, you first need to get a server certificate and key. You have
2116three options: a) ask a CA (buy one, use cacert.org etc.) b) create a
2117self-signed certificate (cheap. check the search engine of your choice,
2118there are many tutorials on the net) or c) make your own CA (tinyca2 is a
2119nice program for that purpose).
2120
2121Then create a file with your private key (in PEM format, see
2122L<AnyEvent::TLS>), followed by the certificate (also in PEM format). The
2123file should then look like this:
2124
2125 -----BEGIN RSA PRIVATE KEY-----
2126 ...header data
2127 ... lots of base64'y-stuff
2128 -----END RSA PRIVATE KEY-----
2129
2130 -----BEGIN CERTIFICATE-----
2131 ... lots of base64'y-stuff
2132 -----END CERTIFICATE-----
2133
2134The important bits are the "PRIVATE KEY" and "CERTIFICATE" parts. Then
2135specify this file as C<cert_file>:
2136
2137 tcp_server undef, $port, sub {
2138 my ($fh) = @_;
2139
2140 my $handle = new AnyEvent::Handle
2141 fh => $fh,
2142 tls => "accept",
2143 tls_ctx => { cert_file => "my-server-keycert.pem" },
2144 ...
2145
2146When you have intermediate CA certificates that your clients might not
2147know about, just append them to the C<cert_file>.
2148
2149=back
2150
2151
2152=head1 SUBCLASSING AnyEvent::Handle
2153
2154In many cases, you might want to subclass AnyEvent::Handle.
2155
2156To make this easier, a given version of AnyEvent::Handle uses these
2157conventions:
2158
2159=over 4
2160
2161=item * all constructor arguments become object members.
2162
2163At least initially, when you pass a C<tls>-argument to the constructor it
2164will end up in C<< $handle->{tls} >>. Those members might be changed or
2165mutated later on (for example C<tls> will hold the TLS connection object).
2166
2167=item * other object member names are prefixed with an C<_>.
2168
2169All object members not explicitly documented (internal use) are prefixed
2170with an underscore character, so the remaining non-C<_>-namespace is free
2171for use for subclasses.
2172
2173=item * all members not documented here and not prefixed with an underscore
2174are free to use in subclasses.
2175
2176Of course, new versions of AnyEvent::Handle may introduce more "public"
2177member variables, but thats just life, at least it is documented.
2178
2179=back
2180
797=head1 AUTHOR 2181=head1 AUTHOR
798 2182
799Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>. 2183Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>.
800 2184
801=cut 2185=cut

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines