1 const char filters_rcs[] = "$Id: filters.c,v 1.114 2009/04/17 11:27:49 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 1.114 2009/04/17 11:27:49 fabiankeil
44 * Petr Pisar's privoxy-3.0.12-ipv6-3.diff.
46 * Revision 1.113 2009/03/08 14:19:23 fabiankeil
47 * Fix justified (but harmless) compiler warnings
48 * on platforms where sizeof(int) < sizeof(long).
50 * Revision 1.112 2009/03/01 18:28:23 fabiankeil
51 * Help clang understand that we aren't dereferencing
54 * Revision 1.111 2008/12/04 18:13:46 fabiankeil
55 * Fix a cparser warning.
57 * Revision 1.110 2008/11/10 16:40:25 fabiankeil
58 * Fix a gcc44 warning.
60 * Revision 1.109 2008/11/08 15:48:41 fabiankeil
61 * Mention actual values when complaining about
62 * the chunk size exceeding the buffer size.
64 * Revision 1.108 2008/05/21 15:35:08 fabiankeil
65 * - Mark csp as immutable for block_acl().
66 * - Remove an obsolete complaint about filter_popups().
68 * Revision 1.107 2008/05/04 17:52:56 fabiankeil
69 * Adjust parse_http_url() call to new prototype.
71 * Revision 1.106 2008/05/03 16:40:44 fabiankeil
72 * Change content_filters_enabled()'s parameter from
73 * csp->action to action so it can be also used in the
74 * CGI code. Don't bother checking if there are filters
75 * loaded, as that's somewhat besides the point.
77 * Revision 1.105 2008/03/28 15:13:39 fabiankeil
78 * Remove inspect-jpegs action.
80 * Revision 1.104 2008/03/27 18:27:24 fabiankeil
81 * Remove kill-popups action.
83 * Revision 1.103 2008/03/06 16:33:45 fabiankeil
84 * If limit-connect isn't used, don't limit CONNECT requests to port 443.
86 * Revision 1.102 2008/03/01 14:00:44 fabiankeil
87 * Let the block action take the reason for the block
88 * as argument and show it on the "blocked" page.
90 * Revision 1.101 2008/02/23 16:57:12 fabiankeil
91 * Rename url_actions() to get_url_actions() and let it
92 * use the standard parameter ordering.
94 * Revision 1.100 2008/02/23 16:33:43 fabiankeil
95 * Let forward_url() use the standard parameter ordering
96 * and mark its second parameter immutable.
98 * Revision 1.99 2008/02/03 13:57:58 fabiankeil
99 * Add SOCKS5 support for forward-override{}.
101 * Revision 1.98 2008/01/04 17:43:45 fabiankeil
102 * Improve the warning messages that get logged if the action files
103 * "enable" filters but no filters of that type have been loaded.
105 * Revision 1.97 2007/11/30 15:37:03 fabiankeil
106 * Use freez instead of free.
108 * Revision 1.96 2007/10/19 16:53:28 fabiankeil
109 * Add helper function to check if any content filters are enabled.
111 * Revision 1.95 2007/10/17 19:31:20 fabiankeil
112 * Omitting the zero chunk that ends the chunk transfer encoding seems
113 * to be the new black. Log the problem and continue filtering anyway.
115 * Revision 1.94 2007/09/29 13:20:20 fabiankeil
116 * Remove two redundant and one useless log messages.
118 * Revision 1.93 2007/09/29 10:21:16 fabiankeil
119 * - Move get_filter_function() from jcc.c to filters.c
120 * so the filter functions can be static.
121 * - Don't bother filtering body-less responses.
123 * Revision 1.92 2007/09/28 16:38:55 fabiankeil
124 * - Execute content filters through execute_content_filter().
125 * - Add prepare_for_filtering() so filter functions don't have to
126 * care about de-chunking and decompression. As a side effect this enables
127 * decompression for gif_deanimate_response() and jpeg_inspect_response().
128 * - Change remove_chunked_transfer_coding()'s return type to jb_err.
129 * Some clowns feel like chunking empty responses in which case
130 * (size == 0) is valid but previously would be interpreted as error.
132 * Revision 1.91 2007/09/02 15:31:20 fabiankeil
133 * Move match_portlist() from filter.c to urlmatch.c.
134 * It's used for url matching, not for filtering.
136 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
137 * Remove newline at the end of a log_error() message.
139 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
140 * #1763173 from Stefan Huehner: declare some more functions static.
142 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
143 * Add forward-override{} to change the forwarding settings through
144 * action sections. This is mainly interesting to forward different
145 * clients differently (for example based on User-Agent or request
148 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
149 * Make sure filters with dynamic jobs actually use them.
151 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
152 * - Introduce dynamic pcrs jobs that can resolve variables.
153 * - Don't run redirect functions more than once,
154 * unless they are activated more than once.
156 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
157 * - Log the content size after decompression in decompress_iob()
158 * instead of pcrs_filter_response().
160 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
161 * Use dedicated header filter actions instead of abusing "filter".
162 * Replace "filter-client-headers" and "filter-client-headers"
163 * with "server-header-filter" and "client-header-filter".
165 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
166 * New config option: enforce-blocks.
168 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
169 * - Fix port handling in acl_addr() and use a temporary acl spec
170 * copy so error messages don't contain a truncated version.
171 * - Log size of iob before and after decompression.
173 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
174 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
175 * - Hide the "Go there anyway" link for blocked CONNECT
176 * requests where going there anyway doesn't work anyway.
178 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
179 * - Save the reason for generating http_responses.
180 * - Block (+block) with status code 403 instead of 404.
181 * - Use a different kludge to remember a failed decompression.
183 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
184 * Search for Max-Forwards headers case-insensitive,
185 * don't generate the "501 unsupported" message for invalid
186 * Max-Forwards values and don't increase negative ones.
188 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
189 * - Add HEAD support to finish_http_response.
190 * - Add error favicon to internal HTML error messages.
192 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
193 * Mark *csp as immutable for is_untrusted_url()
194 * and is_imageurl(). Closes FR 1237736.
196 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
197 * Integrate a modified version of Wil Mahan's
198 * zlib patch (PR #895531).
200 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
201 * Fixed gcc43 conversion warnings,
202 * changed sprintf calls to snprintf.
204 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
205 * Adjust comment in pcrs_filter_response()
206 * to recent pcrs changes. Hohoho.
208 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
209 * Don't crash if pcre returns an error code
210 * that pcrs didn't expect. Fixes BR 1621173.
212 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
213 * Modified is_untrusted_url to complain in case of
214 * write errors and to give a reason when adding new
215 * entries to the trustfile. Closes FR 1097611.
217 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
218 * Skip empty filter files in pcrs_filter_response,
219 * but don't ignore the ones that come afterwards.
220 * Fixes parts of BR 1619208.
222 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
223 * Added some sanity checks for get_last_url().
224 * Fixed possible segfault caused by my last commit.
226 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
227 * Let get_last_url() catch https URLs as well.
229 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
230 * Make sure get_last_url() behaves like advertised
231 * and fast-redirects{} can be combined with redirect{}.
233 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
234 * Implemented +redirect{s@foo@bar@} to generate
235 * a redirect based on a rewritten version of the
238 * Revision 1.66 2006/09/23 13:26:38 roro
239 * Replace TABs by spaces in source code.
241 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
242 * Fix +redirect{}. Didn't work with -fast-redirects.
244 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
245 * Block requests for untrusted URLs with status
246 * code 403 instead of 200.
248 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
249 * Don't free p which is still in use and will be later
250 * freed by free_map(). Don't claim the referrer is unknown
251 * when the client didn't set one.
253 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
254 * Feature request 595948: Re-Filter logging in single line
256 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
257 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
259 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
260 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
261 * with what was really the latest development (the v_3_0_branch branch)
263 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
264 * Multiple filter file support
266 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
267 * A few memory leaks plugged (mostly on error paths)
269 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
270 * Add the ability to check jpeg images for invalid
271 * lengths of comment blocks. Defensive strategy
272 * against the exploit:
273 * Microsoft Security Bulletin MS04-028
274 * Buffer Overrun in JPEG Processing (GDI+) Could
275 * Allow Code Execution (833987)
276 * Enabled with +inspect-jpegs in actions files.
278 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
279 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
281 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
282 * Fixed bug #839859: "See why" link URL now gets url-encoded.
284 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
287 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
288 * Added basic support for OPTIONS and TRACE HTTP methods:
289 * New function direct_response which handles OPTIONS and
290 * TRACE requests whose Max-Forwards header field is zero.
292 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
293 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
295 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
296 * - Bugfix: Executing a filters without jobs no longer results in
297 * turing off *all* filters.
298 * - Security fix: Malicious web servers can't cause a seg fault
299 * through bogus chunk sizes anymore
301 * Revision 1.58 2002/04/24 02:11:17 oes
302 * Jon's multiple AF patch: url_actions now evaluates rules
305 * Revision 1.57 2002/04/08 20:38:34 swa
308 * Revision 1.56 2002/04/05 15:51:24 oes
309 * - bugfix: error-pages now get correct request protocol
310 * - fix for invalid HTML in trust info
312 * Revision 1.55 2002/04/02 16:13:51 oes
313 * Fix: No "Go there anyway" for SSL
315 * Revision 1.54 2002/04/02 14:55:56 oes
316 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
318 * Revision 1.53 2002/03/26 22:29:54 swa
319 * we have a new homepage!
321 * Revision 1.52 2002/03/24 16:35:57 jongfoster
324 * Revision 1.51 2002/03/24 15:23:33 jongfoster
327 * Revision 1.50 2002/03/24 13:25:43 swa
328 * name change related issues
330 * Revision 1.49 2002/03/16 20:29:14 oes
333 * Revision 1.48 2002/03/13 20:25:34 oes
334 * Better logging for content filters
336 * Revision 1.47 2002/03/13 00:30:52 jongfoster
338 * Added option of always sending redirect for imageblock,
339 * currently disabled with #if 0.
341 * Revision 1.46 2002/03/12 01:42:49 oes
342 * Introduced modular filters
344 * Revision 1.45 2002/03/08 16:47:50 oes
345 * Added choice beween GIF and PNG built-in images
347 * Revision 1.44 2002/03/07 03:49:31 oes
348 * - Fixed compiler warnings etc
349 * - Changed built-in images from GIF to PNG
350 * (with regard to Unisys patent issue)
351 * - Added a 4x4 pattern PNG which is less intrusive
352 * than the logo but also clearly marks the deleted banners
354 * Revision 1.43 2002/01/22 23:51:59 jongfoster
355 * Replacing strsav() with the safer string_append().
357 * Adding missing html_encode() to error message generators. Where encoded
358 * and unencoded versions of a string were provided, removing the unencoded
361 * Revision 1.42 2002/01/17 21:00:32 jongfoster
362 * Moving all our URL and URL pattern parsing code to urlmatch.c.
364 * Using a single, simple url_match(pattern,url) function - rather than
365 * the 3-line match routine which was repeated all over the place.
367 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
369 * Using parse_http_url() to parse URLs without faking a HTTP
370 * request line for parse_http_request().
372 * Revision 1.41 2001/11/13 00:14:07 jongfoster
373 * Fixing stupid bug now I've figured out what || means.
374 * (It always returns 0 or 1, not one of it's paramaters.)
376 * Revision 1.40 2001/10/26 17:37:55 oes
377 * - Re-enabled Netscape 200/404 bug workaround in block_url():
378 * - Removed OS/2 special case
379 * - Made block_url() independant from sed() having been run
380 * - Made trust_url independant from sed() having been run
381 * - Made is_imageurl independant from sed() having been run.
382 * It now checks User-Agent: and Accept: by itself.
385 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
386 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
387 * threads to call select() simultaneously. So, it's time to do a real, live,
388 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
389 * (native). Both versions will work, but using __OS2__ offers multi-threading.
391 * Revision 1.38 2001/10/23 21:32:33 jongfoster
392 * Adding error-checking to selected functions
394 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
395 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
396 * filters.c. Added a FIXME in front of the offending code. I'll gladly
397 * put in a better/more robust fix for all parties if one is presented...
398 * It seems that just returning 200 instead of 404 would pretty much fix
399 * it for everyone, but I don't know all the history of the problem.
401 * Revision 1.36 2001/10/10 16:44:16 oes
402 * Added match_portlist function
404 * Revision 1.35 2001/10/07 15:41:23 oes
405 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
407 * New function remove_chunked_transfer_coding that strips chunked
408 * transfer coding to plain and is called by pcrs_filter_response
409 * and gif_deanimate_response if neccessary
411 * Improved handling of zero-change re_filter runs
413 * pcrs_filter_response and gif_deanimate_response now remove
414 * chunked transfer codeing before processing the body.
416 * Revision 1.34 2001/09/20 15:49:36 steudten
418 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
419 * See cgi.c fill_template().
421 * Revision 1.33 2001/09/16 17:05:14 jongfoster
422 * Removing unused #include showarg.h
424 * Revision 1.32 2001/09/16 13:21:27 jongfoster
425 * Changes to use new list functions.
427 * Revision 1.31 2001/09/16 11:38:02 jongfoster
428 * Splitting fill_template() into 2 functions:
429 * template_load() loads the file
430 * template_fill() performs the PCRS regexps.
431 * This is because the CGI edit interface has a "table row"
432 * template which is used many times in the page - this
433 * change means it's only loaded from disk once.
435 * Revision 1.30 2001/09/16 11:00:10 jongfoster
436 * New function alloc_http_response, for symmetry with free_http_response
438 * Revision 1.29 2001/09/13 23:32:40 jongfoster
439 * Moving image data to cgi.c rather than cgi.h
440 * Fixing a GPF under Win32 (and any other OS that protects global
441 * constants from being written to).
443 * Revision 1.28 2001/09/10 10:18:51 oes
444 * Silenced compiler warnings
446 * Revision 1.27 2001/08/05 16:06:20 jongfoster
447 * Modifiying "struct map" so that there are now separate header and
448 * "map_entry" structures. This means that functions which modify a
449 * map no longer need to return a pointer to the modified map.
450 * Also, it no longer reverses the order of the entries (which may be
451 * important with some advanced template substitutions).
453 * Revision 1.26 2001/07/30 22:08:36 jongfoster
454 * Tidying up #defines:
455 * - All feature #defines are now of the form FEATURE_xxx
456 * - Permanently turned off WIN_GUI_EDIT
457 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
459 * Revision 1.25 2001/07/26 10:09:46 oes
460 * Made browser detection a little less naive
462 * Revision 1.24 2001/07/25 17:22:51 oes
463 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
465 * Revision 1.23 2001/07/23 13:40:12 oes
466 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
468 * Revision 1.22 2001/07/18 12:29:34 oes
469 * - Made gif_deanimate_response respect
470 * csp->action->string[ACTION_STRING_DEANIMATE]
471 * - Logging cosmetics
473 * Revision 1.21 2001/07/13 13:59:53 oes
474 * - Introduced gif_deanimate_response which shares the
475 * generic content modification interface of pcrs_filter_response
476 * and acts as a wrapper to deanimate.c:gif_deanimate()
477 * - Renamed re_process_buffer to pcrs_filter_response
478 * - pcrs_filter_response now returns NULL on failiure
479 * - Removed all #ifdef PCRS
481 * Revision 1.20 2001/07/01 17:01:04 oes
482 * Added comments and missing return statement in is_untrusted_url()
484 * Revision 1.19 2001/06/29 21:45:41 oes
485 * Indentation, CRLF->LF, Tab-> Space
487 * Revision 1.18 2001/06/29 13:27:38 oes
488 * - Cleaned up, renamed and reorderd functions
489 * and improved comments
492 * - Ported to CGI platform. Now delivers
493 * http_response or NULL
494 * - Unified HTML and GIF generation (moved image detection
495 * and GIF generation here from jcc.c:chat())
496 * - Fixed HTTP status to:
497 * - 403 (Forbidden) for the "blocked" HTML message
498 * - 200 (OK) for GIF answers
499 * - 302 (Redirect) for redirect to GIF
502 * - Ported to CGI platform. Now delivers
503 * http_response or NULL
504 * - Separated detection of untrusted URL into
505 * (bool)is_untrusted_url
506 * - Added enforcement of untrusted requests
508 * - Moved redirect_url() from cgi.c to here
509 * and ported it to the CGI platform
511 * - Removed logentry from cancelled commit
513 * Revision 1.17 2001/06/09 10:55:28 jongfoster
514 * Changing BUFSIZ ==> BUFFER_SIZE
516 * Revision 1.16 2001/06/07 23:10:26 jongfoster
517 * Allowing unanchored domain patterns to back off and retry
518 * if they partially match. Optimized right-anchored patterns.
519 * Moving ACL and forward files into config file.
520 * Replacing struct gateway with struct forward_spec
522 * Revision 1.15 2001/06/03 19:12:00 oes
523 * extracted-CGI relevant stuff
525 * Revision 1.14 2001/06/01 10:30:55 oes
526 * Added optional left-anchoring to domaincmp
528 * Revision 1.13 2001/05/31 21:21:30 jongfoster
529 * Permissionsfile / actions file changes:
530 * - Changed "permission" to "action" throughout
531 * - changes to file format to allow string parameters
532 * - Moved helper functions to actions.c
534 * Revision 1.12 2001/05/31 17:35:20 oes
536 * - Enhanced domain part globbing with infix and prefix asterisk
537 * matching and optional unanchored operation
539 * Revision 1.11 2001/05/29 11:53:23 oes
540 * "See why" link added to "blocked" page
542 * Revision 1.10 2001/05/29 09:50:24 jongfoster
543 * Unified blocklist/imagelist/permissionslist.
544 * File format is still under discussion, but the internal changes
547 * Also modified interceptor behaviour:
548 * - We now intercept all URLs beginning with one of the following
549 * prefixes (and *only* these prefixes):
551 * * http://ijbswa.sf.net/config/
552 * * http://ijbswa.sourceforge.net/config/
553 * - New interceptors "home page" - go to http://i.j.b/ to see it.
554 * - Internal changes so that intercepted and fast redirect pages
555 * are not replaced with an image.
556 * - Interceptors now have the option to send a binary page direct
557 * to the client. (i.e. ijb-send-banner uses this)
558 * - Implemented show-url-info interceptor. (Which is why I needed
559 * the above interceptors changes - a typical URL is
560 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
561 * The previous mechanism would not have intercepted that, and
562 * if it had been intercepted then it then it would have replaced
565 * Revision 1.9 2001/05/27 22:17:04 oes
567 * - re_process_buffer no longer writes the modified buffer
568 * to the client, which was very ugly. It now returns the
569 * buffer, which it is then written by chat.
571 * - content_length now adjusts the Content-Length: header
572 * for modified documents rather than crunch()ing it.
573 * (Length info in csp->content_length, which is 0 for
574 * unmodified documents)
576 * - For this to work, sed() is called twice when filtering.
578 * Revision 1.8 2001/05/26 17:13:28 jongfoster
579 * Filled in a function comment.
581 * Revision 1.7 2001/05/26 15:26:15 jongfoster
582 * ACL feature now provides more security by immediately dropping
583 * connections from untrusted hosts.
585 * Revision 1.6 2001/05/26 00:28:36 jongfoster
586 * Automatic reloading of config file.
587 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
588 * Most of the global variables have been moved to a new
589 * struct configuration_spec, accessed through csp->config->globalname
590 * Most of the globals remaining are used by the Win32 GUI.
592 * Revision 1.5 2001/05/25 22:34:30 jongfoster
595 * Revision 1.4 2001/05/22 18:46:04 oes
597 * - Enabled filtering banners by size rather than URL
598 * by adding patterns that replace all standard banner
599 * sizes with the "Junkbuster" gif to the re_filterfile
601 * - Enabled filtering WebBugs by providing a pattern
602 * which kills all 1x1 images
604 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
605 * which is selected by the (nonstandard and therefore
606 * capital) letter 'U' in the option string.
607 * It causes the quantifiers to be ungreedy by default.
608 * Appending a ? turns back to greedy (!).
610 * - Added a new interceptor ijb-send-banner, which
611 * sends back the "Junkbuster" gif. Without imagelist or
612 * MSIE detection support, or if tinygif = 1, or the
613 * URL isn't recognized as an imageurl, a lame HTML
614 * explanation is sent instead.
616 * - Added new feature, which permits blocking remote
617 * script redirects and firing back a local redirect
619 * The feature is conditionally compiled, i.e. it
620 * can be disabled with --disable-fast-redirects,
621 * plus it must be activated by a "fast-redirects"
622 * line in the config file, has its own log level
623 * and of course wants to be displayed by show-proxy-args
624 * Note: Boy, all the #ifdefs in 1001 locations and
625 * all the fumbling with configure.in and acconfig.h
626 * were *way* more work than the feature itself :-(
628 * - Because a generic redirect template was needed for
629 * this, tinygif = 3 now uses the same.
631 * - Moved GIFs, and other static HTTP response templates
636 * - Removed some >400 CRs again (Jon, you really worked
639 * Revision 1.3 2001/05/20 16:44:47 jongfoster
640 * Removing last hardcoded Junkbusters.com URLs.
642 * Revision 1.2 2001/05/20 01:21:20 jongfoster
643 * Version 2.9.4 checkin.
644 * - Merged popupfile and cookiefile, and added control over PCRS
645 * filtering, in new "permissionsfile".
646 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
647 * file error you now get a message box (in the Win32 GUI) rather
648 * than the program exiting with no explanation.
649 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
651 * - Removed tabs from "config"
652 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
653 * - Bumped up version number.
655 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
656 * Initial import of version 2.9.3 source tree
659 *********************************************************************/
665 #include <sys/types.h>
671 #ifdef HAVE_GETADDRINFO
673 #include <sys/socket.h>
674 #endif /* def HAVE_GETADDRINFO */
679 #endif /* ndef __OS2__ */
680 #include <netinet/in.h>
682 #include <winsock2.h>
683 #endif /* ndef _WIN32 */
687 #endif /* def __OS2__ */
695 #include "jbsockets.h"
696 #include "miscutil.h"
700 #include "deanimate.h"
701 #include "urlmatch.h"
708 const char filters_h_rcs[] = FILTERS_H_VERSION;
710 /* Fix a problem with Solaris. There should be no effect on other
712 * Solaris's isspace() is a macro which uses it's argument directly
713 * as an array index. Therefore we need to make sure that high-bit
714 * characters generate +ve values, and ideally we also want to make
715 * the argument match the declared parameter type of "int".
717 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
719 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
720 static jb_err prepare_for_filtering(struct client_state *csp);
723 #ifdef HAVE_GETADDRINFO
724 /*********************************************************************
726 * Function : sockaddr_storage_to_ip
728 * Description : Access internal structure of sockaddr_storage
731 * 1 : addr = socket address
732 * 2 : ip = IP address as array of octets in network order
733 * (it points into addr)
734 * 3 : len = length of IP address in octets
735 * 4 : port = port number in network order;
737 * Returns : 0 = no errror; otherwise
739 *********************************************************************/
740 int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, uint8_t **ip,
741 unsigned int *len, in_port_t **port)
748 switch (addr->ss_family)
758 &(( (struct sockaddr_in *) addr)->sin_addr.s_addr);
762 *port = &((struct sockaddr_in *) addr)->sin_port;
773 *ip = ( (struct sockaddr_in6 *) addr)->sin6_addr.s6_addr;
777 *port = &((struct sockaddr_in6 *) addr)->sin6_port;
782 /* Unsupported address family */
790 /*********************************************************************
792 * Function : match_sockaddr
794 * Description : Check whether address matches network (IP address and port)
797 * 1 : network = socket address of subnework
798 * 3 : netmask = network mask as socket address
799 * 2 : address = checked socket address against given network
801 * Returns : 0 = doesn't match; 1 = does match
803 *********************************************************************/
804 int match_sockaddr(const struct sockaddr_storage *network,
805 const struct sockaddr_storage *netmask,
806 const struct sockaddr_storage *address)
808 uint8_t *network_addr, *netmask_addr, *address_addr;
809 unsigned int addr_len;
810 in_port_t *network_port, *netmask_port, *address_port;
813 if (network->ss_family != netmask->ss_family)
815 /* This should never happen */
816 log_error(LOG_LEVEL_ERROR,
817 "Internal error at %s:%llu: network and netmask differ in family",
822 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
823 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
824 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
826 /* Check for family */
827 if (network->ss_family == AF_INET && address->ss_family == AF_INET6 &&
828 IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
830 /* Map AF_INET6 V4MAPPED address into AF_INET */
834 else if (network->ss_family == AF_INET6 && address->ss_family == AF_INET &&
835 IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
837 /* Map AF_INET6 V4MAPPED network into AF_INET */
842 else if (network->ss_family != address->ss_family)
847 /* XXX: Port check is signaled in netmask */
848 if (*netmask_port && *network_port != *address_port)
853 /* TODO: Optimize by checking by words insted of octets */
854 for (i=0; i < addr_len && netmask_addr[i]; i++)
856 if ( (network_addr[i] & netmask_addr[i]) !=
857 (address_addr[i] & netmask_addr[i]) )
865 #endif /* def HAVE_GETADDRINFO */
868 /*********************************************************************
870 * Function : block_acl
872 * Description : Block this request?
873 * Decide yes or no based on ACL file.
876 * 1 : dst = The proxy or gateway address this is going to.
877 * Or NULL to check all possible targets.
878 * 2 : csp = Current client state (buffers, headers, etc...)
879 * Also includes the client IP address.
881 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
883 *********************************************************************/
884 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
886 struct access_control_list *acl = csp->config->acl;
888 /* if not using an access control list, then permit the connection */
894 /* search the list */
898 #ifdef HAVE_GETADDRINFO
899 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
901 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
907 /* Just want to check if they have any access */
908 if (acl->action == ACL_PERMIT)
914 #ifdef HAVE_GETADDRINFO
915 /* XXX: Undefined acl->dst is full of zeros and should be
916 * considered as wildcard address.
917 * sockaddr_storage_to_ip() failes on such dst because of
918 * uknown sa_familly on glibc. However this test is not
921 * So, we signal the acl->dst is wildcard in wildcard_dst.
924 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
926 ((dst->addr & acl->dst->mask) == acl->dst->addr)
927 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
931 if (acl->action == ACL_PERMIT)
949 /*********************************************************************
951 * Function : acl_addr
953 * Description : Called from `load_config' to parse an ACL address.
956 * 1 : aspec = String specifying ACL address.
957 * 2 : aca = struct access_control_addr to fill in.
959 * Returns : 0 => Ok, everything else is an error.
961 *********************************************************************/
962 int acl_addr(const char *aspec, struct access_control_addr *aca)
965 #ifdef HAVE_GETADDRINFO
966 struct addrinfo hints, *result;
968 in_port_t *mask_port;
969 unsigned int addr_len;
972 #endif /* def HAVE_GETADDRINFO */
974 char *acl_spec = NULL;
976 #ifdef HAVE_GETADDRINFO
977 /* FIXME: Depend on ai_family */
985 * Use a temporary acl spec copy so we can log
986 * the unmodified original in case of parse errors.
988 acl_spec = strdup(aspec);
989 if (acl_spec == NULL)
991 /* XXX: This will be logged as parse error. */
995 if ((p = strchr(acl_spec, '/')) != NULL)
998 if (ijb_isdigit(*p) == 0)
1003 masklength = atoi(p);
1006 if ((masklength < 0) ||
1007 #ifdef HAVE_GETADDRINFO
1018 if (*acl_spec == '[' && NULL != (p = strchr(acl_spec, ']')))
1021 memmove(acl_spec, acl_spec + 1, (size_t) (p - acl_spec));
1030 p = strchr(acl_spec, ':');
1033 #ifdef HAVE_GETADDRINFO
1034 memset(&hints, 0, sizeof(struct addrinfo));
1035 hints.ai_family = AF_UNSPEC;
1036 hints.ai_socktype = SOCK_STREAM;
1038 i = getaddrinfo(acl_spec, (p) ? ++p : NULL, &hints, &result);
1043 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s", acl_spec, p,
1048 /* TODO: Allow multihomed hostnames */
1049 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
1050 freeaddrinfo(result);
1057 port = strtol(p, &endptr, 10);
1059 if (port <= 0 || port > 65535 || *endptr != '\0')
1066 aca->port = (unsigned long)port;
1068 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
1071 if (aca->addr == INADDR_NONE)
1073 /* XXX: This will be logged as parse error. */
1076 #endif /* def HAVE_GETADDRINFO */
1078 /* build the netmask */
1079 #ifdef HAVE_GETADDRINFO
1080 /* Clip masklength according current family */
1081 if (aca->addr.ss_family == AF_INET && masklength > 32)
1086 aca->mask.ss_family = aca->addr.ss_family;
1087 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
1094 /* Port number in ACL has been specified, check ports in future */
1098 /* XXX: This could be optimized to operate on whole words instead of octets
1099 * (128-bit CPU could do it in one iteration). */
1100 /* Octets after prefix can be ommitted because of previous initialization
1102 for (i=0; i < addr_len && masklength; i++)
1104 if (masklength >= 8)
1106 mask_data[i] = 0xFF;
1111 /* XXX: This assumes MSB of octet is on the left site. This should be
1112 * true for all architectures or solved on link layer of OSI model. */
1113 mask_data[i] = ~((1 << (8 - masklength)) - 1);
1120 for (i=1; i <= masklength ; i++)
1122 aca->mask |= (1U << (32 - i));
1125 /* now mask off the host portion of the ip address
1126 * (i.e. save on the network portion of the address).
1128 aca->addr = aca->addr & aca->mask;
1129 #endif /* def HAVE_GETADDRINFO */
1134 #endif /* def FEATURE_ACL */
1137 /*********************************************************************
1139 * Function : connect_port_is_forbidden
1141 * Description : Check to see if CONNECT requests to the destination
1142 * port of this request are forbidden. The check is
1143 * independend of the actual request method.
1146 * 1 : csp = Current client state (buffers, headers, etc...)
1148 * Returns : True if yes, false otherwise.
1150 *********************************************************************/
1151 int connect_port_is_forbidden(const struct client_state *csp)
1153 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
1154 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
1159 /*********************************************************************
1161 * Function : block_url
1163 * Description : Called from `chat'. Check to see if we need to block this.
1166 * 1 : csp = Current client state (buffers, headers, etc...)
1168 * Returns : NULL => unblocked, else HTTP block response
1170 *********************************************************************/
1171 struct http_response *block_url(struct client_state *csp)
1173 struct http_response *rsp;
1174 const char *new_content_type = NULL;
1177 * If it's not blocked, don't block it ;-)
1179 if ((csp->action->flags & ACTION_BLOCK) == 0)
1183 if (csp->action->flags & ACTION_REDIRECT)
1185 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
1188 * Else, prepare a response
1190 if (NULL == (rsp = alloc_http_response()))
1192 return cgi_error_memory();
1196 * If it's an image-url, send back an image or redirect
1197 * as specified by the relevant +image action
1199 #ifdef FEATURE_IMAGE_BLOCKING
1200 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
1201 && is_imageurl(csp))
1204 /* determine HOW images should be blocked */
1205 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
1207 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1209 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
1211 #if 1 /* Two alternative strategies, use this one for now: */
1213 /* and handle accordingly: */
1214 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1216 rsp->status = strdup("403 Request blocked by Privoxy");
1217 if (rsp->status == NULL)
1219 free_http_response(rsp);
1220 return cgi_error_memory();
1222 rsp->body = bindup(image_pattern_data, image_pattern_length);
1223 if (rsp->body == NULL)
1225 free_http_response(rsp);
1226 return cgi_error_memory();
1228 rsp->content_length = image_pattern_length;
1230 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1232 free_http_response(rsp);
1233 return cgi_error_memory();
1237 else if (0 == strcmpic(p, "blank"))
1239 rsp->status = strdup("403 Request blocked by Privoxy");
1240 if (rsp->status == NULL)
1242 free_http_response(rsp);
1243 return cgi_error_memory();
1245 rsp->body = bindup(image_blank_data, image_blank_length);
1246 if (rsp->body == NULL)
1248 free_http_response(rsp);
1249 return cgi_error_memory();
1251 rsp->content_length = image_blank_length;
1253 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1255 free_http_response(rsp);
1256 return cgi_error_memory();
1262 rsp->status = strdup("302 Local Redirect from Privoxy");
1263 if (rsp->status == NULL)
1265 free_http_response(rsp);
1266 return cgi_error_memory();
1269 if (enlist_unique_header(rsp->headers, "Location", p))
1271 free_http_response(rsp);
1272 return cgi_error_memory();
1276 #else /* Following code is disabled for now */
1278 /* and handle accordingly: */
1279 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1281 p = CGI_PREFIX "send-banner?type=pattern";
1283 else if (0 == strcmpic(p, "blank"))
1285 p = CGI_PREFIX "send-banner?type=blank";
1287 rsp->status = strdup("302 Local Redirect from Privoxy");
1288 if (rsp->status == NULL)
1290 free_http_response(rsp);
1291 return cgi_error_memory();
1294 if (enlist_unique_header(rsp->headers, "Location", p))
1296 free_http_response(rsp);
1297 return cgi_error_memory();
1299 #endif /* Preceeding code is disabled for now */
1301 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1304 * Send empty document.
