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 1.12 by elmex, Thu May 15 09:03:43 2008 UTC vs.
Revision 1.214 by root, Sun Jan 16 17:12:27 2011 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
17=cut
18
19our $VERSION = '0.02';
20 4
21=head1 SYNOPSIS 5=head1 SYNOPSIS
22 6
23 use AnyEvent; 7 use AnyEvent;
24 use AnyEvent::Handle; 8 use AnyEvent::Handle;
25 9
26 my $cv = AnyEvent->condvar; 10 my $cv = AnyEvent->condvar;
27 11
28 my $ae_fh = AnyEvent::Handle->new (fh => \*STDIN); 12 my $hdl; $hdl = new AnyEvent::Handle
29
30 #TODO
31
32 # or use the constructor to pass the callback:
33
34 my $ae_fh2 =
35 AnyEvent::Handle->new (
36 fh => \*STDIN, 13 fh => \*STDIN,
37 on_eof => sub { 14 on_error => sub {
38 $cv->broadcast; 15 my ($hdl, $fatal, $msg) = @_;
39 }, 16 warn "got error $msg\n";
40 #TODO 17 $hdl->destroy;
18 $cv->send;
41 ); 19 };
42 20
43 $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;
44 32
45=head1 DESCRIPTION 33=head1 DESCRIPTION
46 34
47This 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
48filehandles (and sockets, see L<AnyEvent::Socket> for an easy way to make 36stream-based filehandles (sockets, pipes, and other stream things).
49non-blocking resolves and connects).
50 37
38The L<AnyEvent::Intro> tutorial contains some well-documented
39AnyEvent::Handle examples.
40
51In the following, when the documentation refers to of "bytes" then this 41In the following, where the documentation refers to "bytes", it means
52means characters. As sysread and syswrite are used for all I/O, their 42characters. As sysread and syswrite are used for all I/O, their
53treatment 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.
54 47
55All callbacks will be invoked with the handle object as their first 48All callbacks will be invoked with the handle object as their first
56argument. 49argument.
57 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
80sub MAX_READ_SIZE() { 131072 }
81
58=head1 METHODS 82=head1 METHODS
59 83
60=over 4 84=over 4
61 85
62=item B<new (%args)> 86=item $handle = B<new> AnyEvent::Handle fh => $filehandle, key => value...
63 87
64The constructor supports these arguments (all as key => value pairs). 88The constructor supports these arguments (all as C<< key => value >> pairs).
65 89
66=over 4 90=over 4
67 91
68=item fh => $filehandle [MANDATORY] 92=item fh => $filehandle [C<fh> or C<connect> MANDATORY]
69 93
70The filehandle this L<AnyEvent::Handle> object will operate on. 94The filehandle this L<AnyEvent::Handle> object will operate on.
71
72NOTE: The filehandle will be set to non-blocking (using 95NOTE: The filehandle will be set to non-blocking mode (using
73AnyEvent::Util::fh_nonblocking). 96C<AnyEvent::Util::fh_nonblocking>) by the constructor and needs to stay in
97that mode.
74 98
75=item on_eof => $cb->($self) [MANDATORY] 99=item connect => [$host, $service] [C<fh> or C<connect> MANDATORY]
76 100
77Set the callback to be called on EOF. 101Try to connect to the specified host and service (port), using
102C<AnyEvent::Socket::tcp_connect>. The C<$host> additionally becomes the
103default C<peername>.
78 104
105You have to specify either this parameter, or C<fh>, above.
106
107It is possible to push requests on the read and write queues, and modify
108properties of the stream, even while AnyEvent::Handle is connecting.
109
110When this parameter is specified, then the C<on_prepare>,
111C<on_connect_error> and C<on_connect> callbacks will be called under the
112appropriate circumstances:
113
114=over 4
115
79=item on_error => $cb->($self) 116=item on_prepare => $cb->($handle)
80 117
118This (rarely used) callback is called before a new connection is
119attempted, but after the file handle has been created (you can access that
120file handle via C<< $handle->{fh} >>). It could be used to prepare the
121file handle with parameters required for the actual connect (as opposed to
122settings that can be changed when the connection is already established).
123
124The return value of this callback should be the connect timeout value in
125seconds (or C<0>, or C<undef>, or the empty list, to indicate that the
126default timeout is to be used).
127
128=item on_connect => $cb->($handle, $host, $port, $retry->())
129
130This callback is called when a connection has been successfully established.
131
132The peer's numeric host and port (the socket peername) are passed as
133parameters, together with a retry callback.
134
135If, for some reason, the handle is not acceptable, calling C<$retry>
136will continue with the next connection target (in case of multi-homed
137hosts or SRV records there can be multiple connection endpoints). At the
138time it is called the read and write queues, eof status, tls status and
139similar properties of the handle will have been reset.
140
141In most cases, you should ignore the C<$retry> parameter.
142
143=item on_connect_error => $cb->($handle, $message)
144
145This callback is called when the connection could not be
146established. C<$!> will contain the relevant error code, and C<$message> a
147message describing it (usually the same as C<"$!">).
148
149If this callback isn't specified, then C<on_error> will be called with a
150fatal error instead.
151
152=back
153
154=item on_error => $cb->($handle, $fatal, $message)
155
81This is the fatal error callback, that is called when, well, a fatal error 156This is the error callback, which is called when, well, some error
82ocurs, such as not being able to resolve the hostname, failure to connect 157occured, such as not being able to resolve the hostname, failure to
83or a read error. 158connect, or a read error.
84 159
85The object will not be in a usable state when this callback has been 160Some errors are fatal (which is indicated by C<$fatal> being true). On
86called. 161fatal errors the handle object will be destroyed (by a call to C<< ->
162destroy >>) after invoking the error callback (which means you are free to
163examine the handle object). Examples of fatal errors are an EOF condition
164with active (but unsatisfiable) read watchers (C<EPIPE>) or I/O errors. In
165cases where the other side can close the connection at will, it is
166often easiest to not report C<EPIPE> errors in this callback.
87 167
168AnyEvent::Handle tries to find an appropriate error code for you to check
169against, but in some cases (TLS errors), this does not work well. It is
170recommended to always output the C<$message> argument in human-readable
171error messages (it's usually the same as C<"$!">).
172
173Non-fatal errors can be retried by returning, but it is recommended
174to simply ignore this parameter and instead abondon the handle object
175when this callback is invoked. Examples of non-fatal errors are timeouts
176C<ETIMEDOUT>) or badly-formatted data (C<EBADMSG>).
177
88On callback entrance, the value of C<$!> contains the operating system 178On entry to the callback, the value of C<$!> contains the operating
89error (or C<ENOSPC> or C<EPIPE>). 179system error code (or C<ENOSPC>, C<EPIPE>, C<ETIMEDOUT>, C<EBADMSG> or
180C<EPROTO>).
90 181
91While not mandatory, it is I<highly> recommended to set this callback, as 182While not mandatory, it is I<highly> recommended to set this callback, as
92you will not be notified of errors otherwise. The default simply calls 183you will not be notified of errors otherwise. The default just calls
93die. 184C<croak>.
94 185
95=item on_read => $cb->($self) 186=item on_read => $cb->($handle)
96 187
97This sets the default read callback, which is called when data arrives 188This sets the default read callback, which is called when data arrives
98and no read request is in the queue. 189and no read request is in the queue (unlike read queue callbacks, this
190callback will only be called when at least one octet of data is in the
191read buffer).
99 192
100To access (and remove data from) the read buffer, use the C<< ->rbuf >> 193To access (and remove data from) the read buffer, use the C<< ->rbuf >>
101method or acces sthe C<$self->{rbuf}> member directly. 194method or access the C<< $handle->{rbuf} >> member directly. Note that you
195must not enlarge or modify the read buffer, you can only remove data at
196the beginning from it.
102 197
198You can also call C<< ->push_read (...) >> or any other function that
199modifies the read queue. Or do both. Or ...
200
103When an EOF condition is detected then AnyEvent::Handle will first try to 201When an EOF condition is detected, AnyEvent::Handle will first try to
104feed all the remaining data to the queued callbacks and C<on_read> before 202feed all the remaining data to the queued callbacks and C<on_read> before
105calling the C<on_eof> callback. If no progress can be made, then a fatal 203calling the C<on_eof> callback. If no progress can be made, then a fatal
106error will be raised (with C<$!> set to C<EPIPE>). 204error will be raised (with C<$!> set to C<EPIPE>).
107 205
206Note that, unlike requests in the read queue, an C<on_read> callback
207doesn't mean you I<require> some data: if there is an EOF and there
208are outstanding read requests then an error will be flagged. With an
209C<on_read> callback, the C<on_eof> callback will be invoked.
210
211=item on_eof => $cb->($handle)
212
213Set the callback to be called when an end-of-file condition is detected,
214i.e. in the case of a socket, when the other side has closed the
215connection cleanly, and there are no outstanding read requests in the
216queue (if there are read requests, then an EOF counts as an unexpected
217connection close and will be flagged as an error).
218
219For sockets, this just means that the other side has stopped sending data,
220you can still try to write data, and, in fact, one can return from the EOF
221callback and continue writing data, as only the read part has been shut
222down.
223
224If an EOF condition has been detected but no C<on_eof> callback has been
225set, then a fatal error will be raised with C<$!> set to <0>.
226
108=item on_drain => $cb->() 227=item on_drain => $cb->($handle)
109 228
110This sets the callback that is called when the write buffer becomes empty 229This sets the callback that is called when the write buffer becomes empty
111(or when the callback is set and the buffer is empty already). 230(or immediately if the buffer is empty already).
112 231
113To append to the write buffer, use the C<< ->push_write >> method. 232To append to the write buffer, use the C<< ->push_write >> method.
233
234This callback is useful when you don't want to put all of your write data
235into the queue at once, for example, when you want to write the contents
236of some file to the socket you might not want to read the whole file into
237memory and push it into the queue, but instead only read more data from
238the file when the write queue becomes empty.
239
240=item timeout => $fractional_seconds
241
242=item rtimeout => $fractional_seconds
243
244=item wtimeout => $fractional_seconds
245
246If non-zero, then these enables an "inactivity" timeout: whenever this
247many seconds pass without a successful read or write on the underlying
248file handle (or a call to C<timeout_reset>), the C<on_timeout> callback
249will be invoked (and if that one is missing, a non-fatal C<ETIMEDOUT>
250error will be raised).
251
252There are three variants of the timeouts that work independently
253of each other, for both read and write, just read, and just write:
254C<timeout>, C<rtimeout> and C<wtimeout>, with corresponding callbacks
255C<on_timeout>, C<on_rtimeout> and C<on_wtimeout>, and reset functions
256C<timeout_reset>, C<rtimeout_reset>, and C<wtimeout_reset>.
257
258Note that timeout processing is active even when you do not have
259any outstanding read or write requests: If you plan to keep the connection
260idle then you should disable the timeout temporarily or ignore the timeout
261in the C<on_timeout> callback, in which case AnyEvent::Handle will simply
262restart the timeout.
263
264Zero (the default) disables this timeout.
265
266=item on_timeout => $cb->($handle)
267
268Called whenever the inactivity timeout passes. If you return from this
269callback, then the timeout will be reset as if some activity had happened,
270so this condition is not fatal in any way.
114 271
115=item rbuf_max => <bytes> 272=item rbuf_max => <bytes>
116 273
117If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>) 274If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>)
118when the read buffer ever (strictly) exceeds this size. This is useful to 275when the read buffer ever (strictly) exceeds this size. This is useful to
119avoid denial-of-service attacks. 276avoid some forms of denial-of-service attacks.
120 277
121For example, a server accepting connections from untrusted sources should 278For example, a server accepting connections from untrusted sources should
122be configured to accept only so-and-so much data that it cannot act on 279be configured to accept only so-and-so much data that it cannot act on
123(for example, when expecting a line, an attacker could send an unlimited 280(for example, when expecting a line, an attacker could send an unlimited
124amount of data without a callback ever being called as long as the line 281amount of data without a callback ever being called as long as the line
125isn't finished). 282isn't finished).
