1 |
use Coro; |
2 |
use Coro::Semaphore; |
3 |
use Coro::Event; |
4 |
use Coro::Socket; |
5 |
use Coro::Signal; |
6 |
|
7 |
use HTTP::Date; |
8 |
use POSIX (); |
9 |
|
10 |
use Compress::Zlib (); |
11 |
|
12 |
no utf8; |
13 |
use bytes; |
14 |
|
15 |
# at least on my machine, this thingy serves files |
16 |
# quite a bit faster than apache, ;) |
17 |
# and quite a bit slower than thttpd :( |
18 |
|
19 |
$SIG{PIPE} = 'IGNORE'; |
20 |
|
21 |
our $accesslog; |
22 |
our $errorlog; |
23 |
|
24 |
our $NOW; |
25 |
our $HTTP_NOW; |
26 |
|
27 |
Event->timer(interval => 1, hard => 1, cb => sub { |
28 |
$NOW = time; |
29 |
$HTTP_NOW = time2str $NOW; |
30 |
})->now; |
31 |
|
32 |
if ($ERROR_LOG) { |
33 |
use IO::Handle; |
34 |
open $errorlog, ">>$ERROR_LOG" |
35 |
or die "$ERROR_LOG: $!"; |
36 |
$errorlog->autoflush(1); |
37 |
} |
38 |
|
39 |
if ($ACCESS_LOG) { |
40 |
use IO::Handle; |
41 |
open $accesslog, ">>$ACCESS_LOG" |
42 |
or die "$ACCESS_LOG: $!"; |
43 |
$accesslog->autoflush(1); |
44 |
} |
45 |
|
46 |
sub slog { |
47 |
my $level = shift; |
48 |
my $format = shift; |
49 |
my $NOW = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW); |
50 |
printf "$NOW: $format\n", @_; |
51 |
printf $errorlog "$NOW: $format\n", @_ if $errorlog; |
52 |
} |
53 |
|
54 |
our $connections = new Coro::Semaphore $MAX_CONNECTS || 250; |
55 |
our $httpevent = new Coro::Signal; |
56 |
|
57 |
our $queue_file = new transferqueue $MAX_TRANSFERS; |
58 |
our $queue_index = new transferqueue 10; |
59 |
|
60 |
my @newcons; |
61 |
my @pool; |
62 |
|
63 |
# one "execution thread" |
64 |
sub handler { |
65 |
while () { |
66 |
if (@newcons) { |
67 |
eval { |
68 |
conn->new(@{pop @newcons})->handle; |
69 |
}; |
70 |
slog 1, "$@" if $@ && !ref $@; |
71 |
|
72 |
$httpevent->broadcast; # only for testing, but doesn't matter much |
73 |
|
74 |
$connections->up; |
75 |
} else { |
76 |
last if @pool >= $MAX_POOL; |
77 |
push @pool, $Coro::current; |
78 |
schedule; |
79 |
} |
80 |
} |
81 |
} |
82 |
|
83 |
sub listen_on { |
84 |
my $listen = $_[0]; |
85 |
|
86 |
push @listen_sockets, $listen; |
87 |
|
88 |
# the "main thread" |
89 |
async { |
90 |
slog 1, "accepting connections"; |
91 |
while () { |
92 |
$connections->down; |
93 |
push @newcons, [$listen->accept]; |
94 |
#slog 3, "accepted @$connections ".scalar(@pool); |
95 |
if (@pool) { |
96 |
(pop @pool)->ready; |
97 |
} else { |
98 |
async \&handler; |
99 |
} |
100 |
|
101 |
} |
102 |
}; |
103 |
} |
104 |
|
105 |
my $http_port = new Coro::Socket |
106 |
LocalAddr => $SERVER_HOST, |
107 |
LocalPort => $SERVER_PORT, |
108 |
ReuseAddr => 1, |
109 |
Listen => 50, |
110 |
or die "unable to start server"; |
111 |
|
112 |
listen_on $http_port; |
113 |
|
114 |
if ($SERVER_PORT2) { |
115 |
my $http_port = new Coro::Socket |
116 |
LocalAddr => $SERVER_HOST, |
117 |
LocalPort => $SERVER_PORT2, |
118 |
ReuseAddr => 1, |
119 |
Listen => 50, |
120 |
or die "unable to start server"; |
121 |
|
122 |
listen_on $http_port; |
123 |
} |
124 |
|
125 |
package conn; |
126 |
|
127 |
use Socket; |
128 |
use HTTP::Date; |
129 |
use Convert::Scalar 'weaken'; |
130 |
use Linux::AIO; |
131 |
|
132 |
Linux::AIO::min_parallel $::AIO_PARALLEL; |
133 |
|
134 |
Event->io(fd => Linux::AIO::poll_fileno, |
135 |
poll => 'r', async => 1, |
136 |
cb => \&Linux::AIO::poll_cb); |
137 |
|
138 |
our %conn; # $conn{ip}{self} => connobj |
139 |
our %uri; # $uri{ip}{uri}{self} |
140 |
our %blocked; |
141 |
our %mimetype; |
142 |
|
143 |
sub read_mimetypes { |
144 |
local *M; |
145 |
if (open M, "<mime_types") { |
146 |
while (<M>) { |
147 |
if (/^([^#]\S+)\t+(\S+)$/) { |
148 |
$mimetype{lc $1} = $2; |
149 |
} |
150 |
} |
151 |
} else { |
152 |
print "cannot open mime_types\n"; |
153 |
} |
154 |
} |
155 |
|
156 |
read_mimetypes; |
157 |
|
158 |
sub new { |
159 |
my $class = shift; |
160 |
my $fh = shift; |
161 |
my $peername = shift; |
162 |
my $self = bless { fh => $fh }, $class; |
163 |
my (undef, $iaddr) = unpack_sockaddr_in $peername |
164 |
or $self->err(500, "unable to decode peername"); |
165 |
|
166 |
$self->{remote_addr} = |
167 |
$self->{remote_id} = inet_ntoa $iaddr; |
168 |
|
169 |
$self->{time} = $::NOW; |
170 |
|
171 |
weaken ($Coro::current->{conn} = $self); |
172 |
|
173 |
$::conns++; |
174 |
$::maxconns = $::conns if $::conns > $::maxconns; |
175 |
|
176 |
$self; |
177 |
} |
178 |
|
179 |
sub DESTROY { |
180 |
#my $self = shift; |
181 |
$::conns--; |
182 |
} |
183 |
|
184 |
sub slog { |
185 |
my $self = shift; |
186 |
main::slog($_[0], "$self->{remote_id}> $_[1]"); |
187 |
} |
188 |
|
189 |
sub response { |
190 |
my ($self, $code, $msg, $hdr, $content) = @_; |
191 |
my $res = "HTTP/1.1 $code $msg\015\012"; |
192 |
my $GZ = ""; |
193 |
|
194 |
if (exists $hdr->{Connection}) { |
195 |
if ($hdr->{Connection} =~ /close/) { |
196 |
$self->{h}{connection} = "close" |
197 |
} |
198 |
} else { |
199 |
if ($self->{version} < 1.1) { |
200 |
if ($self->{h}{connection} =~ /keep-alive/i) { |
201 |
$hdr->{Connection} = "Keep-Alive"; |
202 |
} else { |
203 |
$self->{h}{connection} = "close" |
204 |
} |
205 |
} |
206 |
} |
207 |
|
208 |
if ($self->{method} ne "HEAD" |
209 |
&& $self->{h}{"accept-encoding"} =~ /\bgzip\b/ |
210 |
&& 400 < length $content |
211 |
&& $hdr->{"Content-Length"} == length $content |
212 |
&& !exists $hdr->{"Content-Encoding"} |
213 |
) { |
214 |
my $orig = length $content; |
215 |
$hdr->{"Content-Encoding"} = "gzip"; |
216 |
$content = Compress::Zlib::memGzip(\$content); |
217 |
$hdr->{"Content-Length"} = length $content; |
218 |
$GZ = sprintf "GZ%02d", 100 - 100*((length $content) / $orig); |
219 |
} |
220 |
|
221 |
$res .= "Date: $HTTP_NOW\015\012"; |
222 |
|
223 |
while (my ($h, $v) = each %$hdr) { |
224 |
$res .= "$h: $v\015\012" |
225 |
} |
226 |
$res .= "\015\012"; |
227 |
|
228 |
$res .= $content if defined $content and $self->{method} ne "HEAD"; |
229 |
|
230 |
my $log = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW). |
231 |
" $self->{remote_id} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}.$GZ. |
232 |
" \"$self->{h}{referer}\"\n"; |
233 |
|
234 |
print $::accesslog $log if $::accesslog; |
235 |
print STDERR $log; |
236 |
|
237 |
$self->{written} += |
238 |
print {$self->{fh}} $res; |
239 |
} |
240 |
|
241 |
sub err { |
242 |
my $self = shift; |
243 |
my ($code, $msg, $hdr, $content) = @_; |
244 |
|
245 |
unless (defined $content) { |
246 |
$content = "$code $msg\n"; |
247 |
$hdr->{"Content-Type"} = "text/plain"; |
248 |
$hdr->{"Content-Length"} = length $content; |
249 |
} |
250 |
$hdr->{"Connection"} = "close"; |
251 |
|
252 |
$self->response($code, $msg, $hdr, $content); |
253 |
|
254 |
die bless {}, err::; |
255 |
} |
256 |
|
257 |
sub handle { |
258 |
my $self = shift; |
259 |
my $fh = $self->{fh}; |
260 |
|
261 |
my $host; |
262 |
|
263 |
$fh->timeout($::REQ_TIMEOUT); |
264 |
while() { |
265 |
$self->{reqs}++; |
266 |
|
267 |
# read request and parse first line |
268 |
my $req = $fh->readline("\015\012\015\012"); |
269 |
|
270 |
unless (defined $req) { |
271 |
if (exists $self->{version}) { |
272 |
last; |
273 |
} else { |
274 |
$self->err(408, "request timeout"); |
275 |
} |
276 |
} |
277 |
|
278 |
$self->{h} = {}; |
279 |
|
280 |
$fh->timeout($::RES_TIMEOUT); |
281 |
|
282 |
$req =~ /^(?:\015\012)? |
283 |
(GET|HEAD) \040+ |
284 |
([^\040]+) \040+ |
285 |
HTTP\/([0-9]+\.[0-9]+) |
286 |
\015\012/gx |
287 |
or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
288 |
|
289 |
$self->{method} = $1; |
290 |
$self->{uri} = $2; |
291 |
$self->{version} = $3; |
292 |
|
293 |
$3 =~ /^1\./ |
294 |
or $self->err(506, "http protocol version $3 not supported"); |
295 |
|
296 |
# parse headers |
297 |
{ |
298 |
my (%hdr, $h, $v); |
299 |
|
300 |
$hdr{lc $1} .= ",$2" |
301 |
while $req =~ /\G |
302 |
([^:\000-\040]+): |
303 |
[\010\040]* |
304 |
((?: [^\015\012]+ | \015\012[\010\040] )*) |
305 |
\015\012 |
306 |
/gxc; |
307 |
|
308 |
$req =~ /\G\015\012$/ |
309 |
or $self->err(400, "bad request"); |
310 |
|
311 |
$self->{h}{$h} = substr $v, 1 |
312 |
while ($h, $v) = each %hdr; |
313 |
} |
314 |
|
315 |
# remote id should be unique per user |
316 |
my $id = $self->{remote_addr}; |
317 |
|
318 |
if (exists $self->{h}{"client-ip"}) { |
319 |
$id .= "[".$self->{h}{"client-ip"}."]"; |
320 |
} elsif (exists $self->{h}{"x-forwarded-for"}) { |
321 |
$id .= "[".$self->{h}{"x-forwarded-for"}."]"; |
322 |
} |
323 |
|
324 |
$self->{remote_id} = $id; |
325 |
|
326 |
weaken (local $conn{$id}{$self*1} = $self); |
327 |
|
328 |
if ($blocked{$id}) { |
329 |
$self->err_blocked |
330 |
if $blocked{$id}[0] > $::NOW; |
331 |
|
332 |
delete $blocked{$id}; |
333 |
} |
334 |
|
335 |
# find out server name and port |
336 |
if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
337 |
$host = $1; |
338 |
} else { |
339 |
$host = $self->{h}{host}; |
340 |
} |
341 |
|
342 |
if (defined $host) { |
343 |
$self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
344 |
} else { |
345 |
($self->{server_port}, $host) |
346 |
= unpack_sockaddr_in $self->{fh}->sockname |
347 |
or $self->err(500, "unable to get socket name"); |
348 |
$host = inet_ntoa $host; |
349 |
} |
350 |
|
351 |
$self->{server_name} = $host; |
352 |
|
353 |
weaken (local $uri{$id}{$self->{uri}}{$self*1} = $self); |
354 |
|
355 |
eval { |
356 |
$self->map_uri; |
357 |
$self->respond; |
358 |
}; |
359 |
|
360 |
die if $@ && !ref $@; |
361 |
|
362 |
last if $self->{h}{connection} =~ /close/i; |
363 |
|
364 |
$httpevent->broadcast; |
365 |
|
366 |
$fh->timeout($::PER_TIMEOUT); |
367 |
} |
368 |
} |
369 |
|
370 |
sub block { |
371 |
my $self = shift; |
372 |
|
373 |
$blocked{$self->{remote_id}} = [$::NOW + $_[0], $_[1]]; |
374 |
$self->slog(2, "blocked ip $self->{remote_id}"); |
375 |
$self->err_blocked; |
376 |
} |
377 |
|
378 |
# uri => path mapping |
379 |
sub map_uri { |
380 |
my $self = shift; |
381 |
my $host = $self->{server_name}; |
382 |
my $uri = $self->{uri}; |
383 |
|
384 |
# some massaging, also makes it more secure |
385 |
$uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
386 |
$uri =~ s%//+%/%g; |
387 |
$uri =~ s%/\.(?=/|$)%%g; |
388 |
1 while $uri =~ s%/[^/]+/\.\.(?=/|$)%%; |
389 |
|
390 |
$uri =~ m%^/?\.\.(?=/|$)% |
391 |
and $self->err(400, "bad request"); |
392 |
|
393 |
$self->{name} = $uri; |
394 |
|
395 |
# now do the path mapping |
396 |
$self->{path} = "$::DOCROOT/$host$uri"; |
397 |
|
398 |
$self->access_check; |
399 |
} |
400 |
|
401 |
sub _cgi { |
402 |
my $self = shift; |
403 |
my $path = shift; |
404 |
my $fh; |
405 |
|
406 |
# no two-way xxx supported |
407 |
if (0 == fork) { |
408 |
open STDOUT, ">&".fileno($self->{fh}); |
409 |
if (chdir $::DOCROOT) { |
410 |
$ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
411 |
$ENV{HTTP_HOST} = $self->{server_name}; |
412 |
$ENV{HTTP_PORT} = $self->{server_port}; |
413 |
$ENV{SCRIPT_NAME} = $self->{name}; |
414 |
exec $path; |
415 |
} |
416 |
Coro::State::_exit(0); |
417 |
} else { |
418 |
die; |
419 |
} |
420 |
} |
421 |
|
422 |
sub server_hostport { |
423 |
$_[0]{server_port} == 80 |
424 |
? $_[0]{server_name} |
425 |
: "$_[0]{server_name}:$_[0]{server_port}"; |
426 |
} |
427 |
|
428 |
sub respond { |
429 |
my $self = shift; |
430 |
my $path = $self->{path}; |
431 |
|
432 |
if ($self->{name} =~ s%^/internal/([^/]+)%%) { |
433 |
if ($::internal{$1}) { |
434 |
$::internal{$1}->($self); |
435 |
} else { |
436 |
$self->err(404, "not found"); |
437 |
} |
438 |
} else { |
439 |
|
440 |
stat $path |
441 |
or $self->err(404, "not found"); |
442 |
|
443 |
$self->{stat} = [stat _]; |
444 |
|
445 |
# idiotic netscape sends idiotic headers AGAIN |
446 |
my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
447 |
? str2time $1 : 0; |
448 |
|
449 |
if (-d _ && -r _) { |
450 |
# directory |
451 |
if ($path !~ /\/$/) { |
452 |
# create a redirect to get the trailing "/" |
453 |
# we don't try to avoid the :80 |
454 |
$self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
455 |
} else { |
456 |
$ims < $self->{stat}[9] |
457 |
or $self->err(304, "not modified"); |
458 |
|
459 |
if (-r "$path/index.html") { |
460 |
# replace directory "size" by index.html filesize |
461 |
$self->{stat} = [stat ($self->{path} .= "/index.html")]; |
462 |
$self->handle_file($queue_index); |
463 |
} else { |
464 |
$self->handle_dir; |
465 |
} |
466 |
} |
467 |
} elsif (-f _ && -r _) { |
468 |
-x _ and $self->err(403, "forbidden"); |
469 |
|
470 |
if (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
471 |
my $timeout = $::NOW + 10; |
472 |
while (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
473 |
if ($timeout < $::NOW) { |
474 |
$self->block($::BLOCKTIME, "too many connections"); |
475 |
} else { |
476 |
$httpevent->wait; |
477 |
} |
478 |
} |
479 |
} |
480 |
|
481 |
$self->handle_file($queue_file); |
482 |
} else { |
483 |
$self->err(404, "not found"); |
484 |
} |
485 |
} |
486 |
} |
487 |
|
488 |
sub handle_dir { |
489 |
my $self = shift; |
490 |
my $idx = $self->diridx; |
491 |
|
492 |
$self->response(200, "ok", |
493 |
{ |
494 |
