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 1.15 by root, Sat May 17 21:34:15 2008 UTC vs.
Revision 1.200 by root, Tue Oct 12 06:47:54 2010 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 filehandles via AnyEvent 3AnyEvent::Handle - non-blocking I/O on streaming 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 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 36stream-based filehandles (sockets, pipes, and other stream things).
51on sockets see L<AnyEvent::Util>.
52 37
38The L<AnyEvent::Intro> tutorial contains some well-documented
39AnyEvent::Handle examples.
40
53In the following, when the documentation refers to of "bytes" then this 41In the following, where the documentation refers to "bytes", it means
54means characters. As sysread and syswrite are used for all I/O, their 42characters. 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.
44
45At the very minimum, you should specify C<fh> or C<connect>, and the
46C<on_error> callback.
56 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 = $AnyEvent::VERSION;
64
65sub _load_func($) {
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
60=head1 METHODS 80=head1 METHODS
61 81
62=over 4 82=over 4
63 83
64=item B<new (%args)> 84=item $handle = B<new> AnyEvent::Handle fh => $filehandle, key => value...
65 85
66The constructor supports these arguments (all as key => value pairs). 86The constructor supports these arguments (all as C<< key => value >> pairs).
67 87
68=over 4 88=over 4
69 89
70=item fh => $filehandle [MANDATORY] 90=item fh => $filehandle [C<fh> or C<connect> MANDATORY]
71 91
72The filehandle this L<AnyEvent::Handle> object will operate on. 92The filehandle this L<AnyEvent::Handle> object will operate on.
73
74NOTE: The filehandle will be set to non-blocking (using 93NOTE: The filehandle will be set to non-blocking mode (using
75AnyEvent::Util::fh_nonblocking). 94C<AnyEvent::Util::fh_nonblocking>) by the constructor and needs to stay in
95that mode.
76 96
77=item on_eof => $cb->($self) [MANDATORY] 97=item connect => [$host, $service] [C<fh> or C<connect> MANDATORY]
78 98
79Set the callback to be called on EOF. 99Try to connect to the specified host and service (port), using
100C<AnyEvent::Socket::tcp_connect>. The C<$host> additionally becomes the
101default C<peername>.
80 102
103You have to specify either this parameter, or C<fh>, above.
104
105It is possible to push requests on the read and write queues, and modify
106properties of the stream, even while AnyEvent::Handle is connecting.
107
108When this parameter is specified, then the C<on_prepare>,
109C<on_connect_error> and C<on_connect> callbacks will be called under the
110appropriate circumstances:
111
112=over 4
113
81=item on_error => $cb->($self) 114=item on_prepare => $cb->($handle)
82 115
116This (rarely used) callback is called before a new connection is
117attempted, but after the file handle has been created. It could be used to
118prepare the file handle with parameters required for the actual connect
119(as opposed to settings that can be changed when the connection is already
120established).
121
122The return value of this callback should be the connect timeout value in
123seconds (or C<0>, or C<undef>, or the empty list, to indicate that the
124default timeout is to be used).
125
126=item on_connect => $cb->($handle, $host, $port, $retry->())
127
128This callback is called when a connection has been successfully established.
129
130The peer's numeric host and port (the socket peername) are passed as
131parameters, together with a retry callback.
132
133If, for some reason, the handle is not acceptable, calling C<$retry>
134will continue with the next connection target (in case of multi-homed
135hosts or SRV records there can be multiple connection endpoints). At the
136time it is called the read and write queues, eof status, tls status and
137similar properties of the handle will have been reset.
138
139In most cases, you should ignore the C<$retry> parameter.
140
141=item on_connect_error => $cb->($handle, $message)
142
143This callback is called when the connection could not be
144established. C<$!> will contain the relevant error code, and C<$message> a
145message describing it (usually the same as C<"$!">).
146
147If this callback isn't specified, then C<on_error> will be called with a
148fatal error instead.
149
150=back
151
152=item on_error => $cb->($handle, $fatal, $message)
153
83This is the fatal error callback, that is called when, well, a fatal error 154This is the error callback, which is called when, well, some error
84ocurs, such as not being able to resolve the hostname, failure to connect 155occured, such as not being able to resolve the hostname, failure to
85or a read error. 156connect, or a read error.
86 157
87The object will not be in a usable state when this callback has been 158Some errors are fatal (which is indicated by C<$fatal> being true). On
88called. 159fatal errors the handle object will be destroyed (by a call to C<< ->
160destroy >>) after invoking the error callback (which means you are free to
161examine the handle object). Examples of fatal errors are an EOF condition
162with active (but unsatisifable) read watchers (C<EPIPE>) or I/O errors. In
163cases where the other side can close the connection at will, it is
164often easiest to not report C<EPIPE> errors in this callback.
89 165
166AnyEvent::Handle tries to find an appropriate error code for you to check
167against, but in some cases (TLS errors), this does not work well. It is
168recommended to always output the C<$message> argument in human-readable
169error messages (it's usually the same as C<"$!">).
170
171Non-fatal errors can be retried by returning, but it is recommended
172to simply ignore this parameter and instead abondon the handle object
173when this callback is invoked. Examples of non-fatal errors are timeouts
174C<ETIMEDOUT>) or badly-formatted data (C<EBADMSG>).
175
90On callback entrance, the value of C<$!> contains the operating system 176On entry to the callback, the value of C<$!> contains the operating
91error (or C<ENOSPC> or C<EPIPE>). 177system error code (or C<ENOSPC>, C<EPIPE>, C<ETIMEDOUT>, C<EBADMSG> or
178C<EPROTO>).
92 179
93While not mandatory, it is I<highly> recommended to set this callback, as 180While not mandatory, it is I<highly> recommended to set this callback, as
94you will not be notified of errors otherwise. The default simply calls 181you will not be notified of errors otherwise. The default just calls
95die. 182C<croak>.
96 183
97=item on_read => $cb->($self) 184=item on_read => $cb->($handle)
98 185
99This sets the default read callback, which is called when data arrives 186This sets the default read callback, which is called when data arrives
100and no read request is in the queue. 187and no read request is in the queue (unlike read queue callbacks, this
188callback will only be called when at least one octet of data is in the
189read buffer).
101 190
102To access (and remove data from) the read buffer, use the C<< ->rbuf >> 191To access (and remove data from) the read buffer, use the C<< ->rbuf >>
103method or acces sthe C<$self->{rbuf}> member directly. 192method or access the C<< $handle->{rbuf} >> member directly. Note that you
193must not enlarge or modify the read buffer, you can only remove data at
194the beginning from it.
104 195
196You can also call C<< ->push_read (...) >> or any other function that
197modifies the read queue. Or do both. Or ...
198
105When an EOF condition is detected then AnyEvent::Handle will first try to 199When an EOF condition is detected, AnyEvent::Handle will first try to
106feed all the remaining data to the queued callbacks and C<on_read> before 200feed all the remaining data to the queued callbacks and C<on_read> before
107calling the C<on_eof> callback. If no progress can be made, then a fatal 201calling the C<on_eof> callback. If no progress can be made, then a fatal
108error will be raised (with C<$!> set to C<EPIPE>). 202error will be raised (with C<$!> set to C<EPIPE>).
109 203
204Note that, unlike requests in the read queue, an C<on_read> callback
205doesn't mean you I<require> some data: if there is an EOF and there
206are outstanding read requests then an error will be flagged. With an
207C<on_read> callback, the C<on_eof> callback will be invoked.
208
209=item on_eof => $cb->($handle)
210
211Set the callback to be called when an end-of-file condition is detected,
212i.e. in the case of a socket, when the other side has closed the
213connection cleanly, and there are no outstanding read requests in the
214queue (if there are read requests, then an EOF counts as an unexpected
215connection close and will be flagged as an error).
216
217For sockets, this just means that the other side has stopped sending data,
218you can still try to write data, and, in fact, one can return from the EOF
219callback and continue writing data, as only the read part has been shut
220down.
221
222If an EOF condition has been detected but no C<on_eof> callback has been
223set, then a fatal error will be raised with C<$!> set to <0>.
224
110=item on_drain => $cb->() 225=item on_drain => $cb->($handle)
111 226
112This sets the callback that is called when the write buffer becomes empty 227This sets the callback that is called when the write buffer becomes empty
113(or when the callback is set and the buffer is empty already). 228(or immediately if the buffer is empty already).
114 229
115To append to the write buffer, use the C<< ->push_write >> method. 230To append to the write buffer, use the C<< ->push_write >> method.
231
232This callback is useful when you don't want to put all of your write data
233into the queue at once, for example, when you want to write the contents
234of some file to the socket you might not want to read the whole file into
235memory and push it into the queue, but instead only read more data from
236the file when the write queue becomes empty.
237
238=item timeout => $fractional_seconds
239
240=item rtimeout => $fractional_seconds
241
242=item wtimeout => $fractional_seconds
243
244If non-zero, then these enables an "inactivity" timeout: whenever this
245many seconds pass without a successful read or write on the underlying
246file handle (or a call to C<timeout_reset>), the C<on_timeout> callback
247will be invoked (and if that one is missing, a non-fatal C<ETIMEDOUT>
248error will be raised).
249
250There are three variants of the timeouts that work independently
251of each other, for both read and write, just read, and just write:
252C<timeout>, C<rtimeout> and C<wtimeout>, with corresponding callbacks
253C<on_timeout>, C<on_rtimeout> and C<on_wtimeout>, and reset functions
254C<timeout_reset>, C<rtimeout_reset>, and C<wtimeout_reset>.
255
256Note that timeout processing is active even when you do not have
257any outstanding read or write requests: If you plan to keep the connection
258idle then you should disable the timeout temporarily or ignore the timeout
259in the C<on_timeout> callback, in which case AnyEvent::Handle will simply
260restart the timeout.
261
262Zero (the default) disables this timeout.
263
264=item on_timeout => $cb->($handle)
265
266Called whenever the inactivity timeout passes. If you return from this
267callback, then the timeout will be reset as if some activity had happened,
268so this condition is not fatal in any way.
116 269
117=item rbuf_max => <bytes> 270=item rbuf_max => <bytes>
118 271
119If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>) 272If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>)
120when the read buffer ever (strictly) exceeds this size. This is useful to 273when the read buffer ever (strictly) exceeds this size. This is useful to
121avoid denial-of-service attacks. 274avoid some forms of denial-of-service attacks.
122 275
123For example, a server accepting connections from untrusted sources should 276For example, a server accepting connections from untrusted sources should
124be configured to accept only so-and-so much data that it cannot act on 277be configured to accept only so-and-so much data that it cannot act on
125(for example, when expecting a line, an attacker could send an unlimited 278(for example, when expecting a line, an attacker could send an unlimited
126amount of data without a callback ever being called as long as the line 279amount of data without a callback ever being called as long as the line
127isn't finished). 280isn't finished).
128 281
282=item autocork => <boolean>
283
284When disabled (the default), C<push_write> will try to immediately
285write the data to the handle if possible. This avoids having to register
286a write watcher and wait for the next event loop iteration, but can
287be inefficient if you write multiple small chunks (on the wire, this
288disadvantage is usually avoided by your kernel's nagle algorithm, see
289C<no_delay>, but this option can save costly syscalls).
290
291When enabled, writes will always be queued till the next event loop
292iteration. This is efficient when you do many small writes per iteration,
293but less efficient when you do a single write only per iteration (or when
294the write buffer often is full). It also increases write latency.
295
296=item no_delay => <boolean>
297
298When doing small writes on sockets, your operating system kernel might
299wait a bit for more data before actually sending it out. This is called
300the Nagle algorithm, and usually it is beneficial.
301
302In some situations you want as low a delay as possible, which can be
303accomplishd by setting this option to a true value.
304
305The default is your operating system's default behaviour (most likely
306enabled). This option explicitly enables or disables it, if possible.
307
308=item keepalive => <boolean>
309
310Enables (default disable) the SO_KEEPALIVE option on the stream socket:
311normally, TCP connections have no time-out once established, so TCP
312connections, once established, can stay alive forever even when the other
313side has long gone. TCP keepalives are a cheap way to take down long-lived
314TCP connections when the other side becomes unreachable. While the default
315is OS-dependent, TCP keepalives usually kick in after around two hours,
316and, if the other side doesn't reply, take down the TCP connection some 10
317to 15 minutes later.
318
319It is harmless to specify this option for file handles that do not support
320keepalives, and enabling it on connections that are potentially long-lived
321is usually a good idea.
322
323=item oobinline => <boolean>
324
325BSD majorly fucked up the implementation of TCP urgent data. The result
326is that almost no OS implements TCP according to the specs, and every OS
327implements it slightly differently.
328
329If you want to handle TCP urgent data, then setting this flag (the default
330is enabled) gives you the most portable way of getting urgent data, by
331putting it into the stream.
332
333Since BSD emulation of OOB data on top of TCP's urgent data can have
334security implications, AnyEvent::Handle sets this flag automatically
335unless explicitly specified. Note that setting this flag after
336establishing a connection I<may> be a bit too late (data loss could
337already have occured on BSD systems), but at least it will protect you
338from most attacks.
339
129=item read_size => <bytes> 340=item read_size => <bytes>
130 341
131The default read block size (the amount of bytes this module will try to read 342The default read block size (the number of bytes this module will
132on each [loop iteration). Default: C<4096>. 343try to read during each loop iteration, which affects memory
344requirements). Default: C<8192>.
133 345
134=item low_water_mark => <bytes> 346=item low_water_mark => <bytes>
135 347
136Sets the amount of bytes (default: C<0>) that make up an "empty" write 348Sets the number of bytes (default: C<0>) that make up an "empty" write
137buffer: If the write reaches this size or gets even samller it is 349buffer: If the buffer reaches this size or gets even samller it is
138considered empty. 350considered empty.
351
352Sometimes it can be beneficial (for performance reasons) to add data to
353the write buffer before it is fully drained, but this is a rare case, as
354the operating system kernel usually buffers data as well, so the default
355is good in almost all cases.
356
357=item linger => <seconds>
358
359If this is non-zero (default: C<3600>), the destructor of the
360AnyEvent::Handle object will check whether there is still outstanding
361write data and will install a watcher that will write this data to the
362socket. No errors will be reported (this mostly matches how the operating
363system treats outstanding data at socket close time).
364
365This will not work for partial TLS data that could not be encoded
366yet. This data will be lost. Calling the C<stoptls> method in time might
367help.
368
369=item peername => $string
370
371A string used to identify the remote site - usually the DNS hostname
372(I<not> IDN!) used to create the connection, rarely the IP address.
373
374Apart from being useful in error messages, this string is also used in TLS
375peername verification (see C<verify_peername> in L<AnyEvent::TLS>). This
376verification will be skipped when C<peername> is not specified or is
377C<undef>.
378
379=item tls => "accept" | "connect" | Net::SSLeay::SSL object
380
381When this parameter is given, it enables TLS (SSL) mode, that means
382AnyEvent will start a TLS handshake as soon as the connection has been
383established and will transparently encrypt/decrypt data afterwards.
384
385All TLS protocol errors will be signalled as C<EPROTO>, with an
386appropriate error message.
387
388TLS mode requires Net::SSLeay to be installed (it will be loaded
389automatically when you try to create a TLS handle): this module doesn't
390have a dependency on that module, so if your module requires it, you have
391to add the dependency yourself.
392
393Unlike TCP, TLS has a server and client side: for the TLS server side, use
394C<accept>, and for the TLS client side of a connection, use C<connect>
395mode.
396
397You can also provide your own TLS connection object, but you have
398to make sure that you call either C<Net::SSLeay::set_connect_state>
399or C<Net::SSLeay::set_accept_state> on it before you pass it to
400AnyEvent::Handle. Also, this module will take ownership of this connection
401object.
402
403At some future point, AnyEvent::Handle might switch to another TLS
404implementation, then the option to use your own session object will go
405away.
406
407B<IMPORTANT:> since Net::SSLeay "objects" are really only integers,
408passing in the wrong integer will lead to certain crash. This most often
409happens when one uses a stylish C<< tls => 1 >> and is surprised about the
410segmentation fault.
411
412Use the C<< ->starttls >> method if you need to start TLS negotiation later.
413
414=item tls_ctx => $anyevent_tls
415
416Use the given C<AnyEvent::TLS> object to create the new TLS connection
417(unless a connection object was specified directly). If this parameter is
418missing, then AnyEvent::Handle will use C<AnyEvent::Handle::TLS_CTX>.
419
420Instead of an object, you can also specify a hash reference with C<< key
421=> value >> pairs. Those will be passed to L<AnyEvent::TLS> to create a
422new TLS context object.
423
424=item on_starttls => $cb->($handle, $success[, $error_message])
425
426This callback will be invoked when the TLS/SSL handshake has finished. If
427C<$success> is true, then the TLS handshake succeeded, otherwise it failed
428(C<on_stoptls> will not be called in this case).
429
430The session in C<< $handle->{tls} >> can still be examined in this
431callback, even when the handshake was not successful.
432
433TLS handshake failures will not cause C<on_error> to be invoked when this
434callback is in effect, instead, the error message will be passed to C<on_starttls>.
435
436Without this callback, handshake failures lead to C<on_error> being
437called as usual.
438
439Note that you cannot just call C<starttls> again in this callback. If you
440need to do that, start an zero-second timer instead whose callback can
441then call C<< ->starttls >> again.
442
443=item on_stoptls => $cb->($handle)
444
445When a SSLv3/TLS shutdown/close notify/EOF is detected and this callback is
446set, then it will be invoked after freeing the TLS session. If it is not,
447then a TLS shutdown condition will be treated like a normal EOF condition
448on the handle.
449
450The session in C<< $handle->{tls} >> can still be examined in this
451callback.
452
453This callback will only be called on TLS shutdowns, not when the
454underlying handle signals EOF.
455
456=item json => JSON or JSON::XS object
457
458This is the json coder object used by the C<json> read and write types.
459
460If you don't supply it, then AnyEvent::Handle will create and use a
461suitable one (on demand), which will write and expect UTF-8 encoded JSON
462texts.
463
464Note that you are responsible to depend on the JSON module if you want to
465use this functionality, as AnyEvent does not have a dependency itself.
139 466
140=back 467=back
141 468
142=cut 469=cut
143 470
144sub new { 471sub new {
145 my $class = shift; 472 my $class = shift;
146
147 my $self = bless { @_ }, $class; 473 my $self = bless { @_ }, $class;
148 474
149 $self->{fh} or Carp::croak "mandatory argument fh is missing"; 475 if ($self->{fh}) {
476 $self->_start;
477 return unless $self->{fh}; # could be gone by now
478
479 } elsif ($self->{connect}) {
480 require AnyEvent::Socket;
481
482 $self->{peername} = $self->{connect}[0]
483 unless exists $self->{peername};
484
485 $self->{_skip_drain_rbuf} = 1;
486
487 {
488 Scalar::Util::weaken (my $self = $self);
489
490 $self->{_connect} =
491 AnyEvent::Socket::tcp_connect (
492 $self->{connect}[0],
493 $self->{connect}[1],
494 sub {
495 my ($fh, $host, $port, $retry) = @_;
496
497 if ($fh) {
498 $self->{fh} = $fh;
499
500 delete $self->{_skip_drain_rbuf};
501 $self->_start;
502
503 $self->{on_connect}
504 and $self->{on_connect}($self, $host, $port, sub {
505 delete @$self{qw(fh _tw _rtw _wtw _ww _rw _eof _queue rbuf _wbuf tls _tls_rbuf _tls_wbuf)};
506 $self->{_skip_drain_rbuf} = 1;
507 &$retry;
508 });
509
510 } else {
511 if ($self->{on_connect_error}) {
512 $self->{on_connect_error}($self, "$!");
513 $self->destroy;
514 } else {
515 $self->_error ($!, 1);
516 }
517 }
518 },
519 sub {
520 local $self->{fh} = $_[0];
521
522 $self->{on_prepare}
523 ? $self->{on_prepare}->($self)
524 : ()
525 }
526 );
527 }
528
529 } else {
530 Carp::croak "AnyEvent::Handle: either an existing fh or the connect parameter must be specified";
531 }
532
533 $self
534}
535
536sub _start {
537 my ($self) = @_;
538
539 # too many clueless people try to use udp and similar sockets
540 # with AnyEvent::Handle, do them a favour.
541 my $type = getsockopt $self->{fh}, Socket::SOL_SOCKET (), Socket::SO_TYPE ();
542 Carp::croak "AnyEvent::Handle: only stream sockets supported, anything else will NOT work!"
543 if Socket::SOCK_STREAM () != (unpack "I", $type) && defined $type;
150 544
151 AnyEvent::Util::fh_nonblocking $self->{fh}, 1; 545 AnyEvent::Util::fh_nonblocking $self->{fh}, 1;
152 546
153 $self->on_eof ((delete $self->{on_eof} ) or Carp::croak "mandatory argument on_eof is missing"); 547 $self->{_activity} =
548 $self->{_ractivity} =
549 $self->{_wactivity} = AE::now;
154 550
155 $self->on_error (delete $self->{on_error}) if $self->{on_error}; 551 $self->timeout (delete $self->{timeout} ) if $self->{timeout};
552 $self->rtimeout (delete $self->{rtimeout} ) if $self->{rtimeout};
553 $self->wtimeout (delete $self->{wtimeout} ) if $self->{wtimeout};
554
555 $self->no_delay (delete $self->{no_delay} ) if exists $self->{no_delay} && $self->{no_delay};
556 $self->keepalive (delete $self->{keepalive}) if exists $self->{keepalive} && $self->{keepalive};
557
558 $self->oobinline (exists $self->{oobinline} ? delete $self->{oobinline} : 1);
559
560 $self->starttls (delete $self->{tls}, delete $self->{tls_ctx})
561 if $self->{tls};
562
156 $self->on_drain (delete $self->{on_drain}) if $self->{on_drain}; 563 $self->on_drain (delete $self->{on_drain} ) if $self->{on_drain};
157 $self->on_read (delete $self->{on_read} ) if $self->{on_read};
158 564
159 $self->start_read; 565 $self->start_read
566 if $self->{on_read} || @{ $self->{_queue} };
160 567
161 $self 568 $self->_drain_wbuf;
162} 569}
163 570
164sub _shutdown {
165 my ($self) = @_;
166
167 delete $self->{rw};
168 delete $self->{ww};
169 delete $self->{fh};
170}
171
172sub error { 571sub _error {
173 my ($self) = @_; 572 my ($self, $errno, $fatal, $message) = @_;
174 573
175 { 574 $! = $errno;
176 local $!; 575 $message ||= "$!";
177 $self->_shutdown;
178 }
179 576
180 if ($self->{on_error}) { 577 if ($self->{on_error}) {
181 $self->{on_error}($self); 578 $self->{on_error}($self, $fatal, $message);
182 } else { 579 $self->destroy if $fatal;
580 } elsif ($self->{fh} || $self->{connect}) {
581 $self->destroy;
183 die "AnyEvent::Handle uncaught fatal error: $!"; 582 Carp::croak "AnyEvent::Handle uncaught error: $message";
184 } 583 }
185} 584}
186 585
187=item $fh = $handle->fh 586=item $fh = $handle->fh
188 587
189This method returns the filehandle of the L<AnyEvent::Handle> object. 588This method returns the file handle used to create the L<AnyEvent::Handle> object.
190 589
191=cut 590=cut
192 591
193sub fh { $_[0]->{fh} } 592sub fh { $_[0]{fh} }
194 593
195=item $handle->on_error ($cb) 594=item $handle->on_error ($cb)
196 595
197Replace the current C<on_error> callback (see the C<on_error> constructor argument). 596Replace the current C<on_error> callback (see the C<on_error> constructor argument).
198 597
210 609
211sub on_eof { 610sub on_eof {
212 $_[0]{on_eof} = $_[1]; 611 $_[0]{on_eof} = $_[1];
213} 612}
214 613
614=item $handle->on_timeout ($cb)
615
616=item $handle->on_rtimeout ($cb)
617
618=item $handle->on_wtimeout ($cb)
619
620Replace the current C<on_timeout>, C<on_rtimeout> or C<on_wtimeout>
621callback, or disables the callback (but not the timeout) if C<$cb> =
622C<undef>. See the C<timeout> constructor argument and method.
623
624=cut
625
626# see below
627
628=item $handle->autocork ($boolean)
629
630Enables or disables the current autocork behaviour (see C<autocork>
631constructor argument). Changes will only take effect on the next write.
632
633=cut
634
635sub autocork {
636 $_[0]{autocork} = $_[1];
637}
638
639=item $handle->no_delay ($boolean)
640
641Enables or disables the C<no_delay> setting (see constructor argument of
642the same name for details).
643
644=cut
645
646sub no_delay {
647 $_[0]{no_delay} = $_[1];
648
649 setsockopt $_[0]{fh}, Socket::IPPROTO_TCP (), Socket::TCP_NODELAY (), int $_[1]
650 if $_[0]{fh};
651}
652
653=item $handle->keepalive ($boolean)
654
655Enables or disables the C<keepalive> setting (see constructor argument of
656the same name for details).
657
658=cut
659
660sub keepalive {
661 $_[0]{keepalive} = $_[1];
662
663 eval {
664 local $SIG{__DIE__};
665 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
666 if $_[0]{fh};
667 };
668}
669
670=item $handle->oobinline ($boolean)
671
672Enables or disables the C<oobinline> setting (see constructor argument of
673the same name for details).
674
675=cut
676
677sub oobinline {
678 $_[0]{oobinline} = $_[1];
679
680 eval {
681 local $SIG{__DIE__};
682 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_OOBINLINE (), int $_[1]
683 if $_[0]{fh};
684 };
685}
686
687=item $handle->keepalive ($boolean)
688
689Enables or disables the C<keepalive> setting (see constructor argument of
690the same name for details).
691
692=cut
693
694sub keepalive {
695 $_[0]{keepalive} = $_[1];
696
697 eval {
698 local $SIG{__DIE__};
699 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
700 if $_[0]{fh};
701 };
702}
703
704=item $handle->on_starttls ($cb)
705
706Replace the current C<on_starttls> callback (see the C<on_starttls> constructor argument).
707
708=cut
709
710sub on_starttls {
711 $_[0]{on_starttls} = $_[1];
712}
713
714=item $handle->on_stoptls ($cb)
715
716Replace the current C<on_stoptls> callback (see the C<on_stoptls> constructor argument).
717
718=cut
719
720sub on_stoptls {
721 $_[0]{on_stoptls} = $_[1];
722}
723
724=item $handle->rbuf_max ($max_octets)
725
726Configures the C<rbuf_max> setting (C<undef> disables it).
727
728=cut
729
730sub rbuf_max {
731 $_[0]{rbuf_max} = $_[1];
732}
733
734#############################################################################
735
736=item $handle->timeout ($seconds)
737
738=item $handle->rtimeout ($seconds)
739
740=item $handle->wtimeout ($seconds)
741
742Configures (or disables) the inactivity timeout.
743
744=item $handle->timeout_reset
745
746=item $handle->rtimeout_reset
747
748=item $handle->wtimeout_reset
749
750Reset the activity timeout, as if data was received or sent.
751
752These methods are cheap to call.
753
754=cut
755
756for my $dir ("", "r", "w") {
757 my $timeout = "${dir}timeout";
758 my $tw = "_${dir}tw";
759 my $on_timeout = "on_${dir}timeout";
760 my $activity = "_${dir}activity";
761 my $cb;
762
763 *$on_timeout = sub {
764 $_[0]{$on_timeout} = $_[1];
765 };
766
767 *$timeout = sub {
768 my ($self, $new_value) = @_;
769
770 $self->{$timeout} = $new_value;
771 delete $self->{$tw}; &$cb;
772 };
773
774 *{"${dir}timeout_reset"} = sub {
775 $_[0]{$activity} = AE::now;
776 };
777
778 # main workhorse:
779 # reset the timeout watcher, as neccessary
780 # also check for time-outs
781 $cb = sub {
782 my ($self) = @_;
783
784 if ($self->{$timeout} && $self->{fh}) {
785 my $NOW = AE::now;
786
787 # when would the timeout trigger?
788 my $after = $self->{$activity} + $self->{$timeout} - $NOW;
789
790 # now or in the past already?
791 if ($after <= 0) {
792 $self->{$activity} = $NOW;
793
794 if ($self->{$on_timeout}) {
795 $self->{$on_timeout}($self);
796 } else {
797 $self->_error (Errno::ETIMEDOUT);
798 }
799
800 # callback could have changed timeout value, optimise
801 return unless $self->{$timeout};
802
803 # calculate new after
804 $after = $self->{$timeout};
805 }
806
807 Scalar::Util::weaken $self;
808 return unless $self; # ->error could have destroyed $self
809
810 $self->{$tw} ||= AE::timer $after, 0, sub {
811 delete $self->{$tw};
812 $cb->($self);
813 };
814 } else {
815 delete $self->{$tw};
816 }
817 }
818}
819
215############################################################################# 820#############################################################################
216 821
217=back 822=back
218 823
219=head2 WRITE QUEUE 824=head2 WRITE QUEUE
222for reading. 827for reading.
223 828
224The write queue is very simple: you can add data to its end, and 829The write queue is very simple: you can add data to its end, and
225AnyEvent::Handle will automatically try to get rid of it for you. 830AnyEvent::Handle will automatically try to get rid of it for you.
226 831
227When data could be writtena nd the write buffer is shorter then the low 832When data could be written and the write buffer is shorter then the low
228water mark, the C<on_drain> callback will be invoked. 833water mark, the C<on_drain> callback will be invoked.
229 834
230=over 4 835=over 4
231 836
232=item $handle->on_drain ($cb) 837=item $handle->on_drain ($cb)
233 838
234Sets the C<on_drain> callback or clears it (see the description of 839Sets the C<on_drain> callback or clears it (see the description of
235C<on_drain> in the constructor). 840C<on_drain> in the constructor).
236 841
842This method may invoke callbacks (and therefore the handle might be
843destroyed after it returns).
844
237=cut 845=cut
238 846
239sub on_drain { 847sub on_drain {
240 my ($self, $cb) = @_; 848 my ($self, $cb) = @_;
241 849
242 $self->{on_drain} = $cb; 850 $self->{on_drain} = $cb;
243 851
244 $cb->($self) 852 $cb->($self)
245 if $cb && $self->{low_water_mark} >= length $self->{wbuf}; 853 if $cb && $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf});
246} 854}
247 855
248=item $handle->push_write ($data) 856=item $handle->push_write ($data)
249 857
250Queues the given scalar to be written. You can push as much data as you 858Queues the given scalar to be written. You can push as much data as you
251want (only limited by the available memory), as C<AnyEvent::Handle> 859want (only limited by the available memory), as C<AnyEvent::Handle>
252buffers it independently of the kernel. 860buffers it independently of the kernel.
253 861
254=cut 862This method may invoke callbacks (and therefore the handle might be
863destroyed after it returns).
255 864
256sub push_write { 865=cut
866
867sub _drain_wbuf {
257 my ($self, $data) = @_; 868 my ($self) = @_;
258 869
259 $self->{wbuf} .= $data; 870 if (!$self->{_ww} && length $self->{wbuf}) {
260 871
261 unless ($self->{ww}) {
262 Scalar::Util::weaken $self; 872 Scalar::Util::weaken $self;
873
263 my $cb = sub { 874 my $cb = sub {
264 my $len = syswrite $self->{fh}, $self->{wbuf}; 875 my $len = syswrite $self->{fh}, $self->{wbuf};
265 876
266 if ($len > 0) { 877 if (defined $len) {
267 substr $self->{wbuf}, 0, $len, ""; 878 substr $self->{wbuf}, 0, $len, "";
268 879
880 $self->{_activity} = $self->{_wactivity} = AE::now;
269 881
270 $self->{on_drain}($self) 882 $self->{on_drain}($self)
271 if $self->{low_water_mark} >= length $self->{wbuf} 883 if $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf})
272 && $self->{on_drain}; 884 && $self->{on_drain};
273 885
274 delete $self->{ww} unless length $self->{wbuf}; 886 delete $self->{_ww} unless length $self->{wbuf};
275 } elsif ($! != EAGAIN && $! != EINTR) { 887 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
276 $self->error; 888 $self->_error ($!, 1);
277 } 889 }
278 }; 890 };
279 891
280 $self->{ww} = AnyEvent->io (fh => $self->{fh}, poll => "w", cb => $cb); 892 # try to write data immediately
893 $cb->() unless $self->{autocork};
281 894
282 $cb->($self); 895 # if still data left in wbuf, we need to poll
896 $self->{_ww} = AE::io $self->{fh}, 1, $cb
897 if length $self->{wbuf};
283 }; 898 };
284} 899}
900
901our %WH;
902
903# deprecated
904sub register_write_type($$) {
905 $WH{$_[0]} = $_[1];
906}
907
908sub push_write {
909 my $self = shift;
910
911 if (@_ > 1) {
912 my $type = shift;
913
914 @_ = ($WH{$type} ||= _load_func "$type\::anyevent_write_type"
915 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_write")
916 ->($self, @_);
917 }
918
919 # we downgrade here to avoid hard-to-track-down bugs,
920 # and diagnose the problem earlier and better.
