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 |
no utf8; |
11 |
use bytes; |
12 |
|
13 |
# at least on my machine, this thingy serves files |
14 |
# quite a bit faster than apache, ;) |
15 |
# and quite a bit slower than thttpd :( |
16 |
|
17 |
$SIG{PIPE} = 'IGNORE'; |
18 |
|
19 |
our $accesslog; |
20 |
our $errorlog; |
21 |
|
22 |
our $NOW; |
23 |
our $HTTP_NOW; |
24 |
|
25 |
Event->timer(interval => 1, hard => 1, cb => sub { |
26 |
$NOW = time; |
27 |
$HTTP_NOW = time2str $NOW; |
28 |
})->now; |
29 |
|
30 |
if ($ERROR_LOG) { |
31 |
use IO::Handle; |
32 |
open $errorlog, ">>$ERROR_LOG" |
33 |
or die "$ERROR_LOG: $!"; |
34 |
$errorlog->autoflush(1); |
35 |
} |
36 |
|
37 |
if ($ACCESS_LOG) { |
38 |
use IO::Handle; |
39 |
open $accesslog, ">>$ACCESS_LOG" |
40 |
or die "$ACCESS_LOG: $!"; |
41 |
$accesslog->autoflush(1); |
42 |
} |
43 |
|
44 |
sub slog { |
45 |
my $level = shift; |
46 |
my $format = shift; |
47 |
my $NOW = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW); |
48 |
printf "$NOW: $format\n", @_; |
49 |
printf $errorlog "$NOW: $format\n", @_ if $errorlog; |
50 |
} |
51 |
|
52 |
our $connections = new Coro::Semaphore $MAX_CONNECTS || 250; |
53 |
our $httpevent = new Coro::Signal; |
54 |
|
55 |
our $queue_file = new transferqueue slots => $MAX_TRANSFERS, maxsize => 256*1024*1024; |
56 |
our $queue_index = new transferqueue slots => 10; |
57 |
|
58 |
my @newcons; |
59 |
my @pool; |
60 |
|
61 |
# one "execution thread" |
62 |
sub handler { |
63 |
while () { |
64 |
if (@newcons) { |
65 |
eval { |
66 |
conn->new(@{pop @newcons})->handle; |
67 |
}; |
68 |
slog 1, "$@" if $@ && !ref $@; |
69 |
|
70 |
$httpevent->broadcast; # only for testing, but doesn't matter much |
71 |
|
72 |
$connections->up; |
73 |
} else { |
74 |
last if @pool >= $MAX_POOL; |
75 |
push @pool, $Coro::current; |
76 |
schedule; |
77 |
} |
78 |
} |
79 |
} |
80 |
|
81 |
sub listen_on { |
82 |
my $listen = $_[0]; |
83 |
|
84 |
push @listen_sockets, $listen; |
85 |
|
86 |
# the "main thread" |
87 |
async { |
88 |
slog 1, "accepting connections"; |
89 |
while () { |
90 |
$connections->down; |
91 |
push @newcons, [$listen->accept]; |
92 |
#slog 3, "accepted @$connections ".scalar(@pool); |
93 |
if (@pool) { |
94 |
(pop @pool)->ready; |
95 |
} else { |
96 |
async \&handler; |
97 |
} |
98 |
|
99 |
} |
100 |
}; |
101 |
} |
102 |
|
103 |
my $http_port = new Coro::Socket |
104 |
LocalAddr => $SERVER_HOST, |
105 |
LocalPort => $SERVER_PORT, |
106 |
ReuseAddr => 1, |
107 |
Listen => 50, |
108 |
or die "unable to start server"; |
109 |
|
110 |
listen_on $http_port; |
111 |
|
112 |
if ($SERVER_PORT2) { |
113 |
my $http_port = new Coro::Socket |
114 |
LocalAddr => $SERVER_HOST, |
115 |
LocalPort => $SERVER_PORT2, |
116 |
ReuseAddr => 1, |
117 |
Listen => 50, |
118 |
or die "unable to start server"; |
119 |
|
120 |
listen_on $http_port; |
121 |
} |
122 |
|
123 |
package conn; |
124 |
|
125 |
use Socket; |
126 |
use HTTP::Date; |
127 |
use Convert::Scalar 'weaken'; |
128 |
use Linux::AIO; |
129 |
|
130 |
Linux::AIO::min_parallel $::AIO_PARALLEL; |
131 |
|
132 |
Event->io(fd => Linux::AIO::poll_fileno, |
133 |
poll => 'r', async => 1, |
134 |
cb => \&Linux::AIO::poll_cb); |
135 |
|
136 |
our %conn; # $conn{ip}{self} => connobj |
137 |
our %uri; # $uri{ip}{uri}{self} |
138 |
our %blocked; |
139 |
our %mimetype; |
140 |
|
141 |
sub read_mimetypes { |
142 |
local *M; |
143 |
if (open M, "<mime_types") { |
144 |
while (<M>) { |
145 |
if (/^([^#]\S+)\t+(\S+)$/) { |
146 |
$mimetype{lc $1} = $2; |
147 |
} |
148 |
} |
149 |
} else { |
150 |
print "cannot open mime_types\n"; |
151 |
} |
152 |
} |
153 |
|
154 |
read_mimetypes; |
155 |
|
156 |
sub new { |
157 |
my $class = shift; |
158 |
my $fh = shift; |
159 |
my $peername = shift; |
160 |
my $self = bless { fh => $fh }, $class; |
161 |
my (undef, $iaddr) = unpack_sockaddr_in $peername |
162 |
or $self->err(500, "unable to decode peername"); |
163 |
|
164 |
$self->{remote_addr} = |
165 |
$self->{remote_id} = inet_ntoa $iaddr; |
166 |
|
167 |
$self->{time} = $::NOW; |
168 |
|
169 |
weaken ($Coro::current->{conn} = $self); |
170 |
|
171 |
$::conns++; |
172 |
$::maxconns = $::conns if $::conns > $::maxconns; |
173 |
|
174 |
$self; |
175 |
} |
176 |
|
177 |
sub DESTROY { |
178 |
#my $self = shift; |
179 |
$::conns--; |
180 |
} |
181 |
|
182 |
sub slog { |
183 |
my $self = shift; |
184 |
main::slog($_[0], "$self->{remote_id}> $_[1]"); |
185 |
} |
186 |
|
187 |
sub response { |
188 |
my ($self, $code, $msg, $hdr, $content) = @_; |
189 |
my $res = "HTTP/1.1 $code $msg\015\012"; |
190 |
|
191 |
if (exists $hdr->{Connection}) { |
192 |
if ($hdr->{Connection} =~ /close/) { |
193 |
$self->{h}{connection} = "close" |
194 |
} |
195 |
} else { |
196 |
if ($self->{version} < 1.1) { |
197 |
if ($self->{h}{connection} =~ /keep-alive/i) { |
198 |
$hdr->{Connection} = "Keep-Alive"; |
199 |
} else { |
200 |
$self->{h}{connection} = "close" |
201 |
} |
202 |
} |
203 |
} |
204 |
|
205 |
$res .= "Date: $HTTP_NOW\015\012"; |
206 |
|
207 |
while (my ($h, $v) = each %$hdr) { |
208 |
$res .= "$h: $v\015\012" |
209 |
} |
210 |
$res .= "\015\012"; |
211 |
|
212 |
$res .= $content if defined $content and $self->{method} ne "HEAD"; |
213 |
|
214 |
my $log = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $NOW). |
215 |
" $self->{remote_id} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}. |
216 |
" \"$self->{h}{referer}\"\n"; |
217 |
|
218 |
print $accesslog $log if $accesslog; |
219 |
print STDERR $log; |
220 |
|
221 |
$self->{written} += |
222 |
print {$self->{fh}} $res; |
223 |
} |
224 |
|
225 |
sub err { |
226 |
my $self = shift; |
227 |
my ($code, $msg, $hdr, $content) = @_; |
228 |
|
229 |
unless (defined $content) { |
230 |
$content = "$code $msg\n"; |
231 |
$hdr->{"Content-Type"} = "text/plain"; |
232 |
$hdr->{"Content-Length"} = length $content; |
233 |
} |
234 |
$hdr->{"Connection"} = "close"; |
235 |
|
236 |
$self->response($code, $msg, $hdr, $content); |
237 |
|
238 |
die bless {}, err::; |
239 |
} |
240 |
|
241 |
sub handle { |
242 |
my $self = shift; |
243 |
my $fh = $self->{fh}; |
244 |
|
245 |
my $host; |
246 |
|
247 |
$fh->timeout($::REQ_TIMEOUT); |
248 |
while() { |
249 |
$self->{reqs}++; |
250 |
|
251 |
# read request and parse first line |
252 |
my $req = $fh->readline("\015\012\015\012"); |
253 |
|
254 |
unless (defined $req) { |
255 |
if (exists $self->{version}) { |
256 |
last; |
257 |
} else { |
258 |
$self->err(408, "request timeout"); |
259 |
} |
260 |
} |
261 |
|
262 |
$self->{h} = {}; |
263 |
|
264 |
$fh->timeout($::RES_TIMEOUT); |
265 |
|
266 |
$req =~ /^(?:\015\012)? |
267 |
(GET|HEAD) \040+ |
268 |
([^\040]+) \040+ |
269 |
HTTP\/([0-9]+\.[0-9]+) |
270 |
\015\012/gx |
271 |
or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
272 |
|
273 |
$self->{method} = $1; |
274 |
$self->{uri} = $2; |
275 |
$self->{version} = $3; |
276 |
|
277 |
$3 =~ /^1\./ |
278 |
or $self->err(506, "http protocol version $3 not supported"); |
279 |
|
280 |
# parse headers |
281 |
{ |
282 |
my (%hdr, $h, $v); |
283 |
|
284 |
$hdr{lc $1} .= ",$2" |
285 |
while $req =~ /\G |
286 |
([^:\000-\040]+): |
287 |
[\008\040]* |
288 |
((?: [^\015\012]+ | \015\012[\008\040] )*) |
289 |
\015\012 |
290 |
/gxc; |
291 |
|
292 |
$req =~ /\G\015\012$/ |
293 |
or $self->err(400, "bad request"); |
294 |
|
295 |
$self->{h}{$h} = substr $v, 1 |
296 |
while ($h, $v) = each %hdr; |
297 |
} |
298 |
|
299 |
# remote id should be unique per user |
300 |
my $id = $self->{remote_addr}; |
301 |
|
302 |
if (exists $self->{h}{"client-ip"}) { |
303 |
$id .= "[".$self->{h}{"client-ip"}."]"; |
304 |
} elsif (exists $self->{h}{"x-forwarded-for"}) { |
305 |
$id .= "[".$self->{h}{"x-forwarded-for"}."]"; |
306 |
} |
307 |
|
308 |
$self->{remote_id} = $id; |
309 |
|
310 |
weaken (local $conn{$id}{$self*1} = $self); |
311 |
|
312 |
if ($blocked{$id}) { |
313 |
$self->err_blocked |
314 |
if $blocked{$id}[0] > $::NOW; |
315 |
|
316 |
delete $blocked{$id}; |
317 |
} |
318 |
|
319 |
# find out server name and port |
320 |
if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
321 |
$host = $1; |
322 |
} else { |
323 |
$host = $self->{h}{host}; |
324 |
} |
325 |
|
326 |
if (defined $host) { |
327 |
$self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
328 |
} else { |
329 |
($self->{server_port}, $host) |
330 |
= unpack_sockaddr_in $self->{fh}->sockname |
331 |
or $self->err(500, "unable to get socket name"); |
332 |
$host = inet_ntoa $host; |
333 |
} |
334 |
|
335 |
$self->{server_name} = $host; |
336 |
|
337 |
weaken (local $uri{$id}{$self->{uri}}{$self*1} = $self); |
338 |
|
339 |
eval { |
340 |
$self->map_uri; |
341 |
$self->respond; |
342 |
}; |
343 |
|
344 |
die if $@ && !ref $@; |
345 |
|
346 |
last if $self->{h}{connection} =~ /close/i; |
347 |
|
348 |
$httpevent->broadcast; |
349 |
|
350 |
$fh->timeout($::PER_TIMEOUT); |
351 |
} |
352 |
} |
353 |
|
354 |
sub block { |
355 |
my $self = shift; |
356 |
|
357 |
$blocked{$self->{remote_id}} = [$::NOW + $_[0], $_[1]]; |
358 |
$self->slog(2, "blocked ip $self->{remote_id}"); |
359 |
$self->err_blocked; |
360 |
} |
361 |
|
362 |
# uri => path mapping |
363 |
sub map_uri { |
364 |
my $self = shift; |
365 |
my $host = $self->{server_name}; |
366 |
my $uri = $self->{uri}; |
367 |
|
368 |
# some massaging, also makes it more secure |
369 |
$uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
370 |
$uri =~ s%//+%/%g; |
371 |
$uri =~ s%/\.(?=/|$)%%g; |
372 |
1 while $uri =~ s%/[^/]+/\.\.(?=/|$)%%; |
373 |
|
374 |
$uri =~ m%^/?\.\.(?=/|$)% |
375 |
and $self->err(400, "bad request"); |
376 |
|
377 |
$self->{name} = $uri; |
378 |
|
379 |
# now do the path mapping |
380 |
$self->{path} = "$::DOCROOT/$host$uri"; |
381 |
|
382 |
$self->access_check; |
383 |
} |
384 |
|
385 |
sub _cgi { |
386 |
my $self = shift; |
387 |
my $path = shift; |
388 |
my $fh; |
389 |
|
390 |
# no two-way xxx supported |
391 |
if (0 == fork) { |
392 |
open STDOUT, ">&".fileno($self->{fh}); |
393 |
if (chdir $::DOCROOT) { |
394 |
$ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
395 |
$ENV{HTTP_HOST} = $self->{server_name}; |
396 |
$ENV{HTTP_PORT} = $self->{server_port}; |
397 |
$ENV{SCRIPT_NAME} = $self->{name}; |
398 |
exec $path; |
399 |
} |
400 |
Coro::State::_exit(0); |
401 |
} else { |
402 |
die; |
403 |
} |
404 |
} |
405 |
|
406 |
sub server_hostport { |
407 |
$_[0]{server_port} == 80 |
408 |
? $_[0]{server_name} |
409 |
: "$_[0]{server_name}:$_[0]{server_port}"; |
410 |
} |
411 |
|
412 |
sub respond { |
413 |
my $self = shift; |
414 |
my $path = $self->{path}; |
415 |
|
416 |
if ($self->{name} =~ s%^/internal/([^/]+)%%) { |
417 |
if ($::internal{$1}) { |
418 |
$::internal{$1}->($self); |
419 |
} else { |
420 |
$self->err(404, "not found"); |
421 |
} |
422 |
} else { |
423 |
|
424 |
stat $path |
425 |
or $self->err(404, "not found"); |
426 |
|
427 |
$self->{stat} = [stat _]; |
428 |
|
429 |
# idiotic netscape sends idiotic headers AGAIN |
430 |
my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
431 |
? str2time $1 : 0; |
432 |
|
433 |
if (-d _ && -r _) { |
434 |
# directory |
435 |
if ($path !~ /\/$/) { |
436 |
# create a redirect to get the trailing "/" |
437 |
# we don't try to avoid the :80 |
438 |
$self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
439 |
} else { |
440 |
$ims < $self->{stat}[9] |
441 |
or $self->err(304, "not modified"); |
442 |
|
443 |
if (-r "$path/index.html") { |
444 |
# replace directory "size" by index.html filesize |
445 |
$self->{stat} = [stat ($self->{path} .= "/index.html")]; |
446 |
$self->handle_file($queue_index); |
447 |
} else { |
448 |
$self->handle_dir; |
449 |
} |
450 |
} |
451 |
} elsif (-f _ && -r _) { |
452 |
-x _ and $self->err(403, "forbidden"); |
453 |
|
454 |
if (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
455 |
my $timeout = $::NOW + 10; |
456 |
while (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
457 |
if ($timeout < $::NOW) { |
458 |
$self->block($::BLOCKTIME, "too many connections"); |
459 |
} else { |
460 |
$httpevent->wait; |
461 |
} |
462 |
} |
463 |
} |
464 |
|
465 |
$self->handle_file($queue_file); |
466 |
} else { |
467 |
$self->err(404, "not found"); |
468 |
} |
469 |
} |
470 |
} |
471 |
|
472 |
sub handle_dir { |
473 |
my $self = shift; |
474 |
my $idx = $self->diridx; |
475 |
|
476 |
$self->response(200, "ok", |
477 |
{ |
478 |
"Content-Type" => "text/html", |
479 |
"Content-Length" => length $idx, |
480 |
"Last-Modified" => time2str ($self->{stat}[9]), |
481 |
}, |
482 |
$idx); |
483 |
} |
484 |
|
485 |
sub handle_file { |
486 |
my ($self, $queue) = @_; |
487 |
my $length = $self->{stat}[7]; |
488 |
my $hdr = { |
489 |
"Last-Modified" => time2str ((stat _)[9]), |
490 |
}; |
491 |
|
492 |
my @code = (200, "ok"); |
493 |
my ($l, $h); |
494 |
|
495 |
if ($self->{h}{range} =~ /^bytes=(.*)$/) { |
496 |
for (split /,/, $1) { |
497 |
if (/^-(\d+)$/) { |
498 |
($l, $h) = ($length - $1, $length - 1); |
499 |
} elsif (/^(\d+)-(\d*)$/) { |
500 |
($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
501 |
} else { |
502 |
($l, $h) = (0, $length - 1); |
503 |
goto ignore; |
504 |
} |
505 |
goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
506 |
} |
507 |
$hdr->{"Content-Range"} = "bytes */$length"; |
508 |
$hdr->{"Content-Length"} = $length; |
509 |
$self->err(416, "not satisfiable", $hdr, ""); |
510 |
|
511 |
satisfiable: |
512 |
# check for segmented downloads |
513 |
if ($l && $::NO_SEGMENTED) { |
514 |
my $timeout = $::NOW + 15; |
515 |
while (keys %{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { |
516 |
if ($timeout <= $::NOW) { |
517 |
$self->block($::BLOCKTIME, "segmented downloads are forbidden"); |
518 |
#$self->err_segmented_download; |
519 |
} else { |
520 |
$httpevent->wait; |
521 |
} |
522 |
} |
523 |
} |
524 |
|
525 |
$hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
526 |
@code = (206, "partial content"); |
527 |
$length = $h - $l + 1; |
528 |
|
529 |
ignore: |
530 |
} else { |
531 |
($l, $h) = (0, $length - 1); |
532 |
} |
533 |
|
534 |
$self->{path} =~ /\.([^.]+)$/; |
535 |
$hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
536 |
$hdr->{"Content-Length"} = $length; |
537 |
|
538 |
$self->response(@code, $hdr, ""); |
539 |
|
540 |
if ($self->{method} eq "GET") { |
541 |
$self->{time} = $::NOW; |
542 |
|
543 |
my $current = $Coro::current; |
544 |
|
545 |
my ($fh, $buf, $r); |
546 |
|
547 |
open $fh, "<", $self->{path} |
548 |
or die "$self->{path}: late open failure ($!)"; |
549 |
|
550 |
$h -= $l - 1; |
551 |
|
552 |
if (0) { # !AIO |
553 |
if ($l) { |
554 |
sysseek $fh, $l, 0; |
555 |
} |
556 |
} |
557 |
|
558 |
my $transfer = $queue->start_transfer($h); |
559 |
my $locked; |
560 |
my $bufsize = $::WAIT_BUFSIZE; # initial buffer size |
561 |
|
562 |
while ($h > 0) { |
563 |
unless ($locked) { |
564 |
if ($locked ||= $transfer->try($::WAIT_INTERVAL)) { |
565 |
$bufsize = $::BUFSIZE; |
566 |
$self->{time} = $::NOW; |
567 |
} |
568 |
} |
569 |
|
570 |
if ($blocked{$self->{remote_id}}) { |
571 |
$self->{h}{connection} = "close"; |
572 |
die bless {}, err::; |
573 |
} |
574 |
|
575 |
if (0) { # !AIO |
576 |
sysread $fh, $buf, $h > $bufsize ? $bufsize : $h |
577 |
or last; |
578 |
} else { |
579 |
aio_read($fh, $l, ($h > $bufsize ? $bufsize : $h), |
580 |
$buf, 0, sub { |
581 |
$r = $_[0]; |
582 |
Coro::ready($current); |
583 |
}); |
584 |
&Coro::schedule; |
585 |
last unless $r; |
586 |
} |
587 |
my $w = syswrite $self->{fh}, $buf |
588 |
or last; |
589 |
$::written += $w; |
590 |
$self->{written} += $w; |
591 |
$l += $r; |
592 |
} |
593 |
|
594 |
close $fh; |
595 |
} |
596 |
} |
597 |
|
598 |
1; |