1 |
use Coro; |
2 |
use Coro::Semaphore; |
3 |
use Coro::Event; |
4 |
use Coro::Socket; |
5 |
|
6 |
no utf8; |
7 |
use bytes; |
8 |
|
9 |
# at least on my machine, this thingy serves files |
10 |
# quite a bit faster than apache, ;) |
11 |
# and quite a bit slower than thttpd :( |
12 |
|
13 |
$SIG{PIPE} = 'IGNORE'; |
14 |
|
15 |
our $accesslog; |
16 |
|
17 |
if ($ACCESS_LOG) { |
18 |
use IO::Handle; |
19 |
open $accesslog, ">>$ACCESS_LOG" |
20 |
or die "$ACCESS_LOG: $!"; |
21 |
$accesslog->autoflush(1); |
22 |
} |
23 |
|
24 |
sub slog { |
25 |
my $level = shift; |
26 |
my $format = shift; |
27 |
printf "---: $format\n", @_; |
28 |
} |
29 |
|
30 |
my $connections = new Coro::Semaphore $MAX_CONNECTS; |
31 |
|
32 |
my @newcons; |
33 |
my @pool; |
34 |
|
35 |
# one "execution thread" |
36 |
sub handler { |
37 |
while () { |
38 |
my $new = pop @newcons; |
39 |
if ($new) { |
40 |
eval { |
41 |
conn->new(@$new)->handle; |
42 |
}; |
43 |
slog 1, "$@" if $@ && !ref $@; |
44 |
$connections->up; |
45 |
} else { |
46 |
last if @pool >= $MAX_POOL; |
47 |
push @pool, $Coro::current; |
48 |
schedule; |
49 |
} |
50 |
} |
51 |
} |
52 |
|
53 |
my $http_port = new Coro::Socket |
54 |
LocalAddr => $SERVER_HOST, |
55 |
LocalPort => $SERVER_PORT, |
56 |
ReuseAddr => 1, |
57 |
Listen => 50, |
58 |
or die "unable to start server"; |
59 |
|
60 |
push @listen_sockets, $http_port; |
61 |
|
62 |
# the "main thread" |
63 |
async { |
64 |
slog 1, "accepting connections"; |
65 |
while () { |
66 |
$connections->down; |
67 |
push @newcons, [$http_port->accept]; |
68 |
#slog 3, "accepted @$connections ".scalar(@pool); |
69 |
$::NOW = time; |
70 |
if (@pool) { |
71 |
(pop @pool)->ready; |
72 |
} else { |
73 |
async \&handler; |
74 |
} |
75 |
|
76 |
} |
77 |
}; |
78 |
|
79 |
package conn; |
80 |
|
81 |
use Socket; |
82 |
use HTTP::Date; |
83 |
use Convert::Scalar 'weaken'; |
84 |
use Linux::AIO; |
85 |
|
86 |
Linux::AIO::min_parallel $::AIO_PARALLEL; |
87 |
|
88 |
Event->io(fd => Linux::AIO::poll_fileno, |
89 |
poll => 'r', async => 1, |
90 |
cb => \&Linux::AIO::poll_cb); |
91 |
|
92 |
our %conn; # $conn{ip}{self} => connobj |
93 |
our %uri; # $uri{ip}{uri}{self} |
94 |
our %blocked; |
95 |
our %mimetype; |
96 |
|
97 |
sub read_mimetypes { |
98 |
local *M; |
99 |
if (open M, "<mime_types") { |
100 |
while (<M>) { |
101 |
if (/^([^#]\S+)\t+(\S+)$/) { |
102 |
$mimetype{lc $1} = $2; |
103 |
} |
104 |
} |
105 |
} else { |
106 |
print "cannot open mime_types\n"; |
107 |
} |
108 |
} |
109 |
|
110 |
read_mimetypes; |
111 |
|
112 |
sub new { |
113 |
my $class = shift; |
114 |
my $peername = shift; |
115 |
my $fh = shift; |
116 |
my $self = bless { fh => $fh }, $class; |
117 |
my (undef, $iaddr) = unpack_sockaddr_in $peername |
118 |
or $self->err(500, "unable to decode peername"); |
119 |
|
120 |
$self->{remote_addr} = inet_ntoa $iaddr; |
121 |
$self->{time} = $::NOW; |
122 |
|
123 |
# enter ourselves into various lists |
124 |
weaken ($conn{$self->{remote_addr}}{$self*1} = $self); |
125 |
|
126 |
$::conns++; |
127 |
|
128 |
$self; |
129 |
} |
130 |
|
131 |
sub DESTROY { |
132 |
my $self = shift; |
133 |
|
134 |
$::conns--; |
135 |
|
136 |
$self->eoconn; |
137 |
delete $conn{$self->{remote_addr}}{$self*1}; |
138 |
} |
139 |
|
140 |
# end of connection |
141 |
sub eoconn { |
142 |
my $self = shift; |
143 |
delete $uri{$self->{remote_addr}}{$self->{uri}}{$self*1}; |
144 |
} |
145 |
|
146 |
sub slog { |
147 |
my $self = shift; |
148 |
main::slog($_[0], "$self->{remote_addr}> $_[1]"); |
149 |
} |
150 |
|
151 |
sub response { |
152 |
my ($self, $code, $msg, $hdr, $content) = @_; |
153 |
my $res = "HTTP/1.1 $code $msg\015\012"; |
154 |
|
155 |
$self->{h}{connection} ||= $hdr->{Connection}; |
156 |
|
157 |
$res .= "Date: ".(time2str $::NOW)."\015\012"; # slow? nah. :( |
158 |
|
159 |
while (my ($h, $v) = each %$hdr) { |
160 |
$res .= "$h: $v\015\012" |
161 |
} |
162 |
$res .= "\015\012"; |
163 |
|
164 |
$res .= $content if defined $content and $self->{method} ne "HEAD"; |
165 |
|
166 |
my $log = "$self->{remote_addr} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}." \"$self->{h}{referer}\"\n"; |
167 |
|
168 |
print $accesslog $log if $accesslog; |
169 |
print STDERR $log; |
170 |
|
171 |
$self->{written} += |
172 |
print {$self->{fh}} $res; |
173 |
} |
174 |
|
175 |
sub err { |
176 |
my $self = shift; |
177 |
my ($code, $msg, $hdr, $content) = @_; |
178 |
|
179 |
unless (defined $content) { |
180 |
$content = "$code $msg"; |
181 |
$hdr->{"Content-Type"} = "text/plain"; |
182 |
$hdr->{"Content-Length"} = length $content; |
183 |
} |
184 |
$hdr->{"Connection"} = "close"; |
185 |
|
186 |
$self->response($code, $msg, $hdr, $content); |
187 |
|
188 |
die bless {}, err::; |
189 |
} |
190 |
|
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 { |
226 |
my $self = shift; |
227 |
my $fh = $self->{fh}; |
228 |
|
229 |
$fh->timeout($::REQ_TIMEOUT); |
230 |
while() { |
231 |
$self->{reqs}++; |
232 |
|
233 |
# read request and parse first line |
234 |
my $req = $fh->readline("\015\012\015\012"); |
235 |
|
236 |
unless (defined $req) { |
237 |
if (exists $self->{version}) { |
238 |
last; |
239 |
} else { |
240 |
$self->err(408, "request timeout"); |
241 |
} |
242 |
} |
243 |
|
244 |
$self->{h} = {}; |
245 |
|
246 |
$fh->timeout($::RES_TIMEOUT); |
247 |
my $ip = $self->{remote_addr}; |
248 |
|
249 |
if ($blocked{$ip}) { |
250 |
$self->err_blocked($blocked{$ip}) |
251 |
if $blocked{$ip} > $::NOW; |
252 |
|
253 |
delete $blocked{$ip}; |
254 |
} |
255 |
|
256 |
if (%{$conn{$ip}} > $::MAX_CONN_IP) { |
257 |
$self->slog(2, "blocked ip $ip"); |
258 |
$self->err_blocked; |
259 |
} |
260 |
|
261 |
$req =~ /^(?:\015\012)? |
262 |
(GET|HEAD) \040+ |
263 |
([^\040]+) \040+ |
264 |
HTTP\/([0-9]+\.[0-9]+) |
265 |
\015\012/gx |
266 |
or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
267 |
|
268 |
$self->{method} = $1; |
269 |
$self->{uri} = $2; |
270 |
$self->{version} = $3; |
271 |
|
272 |
$3 =~ /^1\./ |
273 |
or $self->err(506, "http protocol version $3 not supported"); |
274 |
|
275 |
# parse headers |
276 |
{ |
277 |
my (%hdr, $h, $v); |
278 |
|
279 |
$hdr{lc $1} .= ",$2" |
280 |
while $req =~ /\G |
281 |
([^:\000-\040]+): |
282 |
[\008\040]* |
283 |
((?: [^\015\012]+ | \015\012[\008\040] )*) |
284 |
\015\012 |
285 |
/gxc; |
286 |
|
287 |
$req =~ /\G\015\012$/ |
288 |
or $self->err(400, "bad request"); |
289 |
|
290 |
$self->{h}{$h} = substr $v, 1 |
291 |
while ($h, $v) = each %hdr; |
292 |
} |
293 |
|
294 |
$self->{server_port} = $self->{h}{host} =~ s/:([0-9]+)$// ? $1 : 80; |
295 |
|
296 |
weaken ($uri{$self->{remote_addr}}{$self->{uri}}{$self*1} = $self); |
297 |
|
298 |
eval { |
299 |
$self->map_uri; |
300 |
$self->respond; |
301 |
}; |
302 |
|
303 |
$self->eoconn; |
304 |
|
305 |
die if $@ && !ref $@; |
306 |
|
307 |
last if $self->{h}{connection} =~ /close/ || $self->{version} lt "1.1"; |
308 |
|
309 |
$self->slog(9, "persistent connection [".$self->{h}{"user-agent"}."][$self->{reqs}]"); |
310 |
$fh->timeout($::PER_TIMEOUT); |
311 |
} |
312 |
} |
313 |
|
314 |
# uri => path mapping |
315 |
sub map_uri { |
316 |
my $self = shift; |
317 |
my $host = $self->{h}{host} || "default"; |
318 |
my $uri = $self->{uri}; |
319 |
|
320 |
# some massaging, also makes it more secure |
321 |
$uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
322 |
$uri =~ s%//+%/%g; |
323 |
$uri =~ s%/\.(?=/|$)%%g; |
324 |
1 while $uri =~ s%/[^/]+/\.\.(?=/|$)%%; |
325 |
|
326 |
$uri =~ m%^/?\.\.(?=/|$)% |
327 |
and $self->err(400, "bad request"); |
328 |
|
329 |
$self->{name} = $uri; |
330 |
|
331 |
# now do the path mapping |
332 |
$self->{path} = "$::DOCROOT/$host$uri"; |
333 |
|
334 |
$self->access_check; |
335 |
} |
336 |
|
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 { |
366 |
my $self = shift; |
367 |
my $path = shift; |
368 |
my $fh; |
369 |
|
370 |
# no two-way xxx supported |
371 |
if (0 == fork) { |
372 |
open STDOUT, ">&".fileno($self->{fh}); |
373 |
if (chdir $::DOCROOT) { |
374 |
$ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
375 |
$ENV{HTTP_HOST} = $self->server_host; |
376 |
$ENV{HTTP_PORT} = $self->{server_host}; |
377 |
$ENV{SCRIPT_NAME} = $self->{name}; |
378 |
exec $path; |
379 |
} |
380 |
Coro::State::_exit(0); |
381 |
} else { |
382 |
} |
383 |
} |
384 |
|
385 |
sub respond { |
386 |
my $self = shift; |
387 |
my $path = $self->{path}; |
388 |
|
389 |
stat $path |
390 |
or $self->err(404, "not found"); |
391 |
|
392 |
$self->{stat} = [stat _]; |
393 |
|
394 |
# idiotic netscape sends idiotic headers AGAIN |
395 |
my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
396 |
? str2time $1 : 0; |
397 |
|
398 |
if (-d _ && -r _) { |
399 |
# directory |
400 |
if ($path !~ /\/$/) { |
401 |
# create a redirect to get the trailing "/" |
402 |
my $host = $self->server_hostport; |
403 |
$self->err(301, "moved permanently", { Location => "http://$host$self->{uri}/" }); |
404 |
} else { |
405 |
$ims < $self->{stat}[9] |
406 |
or $self->err(304, "not modified"); |
407 |
|
408 |
if (-r "$path/index.html") { |
409 |
$self->{path} .= "/index.html"; |
410 |
$self->handle_file; |
411 |
} else { |
412 |
$self->handle_dir; |
413 |
} |
414 |
} |
415 |
} elsif (-f _ && -r _) { |
416 |
