1 | use Coro; |
1 | use Coro; |
2 | use Coro::Semaphore; |
2 | use Coro::Semaphore; |
3 | use Coro::Event; |
3 | use Coro::Event; |
4 | use Coro::Socket; |
4 | use Coro::Socket; |
|
|
5 | |
|
|
6 | use HTTP::Date; |
5 | |
7 | |
6 | no utf8; |
8 | no utf8; |
7 | use bytes; |
9 | use bytes; |
8 | |
10 | |
9 | # at least on my machine, this thingy serves files |
11 | # at least on my machine, this thingy serves files |
… | |
… | |
25 | my $level = shift; |
27 | my $level = shift; |
26 | my $format = shift; |
28 | my $format = shift; |
27 | printf "---: $format\n", @_; |
29 | printf "---: $format\n", @_; |
28 | } |
30 | } |
29 | |
31 | |
30 | my $connections = new Coro::Semaphore $MAX_CONNECTS; |
32 | our $connections = new Coro::Semaphore $MAX_CONNECTS || 250; |
|
|
33 | |
|
|
34 | our $wait_factor = 0.95; |
|
|
35 | |
|
|
36 | our @transfers = ( |
|
|
37 | [(new Coro::Semaphore $MAX_TRANSFERS_SMALL || 50), 600], |
|
|
38 | [(new Coro::Semaphore $MAX_TRANSFERS_LARGE || 50), 600], |
|
|
39 | ); |
31 | |
40 | |
32 | my @newcons; |
41 | my @newcons; |
33 | my @pool; |
42 | my @pool; |
34 | |
43 | |
35 | # one "execution thread" |
44 | # one "execution thread" |
… | |
… | |
57 | Listen => 50, |
66 | Listen => 50, |
58 | or die "unable to start server"; |
67 | or die "unable to start server"; |
59 | |
68 | |
60 | push @listen_sockets, $http_port; |
69 | push @listen_sockets, $http_port; |
61 | |
70 | |
|
|
71 | our $NOW; |
|
|
72 | our $HTTP_NOW; |
|
|
73 | |
|
|
74 | Event->timer(interval => 1, hard => 1, cb => sub { |
|
|
75 | $NOW = time; |
|
|
76 | $HTTP_NOW = time2str $NOW; |
|
|
77 | })->now; |
|
|
78 | |
62 | # the "main thread" |
79 | # the "main thread" |
63 | async { |
80 | async { |
64 | slog 1, "accepting connections"; |
81 | slog 1, "accepting connections"; |
65 | while () { |
82 | while () { |
66 | $connections->down; |
83 | $connections->down; |
67 | push @newcons, [$http_port->accept]; |
84 | push @newcons, [$http_port->accept]; |
68 | #slog 3, "accepted @$connections ".scalar(@pool); |
85 | #slog 3, "accepted @$connections ".scalar(@pool); |
69 | $::NOW = time; |
|
|
70 | if (@pool) { |
86 | if (@pool) { |
71 | (pop @pool)->ready; |
87 | (pop @pool)->ready; |
72 | } else { |
88 | } else { |
73 | async \&handler; |
89 | async \&handler; |
74 | } |
90 | } |
… | |
… | |
82 | use HTTP::Date; |
98 | use HTTP::Date; |
83 | use Convert::Scalar 'weaken'; |
99 | use Convert::Scalar 'weaken'; |
84 | use Linux::AIO; |
100 | use Linux::AIO; |
85 | |
101 | |
86 | Linux::AIO::min_parallel $::AIO_PARALLEL; |
102 | Linux::AIO::min_parallel $::AIO_PARALLEL; |
|
|
103 | |
|
|
104 | my $aio_requests = new Coro::Semaphore $::AIO_PARALLEL * 4; |
87 | |
105 | |
88 | Event->io(fd => Linux::AIO::poll_fileno, |
106 | Event->io(fd => Linux::AIO::poll_fileno, |
89 | poll => 'r', async => 1, |
107 | poll => 'r', async => 1, |
90 | cb => \&Linux::AIO::poll_cb); |
108 | cb => \&Linux::AIO::poll_cb); |
91 | |
109 | |
… | |
… | |
143 | delete $uri{$self->{remote_addr}}{$self->{uri}}{$self*1}; |
161 | delete $uri{$self->{remote_addr}}{$self->{uri}}{$self*1}; |
144 | } |
162 | } |
145 | |
163 | |
146 | sub slog { |
164 | sub slog { |
147 | my $self = shift; |
165 | my $self = shift; |
148 | main::slog($_[0], "$self->{remote_addr}> $_[1]"); |
166 | main::slog($_[0], ($self->{remote_id} || $self->{remote_addr}) ."> $_[1]"); |
149 | } |
167 | } |
150 | |
168 | |
151 | sub response { |
169 | sub response { |
152 | my ($self, $code, $msg, $hdr, $content) = @_; |
170 | my ($self, $code, $msg, $hdr, $content) = @_; |
153 | my $res = "HTTP/1.1 $code $msg\015\012"; |
171 | my $res = "HTTP/1.1 $code $msg\015\012"; |
154 | |
172 | |
155 | $self->{h}{connection} ||= $hdr->{Connection}; |
173 | $self->{h}{connection} ||= $hdr->{Connection}; |
156 | |
174 | |
157 | $res .= "Date: ".(time2str $::NOW)."\015\012"; # slow? nah. :( |
175 | $res .= "Date: $HTTP_NOW\015\012"; |
158 | |
176 | |
159 | while (my ($h, $v) = each %$hdr) { |
177 | while (my ($h, $v) = each %$hdr) { |
160 | $res .= "$h: $v\015\012" |
178 | $res .= "$h: $v\015\012" |
161 | } |
179 | } |
162 | $res .= "\015\012"; |
180 | $res .= "\015\012"; |
… | |
… | |
186 | $self->response($code, $msg, $hdr, $content); |
204 | $self->response($code, $msg, $hdr, $content); |
187 | |
205 | |
188 | die bless {}, err::; |
206 | die bless {}, err::; |
189 | } |
207 | } |
190 | |
208 | |
191 | sub err_blocked { |
|
|
192 | my $self = shift; |
|
|
193 | my $ip = $self->{remote_addr}; |
|
|
194 | my $time = time2str $blocked{$ip} = $::NOW + $::BLOCKTIME; |
|
|
195 | |
|
|
196 | Coro::Event::do_timer(after => 20*rand); |
|
|
197 | |
|
|
198 | $self->err(401, "too many connections", |
|
|
199 | { |
|
|
200 | "Content-Type" => "text/html", |
|
|
201 | "Retry-After" => $::BLOCKTIME, |
|
|
202 | "Warning" => "Please do NOT retry, you have been blocked", |
|
|
203 | "WWW-Authenticate" => "Basic realm=\"Please do NOT retry, you have been blocked\"", |
|
|
204 | "Connection" => "close", |
|
|
205 | }, |
|
|
206 | <<EOF); |
|
|
207 | <html> |
|
|
208 | <head> |
|
|
209 | <title>Too many connections</title> |
|
|
210 | </head> |
|
|
211 | <body bgcolor="#ffffff" text="#000000" link="#0000ff" vlink="#000080" alink="#ff0000"> |
|
|
212 | |
|
|
213 | <p>You have been blocked because you opened too many connections. You |
|
|
214 | may retry at</p> |
|
|
215 | |
|
|
216 | <p><blockquote>$time.</blockquote></p> |
|
|
217 | |
|
|
218 | <p>Until then, each new access will renew the block. You might want to have a |
|
|
219 | look at the <a href="http://www.goof.com/pcg/marc/animefaq.html#connectionlimit">FAQ</a>.</p> |
|
|
220 | |
|
|
221 | </body></html> |
|
|
222 | EOF |
|
|
223 | } |
|
|
224 | |
|
|
225 | sub handle { |
209 | sub handle { |
226 | my $self = shift; |
210 | my $self = shift; |
227 | my $fh = $self->{fh}; |
211 | my $fh = $self->{fh}; |
|
|
212 | |
|
|
213 | my $host; |
228 | |
214 | |
229 | $fh->timeout($::REQ_TIMEOUT); |
215 | $fh->timeout($::REQ_TIMEOUT); |
230 | while() { |
216 | while() { |
231 | $self->{reqs}++; |
217 | $self->{reqs}++; |
232 | |
218 | |
… | |
… | |
252 | |
238 | |
253 | delete $blocked{$ip}; |
239 | delete $blocked{$ip}; |
254 | } |
240 | } |
255 | |
241 | |
256 | if (%{$conn{$ip}} > $::MAX_CONN_IP) { |
242 | if (%{$conn{$ip}} > $::MAX_CONN_IP) { |
|
|
243 | my $delay = 120; |
|
|
244 | while (%{$conn{$ip}} > $::MAX_CONN_IP) { |
|
|
245 | if ($delay <= 0) { |
257 | $self->slog(2, "blocked ip $ip"); |
246 | $self->slog(2, "blocked ip $ip"); |
258 | $self->err_blocked; |
247 | $self->err_blocked; |
|
|
248 | } else { |
|
|
249 | Coro::Event::do_timer(after => 3); |
|
|
250 | $delay -= 3; |
|
|
251 | } |
|
|
252 | } |
259 | } |
253 | } |
260 | |
254 | |
261 | $req =~ /^(?:\015\012)? |
255 | $req =~ /^(?:\015\012)? |
262 | (GET|HEAD) \040+ |
256 | (GET|HEAD) \040+ |
263 | ([^\040]+) \040+ |
257 | ([^\040]+) \040+ |
… | |
… | |
289 | |
283 | |
290 | $self->{h}{$h} = substr $v, 1 |
284 | $self->{h}{$h} = substr $v, 1 |
291 | while ($h, $v) = each %hdr; |
285 | while ($h, $v) = each %hdr; |
292 | } |
286 | } |
293 | |
287 | |
|
|
288 | # find out server name and port |
|
|
289 | if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
|
|
290 | $host = $1; |
|
|
291 | } else { |
|
|
292 | $host = $self->{h}{host}; |
|
|
293 | } |
|
|
294 | |
|
|
295 | if (defined $host) { |
294 | $self->{server_port} = $self->{h}{host} =~ s/:([0-9]+)$// ? $1 : 80; |
296 | $self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
|
|
297 | } else { |
|
|
298 | ($self->{server_port}, $host) |
|
|
299 | = unpack_sockaddr_in $self->{fh}->getsockname |
|
|
300 | or $self->err(500, "unable to get socket name"); |
|
|
301 | $host = inet_ntoa $host; |
|
|
302 | } |
|
|
303 | |
|
|
304 | $self->{server_name} = $host; |
|
|
305 | |
|
|
306 | # remote id should be unique per user |
|
|
307 | $self->{remote_id} = $self->{remote_addr}; |
|
|
308 | |
|
|
309 | if (exists $self->{h}{"client-ip"}) { |
|
|
310 | $self->{remote_id} .= "[".$self->{h}{"client-ip"}."]"; |
|
|
311 | } elsif (exists $self->{h}{"x-forwarded-for"}) { |
|
|
312 | $self->{remote_id} .= "[".$self->{h}{"x-forwarded-for"}."]"; |
|
|
313 | } |
295 | |
314 | |
296 | weaken ($uri{$self->{remote_addr}}{$self->{uri}}{$self*1} = $self); |
315 | weaken ($uri{$self->{remote_addr}}{$self->{uri}}{$self*1} = $self); |
297 | |
316 | |
298 | eval { |
317 | eval { |
299 | $self->map_uri; |
318 | $self->map_uri; |
… | |
… | |
302 | |
321 | |
303 | $self->eoconn; |
322 | $self->eoconn; |
304 | |
323 | |
305 | die if $@ && !ref $@; |
324 | die if $@ && !ref $@; |
306 | |
325 | |
307 | last if $self->{h}{connection} =~ /close/ || $self->{version} lt "1.1"; |
326 | last if $self->{h}{connection} =~ /close/ || $self->{version} < 1.1; |
308 | |
327 | |
309 | $self->slog(9, "persistent connection [".$self->{h}{"user-agent"}."][$self->{reqs}]"); |
|
|
310 | $fh->timeout($::PER_TIMEOUT); |
328 | $fh->timeout($::PER_TIMEOUT); |
311 | } |
329 | } |
312 | } |
330 | } |
313 | |
331 | |
314 | # uri => path mapping |
332 | # uri => path mapping |
315 | sub map_uri { |
333 | sub map_uri { |
316 | my $self = shift; |
334 | my $self = shift; |
317 | my $host = $self->{h}{host} || "default"; |
335 | my $host = $self->{server_name}; |
318 | my $uri = $self->{uri}; |
336 | my $uri = $self->{uri}; |
319 | |
337 | |
320 | # some massaging, also makes it more secure |
338 | # some massaging, also makes it more secure |
321 | $uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
339 | $uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
322 | $uri =~ s%//+%/%g; |
340 | $uri =~ s%//+%/%g; |
… | |
… | |
332 | $self->{path} = "$::DOCROOT/$host$uri"; |
350 | $self->{path} = "$::DOCROOT/$host$uri"; |
333 | |
351 | |
334 | $self->access_check; |
352 | $self->access_check; |
335 | } |
353 | } |
336 | |
354 | |
337 | sub server_address { |
|
|
338 | my $self = shift; |
|
|
339 | my ($port, $iaddr) = unpack_sockaddr_in $self->{fh}->getsockname |
|
|
340 | or $self->err(500, "unable to get socket name"); |
|
|
341 | ((inet_ntoa $iaddr), $port); |
|
|
342 | } |
|
|
343 | |
|
|
344 | sub server_host { |
|
|
345 | my $self = shift; |
|
|
346 | if (exists $self->{h}{host}) { |
|
|
347 | return $self->{h}{host}; |
|
|
348 | } else { |
|
|
349 | return (($self->server_address)[0]); |
|
|
350 | } |
|
|
351 | } |
|
|
352 | |
|
|
353 | sub server_hostport { |
|
|
354 | my $self = shift; |
|
|
355 | my ($host, $port); |
|
|
356 | if (exists $self->{h}{host}) { |
|
|
357 | ($host, $port) = ($self->{h}{host}, $self->{server_port}); |
|
|
358 | } else { |
|
|
359 | ($host, $port) = $self->server_address; |
|
|
360 | } |
|
|
361 | $port = $port == 80 ? "" : ":$port"; |
|
|
362 | $host.$port; |
|
|
363 | } |
|
|
364 | |
|
|
365 | sub _cgi { |
355 | sub _cgi { |
366 | my $self = shift; |
356 | my $self = shift; |
367 | my $path = shift; |
357 | my $path = shift; |
368 | my $fh; |
358 | my $fh; |
369 | |
359 | |
370 | # no two-way xxx supported |
360 | # no two-way xxx supported |
371 | if (0 == fork) { |
361 | if (0 == fork) { |
372 | open STDOUT, ">&".fileno($self->{fh}); |
362 | open STDOUT, ">&".fileno($self->{fh}); |
373 | if (chdir $::DOCROOT) { |
363 | if (chdir $::DOCROOT) { |
374 | $ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
364 | $ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
375 | $ENV{HTTP_HOST} = $self->server_host; |
365 | $ENV{HTTP_HOST} = $self->{server_name}; |
376 | $ENV{HTTP_PORT} = $self->{server_host}; |
366 | $ENV{HTTP_PORT} = $self->{server_port}; |
377 | $ENV{SCRIPT_NAME} = $self->{name}; |
367 | $ENV{SCRIPT_NAME} = $self->{name}; |
378 | exec $path; |
368 | exec $path; |
379 | } |
369 | } |
380 | Coro::State::_exit(0); |
370 | Coro::State::_exit(0); |
381 | } else { |
371 | } else { |
|
|
372 | die; |
382 | } |
373 | } |
|
|
374 | } |
|
|
375 | |
|
|
376 | sub server_hostport { |
|
|
377 | $_[0]{server_port} == 80 |
|
|
378 | ? $_[0]{server_name} |
|
|
379 | : "$_[0]{server_name}:$_[0]{server_port}"; |
383 | } |
380 | } |
384 | |
381 | |
385 | sub respond { |
382 | sub respond { |
386 | my $self = shift; |
383 | my $self = shift; |
387 | my $path = $self->{path}; |
384 | my $path = $self->{path}; |
… | |
… | |
397 | |
394 | |
398 | if (-d _ && -r _) { |
395 | if (-d _ && -r _) { |
399 | # directory |
396 | # directory |
400 | if ($path !~ /\/$/) { |
397 | if ($path !~ /\/$/) { |
401 | # create a redirect to get the trailing "/" |
398 | # create a redirect to get the trailing "/" |
402 | my $host = $self->server_hostport; |
399 | # we don't try to avoid the :80 |
403 | $self->err(301, "moved permanently", { Location => "http://$host$self->{uri}/" }); |
400 | $self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
404 | } else { |
401 | } else { |
405 | $ims < $self->{stat}[9] |
402 | $ims < $self->{stat}[9] |
406 | or $self->err(304, "not modified"); |
403 | or $self->err(304, "not modified"); |
407 | |
404 | |
408 | if (-r "$path/index.html") { |
405 | if (-r "$path/index.html") { |
… | |
… | |
432 | $idx); |
429 | $idx); |
433 | } |
430 | } |
434 | |
431 | |
435 | sub handle_file { |
432 | sub handle_file { |
436 | my $self = shift; |
433 | my $self = shift; |
437 | my $length = -s _; |
434 | my $length = $self->{stat}[7]; |
|
|
435 | my $queue = $::transfers[$length >= $::TRANSFER_SMALL]; |
438 | my $hdr = { |
436 | my $hdr = { |
439 | "Last-Modified" => time2str ((stat _)[9]), |
437 | "Last-Modified" => time2str ((stat _)[9]), |
440 | }; |
438 | }; |
441 | |
439 | |
442 | my @code = (200, "ok"); |
440 | my @code = (200, "ok"); |
… | |
… | |
460 | $self->err(416, "not satisfiable", $hdr, ""); |
458 | $self->err(416, "not satisfiable", $hdr, ""); |
461 | |
459 | |
462 | satisfiable: |
460 | satisfiable: |
463 | # check for segmented downloads |
461 | # check for segmented downloads |
464 | if ($l && $::NO_SEGMENTED) { |
462 | if ($l && $::NO_SEGMENTED) { |
|
|
463 | my $delay = 180; |
465 | if (%{$uri{$self->{remote_addr}}{$self->{uri}}} > 1) { |
464 | while (%{$uri{$self->{remote_addr}}{$self->{uri}}} > 1) { |
466 | $self->err(400, "segmented downloads are not allowed", |
465 | if ($delay <= 0) { |
467 | { "Content-Type" => "text/html", Connection => "close" }, <<EOF); |
466 | $self->err_segmented_download; |
468 | <html> |
467 | } else { |
469 | <head> |
468 | Coro::Event::do_timer(after => 3); $delay -= 3; |
470 | <title>Segmented downloads are not allowed</title> |
469 | } |
471 | </head> |
|
|
472 | <body bgcolor="#ffffff" text="#000000" link="#0000ff" vlink="#000080" alink="#ff0000"> |
|
|
473 | |
|
|
474 | <p>Segmented downloads are not allowed on this server. Please refer to the |
|
|
475 | <a href="http://www.goof.com/pcg/marc/animefaq.html#segmented_downloads">FAQ</a>.</p> |
|
|
476 | |
|
|
477 | </body></html> |
|
|
478 | EOF |
|
|
479 | EOF |
|
|
480 | } |
470 | } |
481 | } |
471 | } |
482 | |
472 | |
483 | $hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
473 | $hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
484 | @code = (206, "partial content"); |
474 | @code = (206, "partial content"); |
… | |
… | |
494 | $hdr->{"Content-Length"} = $length; |
484 | $hdr->{"Content-Length"} = $length; |
495 | |
485 | |
496 | $self->response(@code, $hdr, ""); |
486 | $self->response(@code, $hdr, ""); |
497 | |
487 | |
498 | if ($self->{method} eq "GET") { |
488 | if ($self->{method} eq "GET") { |
|
|
489 | $self->{time} = $::NOW; |
|
|
490 | |
|
|
491 | my $transfer = $queue->[0]->guard; |
|
|
492 | $self->{fh}->writable or return; |
|
|
493 | |
|
|
494 | $queue->[1] = $queue->[1] * $::wait_factor |
|
|
495 | + ($::NOW - $self->{time}) * (1 - $::wait_factor); |
|
|
496 | $self->{time} = $::NOW; |
|
|
497 | |
499 | my ($fh, $buf, $r); |
498 | my ($fh, $buf, $r); |
500 | my $current = $Coro::current; |
499 | my $current = $Coro::current; |
501 | open $fh, "<", $self->{path} |
500 | open $fh, "<", $self->{path} |
502 | or die "$self->{path}: late open failure ($!)"; |
501 | or die "$self->{path}: late open failure ($!)"; |
503 | |
502 | |
… | |
… | |
512 | while ($h > 0) { |
511 | while ($h > 0) { |
513 | if (0) { |
512 | if (0) { |
514 | sysread $fh, $buf, $h > $::BUFSIZE ? $::BUFSIZE : $h |
513 | sysread $fh, $buf, $h > $::BUFSIZE ? $::BUFSIZE : $h |
515 | or last; |
514 | or last; |
516 | } else { |
515 | } else { |
|
|
516 | undef $buf; |
|
|
517 | $aio_requests->down; |
517 | aio_read($fh, $l, ($h > $::BUFSIZE ? $::BUFSIZE : $h), |
518 | aio_read($fh, $l, ($h > $::BUFSIZE ? $::BUFSIZE : $h), |
518 | $buf, 0, sub { |
519 | $buf, 0, sub { |
519 | $r = $_[0]; |
520 | $r = $_[0]; |
520 | $current->ready; |
521 | $current->ready; |
521 | }); |
522 | }); |
522 | &Coro::schedule; |
523 | &Coro::schedule; |
|
|
524 | $aio_requests->up; |
523 | last unless $r; |
525 | last unless $r; |
524 | } |
526 | } |
525 | my $w = $self->{fh}->syswrite($buf) |
527 | my $w = $self->{fh}->syswrite($buf) |
526 | or last; |
528 | or last; |
527 | $::written += $w; |
529 | $::written += $w; |
528 | $self->{written} += $w; |
530 | $self->{written} += $w; |
529 | $l += $r; |
531 | $l += $r; |
530 | } |
532 | } |
531 | } |
|
|
532 | |
533 | |
533 | close $fh; |
534 | close $fh; |
|
|
535 | } |
534 | } |
536 | } |
535 | |
537 | |
536 | 1; |
538 | 1; |