ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/cvsroot/Coro/myhttpd/httpd.pl
(Generate patch)

Comparing cvsroot/Coro/myhttpd/httpd.pl (file contents):
Revision 1.45 by root, Sun Nov 11 03:32:19 2001 UTC vs.
Revision 1.79 by root, Fri Dec 1 03:53:33 2006 UTC

1use Coro; 1use Coro;
2use Coro::Semaphore; 2use Coro::Semaphore;
3use Coro::Event; 3use Coro::Event;
4use Coro::Socket; 4use Coro::Socket;
5use Coro::Signal; 5use Coro::Signal;
6use Coro::AIO ();
6 7
7use HTTP::Date; 8use HTTP::Date;
9use POSIX ();
10
11use Compress::Zlib ();
8 12
9no utf8; 13no utf8;
10use bytes; 14use bytes;
11 15
12# at least on my machine, this thingy serves files 16# at least on my machine, this thingy serves files
14# and quite a bit slower than thttpd :( 18# and quite a bit slower than thttpd :(
15 19
16$SIG{PIPE} = 'IGNORE'; 20$SIG{PIPE} = 'IGNORE';
17 21
18our $accesslog; 22our $accesslog;
23our $errorlog;
24
25our $NOW;
26our $HTTP_NOW;
27
28Event->timer(interval => 1, hard => 1, cb => sub {
29 $NOW = time;
30 $HTTP_NOW = time2str $NOW;
31})->now;
32
33if ($ERROR_LOG) {
34 use IO::Handle;
35 open $errorlog, ">>$ERROR_LOG"
36 or die "$ERROR_LOG: $!";
37 $errorlog->autoflush(1);
38}
19 39
20if ($ACCESS_LOG) { 40if ($ACCESS_LOG) {
21 use IO::Handle; 41 use IO::Handle;
22 open $accesslog, ">>$ACCESS_LOG" 42 open $accesslog, ">>$ACCESS_LOG"
23 or die "$ACCESS_LOG: $!"; 43 or die "$ACCESS_LOG: $!";
25} 45}
26 46
27sub slog { 47sub slog {
28 my $level = shift; 48 my $level = shift;
29 my $format = shift; 49 my $format = shift;
50 my $NOW = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW);
30 printf "---: $format\n", @_; 51 printf "$NOW: $format\n", @_;
52 printf $errorlog "$NOW: $format\n", @_ if $errorlog;
31} 53}
32 54
33our $connections = new Coro::Semaphore $MAX_CONNECTS || 250; 55our $connections = new Coro::Semaphore $MAX_CONNECTS || 250;
34our $httpevent = new Coro::Signal; 56our $httpevent = new Coro::Signal;
35 57
36our $wait_factor = 0.95; 58our $queue_file = new transferqueue $MAX_TRANSFERS;
59our $queue_index = new transferqueue 10;
37 60
38our @transfers = ( 61our $tbf_top = new tbf rate => $TBF_RATE || 100000;
39 [(new Coro::Semaphore $MAX_TRANSFERS_SMALL || 50), 1], 62
40 [(new Coro::Semaphore $MAX_TRANSFERS_LARGE || 50), 1], 63my $unused_bytes = 0;
41); 64my $unused_last = time;
65
66sub unused_bandwidth {
67 $unused_bytes += $_[0];
68 if ($unused_last < $NOW - 30 && $unused_bytes / ($NOW - $unused_last) > 50000) {
69 $unused_last = $NOW;
70 $unused_bytes = 0;
71 $queue_file->force_wake_next;
72 slog 1, "forced filetransfer due to unused bandwidth";
73 }
74}
42 75
43my @newcons; 76my @newcons;
44my @pool; 77my @pool;
45 78
46# one "execution thread" 79# one "execution thread"
49 if (@newcons) { 82 if (@newcons) {
50 eval { 83 eval {
51 conn->new(@{pop @newcons})->handle; 84 conn->new(@{pop @newcons})->handle;
52 }; 85 };
53 slog 1, "$@" if $@ && !ref $@; 86 slog 1, "$@" if $@ && !ref $@;
87
88 $httpevent->broadcast; # only for testing, but doesn't matter much
89
54 $connections->up; 90 $connections->up;
55 } else { 91 } else {
56 last if @pool >= $MAX_POOL; 92 last if @pool >= $MAX_POOL;
57 push @pool, $Coro::current; 93 push @pool, $Coro::current;
58 schedule; 94 schedule;
75 if (@pool) { 111 if (@pool) {
76 (pop @pool)->ready; 112 (pop @pool)->ready;
77 } else { 113 } else {
78 async \&handler; 114 async \&handler;
79 } 115 }
80
81 } 116 }
82 }; 117 };
83} 118}
84 119
85my $http_port = new Coro::Socket 120my $http_port = new Coro::Socket
100 or die "unable to start server"; 135 or die "unable to start server";
101 136
102 listen_on $http_port; 137 listen_on $http_port;
103} 138}
104 139
105our $NOW;
106our $HTTP_NOW;
107
108Event->timer(interval => 1, hard => 1, cb => sub {
109 $NOW = time;
110 $HTTP_NOW = time2str $NOW;
111})->now;
112
113package conn; 140package conn;
114 141
115use Socket; 142use Socket;
116use HTTP::Date; 143use HTTP::Date;
117use Convert::Scalar 'weaken'; 144use Convert::Scalar 'weaken';
118use Linux::AIO; 145use IO::AIO;
119 146
120Linux::AIO::min_parallel $::AIO_PARALLEL; 147IO::AIO::min_parallel $::AIO_PARALLEL;
121 148
122Event->io(fd => Linux::AIO::poll_fileno, 149Event->io (fd => IO::AIO::poll_fileno,
123 poll => 'r', async => 1, 150 poll => 'r', async => 1,
124 cb => \&Linux::AIO::poll_cb); 151 cb => \&IO::AIO::poll_cb);
125 152
126our %conn; # $conn{ip}{self} => connobj 153our %conn; # $conn{ip}{self} => connobj
127our %uri; # $uri{ip}{uri}{self} 154our %uri; # $uri{ip}{uri}{self}
128our %blocked; 155our %blocked;
129our %mimetype; 156our %mimetype;
149 my $peername = shift; 176 my $peername = shift;
150 my $self = bless { fh => $fh }, $class; 177 my $self = bless { fh => $fh }, $class;
151 my (undef, $iaddr) = unpack_sockaddr_in $peername 178 my (undef, $iaddr) = unpack_sockaddr_in $peername
152 or $self->err(500, "unable to decode peername"); 179 or $self->err(500, "unable to decode peername");
153 180
181 $self->{remote_addr} =
154 $self->{remote_addr} = inet_ntoa $iaddr; 182 $self->{remote_id} = inet_ntoa $iaddr;
183
155 $self->{time} = $::NOW; 184 $self->{time} = $::NOW;
156 185
186 weaken ($Coro::current->{conn} = $self);
187
157 $::conns++; 188 $::conns++;
189 $::maxconns = $::conns if $::conns > $::maxconns;
158 190
159 $self; 191 $self;
160} 192}
161 193
162sub DESTROY { 194sub DESTROY {
163 my $self = shift; 195 #my $self = shift;
164 $::conns--; 196 $::conns--;
165 $self->eoconn;
166} 197}
167 198
168# end of connection 199sub prune_cache {
169sub eoconn { 200 my $hash = $_[0];
170 my $self = shift;
171 201
172 # clean up hints 202 for (keys %$hash) {
173 delete $conn{$self->{remote_id}}{$self*1}; 203 if (ref $hash->{$_} eq HASH::) {
174 delete $uri{$self->{remote_id}}{$self->{uri}}{$self*1}; 204 prune_cache($hash->{$_});
175 205 unless (scalar keys %{$hash->{$_}}) {
176 $httpevent->broadcast; 206 delete $hash->{$_};
207 $d2++;
208 }
209 }
210 }
177} 211}
212
213sub prune_caches {
214 prune_cache \%conn;
215 prune_cache \%uri;
216
217 for (keys %blocked) {
218 delete $blocked{$_} unless $blocked{$_}[0] > $::NOW;
219 }
220}
221
222Event->timer(interval => 60, cb => \&prune_caches);
178 223
179sub slog { 224sub slog {
180 my $self = shift; 225 my $self = shift;
181 main::slog($_[0], ($self->{remote_id} || $self->{remote_addr}) ."> $_[1]"); 226 main::slog($_[0], "$self->{remote_id}> $_[1]");
182} 227}
183 228
184sub response { 229sub response {
185 my ($self, $code, $msg, $hdr, $content) = @_; 230 my ($self, $code, $msg, $hdr, $content) = @_;
186 my $res = "HTTP/1.1 $code $msg\015\012"; 231 my $res = "HTTP/1.1 $code $msg\015\012";
232 my $GZ = "";
187 233
234 if (exists $hdr->{Connection}) {
235 if ($hdr->{Connection} =~ /close/) {
188 $self->{h}{connection} = "close" 236 $self->{h}{connection} = "close"
189 if exists $hdr->{Connection} # to avoid "empty" header lines due to vivification 237 }
190 and $hdr->{Connection} =~ /close/; 238 } else {
239 if ($self->{version} < 1.1) {
240 if ($self->{h}{connection} =~ /keep-alive/i) {
241 $hdr->{Connection} = "Keep-Alive";
242 } else {
243 $self->{h}{connection} = "close"
244 }
245 }
246 }
247
248 if ($self->{method} ne "HEAD"
249 && $self->{h}{"accept-encoding"} =~ /\bgzip\b/
250 && 400 < length $content
251 && $hdr->{"Content-Length"} == length $content
252 && !exists $hdr->{"Content-Encoding"}
253 ) {
254 my $orig = length $content;
255 $hdr->{"Content-Encoding"} = "gzip";
256 $content = Compress::Zlib::memGzip(\$content);
257 $hdr->{"Content-Length"} = length $content;
258 $GZ = sprintf "GZ%02d", 100 - 100*((length $content) / $orig);
259 }
191 260
192 $res .= "Date: $HTTP_NOW\015\012"; 261 $res .= "Date: $HTTP_NOW\015\012";
262 $res .= "Server: $::NAME\015\012";
193 263
194 while (my ($h, $v) = each %$hdr) { 264 while (my ($h, $v) = each %$hdr) {
195 $res .= "$h: $v\015\012" 265 $res .= "$h: $v\015\012"
196 } 266 }
197 $res .= "\015\012"; 267 $res .= "\015\012";
198 268
199 $res .= $content if defined $content and $self->{method} ne "HEAD"; 269 $res .= $content if defined $content and $self->{method} ne "HEAD";
200 270
201 my $log = "$self->{remote_addr} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}." \"$self->{h}{referer}\"\n"; 271 my $log = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW).
272 " $self->{remote_id} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}.$GZ.
273 " \"$self->{h}{referer}\"\n";
202 274
203 print $accesslog $log if $accesslog; 275 print $::accesslog $log if $::accesslog;
204 print STDERR $log; 276 print STDERR $log;
205 277
206 $self->{written} += 278 $tbf_top->request(length $res, 1e6);
207 print {$self->{fh}} $res; 279 $self->{written} += print {$self->{fh}} $res;
208} 280}
209 281
210sub err { 282sub err {
211 my $self = shift; 283 my $self = shift;
212 my ($code, $msg, $hdr, $content) = @_; 284 my ($code, $msg, $hdr, $content) = @_;
267 my (%hdr, $h, $v); 339 my (%hdr, $h, $v);
268 340
269 $hdr{lc $1} .= ",$2" 341 $hdr{lc $1} .= ",$2"
270 while $req =~ /\G 342 while $req =~ /\G
271 ([^:\000-\040]+): 343 ([^:\000-\040]+):
272 [\008\040]* 344 [\011\040]*
273 ((?: [^\015\012]+ | \015\012[\008\040] )*) 345 ((?: [^\015\012]+ | \015\012[\011\040] )*)
274 \015\012 346 \015\012
275 /gxc; 347 /gxc;
276 348
277 $req =~ /\G\015\012$/ 349 $req =~ /\G\015\012$/
278 or $self->err(400, "bad request"); 350 or $self->err(400, "bad request");
290 $id .= "[".$self->{h}{"x-forwarded-for"}."]"; 362 $id .= "[".$self->{h}{"x-forwarded-for"}."]";
291 } 363 }
292 364
293 $self->{remote_id} = $id; 365 $self->{remote_id} = $id;
294 366
367 weaken (local $conn{$id}{$self*1} = $self);
368
295 if ($blocked{$id}) { 369 if ($blocked{$id}) {
296 $self->err_blocked($blocked{$id}) 370 $self->err_blocked
297 if $blocked{$id} > $::NOW; 371 if $blocked{$id}[0] > $::NOW;
298 372
299 delete $blocked{$id}; 373 delete $blocked{$id};
300 }
301
302 if (%{$conn{$id}} >= $::MAX_CONN_IP) {
303 my $delay = $::PER_TIMEOUT + $::NOW + 15;
304 while (%{$conn{$id}} >= $::MAX_CONN_IP) {
305 if ($delay < $::NOW) {
306 $self->slog(2, "blocked ip $id");
307 $self->err_blocked;
308 } else {
309 $httpevent->wait;
310 }
311 }
312 } 374 }
313 375
314 # find out server name and port 376 # find out server name and port
315 if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { 377 if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) {
316 $host = $1; 378 $host = $1;
327 $host = inet_ntoa $host; 389 $host = inet_ntoa $host;
328 } 390 }
329 391
330 $self->{server_name} = $host; 392 $self->{server_name} = $host;
331 393
332 # enter ourselves into various lists
333 weaken ($conn{$id}{$self*1} = $self);
334 weaken ($uri{$id}{$self->{uri}}{$self*1} = $self); 394 weaken (local $uri{$id}{$self->{uri}}{$self*1} = $self);
335 395
336 eval { 396 eval {
337 $self->map_uri; 397 $self->map_uri;
338 $self->respond; 398 $self->respond;
339 }; 399 };
340 400
341 $self->eoconn;
342
343 die if $@ && !ref $@; 401 die if $@ && !ref $@;
344 402
345 last if $self->{h}{connection} =~ /close/ || $self->{version} < 1.1; 403 last if $self->{h}{connection} =~ /close/i;
346 404
347 $httpevent->broadcast; 405 $httpevent->broadcast;
348 406
349 $fh->timeout($::PER_TIMEOUT); 407 $fh->timeout($::PER_TIMEOUT);
350 } 408 }
409}
410
411sub block {
412 my $self = shift;
413
414 $blocked{$self->{remote_id}} = [$::NOW + $_[0], $_[1]];
415 $self->slog(2, "blocked ip $self->{remote_id}");
416 $self->err_blocked;
351} 417}
352 418
353# uri => path mapping 419# uri => path mapping
354sub map_uri { 420sub map_uri {
355 my $self = shift; 421 my $self = shift;
402 468
403sub respond { 469sub respond {
404 my $self = shift; 470 my $self = shift;
405 my $path = $self->{path}; 471 my $path = $self->{path};
406 472
407 stat $path 473 if ($self->{name} =~ s%^/internal/([^/]+)%%) {
408 or $self->err(404, "not found"); 474 if ($::internal{$1}) {
409 475 $::internal{$1}->($self);
410 $self->{stat} = [stat _];
411
412 # idiotic netscape sends idiotic headers AGAIN
413 my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/
414 ? str2time $1 : 0;
415
416 if (-d _ && -r _) {
417 # directory
418 if ($path !~ /\/$/) {
419 # create a redirect to get the trailing "/"
420 # we don't try to avoid the :80
421 $self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" });
422 } else { 476 } else {
423 $ims < $self->{stat}[9] 477 $self->err(404, "not found");
478 }
479 } else {
480
481 stat $path
424 or $self->err(304, "not modified"); 482 or $self->err(404, "not found");
425 483
426 if (-r "$path/index.html") { 484 $self->{stat} = [stat _];
427 # replace directory "size" by index.html filesize 485
428 $self->{stat}[7] = (stat ($self->{path} .= "/index.html"))[7]; 486 # idiotic netscape sends idiotic headers AGAIN
429 $self->handle_file; 487 my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/
488 ? str2time $1 : 0;
489
490 if (-d _ && -r _) {
491 # directory
492 if ($path !~ /\/$/) {
493 # create a redirect to get the trailing "/"
494 # we don't try to avoid the :80
495 $self->err(301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" });
430 } else { 496 } else {
497 $ims < $self->{stat}[9]
498 or $self->err(304, "not modified");
499
500 if (-r "$path/index.html") {
501 # replace directory "size" by index.html filesize
502 $self->{stat} = [stat ($self->{path} .= "/index.html")];
503 $self->handle_file($queue_index, $tbf_top);
504 } else {
431 $self->handle_dir; 505 $self->handle_dir;
432 } 506 }
433 } 507 }
434 } elsif (-f _ && -r _) { 508 } elsif (-f _ && -r _) {
435 -x _ and $self->err(403, "forbidden"); 509 -x _ and $self->err(403, "forbidden");
436 $self->handle_file; 510
511 if (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) {
512 my $timeout = $::NOW + 10;
513 while (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) {
514 if ($timeout < $::NOW) {
515 $self->block($::BLOCKTIME, "too many connections");
516 } else {
517 $httpevent->wait;
518 }
519 }
520 }
521
522 $self->handle_file($queue_file, $tbf_top);
437 } else { 523 } else {
438 $self->err(404, "not found"); 524 $self->err(404, "not found");
525 }
439 } 526 }
440} 527}
441 528
442sub handle_dir { 529sub handle_dir {
443 my $self = shift; 530 my $self = shift;
444 my $idx = $self->diridx; 531 my $idx = $self->diridx;
445 532
446 $self->response(200, "ok", 533 $self->response(200, "ok",
447 { 534 {
448 "Content-Type" => "text/html", 535 "Content-Type" => "text/html; charset=utf-8",
449 "Content-Length" => length $idx, 536 "Content-Length" => length $idx,
537 "Last-Modified" => time2str ($self->{stat}[9]),
450 }, 538 },
451 $idx); 539 $idx);
452} 540}
453 541
454sub handle_file { 542sub handle_file {
455 my $self = shift; 543 my ($self, $queue, $tbf) = @_;
456 my $length = $self->{stat}[7]; 544 my $length = $self->{stat}[7];
457 my $queue = $::transfers[$length >= $::TRANSFER_SMALL];
458 my $hdr = { 545 my $hdr = {
459 "Last-Modified" => time2str ((stat _)[9]), 546 "Last-Modified" => time2str ((stat _)[9]),
547 "Accept-Ranges" => "bytes",
460 }; 548 };
461 549
462 my @code = (200, "ok"); 550 my @code = (200, "ok");
463 my ($l, $h); 551 my ($l, $h);
464 552
479 $self->err(416, "not satisfiable", $hdr, ""); 567 $self->err(416, "not satisfiable", $hdr, "");
480 568
481satisfiable: 569satisfiable:
482 # check for segmented downloads 570 # check for segmented downloads
483 if ($l && $::NO_SEGMENTED) { 571 if ($l && $::NO_SEGMENTED) {
484 my $delay = $::NOW + $::PER_TIMEOUT + 15; 572 my $timeout = $::NOW + 15;
485 while (%{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { 573 while (keys %{$uri{$self->{remote_id}}{$self->{uri}}} > 1) {
486 if ($delay <= $::NOW) { 574 if ($timeout <= $::NOW) {
575 $self->block($::BLOCKTIME, "segmented downloads are forbidden");
487 $self->err_segmented_download; 576 #$self->err_segmented_download;
488 } else { 577 } else {
489 $httpevent->broadcast; 578 $httpevent->wait;
490 } 579 }
491 } 580 }
492 } 581 }
493 582
494 $hdr->{"Content-Range"} = "bytes $l-$h/$length"; 583 $hdr->{"Content-Range"} = "bytes $l-$h/$length";
506 595
507 $self->response(@code, $hdr, ""); 596 $self->response(@code, $hdr, "");
508 597
509 if ($self->{method} eq "GET") { 598 if ($self->{method} eq "GET") {
510 $self->{time} = $::NOW; 599 $self->{time} = $::NOW;
511
512 my $fudge = $queue->[0]->waiters;
513 $fudge = $fudge ? ($fudge+1)/$fudge : 1;
514
515 $queue->[1] *= $fudge;
516 my $transfer = $queue->[0]->guard;
517
518 if ($fudge != 1) {
519 $queue->[1] /= $fudge;
520 $queue->[1] = $queue->[1] * $::wait_factor
521 + ($::NOW - $self->{time}) * (1 - $::wait_factor);
522 }
523 $self->{time} = $::NOW; 600 $self->{written} = 0;
524 601
525 $self->{fh}->writable or return; 602 my $current = $Coro::current;
526 603
527 my ($fh, $buf, $r); 604 my ($fh, $buf, $r);
528 my $current = $Coro::current; 605
529 open $fh, "<", $self->{path} 606 open $fh, "<", $self->{path}
530 or die "$self->{path}: late open failure ($!)"; 607 or die "$self->{path}: late open failure ($!)";
531 608
532 $h -= $l - 1; 609 $h -= $l - 1;
533 610
534 if (0) { 611 if (0) { # !AIO
535 if ($l) { 612 if ($l) {
536 sysseek $fh, $l, 0; 613 sysseek $fh, $l, 0;
537 } 614 }
538 } 615 }
616
617 my $transfer = $queue->start_transfer($h);
618 my $locked;
619 my $bufsize = $::WAIT_BUFSIZE; # initial buffer size
539 620
540 while ($h > 0) { 621 while ($h > 0) {
541 if (0) { 622 unless ($locked) {
542 sysread $fh, $buf, $h > $::BUFSIZE ? $::BUFSIZE : $h 623 if ($locked ||= $transfer->try($::WAIT_INTERVAL)) {
624 $bufsize = $::BUFSIZE;
625 $self->{time} = $::NOW;
626 $self->{written} = 0;
627 }
628 }
629
630 if ($blocked{$self->{remote_id}}) {
631 $self->{h}{connection} = "close";
632 die bless {}, err::;
633 }
634
635 Coro::AIO::aio_read $fh, $l, ($h > $bufsize ? $bufsize : $h), $buf, 0
543 or last; 636 or last;
544 } else { 637
545 aio_read($fh, $l, ($h > $::BUFSIZE ? $::BUFSIZE : $h), 638 $tbf->request (length $buf);
546 $buf, 0, sub {
547 $r = $_[0];
548 Coro::ready($current);
549 });
550 &Coro::schedule;
551 last unless $r;
552 }
553 my $w = syswrite $self->{fh}, $buf 639 my $w = syswrite $self->{fh}, $buf
554 or last; 640 or last;
555 $::written += $w; 641 $::written += $w;
556 $self->{written} += $w; 642 $self->{written} += $w;
557 $l += $r; 643 $l += $r;
559 645
560 close $fh; 646 close $fh;
561 } 647 }
562} 648}
563 649
5641; 6501
651

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines