1 |
use Coro; |
2 |
use Coro::Semaphore; |
3 |
use Coro::Event; |
4 |
use Coro::Socket; |
5 |
|
6 |
use HTTP::Date; |
7 |
|
8 |
no utf8; |
9 |
use bytes; |
10 |
|
11 |
# at least on my machine, this thingy serves files |
12 |
# quite a bit faster than apache, ;) |
13 |
# and quite a bit slower than thttpd :( |
14 |
|
15 |
$SIG{PIPE} = 'IGNORE'; |
16 |
|
17 |
our $accesslog; |
18 |
|
19 |
if ($ACCESS_LOG) { |
20 |
use IO::Handle; |
21 |
open $accesslog, ">>$ACCESS_LOG" |
22 |
or die "$ACCESS_LOG: $!"; |
23 |
$accesslog->autoflush(1); |
24 |
} |
25 |
|
26 |
sub slog { |
27 |
my $level = shift; |
28 |
my $format = shift; |
29 |
printf "---: $format\n", @_; |
30 |
} |
31 |
|
32 |
our $connections = new Coro::Semaphore $MAX_CONNECTS || 250; |
33 |
|
34 |
our $wait_factor = 0.95; |
35 |
|
36 |
our @transfers = ( |
37 |
[(new Coro::Semaphore $MAX_TRANSFERS_SMALL || 50), 1], |
38 |
[(new Coro::Semaphore $MAX_TRANSFERS_LARGE || 50), 1], |
39 |
); |
40 |
|
41 |
my @newcons; |
42 |
my @pool; |
43 |
|
44 |
# one "execution thread" |
45 |
sub handler { |
46 |
while () { |
47 |
if (@newcons) { |
48 |
eval { |
49 |
conn->new(@{pop @newcons})->handle; |
50 |
}; |
51 |
slog 1, "$@" if $@ && !ref $@; |
52 |
$connections->up; |
53 |
} else { |
54 |
last if @pool >= $MAX_POOL; |
55 |
push @pool, $Coro::current; |
56 |
schedule; |
57 |
} |
58 |
} |
59 |
} |
60 |
|
61 |
my $http_port = new Coro::Socket |
62 |
LocalAddr => $SERVER_HOST, |
63 |
LocalPort => $SERVER_PORT, |
64 |
ReuseAddr => 1, |
65 |
Listen => 50, |
66 |
or die "unable to start server"; |
67 |
|
68 |
push @listen_sockets, $http_port; |
69 |
|
70 |
our $NOW; |
71 |
our $HTTP_NOW; |
72 |
|
73 |
Event->timer(interval => 1, hard => 1, cb => sub { |
74 |
$NOW = time; |
75 |
$HTTP_NOW = time2str $NOW; |
76 |
})->now; |
77 |
|
78 |
# the "main thread" |
79 |
async { |
80 |
slog 1, "accepting connections"; |
81 |
while () { |
82 |
$connections->down; |
83 |
push @newcons, [$http_port->accept]; |
84 |
#slog 3, "accepted @$connections ".scalar(@pool); |
85 |
if (@pool) { |
86 |
(pop @pool)->ready; |
87 |
} else { |
88 |
async \&handler; |
89 |
} |
90 |
|
91 |
} |
92 |
}; |
93 |
|
94 |
package conn; |
95 |
|
96 |
use Socket; |
97 |
use HTTP::Date; |
98 |
use Convert::Scalar 'weaken'; |
99 |
use Linux::AIO; |
100 |
|
101 |
Linux::AIO::min_parallel $::AIO_PARALLEL; |
102 |
|
103 |
Event->io(fd => Linux::AIO::poll_fileno, |
104 |
poll => 'r', async => 1, |
105 |
cb => \&Linux::AIO::poll_cb); |
106 |
|
107 |
our %conn; # $conn{ip}{self} => connobj |
108 |
our %uri; # $uri{ip}{uri}{self} |
109 |
our %blocked; |
110 |
our %mimetype; |
111 |
|
112 |
sub read_mimetypes { |
113 |
local *M; |
114 |
if (open M, "<mime_types") { |
115 |
while (<M>) { |
116 |
if (/^([^#]\S+)\t+(\S+)$/) { |
117 |
$mimetype{lc $1} = $2; |
118 |
} |
119 |
} |
120 |
} else { |
121 |
print "cannot open mime_types\n"; |
122 |
} |
123 |
} |
124 |
|
125 |
read_mimetypes; |
126 |
|
127 |
sub new { |
128 |
my $class = shift; |
129 |
my $peername = shift; |
130 |
my $fh = shift; |
131 |
my $self = bless { fh => $fh }, $class; |
132 |
my (undef, $iaddr) = unpack_sockaddr_in $peername |
133 |
or $self->err(500, "unable to decode peername"); |
134 |
|
135 |
$self->{remote_addr} = inet_ntoa $iaddr; |
136 |
$self->{time} = $::NOW; |
137 |
|
138 |
$::conns++; |
139 |
|
140 |
$self; |
141 |
} |
142 |
|
143 |
sub DESTROY { |
144 |
my $self = shift; |
145 |
$::conns--; |
146 |
$self->eoconn; |
147 |
} |
148 |
|
149 |
# end of connection |
150 |
sub eoconn { |
151 |
my $self = shift; |
152 |
|
153 |
# clean up hints |
154 |
delete $conn{$self->{remote_id}}{$self*1}; |
155 |
delete $uri{$self->{remote_id}}{$self->{uri}}{$self*1}; |
156 |
} |
157 |
|
158 |
sub slog { |
159 |
my $self = shift; |
160 |
main::slog($_[0], ($self->{remote_id} || $self->{remote_addr}) ."> $_[1]"); |
161 |
} |
162 |
|
163 |
sub response { |
164 |
my ($self, $code, $msg, $hdr, $content) = @_; |
165 |
my $res = "HTTP/1.1 $code $msg\015\012"; |
166 |
|
167 |
$self->{h}{connection} ||= $hdr->{Connection}; |
168 |
|
169 |
$res .= "Date: $HTTP_NOW\015\012"; |
170 |
|
171 |
while (my ($h, $v) = each %$hdr) { |
172 |
$res .= "$h: $v\015\012" |
173 |
} |
174 |
$res .= "\015\012"; |
175 |
|
176 |
$res .= $content if defined $content and $self->{method} ne "HEAD"; |
177 |
|
178 |
my $log = "$self->{remote_addr} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}." \"$self->{h}{referer}\"\n"; |
179 |
|
180 |
print $accesslog $log if $accesslog; |
181 |
print STDERR $log; |
182 |
|
183 |
$self->{written} += |
184 |
print {$self->{fh}} $res; |
185 |
} |
186 |
|
187 |
sub err { |
188 |
my $self = shift; |
189 |
my ($code, $msg, $hdr, $content) = @_; |
190 |
|
191 |
unless (defined $content) { |
192 |
$content = "$code $msg\n"; |
193 |
$hdr->{"Content-Type"} = "text/plain"; |
194 |
$hdr->{"Content-Length"} = length $content; |
195 |
} |
196 |
$hdr->{"Connection"} = "close"; |
197 |
|
198 |
$self->response($code, $msg, $hdr, $content); |
199 |
|
200 |
die bless {}, err::; |
201 |
} |
202 |
|
203 |
sub handle { |
204 |
my $self = shift; |
205 |
my $fh = $self->{fh}; |
206 |
|
207 |
my $host; |
208 |
|
209 |
$fh->timeout($::REQ_TIMEOUT); |
210 |
while() { |
211 |
$self->{reqs}++; |
212 |
|
213 |
# read request and parse first line |
214 |
my $req = $fh->readline("\015\012\015\012"); |
215 |
|
216 |
unless (defined $req) { |
217 |
if (exists $self->{version}) { |
218 |
last; |
219 |
} else { |
220 |
$self->err(408, "request timeout"); |
221 |
} |
222 |
} |
223 |
|
224 |
$self->{h} = {}; |
225 |
|
226 |
$fh->timeout($::RES_TIMEOUT); |
227 |
|
228 |
$req =~ /^(?:\015\012)? |
229 |
(GET|HEAD) \040+ |
230 |
([^\040]+) \040+ |
231 |
HTTP\/([0-9]+\.[0-9]+) |
232 |
\015\012/gx |
233 |
or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
234 |
|
235 |
$self->{method} = $1; |
236 |
$self->{uri} = $2; |
237 |
$self->{version} = $3; |
238 |
|
239 |
$3 =~ /^1\./ |
240 |
or $self->err(506, "http protocol version $3 not supported"); |
241 |
|
242 |
# parse headers |
243 |
{ |
244 |
my (%hdr, $h, $v); |
245 |
|
246 |
$hdr{lc $1} .= ",$2" |
247 |
while $req =~ /\G |
248 |
([^:\000-\040]+): |
249 |
[\008\040]* |
250 |
((?: [^\015\012]+ | \015\012[\008\040] )*) |
251 |
\015\012 |
252 |
/gxc; |
253 |
|
254 |
$req =~ /\G\015\012$/ |
255 |
or $self->err(400, "bad request"); |
256 |
|
257 |
$self->{h}{$h} = substr $v, 1 |
258 |
while ($h, $v) = each %hdr; |
259 |
} |
260 |
|
261 |
# remote id should be unique per user |
262 |
my $id = $self->{remote_addr}; |
263 |
|
264 |
if (exists $self->{h}{"client-ip"}) { |
265 |
$id .= "[".$self->{h}{"client-ip"}."]"; |
266 |
} elsif (exists $self->{h}{"x-forwarded-for"}) { |
267 |
$id .= "[".$self->{h}{"x-forwarded-for"}."]"; |
268 |
} |
269 |
|
270 |
$self->{remote_id} = $id; |
271 |
|
272 |
if ($blocked{$id}) { |
273 |
$self->err_blocked($blocked{$id}) |
274 |
if $blocked{$id} > $::NOW; |
275 |
|
276 |
delete $blocked{$id}; |
277 |
} |
278 |
|
279 |
if (%{$conn{$id}} >= $::MAX_CONN_IP) { |
280 |
my $delay = $::PER_TIMEOUT + 15; |
281 |
while (%{$conn{$id}} >= $::MAX_CONN_IP) { |
282 |
if ($delay <= 0) { |
283 |
$self->slog(2, "blocked ip $id"); |
284 |
$self->err_blocked; |
285 |
} else { |
286 |
Coro::Event::do_timer(after => 4); $delay -= 4; |
287 |
} |
288 |
} |
289 |
} |
290 |
|
291 |
# find out server name and port |
292 |
if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
293 |
$host = $1; |
294 |
} else { |
295 |
$host = $self->{h}{host}; |
296 |
} |
297 |
|
298 |
if (defined $host) { |
299 |
$self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
300 |
} else { |
301 |
($self->{server_port}, $host) |
302 |
= unpack_sockaddr_in $self->{fh}->getsockname |
303 |
or $self->err(500, "unable to get socket name"); |
304 |
$host = inet_ntoa $host; |
305 |
} |
306 |
|
307 |
$self->{server_name} = $host; |
308 |
|
309 |
# enter ourselves into various lists |
310 |
weaken ($conn{$id}{$self*1} = $self); |
311 |
weaken ($uri{$id}{$self->{uri}}{$self*1} = $self); |
312 |
|
313 |
eval { |
314 |
$self->map_uri; |
315 |
$self->respond; |
316 |
}; |
317 |
|
318 |
$self->eoconn; |
319 |
|
320 |
die if $@ && !ref $@; |
321 |
|
322 |
last if $self->{h}{connection} =~ /close/ || $self->{version} < 1.1; |
323 |
|
324 |
$fh->timeout($::PER_TIMEOUT); |
325 |
} |
326 |
} |
327 |
|
328 |
# uri => path mapping |
329 |
sub map_uri { |
330 |
my $self = shift; |
331 |
my $host = $self->{server_name}; |
332 |
my $uri = $self->{uri}; |
333 |
|
334 |
# some massaging, also makes it more secure |
335 |
$uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
336 |
$uri =~ s%//+%/%g; |
337 |
$uri =~ s%/\.(?=/|$)%%g; |
338 |
1 while $uri =~ s%/[^/]+/\.\.(?=/|$)%%; |
339 |
|
340 |
$uri =~ m%^/?\.\.(?=/|$)% |
341 |
and $self->err(400, "bad request"); |
342 |
|
343 |
$self->{name} = $uri; |
344 |
|
345 |
# now do the path mapping |
346 |
$self->{path} = "$::DOCROOT/$host$uri"; |
347 |
|
348 |
$self->access_check; |
349 |
} |
350 |
|
351 |
sub _cgi { |
352 |
my $self = shift; |
353 |
my $path = shift; |
354 |
my $fh; |
355 |
|
356 |
# no two-way xxx supported |
357 |
if (0 == fork) { |
358 |
open STDOUT, ">&".fileno($self->{fh}); |
359 |
if (chdir $::DOCROOT) { |
360 |
$ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
361 |
$ENV{HTTP_HOST} = $self->{server_name}; |
362 |
$ENV{HTTP_PORT} = $self->{server_port}; |
363 |
$ENV{SCRIPT_NAME} = $self->{name}; |
364 |
exec $path; |
365 |
} |
366 |
Coro::State::_exit(0); |
367 |
} else { |
368 |
die; |
369 |
} |
370 |
} |
371 |
|
372 |
sub server_hostport { |
373 |
$_[0]{server_port} == 80 |
374 |
? $_[0]{server_name} |
375 |
: "$_[0]{server_name}:$_[0]{server_port}"; |
376 |
} |
377 |
|
378 |
sub respond { |
379 |
my $self = shift; |
380 |
my $path = $self->{path}; |
381 |
|
382 |
stat $path |
383 |
or $self->err(404, "not found"); |
384 |
|
385 |
$self->{stat} = [stat _]; |
386 |
|
387 |
# idiotic netscape sends idiotic headers AGAIN |
388 |
my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
389 |
? str2time $1 : 0; |
390 |
|
391 |
if (-d _ && -r _) { |
392 |
# directory |
393 |
if ($path !~ /\/$/) { |
394 |
# create a redirect to get the trailing "/" |
395 |
# we don't try to avoid the :80 |
396 |
$self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
397 |
} else { |
398 |
$ims < $self->{stat}[9] |
399 |
or $self->err(304, "not modified"); |
400 |
|
401 |
if (-r "$path/index.html") { |
402 |
$self->{path} .= "/index.html"; |
403 |
$self->handle_file; |
404 |
} else { |
405 |
$self->handle_dir; |
406 |
} |
407 |
} |
408 |
} elsif (-f _ && -r _) { |
409 |
-x _ and $self->err(403, "forbidden"); |
410 |
$self->handle_file; |
411 |
} else { |
412 |
$self->err(404, "not found"); |
413 |
} |
414 |
} |
415 |
|
416 |
sub handle_dir { |
417 |
my $self = shift; |
418 |
my $idx = $self->diridx; |
419 |
|
420 |
$self->response(200, "ok", |
421 |
{ |
422 |
"Content-Type" => "text/html", |
423 |
"Content-Length" => length $idx, |
424 |
}, |
425 |
$idx); |
426 |
} |
427 |
|
428 |
sub handle_file { |
429 |
my $self = shift; |
430 |
my $length = $self->{stat}[7]; |
431 |
my $queue = $::transfers[$length >= $::TRANSFER_SMALL]; |
432 |
my $hdr = { |
433 |
"Last-Modified" => time2str ((stat _)[9]), |
434 |
}; |
435 |
|
436 |
my @code = (200, "ok"); |
437 |
my ($l, $h); |
438 |
|
439 |
if ($self->{h}{range} =~ /^bytes=(.*)$/) { |
440 |
for (split /,/, $1) { |
441 |
if (/^-(\d+)$/) { |
442 |
($l, $h) = ($length - $1, $length - 1); |
443 |
} elsif (/^(\d+)-(\d*)$/) { |
444 |
($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
445 |
} else { |
446 |
($l, $h) = (0, $length - 1); |
447 |
goto ignore; |
448 |
} |
449 |
goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
450 |
} |
451 |
$hdr->{"Content-Range"} = "bytes */$length"; |
452 |
$hdr->{"Content-Length"} = $length; |
453 |
$self->err(416, "not satisfiable", $hdr, ""); |
454 |
|
455 |
satisfiable: |
456 |
# check for segmented downloads |
457 |
if ($l && $::NO_SEGMENTED) { |
458 |
my $delay = $::PER_TIMEOUT + 15; |
459 |
while (%{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { |
460 |
if ($delay <= 0) { |
461 |
$self->err_segmented_download; |
462 |
} else { |
463 |
Coro::Event::do_timer(after => 4); $delay -= 4; |
464 |
} |
465 |
} |
466 |
} |
467 |
|
468 |
$hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
469 |
@code = (206, "partial content"); |
470 |
$length = $h - $l + 1; |
471 |
|
472 |
ignore: |
473 |
} else { |
474 |
($l, $h) = (0, $length - 1); |
475 |
} |
476 |
|
477 |
$self->{path} =~ /\.([^.]+)$/; |
478 |
$hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
479 |
$hdr->{"Content-Length"} = $length; |
480 |
|
481 |
$self->response(@code, $hdr, ""); |
482 |
|
483 |
if ($self->{method} eq "GET") { |
484 |
$self->{time} = $::NOW; |
485 |
|
486 |
my $fudge = $queue->[0]->waiters; |
487 |
$fudge = $fudge ? ($fudge+1)/$fudge : 1; |
488 |
|
489 |
$queue->[1] *= $fudge; |
490 |
my $transfer = $queue->[0]->guard; |
491 |
|
492 |
if ($fudge != 1) { |
493 |
$queue->[1] /= $fudge; |
494 |
$queue->[1] = $queue->[1] * $::wait_factor |
495 |
+ ($::NOW - $self->{time}) * (1 - $::wait_factor); |
496 |
} |
497 |
$self->{time} = $::NOW; |
498 |
|
499 |
$self->{fh}->writable or return; |
500 |
|
501 |
my ($fh, $buf, $r); |
502 |
my $current = $Coro::current; |
503 |
open $fh, "<", $self->{path} |
504 |
or die "$self->{path}: late open failure ($!)"; |
505 |
|
506 |
$h -= $l - 1; |
507 |
|
508 |
if (0) { |
509 |
if ($l) { |
510 |
sysseek $fh, $l, 0; |
511 |
} |
512 |
} |
513 |
|
514 |
while ($h > 0) { |
515 |
if (0) { |
516 |
sysread $fh, $buf, $h > $::BUFSIZE ? $::BUFSIZE : $h |
517 |
or last; |
518 |
} else { |
519 |
aio_read($fh, $l, ($h > $::BUFSIZE ? $::BUFSIZE : $h), |
520 |
$buf, 0, sub { |
521 |
$r = $_[0]; |
522 |
Coro::ready($current); |
523 |
}); |
524 |
&Coro::schedule; |
525 |
last unless $r; |
526 |
} |
527 |
my $w = syswrite $self->{fh}, $buf |
528 |
or last; |
529 |
$::written += $w; |
530 |
$self->{written} += $w; |
531 |
$l += $r; |
532 |
} |
533 |
|
534 |
close $fh; |
535 |
} |
536 |
} |
537 |
|
538 |
1; |