126 283
284=item wbuf_max => <bytes>
285
286If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>)
287when the write buffer ever (strictly) exceeds this size. This is useful to
288avoid some forms of denial-of-service attacks.
289
290Although the units of this parameter is bytes, this is the I<raw> number
291of bytes not yet accepted by the kernel. This can make a difference when
292you e.g. use TLS, as TLS typically makes your write data larger (but it
293can also make it smaller due to compression).
294
295As an example of when this limit is useful, take a chat server that sends
296chat messages to a client. If the client does not read those in a timely
297manner then the send buffer in the server would grow unbounded.
298
299=item autocork => <boolean>
300
301When disabled (the default), C<push_write> will try to immediately
302write the data to the handle if possible. This avoids having to register
303a write watcher and wait for the next event loop iteration, but can
304be inefficient if you write multiple small chunks (on the wire, this
305disadvantage is usually avoided by your kernel's nagle algorithm, see
306C<no_delay>, but this option can save costly syscalls).
307
308When enabled, writes will always be queued till the next event loop
309iteration. This is efficient when you do many small writes per iteration,
310but less efficient when you do a single write only per iteration (or when
311the write buffer often is full). It also increases write latency.
312
313=item no_delay => <boolean>
314
315When doing small writes on sockets, your operating system kernel might
316wait a bit for more data before actually sending it out. This is called
317the Nagle algorithm, and usually it is beneficial.
318
319In some situations you want as low a delay as possible, which can be
320accomplishd by setting this option to a true value.
321
322The default is your operating system's default behaviour (most likely
323enabled). This option explicitly enables or disables it, if possible.
324
325=item keepalive => <boolean>
326
327Enables (default disable) the SO_KEEPALIVE option on the stream socket:
328normally, TCP connections have no time-out once established, so TCP
329connections, once established, can stay alive forever even when the other
330side has long gone. TCP keepalives are a cheap way to take down long-lived
331TCP connections when the other side becomes unreachable. While the default
332is OS-dependent, TCP keepalives usually kick in after around two hours,
333and, if the other side doesn't reply, take down the TCP connection some 10
334to 15 minutes later.
335
336It is harmless to specify this option for file handles that do not support
337keepalives, and enabling it on connections that are potentially long-lived
338is usually a good idea.
339
340=item oobinline => <boolean>
341
342BSD majorly fucked up the implementation of TCP urgent data. The result
343is that almost no OS implements TCP according to the specs, and every OS
344implements it slightly differently.
345
346If you want to handle TCP urgent data, then setting this flag (the default
347is enabled) gives you the most portable way of getting urgent data, by
348putting it into the stream.
349
350Since BSD emulation of OOB data on top of TCP's urgent data can have
351security implications, AnyEvent::Handle sets this flag automatically
352unless explicitly specified. Note that setting this flag after
353establishing a connection I<may> be a bit too late (data loss could
354already have occured on BSD systems), but at least it will protect you
355from most attacks.
356
127=item read_size => <bytes> 357=item read_size => <bytes>
128 358
129The default read block size (the amount of bytes this module will try to read 359The initial read block size, the number of bytes this module will try to
130on each [loop iteration). Default: C<4096>. 360read during each loop iteration. Each handle object will consume at least
361this amount of memory for the read buffer as well, so when handling many
362connections requirements). See also C<max_read_size>. Default: C<2048>.
363
364=item max_read_size => <bytes>
365
366The maximum read buffer size used by the dynamic adjustment
367algorithm: Each time AnyEvent::Handle can read C<read_size> bytes in
368one go it will double C<read_size> up to the maximum given by this
369option. Default: C<131072> or C<read_size>, whichever is higher.
131 370
132=item low_water_mark => <bytes> 371=item low_water_mark => <bytes>
133 372
134Sets the amount of bytes (default: C<0>) that make up an "empty" write 373Sets the number of bytes (default: C<0>) that make up an "empty" write
135buffer: If the write reaches this size or gets even samller it is 374buffer: If the buffer reaches this size or gets even samller it is
136considered empty. 375considered empty.
376
377Sometimes it can be beneficial (for performance reasons) to add data to
378the write buffer before it is fully drained, but this is a rare case, as
379the operating system kernel usually buffers data as well, so the default
380is good in almost all cases.
381
382=item linger => <seconds>
383
384If this is non-zero (default: C<3600>), the destructor of the
385AnyEvent::Handle object will check whether there is still outstanding
386write data and will install a watcher that will write this data to the
387socket. No errors will be reported (this mostly matches how the operating
388system treats outstanding data at socket close time).
389
390This will not work for partial TLS data that could not be encoded
391yet. This data will be lost. Calling the C<stoptls> method in time might
392help.
393
394=item peername => $string
395
396A string used to identify the remote site - usually the DNS hostname
397(I<not> IDN!) used to create the connection, rarely the IP address.
398
399Apart from being useful in error messages, this string is also used in TLS
400peername verification (see C<verify_peername> in L<AnyEvent::TLS>). This
401verification will be skipped when C<peername> is not specified or is
402C<undef>.
403
404=item tls => "accept" | "connect" | Net::SSLeay::SSL object
405
406When this parameter is given, it enables TLS (SSL) mode, that means
407AnyEvent will start a TLS handshake as soon as the connection has been
408established and will transparently encrypt/decrypt data afterwards.
409
410All TLS protocol errors will be signalled as C<EPROTO>, with an
411appropriate error message.
412
413TLS mode requires Net::SSLeay to be installed (it will be loaded
414automatically when you try to create a TLS handle): this module doesn't
415have a dependency on that module, so if your module requires it, you have
416to add the dependency yourself.
417
418Unlike TCP, TLS has a server and client side: for the TLS server side, use
419C<accept>, and for the TLS client side of a connection, use C<connect>
420mode.
421
422You can also provide your own TLS connection object, but you have
423to make sure that you call either C<Net::SSLeay::set_connect_state>
424or C<Net::SSLeay::set_accept_state> on it before you pass it to
425AnyEvent::Handle. Also, this module will take ownership of this connection
426object.
427
428At some future point, AnyEvent::Handle might switch to another TLS
429implementation, then the option to use your own session object will go
430away.
431
432B<IMPORTANT:> since Net::SSLeay "objects" are really only integers,
433passing in the wrong integer will lead to certain crash. This most often
434happens when one uses a stylish C<< tls => 1 >> and is surprised about the
435segmentation fault.
436
437Use the C<< ->starttls >> method if you need to start TLS negotiation later.
438
439=item tls_ctx => $anyevent_tls
440
441Use the given C<AnyEvent::TLS> object to create the new TLS connection
442(unless a connection object was specified directly). If this
443parameter is missing (or C<undef>), then AnyEvent::Handle will use
444C<AnyEvent::Handle::TLS_CTX>.
445
446Instead of an object, you can also specify a hash reference with C<< key
447=> value >> pairs. Those will be passed to L<AnyEvent::TLS> to create a
448new TLS context object.
449
450=item on_starttls => $cb->($handle, $success[, $error_message])
451
452This callback will be invoked when the TLS/SSL handshake has finished. If
453C<$success> is true, then the TLS handshake succeeded, otherwise it failed
454(C<on_stoptls> will not be called in this case).
455
456The session in C<< $handle->{tls} >> can still be examined in this
457callback, even when the handshake was not successful.
458
459TLS handshake failures will not cause C<on_error> to be invoked when this
460callback is in effect, instead, the error message will be passed to C<on_starttls>.
461
462Without this callback, handshake failures lead to C<on_error> being
463called as usual.
464
465Note that you cannot just call C<starttls> again in this callback. If you
466need to do that, start an zero-second timer instead whose callback can
467then call C<< ->starttls >> again.
468
469=item on_stoptls => $cb->($handle)
470
471When a SSLv3/TLS shutdown/close notify/EOF is detected and this callback is
472set, then it will be invoked after freeing the TLS session. If it is not,
473then a TLS shutdown condition will be treated like a normal EOF condition
474on the handle.
475
476The session in C<< $handle->{tls} >> can still be examined in this
477callback.
478
479This callback will only be called on TLS shutdowns, not when the
480underlying handle signals EOF.
481
482=item json => JSON or JSON::XS object
483
484This is the json coder object used by the C<json> read and write types.
485
486If you don't supply it, then AnyEvent::Handle will create and use a
487suitable one (on demand), which will write and expect UTF-8 encoded JSON
488texts.
489
490Note that you are responsible to depend on the JSON module if you want to
491use this functionality, as AnyEvent does not have a dependency itself.
137 492
138=back 493=back
139 494
140=cut 495=cut
141 496
142sub new { 497sub new {
143 my $class = shift; 498 my $class = shift;
144
145 my $self = bless { @_ }, $class; 499 my $self = bless { @_ }, $class;
146 500
147 $self->{fh} or Carp::croak "mandatory argument fh is missing"; 501 if ($self->{fh}) {
502 $self->_start;
503 return unless $self->{fh}; # could be gone by now
504
505 } elsif ($self->{connect}) {
506 require AnyEvent::Socket;
507
508 $self->{peername} = $self->{connect}[0]
509 unless exists $self->{peername};
510
511 $self->{_skip_drain_rbuf} = 1;
512
513 {
514 Scalar::Util::weaken (my $self = $self);
515
516 $self->{_connect} =
517 AnyEvent::Socket::tcp_connect (
518 $self->{connect}[0],
519 $self->{connect}[1],
520 sub {
521 my ($fh, $host, $port, $retry) = @_;
522
523 delete $self->{_connect}; # no longer needed
524
525 if ($fh) {
526 $self->{fh} = $fh;
527
528 delete $self->{_skip_drain_rbuf};
529 $self->_start;
530
531 $self->{on_connect}
532 and $self->{on_connect}($self, $host, $port, sub {
533 delete @$self{qw(fh _tw _rtw _wtw _ww _rw _eof _queue rbuf _wbuf tls _tls_rbuf _tls_wbuf)};
534 $self->{_skip_drain_rbuf} = 1;
535 &$retry;
536 });
537
538 } else {
539 if ($self->{on_connect_error}) {
540 $self->{on_connect_error}($self, "$!");
541 $self->destroy;
542 } else {
543 $self->_error ($!, 1);
544 }
545 }
546 },
547 sub {
548 local $self->{fh} = $_[0];
549
550 $self->{on_prepare}
551 ? $self->{on_prepare}->($self)
552 : ()
553 }
554 );
555 }
556
557 } else {
558 Carp::croak "AnyEvent::Handle: either an existing fh or the connect parameter must be specified";
559 }
560
561 $self
562}
563
564sub _start {
565 my ($self) = @_;
566
567 # too many clueless people try to use udp and similar sockets
568 # with AnyEvent::Handle, do them a favour.
569 my $type = getsockopt $self->{fh}, Socket::SOL_SOCKET (), Socket::SO_TYPE ();
570 Carp::croak "AnyEvent::Handle: only stream sockets supported, anything else will NOT work!"
571 if Socket::SOCK_STREAM () != (unpack "I", $type) && defined $type;
148 572
149 AnyEvent::Util::fh_nonblocking $self->{fh}, 1; 573 AnyEvent::Util::fh_nonblocking $self->{fh}, 1;
150 574
151 $self->on_eof ((delete $self->{on_eof} ) or Carp::croak "mandatory argument on_eof is missing"); 575 $self->{_activity} =
576 $self->{_ractivity} =
577 $self->{_wactivity} = AE::now;
152 578
153 $self->on_error (delete $self->{on_error}) if $self->{on_error}; 579 $self->{read_size} ||= 2048;
580 $self->{max_read_size} = $self->{read_size}
581 if $self->{read_size} > ($self->{max_read_size} || MAX_READ_SIZE);
582
583 $self->timeout (delete $self->{timeout} ) if $self->{timeout};
584 $self->rtimeout (delete $self->{rtimeout} ) if $self->{rtimeout};
585 $self->wtimeout (delete $self->{wtimeout} ) if $self->{wtimeout};
586
587 $self->no_delay (delete $self->{no_delay} ) if exists $self->{no_delay} && $self->{no_delay};
588 $self->keepalive (delete $self->{keepalive}) if exists $self->{keepalive} && $self->{keepalive};
589
590 $self->oobinline (exists $self->{oobinline} ? delete $self->{oobinline} : 1);
591
592 $self->starttls (delete $self->{tls}, delete $self->{tls_ctx})
593 if $self->{tls};
594
154 $self->on_drain (delete $self->{on_drain}) if $self->{on_drain}; 595 $self->on_drain (delete $self->{on_drain} ) if $self->{on_drain};
155 $self->on_read (delete $self->{on_read} ) if $self->{on_read};
156 596
157 $self->start_read; 597 $self->start_read
598 if $self->{on_read} || @{ $self->{_queue} };
158 599
159 $self 600 $self->_drain_wbuf;
160} 601}
161 602
162sub _shutdown {
163 my ($self) = @_;
164
165 delete $self->{rw};
166 delete $self->{ww};
167 delete $self->{fh};
168}
169
170sub error { 603sub _error {
171 my ($self) = @_; 604 my ($self, $errno, $fatal, $message) = @_;
172 605
173 { 606 $! = $errno;
174 local $!; 607 $message ||= "$!";
175 $self->_shutdown;
176 }
177 608
178 if ($self->{on_error}) { 609 if ($self->{on_error}) {
179 $self->{on_error}($self); 610 $self->{on_error}($self, $fatal, $message);
180 } else { 611 $self->destroy if $fatal;
612 } elsif ($self->{fh} || $self->{connect}) {
613 $self->destroy;
181 die "AnyEvent::Handle uncaught fatal error: $!"; 614 Carp::croak "AnyEvent::Handle uncaught error: $message";
182 } 615 }
183} 616}
184 617
185=item $fh = $handle->fh 618=item $fh = $handle->fh
186 619
187This method returns the filehandle of the L<AnyEvent::Handle> object. 620This method returns the file handle used to create the L<AnyEvent::Handle> object.
188 621
189=cut 622=cut
190 623
191sub fh { $_[0]->{fh} } 624sub fh { $_[0]{fh} }
192 625
193=item $handle->on_error ($cb) 626=item $handle->on_error ($cb)
194 627
195Replace the current C<on_error> callback (see the C<on_error> constructor argument). 628Replace the current C<on_error> callback (see the C<on_error> constructor argument).
196 629
208 641
209sub on_eof { 642sub on_eof {
210 $_[0]{on_eof} = $_[1]; 643 $_[0]{on_eof} = $_[1];
211} 644}
212 645
646=item $handle->on_timeout ($cb)
647
648=item $handle->on_rtimeout ($cb)
649
650=item $handle->on_wtimeout ($cb)
651
652Replace the current C<on_timeout>, C<on_rtimeout> or C<on_wtimeout>
653callback, or disables the callback (but not the timeout) if C<$cb> =
654C<undef>. See the C<timeout> constructor argument and method.
655
656=cut
657
658# see below
659
660=item $handle->autocork ($boolean)
661
662Enables or disables the current autocork behaviour (see C<autocork>
663constructor argument). Changes will only take effect on the next write.
664
665=cut
666
667sub autocork {
668 $_[0]{autocork} = $_[1];
669}
670
671=item $handle->no_delay ($boolean)
672
673Enables or disables the C<no_delay> setting (see constructor argument of
674the same name for details).
675
676=cut
677
678sub no_delay {
679 $_[0]{no_delay} = $_[1];
680
681 setsockopt $_[0]{fh}, Socket::IPPROTO_TCP (), Socket::TCP_NODELAY (), int $_[1]
682 if $_[0]{fh};
683}
684
685=item $handle->keepalive ($boolean)
686
687Enables or disables the C<keepalive> setting (see constructor argument of
688the same name for details).
689
690=cut
691
692sub keepalive {
693 $_[0]{keepalive} = $_[1];
694
695 eval {
696 local $SIG{__DIE__};
697 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
698 if $_[0]{fh};
699 };
700}
701
702=item $handle->oobinline ($boolean)
703
704Enables or disables the C<oobinline> setting (see constructor argument of
705the same name for details).
706
707=cut
708
709sub oobinline {
710 $_[0]{oobinline} = $_[1];
711
712 eval {
713 local $SIG{__DIE__};
714 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_OOBINLINE (), int $_[1]
715 if $_[0]{fh};
716 };
717}
718
719=item $handle->keepalive ($boolean)
720
721Enables or disables the C<keepalive> setting (see constructor argument of
722the same name for details).
723
724=cut
725
726sub keepalive {
727 $_[0]{keepalive} = $_[1];
728
729 eval {
730 local $SIG{__DIE__};
731 setsockopt $_[0]{fh}, Socket::SOL_SOCKET (), Socket::SO_KEEPALIVE (), int $_[1]
732 if $_[0]{fh};
733 };
734}
735
736=item $handle->on_starttls ($cb)
737
738Replace the current C<on_starttls> callback (see the C<on_starttls> constructor argument).
739
740=cut
741
742sub on_starttls {
743 $_[0]{on_starttls} = $_[1];
744}
745
746=item $handle->on_stoptls ($cb)
747
748Replace the current C<on_stoptls> callback (see the C<on_stoptls> constructor argument).
749
750=cut
751
752sub on_stoptls {
753 $_[0]{on_stoptls} = $_[1];
754}
755
756=item $handle->rbuf_max ($max_octets)
757
758Configures the C<rbuf_max> setting (C<undef> disables it).
759
760=item $handle->wbuf_max ($max_octets)
761
762Configures the C<wbuf_max> setting (C<undef> disables it).
763
764=cut
765
766sub rbuf_max {
767 $_[0]{rbuf_max} = $_[1];
768}
769
770sub rbuf_max {
771 $_[0]{wbuf_max} = $_[1];
772}
773
774#############################################################################
775
776=item $handle->timeout ($seconds)
777
778=item $handle->rtimeout ($seconds)
779
780=item $handle->wtimeout ($seconds)
781
782Configures (or disables) the inactivity timeout.
783
784=item $handle->timeout_reset
785
786=item $handle->rtimeout_reset
787
788=item $handle->wtimeout_reset
789
790Reset the activity timeout, as if data was received or sent.
791
792These methods are cheap to call.
793
794=cut
795
796for my $dir ("", "r", "w") {
797 my $timeout = "${dir}timeout";
798 my $tw = "_${dir}tw";
799 my $on_timeout = "on_${dir}timeout";
800 my $activity = "_${dir}activity";
801 my $cb;
802
803 *$on_timeout = sub {
804 $_[0]{$on_timeout} = $_[1];
805 };
806
807 *$timeout = sub {
808 my ($self, $new_value) = @_;
809
810 $new_value >= 0
811 or Carp::croak "AnyEvent::Handle->$timeout called with negative timeout ($new_value), caught";
812
813 $self->{$timeout} = $new_value;
814 delete $self->{$tw}; &$cb;
815 };
816
817 *{"${dir}timeout_reset"} = sub {
818 $_[0]{$activity} = AE::now;
819 };
820
821 # main workhorse:
822 # reset the timeout watcher, as neccessary
823 # also check for time-outs
824 $cb = sub {
825 my ($self) = @_;
826
827 if ($self->{$timeout} && $self->{fh}) {
828 my $NOW = AE::now;
829
830 # when would the timeout trigger?
831 my $after = $self->{$activity} + $self->{$timeout} - $NOW;
832
833 # now or in the past already?
834 if ($after <= 0) {
835 $self->{$activity} = $NOW;
836
837 if ($self->{$on_timeout}) {
838 $self->{$on_timeout}($self);
839 } else {
840 $self->_error (Errno::ETIMEDOUT);
841 }
842
843 # callback could have changed timeout value, optimise
844 return unless $self->{$timeout};
845
846 # calculate new after
847 $after = $self->{$timeout};
848 }
849
850 Scalar::Util::weaken $self;
851 return unless $self; # ->error could have destroyed $self
852
853 $self->{$tw} ||= AE::timer $after, 0, sub {
854 delete $self->{$tw};
855 $cb->($self);
856 };
857 } else {
858 delete $self->{$tw};
859 }
860 }
861}
862
213############################################################################# 863#############################################################################
214 864
215=back 865=back
216 866
217=head2 WRITE QUEUE 867=head2 WRITE QUEUE
220for reading. 870for reading.
221 871
222The write queue is very simple: you can add data to its end, and 872The write queue is very simple: you can add data to its end, and
223AnyEvent::Handle will automatically try to get rid of it for you. 873AnyEvent::Handle will automatically try to get rid of it for you.
224 874
225When data could be writtena nd the write buffer is shorter then the low 875When data could be written and the write buffer is shorter then the low
226water mark, the C<on_drain> callback will be invoked. 876water mark, the C<on_drain> callback will be invoked.
227 877
228=over 4 878=over 4
229 879
230=item $handle->on_drain ($cb) 880=item $handle->on_drain ($cb)
231 881
232Sets the C<on_drain> callback or clears it (see the description of 882Sets the C<on_drain> callback or clears it (see the description of
233C<on_drain> in the constructor). 883C<on_drain> in the constructor).
234 884
885This method may invoke callbacks (and therefore the handle might be
886destroyed after it returns).
887
235=cut 888=cut
236 889
237sub on_drain { 890sub on_drain {
238 my ($self, $cb) = @_; 891 my ($self, $cb) = @_;
239 892
240 $self->{on_drain} = $cb; 893 $self->{on_drain} = $cb;
241 894
242 $cb->($self) 895 $cb->($self)
243 if $cb && $self->{low_water_mark} >= length $self->{wbuf}; 896 if $cb && $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf});
244} 897}
245 898
246=item $handle->push_write ($data) 899=item $handle->push_write ($data)
247 900
248Queues the given scalar to be written. You can push as much data as you 901Queues the given scalar to be written. You can push as much data as
249want (only limited by the available memory), as C<AnyEvent::Handle> 902you want (only limited by the available memory and C<wbuf_max>), as
250buffers it independently of the kernel. 903C<AnyEvent::Handle> buffers it independently of the kernel.
251 904
252=cut 905This method may invoke callbacks (and therefore the handle might be
906destroyed after it returns).
253 907
254sub push_write { 908=cut
909
910sub _drain_wbuf {
255 my ($self, $data) = @_; 911 my ($self) = @_;
256 912
257 $self->{wbuf} .= $data; 913 if (!$self->{_ww} && length $self->{wbuf}) {
258 914
259 unless ($self->{ww}) {
260 Scalar::Util::weaken $self; 915 Scalar::Util::weaken $self;
916
261 my $cb = sub { 917 my $cb = sub {
262 my $len = syswrite $self->{fh}, $self->{wbuf}; 918 my $len = syswrite $self->{fh}, $self->{wbuf};
263 919
264 if ($len > 0) { 920 if (defined $len) {
265 substr $self->{wbuf}, 0, $len, ""; 921 substr $self->{wbuf}, 0, $len, "";
266 922
923 $self->{_activity} = $self->{_wactivity} = AE::now;
267 924
268 $self->{on_drain}($self) 925 $self->{on_drain}($self)
269 if $self->{low_water_mark} >= length $self->{wbuf} 926 if $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf})
270 && $self->{on_drain}; 927 && $self->{on_drain};
271 928
272 delete $self->{ww} unless length $self->{wbuf}; 929 delete $self->{_ww} unless length $self->{wbuf};
273 } elsif ($! != EAGAIN && $! != EINTR) { 930 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
274 $self->error; 931 $self->_error ($!, 1);
275 } 932 }
276 }; 933 };
277 934
278 $self->{ww} = AnyEvent->io (fh => $self->{fh}, poll => "w", cb => $cb); 935 # try to write data immediately
936 $cb->() unless $self->{autocork};
279 937
280 $cb->($self); 938 # if still data left in wbuf, we need to poll
939 $self->{_ww} = AE::io $self->{fh}, 1, $cb
940 if length $self->{wbuf};
941
942 if (
943 defined $self->{wbuf_max}
944 && $self->{wbuf_max} < length $self->{wbuf}
945 ) {
946 $self->_error (Errno::ENOSPC, 1), return;
947 }
281 }; 948 };
282} 949}
950
951our %WH;
952
953# deprecated
954sub register_write_type($$) {
955 $WH{$_[0]} = $_[1];
956}
957
958sub push_write {
959 my $self = shift;
960
961 if (@_ > 1) {
962 my $type = shift;
963
964 @_ = ($WH{$type} ||= _load_func "$type\::anyevent_write_type"
965 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_write")
966 ->($self, @_);
967 }
968
969 # we downgrade here to avoid hard-to-track-down bugs,
970 # and diagnose the problem earlier and better.
