1 |
use Coro; |
2 |
use Coro::Semaphore; |
3 |
use Coro::SemaphoreSet; |
4 |
use Coro::EV; |
5 |
use Coro::Socket; |
6 |
use Coro::Signal; |
7 |
use Coro::AIO (); |
8 |
|
9 |
use Fcntl; |
10 |
use HTTP::Date; |
11 |
use POSIX (); |
12 |
|
13 |
use Compress::Zlib (); |
14 |
|
15 |
use common::sense; |
16 |
|
17 |
# at least on my machine, this thingy serves files |
18 |
# quite a bit faster than apache, ;) |
19 |
# and quite a bit slower than thttpd :( |
20 |
|
21 |
$SIG{PIPE} = 'IGNORE'; |
22 |
|
23 |
our $accesslog; |
24 |
our $errorlog; |
25 |
our @listen_sockets; |
26 |
|
27 |
our $NOW; |
28 |
our $HTTP_NOW; |
29 |
|
30 |
our $ERROR_LOG; |
31 |
our $ACCESS_LOG; |
32 |
our $TRANSFER_LOCK = new Coro::SemaphoreSet; # lock to be acquired per ip |
33 |
|
34 |
our $update_time = EV::periodic 0, 1, undef, sub { |
35 |
$NOW = time; |
36 |
$HTTP_NOW = time2str $NOW; |
37 |
}; |
38 |
$update_time->invoke; |
39 |
|
40 |
if ($ERROR_LOG) { |
41 |
use IO::Handle; |
42 |
open $errorlog, ">>$ERROR_LOG" |
43 |
or die "$ERROR_LOG: $!"; |
44 |
$errorlog->autoflush(1); |
45 |
} |
46 |
|
47 |
if ($ACCESS_LOG) { |
48 |
use IO::Handle; |
49 |
open $accesslog, ">>$ACCESS_LOG" |
50 |
or die "$ACCESS_LOG: $!"; |
51 |
$accesslog->autoflush(1); |
52 |
} |
53 |
|
54 |
sub slog { |
55 |
my $level = shift; |
56 |
my $format = shift; |
57 |
my $NOW = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW); |
58 |
printf "$NOW: $format\n", @_; |
59 |
printf $errorlog "$NOW: $format\n", @_ if $errorlog; |
60 |
} |
61 |
|
62 |
our $connections = new Coro::Semaphore $::MAX_CONNECTS || 250; |
63 |
our $httpevent = new Coro::Signal; |
64 |
|
65 |
our $queue_file = new transferqueue $::MAX_TRANSFERS; |
66 |
our $queue_index = new transferqueue 10; |
67 |
|
68 |
our $tbf_top = new tbf rate => $::TBF_RATE || 100000; |
69 |
|
70 |
my $unused_bytes = 0; |
71 |
my $unused_last = time; |
72 |
|
73 |
sub unused_bandwidth { |
74 |
$unused_bytes += $_[0]; |
75 |
if ($unused_last < $NOW - 30 && $unused_bytes / ($NOW - $unused_last) > 50000) { |
76 |
$unused_last = $NOW; |
77 |
$unused_bytes = 0; |
78 |
$queue_file->force_wake_next; |
79 |
slog 1, "forced filetransfer due to unused bandwidth"; |
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 |
my @conn = $listen->accept; |
94 |
#slog 3, "accepted @$connections ".scalar(@pool); |
95 |
|
96 |
async_pool { |
97 |
eval { |
98 |
conn->new (@conn)->handle; |
99 |
}; |
100 |
slog 1, "$@" if $@ && !ref $@; |
101 |
|
102 |
$httpevent->broadcast; # only for testing, but doesn't matter much |
103 |
|
104 |
$connections->up; |
105 |
} |
106 |
} |
107 |
}; |
108 |
} |
109 |
|
110 |
my $http_port = new Coro::Socket |
111 |
LocalAddr => $::SERVER_HOST, |
112 |
LocalPort => $::SERVER_PORT, |
113 |
ReuseAddr => 1, |
114 |
Listen => 50, |
115 |
or die "unable to start server"; |
116 |
|
117 |
listen_on $http_port; |
118 |
|
119 |
if ($::SERVER_PORT2) { |
120 |
my $http_port = new Coro::Socket |
121 |
LocalAddr => $::SERVER_HOST, |
122 |
LocalPort => $::SERVER_PORT2, |
123 |
ReuseAddr => 1, |
124 |
Listen => 50, |
125 |
or die "unable to start server"; |
126 |
|
127 |
listen_on $http_port; |
128 |
} |
129 |
|
130 |
package conn; |
131 |
|
132 |
use common::sense; |
133 |
|
134 |
use Socket; |
135 |
use HTTP::Date; |
136 |
use Convert::Scalar 'weaken'; |
137 |
use IO::AIO; |
138 |
|
139 |
IO::AIO::min_parallel $::AIO_PARALLEL; |
140 |
|
141 |
our $AIO_WATCHER = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb; |
142 |
|
143 |
our %conn; # $conn{ip}{self} => connobj |
144 |
our %uri; # $uri{ip}{uri}{self} |
145 |
our %blocked; |
146 |
our %mimetype; |
147 |
|
148 |
sub read_mimetypes { |
149 |
if (open my $fh, "<mime_types") { |
150 |
while (<$fh>) { |
151 |
if (/^([^#]\S+)\t+(\S+)$/) { |
152 |
$mimetype{lc $1} = $2; |
153 |
} |
154 |
} |
155 |
} else { |
156 |
print "cannot open mime_types\n"; |
157 |
} |
158 |
} |
159 |
|
160 |
read_mimetypes; |
161 |
|
162 |
sub new { |
163 |
my $class = shift; |
164 |
my $fh = shift; |
165 |
my $peername = shift; |
166 |
my $self = bless { fh => $fh }, $class; |
167 |
my (undef, $iaddr) = unpack_sockaddr_in $peername |
168 |
or $self->err (500, "unable to decode peername"); |
169 |
|
170 |
$self->{remote_addr} = |
171 |
$self->{remote_id} = inet_ntoa $iaddr; |
172 |
|
173 |
$self->{time} = $::NOW; |
174 |
|
175 |
weaken ($Coro::current->{conn} = $self); |
176 |
|
177 |
++$::conns; |
178 |
$::maxconns = $::conns if $::conns > $::maxconns; |
179 |
|
180 |
$self |
181 |
} |
182 |
|
183 |
sub DESTROY { |
184 |
my $self = shift; |
185 |
|
186 |
--$::conns; |
187 |
} |
188 |
|
189 |
sub prune_cache { |
190 |
my $hash = $_[0]; |
191 |
|
192 |
for (keys %$hash) { |
193 |
if (ref $hash->{$_} eq HASH::) { |
194 |
prune_cache($hash->{$_}); |
195 |
unless (scalar keys %{$hash->{$_}}) { |
196 |
delete $hash->{$_}; |
197 |
} |
198 |
} |
199 |
} |
200 |
} |
201 |
|
202 |
sub prune_caches { |
203 |
prune_cache \%conn; |
204 |
prune_cache \%uri; |
205 |
|
206 |
for (keys %blocked) { |
207 |
delete $blocked{$_} unless $blocked{$_}[0] > $::NOW; |
208 |
} |
209 |
} |
210 |
|
211 |
our $PRUNE_WATCHER = EV::timer 60, 60, \&prune_caches; |
212 |
|
213 |
sub slog { |
214 |
my $self = shift; |
215 |
main::slog($_[0], "$self->{remote_id}> $_[1]"); |
216 |
} |
217 |
|
218 |
sub response { |
219 |
my ($self, $code, $msg, $hdr, $content) = @_; |
220 |
my $res = "HTTP/1.1 $code $msg\015\012"; |
221 |
my $GZ = ""; |
222 |
|
223 |
if (exists $hdr->{Connection}) { |
224 |
if ($hdr->{Connection} =~ /close/) { |
225 |
$self->{h}{connection} = "close" |
226 |
} |
227 |
} else { |
228 |
if ($self->{version} < 1.1) { |
229 |
if ($self->{h}{connection} =~ /keep-alive/i) { |
230 |
$hdr->{Connection} = "Keep-Alive"; |
231 |
} else { |
232 |
$self->{h}{connection} = "close" |
233 |
} |
234 |
} |
235 |
} |
236 |
|
237 |
if ($self->{method} ne "HEAD" |
238 |
&& $self->{h}{"accept-encoding"} =~ /\bgzip\b/ |
239 |
&& 400 < length $content |
240 |
&& $hdr->{"Content-Length"} == length $content |
241 |
&& !exists $hdr->{"Content-Encoding"} |
242 |
) { |
243 |
my $orig = length $content; |
244 |
$hdr->{"Content-Encoding"} = "gzip"; |
245 |
$content = Compress::Zlib::memGzip(\$content); |
246 |
$hdr->{"Content-Length"} = length $content; |
247 |
$GZ = sprintf "GZ%02d", 100 - 100*((length $content) / $orig); |
248 |
} |
249 |
|
250 |
$res .= "Date: $HTTP_NOW\015\012"; |
251 |
$res .= "Server: $::NAME\015\012"; |
252 |
|
253 |
while (my ($h, $v) = each %$hdr) { |
254 |
$res .= "$h: $v\015\012" |
255 |
} |
256 |
$res .= "\015\012"; |
257 |
|
258 |
$res .= $content if defined $content and $self->{method} ne "HEAD"; |
259 |
|
260 |
my $log = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW). |
261 |
" $self->{remote_id} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}.$GZ. |
262 |
" \"$self->{h}{referer}\"\n"; |
263 |
|
264 |
print $::accesslog $log if $::accesslog; |
265 |
print STDERR $log; |
266 |
|
267 |
$tbf_top->request(length $res, 1e6); |
268 |
$self->{written} += print {$self->{fh}} $res; |
269 |
} |
270 |
|
271 |
sub err { |
272 |
my $self = shift; |
273 |
my ($code, $msg, $hdr, $content) = @_; |
274 |
|
275 |
unless (defined $content) { |
276 |
$content = "$code $msg\n"; |
277 |
$hdr->{"Content-Type"} = "text/plain"; |
278 |
$hdr->{"Content-Length"} = length $content; |
279 |
} |
280 |
$hdr->{"Connection"} = "close"; |
281 |
|
282 |
$self->response ($code, $msg, $hdr, $content); |
283 |
|
284 |
die bless {}, err:: |
285 |
} |
286 |
|
287 |
sub handle { |
288 |
my $self = shift; |
289 |
my $fh = $self->{fh}; |
290 |
|
291 |
my $host; |
292 |
|
293 |
$fh->timeout($::REQ_TIMEOUT); |
294 |
while () { |
295 |
$self->{reqs}++; |
296 |
|
297 |
# read request and parse first line |
298 |
my $req = $fh->readline("\015\012\015\012"); |
299 |
|
300 |
unless (defined $req) { |
301 |
if (exists $self->{version}) { |
302 |
last; |
303 |
} else { |
304 |
$self->err(408, "request timeout"); |
305 |
} |
306 |
} |
307 |
|
308 |
$self->{h} = {}; |
309 |
|
310 |
$fh->timeout($::RES_TIMEOUT); |
311 |
|
312 |
$req =~ /^(?:\015\012)? |
313 |
(GET|HEAD) \040+ |
314 |
([^\040]+) \040+ |
315 |
HTTP\/([0-9]+\.[0-9]+) |
316 |
\015\012/gx |
317 |
or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
318 |
|
319 |
$self->{method} = $1; |
320 |
$self->{uri} = $2; |
321 |
$self->{version} = $3; |
322 |
|
323 |
$3 =~ /^1\./ |
324 |
or $self->err(506, "http protocol version $3 not supported"); |
325 |
|
326 |
# parse headers |
327 |
{ |
328 |
my (%hdr, $h, $v); |
329 |
|
330 |
$hdr{lc $1} .= ",$2" |
331 |
while $req =~ /\G |
332 |
([^:\000-\040]+): |
333 |
[\011\040]* |
334 |
((?: [^\015\012]+ | \015\012[\011\040] )*) |
335 |
\015\012 |
336 |
/gxc; |
337 |
|
338 |
$req =~ /\G\015\012$/ |
339 |
or $self->err(400, "bad request"); |
340 |
|
341 |
$self->{h}{$h} = substr $v, 1 |
342 |
while ($h, $v) = each %hdr; |
343 |
} |
344 |
|
345 |
# remote id should be unique per user |
346 |
my $id = $self->{remote_addr}; |
347 |
|
348 |
if (exists $self->{h}{"client-ip"}) { |
349 |
$id .= "[".$self->{h}{"client-ip"}."]"; |
350 |
} elsif (exists $self->{h}{"x-forwarded-for"}) { |
351 |
$id .= "[".$self->{h}{"x-forwarded-for"}."]"; |
352 |
} |
353 |
|
354 |
$self->{remote_id} = $id; |
355 |
|
356 |
weaken (local $conn{$id}{$self*1} = $self); |
357 |
|
358 |
if ($blocked{$id}) { |
359 |
$self->err_blocked |
360 |
if $blocked{$id}[0] > $::NOW; |
361 |
|
362 |
delete $blocked{$id}; |
363 |
} |
364 |
|
365 |
# find out server name and port |
366 |
if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
367 |
$host = $1; |
368 |
} else { |
369 |
$host = $self->{h}{host}; |
370 |
} |
371 |
|
372 |
if (defined $host) { |
373 |
$self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
374 |
} else { |
375 |
($self->{server_port}, $host) |
376 |
= unpack_sockaddr_in $self->{fh}->sockname |
377 |
or $self->err(500, "unable to get socket name"); |
378 |
$host = inet_ntoa $host; |
379 |
} |
380 |
|
381 |
$self->{server_name} = $host; |
382 |
|
383 |
weaken (local $uri{$id}{$self->{uri}}{$self*1} = $self); |
384 |
|
385 |
eval { |
386 |
$self->map_uri; |
387 |
$self->respond; |
388 |
}; |
389 |
|
390 |
die if $@ && !ref $@; |
391 |
|
392 |
last if $self->{h}{connection} =~ /close/i; |
393 |
|
394 |
$httpevent->broadcast; |
395 |
|
396 |
$fh->timeout($::PER_TIMEOUT); |
397 |
} |
398 |
} |
399 |
|
400 |
sub block { |
401 |
my $self = shift; |
402 |
|
403 |
$blocked{$self->{remote_id}} = [$::NOW + $_[0], $_[1]]; |
404 |
$self->slog(2, "blocked ip $self->{remote_id}"); |
405 |
$self->err_blocked; |
406 |
} |
407 |
|
408 |
# uri => path mapping |
409 |
sub map_uri { |
410 |
my $self = shift; |
411 |
my $host = $self->{server_name}; |
412 |
my $uri = $self->{uri}; |
413 |
|
414 |
$host =~ /[\/\\]/ |
415 |
and $self->err(400, "bad request"); |
416 |
|
417 |
# some massaging, also makes it more secure |
418 |
$uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
419 |
$uri =~ s%//+%/%g; |
420 |
$uri =~ s%/\.(?=/|$)%%g; |
421 |
1 while $uri =~ s%/[^/]+/\.\.(?=/|$)%%; |
422 |
|
423 |
$uri =~ m%^/?\.\.(?=/|$)% |
424 |
and $self->err(400, "bad request"); |
425 |
|
426 |
$self->{name} = $uri; |
427 |
|
428 |
# now do the path mapping |
429 |
$self->{path} = "$::DOCROOT/$host$uri"; |
430 |
|
431 |
$self->access_check; |
432 |
} |
433 |
|
434 |
sub _cgi { |
435 |
my $self = shift; |
436 |
my $path = shift; |
437 |
my $fh; |
438 |
|
439 |
# no two-way xxx supported |
440 |
if (0 == fork) { |
441 |
open STDOUT, ">&".fileno($self->{fh}); |
442 |
if (chdir $::DOCROOT) { |
443 |
$ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
444 |
$ENV{HTTP_HOST} = $self->{server_name}; |
445 |
$ENV{HTTP_PORT} = $self->{server_port}; |
446 |
$ENV{SCRIPT_NAME} = $self->{name}; |
447 |
exec $path; |
448 |
} |
449 |
Coro::State::_exit(0); |
450 |
} else { |
451 |
die; |
452 |
} |
453 |
} |
454 |
|
455 |
sub server_hostport { |
456 |
$_[0]{server_port} == 80 |
457 |
? $_[0]{server_name} |
458 |
: "$_[0]{server_name}:$_[0]{server_port}"; |
459 |
} |
460 |
|
461 |
sub respond { |
462 |
my $self = shift; |
463 |
my $path = $self->{path}; |
464 |
|
465 |
if ($self->{name} =~ s%^/internal/([^/]+)%%) { |
466 |
if ($::internal{$1}) { |
467 |
$::internal{$1}->($self); |
468 |
} else { |
469 |
$self->err (404, "not found"); |
470 |
} |
471 |
} else { |
472 |
|
473 |
Coro::AIO::aio_stat $path |
474 |
and $self->err (404, "not found"); |
475 |
|
476 |
$self->{stat} = [stat _]; |
477 |
|
478 |
# idiotic netscape sends idiotic headers AGAIN |
479 |
my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
480 |
? str2time $1 : 0; |
481 |
|
482 |
if (-d _ && -r _) { |
483 |
# directory |
484 |
if ($path !~ /\/$/) { |
485 |
# create a redirect to get the trailing "/" |
486 |
# we don't try to avoid the :80 |
487 |
$self->err (301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
488 |
} else { |
489 |
$ims < $self->{stat}[9] |
490 |
or $self->err (304, "not modified"); |
491 |
|
492 |
if (-r "$path/index.html") { |
493 |
# replace directory "size" by index.html filesize |
494 |
$self->{stat} = [stat ($self->{path} .= "/index.html")]; |
495 |
$self->handle_file ($queue_index, $tbf_top); |
496 |
} else { |
497 |
$self->handle_dir; |
498 |
} |
499 |
} |
500 |
} elsif (-f _ && -r _) { |
501 |
-x _ and $self->err (403, "forbidden"); |
502 |
|
503 |
if (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
504 |
my $timeout = $::NOW + 10; |
505 |
while (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
506 |
if ($timeout < $::NOW) { |
507 |
$self->block($::BLOCKTIME, "too many connections"); |
508 |
} else { |
509 |
$httpevent->wait; |
510 |
} |
511 |
} |
512 |
} |
513 |
|
514 |
$self->handle_file ($queue_file, $tbf_top); |
515 |
} else { |
516 |
$self->err (404, "not found"); |
517 |
} |
518 |
} |
519 |
} |
520 |
|
521 |
sub handle_dir { |
522 |
my $self = shift; |
523 |
my $idx = $self->diridx; |
524 |
|
525 |
$self->response (200, "ok", |
526 |
{ |
527 |
"Content-Type" => "text/html; charset=utf-8", |
528 |
"Content-Length" => length $idx, |
529 |
"Last-Modified" => time2str ($self->{stat}[9]), |
530 |
}, |
531 |
$idx); |
532 |
} |
533 |
|
534 |
sub handle_file { |
535 |
my ($self, $queue, $tbf) = @_; |
536 |
my $length = $self->{stat}[7]; |
537 |
my $hdr = { |
538 |
"Last-Modified" => time2str ((stat _)[9]), |
539 |
"Accept-Ranges" => "bytes", |
540 |
}; |
541 |
|
542 |
my @code = (200, "ok"); |
543 |
my ($l, $h); |
544 |
|
545 |
if ($self->{h}{range} =~ /^bytes=(.*)$/i) { |
546 |
for (split /,/, $1) { |
547 |
if (/^-(\d+)$/) { |
548 |
($l, $h) = ($length - $1, $length - 1); |
549 |
} elsif (/^(\d+)-(\d*)$/) { |
550 |
($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
551 |
} else { |
552 |
($l, $h) = (0, $length - 1); |
553 |
goto ignore; |
554 |
} |
555 |
goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
556 |
} |
557 |
$hdr->{"Content-Range"} = "bytes */$length"; |
558 |
$hdr->{"Content-Length"} = $length; |
559 |
$self->err (416, "not satisfiable", $hdr, ""); |
560 |
|
561 |
satisfiable: |
562 |
# check for segmented downloads |
563 |
if ($l && $::NO_SEGMENTED) { |
564 |
my $timeout = $::NOW + 15; |
565 |
while (keys %{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { |
566 |
if ($timeout <= $::NOW) { |
567 |
$self->block ($::BLOCKTIME, "segmented downloads are forbidden"); |
568 |
#$self->err_segmented_download; |
569 |
} else { |
570 |
$httpevent->wait; |
571 |
} |
572 |
} |
573 |
} |
574 |
|
575 |
$hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
576 |
@code = (206, "partial content"); |
577 |
$length = $h - $l + 1; |
578 |
|
579 |
ignore: |
580 |
} else { |
581 |
($l, $h) = (0, $length - 1); |
582 |
} |
583 |
|
584 |
$self->{path} =~ /\.([^.]+)$/; |
585 |
$hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
586 |
$hdr->{"Content-Length"} = $length; |
587 |
|
588 |
$self->response (@code, $hdr, ""); |
589 |
|
590 |
if ($self->{method} eq "GET") { |
591 |
$self->{time} = $::NOW; |
592 |
$self->{written} = 0; |
593 |
|
594 |
my $fh = Coro::AIO::aio_open $self->{path}, Fcntl::O_RDONLY, 0 |
595 |
or die "$self->{path}: late open failure ($!)"; |
596 |
|
597 |
$h -= $l - 1; |
598 |
|
599 |
my $transfer = $queue->start_transfer ($h); |
600 |
my $locked; |
601 |
my $bufsize = $::WAIT_BUFSIZE; # initial buffer size |
602 |
|
603 |
while ($h > 0) { |
604 |
Coro::cede; |
605 |
my $transfer_lock = $TRANSFER_LOCK->guard ($self->{remote_id}); |
606 |
|
607 |
unless ($locked) { |
608 |
if ($locked ||= $transfer->try ($::WAIT_INTERVAL)) { |
609 |
$bufsize = $::BUFSIZE; |
610 |
$self->{time} = $::NOW; |
611 |
$self->{written} = 0; |
612 |
} |
613 |
} |
614 |
|
615 |
if ($blocked{$self->{remote_id}}) { |
616 |
$self->{h}{connection} = "close"; |
617 |
die bless {}, err::; |
618 |
} |
619 |
|
620 |
Coro::AIO::aio_read $fh, $l, ($h > $bufsize ? $bufsize : $h), my $buf, 0 |
621 |
or last; |
622 |
|
623 |
$tbf->request (length $buf); |
624 |
my $w = $self->{fh}->syswrite ($buf) |
625 |
or last; |
626 |
$::written += $w; |
627 |
$self->{written} += $w; |
628 |
$l += $w; |
629 |
} |
630 |
|
631 |
close $fh; |
632 |
} |
633 |
} |
634 |
|
635 |
1 |
636 |
|