921
922 if ($self->{tls}) {
923 utf8::downgrade $self->{_tls_wbuf} .= $_[0];
924 &_dotls ($self) if $self->{fh};
925 } else {
926 utf8::downgrade $self->{wbuf} .= $_[0];
927 $self->_drain_wbuf if $self->{fh};
928 }
929}
930
931=item $handle->push_write (type => @args)
932
933Instead of formatting your data yourself, you can also let this module
934do the job by specifying a type and type-specific arguments. You
935can also specify the (fully qualified) name of a package, in which
936case AnyEvent tries to load the package and then expects to find the
937C<anyevent_write_type> function inside (see "custom write types", below).
938
939Predefined types are (if you have ideas for additional types, feel free to
940drop by and tell us):
941
942=over 4
943
944=item netstring => $string
945
946Formats the given value as netstring
947(http://cr.yp.to/proto/netstrings.txt, this is not a recommendation to use them).
948
949=cut
950
951register_write_type netstring => sub {
952 my ($self, $string) = @_;
953
954 (length $string) . ":$string,"
955};
956
957=item packstring => $format, $data
958
959An octet string prefixed with an encoded length. The encoding C<$format>
960uses the same format as a Perl C<pack> format, but must specify a single
961integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
962optional C<!>, C<< < >> or C<< > >> modifier).
963
964=cut
965
966register_write_type packstring => sub {
967 my ($self, $format, $string) = @_;
968
969 pack "$format/a*", $string
970};
971
972=item json => $array_or_hashref
973
974Encodes the given hash or array reference into a JSON object. Unless you
975provide your own JSON object, this means it will be encoded to JSON text
976in UTF-8.
977
978JSON objects (and arrays) are self-delimiting, so you can write JSON at
979one end of a handle and read them at the other end without using any
980additional framing.
981
982The generated JSON text is guaranteed not to contain any newlines: While
983this module doesn't need delimiters after or between JSON texts to be
984able to read them, many other languages depend on that.
985
986A simple RPC protocol that interoperates easily with others is to send
987JSON arrays (or objects, although arrays are usually the better choice as
988they mimic how function argument passing works) and a newline after each
989JSON text:
990
991 $handle->push_write (json => ["method", "arg1", "arg2"]); # whatever
992 $handle->push_write ("\012");
993
994An AnyEvent::Handle receiver would simply use the C<json> read type and
995rely on the fact that the newline will be skipped as leading whitespace:
996
997 $handle->push_read (json => sub { my $array = $_[1]; ... });
998
999Other languages could read single lines terminated by a newline and pass
1000this line into their JSON decoder of choice.
1001
1002=cut
1003
1004sub json_coder() {
1005 eval { require JSON::XS; JSON::XS->new->utf8 }
1006 || do { require JSON; JSON->new->utf8 }
1007}
1008
1009register_write_type json => sub {
1010 my ($self, $ref) = @_;
1011
1012 my $json = $self->{json} ||= json_coder;
1013
1014 $json->encode ($ref)
1015};
1016
1017=item storable => $reference
1018
1019Freezes the given reference using L<Storable> and writes it to the
1020handle. Uses the C<nfreeze> format.
1021
1022=cut
1023
1024register_write_type storable => sub {
1025 my ($self, $ref) = @_;
1026
1027 require Storable;
1028
1029 pack "w/a*", Storable::nfreeze ($ref)
1030};
1031
1032=back
1033
1034=item $handle->push_shutdown
1035
1036Sometimes you know you want to close the socket after writing your data
1037before it was actually written. One way to do that is to replace your
1038C<on_drain> handler by a callback that shuts down the socket (and set
1039C<low_water_mark> to C<0>). This method is a shorthand for just that, and
1040replaces the C<on_drain> callback with:
1041
1042 sub { shutdown $_[0]{fh}, 1 } # for push_shutdown
1043
1044This simply shuts down the write side and signals an EOF condition to the
1045the peer.
1046
1047You can rely on the normal read queue and C<on_eof> handling
1048afterwards. This is the cleanest way to close a connection.
1049
1050This method may invoke callbacks (and therefore the handle might be
1051destroyed after it returns).
1052
1053=cut
1054
1055sub push_shutdown {
1056 my ($self) = @_;
1057
1058 delete $self->{low_water_mark};
1059 $self->on_drain (sub { shutdown $_[0]{fh}, 1 });
1060}
1061
1062=item custom write types - Package::anyevent_write_type $handle, @args
1063
1064Instead of one of the predefined types, you can also specify the name of
1065a package. AnyEvent will try to load the package and then expects to find
1066a function named C<anyevent_write_type> inside. If it isn't found, it
1067progressively tries to load the parent package until it either finds the
1068function (good) or runs out of packages (bad).
1069
1070Whenever the given C<type> is used, C<push_write> will the function with
1071the handle object and the remaining arguments.
1072
1073The function is supposed to return a single octet string that will be
1074appended to the write buffer, so you cna mentally treat this function as a
1075"arguments to on-the-wire-format" converter.
1076
1077Example: implement a custom write type C<join> that joins the remaining
1078arguments using the first one.
1079
1080 $handle->push_write (My::Type => " ", 1,2,3);
1081
1082 # uses the following package, which can be defined in the "My::Type" or in
1083 # the "My" modules to be auto-loaded, or just about anywhere when the
1084 # My::Type::anyevent_write_type is defined before invoking it.
1085
1086 package My::Type;
1087
1088 sub anyevent_write_type {
1089 my ($handle, $delim, @args) = @_;
1090
1091 join $delim, @args
1092 }
1093
1094=cut
285 1095
286############################################################################# 1096#############################################################################
287 1097
288=back 1098=back
289 1099
296ways, the "simple" way, using only C<on_read> and the "complex" way, using 1106ways, the "simple" way, using only C<on_read> and the "complex" way, using
297a queue. 1107a queue.
298 1108
299In the simple case, you just install an C<on_read> callback and whenever 1109In the simple case, you just install an C<on_read> callback and whenever
300new data arrives, it will be called. You can then remove some data (if 1110new data arrives, it will be called. You can then remove some data (if
301enough is there) from the read buffer (C<< $handle->rbuf >>) if you want 1111enough is there) from the read buffer (C<< $handle->rbuf >>). Or you can
302or not. 1112leave the data there if you want to accumulate more (e.g. when only a
1113partial message has been received so far), or change the read queue with
1114e.g. C<push_read>.
303 1115
304In the more complex case, you want to queue multiple callbacks. In this 1116In the more complex case, you want to queue multiple callbacks. In this
305case, AnyEvent::Handle will call the first queued callback each time new 1117case, AnyEvent::Handle will call the first queued callback each time new
306data arrives and removes it when it has done its job (see C<push_read>, 1118data arrives (also the first time it is queued) and remove it when it has
307below). 1119done its job (see C<push_read>, below).
308 1120
309This way you can, for example, push three line-reads, followed by reading 1121This way you can, for example, push three line-reads, followed by reading
310a chunk of data, and AnyEvent::Handle will execute them in order. 1122a chunk of data, and AnyEvent::Handle will execute them in order.
311 1123
312Example 1: EPP protocol parser. EPP sends 4 byte length info, followed by 1124Example 1: EPP protocol parser. EPP sends 4 byte length info, followed by
313the specified number of bytes which give an XML datagram. 1125the specified number of bytes which give an XML datagram.
314 1126
315 # in the default state, expect some header bytes 1127 # in the default state, expect some header bytes
316 $handle->on_read (sub { 1128 $handle->on_read (sub {
317 # some data is here, now queue the length-header-read (4 octets) 1129 # some data is here, now queue the length-header-read (4 octets)
318 shift->unshift_read_chunk (4, sub { 1130 shift->unshift_read (chunk => 4, sub {
319 # header arrived, decode 1131 # header arrived, decode
320 my $len = unpack "N", $_[1]; 1132 my $len = unpack "N", $_[1];
321 1133
322 # now read the payload 1134 # now read the payload
323 shift->unshift_read_chunk ($len, sub { 1135 shift->unshift_read (chunk => $len, sub {
324 my $xml = $_[1]; 1136 my $xml = $_[1];
325 # handle xml 1137 # handle xml
326 }); 1138 });
327 }); 1139 });
328 }); 1140 });
329 1141
330Example 2: Implement a client for a protocol that replies either with 1142Example 2: Implement a client for a protocol that replies either with "OK"
331"OK" and another line or "ERROR" for one request, and 64 bytes for the 1143and another line or "ERROR" for the first request that is sent, and 64
332second request. Due tot he availability of a full queue, we can just 1144bytes for the second request. Due to the availability of a queue, we can
333pipeline sending both requests and manipulate the queue as necessary in 1145just pipeline sending both requests and manipulate the queue as necessary
334the callbacks: 1146in the callbacks.
335 1147
336 # request one 1148When the first callback is called and sees an "OK" response, it will
1149C<unshift> another line-read. This line-read will be queued I<before> the
115064-byte chunk callback.
1151
1152 # request one, returns either "OK + extra line" or "ERROR"
337 $handle->push_write ("request 1\015\012"); 1153 $handle->push_write ("request 1\015\012");
338 1154
339 # we expect "ERROR" or "OK" as response, so push a line read 1155 # we expect "ERROR" or "OK" as response, so push a line read
340 $handle->push_read_line (sub { 1156 $handle->push_read (line => sub {
341 # if we got an "OK", we have to _prepend_ another line, 1157 # if we got an "OK", we have to _prepend_ another line,
342 # so it will be read before the second request reads its 64 bytes 1158 # so it will be read before the second request reads its 64 bytes
343 # which are already in the queue when this callback is called 1159 # which are already in the queue when this callback is called
344 # we don't do this in case we got an error 1160 # we don't do this in case we got an error
345 if ($_[1] eq "OK") { 1161 if ($_[1] eq "OK") {
346 $_[0]->unshift_read_line (sub { 1162 $_[0]->unshift_read (line => sub {
347 my $response = $_[1]; 1163 my $response = $_[1];
348 ... 1164 ...
349 }); 1165 });
350 } 1166 }
351 }); 1167 });
352 1168
353 # request two 1169 # request two, simply returns 64 octets
354 $handle->push_write ("request 2\015\012"); 1170 $handle->push_write ("request 2\015\012");
355 1171
356 # simply read 64 bytes, always 1172 # simply read 64 bytes, always
357 $handle->push_read_chunk (64, sub { 1173 $handle->push_read (chunk => 64, sub {
358 my $response = $_[1]; 1174 my $response = $_[1];
359 ... 1175 ...
360 }); 1176 });
361 1177
362=over 4 1178=over 4
364=cut 1180=cut
365 1181
366sub _drain_rbuf { 1182sub _drain_rbuf {
367 my ($self) = @_; 1183 my ($self) = @_;
368 1184
1185 # avoid recursion
369 return if $self->{in_drain}; 1186 return if $self->{_skip_drain_rbuf};
370 local $self->{in_drain} = 1; 1187 local $self->{_skip_drain_rbuf} = 1;
371 1188
1189 while () {
1190 # we need to use a separate tls read buffer, as we must not receive data while
1191 # we are draining the buffer, and this can only happen with TLS.
1192 $self->{rbuf} .= delete $self->{_tls_rbuf}
1193 if exists $self->{_tls_rbuf};
1194
372 while (my $len = length $self->{rbuf}) { 1195 my $len = length $self->{rbuf};
373 no strict 'refs'; 1196
374 if (my $cb = shift @{ $self->{queue} }) { 1197 if (my $cb = shift @{ $self->{_queue} }) {
375 if (!$cb->($self)) { 1198 unless ($cb->($self)) {
1199 # no progress can be made
1200 # (not enough data and no data forthcoming)
1201 $self->_error (Errno::EPIPE, 1), return
376 if ($self->{eof}) { 1202 if $self->{_eof};
377 # no progress can be made (not enough data and no data forthcoming)
378 $! = &Errno::EPIPE; return $self->error;
379 }
380 1203
381 unshift @{ $self->{queue} }, $cb; 1204 unshift @{ $self->{_queue} }, $cb;
382 return; 1205 last;
383 } 1206 }
384 } elsif ($self->{on_read}) { 1207 } elsif ($self->{on_read}) {
1208 last unless $len;
1209
385 $self->{on_read}($self); 1210 $self->{on_read}($self);
386 1211
387 if ( 1212 if (
388 $self->{eof} # if no further data will arrive
389 && $len == length $self->{rbuf} # and no data has been consumed 1213 $len == length $self->{rbuf} # if no data has been consumed
390 && !@{ $self->{queue} } # and the queue is still empty 1214 && !@{ $self->{_queue} } # and the queue is still empty
391 && $self->{on_read} # and we still want to read data 1215 && $self->{on_read} # but we still have on_read
392 ) { 1216 ) {
1217 # no further data will arrive
393 # then no progress can be made 1218 # so no progress can be made
394 $! = &Errno::EPIPE; return $self->error; 1219 $self->_error (Errno::EPIPE, 1), return
1220 if $self->{_eof};
1221
1222 last; # more data might arrive
395 } 1223 }
396 } else { 1224 } else {
397 # read side becomes idle 1225 # read side becomes idle
398 delete $self->{rw}; 1226 delete $self->{_rw} unless $self->{tls};
399 return; 1227 last;
400 } 1228 }
401 } 1229 }
402 1230
403 if ($self->{eof}) { 1231 if ($self->{_eof}) {
404 $self->_shutdown; 1232 $self->{on_eof}
405 $self->{on_eof}($self); 1233 ? $self->{on_eof}($self)
1234 : $self->_error (0, 1, "Unexpected end-of-file");
1235
1236 return;
1237 }
1238
1239 if (
1240 defined $self->{rbuf_max}
1241 && $self->{rbuf_max} < length $self->{rbuf}
1242 ) {
1243 $self->_error (Errno::ENOSPC, 1), return;
1244 }
1245
1246 # may need to restart read watcher
1247 unless ($self->{_rw}) {
1248 $self->start_read
1249 if $self->{on_read} || @{ $self->{_queue} };
406 } 1250 }
407} 1251}
408 1252
409=item $handle->on_read ($cb) 1253=item $handle->on_read ($cb)
410 1254
411This replaces the currently set C<on_read> callback, or clears it (when 1255This replaces the currently set C<on_read> callback, or clears it (when
412the new callback is C<undef>). See the description of C<on_read> in the 1256the new callback is C<undef>). See the description of C<on_read> in the
413constructor. 1257constructor.
414 1258
1259This method may invoke callbacks (and therefore the handle might be
1260destroyed after it returns).
1261
415=cut 1262=cut
416 1263
417sub on_read { 1264sub on_read {
418 my ($self, $cb) = @_; 1265 my ($self, $cb) = @_;
419 1266
420 $self->{on_read} = $cb; 1267 $self->{on_read} = $cb;
1268 $self->_drain_rbuf if $cb;
421} 1269}
422 1270
423=item $handle->rbuf 1271=item $handle->rbuf
424 1272
425Returns the read buffer (as a modifiable lvalue). 1273Returns the read buffer (as a modifiable lvalue). You can also access the
1274read buffer directly as the C<< ->{rbuf} >> member, if you want (this is
1275much faster, and no less clean).
426 1276
427You can access the read buffer directly as the C<< ->{rbuf} >> member, if 1277The only operation allowed on the read buffer (apart from looking at it)
428you want. 1278is removing data from its beginning. Otherwise modifying or appending to
1279it is not allowed and will lead to hard-to-track-down bugs.
429 1280
430NOTE: The read buffer should only be used or modified if the C<on_read>, 1281NOTE: The read buffer should only be used or modified in the C<on_read>
431C<push_read> or C<unshift_read> methods are used. The other read methods 1282callback or when C<push_read> or C<unshift_read> are used with a single
432automatically manage the read buffer. 1283callback (i.e. untyped). Typed C<push_read> and C<unshift_read> methods
1284will manage the read buffer on their own.
433 1285
434=cut 1286=cut
435 1287
436sub rbuf : lvalue { 1288sub rbuf : lvalue {
437 $_[0]{rbuf} 1289 $_[0]{rbuf}
444Append the given callback to the end of the queue (C<push_read>) or 1296Append the given callback to the end of the queue (C<push_read>) or
445prepend it (C<unshift_read>). 1297prepend it (C<unshift_read>).
446 1298
447The callback is called each time some additional read data arrives. 1299The callback is called each time some additional read data arrives.
448 1300
449It must check wether enough data is in the read buffer already. 1301It must check whether enough data is in the read buffer already.
450 1302
451If not enough data is available, it must return the empty list or a false 1303If not enough data is available, it must return the empty list or a false
452value, in which case it will be called repeatedly until enough data is 1304value, in which case it will be called repeatedly until enough data is
453available (or an error condition is detected). 1305available (or an error condition is detected).
454 1306
455If enough data was available, then the callback must remove all data it is 1307If enough data was available, then the callback must remove all data it is
456interested in (which can be none at all) and return a true value. After returning 1308interested in (which can be none at all) and return a true value. After returning
457true, it will be removed from the queue. 1309true, it will be removed from the queue.
458 1310
1311These methods may invoke callbacks (and therefore the handle might be
1312destroyed after it returns).
1313
459=cut 1314=cut
1315
1316our %RH;
1317
1318sub register_read_type($$) {
1319 $RH{$_[0]} = $_[1];
1320}
460 1321
461sub push_read { 1322sub push_read {
462 my ($self, $cb) = @_; 1323 my $self = shift;
1324 my $cb = pop;
463 1325
1326 if (@_) {
1327 my $type = shift;
1328
1329 $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
1330 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_read")
1331 ->($self, $cb, @_);
1332 }
1333
464 push @{ $self->{queue} }, $cb; 1334 push @{ $self->{_queue} }, $cb;
465 $self->_drain_rbuf; 1335 $self->_drain_rbuf;
466} 1336}
467 1337
468sub unshift_read { 1338sub unshift_read {
469 my ($self, $cb) = @_; 1339 my $self = shift;
1340 my $cb = pop;
470 1341
1342 if (@_) {
1343 my $type = shift;
1344
1345 $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
1346 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::unshift_read")
1347 ->($self, $cb, @_);
1348 }
1349
471 push @{ $self->{queue} }, $cb; 1350 unshift @{ $self->{_queue} }, $cb;
472 $self->_drain_rbuf; 1351 $self->_drain_rbuf;
473} 1352}
474 1353
475=item $handle->push_read_chunk ($len, $cb->($self, $data)) 1354=item $handle->push_read (type => @args, $cb)
476 1355
477=item $handle->unshift_read_chunk ($len, $cb->($self, $data)) 1356=item $handle->unshift_read (type => @args, $cb)
478 1357
479Append the given callback to the end of the queue (C<push_read_chunk>) or 1358Instead of providing a callback that parses the data itself you can chose
480prepend it (C<unshift_read_chunk>). 1359between a number of predefined parsing formats, for chunks of data, lines
1360etc. You can also specify the (fully qualified) name of a package, in
1361which case AnyEvent tries to load the package and then expects to find the
1362C<anyevent_read_type> function inside (see "custom read types", below).
481 1363
482The callback will be called only once C<$len> bytes have been read, and 1364Predefined types are (if you have ideas for additional types, feel free to
483these C<$len> bytes will be passed to the callback. 1365drop by and tell us):
484 1366
485=cut 1367=over 4
486 1368
487sub _read_chunk($$) { 1369=item chunk => $octets, $cb->($handle, $data)
1370
1371Invoke the callback only once C<$octets> bytes have been read. Pass the
1372data read to the callback. The callback will never be called with less
1373data.
1374
1375Example: read 2 bytes.
1376
1377 $handle->push_read (chunk => 2, sub {
1378 warn "yay ", unpack "H*", $_[1];
1379 });
1380
1381=cut
1382
1383register_read_type chunk => sub {
488 my ($self, $len, $cb) = @_; 1384 my ($self, $cb, $len) = @_;
489 1385
490 sub { 1386 sub {
491 $len <= length $_[0]{rbuf} or return; 1387 $len <= length $_[0]{rbuf} or return;
492 $cb->($_[0], substr $_[0]{rbuf}, 0, $len, ""); 1388 $cb->($_[0], substr $_[0]{rbuf}, 0, $len, "");
493 1 1389 1
494 } 1390 }
495} 1391};
496 1392
497sub push_read_chunk { 1393=item line => [$eol, ]$cb->($handle, $line, $eol)
498 $_[0]->push_read (&_read_chunk);
499}
500
501
502sub unshift_read_chunk {
503 $_[0]->unshift_read (&_read_chunk);
504}
505
506=item $handle->push_read_line ([$eol, ]$cb->($self, $line, $eol))
507
508=item $handle->unshift_read_line ([$eol, ]$cb->($self, $line, $eol))
509
510Append the given callback to the end of the queue (C<push_read_line>) or
511prepend it (C<unshift_read_line>).
512 1394
513The callback will be called only once a full line (including the end of 1395The callback will be called only once a full line (including the end of
514line marker, C<$eol>) has been read. This line (excluding the end of line 1396line marker, C<$eol>) has been read. This line (excluding the end of line
515marker) will be passed to the callback as second argument (C<$line>), and 1397marker) will be passed to the callback as second argument (C<$line>), and
516the end of line marker as the third argument (C<$eol>). 1398the end of line marker as the third argument (C<$eol>).
527Partial lines at the end of the stream will never be returned, as they are 1409Partial lines at the end of the stream will never be returned, as they are
528not marked by the end of line marker. 1410not marked by the end of line marker.
529 1411
530=cut 1412=cut
531 1413
532sub _read_line($$) { 1414register_read_type line => sub {
533 my $self = shift; 1415 my ($self, $cb, $eol) = @_;
534 my $cb = pop;
535 my $eol = @_ ? shift : qr|(\015?\012)|;
536 my $pos;
537 1416
1417 if (@_ < 3) {
1418 # this is more than twice as fast as the generic code below
1419 sub {
1420 $_[0]{rbuf} =~ s/^([^\015\012]*)(\015?\012)// or return;
1421
1422 $cb->($_[0], $1, $2);
1423 1
1424 }
1425 } else {
538 $eol = quotemeta $eol unless ref $eol; 1426 $eol = quotemeta $eol unless ref $eol;
539 $eol = qr|^(.*?)($eol)|s; 1427 $eol = qr|^(.*?)($eol)|s;
1428
1429 sub {
1430 $_[0]{rbuf} =~ s/$eol// or return;
1431
1432 $cb->($_[0], $1, $2);
1433 1
1434 }
1435 }
1436};
1437
1438=item regex => $accept[, $reject[, $skip], $cb->($handle, $data)
1439
1440Makes a regex match against the regex object C<$accept> and returns
1441everything up to and including the match.
1442
1443Example: read a single line terminated by '\n'.
1444
1445 $handle->push_read (regex => qr<\n>, sub { ... });
1446
1447If C<$reject> is given and not undef, then it determines when the data is
1448to be rejected: it is matched against the data when the C<$accept> regex
1449does not match and generates an C<EBADMSG> error when it matches. This is
1450useful to quickly reject wrong data (to avoid waiting for a timeout or a
1451receive buffer overflow).
1452
1453Example: expect a single decimal number followed by whitespace, reject
1454anything else (not the use of an anchor).
1455
1456 $handle->push_read (regex => qr<^[0-9]+\s>, qr<[^0-9]>, sub { ... });
1457
1458If C<$skip> is given and not C<undef>, then it will be matched against
1459the receive buffer when neither C<$accept> nor C<$reject> match,
1460and everything preceding and including the match will be accepted
1461unconditionally. This is useful to skip large amounts of data that you
1462know cannot be matched, so that the C<$accept> or C<$reject> regex do not
1463have to start matching from the beginning. This is purely an optimisation
1464and is usually worth it only when you expect more than a few kilobytes.
1465
1466Example: expect a http header, which ends at C<\015\012\015\012>. Since we
1467expect the header to be very large (it isn't in practice, but...), we use
1468a skip regex to skip initial portions. The skip regex is tricky in that
1469it only accepts something not ending in either \015 or \012, as these are
1470required for the accept regex.
1471
1472 $handle->push_read (regex =>
1473 qr<\015\012\015\012>,
1474 undef, # no reject
1475 qr<^.*[^\015\012]>,
1476 sub { ... });
1477
1478=cut
1479
1480register_read_type regex => sub {
1481 my ($self, $cb, $accept, $reject, $skip) = @_;
1482
1483 my $data;
1484 my $rbuf = \$self->{rbuf};
540 1485
541 sub { 1486 sub {
542 $_[0]{rbuf} =~ s/$eol// or return; 1487 # accept
1488 if ($$rbuf =~ $accept) {
1489 $data .= substr $$rbuf, 0, $+[0], "";
1490 $cb->($self, $data);
1491 return 1;
1492 }
1493
1494 # reject
1495 if ($reject && $$rbuf =~ $reject) {
1496 $self->_error (Errno::EBADMSG);
1497 }
543 1498
544 $cb->($_[0], $1, $2); 1499 # skip
1500 if ($skip && $$rbuf =~ $skip) {
1501 $data .= substr $$rbuf, 0, $+[0], "";
1502 }
1503
1504 ()
1505 }
1506};
1507
1508=item netstring => $cb->($handle, $string)
1509
1510A netstring (http://cr.yp.to/proto/netstrings.txt, this is not an endorsement).
1511
1512Throws an error with C<$!> set to EBADMSG on format violations.
1513
1514=cut
1515
1516register_read_type netstring => sub {
1517 my ($self, $cb) = @_;
1518
1519 sub {
1520 unless ($_[0]{rbuf} =~ s/^(0|[1-9][0-9]*)://) {
1521 if ($_[0]{rbuf} =~ /[^0-9]/) {
1522 $self->_error (Errno::EBADMSG);
1523 }
1524 return;
1525 }
1526
1527 my $len = $1;
1528
1529 $self->unshift_read (chunk => $len, sub {
1530 my $string = $_[1];
1531 $_[0]->unshift_read (chunk => 1, sub {
1532 if ($_[1] eq ",") {
1533 $cb->($_[0], $string);
1534 } else {
1535 $self->_error (Errno::EBADMSG);
1536 }
1537 });
1538 });
1539
545 1 1540 1
546 } 1541 }
547} 1542};
548 1543
549sub push_read_line { 1544=item packstring => $format, $cb->($handle, $string)
550 $_[0]->push_read (&_read_line);
551}
552 1545
553sub unshift_read_line { 1546An octet string prefixed with an encoded length. The encoding C<$format>
554 $_[0]->unshift_read (&_read_line); 1547uses the same format as a Perl C<pack> format, but must specify a single
555} 1548integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
1549optional C<!>, C<< < >> or C<< > >> modifier).
1550
1551For example, DNS over TCP uses a prefix of C<n> (2 octet network order),
1552EPP uses a prefix of C<N> (4 octtes).
1553
1554Example: read a block of data prefixed by its length in BER-encoded
1555format (very efficient).
1556
1557 $handle->push_read (packstring => "w", sub {
1558 my ($handle, $data) = @_;
1559 });
1560
1561=cut
1562
1563register_read_type packstring => sub {
1564 my ($self, $cb, $format) = @_;
1565
1566 sub {
1567 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1568 defined (my $len = eval { unpack $format, $_[0]{rbuf} })
1569 or return;
1570
1571 $format = length pack $format, $len;
1572
1573 # bypass unshift if we already have the remaining chunk
1574 if ($format + $len <= length $_[0]{rbuf}) {
1575 my $data = substr $_[0]{rbuf}, $format, $len;
1576 substr $_[0]{rbuf}, 0, $format + $len, "";
1577 $cb->($_[0], $data);
1578 } else {
1579 # remove prefix
1580 substr $_[0]{rbuf}, 0, $format, "";
1581
1582 # read remaining chunk
1583 $_[0]->unshift_read (chunk => $len, $cb);
1584 }
1585
1586 1
1587 }
1588};
1589
1590=item json => $cb->($handle, $hash_or_arrayref)
1591
1592Reads a JSON object or array, decodes it and passes it to the
1593callback. When a parse error occurs, an C<EBADMSG> error will be raised.
1594
1595If a C<json> object was passed to the constructor, then that will be used
1596for the final decode, otherwise it will create a JSON coder expecting UTF-8.
1597
1598This read type uses the incremental parser available with JSON version
15992.09 (and JSON::XS version 2.2) and above. You have to provide a
1600dependency on your own: this module will load the JSON module, but
1601AnyEvent does not depend on it itself.
1602
1603Since JSON texts are fully self-delimiting, the C<json> read and write
1604types are an ideal simple RPC protocol: just exchange JSON datagrams. See
1605the C<json> write type description, above, for an actual example.
1606
1607=cut
1608
1609register_read_type json => sub {
1610 my ($self, $cb) = @_;
1611
1612 my $json = $self->{json} ||= json_coder;
1613
1614 my $data;
1615 my $rbuf = \$self->{rbuf};
1616
1617 sub {
1618 my $ref = eval { $json->incr_parse ($self->{rbuf}) };
1619
1620 if ($ref) {
1621 $self->{rbuf} = $json->incr_text;
1622 $json->incr_text = "";
1623 $cb->($self, $ref);
1624
1625 1
1626 } elsif ($@) {
1627 # error case
1628 $json->incr_skip;
1629
1630 $self->{rbuf} = $json->incr_text;
1631 $json->incr_text = "";
1632
1633 $self->_error (Errno::EBADMSG);
1634
1635 ()
1636 } else {
1637 $self->{rbuf} = "";
1638
1639 ()
1640 }
1641 }
1642};
1643
1644=item storable => $cb->($handle, $ref)
1645
1646Deserialises a L<Storable> frozen representation as written by the
1647C<storable> write type (BER-encoded length prefix followed by nfreeze'd
1648data).
1649
1650Raises C<EBADMSG> error if the data could not be decoded.
1651
1652=cut
1653
1654register_read_type storable => sub {
1655 my ($self, $cb) = @_;
1656
1657 require Storable;
1658
1659 sub {
1660 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1661 defined (my $len = eval { unpack "w", $_[0]{rbuf} })
1662 or return;
1663
1664 my $format = length pack "w", $len;
1665
1666 # bypass unshift if we already have the remaining chunk
1667 if ($format + $len <= length $_[0]{rbuf}) {
1668 my $data = substr $_[0]{rbuf}, $format, $len;
1669 substr $_[0]{rbuf}, 0, $format + $len, "";
1670 $cb->($_[0], Storable::thaw ($data));
1671 } else {
1672 # remove prefix
1673 substr $_[0]{rbuf}, 0, $format, "";
1674
1675 # read remaining chunk
1676 $_[0]->unshift_read (chunk => $len, sub {
1677 if (my $ref = eval { Storable::thaw ($_[1]) }) {
1678 $cb->($_[0], $ref);
1679 } else {
1680 $self->_error (Errno::EBADMSG);
1681 }
1682 });
1683 }
1684
1685 1
1686 }
1687};
1688
1689=back
1690
1691=item custom read types - Package::anyevent_read_type $handle, $cb, @args
1692
1693Instead of one of the predefined types, you can also specify the name
1694of a package. AnyEvent will try to load the package and then expects to
1695find a function named C<anyevent_read_type> inside. If it isn't found, it
1696progressively tries to load the parent package until it either finds the
1697function (good) or runs out of packages (bad).
1698
1699Whenever this type is used, C<push_read> will invoke the function with the
1700handle object, the original callback and the remaining arguments.
1701
1702The function is supposed to return a callback (usually a closure) that
1703works as a plain read callback (see C<< ->push_read ($cb) >>), so you can
1704mentally treat the function as a "configurable read type to read callback"
1705converter.
1706
1707It should invoke the original callback when it is done reading (remember
1708to pass C<$handle> as first argument as all other callbacks do that,
1709although there is no strict requirement on this).
1710
1711For examples, see the source of this module (F<perldoc -m
1712AnyEvent::Handle>, search for C<register_read_type>)).
556 1713
557=item $handle->stop_read 1714=item $handle->stop_read
558 1715
559=item $handle->start_read 1716=item $handle->start_read
560 1717
561In rare cases you actually do not want to read anything form the 1718In rare cases you actually do not want to read anything from the
562socket. In this case you can call C<stop_read>. Neither C<on_read> no 1719socket. In this case you can call C<stop_read>. Neither C<on_read> nor
563any queued callbacks will be executed then. To start readign again, call 1720any queued callbacks will be executed then. To start reading again, call
564C<start_read>. 1721C<start_read>.
1722
1723Note that AnyEvent::Handle will automatically C<start_read> for you when
1724you change the C<on_read> callback or push/unshift a read callback, and it
1725will automatically C<stop_read> for you when neither C<on_read> is set nor
1726there are any read requests in the queue.
1727
1728These methods will have no effect when in TLS mode (as TLS doesn't support
1729half-duplex connections).
565 1730
566=cut 1731=cut
567 1732
568sub stop_read { 1733sub stop_read {
569 my ($self) = @_; 1734 my ($self) = @_;
570 1735
571 delete $self->{rw}; 1736 delete $self->{_rw} unless $self->{tls};
572} 1737}
573 1738
574sub start_read { 1739sub start_read {
575 my ($self) = @_; 1740 my ($self) = @_;
576 1741
577 unless ($self->{rw} || $self->{eof}) { 1742 unless ($self->{_rw} || $self->{_eof} || !$self->{fh}) {
578 Scalar::Util::weaken $self; 1743 Scalar::Util::weaken $self;
579 1744
580 $self->{rw} = AnyEvent->io (fh => $self->{fh}, poll => "r", cb => sub { 1745 $self->{_rw} = AE::io $self->{fh}, 0, sub {
1746 my $rbuf = \($self->{tls} ? my $buf : $self->{rbuf});
581 my $len = sysread $self->{fh}, $self->{rbuf}, $self->{read_size} || 8192, length $self->{rbuf}; 1747 my $len = sysread $self->{fh}, $$rbuf, $self->{read_size} || 8192, length $$rbuf;
582 1748
583 if ($len > 0) { 1749 if ($len > 0) {
584 if (defined $self->{rbuf_max}) { 1750 $self->{_activity} = $self->{_ractivity} = AE::now;
585 if ($self->{rbuf_max} < length $self->{rbuf}) { 1751
586 $! = &Errno::ENOSPC; return $self->error; 1752 if ($self->{tls}) {
587 } 1753 Net::SSLeay::BIO_write ($self->{_rbio}, $$rbuf);
1754
1755 &_dotls ($self);
1756 } else {
1757 $self->_drain_rbuf;
588 } 1758 }
589 1759
590 } elsif (defined $len) { 1760 } elsif (defined $len) {
591 $self->{eof} = 1;
592 delete $self->{rw}; 1761 delete $self->{_rw};
1762 $self->{_eof} = 1;
1763 $self->_drain_rbuf;
593 1764
594 } elsif ($! != EAGAIN && $! != EINTR) { 1765 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
595 return $self->error; 1766 return $self->_error ($!, 1);
596 } 1767 }
1768 };
1769 }
1770}
597 1771
1772our $ERROR_SYSCALL;
1773our $ERROR_WANT_READ;
1774
1775sub _tls_error {
1776 my ($self, $err) = @_;
1777
1778 return $self->_error ($!, 1)
1779 if $err == Net::SSLeay::ERROR_SYSCALL ();
1780
1781 my $err =Net::SSLeay::ERR_error_string (Net::SSLeay::ERR_get_error ());
1782
1783 # reduce error string to look less scary
1784 $err =~ s/^error:[0-9a-fA-F]{8}:[^:]+:([^:]+):/\L$1: /;
1785
1786 if ($self->{_on_starttls}) {
1787 (delete $self->{_on_starttls})->($self, undef, $err);
1788 &_freetls;
1789 } else {
1790 &_freetls;
1791 $self->_error (Errno::EPROTO, 1, $err);
1792 }
1793}
1794
1795# poll the write BIO and send the data if applicable
1796# also decode read data if possible
1797# this is basiclaly our TLS state machine
1798# more efficient implementations are possible with openssl,
1799# but not with the buggy and incomplete Net::SSLeay.
1800sub _dotls {
1801 my ($self) = @_;
1802
1803 my $tmp;
1804
1805 if (length $self->{_tls_wbuf}) {
1806 while (($tmp = Net::SSLeay::write ($self->{tls}, $self->{_tls_wbuf})) > 0) {
1807 substr $self->{_tls_wbuf}, 0, $tmp, "";
1808 }
1809
1810 $tmp = Net::SSLeay::get_error ($self->{tls}, $tmp);
1811 return $self->_tls_error ($tmp)
1812 if $tmp != $ERROR_WANT_READ
1813 && ($tmp != $ERROR_SYSCALL || $!);
1814 }
1815
1816 while (defined ($tmp = Net::SSLeay::read ($self->{tls}))) {
1817 unless (length $tmp) {
1818 $self->{_on_starttls}
1819 and (delete $self->{_on_starttls})->($self, undef, "EOF during handshake"); # ???
1820 &_freetls;
1821
1822 if ($self->{on_stoptls}) {
1823 $self->{on_stoptls}($self);
1824 return;
1825 } else {
1826 # let's treat SSL-eof as we treat normal EOF
1827 delete $self->{_rw};
1828 $self->{_eof} = 1;
1829 }
1830 }
1831
1832 $self->{_tls_rbuf} .= $tmp;
598 $self->_drain_rbuf; 1833 $self->_drain_rbuf;
1834 $self->{tls} or return; # tls session might have gone away in callback
1835 }
1836
1837 $tmp = Net::SSLeay::get_error ($self->{tls}, -1);
1838 return $self->_tls_error ($tmp)
1839 if $tmp != $ERROR_WANT_READ
1840 && ($tmp != $ERROR_SYSCALL || $!);
1841
1842 while (length ($tmp = Net::SSLeay::BIO_read ($self->{_wbio}))) {
1843 $self->{wbuf} .= $tmp;
1844 $self->_drain_wbuf;
1845 $self->{tls} or return; # tls session might have gone away in callback
1846 }
1847
1848 $self->{_on_starttls}
1849 and Net::SSLeay::state ($self->{tls}) == Net::SSLeay::ST_OK ()
1850 and (delete $self->{_on_starttls})->($self, 1, "TLS/SSL connection established");
1851}
1852
1853=item $handle->starttls ($tls[, $tls_ctx])
1854
1855Instead of starting TLS negotiation immediately when the AnyEvent::Handle
1856object is created, you can also do that at a later time by calling
1857C<starttls>.
1858
1859Starting TLS is currently an asynchronous operation - when you push some
1860write data and then call C<< ->starttls >> then TLS negotiation will start
1861immediately, after which the queued write data is then sent.
1862
1863The first argument is the same as the C<tls> constructor argument (either
1864C<"connect">, C<"accept"> or an existing Net::SSLeay object).
1865
1866The second argument is the optional C<AnyEvent::TLS> object that is used
1867when AnyEvent::Handle has to create its own TLS connection object, or
1868a hash reference with C<< key => value >> pairs that will be used to
1869construct a new context.
1870
1871The TLS connection object will end up in C<< $handle->{tls} >>, the TLS
1872context in C<< $handle->{tls_ctx} >> after this call and can be used or
1873changed to your liking. Note that the handshake might have already started
1874when this function returns.
1875
1876Due to bugs in OpenSSL, it might or might not be possible to do multiple
1877handshakes on the same stream. It is best to not attempt to use the
1878stream after stopping TLS.
1879
1880This method may invoke callbacks (and therefore the handle might be
1881destroyed after it returns).
1882
1883=cut
1884
1885our %TLS_CACHE; #TODO not yet documented, should we?
1886
1887sub starttls {
1888 my ($self, $tls, $ctx) = @_;
1889
1890 Carp::croak "It is an error to call starttls on an AnyEvent::Handle object while TLS is already active, caught"
1891 if $self->{tls};
1892
1893 $self->{tls} = $tls;
1894 $self->{tls_ctx} = $ctx if @_ > 2;
1895
1896 return unless $self->{fh};
1897
1898 require Net::SSLeay;
1899
1900 $ERROR_SYSCALL = Net::SSLeay::ERROR_SYSCALL ();
1901 $ERROR_WANT_READ = Net::SSLeay::ERROR_WANT_READ ();
1902
1903 $tls = delete $self->{tls};
1904 $ctx = $self->{tls_ctx};
1905
1906 local $Carp::CarpLevel = 1; # skip ourselves when creating a new context or session
1907
1908 if ("HASH" eq ref $ctx) {
1909 require AnyEvent::TLS;
1910
1911 if ($ctx->{cache}) {
1912 my $key = $ctx+0;
1913 $ctx = $TLS_CACHE{$key} ||= new AnyEvent::TLS %$ctx;
1914 } else {
1915 $ctx = new AnyEvent::TLS %$ctx;
1916 }
1917 }
1918
1919 $self->{tls_ctx} = $ctx || TLS_CTX ();
1920 $self->{tls} = $tls = $self->{tls_ctx}->_get_session ($tls, $self, $self->{peername});
1921
1922 # basically, this is deep magic (because SSL_read should have the same issues)
1923 # but the openssl maintainers basically said: "trust us, it just works".
1924 # (unfortunately, we have to hardcode constants because the abysmally misdesigned
1925 # and mismaintained ssleay-module doesn't even offer them).
1926 # http://www.mail-archive.com/openssl-dev@openssl.org/msg22420.html
1927 #
1928 # in short: this is a mess.
1929 #
1930 # note that we do not try to keep the length constant between writes as we are required to do.
1931 # we assume that most (but not all) of this insanity only applies to non-blocking cases,
1932 # and we drive openssl fully in blocking mode here. Or maybe we don't - openssl seems to
1933 # have identity issues in that area.
1934# Net::SSLeay::CTX_set_mode ($ssl,
1935# (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ENABLE_PARTIAL_WRITE () } || 1)
1936# | (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ACCEPT_MOVING_WRITE_BUFFER () } || 2));
1937 Net::SSLeay::CTX_set_mode ($tls, 1|2);
1938
1939 $self->{_rbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
1940 $self->{_wbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
1941
1942 Net::SSLeay::BIO_write ($self->{_rbio}, delete $self->{rbuf});
1943
1944 Net::SSLeay::set_bio ($tls, $self->{_rbio}, $self->{_wbio});
1945
1946 $self->{_on_starttls} = sub { $_[0]{on_starttls}(@_) }
1947 if $self->{on_starttls};
1948
1949 &_dotls; # need to trigger the initial handshake
1950 $self->start_read; # make sure we actually do read
1951}
1952
1953=item $handle->stoptls
1954
1955Shuts down the SSL connection - this makes a proper EOF handshake by
1956sending a close notify to the other side, but since OpenSSL doesn't
1957support non-blocking shut downs, it is not guaranteed that you can re-use
1958the stream afterwards.
1959
1960This method may invoke callbacks (and therefore the handle might be
1961destroyed after it returns).
1962
1963=cut
1964
1965sub stoptls {
1966 my ($self) = @_;
1967
1968 if ($self->{tls} && $self->{fh}) {
1969 Net::SSLeay::shutdown ($self->{tls});
1970
1971 &_dotls;
1972
1973# # we don't give a shit. no, we do, but we can't. no...#d#
1974# # we, we... have to use openssl :/#d#
1975# &_freetls;#d#
1976 }
1977}
1978
1979sub _freetls {
1980 my ($self) = @_;
1981
1982 return unless $self->{tls};
1983
1984 $self->{tls_ctx}->_put_session (delete $self->{tls})
1985 if $self->{tls} > 0;
1986
1987 delete @$self{qw(_rbio _wbio _tls_wbuf _on_starttls)};
1988}
1989
1990sub DESTROY {
1991 my ($self) = @_;
1992
1993 &_freetls;
1994
1995 my $linger = exists $self->{linger} ? $self->{linger} : 3600;
1996
1997 if ($linger && length $self->{wbuf} && $self->{fh}) {
1998 my $fh = delete $self->{fh};
1999 my $wbuf = delete $self->{wbuf};
2000
2001 my @linger;
2002
2003 push @linger, AE::io $fh, 1, sub {
2004 my $len = syswrite $fh, $wbuf, length $wbuf;
2005
2006 if ($len > 0) {
2007 substr $wbuf, 0, $len, "";
2008 } else {
2009 @linger = (); # end
2010 }
599 }); 2011 };
2012 push @linger, AE::timer $linger, 0, sub {
2013 @linger = ();
2014 };
2015 }
2016}
2017
2018=item $handle->destroy
2019
2020Shuts down the handle object as much as possible - this call ensures that
2021no further callbacks will be invoked and as many resources as possible
2022will be freed. Any method you will call on the handle object after
2023destroying it in this way will be silently ignored (and it will return the
2024empty list).
2025
2026Normally, you can just "forget" any references to an AnyEvent::Handle
2027object and it will simply shut down. This works in fatal error and EOF
2028callbacks, as well as code outside. It does I<NOT> work in a read or write
2029callback, so when you want to destroy the AnyEvent::Handle object from
2030within such an callback. You I<MUST> call C<< ->destroy >> explicitly in
2031that case.
2032
2033Destroying the handle object in this way has the advantage that callbacks
2034will be removed as well, so if those are the only reference holders (as
2035is common), then one doesn't need to do anything special to break any
2036reference cycles.
2037
2038The handle might still linger in the background and write out remaining
2039data, as specified by the C<linger> option, however.
2040
2041=cut
2042
2043sub destroy {
2044 my ($self) = @_;
2045
2046 $self->DESTROY;
2047 %$self = ();
2048 bless $self, "AnyEvent::Handle::destroyed";
2049}
2050
2051sub AnyEvent::Handle::destroyed::AUTOLOAD {
2052 #nop
2053}
2054
2055=item $handle->destroyed
2056
2057Returns false as long as the handle hasn't been destroyed by a call to C<<
2058->destroy >>, true otherwise.
2059
2060Can be useful to decide whether the handle is still valid after some
2061callback possibly destroyed the handle. For example, C<< ->push_write >>,
2062C<< ->starttls >> and other methods can call user callbacks, which in turn
2063can destroy the handle, so work can be avoided by checking sometimes:
2064
2065 $hdl->starttls ("accept");
2066 return if $hdl->destroyed;
2067 $hdl->push_write (...
2068
2069Note that the call to C<push_write> will silently be ignored if the handle
2070has been destroyed, so often you can just ignore the possibility of the
2071handle being destroyed.
2072
2073=cut
2074
2075sub destroyed { 0 }
2076sub AnyEvent::Handle::destroyed::destroyed { 1 }
2077
2078=item AnyEvent::Handle::TLS_CTX
2079
2080This function creates and returns the AnyEvent::TLS object used by default
2081for TLS mode.
2082
2083The context is created by calling L<AnyEvent::TLS> without any arguments.
2084
2085=cut
2086
2087our $TLS_CTX;
2088
2089sub TLS_CTX() {
2090 $TLS_CTX ||= do {
2091 require AnyEvent::TLS;
2092
2093 new AnyEvent::TLS
600 } 2094 }
601} 2095}
602 2096
603=back 2097=back
604 2098
2099
2100=head1 NONFREQUENTLY ASKED QUESTIONS
2101
2102=over 4
2103
2104=item I C<undef> the AnyEvent::Handle reference inside my callback and
2105still get further invocations!
2106
2107That's because AnyEvent::Handle keeps a reference to itself when handling
2108read or write callbacks.
2109
2110It is only safe to "forget" the reference inside EOF or error callbacks,
2111from within all other callbacks, you need to explicitly call the C<<
2112->destroy >> method.
2113
2114=item I get different callback invocations in TLS mode/Why can't I pause
2115reading?
2116
2117Unlike, say, TCP, TLS connections do not consist of two independent
2118communication channels, one for each direction. Or put differently, the
2119read and write directions are not independent of each other: you cannot
2120write data unless you are also prepared to read, and vice versa.
2121
2122This means that, in TLS mode, you might get C<on_error> or C<on_eof>
2123callback invocations when you are not expecting any read data - the reason
2124is that AnyEvent::Handle always reads in TLS mode.
2125
2126During the connection, you have to make sure that you always have a
2127non-empty read-queue, or an C<on_read> watcher. At the end of the
2128connection (or when you no longer want to use it) you can call the
2129C<destroy> method.
2130
2131=item How do I read data until the other side closes the connection?
2132
2133If you just want to read your data into a perl scalar, the easiest way
2134to achieve this is by setting an C<on_read> callback that does nothing,
2135clearing the C<on_eof> callback and in the C<on_error> callback, the data
2136will be in C<$_[0]{rbuf}>:
2137
2138 $handle->on_read (sub { });
2139 $handle->on_eof (undef);
2140 $handle->on_error (sub {
2141 my $data = delete $_[0]{rbuf};
2142 });
2143
2144The reason to use C<on_error> is that TCP connections, due to latencies
2145and packets loss, might get closed quite violently with an error, when in
2146fact all data has been received.
2147
2148It is usually better to use acknowledgements when transferring data,
2149to make sure the other side hasn't just died and you got the data
2150intact. This is also one reason why so many internet protocols have an
2151explicit QUIT command.
2152
2153=item I don't want to destroy the handle too early - how do I wait until
2154all data has been written?
2155
2156After writing your last bits of data, set the C<on_drain> callback
2157and destroy the handle in there - with the default setting of
2158C<low_water_mark> this will be called precisely when all data has been
2159written to the socket:
2160
2161 $handle->push_write (...);
2162 $handle->on_drain (sub {
2163 warn "all data submitted to the kernel\n";
2164 undef $handle;
2165 });
2166
2167If you just want to queue some data and then signal EOF to the other side,
2168consider using C<< ->push_shutdown >> instead.
2169
2170=item I want to contact a TLS/SSL server, I don't care about security.
2171
2172If your TLS server is a pure TLS server (e.g. HTTPS) that only speaks TLS,
2173connect to it and then create the AnyEvent::Handle with the C<tls>
2174parameter:
2175
2176 tcp_connect $host, $port, sub {
2177 my ($fh) = @_;
2178
2179 my $handle = new AnyEvent::Handle
2180 fh => $fh,
2181 tls => "connect",
2182 on_error => sub { ... };
2183
2184 $handle->push_write (...);
2185 };
2186
2187=item I want to contact a TLS/SSL server, I do care about security.
2188
2189Then you should additionally enable certificate verification, including
2190peername verification, if the protocol you use supports it (see
2191L<AnyEvent::TLS>, C<verify_peername>).
2192
2193E.g. for HTTPS:
2194
2195 tcp_connect $host, $port, sub {
2196 my ($fh) = @_;
2197
2198 my $handle = new AnyEvent::Handle
2199 fh => $fh,
2200 peername => $host,
2201 tls => "connect",
2202 tls_ctx => { verify => 1, verify_peername => "https" },
2203 ...
2204
2205Note that you must specify the hostname you connected to (or whatever
2206"peername" the protocol needs) as the C<peername> argument, otherwise no
2207peername verification will be done.
2208
2209The above will use the system-dependent default set of trusted CA
2210certificates. If you want to check against a specific CA, add the
2211C<ca_file> (or C<ca_cert>) arguments to C<tls_ctx>:
2212
2213 tls_ctx => {
2214 verify => 1,
2215 verify_peername => "https",
2216 ca_file => "my-ca-cert.pem",
2217 },
2218
2219=item I want to create a TLS/SSL server, how do I do that?
2220
2221Well, you first need to get a server certificate and key. You have
2222three options: a) ask a CA (buy one, use cacert.org etc.) b) create a
2223self-signed certificate (cheap. check the search engine of your choice,
2224there are many tutorials on the net) or c) make your own CA (tinyca2 is a
2225nice program for that purpose).
2226
2227Then create a file with your private key (in PEM format, see
2228L<AnyEvent::TLS>), followed by the certificate (also in PEM format). The
2229file should then look like this:
2230
2231 -----BEGIN RSA PRIVATE KEY-----
2232 ...header data
2233 ... lots of base64'y-stuff
2234 -----END RSA PRIVATE KEY-----
2235
2236 -----BEGIN CERTIFICATE-----
2237 ... lots of base64'y-stuff
2238 -----END CERTIFICATE-----
2239
2240The important bits are the "PRIVATE KEY" and "CERTIFICATE" parts. Then
2241specify this file as C<cert_file>:
2242
2243 tcp_server undef, $port, sub {
2244 my ($fh) = @_;
2245
2246 my $handle = new AnyEvent::Handle
2247 fh => $fh,
2248 tls => "accept",
2249 tls_ctx => { cert_file => "my-server-keycert.pem" },
2250 ...
2251
2252When you have intermediate CA certificates that your clients might not
2253know about, just append them to the C<cert_file>.
2254
2255=back
2256
2257
2258=head1 SUBCLASSING AnyEvent::Handle
2259
2260In many cases, you might want to subclass AnyEvent::Handle.
2261
2262To make this easier, a given version of AnyEvent::Handle uses these
2263conventions:
2264
2265=over 4
2266
2267=item * all constructor arguments become object members.
2268
2269At least initially, when you pass a C<tls>-argument to the constructor it
2270will end up in C<< $handle->{tls} >>. Those members might be changed or
2271mutated later on (for example C<tls> will hold the TLS connection object).
2272
2273=item * other object member names are prefixed with an C<_>.
2274
2275All object members not explicitly documented (internal use) are prefixed
2276with an underscore character, so the remaining non-C<_>-namespace is free
2277for use for subclasses.
2278
2279=item * all members not documented here and not prefixed with an underscore
2280are free to use in subclasses.
2281
2282Of course, new versions of AnyEvent::Handle may introduce more "public"
2283member variables, but that's just life. At least it is documented.
2284
2285=back
2286
605=head1 AUTHOR 2287=head1 AUTHOR
606 2288
607Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>. 2289Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>.
608 2290
609=cut 2291=cut

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines