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 $accesslog; |
20 |
|
21 |
if ($ACCESS_LOG) { |
22 |
use IO::Handle; |
23 |
open $accesslog, ">>$ACCESS_LOG" |
24 |
or die "$ACCESS_LOG: $!"; |
25 |
$accesslog->autoflush(1); |
26 |
} |
27 |
|
28 |
sub slog { |
29 |
my $level = shift; |
30 |
my $format = shift; |
31 |
printf "---: $format\n", @_; |
32 |
} |
33 |
|
34 |
our $connections = new Coro::Semaphore $MAX_CONNECTS || 250; |
35 |
our $httpevent = new Coro::Signal; |
36 |
|
37 |
our $wait_factor = 0.95; |
38 |
|
39 |
our @transfers = ( |
40 |
(new Coro::Semaphore $MAX_TRANSFERS_SMALL), |
41 |
(new Coro::Semaphore $MAX_TRANSFERS_LARGE), |
42 |
); |
43 |
|
44 |
my @newcons; |
45 |
my @pool; |
46 |
|
47 |
# one "execution thread" |
48 |
sub handler { |
49 |
while () { |
50 |
if (@newcons) { |
51 |
eval { |
52 |
conn->new(@{pop @newcons})->handle; |
53 |
}; |
54 |
slog 1, "$@" if $@ && !ref $@; |
55 |
$connections->up; |
56 |
} else { |
57 |
last if @pool >= $MAX_POOL; |
58 |
push @pool, $Coro::current; |
59 |
schedule; |
60 |
} |
61 |
} |
62 |
} |
63 |
|
64 |
sub listen_on { |
65 |
my $listen = $_[0]; |
66 |
|
67 |
push @listen_sockets, $listen; |
68 |
|
69 |
# the "main thread" |
70 |
async { |
71 |
slog 1, "accepting connections"; |
72 |
while () { |
73 |
$connections->down; |
74 |
push @newcons, [$listen->accept]; |
75 |
#slog 3, "accepted @$connections ".scalar(@pool); |
76 |
if (@pool) { |
77 |
(pop @pool)->ready; |
78 |
} else { |
79 |
async \&handler; |
80 |
} |
81 |
|
82 |
} |
83 |
}; |
84 |
} |
85 |
|
86 |
my $http_port = new Coro::Socket |
87 |
LocalAddr => $SERVER_HOST, |
88 |
LocalPort => $SERVER_PORT, |
89 |
ReuseAddr => 1, |
90 |
Listen => 50, |
91 |
or die "unable to start server"; |
92 |
|
93 |
listen_on $http_port; |
94 |
|
95 |
if ($SERVER_PORT2) { |
96 |
my $http_port = new Coro::Socket |
97 |
LocalAddr => $SERVER_HOST, |
98 |
LocalPort => $SERVER_PORT2, |
99 |
ReuseAddr => 1, |
100 |
Listen => 50, |
101 |
or die "unable to start server"; |
102 |
|
103 |
listen_on $http_port; |
104 |
} |
105 |
|
106 |
our $NOW; |
107 |
our $HTTP_NOW; |
108 |
|
109 |
Event->timer(interval => 1, hard => 1, cb => sub { |
110 |
$NOW = time; |
111 |
$HTTP_NOW = time2str $NOW; |
112 |
})->now; |
113 |
|
114 |
package conn; |
115 |
|
116 |
use Socket; |
117 |
use HTTP::Date; |
118 |
use Convert::Scalar 'weaken'; |
119 |
use Linux::AIO; |
120 |
|
121 |
Linux::AIO::min_parallel $::AIO_PARALLEL; |
122 |
|
123 |
Event->io(fd => Linux::AIO::poll_fileno, |
124 |
poll => 'r', async => 1, |
125 |
cb => \&Linux::AIO::poll_cb); |
126 |
|
127 |
our %conn; # $conn{ip}{self} => connobj |
128 |
our %uri; # $uri{ip}{uri}{self} |
129 |
our %blocked; |
130 |
our %mimetype; |
131 |
|
132 |
sub read_mimetypes { |
133 |
local *M; |
134 |
if (open M, "<mime_types") { |
135 |
while (<M>) { |
136 |
if (/^([^#]\S+)\t+(\S+)$/) { |
137 |
$mimetype{lc $1} = $2; |
138 |
} |
139 |
} |
140 |
} else { |
141 |
print "cannot open mime_types\n"; |
142 |
} |
143 |
} |
144 |
|
145 |
read_mimetypes; |
146 |
|
147 |
sub new { |
148 |
my $class = shift; |
149 |
my $fh = shift; |
150 |
my $peername = shift; |
151 |
my $self = bless { fh => $fh }, $class; |
152 |
my (undef, $iaddr) = unpack_sockaddr_in $peername |
153 |
or $self->err(500, "unable to decode peername"); |
154 |
|
155 |
$self->{remote_addr} = |
156 |
$self->{remote_id} = inet_ntoa $iaddr; |
157 |
$self->{time} = $::NOW; |
158 |
|
159 |
weaken ($Coro::current->{conn} = $self); |
160 |
|
161 |
$::conns++; |
162 |
|
163 |
$self; |
164 |
} |
165 |
|
166 |
sub DESTROY { |
167 |
my $self = shift; |
168 |
$::conns--; |
169 |
$self->eoconn; |
170 |
} |
171 |
|
172 |
# end of connection |
173 |
sub eoconn { |
174 |
my $self = shift; |
175 |
|
176 |
# clean up hints |
177 |
delete $conn{$self->{remote_id}}{$self*1}; |
178 |
delete $uri{$self->{remote_id}}{$self->{uri}}{$self*1}; |
179 |
|
180 |
$httpevent->broadcast; |
181 |
} |
182 |
|
183 |
sub slog { |
184 |
my $self = shift; |
185 |
main::slog($_[0], "$self->{remote_id}> $_[1]"); |
186 |
} |
187 |
|
188 |
sub response { |
189 |
my ($self, $code, $msg, $hdr, $content) = @_; |
190 |
my $res = "HTTP/1.1 $code $msg\015\012"; |
191 |
|
192 |
if (exists $hdr->{Connection}) { |
193 |
if ($hdr->{Connection} =~ /close/) { |
194 |
$self->{h}{connection} = "close" |
195 |
} |
196 |
} else { |
197 |
if ($self->{version} < 1.1) { |
198 |
if ($self->{h}{connection} =~ /keep-alive/i) { |
199 |
$hdr->{Connection} = "Keep-Alive"; |
200 |
} else { |
201 |
$self->{h}{connection} = "close" |
202 |
} |
203 |
} |
204 |
} |
205 |
|
206 |
$res .= "Date: $HTTP_NOW\015\012"; |
207 |
|
208 |
while (my ($h, $v) = each %$hdr) { |
209 |
$res .= "$h: $v\015\012" |
210 |
} |
211 |
$res .= "\015\012"; |
212 |
|
213 |
$res .= $content if defined $content and $self->{method} ne "HEAD"; |
214 |
|
215 |
my $log = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $NOW). |
216 |
" $self->{remote_id} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}. |
217 |
" \"$self->{h}{referer}\"\n"; |
218 |
|
219 |
print $accesslog $log if $accesslog; |
220 |
print STDERR $log; |
221 |
|
222 |
$self->{written} += |
223 |
print {$self->{fh}} $res; |
224 |
} |
225 |
|
226 |
sub err { |
227 |
my $self = shift; |
228 |
my ($code, $msg, $hdr, $content) = @_; |
229 |
|
230 |
unless (defined $content) { |
231 |
$content = "$code $msg\n"; |
232 |
$hdr->{"Content-Type"} = "text/plain"; |
233 |
$hdr->{"Content-Length"} = length $content; |
234 |
} |
235 |
$hdr->{"Connection"} = "close"; |
236 |
|
237 |
$self->response($code, $msg, $hdr, $content); |
238 |
|
239 |
die bless {}, err::; |
240 |
} |
241 |
|
242 |
sub handle { |
243 |
my $self = shift; |
244 |
my $fh = $self->{fh}; |
245 |
|
246 |
my $host; |
247 |
|
248 |
$fh->timeout($::REQ_TIMEOUT); |
249 |
while() { |
250 |
$self->{reqs}++; |
251 |
|
252 |
# read request and parse first line |
253 |
my $req = $fh->readline("\015\012\015\012"); |
254 |
|
255 |
unless (defined $req) { |
256 |
if (exists $self->{version}) { |
257 |
last; |
258 |
} else { |
259 |
$self->err(408, "request timeout"); |
260 |
} |
261 |
} |
262 |
|
263 |
$self->{h} = {}; |
264 |
|
265 |
$fh->timeout($::RES_TIMEOUT); |
266 |
|
267 |
$req =~ /^(?:\015\012)? |
268 |
(GET|HEAD) \040+ |
269 |
([^\040]+) \040+ |
270 |
HTTP\/([0-9]+\.[0-9]+) |
271 |
\015\012/gx |
272 |
or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
273 |
|
274 |
$self->{method} = $1; |
275 |
$self->{uri} = $2; |
276 |
$self->{version} = $3; |
277 |
|
278 |
$3 =~ /^1\./ |
279 |
or $self->err(506, "http protocol version $3 not supported"); |
280 |
|
281 |
# parse headers |
282 |
{ |
283 |
my (%hdr, $h, $v); |
284 |
|
285 |
$hdr{lc $1} .= ",$2" |
286 |
while $req =~ /\G |
287 |
([^:\000-\040]+): |
288 |
[\008\040]* |
289 |
((?: [^\015\012]+ | \015\012[\008\040] )*) |
290 |
\015\012 |
291 |
/gxc; |
292 |
|
293 |
$req =~ /\G\015\012$/ |
294 |
or $self->err(400, "bad request"); |
295 |
|
296 |
$self->{h}{$h} = substr $v, 1 |
297 |
while ($h, $v) = each %hdr; |
298 |
} |
299 |
|
300 |
# remote id should be unique per user |
301 |
my $id = $self->{remote_addr}; |
302 |
|
303 |
if (exists $self->{h}{"client-ip"}) { |
304 |
$id .= "[".$self->{h}{"client-ip"}."]"; |
305 |
} elsif (exists $self->{h}{"x-forwarded-for"}) { |
306 |
$id .= "[".$self->{h}{"x-forwarded-for"}."]"; |
307 |
} |
308 |
|
309 |
$self->{remote_id} = $id; |
310 |
|
311 |
if ($blocked{$id}) { |
312 |
$self->err_blocked($blocked{$id}) |
313 |
if $blocked{$id} > $::NOW; |
314 |
|
315 |
delete $blocked{$id}; |
316 |
} |
317 |
|
318 |
if (%{$conn{$id}} >= $::MAX_CONN_IP) { |
319 |
my $delay = $::PER_TIMEOUT + $::NOW + 15; |
320 |
while (%{$conn{$id}} >= $::MAX_CONN_IP) { |
321 |
if ($delay < $::NOW) { |
322 |
$self->slog(2, "blocked ip $id"); |
323 |
$self->err_blocked; |
324 |
} else { |
325 |
$httpevent->wait; |
326 |
} |
327 |
} |
328 |
} |
329 |
|
330 |
# find out server name and port |
331 |
if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
332 |
$host = $1; |
333 |
} else { |
334 |
$host = $self->{h}{host}; |
335 |
} |
336 |
|
337 |
if (defined $host) { |
338 |
$self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
339 |
} else { |
340 |
($self->{server_port}, $host) |
341 |
= unpack_sockaddr_in $self->{fh}->sockname |
342 |
or $self->err(500, "unable to get socket name"); |
343 |
$host = inet_ntoa $host; |
344 |
} |
345 |
|
346 |
$self->{server_name} = $host; |
347 |
|
348 |
# enter ourselves into various lists |
349 |
weaken ($conn{$id}{$self*1} = $self); |
350 |
weaken ($uri{$id}{$self->{uri}}{$self*1} = $self); |
351 |
|
352 |
eval { |
353 |
$self->map_uri; |
354 |
$self->respond; |
355 |
}; |
356 |
|
357 |
$self->eoconn; |
358 |
|
359 |
die if $@ && !ref $@; |
360 |
|
361 |
last if $self->{h}{connection} =~ /close/; |
362 |
|
363 |
$httpevent->broadcast; |
364 |
|
365 |
$fh->timeout($::PER_TIMEOUT); |
366 |
} |
367 |
} |
368 |
|
369 |
# uri => path mapping |
370 |
sub map_uri { |
371 |
my $self = shift; |
372 |
my $host = $self->{server_name}; |
373 |
my $uri = $self->{uri}; |
374 |
|
375 |
# some massaging, also makes it more secure |
376 |
$uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
377 |
$uri =~ s%//+%/%g; |
378 |
$uri =~ s%/\.(?=/|$)%%g; |
379 |
1 while $uri =~ s%/[^/]+/\.\.(?=/|$)%%; |
380 |
|
381 |
$uri =~ m%^/?\.\.(?=/|$)% |
382 |
and $self->err(400, "bad request"); |
383 |
|
384 |
$self->{name} = $uri; |
385 |
|
386 |
# now do the path mapping |
387 |
$self->{path} = "$::DOCROOT/$host$uri"; |
388 |
|
389 |
$self->access_check; |
390 |
} |
391 |
|
392 |
sub _cgi { |
393 |
my $self = shift; |
394 |
my $path = shift; |
395 |
my $fh; |
396 |
|
397 |
# no two-way xxx supported |
398 |
if (0 == fork) { |
399 |
open STDOUT, ">&".fileno($self->{fh}); |
400 |
if (chdir $::DOCROOT) { |
401 |
$ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
402 |
$ENV{HTTP_HOST} = $self->{server_name}; |
403 |
$ENV{HTTP_PORT} = $self->{server_port}; |
404 |
$ENV{SCRIPT_NAME} = $self->{name}; |
405 |
exec $path; |
406 |
} |
407 |
Coro::State::_exit(0); |
408 |
} else { |
409 |
die; |
410 |
} |
411 |
} |
412 |
|
413 |
sub server_hostport { |
414 |
$_[0]{server_port} == 80 |
415 |
? $_[0]{server_name} |
416 |
: "$_[0]{server_name}:$_[0]{server_port}"; |
417 |
} |
418 |
|
419 |
sub respond { |
420 |
my $self = shift; |
421 |
my $path = $self->{path}; |
422 |
|
423 |
if ($self->{name} =~ s%^/internal/([^/]+)%%) { |
424 |
if ($::internal{$1}) { |
425 |
$::internal{$1}->($self); |
426 |
} else { |
427 |
$self->err(404, "not found"); |
428 |
} |
429 |
} else { |
430 |
|
431 |
stat $path |
432 |
or $self->err(404, "not found"); |
433 |
|
434 |
$self->{stat} = [stat _]; |
435 |
|
436 |
# idiotic netscape sends idiotic headers AGAIN |
437 |
my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
438 |
? str2time $1 : 0; |
439 |
|
440 |
if (-d _ && -r _) { |
441 |
# directory |
442 |
if ($path !~ /\/$/) { |
443 |
# create a redirect to get the trailing "/" |
444 |
# we don't try to avoid the :80 |
445 |
$self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
446 |
} else { |
447 |
$ims < $self->{stat}[9] |
448 |
or $self->err(304, "not modified"); |
449 |
|
450 |
if (-r "$path/index.html") { |
451 |
# replace directory "size" by index.html filesize |
452 |
$self->{stat}[7] = (stat ($self->{path} .= "/index.html"))[7]; |
453 |
$self->handle_file; |
454 |
} else { |
455 |
$self->handle_dir; |
456 |
} |
457 |
} |
458 |
} elsif (-f _ && -r _) { |
459 |
-x _ and $self->err(403, "forbidden"); |
460 |
$self->handle_file; |
461 |
} else { |
462 |
$self->err(404, "not found"); |
463 |
} |
464 |
} |
465 |
} |
466 |
|
467 |
sub handle_dir { |
468 |
my $self = shift; |
469 |
my $idx = $self->diridx; |
470 |
|
471 |
$self->response(200, "ok", |
472 |
{ |
473 |
"Content-Type" => "text/html", |
474 |
"Content-Length" => length $idx, |
475 |
#d# directories change all the time, so X- |
476 |
"X-Last-Modified" => time2str ((stat _)[9]), |
477 |
}, |
478 |
$idx); |
479 |
} |
480 |
|
481 |
sub handle_file { |
482 |
my $self = shift; |
483 |
my $length = $self->{stat}[7]; |
484 |
my $queue = $::transfers[$length >= $::TRANSFER_SMALL]; |
485 |
my $hdr = { |
486 |
"Last-Modified" => time2str ((stat _)[9]), |
487 |
}; |
488 |
|
489 |
my @code = (200, "ok"); |
490 |
my ($l, $h); |
491 |
|
492 |
if ($self->{h}{range} =~ /^bytes=(.*)$/) { |
493 |
for (split /,/, $1) { |
494 |
if (/^-(\d+)$/) { |
495 |
($l, $h) = ($length - $1, $length - 1); |
496 |
} elsif (/^(\d+)-(\d*)$/) { |
497 |
($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
498 |
} else { |
499 |
($l, $h) = (0, $length - 1); |
500 |
goto ignore; |
501 |
} |
502 |
goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
503 |
} |
504 |
$hdr->{"Content-Range"} = "bytes */$length"; |
505 |
$hdr->{"Content-Length"} = $length; |
506 |
$self->err(416, "not satisfiable", $hdr, ""); |
507 |
|
508 |
satisfiable: |
509 |
# check for segmented downloads |
510 |
if ($l && $::NO_SEGMENTED) { |
511 |
my $delay = $::NOW + $::PER_TIMEOUT + 15; |
512 |
while (%{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { |
513 |
if ($delay <= $::NOW) { |
514 |
$self->err_segmented_download; |
515 |
} else { |
516 |
$httpevent->wait; |
517 |
} |
518 |
} |
519 |
} |
520 |
|
521 |
$hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
522 |
@code = (206, "partial content"); |
523 |
$length = $h - $l + 1; |
524 |
|
525 |
ignore: |
526 |
} else { |
527 |
($l, $h) = (0, $length - 1); |
528 |
} |
529 |
|
530 |
$self->{path} =~ /\.([^.]+)$/; |
531 |
$hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
532 |
$hdr->{"Content-Length"} = $length; |
533 |
|
534 |
$self->response(@code, $hdr, ""); |
535 |
|
536 |
if ($self->{method} eq "GET") { |
537 |
$self->{time} = $::NOW; |
538 |
|
539 |
my $current = $Coro::current; |
540 |
|
541 |
my ($fh, $buf, $r); |
542 |
|
543 |
open $fh, "<", $self->{path} |
544 |
or die "$self->{path}: late open failure ($!)"; |
545 |
|
546 |
$h -= $l - 1; |
547 |
|
548 |
if (0) { # !AIO |
549 |
if ($l) { |
550 |
sysseek $fh, $l, 0; |
551 |
} |
552 |
} |
553 |
|
554 |
my $transfer; # transfer guard |
555 |
my $bufsize = $::WAIT_BUFSIZE; # initial buffer size |
556 |
|
557 |
$self->{time} = $::NOW; |
558 |
|
559 |
while ($h > 0) { |
560 |
unless ($transfer) { |
561 |
if ($transfer ||= $queue->timed_guard($::WAIT_INTERVAL)) { |
562 |
$bufsize = $::BUFSIZE; |
563 |
$self->{time} = $::NOW; |
564 |
} |
565 |
} |
566 |
|
567 |
if (0) { # !AIO |
568 |
sysread $fh, $buf, $h > $bufsize ? $bufsize : $h |
569 |
or last; |
570 |
} else { |
571 |
aio_read($fh, $l, ($h > $bufsize ? $bufsize : $h), |
572 |
$buf, 0, sub { |
573 |
$r = $_[0]; |
574 |
Coro::ready($current); |
575 |
}); |
576 |
&Coro::schedule; |
577 |
last unless $r; |
578 |
} |
579 |
my $w = syswrite $self->{fh}, $buf |
580 |
or last; |
581 |
$::written += $w; |
582 |
$self->{written} += $w; |
583 |
$l += $r; |
584 |
} |
585 |
|
586 |
close $fh; |
587 |
} |
588 |
} |
589 |
|
590 |
1; |