-x _ and $self->err(403, "forbidden"); |
417 |
$self->handle_file; |
418 |
} else { |
419 |
$self->err(404, "not found"); |
420 |
} |
421 |
} |
422 |
|
423 |
sub handle_dir { |
424 |
my $self = shift; |
425 |
my $idx = $self->diridx; |
426 |
|
427 |
$self->response(200, "ok", |
428 |
{ |
429 |
"Content-Type" => "text/html", |
430 |
"Content-Length" => length $idx, |
431 |
}, |
432 |
$idx); |
433 |
} |
434 |
|
435 |
sub handle_file { |
436 |
my $self = shift; |
437 |
my $length = -s _; |
438 |
my $hdr = { |
439 |
"Last-Modified" => time2str ((stat _)[9]), |
440 |
}; |
441 |
|
442 |
my @code = (200, "ok"); |
443 |
my ($l, $h); |
444 |
|
445 |
if ($self->{h}{range} =~ /^bytes=(.*)$/) { |
446 |
for (split /,/, $1) { |
447 |
if (/^-(\d+)$/) { |
448 |
($l, $h) = ($length - $1, $length - 1); |
449 |
} elsif (/^(\d+)-(\d*)$/) { |
450 |
($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
451 |
} else { |
452 |
($l, $h) = (0, $length - 1); |
453 |
goto ignore; |
454 |
} |
455 |
goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
456 |
} |
457 |
$hdr->{"Content-Range"} = "bytes */$length"; |
458 |
$hdr->{"Content-Length"} = $length; |
459 |
$self->slog(9, "not satisfiable($self->{h}{range}|".$self->{h}{"user-agent"}.")"); |
460 |
$self->err(416, "not satisfiable", $hdr, ""); |
461 |
|
462 |
satisfiable: |
463 |
# check for segmented downloads |
464 |
if ($l && $::NO_SEGMENTED) { |
465 |
if (%{$uri{$self->{remote_addr}}{$self->{uri}}} > 1) { |
466 |
$self->err(400, "segmented downloads are not allowed", |
467 |
{ "Content-Type" => "text/html", Connection => "close" }, <<EOF); |
468 |
<html> |
469 |
<head> |
470 |
<title>Segmented downloads are not allowed</title> |
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 |
} |
481 |
} |
482 |
|
483 |
$hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
484 |
@code = (206, "partial content"); |
485 |
$length = $h - $l + 1; |
486 |
|
487 |
ignore: |
488 |
} else { |
489 |
($l, $h) = (0, $length - 1); |
490 |
} |
491 |
|
492 |
$self->{path} =~ /\.([^.]+)$/; |
493 |
$hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
494 |
$hdr->{"Content-Length"} = $length; |
495 |
|
496 |
$self->response(@code, $hdr, ""); |
497 |
|
498 |
if ($self->{method} eq "GET") { |
499 |
my ($fh, $buf, $r); |
500 |
my $current = $Coro::current; |
501 |
open $fh, "<", $self->{path} |
502 |
or die "$self->{path}: late open failure ($!)"; |
503 |
|
504 |
$h -= $l - 1; |
505 |
|
506 |
if (0) { |
507 |
if ($l) { |
508 |
sysseek $fh, $l, 0; |
509 |
} |
510 |
} |
511 |
|
512 |
while ($h > 0) { |
513 |
if (0) { |
514 |
sysread $fh, $buf, $h > $::BUFSIZE ? $::BUFSIZE : $h |
515 |
or last; |
516 |
} else { |
517 |
aio_read($fh, $l, ($h > $::BUFSIZE ? $::BUFSIZE : $h), |
518 |
$buf, 0, sub { |
519 |
$r = $_[0]; |
520 |
$current->ready; |
521 |
}); |
522 |
&Coro::schedule; |
523 |
last unless $r; |
524 |
} |
525 |
my $w = $self->{fh}->syswrite($buf) |
526 |
or last; |
527 |
$::written += $w; |
528 |
$self->{written} += $w; |
529 |
$l += $r; |
530 |
} |
531 |
} |
532 |
|
533 |
close $fh; |
534 |
} |
535 |
|
536 |
1; |