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 |
|
146 |
$::conns--; |
147 |
|
148 |
$self->eoconn; |
149 |
} |
150 |
|
151 |
# end of connection |
152 |
sub eoconn { |
153 |
my $self = shift; |
154 |
|
155 |
# clean up hints |
156 |
delete $conn{$self->{remote_id}}{$self*1}; |
157 |
delete $uri{$self->{remote_id}}{$self->{uri}}{$self*1}; |
158 |
} |
159 |
|
160 |
sub slog { |
161 |
my $self = shift; |
162 |
main::slog($_[0], ($self->{remote_id} || $self->{remote_addr}) ."> $_[1]"); |
163 |
} |
164 |
|
165 |
sub response { |
166 |
my ($self, $code, $msg, $hdr, $content) = @_; |
167 |
my $res = "HTTP/1.1 $code $msg\015\012"; |
168 |
|
169 |
$self->{h}{connection} ||= $hdr->{Connection}; |
170 |
|
171 |
$res .= "Date: $HTTP_NOW\015\012"; |
172 |
|
173 |
while (my ($h, $v) = each %$hdr) { |
174 |
$res .= "$h: $v\015\012" |
175 |
} |
176 |
$res .= "\015\012"; |
177 |
|
178 |
$res .= $content if defined $content and $self->{method} ne "HEAD"; |
179 |
|
180 |
my $log = "$self->{remote_addr} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}." \"$self->{h}{referer}\"\n"; |
181 |
|
182 |
print $accesslog $log if $accesslog; |
183 |
print STDERR $log; |
184 |
|
185 |
$self->{written} += |
186 |
print {$self->{fh}} $res; |
187 |
} |
188 |
|
189 |
sub err { |
190 |
my $self = shift; |
191 |
my ($code, $msg, $hdr, $content) = @_; |
192 |
|
193 |
unless (defined $content) { |
194 |
$content = "$code $msg\n"; |
195 |
$hdr->{"Content-Type"} = "text/plain"; |
196 |
$hdr->{"Content-Length"} = length $content; |
197 |
} |
198 |
$hdr->{"Connection"} = "close"; |
199 |
|
200 |
$self->response($code, $msg, $hdr, $content); |
201 |
|
202 |
die bless {}, err::; |
203 |
} |
204 |
|
205 |
sub handle { |
206 |
my $self = shift; |
207 |
my $fh = $self->{fh}; |
208 |
|
209 |
my $host; |
210 |
|
211 |
$fh->timeout($::REQ_TIMEOUT); |
212 |
while() { |
213 |
$self->{reqs}++; |
214 |
|
215 |
# read request and parse first line |
216 |
my $req = $fh->readline("\015\012\015\012"); |
217 |
|
218 |
unless (defined $req) { |
219 |
if (exists $self->{version}) { |
220 |
last; |
221 |
} else { |
222 |
$self->err(408, "request timeout"); |
223 |
} |
224 |
} |
225 |
|
226 |
$self->{h} = {}; |
227 |
|
228 |
$fh->timeout($::RES_TIMEOUT); |
229 |
|
230 |
$req =~ /^(?:\015\012)? |
231 |
(GET|HEAD) \040+ |
232 |
([^\040]+) \040+ |
233 |
HTTP\/([0-9]+\.[0-9]+) |
234 |
\015\012/gx |
235 |
or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
236 |
|
237 |
$self->{method} = $1; |
238 |
$self->{uri} = $2; |
239 |
$self->{version} = $3; |
240 |
|
241 |
$3 =~ /^1\./ |
242 |
or $self->err(506, "http protocol version $3 not supported"); |
243 |
|
244 |
# parse headers |
245 |
{ |
246 |
my (%hdr, $h, $v); |
247 |
|
248 |
$hdr{lc $1} .= ",$2" |
249 |
while $req =~ /\G |
250 |
([^:\000-\040]+): |
251 |
[\008\040]* |
252 |
((?: [^\015\012]+ | \015\012[\008\040] )*) |
253 |
\015\012 |
254 |
/gxc; |
255 |
|
256 |
$req =~ /\G\015\012$/ |
257 |
or $self->err(400, "bad request"); |
258 |
|
259 |
$self->{h}{$h} = substr $v, 1 |
260 |
while ($h, $v) = each %hdr; |
261 |
} |
262 |
|
263 |
# remote id should be unique per user |
264 |
my $id = $self->{remote_addr}; |
265 |
|
266 |
if (exists $self->{h}{"client-ip"}) { |
267 |
$id .= "[".$self->{h}{"client-ip"}."]"; |
268 |
} elsif (exists $self->{h}{"x-forwarded-for"}) { |
269 |
$id .= "[".$self->{h}{"x-forwarded-for"}."]"; |
270 |
} |
271 |
|
272 |
$self->{remote_id} = $id; |
273 |
|
274 |
if ($blocked{$id}) { |
275 |
$self->err_blocked($blocked{$id}) |
276 |
if $blocked{$id} > $::NOW; |
277 |
|
278 |
delete $blocked{$id}; |
279 |
} |
280 |
|
281 |
if (%{$conn{$id}} >= $::MAX_CONN_IP) { |
282 |
my $delay = $::PER_TIMEOUT + 15; |
283 |
while (%{$conn{$id}} >= $::MAX_CONN_IP) { |
284 |
if ($delay <= 0) { |
285 |
$self->slog(2, "blocked ip $id"); |
286 |
$self->err_blocked; |
287 |
} else { |
288 |
Coro::Event::do_timer(after => 4); $delay -= 4; |
289 |
} |
290 |
} |
291 |
} |
292 |
|
293 |
# find out server name and port |
294 |
if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
295 |
$host = $1; |
296 |
} else { |
297 |
$host = $self->{h}{host}; |
298 |
} |
299 |
|
300 |
if (defined $host) { |
301 |
$self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
302 |
} else { |
303 |
($self->{server_port}, $host) |
304 |
= unpack_sockaddr_in $self->{fh}->getsockname |
305 |
or $self->err(500, "unable to get socket name"); |
306 |
$host = inet_ntoa $host; |
307 |
} |
308 |
|
309 |
$self->{server_name} = $host; |
310 |
|
311 |
# enter ourselves into various lists |
312 |
weaken ($conn{$id}{$self*1} = $self); |
313 |
weaken ($uri{$id}{$self->{uri}}{$self*1} = $self); |
314 |
|
315 |
eval { |
316 |
$self->map_uri; |
317 |
$self->respond; |
318 |
}; |
319 |
|
320 |
$self->eoconn; |
321 |
|
322 |
die if $@ && !ref $@; |
323 |
|
324 |
last if $self->{h}{connection} =~ /close/ || $self->{version} < 1.1; |
325 |
|
326 |
$fh->timeout($::PER_TIMEOUT); |
327 |
} |
328 |
} |
329 |
|
330 |
# uri => path mapping |
331 |
sub map_uri { |
332 |
my $self = shift; |
333 |
my $host = $self->{server_name}; |
334 |
my $uri = $self->{uri}; |
335 |
|
336 |
# some massaging, also makes it more secure |
337 |
$uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
338 |
$uri =~ s%//+%/%g; |
339 |
$uri =~ s%/\.(?=/|$)%%g; |
340 |
1 while $uri =~ s%/[^/]+/\.\.(?=/|$)%%; |
341 |
|
342 |
$uri =~ m%^/?\.\.(?=/|$)% |
343 |
and $self->err(400, "bad request"); |
344 |
|
345 |
$self->{name} = $uri; |
346 |
|
347 |
# now do the path mapping |
348 |
$self->{path} = "$::DOCROOT/$host$uri"; |
349 |
|
350 |
$self->access_check; |
351 |
} |
352 |
|
353 |
sub _cgi { |
354 |
my $self = shift; |
355 |
my $path = shift; |
356 |
my $fh; |
357 |
|
358 |
# no two-way xxx supported |
359 |
if (0 == fork) { |
360 |
open STDOUT, ">&".fileno($self->{fh}); |
361 |
if (chdir $::DOCROOT) { |
362 |
$ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
363 |
$ENV{HTTP_HOST} = $self->{server_name}; |
364 |
$ENV{HTTP_PORT} = $self->{server_port}; |
365 |
$ENV{SCRIPT_NAME} = $self->{name}; |
366 |
exec $path; |
367 |
} |
368 |
Coro::State::_exit(0); |
369 |
} else { |
370 |
die; |
371 |
} |
372 |
} |
373 |
|
374 |
sub server_hostport { |
375 |
$_[0]{server_port} == 80 |
376 |
? $_[0]{server_name} |
377 |
: "$_[0]{server_name}:$_[0]{server_port}"; |
378 |
} |
379 |
|
380 |
sub respond { |
381 |
my $self = shift; |
382 |
my $path = $self->{path}; |
383 |
|
384 |
stat $path |
385 |
or $self->err(404, "not found"); |
386 |
|
387 |
$self->{stat} = [stat _]; |
388 |
|
389 |
# idiotic netscape sends idiotic headers AGAIN |
390 |
my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
391 |
? str2time $1 : 0; |
392 |
|
393 |
if (-d _ && -r _) { |
394 |
# directory |
395 |
if ($path !~ /\/$/) { |
396 |
# create a redirect to get the trailing "/" |
397 |
# we don't try to avoid the :80 |
398 |
$self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
399 |
} else { |
400 |
$ims < $self->{stat}[9] |
401 |
or $self->err(304, "not modified"); |
402 |
|
403 |
if (-r "$path/index.html") { |
404 |
$self->{path} .= "/index.html"; |
405 |
$self->handle_file; |
406 |
} else { |
407 |
$self->handle_dir; |
408 |
} |
409 |
} |
410 |
} elsif (-f _ && -r _) { |
411 |
-x _ and $self->err(403, "forbidden"); |
412 |
$self->handle_file; |
413 |
} else { |
414 |
$self->err(404, "not found"); |
415 |
} |
416 |
} |
417 |
|
418 |
sub handle_dir { |
419 |
my $self = shift; |
420 |
my $idx = $self->diridx; |
421 |
|
422 |
$self->response(200, "ok", |
423 |
{ |
424 |
"Content-Type" => "text/html", |
425 |
"Content-Length" => length $idx, |
426 |
}, |
427 |
$idx); |
428 |
} |
429 |
|
430 |
sub handle_file { |
431 |
my $self = shift; |
432 |
my $length = $self->{stat}[7]; |
433 |
my $queue = $::transfers[$length >= $::TRANSFER_SMALL]; |
434 |
my $hdr = { |
435 |
"Last-Modified" => time2str ((stat _)[9]), |
436 |
}; |
437 |
|
438 |
my @code = (200, "ok"); |
439 |
my ($l, $h); |
440 |
|
441 |
if ($self->{h}{range} =~ /^bytes=(.*)$/) { |
442 |
for (split /,/, $1) { |
443 |
if (/^-(\d+)$/) { |
444 |
($l, $h) = ($length - $1, $length - 1); |
445 |
} elsif (/^(\d+)-(\d*)$/) { |
446 |
($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
447 |
} else { |
448 |
($l, $h) = (0, $length - 1); |
449 |
goto ignore; |
450 |
} |
451 |
goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
452 |
} |
453 |
$hdr->{"Content-Range"} = "bytes */$length"; |
454 |
$hdr->{"Content-Length"} = $length; |
455 |
$self->err(416, "not satisfiable", $hdr, ""); |
456 |
|
457 |
satisfiable: |
458 |
# check for segmented downloads |
459 |
if ($l && $::NO_SEGMENTED) { |
460 |
my $delay = $::PER_TIMEOUT + 15; |
461 |
while (%{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { |
462 |
if ($delay <= 0) { |
463 |
$self->err_segmented_download; |
464 |
} else { |
465 |
Coro::Event::do_timer(after => 4); $delay -= 4; |
466 |
} |
467 |
} |
468 |
} |
469 |
|
470 |
$hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
471 |
@code = (206, "partial content"); |
472 |
$length = $h - $l + 1; |
473 |
|
474 |
ignore: |
475 |
} else { |
476 |
($l, $h) = (0, $length - 1); |
477 |
} |
478 |
|
479 |
$self->{path} =~ /\.([^.]+)$/; |
480 |
$hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
481 |
$hdr->{"Content-Length"} = $length; |
482 |
|
483 |
$self->response(@code, $hdr, ""); |
484 |
|
485 |
if ($self->{method} eq "GET") { |
486 |
$self->{time} = $::NOW; |
487 |
|
488 |
my $fudge = $queue->[0]->waiters; |
489 |
$fudge = $fudge ? ($fudge+1)/$fudge : 1; |
490 |
|
491 |
$queue->[1] *= $fudge; |
492 |
my $transfer = $queue->[0]->guard; |
493 |
|
494 |
if ($fudge != 1) { |
495 |
$queue->[1] /= $fudge; |
496 |
$queue->[1] = $queue->[1] * $::wait_factor |
497 |
+ ($::NOW - $self->{time}) * (1 - $::wait_factor); |
498 |
} |
499 |
$self->{time} = $::NOW; |
500 |
|
501 |
$self->{fh}->writable or return; |
502 |
|
503 |
my ($fh, $buf, $r); |
504 |
my $current = $Coro::current; |
505 |
open $fh, "<", $self->{path} |
506 |
or die "$self->{path}: late open failure ($!)"; |
507 |
|
508 |
$h -= $l - 1; |
509 |
|
510 |
if (0) { |
511 |
if ($l) { |
512 |
sysseek $fh, $l, 0; |
513 |
} |
514 |
} |
515 |
|
516 |
while ($h > 0) { |
517 |
if (0) { |
518 |
sysread $fh, $buf, $h > $::BUFSIZE ? $::BUFSIZE : $h |
519 |
or last; |
520 |
} else { |
521 |
aio_read($fh, $l, ($h > $::BUFSIZE ? $::BUFSIZE : $h), |
522 |
$buf, 0, sub { |
523 |
$r = $_[0]; |
524 |
Coro::ready($current); |
525 |
}); |
526 |
&Coro::schedule; |
527 |
last unless $r; |
528 |
} |
529 |
my $w = syswrite $self->{fh}, $buf |
530 |
or last; |
531 |
$::written += $w; |
532 |
$self->{written} += $w; |
533 |
$l += $r; |
534 |
} |
535 |
|
536 |
close $fh; |
537 |
} |
538 |
} |
539 |
|
540 |
1; |