971
972 if ($self->{tls}) {
973 utf8::downgrade $self->{_tls_wbuf} .= $_[0];
974 &_dotls ($self) if $self->{fh};
975 } else {
976 utf8::downgrade $self->{wbuf} .= $_[0];
977 $self->_drain_wbuf if $self->{fh};
978 }
979}
980
981=item $handle->push_write (type => @args)
982
983Instead of formatting your data yourself, you can also let this module
984do the job by specifying a type and type-specific arguments. You
985can also specify the (fully qualified) name of a package, in which
986case AnyEvent tries to load the package and then expects to find the
987C<anyevent_write_type> function inside (see "custom write types", below).
988
989Predefined types are (if you have ideas for additional types, feel free to
990drop by and tell us):
991
992=over 4
993
994=item netstring => $string
995
996Formats the given value as netstring
997(http://cr.yp.to/proto/netstrings.txt, this is not a recommendation to use them).
998
999=cut
1000
1001register_write_type netstring => sub {
1002 my ($self, $string) = @_;
1003
1004 (length $string) . ":$string,"
1005};
1006
1007=item packstring => $format, $data
1008
1009An octet string prefixed with an encoded length. The encoding C<$format>
1010uses the same format as a Perl C<pack> format, but must specify a single
1011integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
1012optional C<!>, C<< < >> or C<< > >> modifier).
1013
1014=cut
1015
1016register_write_type packstring => sub {
1017 my ($self, $format, $string) = @_;
1018
1019 pack "$format/a*", $string
1020};
1021
1022=item json => $array_or_hashref
1023
1024Encodes the given hash or array reference into a JSON object. Unless you
1025provide your own JSON object, this means it will be encoded to JSON text
1026in UTF-8.
1027
1028JSON objects (and arrays) are self-delimiting, so you can write JSON at
1029one end of a handle and read them at the other end without using any
1030additional framing.
1031
1032The generated JSON text is guaranteed not to contain any newlines: While
1033this module doesn't need delimiters after or between JSON texts to be
1034able to read them, many other languages depend on that.
1035
1036A simple RPC protocol that interoperates easily with others is to send
1037JSON arrays (or objects, although arrays are usually the better choice as
1038they mimic how function argument passing works) and a newline after each
1039JSON text:
1040
1041 $handle->push_write (json => ["method", "arg1", "arg2"]); # whatever
1042 $handle->push_write ("\012");
1043
1044An AnyEvent::Handle receiver would simply use the C<json> read type and
1045rely on the fact that the newline will be skipped as leading whitespace:
1046
1047 $handle->push_read (json => sub { my $array = $_[1]; ... });
1048
1049Other languages could read single lines terminated by a newline and pass
1050this line into their JSON decoder of choice.
1051
1052=cut
1053
1054sub json_coder() {
1055 eval { require JSON::XS; JSON::XS->new->utf8 }
1056 || do { require JSON; JSON->new->utf8 }
1057}
1058
1059register_write_type json => sub {
1060 my ($self, $ref) = @_;
1061
1062 my $json = $self->{json} ||= json_coder;
1063
1064 $json->encode ($ref)
1065};
1066
1067=item storable => $reference
1068
1069Freezes the given reference using L<Storable> and writes it to the
1070handle. Uses the C<nfreeze> format.
1071
1072=cut
1073
1074register_write_type storable => sub {
1075 my ($self, $ref) = @_;
1076
1077 require Storable;
1078
1079 pack "w/a*", Storable::nfreeze ($ref)
1080};
1081
1082=back
1083
1084=item $handle->push_shutdown
1085
1086Sometimes you know you want to close the socket after writing your data
1087before it was actually written. One way to do that is to replace your
1088C<on_drain> handler by a callback that shuts down the socket (and set
1089C<low_water_mark> to C<0>). This method is a shorthand for just that, and
1090replaces the C<on_drain> callback with:
1091
1092 sub { shutdown $_[0]{fh}, 1 }
1093
1094This simply shuts down the write side and signals an EOF condition to the
1095the peer.
1096
1097You can rely on the normal read queue and C<on_eof> handling
1098afterwards. This is the cleanest way to close a connection.
1099
1100This method may invoke callbacks (and therefore the handle might be
1101destroyed after it returns).
1102
1103=cut
1104
1105sub push_shutdown {
1106 my ($self) = @_;
1107
1108 delete $self->{low_water_mark};
1109 $self->on_drain (sub { shutdown $_[0]{fh}, 1 });
1110}
1111
1112=item custom write types - Package::anyevent_write_type $handle, @args
1113
1114Instead of one of the predefined types, you can also specify the name of
1115a package. AnyEvent will try to load the package and then expects to find
1116a function named C<anyevent_write_type> inside. If it isn't found, it
1117progressively tries to load the parent package until it either finds the
1118function (good) or runs out of packages (bad).
1119
1120Whenever the given C<type> is used, C<push_write> will the function with
1121the handle object and the remaining arguments.
1122
1123The function is supposed to return a single octet string that will be
1124appended to the write buffer, so you cna mentally treat this function as a
1125"arguments to on-the-wire-format" converter.
1126
1127Example: implement a custom write type C<join> that joins the remaining
1128arguments using the first one.
1129
1130 $handle->push_write (My::Type => " ", 1,2,3);
1131
1132 # uses the following package, which can be defined in the "My::Type" or in
1133 # the "My" modules to be auto-loaded, or just about anywhere when the
1134 # My::Type::anyevent_write_type is defined before invoking it.
1135
1136 package My::Type;
1137
1138 sub anyevent_write_type {
1139 my ($handle, $delim, @args) = @_;
1140
1141 join $delim, @args
1142 }
1143
1144=cut
283 1145
284############################################################################# 1146#############################################################################
285 1147
286=back 1148=back
287 1149
294ways, the "simple" way, using only C<on_read> and the "complex" way, using 1156ways, the "simple" way, using only C<on_read> and the "complex" way, using
295a queue. 1157a queue.
296 1158
297In the simple case, you just install an C<on_read> callback and whenever 1159In the simple case, you just install an C<on_read> callback and whenever
298new data arrives, it will be called. You can then remove some data (if 1160new data arrives, it will be called. You can then remove some data (if
299enough is there) from the read buffer (C<< $handle->rbuf >>) if you want 1161enough is there) from the read buffer (C<< $handle->rbuf >>). Or you can
300or not. 1162leave the data there if you want to accumulate more (e.g. when only a
1163partial message has been received so far), or change the read queue with
1164e.g. C<push_read>.
301 1165
302In the more complex case, you want to queue multiple callbacks. In this 1166In the more complex case, you want to queue multiple callbacks. In this
303case, AnyEvent::Handle will call the first queued callback each time new 1167case, AnyEvent::Handle will call the first queued callback each time new
304data arrives and removes it when it has done its job (see C<push_read>, 1168data arrives (also the first time it is queued) and remove it when it has
305below). 1169done its job (see C<push_read>, below).
306 1170
307This way you can, for example, push three line-reads, followed by reading 1171This way you can, for example, push three line-reads, followed by reading
308a chunk of data, and AnyEvent::Handle will execute them in order. 1172a chunk of data, and AnyEvent::Handle will execute them in order.
309 1173
310Example 1: EPP protocol parser. EPP sends 4 byte length info, followed by 1174Example 1: EPP protocol parser. EPP sends 4 byte length info, followed by
311the specified number of bytes which give an XML datagram. 1175the specified number of bytes which give an XML datagram.
312 1176
313 # in the default state, expect some header bytes 1177 # in the default state, expect some header bytes
314 $handle->on_read (sub { 1178 $handle->on_read (sub {
315 # some data is here, now queue the length-header-read (4 octets) 1179 # some data is here, now queue the length-header-read (4 octets)
316 shift->unshift_read_chunk (4, sub { 1180 shift->unshift_read (chunk => 4, sub {
317 # header arrived, decode 1181 # header arrived, decode
318 my $len = unpack "N", $_[1]; 1182 my $len = unpack "N", $_[1];
319 1183
320 # now read the payload 1184 # now read the payload
321 shift->unshift_read_chunk ($len, sub { 1185 shift->unshift_read (chunk => $len, sub {
322 my $xml = $_[1]; 1186 my $xml = $_[1];
323 # handle xml 1187 # handle xml
324 }); 1188 });
325 }); 1189 });
326 }); 1190 });
327 1191
328Example 2: Implement a client for a protocol that replies either with 1192Example 2: Implement a client for a protocol that replies either with "OK"
329"OK" and another line or "ERROR" for one request, and 64 bytes for the 1193and another line or "ERROR" for the first request that is sent, and 64
330second request. Due tot he availability of a full queue, we can just 1194bytes for the second request. Due to the availability of a queue, we can
331pipeline sending both requests and manipulate the queue as necessary in 1195just pipeline sending both requests and manipulate the queue as necessary
332the callbacks: 1196in the callbacks.
333 1197
334 # request one 1198When the first callback is called and sees an "OK" response, it will
1199C<unshift> another line-read. This line-read will be queued I<before> the
120064-byte chunk callback.
1201
1202 # request one, returns either "OK + extra line" or "ERROR"
335 $handle->push_write ("request 1\015\012"); 1203 $handle->push_write ("request 1\015\012");
336 1204
337 # we expect "ERROR" or "OK" as response, so push a line read 1205 # we expect "ERROR" or "OK" as response, so push a line read
338 $handle->push_read_line (sub { 1206 $handle->push_read (line => sub {
339 # if we got an "OK", we have to _prepend_ another line, 1207 # if we got an "OK", we have to _prepend_ another line,
340 # so it will be read before the second request reads its 64 bytes 1208 # so it will be read before the second request reads its 64 bytes
341 # which are already in the queue when this callback is called 1209 # which are already in the queue when this callback is called
342 # we don't do this in case we got an error 1210 # we don't do this in case we got an error
343 if ($_[1] eq "OK") { 1211 if ($_[1] eq "OK") {
344 $_[0]->unshift_read_line (sub { 1212 $_[0]->unshift_read (line => sub {
345 my $response = $_[1]; 1213 my $response = $_[1];
346 ... 1214 ...
347 }); 1215 });
348 } 1216 }
349 }); 1217 });
350 1218
351 # request two 1219 # request two, simply returns 64 octets
352 $handle->push_write ("request 2\015\012"); 1220 $handle->push_write ("request 2\015\012");
353 1221
354 # simply read 64 bytes, always 1222 # simply read 64 bytes, always
355 $handle->push_read_chunk (64, sub { 1223 $handle->push_read (chunk => 64, sub {
356 my $response = $_[1]; 1224 my $response = $_[1];
357 ... 1225 ...
358 }); 1226 });
359 1227
360=over 4 1228=over 4
362=cut 1230=cut
363 1231
364sub _drain_rbuf { 1232sub _drain_rbuf {
365 my ($self) = @_; 1233 my ($self) = @_;
366 1234
1235 # avoid recursion
367 return if $self->{in_drain}; 1236 return if $self->{_skip_drain_rbuf};
368 local $self->{in_drain} = 1; 1237 local $self->{_skip_drain_rbuf} = 1;
369 1238
1239 while () {
1240 # we need to use a separate tls read buffer, as we must not receive data while
1241 # we are draining the buffer, and this can only happen with TLS.
1242 $self->{rbuf} .= delete $self->{_tls_rbuf}
1243 if exists $self->{_tls_rbuf};
1244
370 while (my $len = length $self->{rbuf}) { 1245 my $len = length $self->{rbuf};
371 no strict 'refs'; 1246
372 if (my $cb = shift @{ $self->{queue} }) { 1247 if (my $cb = shift @{ $self->{_queue} }) {
373 if (!$cb->($self)) { 1248 unless ($cb->($self)) {
1249 # no progress can be made
1250 # (not enough data and no data forthcoming)
1251 $self->_error (Errno::EPIPE, 1), return
374 if ($self->{eof}) { 1252 if $self->{_eof};
375 # no progress can be made (not enough data and no data forthcoming)
376 $! = &Errno::EPIPE; return $self->error;
377 }
378 1253
379 unshift @{ $self->{queue} }, $cb; 1254 unshift @{ $self->{_queue} }, $cb;
380 return; 1255 last;
381 } 1256 }
382 } elsif ($self->{on_read}) { 1257 } elsif ($self->{on_read}) {
1258 last unless $len;
1259
383 $self->{on_read}($self); 1260 $self->{on_read}($self);
384 1261
385 if ( 1262 if (
386 $self->{eof} # if no further data will arrive
387 && $len == length $self->{rbuf} # and no data has been consumed 1263 $len == length $self->{rbuf} # if no data has been consumed
388 && !@{ $self->{queue} } # and the queue is still empty 1264 && !@{ $self->{_queue} } # and the queue is still empty
389 && $self->{on_read} # and we still want to read data 1265 && $self->{on_read} # but we still have on_read
390 ) { 1266 ) {
1267 # no further data will arrive
391 # then no progress can be made 1268 # so no progress can be made
392 $! = &Errno::EPIPE; return $self->error; 1269 $self->_error (Errno::EPIPE, 1), return
1270 if $self->{_eof};
1271
1272 last; # more data might arrive
393 } 1273 }
394 } else { 1274 } else {
395 # read side becomes idle 1275 # read side becomes idle
396 delete $self->{rw}; 1276 delete $self->{_rw} unless $self->{tls};
397 return; 1277 last;
398 } 1278 }
399 } 1279 }
400 1280
401 if ($self->{eof}) { 1281 if ($self->{_eof}) {
402 $self->_shutdown; 1282 $self->{on_eof}
403 $self->{on_eof}($self); 1283 ? $self->{on_eof}($self)
1284 : $self->_error (0, 1, "Unexpected end-of-file");
1285
1286 return;
1287 }
1288
1289 if (
1290 defined $self->{rbuf_max}
1291 && $self->{rbuf_max} < length $self->{rbuf}
1292 ) {
1293 $self->_error (Errno::ENOSPC, 1), return;
1294 }
1295
1296 # may need to restart read watcher
1297 unless ($self->{_rw}) {
1298 $self->start_read
1299 if $self->{on_read} || @{ $self->{_queue} };
404 } 1300 }
405} 1301}
406 1302
407=item $handle->on_read ($cb) 1303=item $handle->on_read ($cb)
408 1304
409This replaces the currently set C<on_read> callback, or clears it (when 1305This replaces the currently set C<on_read> callback, or clears it (when
410the new callback is C<undef>). See the description of C<on_read> in the 1306the new callback is C<undef>). See the description of C<on_read> in the
411constructor. 1307constructor.
412 1308
1309This method may invoke callbacks (and therefore the handle might be
1310destroyed after it returns).
1311
413=cut 1312=cut
414 1313
415sub on_read { 1314sub on_read {
416 my ($self, $cb) = @_; 1315 my ($self, $cb) = @_;
417 1316
418 $self->{on_read} = $cb; 1317 $self->{on_read} = $cb;
1318 $self->_drain_rbuf if $cb;
419} 1319}
420 1320
421=item $handle->rbuf 1321=item $handle->rbuf
422 1322
423Returns the read buffer (as a modifiable lvalue). 1323Returns the read buffer (as a modifiable lvalue). You can also access the
1324read buffer directly as the C<< ->{rbuf} >> member, if you want (this is
1325much faster, and no less clean).
424 1326
425You can access the read buffer directly as the C<< ->{rbuf} >> member, if 1327The only operation allowed on the read buffer (apart from looking at it)
426you want. 1328is removing data from its beginning. Otherwise modifying or appending to
1329it is not allowed and will lead to hard-to-track-down bugs.
427 1330
428NOTE: The read buffer should only be used or modified if the C<on_read>, 1331NOTE: The read buffer should only be used or modified in the C<on_read>
429C<push_read> or C<unshift_read> methods are used. The other read methods 1332callback or when C<push_read> or C<unshift_read> are used with a single
430automatically manage the read buffer. 1333callback (i.e. untyped). Typed C<push_read> and C<unshift_read> methods
1334will manage the read buffer on their own.
431 1335
432=cut 1336=cut
433 1337
434sub rbuf : lvalue { 1338sub rbuf : lvalue {
435 $_[0]{rbuf} 1339 $_[0]{rbuf}
442Append the given callback to the end of the queue (C<push_read>) or 1346Append the given callback to the end of the queue (C<push_read>) or
443prepend it (C<unshift_read>). 1347prepend it (C<unshift_read>).
444 1348
445The callback is called each time some additional read data arrives. 1349The callback is called each time some additional read data arrives.
446 1350
447It must check wether enough data is in the read buffer already. 1351It must check whether enough data is in the read buffer already.
448 1352
449If not enough data is available, it must return the empty list or a false 1353If not enough data is available, it must return the empty list or a false
450value, in which case it will be called repeatedly until enough data is 1354value, in which case it will be called repeatedly until enough data is
451available (or an error condition is detected). 1355available (or an error condition is detected).
452 1356
453If enough data was available, then the callback must remove all data it is 1357If enough data was available, then the callback must remove all data it is
454interested in (which can be none at all) and return a true value. After returning 1358interested in (which can be none at all) and return a true value. After returning
455true, it will be removed from the queue. 1359true, it will be removed from the queue.
456 1360
1361These methods may invoke callbacks (and therefore the handle might be
1362destroyed after it returns).
1363
457=cut 1364=cut
1365
1366our %RH;
1367
1368sub register_read_type($$) {
1369 $RH{$_[0]} = $_[1];
1370}
458 1371
459sub push_read { 1372sub push_read {
460 my ($self, $cb) = @_; 1373 my $self = shift;
1374 my $cb = pop;
461 1375
1376 if (@_) {
1377 my $type = shift;
1378
1379 $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
1380 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_read")
1381 ->($self, $cb, @_);
1382 }
1383
462 push @{ $self->{queue} }, $cb; 1384 push @{ $self->{_queue} }, $cb;
463 $self->_drain_rbuf; 1385 $self->_drain_rbuf;
464} 1386}
465 1387
466sub unshift_read { 1388sub unshift_read {
467 my ($self, $cb) = @_; 1389 my $self = shift;
1390 my $cb = pop;
468 1391
1392 if (@_) {
1393 my $type = shift;
1394
1395 $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
1396 or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::unshift_read")
1397 ->($self, $cb, @_);
1398 }
1399
469 push @{ $self->{queue} }, $cb; 1400 unshift @{ $self->{_queue} }, $cb;
470 $self->_drain_rbuf; 1401 $self->_drain_rbuf;
471} 1402}
472 1403
473=item $handle->push_read_chunk ($len, $cb->($self, $data)) 1404=item $handle->push_read (type => @args, $cb)
474 1405
475=item $handle->unshift_read_chunk ($len, $cb->($self, $data)) 1406=item $handle->unshift_read (type => @args, $cb)
476 1407
477Append the given callback to the end of the queue (C<push_read_chunk>) or 1408Instead of providing a callback that parses the data itself you can chose
478prepend it (C<unshift_read_chunk>). 1409between a number of predefined parsing formats, for chunks of data, lines
1410etc. You can also specify the (fully qualified) name of a package, in
1411which case AnyEvent tries to load the package and then expects to find the
1412C<anyevent_read_type> function inside (see "custom read types", below).
479 1413
480The callback will be called only once C<$len> bytes have been read, and 1414Predefined types are (if you have ideas for additional types, feel free to
481these C<$len> bytes will be passed to the callback. 1415drop by and tell us):
482 1416
483=cut 1417=over 4
484 1418
485sub _read_chunk($$) { 1419=item chunk => $octets, $cb->($handle, $data)
1420
1421Invoke the callback only once C<$octets> bytes have been read. Pass the
1422data read to the callback. The callback will never be called with less
1423data.
1424
1425Example: read 2 bytes.
1426
1427 $handle->push_read (chunk => 2, sub {
1428 warn "yay ", unpack "H*", $_[1];
1429 });
1430
1431=cut
1432
1433register_read_type chunk => sub {
486 my ($self, $len, $cb) = @_; 1434 my ($self, $cb, $len) = @_;
487 1435
488 sub { 1436 sub {
489 $len <= length $_[0]{rbuf} or return; 1437 $len <= length $_[0]{rbuf} or return;
490 $cb->($_[0], substr $_[0]{rbuf}, 0, $len, ""); 1438 $cb->($_[0], substr $_[0]{rbuf}, 0, $len, "");
491 1 1439 1
492 } 1440 }
493} 1441};
494 1442
495sub push_read_chunk { 1443=item line => [$eol, ]$cb->($handle, $line, $eol)
496 $_[0]->push_read (&_read_chunk);
497}
498
499
500sub unshift_read_chunk {
501 $_[0]->unshift_read (&_read_chunk);
502}
503
504=item $handle->push_read_line ([$eol, ]$cb->($self, $line, $eol))
505
506=item $handle->unshift_read_line ([$eol, ]$cb->($self, $line, $eol))
507
508Append the given callback to the end of the queue (C<push_read_line>) or
509prepend it (C<unshift_read_line>).
510 1444
511The callback will be called only once a full line (including the end of 1445The callback will be called only once a full line (including the end of
512line marker, C<$eol>) has been read. This line (excluding the end of line 1446line marker, C<$eol>) has been read. This line (excluding the end of line
513marker) will be passed to the callback as second argument (C<$line>), and 1447marker) will be passed to the callback as second argument (C<$line>), and
514the end of line marker as the third argument (C<$eol>). 1448the end of line marker as the third argument (C<$eol>).
525Partial lines at the end of the stream will never be returned, as they are 1459Partial lines at the end of the stream will never be returned, as they are
526not marked by the end of line marker. 1460not marked by the end of line marker.
527 1461
528=cut 1462=cut
529 1463
530sub _read_line($$) { 1464register_read_type line => sub {
531 my $self = shift; 1465 my ($self, $cb, $eol) = @_;
532 my $cb = pop;
533 my $eol = @_ ? shift : qr|(\015?\012)|;
534 my $pos;
535 1466
1467 if (@_ < 3) {
1468 # this is more than twice as fast as the generic code below
1469 sub {
1470 $_[0]{rbuf} =~ s/^([^\015\012]*)(\015?\012)// or return;
1471
1472 $cb->($_[0], $1, $2);
1473 1
1474 }
1475 } else {
536 $eol = qr|(\Q$eol\E)| unless ref $eol; 1476 $eol = quotemeta $eol unless ref $eol;
537 $eol = qr|^(.*?)($eol)|; 1477 $eol = qr|^(.*?)($eol)|s;
1478
1479 sub {
1480 $_[0]{rbuf} =~ s/$eol// or return;
1481
1482 $cb->($_[0], $1, $2);
1483 1
1484 }
1485 }
1486};
1487
1488=item regex => $accept[, $reject[, $skip], $cb->($handle, $data)
1489
1490Makes a regex match against the regex object C<$accept> and returns
1491everything up to and including the match.
