1 const char filters_rcs[] = "$Id: filters.c,v 1.115 2009/04/17 11:29:18 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `acl_addr', `add_stats', `block_acl', `block_imageurl',
9 * `block_url', `url_actions', `domain_split',
10 * `filter_popups', `forward_url', 'redirect_url',
11 * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
12 * `ijb_send_banner', `trust_url', `gif_deanimate_response',
13 * `execute_single_pcrs_command', `rewrite_url',
16 * Copyright : Written by and Copyright (C) 2001, 2004-2008 the SourceForge
17 * Privoxy team. http://www.privoxy.org/
19 * Based on the Internet Junkbuster originally written
20 * by and Copyright (C) 1997 Anonymous Coders and
21 * Junkbusters Corporation. http://www.junkbusters.com
23 * This program is free software; you can redistribute it
24 * and/or modify it under the terms of the GNU General
25 * Public License as published by the Free Software
26 * Foundation; either version 2 of the License, or (at
27 * your option) any later version.
29 * This program is distributed in the hope that it will
30 * be useful, but WITHOUT ANY WARRANTY; without even the
31 * implied warranty of MERCHANTABILITY or FITNESS FOR A
32 * PARTICULAR PURPOSE. See the GNU General Public
33 * License for more details.
35 * The GNU General Public License should be included with
36 * this file. If not, you can view it at
37 * http://www.gnu.org/copyleft/gpl.html
38 * or write to the Free Software Foundation, Inc., 59
39 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
43 * Revision 1.115 2009/04/17 11:29:18 fabiankeil
44 * Compile fix for BSD libc.
46 * Revision 1.114 2009/04/17 11:27:49 fabiankeil
47 * Petr Pisar's privoxy-3.0.12-ipv6-3.diff.
49 * Revision 1.113 2009/03/08 14:19:23 fabiankeil
50 * Fix justified (but harmless) compiler warnings
51 * on platforms where sizeof(int) < sizeof(long).
53 * Revision 1.112 2009/03/01 18:28:23 fabiankeil
54 * Help clang understand that we aren't dereferencing
57 * Revision 1.111 2008/12/04 18:13:46 fabiankeil
58 * Fix a cparser warning.
60 * Revision 1.110 2008/11/10 16:40:25 fabiankeil
61 * Fix a gcc44 warning.
63 * Revision 1.109 2008/11/08 15:48:41 fabiankeil
64 * Mention actual values when complaining about
65 * the chunk size exceeding the buffer size.
67 * Revision 1.108 2008/05/21 15:35:08 fabiankeil
68 * - Mark csp as immutable for block_acl().
69 * - Remove an obsolete complaint about filter_popups().
71 * Revision 1.107 2008/05/04 17:52:56 fabiankeil
72 * Adjust parse_http_url() call to new prototype.
74 * Revision 1.106 2008/05/03 16:40:44 fabiankeil
75 * Change content_filters_enabled()'s parameter from
76 * csp->action to action so it can be also used in the
77 * CGI code. Don't bother checking if there are filters
78 * loaded, as that's somewhat besides the point.
80 * Revision 1.105 2008/03/28 15:13:39 fabiankeil
81 * Remove inspect-jpegs action.
83 * Revision 1.104 2008/03/27 18:27:24 fabiankeil
84 * Remove kill-popups action.
86 * Revision 1.103 2008/03/06 16:33:45 fabiankeil
87 * If limit-connect isn't used, don't limit CONNECT requests to port 443.
89 * Revision 1.102 2008/03/01 14:00:44 fabiankeil
90 * Let the block action take the reason for the block
91 * as argument and show it on the "blocked" page.
93 * Revision 1.101 2008/02/23 16:57:12 fabiankeil
94 * Rename url_actions() to get_url_actions() and let it
95 * use the standard parameter ordering.
97 * Revision 1.100 2008/02/23 16:33:43 fabiankeil
98 * Let forward_url() use the standard parameter ordering
99 * and mark its second parameter immutable.
101 * Revision 1.99 2008/02/03 13:57:58 fabiankeil
102 * Add SOCKS5 support for forward-override{}.
104 * Revision 1.98 2008/01/04 17:43:45 fabiankeil
105 * Improve the warning messages that get logged if the action files
106 * "enable" filters but no filters of that type have been loaded.
108 * Revision 1.97 2007/11/30 15:37:03 fabiankeil
109 * Use freez instead of free.
111 * Revision 1.96 2007/10/19 16:53:28 fabiankeil
112 * Add helper function to check if any content filters are enabled.
114 * Revision 1.95 2007/10/17 19:31:20 fabiankeil
115 * Omitting the zero chunk that ends the chunk transfer encoding seems
116 * to be the new black. Log the problem and continue filtering anyway.
118 * Revision 1.94 2007/09/29 13:20:20 fabiankeil
119 * Remove two redundant and one useless log messages.
121 * Revision 1.93 2007/09/29 10:21:16 fabiankeil
122 * - Move get_filter_function() from jcc.c to filters.c
123 * so the filter functions can be static.
124 * - Don't bother filtering body-less responses.
126 * Revision 1.92 2007/09/28 16:38:55 fabiankeil
127 * - Execute content filters through execute_content_filter().
128 * - Add prepare_for_filtering() so filter functions don't have to
129 * care about de-chunking and decompression. As a side effect this enables
130 * decompression for gif_deanimate_response() and jpeg_inspect_response().
131 * - Change remove_chunked_transfer_coding()'s return type to jb_err.
132 * Some clowns feel like chunking empty responses in which case
133 * (size == 0) is valid but previously would be interpreted as error.
135 * Revision 1.91 2007/09/02 15:31:20 fabiankeil
136 * Move match_portlist() from filter.c to urlmatch.c.
137 * It's used for url matching, not for filtering.
139 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
140 * Remove newline at the end of a log_error() message.
142 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
143 * #1763173 from Stefan Huehner: declare some more functions static.
145 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
146 * Add forward-override{} to change the forwarding settings through
147 * action sections. This is mainly interesting to forward different
148 * clients differently (for example based on User-Agent or request
151 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
152 * Make sure filters with dynamic jobs actually use them.
154 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
155 * - Introduce dynamic pcrs jobs that can resolve variables.
156 * - Don't run redirect functions more than once,
157 * unless they are activated more than once.
159 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
160 * - Log the content size after decompression in decompress_iob()
161 * instead of pcrs_filter_response().
163 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
164 * Use dedicated header filter actions instead of abusing "filter".
165 * Replace "filter-client-headers" and "filter-client-headers"
166 * with "server-header-filter" and "client-header-filter".
168 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
169 * New config option: enforce-blocks.
171 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
172 * - Fix port handling in acl_addr() and use a temporary acl spec
173 * copy so error messages don't contain a truncated version.
174 * - Log size of iob before and after decompression.
176 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
177 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
178 * - Hide the "Go there anyway" link for blocked CONNECT
179 * requests where going there anyway doesn't work anyway.
181 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
182 * - Save the reason for generating http_responses.
183 * - Block (+block) with status code 403 instead of 404.
184 * - Use a different kludge to remember a failed decompression.
186 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
187 * Search for Max-Forwards headers case-insensitive,
188 * don't generate the "501 unsupported" message for invalid
189 * Max-Forwards values and don't increase negative ones.
191 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
192 * - Add HEAD support to finish_http_response.
193 * - Add error favicon to internal HTML error messages.
195 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
196 * Mark *csp as immutable for is_untrusted_url()
197 * and is_imageurl(). Closes FR 1237736.
199 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
200 * Integrate a modified version of Wil Mahan's
201 * zlib patch (PR #895531).
203 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
204 * Fixed gcc43 conversion warnings,
205 * changed sprintf calls to snprintf.
207 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
208 * Adjust comment in pcrs_filter_response()
209 * to recent pcrs changes. Hohoho.
211 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
212 * Don't crash if pcre returns an error code
213 * that pcrs didn't expect. Fixes BR 1621173.
215 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
216 * Modified is_untrusted_url to complain in case of
217 * write errors and to give a reason when adding new
218 * entries to the trustfile. Closes FR 1097611.
220 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
221 * Skip empty filter files in pcrs_filter_response,
222 * but don't ignore the ones that come afterwards.
223 * Fixes parts of BR 1619208.
225 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
226 * Added some sanity checks for get_last_url().