"Content-Type" => "text/html", |
495 |
"Content-Length" => length $idx, |
496 |
"Last-Modified" => time2str ($self->{stat}[9]), |
497 |
}, |
498 |
$idx); |
499 |
} |
500 |
|
501 |
sub handle_file { |
502 |
my ($self, $queue) = @_; |
503 |
my $length = $self->{stat}[7]; |
504 |
my $hdr = { |
505 |
"Last-Modified" => time2str ((stat _)[9]), |
506 |
}; |
507 |
|
508 |
my @code = (200, "ok"); |
509 |
my ($l, $h); |
510 |
|
511 |
if ($self->{h}{range} =~ /^bytes=(.*)$/) { |
512 |
for (split /,/, $1) { |
513 |
if (/^-(\d+)$/) { |
514 |
($l, $h) = ($length - $1, $length - 1); |
515 |
} elsif (/^(\d+)-(\d*)$/) { |
516 |
($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
517 |
} else { |
518 |
($l, $h) = (0, $length - 1); |
519 |
goto ignore; |
520 |
} |
521 |
goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
522 |
} |
523 |
$hdr->{"Content-Range"} = "bytes */$length"; |
524 |
$hdr->{"Content-Length"} = $length; |
525 |
$self->err(416, "not satisfiable", $hdr, ""); |
526 |
|
527 |
satisfiable: |
528 |
# check for segmented downloads |
529 |
if ($l && $::NO_SEGMENTED) { |
530 |
my $timeout = $::NOW + 15; |
531 |
while (keys %{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { |
532 |
if ($timeout <= $::NOW) { |
533 |
$self->block($::BLOCKTIME, "segmented downloads are forbidden"); |
534 |
#$self->err_segmented_download; |
535 |
} else { |
536 |
$httpevent->wait; |
537 |
} |
538 |
} |
539 |
} |
540 |
|
541 |
$hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
542 |
@code = (206, "partial content"); |
543 |
$length = $h - $l + 1; |
544 |
|
545 |
ignore: |
546 |
} else { |
547 |
($l, $h) = (0, $length - 1); |
548 |
} |
549 |
|
550 |
$self->{path} =~ /\.([^.]+)$/; |
551 |
$hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
552 |
$hdr->{"Content-Length"} = $length; |
553 |
|
554 |
$self->response(@code, $hdr, ""); |
555 |
|
556 |
if ($self->{method} eq "GET") { |
557 |
$self->{time} = $::NOW; |
558 |
|
559 |
my $current = $Coro::current; |
560 |
|
561 |
my ($fh, $buf, $r); |
562 |
|
563 |
open $fh, "<", $self->{path} |
564 |
or die "$self->{path}: late open failure ($!)"; |
565 |
|
566 |
$h -= $l - 1; |
567 |
|
568 |
if (0) { # !AIO |
569 |
if ($l) { |
570 |
sysseek $fh, $l, 0; |
571 |
} |
572 |
} |
573 |
|
574 |
my $transfer = $queue->start_transfer($h); |
575 |
my $locked; |
576 |
my $bufsize = $::WAIT_BUFSIZE; # initial buffer size |
577 |
|
578 |
while ($h > 0) { |
579 |
unless ($locked) { |
580 |
if ($locked ||= $transfer->try($::WAIT_INTERVAL)) { |
581 |
$bufsize = $::BUFSIZE; |
582 |
$self->{time} = $::NOW; |
583 |
} |
584 |
} |
585 |
|
586 |
if ($blocked{$self->{remote_id}}) { |
587 |
$self->{h}{connection} = "close"; |
588 |
die bless {}, err::; |
589 |
} |
590 |
|
591 |
if (0) { # !AIO |
592 |
sysread $fh, $buf, $h > $bufsize ? $bufsize : $h |
593 |
or last; |
594 |
} else { |
595 |
aio_read($fh, $l, ($h > $bufsize ? $bufsize : $h), |
596 |
$buf, 0, sub { |
597 |
$r = $_[0]; |
598 |
Coro::ready($current); |
599 |
}); |
600 |
&Coro::schedule; |
601 |
last unless $r; |
602 |
} |
603 |
my $w = syswrite $self->{fh}, $buf |
604 |
or last; |
605 |
$::written += $w; |
606 |
$self->{written} += $w; |
607 |
$l += $r; |
608 |
} |
609 |
|
610 |
close $fh; |
611 |
} |
612 |
} |
613 |
|
614 |
1; |