1492
1493Example: read a single line terminated by '\n'.
1494
1495 $handle->push_read (regex => qr<\n>, sub { ... });
1496
1497If C<$reject> is given and not undef, then it determines when the data is
1498to be rejected: it is matched against the data when the C<$accept> regex
1499does not match and generates an C<EBADMSG> error when it matches. This is
1500useful to quickly reject wrong data (to avoid waiting for a timeout or a
1501receive buffer overflow).
1502
1503Example: expect a single decimal number followed by whitespace, reject
1504anything else (not the use of an anchor).
1505
1506 $handle->push_read (regex => qr<^[0-9]+\s>, qr<[^0-9]>, sub { ... });
1507
1508If C<$skip> is given and not C<undef>, then it will be matched against
1509the receive buffer when neither C<$accept> nor C<$reject> match,
1510and everything preceding and including the match will be accepted
1511unconditionally. This is useful to skip large amounts of data that you
1512know cannot be matched, so that the C<$accept> or C<$reject> regex do not
1513have to start matching from the beginning. This is purely an optimisation
1514and is usually worth it only when you expect more than a few kilobytes.
1515
1516Example: expect a http header, which ends at C<\015\012\015\012>. Since we
1517expect the header to be very large (it isn't in practice, but...), we use
1518a skip regex to skip initial portions. The skip regex is tricky in that
1519it only accepts something not ending in either \015 or \012, as these are
1520required for the accept regex.
1521
1522 $handle->push_read (regex =>
1523 qr<\015\012\015\012>,
1524 undef, # no reject
1525 qr<^.*[^\015\012]>,
1526 sub { ... });
1527
1528=cut
1529
1530register_read_type regex => sub {
1531 my ($self, $cb, $accept, $reject, $skip) = @_;
1532
1533 my $data;
1534 my $rbuf = \$self->{rbuf};
538 1535
539 sub { 1536 sub {
540 $_[0]{rbuf} =~ s/$eol// or return; 1537 # accept
1538 if ($$rbuf =~ $accept) {
1539 $data .= substr $$rbuf, 0, $+[0], "";
1540 $cb->($self, $data);
1541 return 1;
1542 }
1543
1544 # reject
1545 if ($reject && $$rbuf =~ $reject) {
1546 $self->_error (Errno::EBADMSG);
1547 }
541 1548
542 $cb->($_[0], $1, $2); 1549 # skip
1550 if ($skip && $$rbuf =~ $skip) {
1551 $data .= substr $$rbuf, 0, $+[0], "";
1552 }
1553
1554 ()
1555 }
1556};
1557
1558=item netstring => $cb->($handle, $string)
1559
1560A netstring (http://cr.yp.to/proto/netstrings.txt, this is not an endorsement).
1561
1562Throws an error with C<$!> set to EBADMSG on format violations.
1563
1564=cut
1565
1566register_read_type netstring => sub {
1567 my ($self, $cb) = @_;
1568
1569 sub {
1570 unless ($_[0]{rbuf} =~ s/^(0|[1-9][0-9]*)://) {
1571 if ($_[0]{rbuf} =~ /[^0-9]/) {
1572 $self->_error (Errno::EBADMSG);
1573 }
1574 return;
1575 }
1576
1577 my $len = $1;
1578
1579 $self->unshift_read (chunk => $len, sub {
1580 my $string = $_[1];
1581 $_[0]->unshift_read (chunk => 1, sub {
1582 if ($_[1] eq ",") {
1583 $cb->($_[0], $string);
1584 } else {
1585 $self->_error (Errno::EBADMSG);
1586 }
1587 });
1588 });
1589
543 1 1590 1
544 } 1591 }
545} 1592};
546 1593
547sub push_read_line { 1594=item packstring => $format, $cb->($handle, $string)
548 $_[0]->push_read (&_read_line);
549}
550 1595
551sub unshift_read_line { 1596An octet string prefixed with an encoded length. The encoding C<$format>
552 $_[0]->unshift_read (&_read_line); 1597uses the same format as a Perl C<pack> format, but must specify a single
553} 1598integer only (only one of C<cCsSlLqQiInNvVjJw> is allowed, plus an
1599optional C<!>, C<< < >> or C<< > >> modifier).
1600
1601For example, DNS over TCP uses a prefix of C<n> (2 octet network order),
1602EPP uses a prefix of C<N> (4 octtes).
1603
1604Example: read a block of data prefixed by its length in BER-encoded
1605format (very efficient).
1606
1607 $handle->push_read (packstring => "w", sub {
1608 my ($handle, $data) = @_;
1609 });
1610
1611=cut
1612
1613register_read_type packstring => sub {
1614 my ($self, $cb, $format) = @_;
1615
1616 sub {
1617 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1618 defined (my $len = eval { unpack $format, $_[0]{rbuf} })
1619 or return;
1620
1621 $format = length pack $format, $len;
1622
1623 # bypass unshift if we already have the remaining chunk
1624 if ($format + $len <= length $_[0]{rbuf}) {
1625 my $data = substr $_[0]{rbuf}, $format, $len;
1626 substr $_[0]{rbuf}, 0, $format + $len, "";
1627 $cb->($_[0], $data);
1628 } else {
1629 # remove prefix
1630 substr $_[0]{rbuf}, 0, $format, "";
1631
1632 # read remaining chunk
1633 $_[0]->unshift_read (chunk => $len, $cb);
1634 }
1635
1636 1
1637 }
1638};
1639
1640=item json => $cb->($handle, $hash_or_arrayref)
1641
1642Reads a JSON object or array, decodes it and passes it to the
1643callback. When a parse error occurs, an C<EBADMSG> error will be raised.
1644
1645If a C<json> object was passed to the constructor, then that will be used
1646for the final decode, otherwise it will create a JSON coder expecting UTF-8.
1647
1648This read type uses the incremental parser available with JSON version
16492.09 (and JSON::XS version 2.2) and above. You have to provide a
1650dependency on your own: this module will load the JSON module, but
1651AnyEvent does not depend on it itself.
1652
1653Since JSON texts are fully self-delimiting, the C<json> read and write
1654types are an ideal simple RPC protocol: just exchange JSON datagrams. See
1655the C<json> write type description, above, for an actual example.
1656
1657=cut
1658
1659register_read_type json => sub {
1660 my ($self, $cb) = @_;
1661
1662 my $json = $self->{json} ||= json_coder;
1663
1664 my $data;
1665 my $rbuf = \$self->{rbuf};
1666
1667 sub {
1668 my $ref = eval { $json->incr_parse ($self->{rbuf}) };
1669
1670 if ($ref) {
1671 $self->{rbuf} = $json->incr_text;
1672 $json->incr_text = "";
1673 $cb->($self, $ref);
1674
1675 1
1676 } elsif ($@) {
1677 # error case
1678 $json->incr_skip;
1679
1680 $self->{rbuf} = $json->incr_text;
1681 $json->incr_text = "";
1682
1683 $self->_error (Errno::EBADMSG);
1684
1685 ()
1686 } else {
1687 $self->{rbuf} = "";
1688
1689 ()
1690 }
1691 }
1692};
1693
1694=item storable => $cb->($handle, $ref)
1695
1696Deserialises a L<Storable> frozen representation as written by the
1697C<storable> write type (BER-encoded length prefix followed by nfreeze'd
1698data).
1699
1700Raises C<EBADMSG> error if the data could not be decoded.
1701
1702=cut
1703
1704register_read_type storable => sub {
1705 my ($self, $cb) = @_;
1706
1707 require Storable;
1708
1709 sub {
1710 # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
1711 defined (my $len = eval { unpack "w", $_[0]{rbuf} })
1712 or return;
1713
1714 my $format = length pack "w", $len;
1715
1716 # bypass unshift if we already have the remaining chunk
1717 if ($format + $len <= length $_[0]{rbuf}) {
1718 my $data = substr $_[0]{rbuf}, $format, $len;
1719 substr $_[0]{rbuf}, 0, $format + $len, "";
1720 $cb->($_[0], Storable::thaw ($data));
1721 } else {
1722 # remove prefix
1723 substr $_[0]{rbuf}, 0, $format, "";
1724
1725 # read remaining chunk
1726 $_[0]->unshift_read (chunk => $len, sub {
1727 if (my $ref = eval { Storable::thaw ($_[1]) }) {
1728 $cb->($_[0], $ref);
1729 } else {
1730 $self->_error (Errno::EBADMSG);
1731 }
1732 });
1733 }
1734
1735 1
1736 }
1737};
1738
1739=back
1740
1741=item custom read types - Package::anyevent_read_type $handle, $cb, @args
1742
1743Instead of one of the predefined types, you can also specify the name
1744of a package. AnyEvent will try to load the package and then expects to
1745find a function named C<anyevent_read_type> inside. If it isn't found, it
1746progressively tries to load the parent package until it either finds the
1747function (good) or runs out of packages (bad).
1748
1749Whenever this type is used, C<push_read> will invoke the function with the
1750handle object, the original callback and the remaining arguments.
1751
1752The function is supposed to return a callback (usually a closure) that
1753works as a plain read callback (see C<< ->push_read ($cb) >>), so you can
1754mentally treat the function as a "configurable read type to read callback"
1755converter.
1756
1757It should invoke the original callback when it is done reading (remember
1758to pass C<$handle> as first argument as all other callbacks do that,
1759although there is no strict requirement on this).
1760
1761For examples, see the source of this module (F<perldoc -m
1762AnyEvent::Handle>, search for C<register_read_type>)).
554 1763
555=item $handle->stop_read 1764=item $handle->stop_read
556 1765
557=item $handle->start_read 1766=item $handle->start_read
558 1767
559In rare cases you actually do not want to read anything form the 1768In rare cases you actually do not want to read anything from the
560socket. In this case you can call C<stop_read>. Neither C<on_read> no 1769socket. In this case you can call C<stop_read>. Neither C<on_read> nor
561any queued callbacks will be executed then. To start readign again, call 1770any queued callbacks will be executed then. To start reading again, call
562C<start_read>. 1771C<start_read>.
1772
1773Note that AnyEvent::Handle will automatically C<start_read> for you when
1774you change the C<on_read> callback or push/unshift a read callback, and it
1775will automatically C<stop_read> for you when neither C<on_read> is set nor
1776there are any read requests in the queue.
1777
1778In older versions of this module (<= 5.3), these methods had no effect,
1779as TLS does not support half-duplex connections. In current versions they
1780work as expected, as this behaviour is required to avoid certain resource
1781attacks, where the program would be forced to read (and buffer) arbitrary
1782amounts of data before being able to send some data. The drawback is that
1783some readings of the the SSL/TLS specifications basically require this
1784attack to be working, as SSL/TLS implementations might stall sending data
1785during a rehandshake.
1786
1787As a guideline, during the initial handshake, you should not stop reading,
1788and as a client, it might cause problems, depending on your applciation.