227 * Fixed possible segfault caused by my last commit.
229 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
230 * Let get_last_url() catch https URLs as well.
232 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
233 * Make sure get_last_url() behaves like advertised
234 * and fast-redirects{} can be combined with redirect{}.
236 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
237 * Implemented +redirect{s@foo@bar@} to generate
238 * a redirect based on a rewritten version of the
241 * Revision 1.66 2006/09/23 13:26:38 roro
242 * Replace TABs by spaces in source code.
244 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
245 * Fix +redirect{}. Didn't work with -fast-redirects.
247 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
248 * Block requests for untrusted URLs with status
249 * code 403 instead of 200.
251 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
252 * Don't free p which is still in use and will be later
253 * freed by free_map(). Don't claim the referrer is unknown
254 * when the client didn't set one.
256 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
257 * Feature request 595948: Re-Filter logging in single line
259 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
260 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
262 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
263 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
264 * with what was really the latest development (the v_3_0_branch branch)
266 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
267 * Multiple filter file support
269 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
270 * A few memory leaks plugged (mostly on error paths)
272 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
273 * Add the ability to check jpeg images for invalid
274 * lengths of comment blocks. Defensive strategy
275 * against the exploit:
276 * Microsoft Security Bulletin MS04-028
277 * Buffer Overrun in JPEG Processing (GDI+) Could
278 * Allow Code Execution (833987)
279 * Enabled with +inspect-jpegs in actions files.
281 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
282 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
284 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
285 * Fixed bug #839859: "See why" link URL now gets url-encoded.
287 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
290 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
291 * Added basic support for OPTIONS and TRACE HTTP methods:
292 * New function direct_response which handles OPTIONS and
293 * TRACE requests whose Max-Forwards header field is zero.
295 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
296 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
298 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
299 * - Bugfix: Executing a filters without jobs no longer results in
300 * turing off *all* filters.
301 * - Security fix: Malicious web servers can't cause a seg fault
302 * through bogus chunk sizes anymore
304 * Revision 1.58 2002/04/24 02:11:17 oes
305 * Jon's multiple AF patch: url_actions now evaluates rules
308 * Revision 1.57 2002/04/08 20:38:34 swa
311 * Revision 1.56 2002/04/05 15:51:24 oes
312 * - bugfix: error-pages now get correct request protocol
313 * - fix for invalid HTML in trust info
315 * Revision 1.55 2002/04/02 16:13:51 oes
316 * Fix: No "Go there anyway" for SSL
318 * Revision 1.54 2002/04/02 14:55:56 oes
319 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
321 * Revision 1.53 2002/03/26 22:29:54 swa
322 * we have a new homepage!
324 * Revision 1.52 2002/03/24 16:35:57 jongfoster
327 * Revision 1.51 2002/03/24 15:23:33 jongfoster
330 * Revision 1.50 2002/03/24 13:25:43 swa
331 * name change related issues
333 * Revision 1.49 2002/03/16 20:29:14 oes
336 * Revision 1.48 2002/03/13 20:25:34 oes
337 * Better logging for content filters
339 * Revision 1.47 2002/03/13 00:30:52 jongfoster
341 * Added option of always sending redirect for imageblock,
342 * currently disabled with #if 0.
344 * Revision 1.46 2002/03/12 01:42:49 oes
345 * Introduced modular filters
347 * Revision 1.45 2002/03/08 16:47:50 oes
348 * Added choice beween GIF and PNG built-in images
350 * Revision 1.44 2002/03/07 03:49:31 oes
351 * - Fixed compiler warnings etc
352 * - Changed built-in images from GIF to PNG
353 * (with regard to Unisys patent issue)
354 * - Added a 4x4 pattern PNG which is less intrusive
355 * than the logo but also clearly marks the deleted banners
357 * Revision 1.43 2002/01/22 23:51:59 jongfoster
358 * Replacing strsav() with the safer string_append().
360 * Adding missing html_encode() to error message generators. Where encoded
361 * and unencoded versions of a string were provided, removing the unencoded
364 * Revision 1.42 2002/01/17 21:00:32 jongfoster
365 * Moving all our URL and URL pattern parsing code to urlmatch.c.
367 * Using a single, simple url_match(pattern,url) function - rather than
368 * the 3-line match routine which was repeated all over the place.
370 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
372 * Using parse_http_url() to parse URLs without faking a HTTP
373 * request line for parse_http_request().
375 * Revision 1.41 2001/11/13 00:14:07 jongfoster
376 * Fixing stupid bug now I've figured out what || means.
377 * (It always returns 0 or 1, not one of it's paramaters.)
379 * Revision 1.40 2001/10/26 17:37:55 oes
380 * - Re-enabled Netscape 200/404 bug workaround in block_url():
381 * - Removed OS/2 special case
382 * - Made block_url() independant from sed() having been run
383 * - Made trust_url independant from sed() having been run
384 * - Made is_imageurl independant from sed() having been run.
385 * It now checks User-Agent: and Accept: by itself.
388 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
389 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
390 * threads to call select() simultaneously. So, it's time to do a real, live,
391 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
392 * (native). Both versions will work, but using __OS2__ offers multi-threading.
394 * Revision 1.38 2001/10/23 21:32:33 jongfoster
395 * Adding error-checking to selected functions
397 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
398 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
399 * filters.c. Added a FIXME in front of the offending code. I'll gladly
400 * put in a better/more robust fix for all parties if one is presented...
401 * It seems that just returning 200 instead of 404 would pretty much fix
402 * it for everyone, but I don't know all the history of the problem.
404 * Revision 1.36 2001/10/10 16:44:16 oes
405 * Added match_portlist function
407 * Revision 1.35 2001/10/07 15:41:23 oes
408 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
410 * New function remove_chunked_transfer_coding that strips chunked
411 * transfer coding to plain and is called by pcrs_filter_response
412 * and gif_deanimate_response if neccessary
414 * Improved handling of zero-change re_filter runs
416 * pcrs_filter_response and gif_deanimate_response now remove
417 * chunked transfer codeing before processing the body.
419 * Revision 1.34 2001/09/20 15:49:36 steudten
421 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
422 * See cgi.c fill_template().
424 * Revision 1.33 2001/09/16 17:05:14 jongfoster
425 * Removing unused #include showarg.h
427 * Revision 1.32 2001/09/16 13:21:27 jongfoster
428 * Changes to use new list functions.
430 * Revision 1.31 2001/09/16 11:38:02 jongfoster
431 * Splitting fill_template() into 2 functions:
432 * template_load() loads the file
433 * template_fill() performs the PCRS regexps.
434 * This is because the CGI edit interface has a "table row"
435 * template which is used many times in the page - this
436 * change means it's only loaded from disk once.
438 * Revision 1.30 2001/09/16 11:00:10 jongfoster
439 * New function alloc_http_response, for symmetry with free_http_response
441 * Revision 1.29 2001/09/13 23:32:40 jongfoster
442 * Moving image data to cgi.c rather than cgi.h
443 * Fixing a GPF under Win32 (and any other OS that protects global
444 * constants from being written to).
446 * Revision 1.28 2001/09/10 10:18:51 oes
447 * Silenced compiler warnings
449 * Revision 1.27 2001/08/05 16:06:20 jongfoster
450 * Modifiying "struct map" so that there are now separate header and
451 * "map_entry" structures. This means that functions which modify a
452 * map no longer need to return a pointer to the modified map.
453 * Also, it no longer reverses the order of the entries (which may be
454 * important with some advanced template substitutions).
456 * Revision 1.26 2001/07/30 22:08:36 jongfoster
457 * Tidying up #defines:
458 * - All feature #defines are now of the form FEATURE_xxx
459 * - Permanently turned off WIN_GUI_EDIT
460 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
462 * Revision 1.25 2001/07/26 10:09:46 oes
463 * Made browser detection a little less naive
465 * Revision 1.24 2001/07/25 17:22:51 oes
466 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
468 * Revision 1.23 2001/07/23 13:40:12 oes
469 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
471 * Revision 1.22 2001/07/18 12:29:34 oes
472 * - Made gif_deanimate_response respect
473 * csp->action->string[ACTION_STRING_DEANIMATE]
474 * - Logging cosmetics
476 * Revision 1.21 2001/07/13 13:59:53 oes
477 * - Introduced gif_deanimate_response which shares the
478 * generic content modification interface of pcrs_filter_response
479 * and acts as a wrapper to deanimate.c:gif_deanimate()
480 * - Renamed re_process_buffer to pcrs_filter_response
481 * - pcrs_filter_response now returns NULL on failiure
482 * - Removed all #ifdef PCRS
484 * Revision 1.20 2001/07/01 17:01:04 oes
485 * Added comments and missing return statement in is_untrusted_url()
487 * Revision 1.19 2001/06/29 21:45:41 oes
488 * Indentation, CRLF->LF, Tab-> Space
490 * Revision 1.18 2001/06/29 13:27:38 oes
491 * - Cleaned up, renamed and reorderd functions
492 * and improved comments
495 * - Ported to CGI platform. Now delivers
496 * http_response or NULL
497 * - Unified HTML and GIF generation (moved image detection
498 * and GIF generation here from jcc.c:chat())
499 * - Fixed HTTP status to:
500 * - 403 (Forbidden) for the "blocked" HTML message
501 * - 200 (OK) for GIF answers
502 * - 302 (Redirect) for redirect to GIF
505 * - Ported to CGI platform. Now delivers
506 * http_response or NULL
507 * - Separated detection of untrusted URL into
508 * (bool)is_untrusted_url
509 * - Added enforcement of untrusted requests
511 * - Moved redirect_url() from cgi.c to here
512 * and ported it to the CGI platform
514 * - Removed logentry from cancelled commit
516 * Revision 1.17 2001/06/09 10:55:28 jongfoster
517 * Changing BUFSIZ ==> BUFFER_SIZE
519 * Revision 1.16 2001/06/07 23:10:26 jongfoster
520 * Allowing unanchored domain patterns to back off and retry
521 * if they partially match. Optimized right-anchored patterns.
522 * Moving ACL and forward files into config file.
523 * Replacing struct gateway with struct forward_spec
525 * Revision 1.15 2001/06/03 19:12:00 oes
526 * extracted-CGI relevant stuff
528 * Revision 1.14 2001/06/01 10:30:55 oes
529 * Added optional left-anchoring to domaincmp
531 * Revision 1.13 2001/05/31 21:21:30 jongfoster
532 * Permissionsfile / actions file changes:
533 * - Changed "permission" to "action" throughout
534 * - changes to file format to allow string parameters
535 * - Moved helper functions to actions.c
537 * Revision 1.12 2001/05/31 17:35:20 oes
539 * - Enhanced domain part globbing with infix and prefix asterisk
540 * matching and optional unanchored operation
542 * Revision 1.11 2001/05/29 11:53:23 oes
543 * "See why" link added to "blocked" page
545 * Revision 1.10 2001/05/29 09:50:24 jongfoster
546 * Unified blocklist/imagelist/permissionslist.
547 * File format is still under discussion, but the internal changes
550 * Also modified interceptor behaviour:
551 * - We now intercept all URLs beginning with one of the following
552 * prefixes (and *only* these prefixes):
554 * * http://ijbswa.sf.net/config/
555 * * http://ijbswa.sourceforge.net/config/
556 * - New interceptors "home page" - go to http://i.j.b/ to see it.
557 * - Internal changes so that intercepted and fast redirect pages
558 * are not replaced with an image.
559 * - Interceptors now have the option to send a binary page direct
560 * to the client. (i.e. ijb-send-banner uses this)
561 * - Implemented show-url-info interceptor. (Which is why I needed
562 * the above interceptors changes - a typical URL is
563 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
564 * The previous mechanism would not have intercepted that, and
565 * if it had been intercepted then it then it would have replaced
568 * Revision 1.9 2001/05/27 22:17:04 oes
570 * - re_process_buffer no longer writes the modified buffer
571 * to the client, which was very ugly. It now returns the
572 * buffer, which it is then written by chat.
574 * - content_length now adjusts the Content-Length: header
575 * for modified documents rather than crunch()ing it.
576 * (Length info in csp->content_length, which is 0 for
577 * unmodified documents)
579 * - For this to work, sed() is called twice when filtering.
581 * Revision 1.8 2001/05/26 17:13:28 jongfoster
582 * Filled in a function comment.
584 * Revision 1.7 2001/05/26 15:26:15 jongfoster
585 * ACL feature now provides more security by immediately dropping
586 * connections from untrusted hosts.
588 * Revision 1.6 2001/05/26 00:28:36 jongfoster
589 * Automatic reloading of config file.
590 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
591 * Most of the global variables have been moved to a new
592 * struct configuration_spec, accessed through csp->config->globalname
593 * Most of the globals remaining are used by the Win32 GUI.
595 * Revision 1.5 2001/05/25 22:34:30 jongfoster
598 * Revision 1.4 2001/05/22 18:46:04 oes
600 * - Enabled filtering banners by size rather than URL
601 * by adding patterns that replace all standard banner
602 * sizes with the "Junkbuster" gif to the re_filterfile
604 * - Enabled filtering WebBugs by providing a pattern
605 * which kills all 1x1 images
607 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
608 * which is selected by the (nonstandard and therefore
609 * capital) letter 'U' in the option string.
610 * It causes the quantifiers to be ungreedy by default.
611 * Appending a ? turns back to greedy (!).
613 * - Added a new interceptor ijb-send-banner, which
614 * sends back the "Junkbuster" gif. Without imagelist or
615 * MSIE detection support, or if tinygif = 1, or the
616 * URL isn't recognized as an imageurl, a lame HTML
617 * explanation is sent instead.
619 * - Added new feature, which permits blocking remote
620 * script redirects and firing back a local redirect
622 * The feature is conditionally compiled, i.e. it
623 * can be disabled with --disable-fast-redirects,
624 * plus it must be activated by a "fast-redirects"
625 * line in the config file, has its own log level
626 * and of course wants to be displayed by show-proxy-args
627 * Note: Boy, all the #ifdefs in 1001 locations and
628 * all the fumbling with configure.in and acconfig.h
629 * were *way* more work than the feature itself :-(
631 * - Because a generic redirect template was needed for
632 * this, tinygif = 3 now uses the same.
634 * - Moved GIFs, and other static HTTP response templates
639 * - Removed some >400 CRs again (Jon, you really worked
642 * Revision 1.3 2001/05/20 16:44:47 jongfoster
643 * Removing last hardcoded Junkbusters.com URLs.
645 * Revision 1.2 2001/05/20 01:21:20 jongfoster
646 * Version 2.9.4 checkin.
647 * - Merged popupfile and cookiefile, and added control over PCRS
648 * filtering, in new "permissionsfile".
649 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
650 * file error you now get a message box (in the Win32 GUI) rather
651 * than the program exiting with no explanation.
652 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
654 * - Removed tabs from "config"
655 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
656 * - Bumped up version number.
658 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
659 * Initial import of version 2.9.3 source tree
662 *********************************************************************/
668 #include <sys/types.h>
674 #ifdef HAVE_GETADDRINFO
676 #include <sys/socket.h>
677 #endif /* def HAVE_GETADDRINFO */
682 #endif /* ndef __OS2__ */
683 #include <netinet/in.h>
685 #include <winsock2.h>
686 #endif /* ndef _WIN32 */
690 #endif /* def __OS2__ */
698 #include "jbsockets.h"
699 #include "miscutil.h"
703 #include "deanimate.h"
704 #include "urlmatch.h"
711 const char filters_h_rcs[] = FILTERS_H_VERSION;
713 /* Fix a problem with Solaris. There should be no effect on other
715 * Solaris's isspace() is a macro which uses it's argument directly
716 * as an array index. Therefore we need to make sure that high-bit
717 * characters generate +ve values, and ideally we also want to make
718 * the argument match the declared parameter type of "int".
720 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
722 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
723 static jb_err prepare_for_filtering(struct client_state *csp);
726 #ifdef HAVE_GETADDRINFO
727 /*********************************************************************
729 * Function : sockaddr_storage_to_ip
731 * Description : Access internal structure of sockaddr_storage
734 * 1 : addr = socket address
735 * 2 : ip = IP address as array of octets in network order
736 * (it points into addr)
737 * 3 : len = length of IP address in octets
738 * 4 : port = port number in network order;
740 * Returns : 0 = no errror; -1 otherwise.
742 *********************************************************************/
743 int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, uint8_t **ip,
744 unsigned int *len, in_port_t **port)
751 switch (addr->ss_family)
761 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
765 *port = &((struct sockaddr_in *)addr)->sin_port;
776 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
780 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
785 /* Unsupported address family */
793 /*********************************************************************
795 * Function : match_sockaddr
797 * Description : Check whether address matches network (IP address and port)
800 * 1 : network = socket address of subnework
801 * 2 : netmask = network mask as socket address
802 * 3 : address = checked socket address against given network
804 * Returns : 0 = doesn't match; 1 = does match
806 *********************************************************************/
807 int match_sockaddr(const struct sockaddr_storage *network,
808 const struct sockaddr_storage *netmask,
809 const struct sockaddr_storage *address)
811 uint8_t *network_addr, *netmask_addr, *address_addr;
812 unsigned int addr_len;
813 in_port_t *network_port, *netmask_port, *address_port;
816 if (network->ss_family != netmask->ss_family)
818 /* This should never happen */
819 log_error(LOG_LEVEL_ERROR,
820 "Internal error at %s:%llu: network and netmask differ in family",
825 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
826 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
827 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
829 /* Check for family */
830 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
831 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
833 /* Map AF_INET6 V4MAPPED address into AF_INET */
837 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
838 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
840 /* Map AF_INET6 V4MAPPED network into AF_INET */
845 else if (network->ss_family != address->ss_family)
850 /* XXX: Port check is signaled in netmask */
851 if (*netmask_port && *network_port != *address_port)
856 /* TODO: Optimize by checking by words insted of octets */
857 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
859 if ((network_addr[i] & netmask_addr[i]) !=
860 (address_addr[i] & netmask_addr[i]))
868 #endif /* def HAVE_GETADDRINFO */
871 /*********************************************************************
873 * Function : block_acl
875 * Description : Block this request?
876 * Decide yes or no based on ACL file.
879 * 1 : dst = The proxy or gateway address this is going to.
880 * Or NULL to check all possible targets.
881 * 2 : csp = Current client state (buffers, headers, etc...)
882 * Also includes the client IP address.
884 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
886 *********************************************************************/
887 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
889 struct access_control_list *acl = csp->config->acl;
891 /* if not using an access control list, then permit the connection */
897 /* search the list */
901 #ifdef HAVE_GETADDRINFO
902 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
904 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
910 /* Just want to check if they have any access */
911 if (acl->action == ACL_PERMIT)
917 #ifdef HAVE_GETADDRINFO
919 * XXX: An undefined acl->dst is full of zeros and should be
920 * considered a wildcard address. sockaddr_storage_to_ip()
921 * fails on such destinations because of unknown sa_familly
922 * (glibc only?). However this test is not portable.
924 * So, we signal the acl->dst is wildcard in wildcard_dst.
927 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
929 ((dst->addr & acl->dst->mask) == acl->dst->addr)
930 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
934 if (acl->action == ACL_PERMIT)
952 /*********************************************************************
954 * Function : acl_addr
956 * Description : Called from `load_config' to parse an ACL address.
959 * 1 : aspec = String specifying ACL address.
960 * 2 : aca = struct access_control_addr to fill in.
962 * Returns : 0 => Ok, everything else is an error.
964 *********************************************************************/
965 int acl_addr(const char *aspec, struct access_control_addr *aca)
968 #ifdef HAVE_GETADDRINFO
969 struct addrinfo hints, *result;
971 in_port_t *mask_port;
972 unsigned int addr_len;
975 #endif /* def HAVE_GETADDRINFO */
977 char *acl_spec = NULL;
979 #ifdef HAVE_GETADDRINFO
980 /* XXX: Depend on ai_family */
988 * Use a temporary acl spec copy so we can log
989 * the unmodified original in case of parse errors.
991 acl_spec = strdup(aspec);
992 if (acl_spec == NULL)
994 /* XXX: This will be logged as parse error. */
998 if ((p = strchr(acl_spec, '/')) != NULL)
1001 if (ijb_isdigit(*p) == 0)
1006 masklength = atoi(p);
1009 if ((masklength < 0) ||
1010 #ifdef HAVE_GETADDRINFO
1021 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
1024 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
1033 p = strchr(acl_spec, ':');
1036 #ifdef HAVE_GETADDRINFO
1037 memset(&hints, 0, sizeof(struct addrinfo));
1038 hints.ai_family = AF_UNSPEC;
1039 hints.ai_socktype = SOCK_STREAM;
1041 i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result);
1046 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
1047 acl_spec, p, gai_strerror(i));
1051 /* TODO: Allow multihomed hostnames */
1052 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
1053 freeaddrinfo(result);
1060 port = strtol(p, &endptr, 10);
1062 if (port <= 0 || port > 65535 || *endptr != '\0')
1069 aca->port = (unsigned long)port;
1071 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
1074 if (aca->addr == INADDR_NONE)
1076 /* XXX: This will be logged as parse error. */
1079 #endif /* def HAVE_GETADDRINFO */
1081 /* build the netmask */
1082 #ifdef HAVE_GETADDRINFO
1083 /* Clip masklength according to current family. */
1084 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
1089 aca->mask.ss_family = aca->addr.ss_family;
1090 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
1097 /* ACL contains a port number, check ports in the future. */
1102 * XXX: This could be optimized to operate on whole words instead
1103 * of octets (128-bit CPU could do it in one iteration).
1106 * Octets after prefix can be ommitted because of
1107 * previous initialization to zeros.
1109 for (i = 0; (i < addr_len) && masklength; i++)
1111 if (masklength >= 8)
1113 mask_data[i] = 0xFF;
1119 * XXX: This assumes MSB of octet is on the left side.
1120 * This should be true for all architectures or solved
1121 * by the link layer.
1123 mask_data[i] = ~((1 << (8 - masklength)) - 1);
1130 for (i=1; i <= masklength ; i++)
1132 aca->mask |= (1U << (32 - i));
1135 /* now mask off the host portion of the ip address
1136 * (i.e. save on the network portion of the address).
1138 aca->addr = aca->addr & aca->mask;
1139 #endif /* def HAVE_GETADDRINFO */
1144 #endif /* def FEATURE_ACL */
1147 /*********************************************************************
1149 * Function : connect_port_is_forbidden
1151 * Description : Check to see if CONNECT requests to the destination
1152 * port of this request are forbidden. The check is
1153 * independend of the actual request method.
1156 * 1 : csp = Current client state (buffers, headers, etc...)
1158 * Returns : True if yes, false otherwise.
1160 *********************************************************************/
1161 int connect_port_is_forbidden(const struct client_state *csp)
1163 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
1164 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
1169 /*********************************************************************
1171 * Function : block_url
1173 * Description : Called from `chat'. Check to see if we need to block this.
1176 * 1 : csp = Current client state (buffers, headers, etc...)
1178 * Returns : NULL => unblocked, else HTTP block response
1180 *********************************************************************/
1181 struct http_response *block_url(struct client_state *csp)
1183 struct http_response *rsp;
1184 const char *new_content_type = NULL;
1187 * If it's not blocked, don't block it ;-)
1189 if ((csp->action->flags & ACTION_BLOCK) == 0)
1193 if (csp->action->flags & ACTION_REDIRECT)
1195 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
1198 * Else, prepare a response
1200 if (NULL == (rsp = alloc_http_response()))
1202 return cgi_error_memory();
1206 * If it's an image-url, send back an image or redirect
1207 * as specified by the relevant +image action
1209 #ifdef FEATURE_IMAGE_BLOCKING
1210 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
1211 && is_imageurl(csp))
1214 /* determine HOW images should be blocked */
1215 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
1217 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1219 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
1221 #if 1 /* Two alternative strategies, use this one for now: */
1223 /* and handle accordingly: */
1224 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1226 rsp->status = strdup("403 Request blocked by Privoxy");
1227 if (rsp->status == NULL)
1229 free_http_response(rsp);
1230 return cgi_error_memory();
1232 rsp->body = bindup(image_pattern_data, image_pattern_length);
1233 if (rsp->body == NULL)
1235 free_http_response(rsp);
1236 return cgi_error_memory();
1238 rsp->content_length = image_pattern_length;
1240 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1242 free_http_response(rsp);
1243 return cgi_error_memory();
1247 else if (0 == strcmpic(p, "blank"))
1249 rsp->status = strdup("403 Request blocked by Privoxy");
1250 if (rsp->status == NULL)
1252 free_http_response(rsp);
1253 return cgi_error_memory();
1255 rsp->body = bindup(image_blank_data, image_blank_length);
1256 if (rsp->body == NULL)
1258 free_http_response(rsp);
1259 return cgi_error_memory();
1261 rsp->content_length = image_blank_length;
1263 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1265 free_http_response(rsp);
1266 return cgi_error_memory();
1272 rsp->status = strdup("302 Local Redirect from Privoxy");
1273 if (rsp->status == NULL)
1275 free_http_response(rsp);
1276 return cgi_error_memory();
1279 if (enlist_unique_header(rsp->headers, "Location", p))
1281 free_http_response(rsp);
1282 return cgi_error_memory();
1286 #else /* Following code is disabled for now */
1288 /* and handle accordingly: */
1289 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1291 p = CGI_PREFIX "send-banner?type=pattern";
1293 else if (0 == strcmpic(p, "blank"))
1295 p = CGI_PREFIX "send-banner?type=blank";
1297 rsp->status = strdup("302 Local Redirect from Privoxy");
1298 if (rsp->status == NULL)
1300 free_http_response(rsp);
1301 return cgi_error_memory();
1304 if (enlist_unique_header(rsp->headers, "Location", p))
1306 free_http_response(rsp);
1307 return cgi_error_memory();
1309 #endif /* Preceeding code is disabled for now */
1311 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1314 * Send empty document.
1316 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1319 rsp->body = strdup(" ");
1320 rsp->content_length = 1;
1322 rsp->status = strdup("403 Request blocked by Privoxy");
1323 if (rsp->status == NULL)
1325 free_http_response(rsp);
1326 return cgi_error_memory();
1328 if (new_content_type != 0)
1330 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1331 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1333 free_http_response(rsp);
1334 return cgi_error_memory();
1339 #endif /* def FEATURE_IMAGE_BLOCKING */
1342 * Else, generate an HTML "blocked" message:
1346 struct map * exports;
1350 * Workaround for stupid Netscape bug which prevents
1351 * pages from being displayed if loading a referenced
1352 * JavaScript or style sheet fails. So make it appear
1353 * as if it succeeded.
1355 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1356 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1357 && !strstr(p, "Gecko") /* save Mozilla, */
1358 && !strstr(p, "compatible") /* MSIE */
1359 && !strstr(p, "Opera")) /* and Opera. */
1361 rsp->status = strdup("200 Request for blocked URL");
1365 rsp->status = strdup("403 Request for blocked URL");
1368 if (rsp->status == NULL)
1370 free_http_response(rsp);
1371 return cgi_error_memory();
1374 exports = default_exports(csp, NULL);
1375 if (exports == NULL)
1377 free_http_response(rsp);
1378 return cgi_error_memory();
1381 #ifdef FEATURE_FORCE_LOAD
1382 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1384 * Export the force conditional block killer if
1386 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1387 * - Privoxy is configured to enforce blocks, or
1388 * - it's a CONNECT request and enforcing wouldn't work anyway.
1390 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1391 || (0 == strcmpic(csp->http->gpc, "connect")))
1392 #endif /* ndef FEATURE_FORCE_LOAD */
1394 err = map_block_killer(exports, "force-support");
1397 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1398 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1399 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1400 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1403 const char *block_reason;
1404 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
1406 block_reason = csp->action->string[ACTION_STRING_BLOCK];
1410 assert(connect_port_is_forbidden(csp));
1411 block_reason = "Forbidden CONNECT port.";
1413 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
1418 free_http_response(rsp);
1419 return cgi_error_memory();
1422 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1425 free_http_response(rsp);
1426 return cgi_error_memory();
1429 rsp->reason = RSP_REASON_BLOCKED;
1431 return finish_http_response(csp, rsp);
1436 #ifdef FEATURE_TRUST
1437 /*********************************************************************
1439 * Function : trust_url FIXME: I should be called distrust_url
1441 * Description : Calls is_untrusted_url to determine if the URL is trusted
1442 * and if not, returns a HTTP 403 response with a reject message.
1445 * 1 : csp = Current client state (buffers, headers, etc...)
1447 * Returns : NULL => trusted, else http_response.
1449 *********************************************************************/
1450 struct http_response *trust_url(struct client_state *csp)
1452 struct http_response *rsp;
1453 struct map * exports;
1454 char buf[BUFFER_SIZE];
1456 struct url_spec **tl;
1461 * Don't bother to work on trusted URLs
1463 if (!is_untrusted_url(csp))
1469 * Else, prepare a response:
1471 if (NULL == (rsp = alloc_http_response()))
1473 return cgi_error_memory();
1476 rsp->status = strdup("403 Request blocked by Privoxy");
1477 exports = default_exports(csp, NULL);
1478 if (exports == NULL || rsp->status == NULL)
1480 free_http_response(rsp);
1481 return cgi_error_memory();
1485 * Export the protocol, host, port, and referrer information
1487 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1488 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1489 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1491 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1493 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1497 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1503 free_http_response(rsp);
1504 return cgi_error_memory();
1508 * Export the trust list
1511 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1513 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1514 string_append(&p, buf);
1516 err = map(exports, "trusted-referrers", 1, p, 0);
1521 free_http_response(rsp);
1522 return cgi_error_memory();
1526 * Export the trust info, if available
1528 if (csp->config->trust_info->first)
1530 struct list_entry *l;
1533 for (l = csp->config->trust_info->first; l ; l = l->next)
1535 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1536 string_append(&p, buf);
1538 err = map(exports, "trust-info", 1, p, 0);
1542 err = map_block_killer(exports, "have-trust-info");
1548 free_http_response(rsp);
1549 return cgi_error_memory();
1553 * Export the force conditional block killer if
1555 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1556 * - Privoxy is configured to enforce blocks, or
1557 * - it's a CONNECT request and enforcing wouldn't work anyway.
1559 #ifdef FEATURE_FORCE_LOAD
1560 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1561 || (0 == strcmpic(csp->http->gpc, "connect")))
1563 err = map_block_killer(exports, "force-support");
1567 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1569 #else /* ifndef FEATURE_FORCE_LOAD */
1570 err = map_block_killer(exports, "force-support");
1571 #endif /* ndef FEATURE_FORCE_LOAD */
1576 free_http_response(rsp);
1577 return cgi_error_memory();
1581 * Build the response
1583 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1586 free_http_response(rsp);
1587 return cgi_error_memory();
1589 rsp->reason = RSP_REASON_UNTRUSTED;
1591 return finish_http_response(csp, rsp);
1593 #endif /* def FEATURE_TRUST */
1596 /*********************************************************************
1598 * Function : compile_dynamic_pcrs_job_list
1600 * Description : Compiles a dynamic pcrs job list (one with variables
1601 * resolved at request time)
1604 * 1 : csp = Current client state (buffers, headers, etc...)
1605 * 2 : b = The filter list to compile
1607 * Returns : NULL in case of errors, otherwise the
1610 *********************************************************************/
1611 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1613 struct list_entry *pattern;
1614 pcrs_job *job_list = NULL;
1615 pcrs_job *dummy = NULL;
1616 pcrs_job *lastjob = NULL;
1619 const struct pcrs_variable variables[] =
1621 {"url", csp->http->url, 1},
1622 {"path", csp->http->path, 1},
1623 {"host", csp->http->host, 1},
1624 {"origin", csp->ip_addr_str, 1},
1628 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1630 assert(pattern->str != NULL);
1632 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1636 log_error(LOG_LEVEL_ERROR,
1637 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1638 pattern->str, b->name, pcrs_strerror(error));
1643 if (error == PCRS_WARN_TRUNCATION)
1645 log_error(LOG_LEVEL_ERROR,
1646 "At least one of the variables in \'%s\' had to "
1647 "be truncated before compilation", pattern->str);
1649 if (job_list == NULL)
1655 lastjob->next = dummy;
1665 /*********************************************************************
1667 * Function : rewrite_url
1669 * Description : Rewrites a URL with a single pcrs command
1670 * and returns the result if it differs from the
1671 * original and isn't obviously invalid.
1674 * 1 : old_url = URL to rewrite.
1675 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1678 * Returns : NULL if the pcrs_command didn't change the url, or
1679 * the result of the modification.
1681 *********************************************************************/
1682 char *rewrite_url(char *old_url, const char *pcrs_command)
1684 char *new_url = NULL;
1688 assert(pcrs_command);
1690 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1694 log_error(LOG_LEVEL_REDIRECTS,
1695 "pcrs command \"%s\" didn't change \"%s\".",
1696 pcrs_command, old_url);
1701 log_error(LOG_LEVEL_REDIRECTS,
1702 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1703 pcrs_command, old_url, pcrs_strerror(hits));
1706 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1708 log_error(LOG_LEVEL_ERROR,
1709 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1710 "but the result doesn't look like a valid URL and will be ignored.",
1711 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1716 log_error(LOG_LEVEL_REDIRECTS,
1717 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1718 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1726 #ifdef FEATURE_FAST_REDIRECTS
1727 /*********************************************************************
1729 * Function : get_last_url
1731 * Description : Search for the last URL inside a string.
1732 * If the string already is a URL, it will
1733 * be the first URL found.
1736 * 1 : subject = the string to check
1737 * 2 : redirect_mode = +fast-redirect{} mode
1739 * Returns : NULL if no URL was found, or
1740 * the last URL found.
1742 *********************************************************************/
1743 char *get_last_url(char *subject, const char *redirect_mode)
1745 char *new_url = NULL;
1749 assert(redirect_mode);
1751 subject = strdup(subject);
1752 if (subject == NULL)
1754 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1758 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1760 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1761 new_url = url_decode(subject);
1762 if (new_url != NULL)
1769 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1773 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1776 * Find the last URL encoded in the request
1779 while ((tmp = strstr(tmp, "http://")) != NULL)
1783 tmp = (new_url != NULL) ? new_url : subject;
1784 while ((tmp = strstr(tmp, "https://")) != NULL)
1789 if ((new_url != NULL)
1790 && ( (new_url != subject)
1791 || (0 == strncmpic(subject, "http://", 7))
1792 || (0 == strncmpic(subject, "https://", 8))
1796 * Return new URL if we found a redirect
1797 * or if the subject already was a URL.
1799 * The second case makes sure that we can
1800 * chain get_last_url after another redirection check
1801 * (like rewrite_url) without losing earlier redirects.
1803 new_url = strdup(new_url);
1812 #endif /* def FEATURE_FAST_REDIRECTS */
1815 /*********************************************************************
1817 * Function : redirect_url
1819 * Description : Checks if Privoxy should answer the request with
1820 * a HTTP redirect and generates the redirect if
1824 * 1 : csp = Current client state (buffers, headers, etc...)
1826 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1828 *********************************************************************/
1829 struct http_response *redirect_url(struct client_state *csp)
1831 struct http_response *rsp;
1832 #ifdef FEATURE_FAST_REDIRECTS
1834 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1835 * as compile-time option? The user can easily disable
1836 * it in his action file.
1838 char * redirect_mode;
1839 #endif /* def FEATURE_FAST_REDIRECTS */
1840 char *old_url = NULL;
1841 char *new_url = NULL;
1842 char *redirection_string;
1844 if ((csp->action->flags & ACTION_REDIRECT))
1846 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1849 * If the redirection string begins with 's',
1850 * assume it's a pcrs command, otherwise treat it as
1851 * properly formatted URL and use it for the redirection
1854 * According to RFC 2616 section 14.30 the URL
1855 * has to be absolute and if the user tries:
1856 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1857 * she would get undefined results anyway.
1861 if (*redirection_string == 's')
1863 old_url = csp->http->url;
1864 new_url = rewrite_url(old_url, redirection_string);
1868 log_error(LOG_LEVEL_REDIRECTS,
1869 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1870 redirection_string);
1871 new_url = strdup(redirection_string);
1875 #ifdef FEATURE_FAST_REDIRECTS
1876 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1878 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1881 * If it exists, use the previously rewritten URL as input
1882 * otherwise just use the old path.
1884 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1885 new_url = get_last_url(old_url, redirect_mode);
1890 * Disable redirect checkers, so that they
1891 * will be only run more than once if the user
1892 * also enables them through tags.
1894 * From a performance point of view
1895 * it doesn't matter, but the duplicated
1896 * log messages are annoying.
1898 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1899 #endif /* def FEATURE_FAST_REDIRECTS */
1900 csp->action->flags &= ~ACTION_REDIRECT;
1902 /* Did any redirect action trigger? */
1905 if (0 == strcmpic(new_url, csp->http->url))
1907 log_error(LOG_LEVEL_ERROR,
1908 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1909 csp->http->url, new_url);
1914 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1916 if (NULL == (rsp = alloc_http_response()))
1919 return cgi_error_memory();
1922 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1923 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1926 free_http_response(rsp);
1927 return cgi_error_memory();
1929 rsp->reason = RSP_REASON_REDIRECTED;
1932 return finish_http_response(csp, rsp);
1936 /* Only reached if no redirect is required */
1942 #ifdef FEATURE_IMAGE_BLOCKING
1943 /*********************************************************************
1945 * Function : is_imageurl
1947 * Description : Given a URL, decide whether it is an image or not,
1948 * using either the info from a previous +image action
1949 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1950 * is MSIE and not on a Mac, tell from the browser's accept
1954 * 1 : csp = Current client state (buffers, headers, etc...)
1956 * Returns : True (nonzero) if URL is an image, false (0)
1959 *********************************************************************/
1960 int is_imageurl(const struct client_state *csp)
1962 #ifdef FEATURE_IMAGE_DETECT_MSIE
1965 tmp = get_header_value(csp->headers, "User-Agent:");
1966 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1968 tmp = get_header_value(csp->headers, "Accept:");
1969 if (tmp && strstr(tmp, "image/gif"))
1971 /* Client will accept HTML. If this seems counterintuitive,
1981 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1983 return ((csp->action->flags & ACTION_IMAGE) != 0);
1986 #endif /* def FEATURE_IMAGE_BLOCKING */
1989 #ifdef FEATURE_TRUST
1990 /*********************************************************************
1992 * Function : is_untrusted_url
1994 * Description : Should we "distrust" this URL (and block it)?
1996 * Yes if it matches a line in the trustfile, or if the
1997 * referrer matches a line starting with "+" in the
2002 * 1 : csp = Current client state (buffers, headers, etc...)
2004 * Returns : 0 => trusted, 1 => untrusted
2006 *********************************************************************/
2007 int is_untrusted_url(const struct client_state *csp)
2009 struct file_list *fl;
2010 struct block_spec *b;
2011 struct url_spec **trusted_url;
2012 struct http_request rhttp[1];
2013 const char * referer;
2017 * If we don't have a trustlist, we trust everybody
2019 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
2024 memset(rhttp, '\0', sizeof(*rhttp));
2027 * Do we trust the request URL itself?
2029 for (b = b->next; b ; b = b->next)
2031 if (url_match(b->url, csp->http))
2037 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
2039 /* no referrer was supplied */
2045 * If not, do we maybe trust its referrer?
2047 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
2053 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
2055 if (url_match(*trusted_url, rhttp))
2057 /* if the URL's referrer is from a trusted referrer, then
2058 * add the target spec to the trustfile as an unblocked
2059 * domain and return 0 (which means it's OK).
2064 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
2068 char * new_entry = strdup("~");
2070 string_append(&new_entry, csp->http->hostport);
2072 path = csp->http->path;
2073 if ( (path[0] == '/')
2075 && ((path_end = strchr(path + 2, '/')) != NULL))
2077 /* since this path points into a user's home space
2078 * be sure to include this spec in the trustfile.
2080 long path_len = path_end - path; /* save offset */
2081 path = strdup(path); /* Copy string */
2084 path_end = path + path_len; /* regenerate ptr to new buffer */
2085 *(path_end + 1) = '\0'; /* Truncate path after '/' */
2087 string_join(&new_entry, path);
2091 * Give a reason for generating this entry.
2093 string_append(&new_entry, " # Trusted referrer was: ");
2094 string_append(&new_entry, referer);
2096 if (new_entry != NULL)
2098 if (-1 == fprintf(fp, "%s\n", new_entry))
2100 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
2101 new_entry, csp->config->trustfile);
2107 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
2108 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
2115 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
2116 csp->http->hostport, csp->config->trustfile);
2124 #endif /* def FEATURE_TRUST */
2127 /*********************************************************************
2129 * Function : pcrs_filter_response
2131 * Description : Execute all text substitutions from all applying
2132 * +filter actions on the text buffer that's been
2133 * accumulated in csp->iob->buf.
2136 * 1 : csp = Current client state (buffers, headers, etc...)
2138 * Returns : a pointer to the (newly allocated) modified buffer.
2139 * or NULL if there were no hits or something went wrong
2141 *********************************************************************/
2142 static char *pcrs_filter_response(struct client_state *csp)
2145 size_t size, prev_size;
2151 struct file_list *fl;
2152 struct re_filterfile_spec *b;
2153 struct list_entry *filtername;
2155 int i, found_filters = 0;
2160 if (csp->iob->cur >= csp->iob->eod)
2166 * Need to check the set of re_filterfiles...
2168 for (i = 0; i < MAX_AF_FILES; i++)
2181 if (0 == found_filters)
2183 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2184 "content filtering enabled, but no content filters available.");
2188 size = (size_t)(csp->iob->eod - csp->iob->cur);
2189 old = csp->iob->cur;
2191 for (i = 0; i < MAX_AF_FILES; i++)
2194 if ((NULL == fl) || (NULL == fl->f))
2197 * Either there are no filter files
2198 * left, or this filter file just
2199 * contains no valid filters.
2201 * Continue to be sure we don't miss
2202 * valid filter files that are chained
2203 * after empty or invalid ones.
2208 * For all applying +filter actions, look if a filter by that
2209 * name exists and if yes, execute it's pcrs_joblist on the
2212 for (b = fl->f; b; b = b->next)
2214 if (b->type != FT_CONTENT_FILTER)
2216 /* Skip header filters */
2220 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
2221 filtername ; filtername = filtername->next)
2223 if (strcmp(b->name, filtername->str) == 0)
2225 int current_hits = 0; /* Number of hits caused by this filter */
2226 int job_number = 0; /* Which job we're currently executing */
2227 int job_hits = 0; /* How many hits the current job caused */
2228 pcrs_job *joblist = b->joblist;
2230 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2232 if (NULL == joblist)
2234 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2239 /* Apply all jobs from the joblist */
2240 for (job = joblist; NULL != job; job = job->next)
2243 job_hits = pcrs_execute(job, old, size, &new, &size);
2248 * That went well. Continue filtering
2249 * and use the result of this job as
2250 * input for the next one.
2252 current_hits += job_hits;
2253 if (old != csp->iob->cur)
2262 * This job caused an unexpected error. Inform the user
2263 * and skip the rest of the jobs in this filter. We could
2264 * continue with the next job, but usually the jobs
2265 * depend on each other or are similar enough to
2266 * fail for the same reason.
2268 * At the moment our pcrs expects the error codes of pcre 3.4,
2269 * but newer pcre versions can return additional error codes.
2270 * As a result pcrs_strerror()'s error message might be
2271 * "Unknown error ...", therefore we print the numerical value
2274 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2275 * should we use LOG_LEVEL_RE_FILTER instead?
2277 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2278 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2283 if (b->dynamic) pcrs_free_joblist(joblist);
2285 log_error(LOG_LEVEL_RE_FILTER,
2286 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2287 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2289 hits += current_hits;
2296 * If there were no hits, destroy our copy and let
2297 * chat() use the original in csp->iob
2305 csp->flags |= CSP_FLAG_MODIFIED;
2306 csp->content_length = size;
2314 /*********************************************************************
2316 * Function : gif_deanimate_response
2318 * Description : Deanimate the GIF image that has been accumulated in
2319 * csp->iob->buf, set csp->content_length to the modified
2320 * size and raise the CSP_FLAG_MODIFIED flag.
2323 * 1 : csp = Current client state (buffers, headers, etc...)
2325 * Returns : a pointer to the (newly allocated) modified buffer.
2326 * or NULL in case something went wrong.
2328 *********************************************************************/
2329 static char *gif_deanimate_response(struct client_state *csp)
2331 struct binbuffer *in, *out;
2335 size = (size_t)(csp->iob->eod - csp->iob->cur);
2337 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2338 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2340 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2344 in->buffer = csp->iob->cur;
2347 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2349 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2356 if ((int)size == out->offset)
2358 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2362 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2364 csp->content_length = out->offset;
2365 csp->flags |= CSP_FLAG_MODIFIED;
2375 /*********************************************************************
2377 * Function : get_filter_function
2379 * Description : Decides which content filter function has
2380 * to be applied (if any).
2383 * 1 : csp = Current client state (buffers, headers, etc...)
2385 * Returns : The content filter function to run, or
2386 * NULL if no content filter is active
2388 *********************************************************************/
2389 filter_function_ptr get_filter_function(struct client_state *csp)
2391 filter_function_ptr filter_function = NULL;
2394 * Are we enabling text mode by force?
2396 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2399 * Do we really have to?
2401 if (csp->content_type & CT_TEXT)
2403 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2407 csp->content_type |= CT_TEXT;
2408 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2412 if (!(csp->content_type & CT_DECLARED))
2415 * The server didn't bother to declare a MIME-Type.
2416 * Assume it's text that can be filtered.
2418 * This also regulary happens with 304 responses,
2419 * therefore logging anything here would cause
2422 csp->content_type |= CT_TEXT;
2426 * Choose the applying filter function based on
2427 * the content type and action settings.
2429 if ((csp->content_type & CT_TEXT) &&
2430 (csp->rlist != NULL) &&
2431 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2433 filter_function = pcrs_filter_response;
2435 else if ((csp->content_type & CT_GIF) &&
2436 (csp->action->flags & ACTION_DEANIMATE))
2438 filter_function = gif_deanimate_response;
2441 return filter_function;
2445 /*********************************************************************
2447 * Function : remove_chunked_transfer_coding
2449 * Description : In-situ remove the "chunked" transfer coding as defined
2450 * in rfc2616 from a buffer.
2453 * 1 : buffer = Pointer to the text buffer
2454 * 2 : size = In: Number of bytes to be processed,
2455 * Out: Number of bytes after de-chunking.
2456 * (undefined in case of errors)
2458 * Returns : JB_ERR_OK for success,
2459 * JB_ERR_PARSE otherwise
2461 *********************************************************************/
2462 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2465 unsigned int chunksize = 0;
2466 char *from_p, *to_p;
2469 from_p = to_p = buffer;
2471 if (sscanf(buffer, "%x", &chunksize) != 1)
2473 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2474 return JB_ERR_PARSE;
2477 while (chunksize > 0U)
2479 if (NULL == (from_p = strstr(from_p, "\r\n")))
2481 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2482 return JB_ERR_PARSE;
2485 if ((newsize += chunksize) >= *size)
2487 log_error(LOG_LEVEL_ERROR,
2488 "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding",
2490 return JB_ERR_PARSE;
2494 memmove(to_p, from_p, (size_t) chunksize);
2495 to_p = buffer + newsize;
2496 from_p += chunksize + 2;
2498 if (sscanf(from_p, "%x", &chunksize) != 1)
2500 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2505 /* XXX: Should get its own loglevel. */
2506 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2515 /*********************************************************************
2517 * Function : prepare_for_filtering
2519 * Description : If necessary, de-chunks and decompresses
2520 * the content so it can get filterd.
2523 * 1 : csp = Current client state (buffers, headers, etc...)
2525 * Returns : JB_ERR_OK for success,
2526 * JB_ERR_PARSE otherwise
2528 *********************************************************************/
2529 static jb_err prepare_for_filtering(struct client_state *csp)
2531 jb_err err = JB_ERR_OK;
2534 * If the body has a "chunked" transfer-encoding,
2535 * get rid of it, adjusting size and iob->eod
2537 if (csp->flags & CSP_FLAG_CHUNKED)
2539 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2541 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2542 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2543 if (JB_ERR_OK == err)
2545 csp->iob->eod = csp->iob->cur + size;
2546 csp->flags |= CSP_FLAG_MODIFIED;
2550 return JB_ERR_PARSE;
2556 * If the body has a supported transfer-encoding,
2557 * decompress it, adjusting size and iob->eod.
2559 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2561 if (0 == csp->iob->eod - csp->iob->cur)
2563 /* Nothing left after de-chunking. */
2567 err = decompress_iob(csp);
2569 if (JB_ERR_OK == err)
2571 csp->flags |= CSP_FLAG_MODIFIED;
2572 csp->content_type &= ~CT_TABOO;
2577 * Unset CT_GZIP and CT_DEFLATE to remember not
2578 * to modify the Content-Encoding header later.
2580 csp->content_type &= ~CT_GZIP;
2581 csp->content_type &= ~CT_DEFLATE;
2590 /*********************************************************************
2592 * Function : execute_content_filter
2594 * Description : Executes a given content filter.
2597 * 1 : csp = Current client state (buffers, headers, etc...)
2598 * 2 : content_filter = The filter function to execute
2600 * Returns : Pointer to the modified buffer, or
2601 * NULL if filtering failed or wasn't necessary.
2603 *********************************************************************/
2604 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2606 if (0 == csp->iob->eod - csp->iob->cur)
2609 * No content (probably status code 301, 302 ...),
2610 * no filtering necessary.
2615 if (JB_ERR_OK != prepare_for_filtering(csp))
2618 * failed to de-chunk or decompress.
2623 if (0 == csp->iob->eod - csp->iob->cur)
2626 * Clown alarm: chunked and/or compressed nothing delivered.
2631 return ((*content_filter)(csp));
2635 /*********************************************************************
2637 * Function : get_url_actions
2639 * Description : Gets the actions for this URL.
2642 * 1 : csp = Current client state (buffers, headers, etc...)
2643 * 2 : http = http_request request for blocked URLs
2647 *********************************************************************/
2648 void get_url_actions(struct client_state *csp, struct http_request *http)
2650 struct file_list *fl;
2651 struct url_actions *b;
2654 init_current_action(csp->action);
2656 for (i = 0; i < MAX_AF_FILES; i++)
2658 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2663 apply_url_actions(csp->action, http, b);
2670 /*********************************************************************
2672 * Function : apply_url_actions
2674 * Description : Applies a list of URL actions.
2677 * 1 : action = Destination.
2678 * 2 : http = Current URL
2679 * 3 : b = list of URL actions to apply
2683 *********************************************************************/
2684 void apply_url_actions(struct current_action_spec *action,
2685 struct http_request *http,
2686 struct url_actions *b)
2690 /* Should never happen */
2694 for (b = b->next; NULL != b; b = b->next)
2696 if (url_match(b->url, http))
2698 merge_current_action(action, b->action);
2704 /*********************************************************************
2706 * Function : get_forward_override_settings
2708 * Description : Returns forward settings as specified with the
2709 * forward-override{} action. forward-override accepts
2710 * forward lines similar to the one used in the
2711 * configuration file, but without the URL pattern.
2717 * in the configuration file can be replaced with
2718 * the action section:
2720 * {+forward-override{forward .}}
2724 * 1 : csp = Current client state (buffers, headers, etc...)
2726 * Returns : Pointer to forwarding structure in case of success.
2727 * Invalid syntax is fatal.
2729 *********************************************************************/
2730 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2732 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2733 char forward_settings[BUFFER_SIZE];
2734 char *http_parent = NULL;
2735 /* variable names were chosen for consistency reasons. */
2736 struct forward_spec *fwd = NULL;
2740 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2741 /* Should be enforced by load_one_actions_file() */
2742 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2744 /* Create a copy ssplit can modify */
2745 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2747 if (NULL != csp->fwd)
2750 * XXX: Currently necessary to prevent memory
2751 * leaks when the show-url-info cgi page is visited.
2753 unload_forward_spec(csp->fwd);
2757 * allocate a new forward node, valid only for
2758 * the lifetime of this request. Save its location
2759 * in csp as well, so sweep() can free it later on.
2761 fwd = csp->fwd = zalloc(sizeof(*fwd));
2764 log_error(LOG_LEVEL_FATAL,
2765 "can't allocate memory for forward-override{%s}", forward_override_line);
2766 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2770 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2771 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2773 fwd->type = SOCKS_NONE;
2775 /* Parse the parent HTTP proxy host:port */
2776 http_parent = vec[1];
2779 else if (vec_count == 3)
2781 char *socks_proxy = NULL;
2783 if (!strcasecmp(vec[0], "forward-socks4"))
2785 fwd->type = SOCKS_4;
2786 socks_proxy = vec[1];
2788 else if (!strcasecmp(vec[0], "forward-socks4a"))
2790 fwd->type = SOCKS_4A;
2791 socks_proxy = vec[1];
2793 else if (!strcasecmp(vec[0], "forward-socks5"))
2795 fwd->type = SOCKS_5;
2796 socks_proxy = vec[1];
2799 if (NULL != socks_proxy)
2801 /* Parse the SOCKS proxy host[:port] */
2802 fwd->gateway_host = strdup(socks_proxy);
2804 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2806 *socks_proxy++ = '\0';
2807 fwd->gateway_port = (int)strtol(socks_proxy, NULL, 0);
2810 if (fwd->gateway_port <= 0)
2812 fwd->gateway_port = 1080;
2815 http_parent = vec[2];
2819 if (NULL == http_parent)
2821 log_error(LOG_LEVEL_FATAL,
2822 "Invalid forward-override syntax in: %s", forward_override_line);
2823 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2826 /* Parse http forwarding settings */
2827 if (strcmp(http_parent, ".") != 0)
2829 fwd->forward_host = strdup(http_parent);
2831 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2833 *http_parent++ = '\0';
2834 fwd->forward_port = (int)strtol(http_parent, NULL, 0);
2837 if (fwd->forward_port <= 0)
2839 fwd->forward_port = 8000;
2843 assert (NULL != fwd);
2845 log_error(LOG_LEVEL_CONNECT,
2846 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2852 /*********************************************************************
2854 * Function : forward_url
2856 * Description : Should we forward this to another proxy?
2859 * 1 : csp = Current client state (buffers, headers, etc...)
2860 * 2 : http = http_request request for current URL
2862 * Returns : Pointer to forwarding information.
2864 *********************************************************************/
2865 const struct forward_spec *forward_url(struct client_state *csp,
2866 const struct http_request *http)
2868 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2869 struct forward_spec *fwd = csp->config->forward;
2871 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2873 return get_forward_override_settings(csp);
2883 if (url_match(fwd->url, http))
2894 /*********************************************************************
2896 * Function : direct_response
2898 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2899 * request and if so, return a HTTP 501 to the client.
2901 * FIXME: I have a stupid name and I should handle the
2902 * requests properly. Still, what we do here is rfc-
2903 * compliant, whereas ignoring or forwarding are not.
2906 * 1 : csp = Current client state (buffers, headers, etc...)
2908 * Returns : http_response if , NULL if nonmatch or handler fail
2910 *********************************************************************/
2911 struct http_response *direct_response(struct client_state *csp)
2913 struct http_response *rsp;
2914 struct list_entry *p;
2916 if ((0 == strcmpic(csp->http->gpc, "trace"))
2917 || (0 == strcmpic(csp->http->gpc, "options")))
2919 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2921 if (!strncmpic("Max-Forwards:", p->str, 13))
2923 unsigned int max_forwards;
2926 * If it's a Max-Forwards value of zero,
2927 * we have to intercept the request.
2929 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2932 * FIXME: We could handle at least TRACE here,
2933 * but that would require a verbatim copy of
2934 * the request which we don't have anymore
2936 log_error(LOG_LEVEL_HEADER,
2937 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2940 /* Get mem for response or fail*/
2941 if (NULL == (rsp = alloc_http_response()))
2943 return cgi_error_memory();
2946 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2948 free_http_response(rsp);
2949 return cgi_error_memory();
2953 rsp->reason = RSP_REASON_UNSUPPORTED;
2955 return(finish_http_response(csp, rsp));
2964 /*********************************************************************
2966 * Function : content_filters_enabled
2968 * Description : Checks whether there are any content filters
2969 * enabled for the current request.
2972 * 1 : action = Action spec to check.
2974 * Returns : TRUE for yes, FALSE otherwise
2976 *********************************************************************/
2977 int content_filters_enabled(const struct current_action_spec *action)
2979 return ((action->flags & ACTION_DEANIMATE) ||
2980 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));