1306 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1309 rsp->body = strdup(" ");
1310 rsp->content_length = 1;
1312 rsp->status = strdup("403 Request blocked by Privoxy");
1313 if (rsp->status == NULL)
1315 free_http_response(rsp);
1316 return cgi_error_memory();
1318 if (new_content_type != 0)
1320 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1321 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1323 free_http_response(rsp);
1324 return cgi_error_memory();
1329 #endif /* def FEATURE_IMAGE_BLOCKING */
1332 * Else, generate an HTML "blocked" message:
1336 struct map * exports;
1340 * Workaround for stupid Netscape bug which prevents
1341 * pages from being displayed if loading a referenced
1342 * JavaScript or style sheet fails. So make it appear
1343 * as if it succeeded.
1345 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1346 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1347 && !strstr(p, "Gecko") /* save Mozilla, */
1348 && !strstr(p, "compatible") /* MSIE */
1349 && !strstr(p, "Opera")) /* and Opera. */
1351 rsp->status = strdup("200 Request for blocked URL");
1355 rsp->status = strdup("403 Request for blocked URL");
1358 if (rsp->status == NULL)
1360 free_http_response(rsp);
1361 return cgi_error_memory();
1364 exports = default_exports(csp, NULL);
1365 if (exports == NULL)
1367 free_http_response(rsp);
1368 return cgi_error_memory();
1371 #ifdef FEATURE_FORCE_LOAD
1372 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1374 * Export the force conditional block killer if
1376 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1377 * - Privoxy is configured to enforce blocks, or
1378 * - it's a CONNECT request and enforcing wouldn't work anyway.
1380 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1381 || (0 == strcmpic(csp->http->gpc, "connect")))
1382 #endif /* ndef FEATURE_FORCE_LOAD */
1384 err = map_block_killer(exports, "force-support");
1387 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1388 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1389 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1390 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1393 const char *block_reason;
1394 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
1396 block_reason = csp->action->string[ACTION_STRING_BLOCK];
1400 assert(connect_port_is_forbidden(csp));
1401 block_reason = "Forbidden CONNECT port.";
1403 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
1408 free_http_response(rsp);
1409 return cgi_error_memory();
1412 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1415 free_http_response(rsp);
1416 return cgi_error_memory();
1419 rsp->reason = RSP_REASON_BLOCKED;
1421 return finish_http_response(csp, rsp);
1426 #ifdef FEATURE_TRUST
1427 /*********************************************************************
1429 * Function : trust_url FIXME: I should be called distrust_url
1431 * Description : Calls is_untrusted_url to determine if the URL is trusted
1432 * and if not, returns a HTTP 403 response with a reject message.
1435 * 1 : csp = Current client state (buffers, headers, etc...)
1437 * Returns : NULL => trusted, else http_response.
1439 *********************************************************************/
1440 struct http_response *trust_url(struct client_state *csp)
1442 struct http_response *rsp;
1443 struct map * exports;
1444 char buf[BUFFER_SIZE];
1446 struct url_spec **tl;
1451 * Don't bother to work on trusted URLs
1453 if (!is_untrusted_url(csp))
1459 * Else, prepare a response:
1461 if (NULL == (rsp = alloc_http_response()))
1463 return cgi_error_memory();
1466 rsp->status = strdup("403 Request blocked by Privoxy");
1467 exports = default_exports(csp, NULL);
1468 if (exports == NULL || rsp->status == NULL)
1470 free_http_response(rsp);
1471 return cgi_error_memory();
1475 * Export the protocol, host, port, and referrer information
1477 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1478 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1479 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1481 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1483 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1487 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1493 free_http_response(rsp);
1494 return cgi_error_memory();
1498 * Export the trust list
1501 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1503 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1504 string_append(&p, buf);
1506 err = map(exports, "trusted-referrers", 1, p, 0);
1511 free_http_response(rsp);
1512 return cgi_error_memory();
1516 * Export the trust info, if available
1518 if (csp->config->trust_info->first)
1520 struct list_entry *l;
1523 for (l = csp->config->trust_info->first; l ; l = l->next)
1525 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1526 string_append(&p, buf);
1528 err = map(exports, "trust-info", 1, p, 0);
1532 err = map_block_killer(exports, "have-trust-info");
1538 free_http_response(rsp);
1539 return cgi_error_memory();
1543 * Export the force conditional block killer if
1545 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1546 * - Privoxy is configured to enforce blocks, or
1547 * - it's a CONNECT request and enforcing wouldn't work anyway.
1549 #ifdef FEATURE_FORCE_LOAD
1550 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1551 || (0 == strcmpic(csp->http->gpc, "connect")))
1553 err = map_block_killer(exports, "force-support");
1557 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1559 #else /* ifndef FEATURE_FORCE_LOAD */
1560 err = map_block_killer(exports, "force-support");
1561 #endif /* ndef FEATURE_FORCE_LOAD */
1566 free_http_response(rsp);
1567 return cgi_error_memory();
1571 * Build the response
1573 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1576 free_http_response(rsp);
1577 return cgi_error_memory();
1579 rsp->reason = RSP_REASON_UNTRUSTED;
1581 return finish_http_response(csp, rsp);
1583 #endif /* def FEATURE_TRUST */
1586 /*********************************************************************
1588 * Function : compile_dynamic_pcrs_job_list
1590 * Description : Compiles a dynamic pcrs job list (one with variables
1591 * resolved at request time)
1594 * 1 : csp = Current client state (buffers, headers, etc...)
1595 * 2 : b = The filter list to compile
1597 * Returns : NULL in case of errors, otherwise the
1600 *********************************************************************/
1601 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1603 struct list_entry *pattern;
1604 pcrs_job *job_list = NULL;
1605 pcrs_job *dummy = NULL;
1606 pcrs_job *lastjob = NULL;
1609 const struct pcrs_variable variables[] =
1611 {"url", csp->http->url, 1},
1612 {"path", csp->http->path, 1},
1613 {"host", csp->http->host, 1},
1614 {"origin", csp->ip_addr_str, 1},
1618 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1620 assert(pattern->str != NULL);
1622 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1626 log_error(LOG_LEVEL_ERROR,
1627 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1628 pattern->str, b->name, pcrs_strerror(error));
1633 if (error == PCRS_WARN_TRUNCATION)
1635 log_error(LOG_LEVEL_ERROR,
1636 "At least one of the variables in \'%s\' had to "
1637 "be truncated before compilation", pattern->str);
1639 if (job_list == NULL)
1645 lastjob->next = dummy;
1655 /*********************************************************************
1657 * Function : rewrite_url
1659 * Description : Rewrites a URL with a single pcrs command
1660 * and returns the result if it differs from the
1661 * original and isn't obviously invalid.
1664 * 1 : old_url = URL to rewrite.
1665 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1668 * Returns : NULL if the pcrs_command didn't change the url, or
1669 * the result of the modification.
1671 *********************************************************************/
1672 char *rewrite_url(char *old_url, const char *pcrs_command)
1674 char *new_url = NULL;
1678 assert(pcrs_command);
1680 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1684 log_error(LOG_LEVEL_REDIRECTS,
1685 "pcrs command \"%s\" didn't change \"%s\".",
1686 pcrs_command, old_url);
1691 log_error(LOG_LEVEL_REDIRECTS,
1692 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1693 pcrs_command, old_url, pcrs_strerror(hits));
1696 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1698 log_error(LOG_LEVEL_ERROR,
1699 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1700 "but the result doesn't look like a valid URL and will be ignored.",
1701 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1706 log_error(LOG_LEVEL_REDIRECTS,
1707 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1708 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1716 #ifdef FEATURE_FAST_REDIRECTS
1717 /*********************************************************************
1719 * Function : get_last_url
1721 * Description : Search for the last URL inside a string.
1722 * If the string already is a URL, it will
1723 * be the first URL found.
1726 * 1 : subject = the string to check
1727 * 2 : redirect_mode = +fast-redirect{} mode
1729 * Returns : NULL if no URL was found, or
1730 * the last URL found.
1732 *********************************************************************/
1733 char *get_last_url(char *subject, const char *redirect_mode)
1735 char *new_url = NULL;
1739 assert(redirect_mode);
1741 subject = strdup(subject);
1742 if (subject == NULL)
1744 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1748 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1750 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1751 new_url = url_decode(subject);
1752 if (new_url != NULL)
1759 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1763 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1766 * Find the last URL encoded in the request
1769 while ((tmp = strstr(tmp, "http://")) != NULL)
1773 tmp = (new_url != NULL) ? new_url : subject;
1774 while ((tmp = strstr(tmp, "https://")) != NULL)
1779 if ((new_url != NULL)
1780 && ( (new_url != subject)
1781 || (0 == strncmpic(subject, "http://", 7))
1782 || (0 == strncmpic(subject, "https://", 8))
1786 * Return new URL if we found a redirect
1787 * or if the subject already was a URL.
1789 * The second case makes sure that we can
1790 * chain get_last_url after another redirection check
1791 * (like rewrite_url) without losing earlier redirects.
1793 new_url = strdup(new_url);
1802 #endif /* def FEATURE_FAST_REDIRECTS */
1805 /*********************************************************************
1807 * Function : redirect_url
1809 * Description : Checks if Privoxy should answer the request with
1810 * a HTTP redirect and generates the redirect if
1814 * 1 : csp = Current client state (buffers, headers, etc...)
1816 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1818 *********************************************************************/
1819 struct http_response *redirect_url(struct client_state *csp)
1821 struct http_response *rsp;
1822 #ifdef FEATURE_FAST_REDIRECTS
1824 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1825 * as compile-time option? The user can easily disable
1826 * it in his action file.
1828 char * redirect_mode;
1829 #endif /* def FEATURE_FAST_REDIRECTS */
1830 char *old_url = NULL;
1831 char *new_url = NULL;
1832 char *redirection_string;
1834 if ((csp->action->flags & ACTION_REDIRECT))
1836 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1839 * If the redirection string begins with 's',
1840 * assume it's a pcrs command, otherwise treat it as
1841 * properly formatted URL and use it for the redirection
1844 * According to RFC 2616 section 14.30 the URL
1845 * has to be absolute and if the user tries:
1846 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1847 * she would get undefined results anyway.
1851 if (*redirection_string == 's')
1853 old_url = csp->http->url;
1854 new_url = rewrite_url(old_url, redirection_string);
1858 log_error(LOG_LEVEL_REDIRECTS,
1859 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1860 redirection_string);
1861 new_url = strdup(redirection_string);
1865 #ifdef FEATURE_FAST_REDIRECTS
1866 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1868 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1871 * If it exists, use the previously rewritten URL as input
1872 * otherwise just use the old path.
1874 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1875 new_url = get_last_url(old_url, redirect_mode);
1880 * Disable redirect checkers, so that they
1881 * will be only run more than once if the user
1882 * also enables them through tags.
1884 * From a performance point of view
1885 * it doesn't matter, but the duplicated
1886 * log messages are annoying.
1888 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1889 #endif /* def FEATURE_FAST_REDIRECTS */
1890 csp->action->flags &= ~ACTION_REDIRECT;
1892 /* Did any redirect action trigger? */
1895 if (0 == strcmpic(new_url, csp->http->url))
1897 log_error(LOG_LEVEL_ERROR,
1898 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1899 csp->http->url, new_url);
1904 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1906 if (NULL == (rsp = alloc_http_response()))
1909 return cgi_error_memory();
1912 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1913 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1916 free_http_response(rsp);
1917 return cgi_error_memory();
1919 rsp->reason = RSP_REASON_REDIRECTED;
1922 return finish_http_response(csp, rsp);
1926 /* Only reached if no redirect is required */
1932 #ifdef FEATURE_IMAGE_BLOCKING
1933 /*********************************************************************
1935 * Function : is_imageurl
1937 * Description : Given a URL, decide whether it is an image or not,
1938 * using either the info from a previous +image action
1939 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1940 * is MSIE and not on a Mac, tell from the browser's accept
1944 * 1 : csp = Current client state (buffers, headers, etc...)
1946 * Returns : True (nonzero) if URL is an image, false (0)
1949 *********************************************************************/
1950 int is_imageurl(const struct client_state *csp)
1952 #ifdef FEATURE_IMAGE_DETECT_MSIE
1955 tmp = get_header_value(csp->headers, "User-Agent:");
1956 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1958 tmp = get_header_value(csp->headers, "Accept:");
1959 if (tmp && strstr(tmp, "image/gif"))
1961 /* Client will accept HTML. If this seems counterintuitive,
1971 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1973 return ((csp->action->flags & ACTION_IMAGE) != 0);
1976 #endif /* def FEATURE_IMAGE_BLOCKING */
1979 #ifdef FEATURE_TRUST
1980 /*********************************************************************
1982 * Function : is_untrusted_url
1984 * Description : Should we "distrust" this URL (and block it)?
1986 * Yes if it matches a line in the trustfile, or if the
1987 * referrer matches a line starting with "+" in the
1992 * 1 : csp = Current client state (buffers, headers, etc...)
1994 * Returns : 0 => trusted, 1 => untrusted
1996 *********************************************************************/
1997 int is_untrusted_url(const struct client_state *csp)
1999 struct file_list *fl;
2000 struct block_spec *b;
2001 struct url_spec **trusted_url;
2002 struct http_request rhttp[1];
2003 const char * referer;
2007 * If we don't have a trustlist, we trust everybody
2009 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
2014 memset(rhttp, '\0', sizeof(*rhttp));
2017 * Do we trust the request URL itself?
2019 for (b = b->next; b ; b = b->next)
2021 if (url_match(b->url, csp->http))
2027 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
2029 /* no referrer was supplied */
2035 * If not, do we maybe trust its referrer?
2037 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
2043 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
2045 if (url_match(*trusted_url, rhttp))
2047 /* if the URL's referrer is from a trusted referrer, then
2048 * add the target spec to the trustfile as an unblocked
2049 * domain and return 0 (which means it's OK).
2054 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
2058 char * new_entry = strdup("~");
2060 string_append(&new_entry, csp->http->hostport);
2062 path = csp->http->path;
2063 if ( (path[0] == '/')
2065 && ((path_end = strchr(path + 2, '/')) != NULL))
2067 /* since this path points into a user's home space
2068 * be sure to include this spec in the trustfile.
2070 long path_len = path_end - path; /* save offset */
2071 path = strdup(path); /* Copy string */
2074 path_end = path + path_len; /* regenerate ptr to new buffer */
2075 *(path_end + 1) = '\0'; /* Truncate path after '/' */
2077 string_join(&new_entry, path);
2081 * Give a reason for generating this entry.
2083 string_append(&new_entry, " # Trusted referrer was: ");
2084 string_append(&new_entry, referer);
2086 if (new_entry != NULL)
2088 if (-1 == fprintf(fp, "%s\n", new_entry))
2090 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
2091 new_entry, csp->config->trustfile);
2097 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
2098 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
2105 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
2106 csp->http->hostport, csp->config->trustfile);
2114 #endif /* def FEATURE_TRUST */
2117 /*********************************************************************
2119 * Function : pcrs_filter_response
2121 * Description : Execute all text substitutions from all applying
2122 * +filter actions on the text buffer that's been
2123 * accumulated in csp->iob->buf.
2126 * 1 : csp = Current client state (buffers, headers, etc...)
2128 * Returns : a pointer to the (newly allocated) modified buffer.
2129 * or NULL if there were no hits or something went wrong
2131 *********************************************************************/
2132 static char *pcrs_filter_response(struct client_state *csp)
2135 size_t size, prev_size;
2141 struct file_list *fl;
2142 struct re_filterfile_spec *b;
2143 struct list_entry *filtername;
2145 int i, found_filters = 0;
2150 if (csp->iob->cur >= csp->iob->eod)
2156 * Need to check the set of re_filterfiles...
2158 for (i = 0; i < MAX_AF_FILES; i++)
2171 if (0 == found_filters)
2173 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2174 "content filtering enabled, but no content filters available.");
2178 size = (size_t)(csp->iob->eod - csp->iob->cur);
2179 old = csp->iob->cur;
2181 for (i = 0; i < MAX_AF_FILES; i++)
2184 if ((NULL == fl) || (NULL == fl->f))
2187 * Either there are no filter files
2188 * left, or this filter file just
2189 * contains no valid filters.
2191 * Continue to be sure we don't miss
2192 * valid filter files that are chained
2193 * after empty or invalid ones.
2198 * For all applying +filter actions, look if a filter by that
2199 * name exists and if yes, execute it's pcrs_joblist on the
2202 for (b = fl->f; b; b = b->next)
2204 if (b->type != FT_CONTENT_FILTER)
2206 /* Skip header filters */
2210 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
2211 filtername ; filtername = filtername->next)
2213 if (strcmp(b->name, filtername->str) == 0)
2215 int current_hits = 0; /* Number of hits caused by this filter */
2216 int job_number = 0; /* Which job we're currently executing */
2217 int job_hits = 0; /* How many hits the current job caused */
2218 pcrs_job *joblist = b->joblist;
2220 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2222 if (NULL == joblist)
2224 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2229 /* Apply all jobs from the joblist */
2230 for (job = joblist; NULL != job; job = job->next)
2233 job_hits = pcrs_execute(job, old, size, &new, &size);
2238 * That went well. Continue filtering
2239 * and use the result of this job as
2240 * input for the next one.
2242 current_hits += job_hits;
2243 if (old != csp->iob->cur)
2252 * This job caused an unexpected error. Inform the user
2253 * and skip the rest of the jobs in this filter. We could
2254 * continue with the next job, but usually the jobs
2255 * depend on each other or are similar enough to
2256 * fail for the same reason.
2258 * At the moment our pcrs expects the error codes of pcre 3.4,
2259 * but newer pcre versions can return additional error codes.
2260 * As a result pcrs_strerror()'s error message might be
2261 * "Unknown error ...", therefore we print the numerical value
2264 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2265 * should we use LOG_LEVEL_RE_FILTER instead?
2267 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2268 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2273 if (b->dynamic) pcrs_free_joblist(joblist);
2275 log_error(LOG_LEVEL_RE_FILTER,
2276 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2277 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2279 hits += current_hits;
2286 * If there were no hits, destroy our copy and let
2287 * chat() use the original in csp->iob
2295 csp->flags |= CSP_FLAG_MODIFIED;
2296 csp->content_length = size;
2304 /*********************************************************************
2306 * Function : gif_deanimate_response
2308 * Description : Deanimate the GIF image that has been accumulated in
2309 * csp->iob->buf, set csp->content_length to the modified
2310 * size and raise the CSP_FLAG_MODIFIED flag.
2313 * 1 : csp = Current client state (buffers, headers, etc...)
2315 * Returns : a pointer to the (newly allocated) modified buffer.
2316 * or NULL in case something went wrong.
2318 *********************************************************************/
2319 static char *gif_deanimate_response(struct client_state *csp)
2321 struct binbuffer *in, *out;
2325 size = (size_t)(csp->iob->eod - csp->iob->cur);
2327 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2328 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2330 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2334 in->buffer = csp->iob->cur;
2337 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2339 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2346 if ((int)size == out->offset)
2348 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2352 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2354 csp->content_length = out->offset;
2355 csp->flags |= CSP_FLAG_MODIFIED;
2365 /*********************************************************************
2367 * Function : get_filter_function
2369 * Description : Decides which content filter function has
2370 * to be applied (if any).
2373 * 1 : csp = Current client state (buffers, headers, etc...)
2375 * Returns : The content filter function to run, or
2376 * NULL if no content filter is active
2378 *********************************************************************/
2379 filter_function_ptr get_filter_function(struct client_state *csp)
2381 filter_function_ptr filter_function = NULL;
2384 * Are we enabling text mode by force?
2386 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2389 * Do we really have to?
2391 if (csp->content_type & CT_TEXT)
2393 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2397 csp->content_type |= CT_TEXT;
2398 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2402 if (!(csp->content_type & CT_DECLARED))
2405 * The server didn't bother to declare a MIME-Type.
2406 * Assume it's text that can be filtered.
2408 * This also regulary happens with 304 responses,
2409 * therefore logging anything here would cause
2412 csp->content_type |= CT_TEXT;
2416 * Choose the applying filter function based on
2417 * the content type and action settings.
2419 if ((csp->content_type & CT_TEXT) &&
2420 (csp->rlist != NULL) &&
2421 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2423 filter_function = pcrs_filter_response;
2425 else if ((csp->content_type & CT_GIF) &&
2426 (csp->action->flags & ACTION_DEANIMATE))
2428 filter_function = gif_deanimate_response;
2431 return filter_function;
2435 /*********************************************************************
2437 * Function : remove_chunked_transfer_coding
2439 * Description : In-situ remove the "chunked" transfer coding as defined
2440 * in rfc2616 from a buffer.
2443 * 1 : buffer = Pointer to the text buffer
2444 * 2 : size = In: Number of bytes to be processed,
2445 * Out: Number of bytes after de-chunking.
2446 * (undefined in case of errors)
2448 * Returns : JB_ERR_OK for success,
2449 * JB_ERR_PARSE otherwise
2451 *********************************************************************/
2452 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2455 unsigned int chunksize = 0;
2456 char *from_p, *to_p;
2459 from_p = to_p = buffer;
2461 if (sscanf(buffer, "%x", &chunksize) != 1)
2463 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2464 return JB_ERR_PARSE;
2467 while (chunksize > 0U)
2469 if (NULL == (from_p = strstr(from_p, "\r\n")))
2471 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2472 return JB_ERR_PARSE;
2475 if ((newsize += chunksize) >= *size)
2477 log_error(LOG_LEVEL_ERROR,
2478 "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding",
2480 return JB_ERR_PARSE;
2484 memmove(to_p, from_p, (size_t) chunksize);
2485 to_p = buffer + newsize;
2486 from_p += chunksize + 2;
2488 if (sscanf(from_p, "%x", &chunksize) != 1)
2490 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2495 /* XXX: Should get its own loglevel. */
2496 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2505 /*********************************************************************
2507 * Function : prepare_for_filtering
2509 * Description : If necessary, de-chunks and decompresses
2510 * the content so it can get filterd.
2513 * 1 : csp = Current client state (buffers, headers, etc...)
2515 * Returns : JB_ERR_OK for success,
2516 * JB_ERR_PARSE otherwise
2518 *********************************************************************/
2519 static jb_err prepare_for_filtering(struct client_state *csp)
2521 jb_err err = JB_ERR_OK;
2524 * If the body has a "chunked" transfer-encoding,
2525 * get rid of it, adjusting size and iob->eod
2527 if (csp->flags & CSP_FLAG_CHUNKED)
2529 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2531 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2532 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2533 if (JB_ERR_OK == err)
2535 csp->iob->eod = csp->iob->cur + size;
2536 csp->flags |= CSP_FLAG_MODIFIED;
2540 return JB_ERR_PARSE;
2546 * If the body has a supported transfer-encoding,
2547 * decompress it, adjusting size and iob->eod.
2549 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2551 if (0 == csp->iob->eod - csp->iob->cur)
2553 /* Nothing left after de-chunking. */
2557 err = decompress_iob(csp);
2559 if (JB_ERR_OK == err)
2561 csp->flags |= CSP_FLAG_MODIFIED;
2562 csp->content_type &= ~CT_TABOO;
2567 * Unset CT_GZIP and CT_DEFLATE to remember not
2568 * to modify the Content-Encoding header later.
2570 csp->content_type &= ~CT_GZIP;
2571 csp->content_type &= ~CT_DEFLATE;
2580 /*********************************************************************
2582 * Function : execute_content_filter
2584 * Description : Executes a given content filter.
2587 * 1 : csp = Current client state (buffers, headers, etc...)
2588 * 2 : content_filter = The filter function to execute
2590 * Returns : Pointer to the modified buffer, or
2591 * NULL if filtering failed or wasn't necessary.
2593 *********************************************************************/
2594 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2596 if (0 == csp->iob->eod - csp->iob->cur)
2599 * No content (probably status code 301, 302 ...),
2600 * no filtering necessary.
2605 if (JB_ERR_OK != prepare_for_filtering(csp))
2608 * failed to de-chunk or decompress.
2613 if (0 == csp->iob->eod - csp->iob->cur)
2616 * Clown alarm: chunked and/or compressed nothing delivered.
2621 return ((*content_filter)(csp));
2625 /*********************************************************************
2627 * Function : get_url_actions
2629 * Description : Gets the actions for this URL.
2632 * 1 : csp = Current client state (buffers, headers, etc...)
2633 * 2 : http = http_request request for blocked URLs
2637 *********************************************************************/
2638 void get_url_actions(struct client_state *csp, struct http_request *http)
2640 struct file_list *fl;
2641 struct url_actions *b;
2644 init_current_action(csp->action);
2646 for (i = 0; i < MAX_AF_FILES; i++)
2648 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2653 apply_url_actions(csp->action, http, b);
2660 /*********************************************************************
2662 * Function : apply_url_actions
2664 * Description : Applies a list of URL actions.
2667 * 1 : action = Destination.
2668 * 2 : http = Current URL
2669 * 3 : b = list of URL actions to apply
2673 *********************************************************************/
2674 void apply_url_actions(struct current_action_spec *action,
2675 struct http_request *http,
2676 struct url_actions *b)
2680 /* Should never happen */
2684 for (b = b->next; NULL != b; b = b->next)
2686 if (url_match(b->url, http))
2688 merge_current_action(action, b->action);
2694 /*********************************************************************
2696 * Function : get_forward_override_settings
2698 * Description : Returns forward settings as specified with the
2699 * forward-override{} action. forward-override accepts
2700 * forward lines similar to the one used in the
2701 * configuration file, but without the URL pattern.
2707 * in the configuration file can be replaced with
2708 * the action section:
2710 * {+forward-override{forward .}}
2714 * 1 : csp = Current client state (buffers, headers, etc...)
2716 * Returns : Pointer to forwarding structure in case of success.
2717 * Invalid syntax is fatal.
2719 *********************************************************************/
2720 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2722 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2723 char forward_settings[BUFFER_SIZE];
2724 char *http_parent = NULL;
2725 /* variable names were chosen for consistency reasons. */
2726 struct forward_spec *fwd = NULL;
2730 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2731 /* Should be enforced by load_one_actions_file() */
2732 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2734 /* Create a copy ssplit can modify */
2735 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2737 if (NULL != csp->fwd)
2740 * XXX: Currently necessary to prevent memory
2741 * leaks when the show-url-info cgi page is visited.
2743 unload_forward_spec(csp->fwd);
2747 * allocate a new forward node, valid only for
2748 * the lifetime of this request. Save its location
2749 * in csp as well, so sweep() can free it later on.
2751 fwd = csp->fwd = zalloc(sizeof(*fwd));
2754 log_error(LOG_LEVEL_FATAL,
2755 "can't allocate memory for forward-override{%s}", forward_override_line);
2756 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2760 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2761 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2763 fwd->type = SOCKS_NONE;
2765 /* Parse the parent HTTP proxy host:port */
2766 http_parent = vec[1];
2769 else if (vec_count == 3)
2771 char *socks_proxy = NULL;
2773 if (!strcasecmp(vec[0], "forward-socks4"))
2775 fwd->type = SOCKS_4;
2776 socks_proxy = vec[1];
2778 else if (!strcasecmp(vec[0], "forward-socks4a"))
2780 fwd->type = SOCKS_4A;
2781 socks_proxy = vec[1];
2783 else if (!strcasecmp(vec[0], "forward-socks5"))
2785 fwd->type = SOCKS_5;
2786 socks_proxy = vec[1];
2789 if (NULL != socks_proxy)
2791 /* Parse the SOCKS proxy host[:port] */
2792 fwd->gateway_host = strdup(socks_proxy);
2794 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2796 *socks_proxy++ = '\0';
2797 fwd->gateway_port = (int)strtol(socks_proxy, NULL, 0);
2800 if (fwd->gateway_port <= 0)
2802 fwd->gateway_port = 1080;
2805 http_parent = vec[2];
2809 if (NULL == http_parent)
2811 log_error(LOG_LEVEL_FATAL,
2812 "Invalid forward-override syntax in: %s", forward_override_line);
2813 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2816 /* Parse http forwarding settings */
2817 if (strcmp(http_parent, ".") != 0)
2819 fwd->forward_host = strdup(http_parent);
2821 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2823 *http_parent++ = '\0';
2824 fwd->forward_port = (int)strtol(http_parent, NULL, 0);
2827 if (fwd->forward_port <= 0)
2829 fwd->forward_port = 8000;
2833 assert (NULL != fwd);
2835 log_error(LOG_LEVEL_CONNECT,
2836 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2842 /*********************************************************************
2844 * Function : forward_url
2846 * Description : Should we forward this to another proxy?
2849 * 1 : csp = Current client state (buffers, headers, etc...)
2850 * 2 : http = http_request request for current URL
2852 * Returns : Pointer to forwarding information.
2854 *********************************************************************/
2855 const struct forward_spec *forward_url(struct client_state *csp,
2856 const struct http_request *http)
2858 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2859 struct forward_spec *fwd = csp->config->forward;
2861 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2863 return get_forward_override_settings(csp);
2873 if (url_match(fwd->url, http))
2884 /*********************************************************************
2886 * Function : direct_response
2888 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2889 * request and if so, return a HTTP 501 to the client.
2891 * FIXME: I have a stupid name and I should handle the
2892 * requests properly. Still, what we do here is rfc-
2893 * compliant, whereas ignoring or forwarding are not.
2896 * 1 : csp = Current client state (buffers, headers, etc...)
2898 * Returns : http_response if , NULL if nonmatch or handler fail
2900 *********************************************************************/
2901 struct http_response *direct_response(struct client_state *csp)
2903 struct http_response *rsp;
2904 struct list_entry *p;
2906 if ((0 == strcmpic(csp->http->gpc, "trace"))
2907 || (0 == strcmpic(csp->http->gpc, "options")))
2909 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2911 if (!strncmpic("Max-Forwards:", p->str, 13))
2913 unsigned int max_forwards;
2916 * If it's a Max-Forwards value of zero,
2917 * we have to intercept the request.
2919 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2922 * FIXME: We could handle at least TRACE here,
2923 * but that would require a verbatim copy of
2924 * the request which we don't have anymore
2926 log_error(LOG_LEVEL_HEADER,
2927 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2930 /* Get mem for response or fail*/
2931 if (NULL == (rsp = alloc_http_response()))
2933 return cgi_error_memory();
2936 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2938 free_http_response(rsp);
2939 return cgi_error_memory();
2943 rsp->reason = RSP_REASON_UNSUPPORTED;
2945 return(finish_http_response(csp, rsp));
2954 /*********************************************************************
2956 * Function : content_filters_enabled
2958 * Description : Checks whether there are any content filters
2959 * enabled for the current request.
2962 * 1 : action = Action spec to check.
2964 * Returns : TRUE for yes, FALSE otherwise
2966 *********************************************************************/
2967 int content_filters_enabled(const struct current_action_spec *action)
2969 return ((action->flags & ACTION_DEANIMATE) ||
2970 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));
2978 vim:softtabstop=3 shiftwidth=3