TCMS.pm 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423
  1. package TCMS;
  2. use strict;
  3. use warnings;
  4. no warnings 'experimental';
  5. use feature qw{signatures state};
  6. use Date::Format qw{strftime};
  7. use HTTP::Body ();
  8. use URL::Encode ();
  9. use Text::Xslate ();
  10. use Plack::MIME ();
  11. use Mojo::File ();
  12. use DateTime::Format::HTTP();
  13. use CGI::Cookie ();
  14. use File::Basename();
  15. use IO::Compress::Gzip();
  16. use Time::HiRes qw{gettimeofday tv_interval};
  17. use HTTP::Parser::XS qw{HEADERS_AS_HASHREF};
  18. use List::Util;
  19. #Grab our custom routes
  20. use lib 'lib';
  21. use Trog::Routes::HTML;
  22. use Trog::Routes::JSON;
  23. use Trog::Auth;
  24. use Trog::Utils;
  25. use Trog::Config;
  26. use Trog::Data;
  27. use Trog::Vars;
  28. # Troglodyne philosophy - simple as possible
  29. # Import the routes
  30. my $conf = Trog::Config::get();
  31. my $data = Trog::Data->new($conf);
  32. my %roots = $data->routes();
  33. my %routes = %Trog::Routes::HTML::routes;
  34. @routes{keys(%Trog::Routes::JSON::routes)} = values(%Trog::Routes::JSON::routes);
  35. @routes{keys(%roots)} = values(%roots);
  36. my %aliases = $data->aliases();
  37. # XXX this is built progressively across the forks, leading to inconsistent behavior.
  38. # This should eventually be pre-filled from DB.
  39. my %etags;
  40. #1MB chunks
  41. my $CHUNK_SIZE = 1024000;
  42. my $CHUNK_SEP = 'tCMSep666YOLO42069';
  43. #Stuff that isn't in upstream finders
  44. my %extra_types = (
  45. '.docx' => 'application/vnd.openxmlformats-officedocument.wordprocessingml.document',
  46. );
  47. =head2 app()
  48. Dispatches requests based on %routes built above.
  49. The dispatcher here does *not* do anything with the authn/authz data. It sets those in the 'user' and 'acls' parameters of the query object passed to routes.
  50. If a path passed is not a defined route (or regex route), but exists as a file under www/, it will be served up immediately.
  51. =cut
  52. sub app {
  53. # Start the server timing clock
  54. my $start = [gettimeofday];
  55. my $env = shift;
  56. return _toolong() if length($env->{REQUEST_URI}) > 2048;
  57. # Check eTags. If we don't know about it, just assume it's good and lazily fill the cache
  58. # XXX yes, this allows cache poisoning...but only for logged in users!
  59. if ($env->{HTTP_IF_NONE_MATCH}) {
  60. return [304, [], ['']] if $env->{HTTP_IF_NONE_MATCH} eq ($etags{$env->{REQUEST_URI}} || '');
  61. $etags{$env->{REQUEST_URI}} = $env->{HTTP_IF_NONE_MATCH} unless exists $etags{$env->{REQUEST_URI}};
  62. }
  63. my $last_fetch = 0;
  64. if ($env->{HTTP_IF_MODIFIED_SINCE}) {
  65. $last_fetch = DateTime::Format::HTTP->parse_datetime($env->{HTTP_IF_MODIFIED_SINCE})->epoch();
  66. }
  67. #XXX Don't use statics anything that has a search query
  68. # On one hand, I don't want to DOS the disk, but I'd also like some like ?rss...
  69. # Should probably turn those into aliases.
  70. my $has_query = !!$env->{QUERY_STRING};
  71. my $query = {};
  72. $query = URL::Encode::url_params_mixed($env->{QUERY_STRING}) if $env->{QUERY_STRING};
  73. #Actually parse the POSTDATA and dump it into the QUERY object if this is a POST
  74. if ($env->{REQUEST_METHOD} eq 'POST') {
  75. my $body = HTTP::Body->new( $env->{CONTENT_TYPE}, $env->{CONTENT_LENGTH} );
  76. while ( $env->{'psgi.input'}->read(my $buf, $CHUNK_SIZE) ) {
  77. $body->add($buf);
  78. }
  79. @$query{keys(%{$body->param})} = values(%{$body->param});
  80. @$query{keys(%{$body->upload})} = values(%{$body->upload});
  81. }
  82. # Grab the list of ACLs we want to add to a post, if any.
  83. $query->{acls} = [$query->{acls}] if ($query->{acls} && ref $query->{acls} ne 'ARRAY');
  84. my $path = $env->{PATH_INFO};
  85. $path = '/index' if $path eq '/';
  86. # Translate alias paths into their actual path
  87. $path = $aliases{$path} if exists $aliases{$path};
  88. # Figure out if we want compression or not
  89. my $alist = $env->{HTTP_ACCEPT_ENCODING} || '';
  90. $alist =~ s/\s//g;
  91. my @accept_encodings;
  92. @accept_encodings = split(/,/, $alist);
  93. my $deflate = grep { 'gzip' eq $_ } @accept_encodings;
  94. # Collapse multiple slashes in the path
  95. $path =~ s/[\/]+/\//g;
  96. # Let's open up our default route before we bother to see if users even exist
  97. return $routes{default}{callback}->($query) unless -f "config/setup";
  98. my $cookies = {};
  99. if ($env->{HTTP_COOKIE}) {
  100. $cookies = CGI::Cookie->parse($env->{HTTP_COOKIE});
  101. }
  102. my $active_user = '';
  103. if (exists $cookies->{tcmslogin}) {
  104. $active_user = Trog::Auth::session2user($cookies->{tcmslogin}->value);
  105. }
  106. $query->{user_acls} = [];
  107. $query->{user_acls} = Trog::Auth::acls4user($active_user) // [] if $active_user;
  108. # Filter out passed ACLs which are naughty
  109. my $is_admin = grep { $_ eq 'admin' } @{$query->{user_acls}};
  110. @{$query->{acls}} = grep { $_ ne 'admin' } @{$query->{acls}} unless $is_admin;
  111. # Disallow any paths that are naughty ( starman auto-removes .. up-traversal)
  112. if (index($path,'/templates') == 0 || index($path, '/statics') == 0 || $path =~ m/.*(\.psgi|\.pm)$/i ) {
  113. return _forbidden($query);
  114. }
  115. my $streaming = $env->{'psgi.streaming'};
  116. $query->{streaming} = $streaming;
  117. # If we have a static render, just use it instead (These will ALWAYS be correct, data saves invalidate this)
  118. # TODO: make this key on admin INSTEAD of active user when we add non-admin users.
  119. $query->{start} = $start;
  120. if (!$active_user && !$has_query) {
  121. return _static("$path.z",$start, $streaming) if -f "www/statics/$path.z" && $deflate;
  122. return _static($path,$start, $streaming) if -f "www/statics/$path";
  123. }
  124. # Handle HTTP range/streaming requests
  125. my $range = $env->{HTTP_RANGE} || "bytes=0-" if $env->{HTTP_RANGE} || $env->{HTTP_IF_RANGE};
  126. my @ranges;
  127. if ($range) {
  128. $range =~ s/bytes=//g;
  129. push(@ranges, map {
  130. [split(/-/, $_)];
  131. #$tuples[1] //= $tuples[0] + $CHUNK_SIZE;
  132. #\@tuples
  133. } split(/,/, $range) );
  134. }
  135. return _serve("www/$path", $start, $streaming, \@ranges, $last_fetch, $deflate) if -f "www/$path";
  136. return _serve("totp/$path", $start, $streaming, \@ranges, $last_fetch, $deflate) if -f "totp/$path" && $active_user;
  137. #Handle regex/capture routes
  138. if (!exists $routes{$path}) {
  139. my @captures;
  140. foreach my $pattern (keys(%routes)) {
  141. @captures = $path =~ m/^$pattern$/;
  142. if (@captures) {
  143. $path = $pattern;
  144. foreach my $field (@{$routes{$path}{captures}}) {
  145. $routes{$path}{data} //= {};
  146. $routes{$path}{data}{$field} = shift @captures;
  147. }
  148. last;
  149. }
  150. }
  151. }
  152. $query->{deflate} = $deflate;
  153. $query->{user} = $active_user;
  154. return _forbidden($query) if $routes{$path}{auth} && !$active_user;
  155. return _notfound($query) unless exists $routes{$path};
  156. return _badrequest($query) unless grep { $env->{REQUEST_METHOD} eq $_ } ($routes{$path}{method} || '','HEAD');
  157. @{$query}{keys(%{$routes{$path}{'data'}})} = values(%{$routes{$path}{'data'}}) if ref $routes{$path}{'data'} eq 'HASH' && %{$routes{$path}{'data'}};
  158. #Set various things we don't want overridden
  159. $query->{body} = '';
  160. $query->{dnt} = $env->{HTTP_DNT};
  161. $query->{user} = $active_user;
  162. $query->{domain} = $env->{HTTP_X_FORWARDED_HOST} || $env->{HTTP_HOST};
  163. $query->{route} = $path;
  164. $query->{scheme} = $env->{'psgi.url_scheme'} // 'http';
  165. $query->{social_meta} = 1;
  166. $query->{primary_post} = {};
  167. $query->{has_query} = $has_query;
  168. #XXX there is a trick to now use strict refs, but I don't remember it right at the moment
  169. {
  170. no strict 'refs';
  171. my $output = $routes{$path}{callback}->($query);
  172. # Append server-timing headers
  173. my $tot = tv_interval($start) * 1000;
  174. push(@{$output->[1]}, 'Server-Timing' => "app;dur=$tot");
  175. return $output;
  176. }
  177. };
  178. sub _generic($type, $query) {
  179. return _static("$type.z",$query->{start}, $query->{streaming}) if -f "www/statics/$type.z";
  180. return _static($type, $query->{start}, $query->{streaming}) if -f "www/statics/$type";
  181. my %lookup = (
  182. notfound => \&Trog::Routes::HTML::notfound,
  183. forbidden => \&Trog::Routes::HTML::forbidden,
  184. badrequest => \&Trog::Routes::HTML::badrequest,
  185. toolong => \&Trog::Routes::HTML::toolong,
  186. );
  187. return $lookup{$type}->($query);
  188. }
  189. sub _notfound ($query) {
  190. return _generic('notfound', $query);
  191. }
  192. sub _forbidden($query) {
  193. return _generic('forbidden', $query);
  194. }
  195. sub _badrequest($query) {
  196. return _generic('badrequest', $query);
  197. }
  198. sub _toolong() {
  199. return _generic('toolong', {});
  200. }
  201. sub _static($path,$start,$streaming,$last_fetch=0) {
  202. # XXX because of psgi I can't just vomit the file directly
  203. if (open(my $fh, '<', "www/statics/$path")) {
  204. my $headers = '';
  205. # NOTE: this is relying on while advancing the file pointer
  206. while (<$fh>) {
  207. last if $_ eq "\n";
  208. $headers .= $_;
  209. }
  210. my(undef, undef, $status, undef, $headers_parsed) = HTTP::Parser::XS::parse_http_response("$headers\n", HEADERS_AS_HASHREF);
  211. #XXX need to put this into the file itself
  212. my $mt = (stat($fh))[9];
  213. my @gm = gmtime($mt);
  214. my $now_string = strftime( "%a, %d %b %Y %H:%M:%S GMT", @gm );
  215. my $code = $mt > $last_fetch ? $status : 304;
  216. $headers_parsed->{"Last-Modified"} = $now_string;
  217. # Append server-timing headers
  218. my $tot = tv_interval($start) * 1000;
  219. $headers_parsed->{'Server-Timing'} = "static;dur=$tot";
  220. #XXX uwsgi just opens the file *again* when we already have a filehandle if it has a path.
  221. # starman by comparison doesn't violate the principle of least astonishment here.
  222. # This is probably a performance optimization, but makes the kind of micromanagement I need to do inconvenient.
  223. # As such, we will just return a stream.
  224. return sub {
  225. my $responder = shift;
  226. #push(@headers, 'Content-Length' => $sz);
  227. my $writer = $responder->([ $code, [%$headers_parsed]]);
  228. while ( $fh->read( my $buf, $CHUNK_SIZE) ) {
  229. $writer->write($buf);
  230. }
  231. close $fh;
  232. $writer->close;
  233. } if $streaming;
  234. return [$code, [%$headers_parsed], $fh];
  235. }
  236. return [ 403, ['Content-Type' => $Trog::Vars::content_types{plain}], ["STAY OUT YOU RED MENACE"]];
  237. }
  238. sub _range ($fh, $ranges, $sz, %headers) {
  239. # Set mode
  240. my $primary_ct = "Content-Type: $headers{'Content-type'}";
  241. my $is_multipart = scalar(@$ranges) > 1;
  242. if ( $is_multipart ) {
  243. $headers{'Content-type'} = "multipart/byteranges; boundary=$CHUNK_SEP";
  244. }
  245. my $code = 206;
  246. my $fc = '';
  247. # Calculate the content-length up-front. We have to fix unspecified lengths first, and reject bad requests.
  248. foreach my $range (@$ranges) {
  249. $range->[1] //= $sz-1;
  250. return [416, [%headers], ["Requested range not satisfiable"]] if $range->[0] > $sz || $range->[0] < 0 || $range->[1] < 0 || $range->[0] > $range->[1];
  251. }
  252. $headers{'Content-Length'} = List::Util::sum(map { my $arr=$_; $arr->[1]+1, -$arr->[0] } @$ranges);
  253. #XXX Add the entity header lengths to the value - should hash-ify this to DRY
  254. if ($is_multipart) {
  255. foreach my $range (@$ranges) {
  256. $headers{'Content-Length'} += length("$fc--$CHUNK_SEP\n$primary_ct\nContent-Range: bytes $range->[0]-$range->[1]/$sz\n\n" );
  257. $fc = "\n";
  258. }
  259. $headers{'Content-Length'} += length( "\n--$CHUNK_SEP\--\n" );
  260. $fc = '';
  261. }
  262. return sub {
  263. my $responder = shift;
  264. my $writer;
  265. foreach my $range (@$ranges) {
  266. $headers{'Content-Range'} = "bytes $range->[0]-$range->[1]/$sz" unless $is_multipart;
  267. $writer //= $responder->([ $code, [%headers]]);
  268. $writer->write( "$fc--$CHUNK_SEP\n$primary_ct\nContent-Range: bytes $range->[0]-$range->[1]/$sz\n\n" ) if $is_multipart;
  269. $fc = "\n";
  270. my $len = List::Util::min($sz,$range->[1]+1) - $range->[0];
  271. $fh->seek( $range->[0], 0);
  272. while ($len) {
  273. $fh->read(my $buf, List::Util::min($len,$CHUNK_SIZE) );
  274. $writer->write($buf);
  275. # Adjust for amount written
  276. $len = List::Util::max($len - $CHUNK_SIZE, 0);
  277. }
  278. }
  279. $fh->close();
  280. $writer->write( "\n--$CHUNK_SEP\--\n" ) if $is_multipart;
  281. $writer->close;
  282. };
  283. }
  284. sub _serve ($path, $start, $streaming, $ranges, $last_fetch=0, $deflate=0) {
  285. my $mf = Mojo::File->new($path);
  286. my $ext = '.'.$mf->extname();
  287. my $ft;
  288. if ($ext) {
  289. $ft = Plack::MIME->mime_type($ext) if $ext;
  290. $ft ||= $extra_types{$ext} if exists $extra_types{$ext};
  291. }
  292. $ft ||= $Trog::Vars::content_types{plain};
  293. my $ct = 'Content-type';
  294. my @headers = ($ct => $ft);
  295. #TODO use static Cache-Control for everything but JS/CSS?
  296. push(@headers,'Cache-control' => $Trog::Vars::cache_control{revalidate});
  297. push(@headers,'Accept-Ranges' => 'bytes');
  298. my $mt = (stat($path))[9];
  299. my $sz = (stat(_))[7];
  300. my @gm = gmtime($mt);
  301. my $now_string = strftime( "%a, %d %b %Y %H:%M:%S GMT", @gm );
  302. my $code = $mt > $last_fetch ? 200 : 304;
  303. push(@headers, "Last-Modified" => $now_string);
  304. push(@headers, 'Vary' => 'Accept-Encoding');
  305. if (open(my $fh, '<', $path)) {
  306. return _range($fh, $ranges, $sz, @headers) if @$ranges && $streaming;
  307. # Transfer-encoding: chunked
  308. return sub {
  309. my $responder = shift;
  310. push(@headers, 'Content-Length' => $sz);
  311. my $writer = $responder->([ $code, \@headers]);
  312. while ( $fh->read( my $buf, $CHUNK_SIZE) ) {
  313. $writer->write($buf);
  314. }
  315. close $fh;
  316. $writer->close;
  317. } if $streaming && $sz > $CHUNK_SIZE;
  318. #Return data in the event the caller does not support deflate
  319. if (!$deflate) {
  320. push( @headers, "Content-Length" => $sz );
  321. # Append server-timing headers
  322. my $tot = tv_interval($start) * 1000;
  323. push(@headers, 'Server-Timing' => "file;dur=$tot");
  324. return [ $code, \@headers, $fh];
  325. }
  326. #Compress everything less than 1MB
  327. push( @headers, "Content-Encoding" => "gzip" );
  328. my $dfh;
  329. IO::Compress::Gzip::gzip( $fh => \$dfh );
  330. print $IO::Compress::Gzip::GzipError if $IO::Compress::Gzip::GzipError;
  331. push( @headers, "Content-Length" => length($dfh) );
  332. # Append server-timing headers
  333. my $tot = tv_interval($start) * 1000;
  334. push(@headers, 'Server-Timing' => "file;dur=$tot");
  335. return [ $code, \@headers, [$dfh]];
  336. }
  337. return [ 403, [$ct => $Trog::Vars::content_types{plain}], ["STAY OUT YOU RED MENACE"]];
  338. }
  339. 1;