563 1789
564=cut 1790=cut
565 1791
566sub stop_read { 1792sub stop_read {
567 my ($self) = @_; 1793 my ($self) = @_;
568 1794
569 delete $self->{rw}; 1795 delete $self->{_rw};
570} 1796}
571 1797
572sub start_read { 1798sub start_read {
573 my ($self) = @_; 1799 my ($self) = @_;
574 1800
575 unless ($self->{rw} || $self->{eof}) { 1801 unless ($self->{_rw} || $self->{_eof} || !$self->{fh}) {
576 Scalar::Util::weaken $self; 1802 Scalar::Util::weaken $self;
577 1803
578 $self->{rw} = AnyEvent->io (fh => $self->{fh}, poll => "r", cb => sub { 1804 $self->{_rw} = AE::io $self->{fh}, 0, sub {
1805 my $rbuf = \($self->{tls} ? my $buf : $self->{rbuf});
579 my $len = sysread $self->{fh}, $self->{rbuf}, $self->{read_size} || 8192, length $self->{rbuf}; 1806 my $len = sysread $self->{fh}, $$rbuf, $self->{read_size}, length $$rbuf;
580 1807
581 if ($len > 0) { 1808 if ($len > 0) {
582 if (defined $self->{rbuf_max}) { 1809 $self->{_activity} = $self->{_ractivity} = AE::now;
583 if ($self->{rbuf_max} < length $self->{rbuf}) { 1810
584 $! = &Errno::ENOSPC; return $self->error; 1811 if ($self->{tls}) {
585 } 1812 Net::SSLeay::BIO_write ($self->{_rbio}, $$rbuf);
1813
1814 &_dotls ($self);
1815 } else {
1816 $self->_drain_rbuf;
586 } 1817 }
587 1818
1819 if ($len == $self->{read_size}) {
1820 $self->{read_size} *= 2;
1821 $self->{read_size} = $self->{max_read_size} || MAX_READ_SIZE
1822 if $self->{read_size} > ($self->{max_read_size} || MAX_READ_SIZE);
1823 }
1824
588 } elsif (defined $len) { 1825 } elsif (defined $len) {
589 $self->{eof} = 1;
590 delete $self->{rw}; 1826 delete $self->{_rw};
1827 $self->{_eof} = 1;
1828 $self->_drain_rbuf;
591 1829
592 } elsif ($! != EAGAIN && $! != EINTR) { 1830 } elsif ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK) {
593 return $self->error; 1831 return $self->_error ($!, 1);
594 } 1832 }
1833 };
1834 }
1835}
595 1836
1837our $ERROR_SYSCALL;
1838our $ERROR_WANT_READ;
1839
1840sub _tls_error {
1841 my ($self, $err) = @_;
1842
1843 return $self->_error ($!, 1)
1844 if $err == Net::SSLeay::ERROR_SYSCALL ();
1845
1846 my $err =Net::SSLeay::ERR_error_string (Net::SSLeay::ERR_get_error ());
1847
1848 # reduce error string to look less scary
1849 $err =~ s/^error:[0-9a-fA-F]{8}:[^:]+:([^:]+):/\L$1: /;
1850
1851 if ($self->{_on_starttls}) {
1852 (delete $self->{_on_starttls})->($self, undef, $err);
1853 &_freetls;
1854 } else {
1855 &_freetls;
1856 $self->_error (Errno::EPROTO, 1, $err);
1857 }
1858}
1859
1860# poll the write BIO and send the data if applicable
1861# also decode read data if possible
1862# this is basiclaly our TLS state machine
1863# more efficient implementations are possible with openssl,
1864# but not with the buggy and incomplete Net::SSLeay.
1865sub _dotls {
1866 my ($self) = @_;
1867
1868 my $tmp;
1869
1870 if (length $self->{_tls_wbuf}) {
1871 while (($tmp = Net::SSLeay::write ($self->{tls}, $self->{_tls_wbuf})) > 0) {
1872 substr $self->{_tls_wbuf}, 0, $tmp, "";
1873 }
1874
1875 $tmp = Net::SSLeay::get_error ($self->{tls}, $tmp);
1876 return $self->_tls_error ($tmp)
1877 if $tmp != $ERROR_WANT_READ
1878 && ($tmp != $ERROR_SYSCALL || $!);
1879 }
1880
1881 while (defined ($tmp = Net::SSLeay::read ($self->{tls}))) {
1882 unless (length $tmp) {
1883 $self->{_on_starttls}
1884 and (delete $self->{_on_starttls})->($self, undef, "EOF during handshake"); # ???
1885 &_freetls;
1886
1887 if ($self->{on_stoptls}) {
1888 $self->{on_stoptls}($self);
1889 return;
1890 } else {
1891 # let's treat SSL-eof as we treat normal EOF
1892 delete $self->{_rw};
1893 $self->{_eof} = 1;
1894 }
1895 }
1896
1897 $self->{_tls_rbuf} .= $tmp;
596 $self->_drain_rbuf; 1898 $self->_drain_rbuf;
1899 $self->{tls} or return; # tls session might have gone away in callback
1900 }
1901
1902 $tmp = Net::SSLeay::get_error ($self->{tls}, -1);
1903 return $self->_tls_error ($tmp)
1904 if $tmp != $ERROR_WANT_READ
1905 && ($tmp != $ERROR_SYSCALL || $!);
1906
1907 while (length ($tmp = Net::SSLeay::BIO_read ($self->{_wbio}))) {
1908 $self->{wbuf} .= $tmp;
1909 $self->_drain_wbuf;
1910 $self->{tls} or return; # tls session might have gone away in callback
1911 }
1912
1913 $self->{_on_starttls}
1914 and Net::SSLeay::state ($self->{tls}) == Net::SSLeay::ST_OK ()
1915 and (delete $self->{_on_starttls})->($self, 1, "TLS/SSL connection established");
1916}
1917
1918=item $handle->starttls ($tls[, $tls_ctx])
1919
1920Instead of starting TLS negotiation immediately when the AnyEvent::Handle
1921object is created, you can also do that at a later time by calling
1922C<starttls>.
1923
1924Starting TLS is currently an asynchronous operation - when you push some
1925write data and then call C<< ->starttls >> then TLS negotiation will start
1926immediately, after which the queued write data is then sent.
1927
1928The first argument is the same as the C<tls> constructor argument (either
1929C<"connect">, C<"accept"> or an existing Net::SSLeay object).
1930
1931The second argument is the optional C<AnyEvent::TLS> object that is used
1932when AnyEvent::Handle has to create its own TLS connection object, or
1933a hash reference with C<< key => value >> pairs that will be used to
1934construct a new context.
1935
1936The TLS connection object will end up in C<< $handle->{tls} >>, the TLS
1937context in C<< $handle->{tls_ctx} >> after this call and can be used or
1938changed to your liking. Note that the handshake might have already started
1939when this function returns.
1940
1941Due to bugs in OpenSSL, it might or might not be possible to do multiple
1942handshakes on the same stream. It is best to not attempt to use the
1943stream after stopping TLS.
1944
1945This method may invoke callbacks (and therefore the handle might be
1946destroyed after it returns).
1947
1948=cut
1949
1950our %TLS_CACHE; #TODO not yet documented, should we?
1951
1952sub starttls {
1953 my ($self, $tls, $ctx) = @_;
1954
1955 Carp::croak "It is an error to call starttls on an AnyEvent::Handle object while TLS is already active, caught"
1956 if $self->{tls};
1957
1958 $self->{tls} = $tls;
1959 $self->{tls_ctx} = $ctx if @_ > 2;
1960
1961 return unless $self->{fh};
1962
1963 require Net::SSLeay;
1964
1965 $ERROR_SYSCALL = Net::SSLeay::ERROR_SYSCALL ();
1966 $ERROR_WANT_READ = Net::SSLeay::ERROR_WANT_READ ();
1967
1968 $tls = delete $self->{tls};
1969 $ctx = $self->{tls_ctx};
1970
1971 local $Carp::CarpLevel = 1; # skip ourselves when creating a new context or session
1972
1973 if ("HASH" eq ref $ctx) {
1974 require AnyEvent::TLS;
1975
1976 if ($ctx->{cache}) {
1977 my $key = $ctx+0;
1978 $ctx = $TLS_CACHE{$key} ||= new AnyEvent::TLS %$ctx;
1979 } else {
1980 $ctx = new AnyEvent::TLS %$ctx;
1981 }
1982 }
1983
1984 $self->{tls_ctx} = $ctx || TLS_CTX ();
1985 $self->{tls} = $tls = $self->{tls_ctx}->_get_session ($tls, $self, $self->{peername});
1986
1987 # basically, this is deep magic (because SSL_read should have the same issues)
1988 # but the openssl maintainers basically said: "trust us, it just works".
1989 # (unfortunately, we have to hardcode constants because the abysmally misdesigned
1990 # and mismaintained ssleay-module doesn't even offer them).
1991 # http://www.mail-archive.com/openssl-dev@openssl.org/msg22420.html
1992 #
1993 # in short: this is a mess.
1994 #
1995 # note that we do not try to keep the length constant between writes as we are required to do.
1996 # we assume that most (but not all) of this insanity only applies to non-blocking cases,
1997 # and we drive openssl fully in blocking mode here. Or maybe we don't - openssl seems to
1998 # have identity issues in that area.
1999# Net::SSLeay::CTX_set_mode ($ssl,
2000# (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ENABLE_PARTIAL_WRITE () } || 1)
2001# | (eval { local $SIG{__DIE__}; Net::SSLeay::MODE_ACCEPT_MOVING_WRITE_BUFFER () } || 2));
2002 Net::SSLeay::CTX_set_mode ($tls, 1|2);
2003
2004 $self->{_rbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
2005 $self->{_wbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
2006
2007 Net::SSLeay::BIO_write ($self->{_rbio}, delete $self->{rbuf});
2008
2009 Net::SSLeay::set_bio ($tls, $self->{_rbio}, $self->{_wbio});
2010
2011 $self->{_on_starttls} = sub { $_[0]{on_starttls}(@_) }
2012 if $self->{on_starttls};
2013
2014 &_dotls; # need to trigger the initial handshake
2015 $self->start_read; # make sure we actually do read
2016}
2017
2018=item $handle->stoptls
2019
2020Shuts down the SSL connection - this makes a proper EOF handshake by
2021sending a close notify to the other side, but since OpenSSL doesn't
2022support non-blocking shut downs, it is not guaranteed that you can re-use
2023the stream afterwards.
2024
2025This method may invoke callbacks (and therefore the handle might be
2026destroyed after it returns).
2027
2028=cut
2029
2030sub stoptls {
2031 my ($self) = @_;
2032
2033 if ($self->{tls} && $self->{fh}) {
2034 Net::SSLeay::shutdown ($self->{tls});
2035
2036 &_dotls;
2037
2038# # we don't give a shit. no, we do, but we can't. no...#d#
2039# # we, we... have to use openssl :/#d#
2040# &_freetls;#d#
2041 }
2042}
2043
2044sub _freetls {
2045 my ($self) = @_;
2046
2047 return unless $self->{tls};
2048
2049 $self->{tls_ctx}->_put_session (delete $self->{tls})
2050 if $self->{tls} > 0;
2051
2052 delete @$self{qw(_rbio _wbio _tls_wbuf _on_starttls)};
2053}
2054
2055sub DESTROY {
2056 my ($self) = @_;
2057
2058 &_freetls;
2059
2060 my $linger = exists $self->{linger} ? $self->{linger} : 3600;
2061
2062 if ($linger && length $self->{wbuf} && $self->{fh}) {
2063 my $fh = delete $self->{fh};
2064 my $wbuf = delete $self->{wbuf};
2065
2066 my @linger;
2067
2068 push @linger, AE::io $fh, 1, sub {
2069 my $len = syswrite $fh, $wbuf, length $wbuf;
2070
2071 if ($len > 0) {
2072 substr $wbuf, 0, $len, "";
2073 } elsif (defined $len || ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK)) {
2074 @linger = (); # end
2075 }
2076 };
2077 push @linger, AE::timer $linger, 0, sub {
2078 @linger = ();
2079 };
2080 }
2081}
2082
2083=item $handle->destroy
2084
2085Shuts down the handle object as much as possible - this call ensures that
2086no further callbacks will be invoked and as many resources as possible
2087will be freed. Any method you will call on the handle object after
2088destroying it in this way will be silently ignored (and it will return the
2089empty list).
2090
2091Normally, you can just "forget" any references to an AnyEvent::Handle
2092object and it will simply shut down. This works in fatal error and EOF
2093callbacks, as well as code outside. It does I<NOT> work in a read or write
2094callback, so when you want to destroy the AnyEvent::Handle object from
2095within such an callback. You I<MUST> call C<< ->destroy >> explicitly in
2096that case.
2097
2098Destroying the handle object in this way has the advantage that callbacks
2099will be removed as well, so if those are the only reference holders (as
2100is common), then one doesn't need to do anything special to break any
2101reference cycles.
2102
2103The handle might still linger in the background and write out remaining
2104data, as specified by the C<linger> option, however.
2105
2106=cut
2107
2108sub destroy {
2109 my ($self) = @_;
2110
2111 $self->DESTROY;
2112 %$self = ();
2113 bless $self, "AnyEvent::Handle::destroyed";
2114}
2115
2116sub AnyEvent::Handle::destroyed::AUTOLOAD {
2117 #nop
2118}
2119
2120=item $handle->destroyed
2121
2122Returns false as long as the handle hasn't been destroyed by a call to C<<
2123->destroy >>, true otherwise.
2124
2125Can be useful to decide whether the handle is still valid after some
2126callback possibly destroyed the handle. For example, C<< ->push_write >>,
2127C<< ->starttls >> and other methods can call user callbacks, which in turn
2128can destroy the handle, so work can be avoided by checking sometimes:
2129
2130 $hdl->starttls ("accept");
2131 return if $hdl->destroyed;
2132 $hdl->push_write (...
2133
2134Note that the call to C<push_write> will silently be ignored if the handle
2135has been destroyed, so often you can just ignore the possibility of the
2136handle being destroyed.
2137
2138=cut
2139
2140sub destroyed { 0 }
2141sub AnyEvent::Handle::destroyed::destroyed { 1 }
2142
2143=item AnyEvent::Handle::TLS_CTX
2144
2145This function creates and returns the AnyEvent::TLS object used by default
2146for TLS mode.
2147
2148The context is created by calling L<AnyEvent::TLS> without any arguments.
2149
2150=cut
2151
2152our $TLS_CTX;
2153
2154sub TLS_CTX() {
2155 $TLS_CTX ||= do {
2156 require AnyEvent::TLS;
2157
2158 new AnyEvent::TLS
2159 }
2160}
2161
2162=back
2163
2164
2165=head1 NONFREQUENTLY ASKED QUESTIONS
2166
2167=over 4
2168
2169=item I C<undef> the AnyEvent::Handle reference inside my callback and
2170still get further invocations!
2171
2172That's because AnyEvent::Handle keeps a reference to itself when handling
2173read or write callbacks.
2174
2175It is only safe to "forget" the reference inside EOF or error callbacks,
2176from within all other callbacks, you need to explicitly call the C<<
2177->destroy >> method.
2178
2179=item Why is my C<on_eof> callback never called?
2180
2181Probably because your C<on_error> callback is being called instead: When
2182you have outstanding requests in your read queue, then an EOF is
2183considered an error as you clearly expected some data.
2184
2185To avoid this, make sure you have an empty read queue whenever your handle
2186is supposed to be "idle" (i.e. connection closes are O.K.). You cna set
2187an C<on_read> handler that simply pushes the first read requests in the
2188queue.
2189
2190See also the next question, which explains this in a bit more detail.
2191
2192=item How can I serve requests in a loop?
2193
2194Most protocols consist of some setup phase (authentication for example)
2195followed by a request handling phase, where the server waits for requests
2196and handles them, in a loop.
2197
2198There are two important variants: The first (traditional, better) variant
2199handles requests until the server gets some QUIT command, causing it to
2200close the connection first (highly desirable for a busy TCP server). A
2201client dropping the connection is an error, which means this variant can
2202detect an unexpected detection close.
2203
2204To handle this case, always make sure you have a on-empty read queue, by
2205pushing the "read request start" handler on it:
2206
2207 # we assume a request starts with a single line
2208 my @start_request; @start_request = (line => sub {
2209 my ($hdl, $line) = @_;
2210
2211 ... handle request
2212
2213 # push next request read, possibly from a nested callback
2214 $hdl->push_read (@start_request);
2215 });
2216
2217 # auth done, now go into request handling loop
2218 # now push the first @start_request
2219 $hdl->push_read (@start_request);
2220
2221By always having an outstanding C<push_read>, the handle always expects
2222some data and raises the C<EPIPE> error when the connction is dropped
2223unexpectedly.
2224
2225The second variant is a protocol where the client can drop the connection
2226at any time. For TCP, this means that the server machine may run out of
2227sockets easier, and in general, it means you cnanot distinguish a protocl
2228failure/client crash from a normal connection close. Nevertheless, these
2229kinds of protocols are common (and sometimes even the best solution to the
2230problem).
2231
2232Having an outstanding read request at all times is possible if you ignore
2233C<EPIPE> errors, but this doesn't help with when the client drops the
2234connection during a request, which would still be an error.
2235
2236A better solution is to push the initial request read in an C<on_read>
2237callback. This avoids an error, as when the server doesn't expect data
2238(i.e. is idly waiting for the next request, an EOF will not raise an
2239error, but simply result in an C<on_eof> callback. It is also a bit slower
2240and simpler:
2241
2242 # auth done, now go into request handling loop
2243 $hdl->on_read (sub {
2244 my ($hdl) = @_;
2245
2246 # called each time we receive data but the read queue is empty
2247 # simply start read the request
2248
2249 $hdl->push_read (line => sub {
2250 my ($hdl, $line) = @_;
2251
2252 ... handle request
2253
2254 # do nothing special when the request has been handled, just
2255 # let the request queue go empty.
597 }); 2256 });
2257 });
2258
2259=item I get different callback invocations in TLS mode/Why can't I pause
2260reading?
2261
2262Unlike, say, TCP, TLS connections do not consist of two independent
2263communication channels, one for each direction. Or put differently, the
2264read and write directions are not independent of each other: you cannot
2265write data unless you are also prepared to read, and vice versa.
2266
2267This means that, in TLS mode, you might get C<on_error> or C<on_eof>
2268callback invocations when you are not expecting any read data - the reason
2269is that AnyEvent::Handle always reads in TLS mode.
2270
2271During the connection, you have to make sure that you always have a
2272non-empty read-queue, or an C<on_read> watcher. At the end of the
2273connection (or when you no longer want to use it) you can call the
2274C<destroy> method.
2275
2276=item How do I read data until the other side closes the connection?
2277
2278If you just want to read your data into a perl scalar, the easiest way
2279to achieve this is by setting an C<on_read> callback that does nothing,
2280clearing the C<on_eof> callback and in the C<on_error> callback, the data
2281will be in C<$_[0]{rbuf}>:
2282
2283 $handle->on_read (sub { });
2284 $handle->on_eof (undef);
2285 $handle->on_error (sub {
2286 my $data = delete $_[0]{rbuf};
2287 });
2288
2289The reason to use C<on_error> is that TCP connections, due to latencies
2290and packets loss, might get closed quite violently with an error, when in
2291fact all data has been received.
2292
2293It is usually better to use acknowledgements when transferring data,
2294to make sure the other side hasn't just died and you got the data
2295intact. This is also one reason why so many internet protocols have an
2296explicit QUIT command.
2297
2298=item I don't want to destroy the handle too early - how do I wait until
2299all data has been written?
2300
2301After writing your last bits of data, set the C<on_drain> callback
2302and destroy the handle in there - with the default setting of
2303C<low_water_mark> this will be called precisely when all data has been
2304written to the socket:
2305
2306 $handle->push_write (...);
2307 $handle->on_drain (sub {
2308 warn "all data submitted to the kernel\n";
2309 undef $handle;
2310 });
2311
2312If you just want to queue some data and then signal EOF to the other side,
2313consider using C<< ->push_shutdown >> instead.
2314
2315=item I want to contact a TLS/SSL server, I don't care about security.
2316
2317If your TLS server is a pure TLS server (e.g. HTTPS) that only speaks TLS,
2318connect to it and then create the AnyEvent::Handle with the C<tls>
2319parameter:
2320
2321 tcp_connect $host, $port, sub {
2322 my ($fh) = @_;
2323
2324 my $handle = new AnyEvent::Handle
2325 fh => $fh,
2326 tls => "connect",
2327 on_error => sub { ... };
2328
2329 $handle->push_write (...);
598 } 2330 };
599} 2331
2332=item I want to contact a TLS/SSL server, I do care about security.
2333
2334Then you should additionally enable certificate verification, including
2335peername verification, if the protocol you use supports it (see
2336L<AnyEvent::TLS>, C<verify_peername>).
2337
2338E.g. for HTTPS:
2339
2340 tcp_connect $host, $port, sub {
2341 my ($fh) = @_;
2342
2343 my $handle = new AnyEvent::Handle
2344 fh => $fh,
2345 peername => $host,
2346 tls => "connect",
2347 tls_ctx => { verify => 1, verify_peername => "https" },
2348 ...
2349
2350Note that you must specify the hostname you connected to (or whatever
2351"peername" the protocol needs) as the C<peername> argument, otherwise no
2352peername verification will be done.
2353
2354The above will use the system-dependent default set of trusted CA
2355certificates. If you want to check against a specific CA, add the
2356C<ca_file> (or C<ca_cert>) arguments to C<tls_ctx>:
2357
2358 tls_ctx => {
2359 verify => 1,
2360 verify_peername => "https",
2361 ca_file => "my-ca-cert.pem",
2362 },
2363
2364=item I want to create a TLS/SSL server, how do I do that?
2365
2366Well, you first need to get a server certificate and key. You have
2367three options: a) ask a CA (buy one, use cacert.org etc.) b) create a
2368self-signed certificate (cheap. check the search engine of your choice,
2369there are many tutorials on the net) or c) make your own CA (tinyca2 is a
2370nice program for that purpose).
2371
2372Then create a file with your private key (in PEM format, see
2373L<AnyEvent::TLS>), followed by the certificate (also in PEM format). The
2374file should then look like this:
2375
2376 -----BEGIN RSA PRIVATE KEY-----
2377 ...header data
2378 ... lots of base64'y-stuff
2379 -----END RSA PRIVATE KEY-----
2380
2381 -----BEGIN CERTIFICATE-----
2382 ... lots of base64'y-stuff
2383 -----END CERTIFICATE-----
2384
2385The important bits are the "PRIVATE KEY" and "CERTIFICATE" parts. Then
2386specify this file as C<cert_file>:
2387
2388 tcp_server undef, $port, sub {
2389 my ($fh) = @_;
2390
2391 my $handle = new AnyEvent::Handle
2392 fh => $fh,
2393 tls => "accept",
2394 tls_ctx => { cert_file => "my-server-keycert.pem" },
2395 ...
2396
2397When you have intermediate CA certificates that your clients might not
2398know about, just append them to the C<cert_file>.
600 2399
601=back 2400=back
602 2401
2402
2403=head1 SUBCLASSING AnyEvent::Handle
2404
2405In many cases, you might want to subclass AnyEvent::Handle.
2406
2407To make this easier, a given version of AnyEvent::Handle uses these
2408conventions:
2409
2410=over 4
2411
2412=item * all constructor arguments become object members.
2413
2414At least initially, when you pass a C<tls>-argument to the constructor it
2415will end up in C<< $handle->{tls} >>. Those members might be changed or
2416mutated later on (for example C<tls> will hold the TLS connection object).
2417
2418=item * other object member names are prefixed with an C<_>.
2419
2420All object members not explicitly documented (internal use) are prefixed
2421with an underscore character, so the remaining non-C<_>-namespace is free
2422for use for subclasses.
2423
2424=item * all members not documented here and not prefixed with an underscore
2425are free to use in subclasses.
2426
2427Of course, new versions of AnyEvent::Handle may introduce more "public"
2428member variables, but that's just life. At least it is documented.
2429
2430=back
2431
603=head1 AUTHOR 2432=head1 AUTHOR
604 2433
605Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>. 2434Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>.
606 2435
607=cut 2436=cut

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines