1 | use Coro; |
1 | use Coro; |
2 | use Coro::Semaphore; |
2 | use Coro::Semaphore; |
|
|
3 | use Coro::SemaphoreSet; |
3 | use Coro::Event; |
4 | use Coro::EV; |
4 | use Coro::Socket; |
5 | use Coro::Socket; |
|
|
6 | use Coro::Signal; |
|
|
7 | use Coro::AIO (); |
5 | |
8 | |
6 | no utf8; |
9 | use Fcntl; |
7 | use bytes; |
10 | use HTTP::Date; |
|
|
11 | use POSIX (); |
|
|
12 | |
|
|
13 | use Compress::Zlib (); |
|
|
14 | |
|
|
15 | use common::sense; |
8 | |
16 | |
9 | # at least on my machine, this thingy serves files |
17 | # at least on my machine, this thingy serves files |
10 | # quite a bit faster than apache, ;) |
18 | # quite a bit faster than apache, ;) |
11 | # and quite a bit slower than thttpd :( |
19 | # and quite a bit slower than thttpd :( |
12 | |
20 | |
13 | $SIG{PIPE} = 'IGNORE'; |
21 | $SIG{PIPE} = 'IGNORE'; |
14 | |
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 | |
15 | sub slog { |
54 | sub slog { |
16 | my $level = shift; |
55 | my $level = shift; |
17 | my $format = shift; |
56 | my $format = shift; |
|
|
57 | |
|
|
58 | $format = sprintf $format, @_ if @_; |
|
|
59 | |
|
|
60 | my $NOW = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW); |
18 | printf "---: $format\n", @_; |
61 | print "$NOW: $format\n"; |
|
|
62 | print $errorlog "$NOW: $format\n", @_ if $errorlog; |
19 | } |
63 | } |
20 | |
64 | |
21 | my $connections = new Coro::Semaphore $MAX_CONNECTS; |
65 | our $connections = new Coro::Semaphore $::MAX_CONNECTS || 250; |
|
|
66 | our $httpevent = new Coro::Signal; |
22 | |
67 | |
23 | my @fh; |
68 | our $queue_file = new transferqueue $::MAX_TRANSFERS; |
24 | my @pool; |
69 | our $queue_index = new transferqueue 10; |
25 | |
70 | |
26 | # one "execution thread" |
71 | our $tbf_top = new tbf rate => $::TBF_RATE || 100000; |
27 | sub handler { |
72 | |
|
|
73 | my $unused_bytes = 0; |
|
|
74 | my $unused_last = time; |
|
|
75 | |
|
|
76 | sub unused_bandwidth { |
|
|
77 | $unused_bytes += $_[0]; |
|
|
78 | if ($unused_last < $NOW - 30 && $unused_bytes / ($NOW - $unused_last) > 50000) { |
|
|
79 | $unused_last = $NOW; |
|
|
80 | $unused_bytes = 0; |
|
|
81 | $queue_file->force_wake_next |
|
|
82 | and slog 1, "forced filetransfer due to unused bandwidth"; |
|
|
83 | } |
|
|
84 | } |
|
|
85 | |
|
|
86 | sub listen_on { |
|
|
87 | my $listen = $_[0]; |
|
|
88 | |
|
|
89 | push @listen_sockets, $listen; |
|
|
90 | |
|
|
91 | # the "main thread" |
|
|
92 | async { |
|
|
93 | slog 1, "accepting connections"; |
28 | while () { |
94 | while () { |
29 | my $fh = pop @fh; |
95 | $connections->down; |
30 | if ($fh) { |
96 | my @conn = $listen->accept; |
|
|
97 | #slog 3, "accepted @$connections ".scalar(@pool); |
|
|
98 | |
|
|
99 | async_pool { |
31 | eval { |
100 | eval { |
32 | conn->new($fh)->handle; |
101 | conn->new (@conn)->handle; |
33 | }; |
102 | }; |
34 | close $fh; |
|
|
35 | slog 1, "$@" if $@ && !ref $@; |
103 | slog 1, "$@" if $@ && !ref $@; |
|
|
104 | |
|
|
105 | $httpevent->broadcast; # only for testing, but doesn't matter much |
|
|
106 | |
36 | $connections->up; |
107 | $connections->up; |
37 | } else { |
|
|
38 | last if @pool >= $MAX_POOL; |
|
|
39 | push @pool, $Coro::current; |
|
|
40 | schedule; |
|
|
41 | } |
108 | } |
|
|
109 | } |
42 | } |
110 | }; |
43 | } |
111 | } |
44 | |
112 | |
45 | my $http_port = new Coro::Socket |
113 | my $http_port = new Coro::Socket |
46 | LocalAddr => $SERVER_HOST, |
114 | LocalAddr => $::SERVER_HOST, |
47 | LocalPort => $SERVER_PORT, |
115 | LocalPort => $::SERVER_PORT, |
48 | ReuseAddr => 1, |
116 | ReuseAddr => 1, |
49 | Listen => 1, |
117 | Listen => 50, |
50 | or die "unable to start server"; |
118 | or die "unable to start server"; |
51 | |
119 | |
52 | push @listen_sockets, $http_port; |
120 | listen_on $http_port; |
53 | |
121 | |
54 | # the "main thread" |
122 | if ($::SERVER_PORT2) { |
55 | async { |
123 | my $http_port = new Coro::Socket |
56 | slog 1, "accepting connections"; |
124 | LocalAddr => $::SERVER_HOST, |
57 | while () { |
125 | LocalPort => $::SERVER_PORT2, |
58 | $connections->down; |
126 | ReuseAddr => 1, |
59 | push @fh, $http_port->accept; |
127 | Listen => 50, |
60 | #slog 3, "accepted @$connections ".scalar(@pool); |
128 | or die "unable to start server"; |
61 | $::NOW = time; |
|
|
62 | if (@pool) { |
|
|
63 | (pop @pool)->ready; |
|
|
64 | } else { |
|
|
65 | async \&handler; |
|
|
66 | } |
|
|
67 | |
129 | |
68 | } |
130 | listen_on $http_port; |
69 | }; |
131 | } |
70 | |
132 | |
71 | package conn; |
133 | package conn; |
|
|
134 | |
|
|
135 | use common::sense; |
72 | |
136 | |
73 | use Socket; |
137 | use Socket; |
74 | use HTTP::Date; |
138 | use HTTP::Date; |
75 | use Convert::Scalar 'weaken'; |
139 | use Convert::Scalar 'weaken'; |
|
|
140 | use IO::AIO; |
76 | |
141 | |
|
|
142 | IO::AIO::min_parallel $::AIO_PARALLEL; |
|
|
143 | |
|
|
144 | our $AIO_WATCHER = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb; |
|
|
145 | |
77 | our %conn; # $conn{ip}{fh} => connobj |
146 | our %conn; # $conn{ip}{self} => connobj |
|
|
147 | our %uri; # $uri{ip}{uri}{self} |
78 | our %blocked; |
148 | our %blocked; |
|
|
149 | our %mimetype; |
|
|
150 | |
|
|
151 | sub read_mimetypes { |
|
|
152 | if (open my $fh, "<mime_types") { |
|
|
153 | while (<$fh>) { |
|
|
154 | if (/^([^#]\S+)\t+(\S+)$/) { |
|
|
155 | $mimetype{lc $1} = $2; |
|
|
156 | } |
|
|
157 | } |
|
|
158 | } else { |
|
|
159 | print "cannot open mime_types\n"; |
|
|
160 | } |
|
|
161 | } |
|
|
162 | |
|
|
163 | read_mimetypes; |
79 | |
164 | |
80 | sub new { |
165 | sub new { |
81 | my $class = shift; |
166 | my $class = shift; |
82 | my $fh = shift; |
167 | my $fh = shift; |
|
|
168 | my $peername = shift; |
83 | my $self = bless { fh => $fh }, $class; |
169 | my $self = bless { fh => $fh }, $class; |
84 | my (undef, $iaddr) = unpack_sockaddr_in $fh->getpeername |
170 | my (undef, $iaddr) = unpack_sockaddr_in $peername |
85 | or $self->err(500, "unable to get peername"); |
171 | or $self->err (500, "unable to decode peername"); |
|
|
172 | |
|
|
173 | $self->{remote_addr} = |
86 | $self->{remote_addr} = inet_ntoa $iaddr; |
174 | $self->{remote_id} = inet_ntoa $iaddr; |
87 | |
175 | |
88 | # enter ourselves into various lists |
176 | $self->{time} = $::NOW; |
89 | weaken ($conn{$self->{remote_addr}}{$self*1} = $self); |
|
|
90 | |
177 | |
|
|
178 | weaken ($Coro::current->{conn} = $self); |
|
|
179 | |
|
|
180 | ++$::conns; |
|
|
181 | $::maxconns = $::conns if $::conns > $::maxconns; |
|
|
182 | |
91 | $self; |
183 | $self |
92 | } |
184 | } |
93 | |
185 | |
94 | sub DESTROY { |
186 | sub DESTROY { |
95 | my $self = shift; |
187 | my $self = shift; |
96 | delete $conn{$self->{remote_addr}}{$self*1}; |
188 | |
97 | delete $uri{$self->{uri}}{$self*1}; |
189 | --$::conns; |
98 | } |
190 | } |
|
|
191 | |
|
|
192 | sub prune_cache { |
|
|
193 | my $hash = $_[0]; |
|
|
194 | |
|
|
195 | for (keys %$hash) { |
|
|
196 | if (ref $hash->{$_} eq HASH::) { |
|
|
197 | prune_cache($hash->{$_}); |
|
|
198 | unless (scalar keys %{$hash->{$_}}) { |
|
|
199 | delete $hash->{$_}; |
|
|
200 | } |
|
|
201 | } |
|
|
202 | } |
|
|
203 | } |
|
|
204 | |
|
|
205 | sub prune_caches { |
|
|
206 | prune_cache \%conn; |
|
|
207 | prune_cache \%uri; |
|
|
208 | |
|
|
209 | for (keys %blocked) { |
|
|
210 | delete $blocked{$_} unless $blocked{$_}[0] > $::NOW; |
|
|
211 | } |
|
|
212 | } |
|
|
213 | |
|
|
214 | our $PRUNE_WATCHER = EV::timer 60, 60, \&prune_caches; |
99 | |
215 | |
100 | sub slog { |
216 | sub slog { |
101 | my $self = shift; |
217 | my $self = shift; |
102 | main::slog($_[0], "$self->{remote_addr}> $_[1]"); |
218 | main::slog($_[0], "$self->{remote_id}> $_[1]"); |
103 | } |
219 | } |
104 | |
220 | |
105 | sub response { |
221 | sub response { |
106 | my ($self, $code, $msg, $hdr, $content) = @_; |
222 | my ($self, $code, $msg, $hdr, $content) = @_; |
107 | my $res = "HTTP/1.0 $code $msg\015\012"; |
223 | my $res = "HTTP/1.1 $code $msg\015\012"; |
|
|
224 | my $GZ = ""; |
108 | |
225 | |
109 | $res .= "Connection: close\015\012"; |
226 | if (exists $hdr->{Connection}) { |
110 | $res .= "Date: ".(time2str $::NOW)."\015\012"; # slow? nah. :( |
227 | if ($hdr->{Connection} =~ /close/) { |
|
|
228 | $self->{h}{connection} = "close" |
|
|
229 | } |
|
|
230 | } else { |
|
|
231 | if ($self->{version} < 1.1) { |
|
|
232 | if ($self->{h}{connection} =~ /keep-alive/i) { |
|
|
233 | $hdr->{Connection} = "Keep-Alive"; |
|
|
234 | } else { |
|
|
235 | $self->{h}{connection} = "close" |
|
|
236 | } |
|
|
237 | } |
|
|
238 | } |
|
|
239 | |
|
|
240 | if ($self->{method} ne "HEAD" |
|
|
241 | && $self->{h}{"accept-encoding"} =~ /\bgzip\b/ |
|
|
242 | && 400 < length $content |
|
|
243 | && $hdr->{"Content-Length"} == length $content |
|
|
244 | && !exists $hdr->{"Content-Encoding"} |
|
|
245 | ) { |
|
|
246 | my $orig = length $content; |
|
|
247 | $hdr->{"Content-Encoding"} = "gzip"; |
|
|
248 | $content = Compress::Zlib::memGzip(\$content); |
|
|
249 | $hdr->{"Content-Length"} = length $content; |
|
|
250 | $GZ = sprintf "GZ%02d", 100 - 100*((length $content) / $orig); |
|
|
251 | } |
|
|
252 | |
|
|
253 | $res .= "Date: $HTTP_NOW\015\012"; |
|
|
254 | $res .= "Server: $::NAME\015\012"; |
111 | |
255 | |
112 | while (my ($h, $v) = each %$hdr) { |
256 | while (my ($h, $v) = each %$hdr) { |
113 | $res .= "$h: $v\015\012" |
257 | $res .= "$h: $v\015\012" |
114 | } |
258 | } |
|
|
259 | $res .= "\015\012"; |
115 | |
260 | |
116 | $res .= "\015\012$content" if defined $content; |
261 | $res .= $content if defined $content and $self->{method} ne "HEAD"; |
117 | |
262 | |
118 | print STDERR "$self->{remote_addr} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}." \"$self->{h}{referer}\"\n";#d# |
263 | my $log = (POSIX::strftime "%Y-%m-%d %H:%M:%S", gmtime $::NOW). |
|
|
264 | " $self->{remote_id} \"$self->{uri}\" $code ".$hdr->{"Content-Length"}.$GZ. |
|
|
265 | " \"$self->{h}{referer}\"\n"; |
119 | |
266 | |
120 | print {$self->{fh}} $res; |
267 | print $::accesslog $log if $::accesslog; |
|
|
268 | print STDERR $log; |
|
|
269 | |
|
|
270 | $tbf_top->request(length $res, 1e6); |
|
|
271 | $self->{written} += print {$self->{fh}} $res; |
121 | } |
272 | } |
122 | |
273 | |
123 | sub err { |
274 | sub err { |
124 | my $self = shift; |
275 | my $self = shift; |
125 | my ($code, $msg, $hdr, $content) = @_; |
276 | my ($code, $msg, $hdr, $content) = @_; |
126 | |
277 | |
127 | unless (defined $content) { |
278 | unless (defined $content) { |
128 | $content = "$code $msg"; |
279 | $content = "$code $msg\n"; |
129 | $hdr->{"Content-Type"} = "text/plain"; |
280 | $hdr->{"Content-Type"} = "text/plain"; |
130 | $hdr->{"Content-Length"} = length $content; |
281 | $hdr->{"Content-Length"} = length $content; |
131 | } |
282 | } |
|
|
283 | $hdr->{"Connection"} = "close"; |
132 | |
284 | |
133 | $self->response($code, $msg, $hdr, $content); |
285 | $self->response ($code, $msg, $hdr, $content); |
134 | |
286 | |
135 | die bless {}, err::; |
287 | die bless {}, err:: |
136 | } |
|
|
137 | |
|
|
138 | sub err_blocked { |
|
|
139 | my $self = shift; |
|
|
140 | my $ip = $self->{remote_addr}; |
|
|
141 | my $time = time2str $blocked{$ip} = $::NOW + $::BLOCKTIME; |
|
|
142 | $self->err(403, "too many connections", |
|
|
143 | { |
|
|
144 | "Content-Type" => "text/html", |
|
|
145 | "Retry-After" => $::BLOCKTIME |
|
|
146 | }, |
|
|
147 | <<EOF); |
|
|
148 | <html><p> |
|
|
149 | You have been blocked because you opened too many connections. You |
|
|
150 | may retry at</p> |
|
|
151 | |
|
|
152 | <p><blockquote>$time.</blockquote></p> |
|
|
153 | |
|
|
154 | <p>Until then, each new access will renew the block. You might want to have a |
|
|
155 | look at the <a href="http://www.goof.com/pcg/marc/animefaq.html">FAQ</a>.</p> |
|
|
156 | </html> |
|
|
157 | EOF |
|
|
158 | } |
288 | } |
159 | |
289 | |
160 | sub handle { |
290 | sub handle { |
161 | my $self = shift; |
291 | my $self = shift; |
162 | my $fh = $self->{fh}; |
292 | my $fh = $self->{fh}; |
163 | |
293 | |
|
|
294 | my $host; |
|
|
295 | |
|
|
296 | $fh->timeout($::REQ_TIMEOUT); |
164 | #while() { |
297 | while () { |
|
|
298 | $self->{reqs}++; |
|
|
299 | |
|
|
300 | # read request and parse first line |
|
|
301 | my $req = $fh->readline("\015\012\015\012"); |
|
|
302 | |
|
|
303 | unless (defined $req) { |
|
|
304 | if (exists $self->{version}) { |
|
|
305 | last; |
|
|
306 | } else { |
|
|
307 | $self->err(408, "request timeout"); |
|
|
308 | } |
|
|
309 | } |
|
|
310 | |
165 | $self->{h} = {}; |
311 | $self->{h} = {}; |
166 | |
312 | |
167 | # read request and parse first line |
|
|
168 | $fh->timeout($::REQ_TIMEOUT); |
|
|
169 | my $req = $fh->readline("\015\012\015\012"); |
|
|
170 | $fh->timeout($::RES_TIMEOUT); |
313 | $fh->timeout($::RES_TIMEOUT); |
171 | |
|
|
172 | defined $req or |
|
|
173 | $self->err(408, "request timeout"); |
|
|
174 | |
|
|
175 | my $ip = $self->{remote_addr}; |
|
|
176 | |
|
|
177 | if ($blocked{$ip}) { |
|
|
178 | $self->err_blocked($blocked{$ip}) |
|
|
179 | if $blocked{$ip} > $::NOW; |
|
|
180 | |
|
|
181 | delete $blocked{$ip}; |
|
|
182 | } |
|
|
183 | |
|
|
184 | if (%{$conn{$ip}} > $::MAX_CONN_IP) { |
|
|
185 | $self->slog("blocked ip $ip"); |
|
|
186 | $self->err_blocked; |
|
|
187 | } |
|
|
188 | |
314 | |
189 | $req =~ /^(?:\015\012)? |
315 | $req =~ /^(?:\015\012)? |
190 | (GET|HEAD) \040+ |
316 | (GET|HEAD) \040+ |
191 | ([^\040]+) \040+ |
317 | ([^\040]+) \040+ |
192 | HTTP\/([0-9]+\.[0-9]+) |
318 | HTTP\/([0-9]+\.[0-9]+) |
193 | \015\012/gx |
319 | \015\012/gx |
194 | or $self->err(403, "method not allowed", { Allow => "GET,HEAD" }); |
320 | or $self->err(405, "method not allowed", { Allow => "GET,HEAD" }); |
195 | |
|
|
196 | $2 ne "1.0" |
|
|
197 | or $self->err(506, "http protocol version not supported"); |
|
|
198 | |
321 | |
199 | $self->{method} = $1; |
322 | $self->{method} = $1; |
200 | $self->{uri} = $2; |
323 | $self->{uri} = $2; |
|
|
324 | $self->{version} = $3; |
|
|
325 | |
|
|
326 | $3 =~ /^1\./ |
|
|
327 | or $self->err(506, "http protocol version $3 not supported"); |
201 | |
328 | |
202 | # parse headers |
329 | # parse headers |
203 | { |
330 | { |
204 | my (%hdr, $h, $v); |
331 | my (%hdr, $h, $v); |
205 | |
332 | |
206 | $hdr{lc $1} .= ",$2" |
333 | $hdr{lc $1} .= ",$2" |
207 | while $req =~ /\G |
334 | while $req =~ /\G |
208 | ([^:\000-\040]+): |
335 | ([^:\000-\040]+): |
209 | [\008\040]* |
336 | [\011\040]* |
210 | ((?: [^\015\012]+ | \015\012[\008\040] )*) |
337 | ((?: [^\015\012]+ | \015\012[\011\040] )*) |
211 | \015\012 |
338 | \015\012 |
212 | /gxc; |
339 | /gxc; |
213 | |
340 | |
214 | $req =~ /\G\015\012$/ |
341 | $req =~ /\G\015\012$/ |
215 | or $self->err(400, "bad request"); |
342 | or $self->err(400, "bad request"); |
216 | |
343 | |
217 | $self->{h}{$h} = substr $v, 1 |
344 | $self->{h}{$h} = substr $v, 1 |
218 | while ($h, $v) = each %hdr; |
345 | while ($h, $v) = each %hdr; |
219 | } |
346 | } |
220 | |
347 | |
|
|
348 | # remote id should be unique per user |
|
|
349 | my $id = $self->{remote_addr}; |
|
|
350 | |
|
|
351 | if (exists $self->{h}{"client-ip"}) { |
|
|
352 | $id .= "[".$self->{h}{"client-ip"}."]"; |
|
|
353 | } elsif (exists $self->{h}{"x-forwarded-for"}) { |
|
|
354 | $id .= "[".$self->{h}{"x-forwarded-for"}."]"; |
|
|
355 | } |
|
|
356 | |
|
|
357 | $self->{remote_id} = $id; |
|
|
358 | |
|
|
359 | weaken (local $conn{$id}{$self*1} = $self); |
|
|
360 | |
|
|
361 | if ($blocked{$id}) { |
|
|
362 | $self->err_blocked |
|
|
363 | if $blocked{$id}[0] > $::NOW; |
|
|
364 | |
|
|
365 | delete $blocked{$id}; |
|
|
366 | } |
|
|
367 | |
|
|
368 | # find out server name and port |
|
|
369 | if ($self->{uri} =~ s/^http:\/\/([^\/?#]*)//i) { |
|
|
370 | $host = $1; |
|
|
371 | } else { |
|
|
372 | $host = $self->{h}{host}; |
|
|
373 | } |
|
|
374 | |
|
|
375 | if (defined $host) { |
221 | $self->{server_port} = $self->{h}{host} =~ s/:([0-9]+)$// ? $1 : 80; |
376 | $self->{server_port} = $host =~ s/:([0-9]+)$// ? $1 : 80; |
|
|
377 | } else { |
|
|
378 | ($self->{server_port}, $host) |
|
|
379 | = unpack_sockaddr_in $self->{fh}->sockname |
|
|
380 | or $self->err(500, "unable to get socket name"); |
|
|
381 | $host = inet_ntoa $host; |
|
|
382 | } |
222 | |
383 | |
|
|
384 | $self->{server_name} = $host; |
|
|
385 | |
223 | weaken ($uri{$self->{uri}}{$self*1} = $self); |
386 | weaken (local $uri{$id}{$self->{uri}}{$self*1} = $self); |
224 | |
387 | |
|
|
388 | eval { |
225 | $self->map_uri; |
389 | $self->map_uri; |
226 | $self->respond; |
390 | $self->respond; |
|
|
391 | }; |
|
|
392 | |
|
|
393 | die if $@ && !ref $@; |
|
|
394 | |
|
|
395 | last if $self->{h}{connection} =~ /close/i; |
|
|
396 | |
|
|
397 | $httpevent->broadcast; |
|
|
398 | |
|
|
399 | $fh->timeout($::PER_TIMEOUT); |
227 | #} |
400 | } |
|
|
401 | } |
|
|
402 | |
|
|
403 | sub block { |
|
|
404 | my $self = shift; |
|
|
405 | |
|
|
406 | $blocked{$self->{remote_id}} = [$::NOW + $_[0], $_[1]]; |
|
|
407 | $self->slog(2, "blocked ip $self->{remote_id}"); |
|
|
408 | $self->err_blocked; |
228 | } |
409 | } |
229 | |
410 | |
230 | # uri => path mapping |
411 | # uri => path mapping |
231 | sub map_uri { |
412 | sub map_uri { |
232 | my $self = shift; |
413 | my $self = shift; |
233 | my $host = $self->{h}{host} || "default"; |
414 | my $host = $self->{server_name}; |
234 | my $uri = $self->{uri}; |
415 | my $uri = $self->{uri}; |
|
|
416 | |
|
|
417 | $host =~ /[\/\\]/ |
|
|
418 | and $self->err(400, "bad request"); |
235 | |
419 | |
236 | # some massaging, also makes it more secure |
420 | # some massaging, also makes it more secure |
237 | $uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
421 | $uri =~ s/%([0-9a-fA-F][0-9a-fA-F])/chr hex $1/ge; |
238 | $uri =~ s%//+%/%g; |
422 | $uri =~ s%//+%/%g; |
239 | $uri =~ s%/\.(?=/|$)%%g; |
423 | $uri =~ s%/\.(?=/|$)%%g; |
… | |
… | |
244 | |
428 | |
245 | $self->{name} = $uri; |
429 | $self->{name} = $uri; |
246 | |
430 | |
247 | # now do the path mapping |
431 | # now do the path mapping |
248 | $self->{path} = "$::DOCROOT/$host$uri"; |
432 | $self->{path} = "$::DOCROOT/$host$uri"; |
249 | } |
|
|
250 | |
433 | |
251 | sub server_address { |
434 | $self->access_check; |
252 | my $self = shift; |
|
|
253 | my ($port, $iaddr) = unpack_sockaddr_in $self->{fh}->getsockname |
|
|
254 | or $self->err(500, "unable to get socket name"); |
|
|
255 | ((inet_ntoa $iaddr), $port); |
|
|
256 | } |
|
|
257 | |
|
|
258 | sub server_host { |
|
|
259 | my $self = shift; |
|
|
260 | if (exists $self->{h}{host}) { |
|
|
261 | return $self->{h}{host}; |
|
|
262 | } else { |
|
|
263 | return (($self->server_address)[0]); |
|
|
264 | } |
|
|
265 | } |
|
|
266 | |
|
|
267 | sub server_hostport { |
|
|
268 | my $self = shift; |
|
|
269 | my ($host, $port); |
|
|
270 | if (exists $self->{h}{host}) { |
|
|
271 | ($host, $port) = ($self->{h}{host}, $self->{server_port}); |
|
|
272 | } else { |
|
|
273 | ($host, $port) = $self->server_address; |
|
|
274 | } |
|
|
275 | $port = $port == 80 ? "" : ":$port"; |
|
|
276 | $host.$port; |
|
|
277 | } |
435 | } |
278 | |
436 | |
279 | sub _cgi { |
437 | sub _cgi { |
280 | my $self = shift; |
438 | my $self = shift; |
281 | my $path = shift; |
439 | my $path = shift; |
… | |
… | |
284 | # no two-way xxx supported |
442 | # no two-way xxx supported |
285 | if (0 == fork) { |
443 | if (0 == fork) { |
286 | open STDOUT, ">&".fileno($self->{fh}); |
444 | open STDOUT, ">&".fileno($self->{fh}); |
287 | if (chdir $::DOCROOT) { |
445 | if (chdir $::DOCROOT) { |
288 | $ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
446 | $ENV{SERVER_SOFTWARE} = "thttpd-myhttpd"; # we are thttpd-alike |
289 | $ENV{HTTP_HOST} = $self->server_host; |
447 | $ENV{HTTP_HOST} = $self->{server_name}; |
290 | $ENV{HTTP_PORT} = $self->{server_host}; |
448 | $ENV{HTTP_PORT} = $self->{server_port}; |
291 | $ENV{SCRIPT_NAME} = $self->{name}; |
449 | $ENV{SCRIPT_NAME} = $self->{name}; |
292 | exec $::INDEXPROG; |
450 | exec $path; |
293 | } |
451 | } |
294 | Coro::State::_exit(0); |
452 | Coro::State::_exit(0); |
295 | } else { |
453 | } else { |
|
|
454 | die; |
296 | } |
455 | } |
|
|
456 | } |
|
|
457 | |
|
|
458 | sub server_hostport { |
|
|
459 | $_[0]{server_port} == 80 |
|
|
460 | ? $_[0]{server_name} |
|
|
461 | : "$_[0]{server_name}:$_[0]{server_port}"; |
297 | } |
462 | } |
298 | |
463 | |
299 | sub respond { |
464 | sub respond { |
300 | my $self = shift; |
465 | my $self = shift; |
301 | my $path = $self->{path}; |
466 | my $path = $self->{path}; |
302 | |
467 | |
303 | stat $path |
468 | if ($self->{name} =~ s%^/internal/([^/]+)%%) { |
304 | or $self->err(404, "not found"); |
469 | if ($::internal{$1}) { |
305 | |
470 | $::internal{$1}->($self); |
306 | # idiotic netscape sends idiotic headers AGAIN |
|
|
307 | my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
|
|
308 | ? str2time $1 : 0; |
|
|
309 | |
|
|
310 | if (-d _ && -r _) { |
|
|
311 | # directory |
|
|
312 | if ($path !~ /\/$/) { |
|
|
313 | # create a redirect to get the trailing "/" |
|
|
314 | my $host = $self->server_hostport; |
|
|
315 | $self->err(301, "moved permanently", { Location => "http://$host$self->{uri}/" }); |
|
|
316 | } else { |
471 | } else { |
317 | $ims < (stat _)[9] |
472 | $self->err (404, "not found"); |
|
|
473 | } |
|
|
474 | } else { |
|
|
475 | |
|
|
476 | Coro::AIO::aio_stat $path |
|
|
477 | and $self->err (404, "not found"); |
|
|
478 | |
|
|
479 | $self->{stat} = [stat _]; |
|
|
480 | |
|
|
481 | # idiotic netscape sends idiotic headers AGAIN |
|
|
482 | my $ims = $self->{h}{"if-modified-since"} =~ /^([^;]+)/ |
|
|
483 | ? str2time $1 : 0; |
|
|
484 | |
|
|
485 | if (-d _ && -r _) { |
|
|
486 | # directory |
|
|
487 | if ($path !~ /\/$/) { |
|
|
488 | # create a redirect to get the trailing "/" |
|
|
489 | # we don't try to avoid the :80 |
|
|
490 | $self->err (301, "moved permanently", { Location => "http://".$self->server_hostport."$self->{uri}/" }); |
|
|
491 | } else { |
|
|
492 | $ims < $self->{stat}[9] |
318 | or $self->err(304, "not modified"); |
493 | or $self->err (304, "not modified"); |
319 | |
494 | |
320 | if ($self->{method} eq "GET") { |
|
|
321 | if (-r "$path/index.html") { |
495 | if (-r "$path/index.html") { |
|
|
496 | # replace directory "size" by index.html filesize |
322 | $self->{path} .= "/index.html"; |
497 | $self->{stat} = [stat ($self->{path} .= "/index.html")]; |
323 | $self->handle_file; |
498 | $self->handle_file ($queue_index, $tbf_top); |
324 | } else { |
499 | } else { |
325 | $self->handle_dir; |
500 | $self->handle_dir; |
326 | } |
501 | } |
327 | } |
502 | } |
328 | } |
|
|
329 | } elsif (-f _ && -r _) { |
503 | } elsif (-f _ && -r _) { |
330 | -x _ and $self->err(403, "forbidden"); |
504 | -x _ and $self->err (403, "forbidden"); |
331 | $self->handle_file; |
505 | |
|
|
506 | if (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
|
|
507 | my $timeout = $::NOW + 10; |
|
|
508 | while (keys %{$conn{$self->{remote_id}}} > $::MAX_TRANSFERS_IP) { |
|
|
509 | if ($timeout < $::NOW) { |
|
|
510 | $self->block($::BLOCKTIME, "too many connections"); |
|
|
511 | } else { |
|
|
512 | $httpevent->wait; |
|
|
513 | } |
|
|
514 | } |
|
|
515 | } |
|
|
516 | |
|
|
517 | $self->handle_file ($queue_file, $tbf_top); |
332 | } else { |
518 | } else { |
333 | $self->err(404, "not found"); |
519 | $self->err (404, "not found"); |
|
|
520 | } |
334 | } |
521 | } |
335 | } |
522 | } |
336 | |
523 | |
337 | sub handle_dir { |
524 | sub handle_dir { |
338 | my $self = shift; |
525 | my $self = shift; |
339 | $self->_cgi($::INDEXPROG); |
526 | my $idx = $self->diridx; |
|
|
527 | |
|
|
528 | $self->response (200, "ok", |
|
|
529 | { |
|
|
530 | "Content-Type" => "text/html; charset=utf-8", |
|
|
531 | "Content-Length" => length $idx, |
|
|
532 | "Last-Modified" => time2str ($self->{stat}[9]), |
|
|
533 | }, |
|
|
534 | $idx); |
340 | } |
535 | } |
341 | |
536 | |
342 | sub handle_file { |
537 | sub handle_file { |
343 | my $self = shift; |
538 | my ($self, $queue, $tbf) = @_; |
344 | my $length = -s _; |
539 | my $length = $self->{stat}[7]; |
345 | my $hdr = { |
540 | my $hdr = { |
346 | "Last-Modified" => time2str ((stat _)[9]), |
541 | "Last-Modified" => time2str ((stat _)[9]), |
|
|
542 | "Accept-Ranges" => "bytes", |
347 | }; |
543 | }; |
348 | |
544 | |
349 | my @code = (200, "ok"); |
545 | my @code = (200, "ok"); |
350 | my ($l, $h); |
546 | my ($l, $h); |
351 | |
547 | |
352 | if ($self->{h}{range} =~ /^bytes=(.*)$/) { |
548 | if ($self->{h}{range} =~ /^bytes=(.*)$/i) { |
353 | for (split /,/, $1) { |
549 | for (split /,/, $1) { |
354 | if (/^-(\d+)$/) { |
550 | if (/^-(\d+)$/) { |
355 | ($l, $h) = ($length - $1, $length - 1); |
551 | ($l, $h) = ($length - $1, $length - 1); |
356 | } elsif (/^(\d+)-(\d*)$/) { |
552 | } elsif (/^(\d+)-(\d*)$/) { |
357 | ($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
553 | ($l, $h) = ($1, ($2 ne "" || $2 >= $length) ? $2 : $length - 1); |
358 | } else { |
554 | } else { |
359 | ($l, $h) = (0, $length - 1); |
555 | ($l, $h) = (0, $length - 1); |
360 | goto ignore; |
556 | goto ignore; |
361 | } |
557 | } |
362 | goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h > $l; |
558 | goto satisfiable if $l >= 0 && $l < $length && $h >= 0 && $h >= $l; |
363 | } |
559 | } |
364 | $hdr->{"Content-Range"} = "bytes */$length"; |
560 | $hdr->{"Content-Range"} = "bytes */$length"; |
|
|
561 | $hdr->{"Content-Length"} = $length; |
365 | $self->err(416, "not satisfiable", $hdr); |
562 | $self->err (416, "not satisfiable", $hdr, ""); |
366 | |
563 | |
367 | satisfiable: |
564 | satisfiable: |
368 | # check for segmented downloads |
565 | # check for segmented downloads |
369 | if ($l && $NO_SEGMENTED) { |
566 | if ($l && $::NO_SEGMENTED) { |
370 | if (%{$uri{$self->{uri}}} > 1) { |
567 | my $timeout = $::NOW + 15; |
371 | $self->slog("segmented download refused\n"); |
568 | while (keys %{$uri{$self->{remote_id}}{$self->{uri}}} > 1) { |
372 | $self->err(400, "segmented downloads are not allowed"); |
569 | if ($timeout <= $::NOW) { |
|
|
570 | $self->block ($::BLOCKTIME, "segmented downloads are forbidden"); |
|
|
571 | #$self->err_segmented_download; |
|
|
572 | } else { |
|
|
573 | $httpevent->wait; |
|
|
574 | } |
373 | } |
575 | } |
374 | } |
576 | } |
375 | |
577 | |
376 | $hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
578 | $hdr->{"Content-Range"} = "bytes $l-$h/$length"; |
377 | @code = (206, "partial content"); |
579 | @code = (206, "partial content"); |
… | |
… | |
380 | ignore: |
582 | ignore: |
381 | } else { |
583 | } else { |
382 | ($l, $h) = (0, $length - 1); |
584 | ($l, $h) = (0, $length - 1); |
383 | } |
585 | } |
384 | |
586 | |
385 | if ($self->{path} =~ /\.html$/) { |
587 | $self->{path} =~ /\.([^.]+)$/; |
386 | $hdr->{"Content-Type"} = "text/html"; |
|
|
387 | } else { |
|
|
388 | $hdr->{"Content-Type"} = "application/octet-stream"; |
588 | $hdr->{"Content-Type"} = $mimetype{lc $1} || "application/octet-stream"; |
389 | } |
|
|
390 | |
|
|
391 | $hdr->{"Content-Length"} = $length; |
589 | $hdr->{"Content-Length"} = $length; |
392 | |
590 | |
393 | $self->response(@code, $hdr, ""); |
591 | $self->response (@code, $hdr, ""); |
394 | |
592 | |
395 | if ($self->{method} eq "GET") { |
593 | if ($self->{method} eq "GET") { |
396 | my ($fh, $buf); |
594 | $self->{time} = $::NOW; |
397 | open $fh, "<", $self->{path} |
595 | $self->{written} = 0; |
|
|
596 | |
|
|
597 | my $fh = Coro::AIO::aio_open $self->{path}, Fcntl::O_RDONLY, 0 |
398 | or die "$self->{path}: late open failure ($!)"; |
598 | or die "$self->{path}: late open failure ($!)"; |
399 | |
599 | |
400 | if ($l) { |
|
|
401 | sysseek $fh, $l, 0 |
|
|
402 | or die "$self->{path}: cannot seek to $l ($!)"; |
|
|
403 | } |
|
|
404 | |
|
|
405 | $h -= $l - 1; |
600 | $h -= $l - 1; |
406 | |
601 | |
|
|
602 | my $transfer = $queue->start_transfer ($h); |
|
|
603 | my $locked; |
|
|
604 | my $bufsize = $::WAIT_BUFSIZE; # initial buffer size |
|
|
605 | |
407 | while ($h > 0) { |
606 | while ($h > 0) { |
408 | $h -= sysread $fh, $buf, $h > $::BUFSIZE ? $::BUFSIZE : $h; |
607 | Coro::cede; |
409 | $self->{fh}->syswrite($buf) |
608 | my $transfer_lock = $TRANSFER_LOCK->guard ($self->{remote_id}); |
|
|
609 | |
|
|
610 | unless ($locked) { |
|
|
611 | if ($locked ||= $transfer->try ($::WAIT_INTERVAL)) { |
|
|
612 | $bufsize = $::BUFSIZE; |
|
|
613 | $self->{time} = $::NOW; |
|
|
614 | $self->{written} = 0; |
|
|
615 | } |
|
|
616 | } |
|
|
617 | |
|
|
618 | if ($blocked{$self->{remote_id}}) { |
|
|
619 | $self->{h}{connection} = "close"; |
|
|
620 | die bless {}, err::; |
|
|
621 | } |
|
|
622 | |
|
|
623 | Coro::AIO::aio_read $fh, $l, ($h > $bufsize ? $bufsize : $h), my $buf, 0 |
410 | or last; |
624 | or last; |
411 | } |
|
|
412 | } |
|
|
413 | |
625 | |
|
|
626 | $tbf->request (length $buf); |
|
|
627 | my $w = $self->{fh}->syswrite ($buf) |
|
|
628 | or last; |
|
|
629 | $::written += $w; |
|
|
630 | $self->{written} += $w; |
|
|
631 | $l += $w; |
|
|
632 | } |
|
|
633 | |
414 | close $fh; |
634 | close $fh; |
|
|
635 | } |
415 | } |
636 | } |
416 | |
637 | |
417 | 1; |
638 | 1 |
|
|
639 | |