TCMS.pm 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325
  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 Sys::Hostname();
  8. use HTTP::Body ();
  9. use URL::Encode ();
  10. use Text::Xslate ();
  11. use Plack::MIME ();
  12. use Mojo::File ();
  13. use DateTime::Format::HTTP();
  14. use CGI::Cookie ();
  15. use File::Basename();
  16. use IO::Compress::Gzip();
  17. use Time::HiRes qw{gettimeofday tv_interval};
  18. use HTTP::Parser::XS qw{HEADERS_AS_HASHREF};
  19. use List::Util;
  20. use URI();
  21. #Grab our custom routes
  22. use FindBin::libs;
  23. use Trog::Routes::HTML;
  24. use Trog::Routes::JSON;
  25. use Trog::Log qw{:all};
  26. use Trog::Auth;
  27. use Trog::Utils;
  28. use Trog::Config;
  29. use Trog::Data;
  30. use Trog::Vars;
  31. use Trog::FileHandler;
  32. # Troglodyne philosophy - simple as possible
  33. # Import the routes
  34. my $conf = Trog::Config::get();
  35. my $data = Trog::Data->new($conf);
  36. my %roots = $data->routes();
  37. my %routes = %Trog::Routes::HTML::routes;
  38. @routes{ keys(%Trog::Routes::JSON::routes) } = values(%Trog::Routes::JSON::routes);
  39. @routes{ keys(%roots) } = values(%roots);
  40. my %aliases = $data->aliases();
  41. # XXX this is built progressively across the forks, leading to inconsistent behavior.
  42. # This should eventually be pre-filled from DB.
  43. my %etags;
  44. =head2 app()
  45. Dispatches requests based on %routes built above.
  46. 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.
  47. 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.
  48. =cut
  49. sub app {
  50. # Start the server timing clock
  51. my $start = [gettimeofday];
  52. my $env = shift;
  53. return _toolong() if length( $env->{REQUEST_URI} ) > 2048;
  54. my $requestid = Trog::Utils::uuid();
  55. Trog::Log::uuid($requestid);
  56. # Check eTags. If we don't know about it, just assume it's good and lazily fill the cache
  57. # XXX yes, this allows cache poisoning...but only for logged in users!
  58. if ( $env->{HTTP_IF_NONE_MATCH} ) {
  59. return [ 304, [], [''] ] if $env->{HTTP_IF_NONE_MATCH} eq ( $etags{ $env->{REQUEST_URI} } || '' );
  60. $etags{ $env->{REQUEST_URI} } = $env->{HTTP_IF_NONE_MATCH} unless exists $etags{ $env->{REQUEST_URI} };
  61. }
  62. my $last_fetch = 0;
  63. if ( $env->{HTTP_IF_MODIFIED_SINCE} ) {
  64. $last_fetch = DateTime::Format::HTTP->parse_datetime( $env->{HTTP_IF_MODIFIED_SINCE} )->epoch();
  65. }
  66. #XXX Don't use statics anything that has a search query
  67. # On one hand, I don't want to DOS the disk, but I'd also like some like ?rss...
  68. # Should probably turn those into aliases.
  69. my $has_query = !!$env->{QUERY_STRING};
  70. my $query = {};
  71. $query = URL::Encode::url_params_mixed( $env->{QUERY_STRING} ) if $env->{QUERY_STRING};
  72. #Actually parse the POSTDATA and dump it into the QUERY object if this is a POST
  73. if ( $env->{REQUEST_METHOD} eq 'POST' ) {
  74. my $body = HTTP::Body->new( $env->{CONTENT_TYPE}, $env->{CONTENT_LENGTH} );
  75. while ( $env->{'psgi.input'}->read( my $buf, $Trog::Vars::CHUNK_SIZE ) ) {
  76. $body->add($buf);
  77. }
  78. @$query{ keys( %{ $body->param } ) } = values( %{ $body->param } );
  79. @$query{ keys( %{ $body->upload } ) } = values( %{ $body->upload } );
  80. }
  81. # Grab the list of ACLs we want to add to a post, if any.
  82. $query->{acls} = [ $query->{acls} ] if ( $query->{acls} && ref $query->{acls} ne 'ARRAY' );
  83. my $path = $env->{PATH_INFO};
  84. $path = '/index' if $path eq '/';
  85. #XXX this is hardcoded in browsers, so just rewrite the path
  86. $path = '/img/icon/favicon.ico' if $path eq '/favicon.ico';
  87. # Translate alias paths into their actual path
  88. $path = $aliases{$path} if exists $aliases{$path};
  89. # Figure out if we want compression or not
  90. my $alist = $env->{HTTP_ACCEPT_ENCODING} || '';
  91. $alist =~ s/\s//g;
  92. my @accept_encodings;
  93. @accept_encodings = split( /,/, $alist );
  94. my $deflate = grep { 'gzip' eq $_ } @accept_encodings;
  95. # Collapse multiple slashes in the path
  96. $path =~ s/[\/]+/\//g;
  97. # Let's open up our default route before we bother to see if users even exist
  98. return $routes{default}{callback}->($query) unless -f "config/setup";
  99. my $cookies = {};
  100. if ( $env->{HTTP_COOKIE} ) {
  101. $cookies = CGI::Cookie->parse( $env->{HTTP_COOKIE} );
  102. }
  103. # Set the IP of the request so we can fail2ban
  104. $Trog::Log::ip = $env->{HTTP_X_FORWARDED_FOR} || $env->{REMOTE_ADDR};
  105. my $active_user = '';
  106. $Trog::Log::user = 'nobody';
  107. if ( exists $cookies->{tcmslogin} ) {
  108. $active_user = Trog::Auth::session2user( $cookies->{tcmslogin}->value );
  109. $Trog::Log::user = $active_user if $active_user;
  110. }
  111. $query->{user_acls} = [];
  112. $query->{user_acls} = Trog::Auth::acls4user($active_user) // [] if $active_user;
  113. # Filter out passed ACLs which are naughty
  114. my $is_admin = grep { $_ eq 'admin' } @{ $query->{user_acls} };
  115. @{ $query->{acls} } = grep { $_ ne 'admin' } @{ $query->{acls} } unless $is_admin;
  116. # Ensure any short-circuit routes can log the request
  117. $query->{method} = $env->{REQUEST_METHOD};
  118. $query->{route} = $path;
  119. # Disallow any paths that are naughty ( starman auto-removes .. up-traversal)
  120. if ( index( $path, '/templates' ) == 0 || index( $path, '/statics' ) == 0 || $path =~ m/.*(\.psgi|\.pm)$/i ) {
  121. return _forbidden($query);
  122. }
  123. my $streaming = $env->{'psgi.streaming'};
  124. $query->{streaming} = $streaming;
  125. # If we have a static render, just use it instead (These will ALWAYS be correct, data saves invalidate this)
  126. # TODO: make this key on admin INSTEAD of active user when we add non-admin users.
  127. $query->{start} = $start;
  128. if ( !$active_user && !$has_query ) {
  129. return _static( "$path.z", $start, $streaming ) if -f "www/statics/$path.z" && $deflate;
  130. return _static( $path, $start, $streaming ) if -f "www/statics/$path";
  131. }
  132. # Handle HTTP range/streaming requests
  133. my $range = $env->{HTTP_RANGE} || "bytes=0-" if $env->{HTTP_RANGE} || $env->{HTTP_IF_RANGE};
  134. my @ranges;
  135. if ($range) {
  136. $range =~ s/bytes=//g;
  137. push(
  138. @ranges,
  139. map {
  140. [ split( /-/, $_ ) ];
  141. #$tuples[1] //= $tuples[0] + $Trog::Vars::CHUNK_SIZE;
  142. #\@tuples
  143. } split( /,/, $range )
  144. );
  145. }
  146. return Trog::FileHandler::serve( "www/$path", $start, $streaming, \@ranges, $last_fetch, $deflate ) if -f "www/$path";
  147. return Trog::FileHandler::serve( "totp/$path", $start, $streaming, \@ranges, $last_fetch, $deflate ) if -f "totp/$path" && $active_user;
  148. #Handle regex/capture routes
  149. if ( !exists $routes{$path} ) {
  150. my @captures;
  151. foreach my $pattern ( keys(%routes) ) {
  152. @captures = $path =~ m/^$pattern$/;
  153. if (@captures) {
  154. $path = $pattern;
  155. foreach my $field ( @{ $routes{$path}{captures} } ) {
  156. $routes{$path}{data} //= {};
  157. $routes{$path}{data}{$field} = shift @captures;
  158. }
  159. last;
  160. }
  161. }
  162. }
  163. $query->{deflate} = $deflate;
  164. $query->{user} = $active_user;
  165. return _forbidden($query) if exists $routes{$path}{auth} && !$active_user;
  166. return _notfound($query) unless $routes{$path} && ref $routes{$path} eq 'HASH' && keys(%{$routes{$path}});
  167. return _badrequest($query) unless grep { $env->{REQUEST_METHOD} eq $_ } ( $routes{$path}{method} || '', 'HEAD' );
  168. @{$query}{ keys( %{ $routes{$path}{'data'} } ) } = values( %{ $routes{$path}{'data'} } ) if ref $routes{$path}{'data'} eq 'HASH' && %{ $routes{$path}{'data'} };
  169. #Set various things we don't want overridden
  170. $query->{body} = '';
  171. $query->{dnt} = $env->{HTTP_DNT};
  172. $query->{user} = $active_user;
  173. $query->{domain} = eval { Sys::Hostname::hostname() } // $env->{HTTP_X_FORWARDED_HOST} || $env->{HTTP_HOST};
  174. $query->{route} = $path;
  175. $query->{scheme} = $env->{'psgi.url_scheme'} // 'http';
  176. $query->{social_meta} = 1;
  177. $query->{primary_post} = {};
  178. $query->{has_query} = $has_query;
  179. $query->{port} = $env->{HTTP_X_FORWARDED_PORT} // $env->{HTTP_PORT};
  180. # Redirecting somewhere naughty not allow
  181. $query->{to} = URI->new($query->{to} // '')->path() || $query->{to} if $query->{to};
  182. #XXX there is a trick to now use strict refs, but I don't remember it right at the moment
  183. {
  184. no strict 'refs';
  185. my $output = $routes{$path}{callback}->($query);
  186. die "$path returned no data!" unless ref $output eq 'ARRAY' && @$output == 3;
  187. INFO("$env->{REQUEST_METHOD} $output->[0] $path");
  188. # Append server-timing headers
  189. my $tot = tv_interval($start) * 1000;
  190. push( @{ $output->[1] }, 'Server-Timing' => "app;dur=$tot" );
  191. return $output;
  192. }
  193. }
  194. sub _generic ( $type, $query ) {
  195. return _static( "$type.z", $query->{start}, $query->{streaming} ) if -f "www/statics/$type.z";
  196. return _static( $type, $query->{start}, $query->{streaming} ) if -f "www/statics/$type";
  197. my %lookup = (
  198. notfound => \&Trog::Routes::HTML::notfound,
  199. forbidden => \&Trog::Routes::HTML::forbidden,
  200. badrequest => \&Trog::Routes::HTML::badrequest,
  201. toolong => \&Trog::Routes::HTML::toolong,
  202. );
  203. return $lookup{$type}->($query);
  204. }
  205. sub _notfound ($query) {
  206. return _generic( 'notfound', $query );
  207. }
  208. sub _forbidden ($query) {
  209. return _generic( 'forbidden', $query );
  210. }
  211. sub _badrequest ($query) {
  212. return _generic( 'badrequest', $query );
  213. }
  214. sub _toolong() {
  215. return _generic( 'toolong', {} );
  216. }
  217. sub _static ( $path, $start, $streaming, $last_fetch = 0 ) {
  218. DEBUG("Rendering static for $path");
  219. # XXX because of psgi I can't just vomit the file directly
  220. if ( open( my $fh, '<', "www/statics/$path" ) ) {
  221. my $headers = '';
  222. # NOTE: this is relying on while advancing the file pointer
  223. while (<$fh>) {
  224. last if $_ eq "\n";
  225. $headers .= $_;
  226. }
  227. my ( undef, undef, $status, undef, $headers_parsed ) = HTTP::Parser::XS::parse_http_response( "$headers\n", HEADERS_AS_HASHREF );
  228. #XXX need to put this into the file itself
  229. my $mt = ( stat($fh) )[9];
  230. my @gm = gmtime($mt);
  231. my $now_string = strftime( "%a, %d %b %Y %H:%M:%S GMT", @gm );
  232. my $code = $mt > $last_fetch ? $status : 304;
  233. $headers_parsed->{"Last-Modified"} = $now_string;
  234. # Append server-timing headers
  235. my $tot = tv_interval($start) * 1000;
  236. $headers_parsed->{'Server-Timing'} = "static;dur=$tot";
  237. #XXX uwsgi just opens the file *again* when we already have a filehandle if it has a path.
  238. # starman by comparison doesn't violate the principle of least astonishment here.
  239. # This is probably a performance optimization, but makes the kind of micromanagement I need to do inconvenient.
  240. # As such, we will just return a stream.
  241. return sub {
  242. my $responder = shift;
  243. #push(@headers, 'Content-Length' => $sz);
  244. my $writer = $responder->( [ $code, [%$headers_parsed] ] );
  245. while ( $fh->read( my $buf, $Trog::Vars::CHUNK_SIZE ) ) {
  246. $writer->write($buf);
  247. }
  248. close $fh;
  249. $writer->close;
  250. }
  251. if $streaming;
  252. return [ $code, [%$headers_parsed], $fh ];
  253. }
  254. return [ 403, [ 'Content-Type' => $Trog::Vars::content_types{text} ], ["STAY OUT YOU RED MENACE"] ];
  255. }
  256. 1;