1 const char filters_rcs[] = "$Id: filters.c,v 1.113 2009/03/08 14:19:23 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 1.113 2009/03/08 14:19:23 fabiankeil
44 * Fix justified (but harmless) compiler warnings
45 * on platforms where sizeof(int) < sizeof(long).
47 * Revision 1.112 2009/03/01 18:28:23 fabiankeil
48 * Help clang understand that we aren't dereferencing
51 * Revision 1.111 2008/12/04 18:13:46 fabiankeil
52 * Fix a cparser warning.
54 * Revision 1.110 2008/11/10 16:40:25 fabiankeil
55 * Fix a gcc44 warning.
57 * Revision 1.109 2008/11/08 15:48:41 fabiankeil
58 * Mention actual values when complaining about
59 * the chunk size exceeding the buffer size.
61 * Revision 1.108 2008/05/21 15:35:08 fabiankeil
62 * - Mark csp as immutable for block_acl().
63 * - Remove an obsolete complaint about filter_popups().
65 * Revision 1.107 2008/05/04 17:52:56 fabiankeil
66 * Adjust parse_http_url() call to new prototype.
68 * Revision 1.106 2008/05/03 16:40:44 fabiankeil
69 * Change content_filters_enabled()'s parameter from
70 * csp->action to action so it can be also used in the
71 * CGI code. Don't bother checking if there are filters
72 * loaded, as that's somewhat besides the point.
74 * Revision 1.105 2008/03/28 15:13:39 fabiankeil
75 * Remove inspect-jpegs action.
77 * Revision 1.104 2008/03/27 18:27:24 fabiankeil
78 * Remove kill-popups action.
80 * Revision 1.103 2008/03/06 16:33:45 fabiankeil
81 * If limit-connect isn't used, don't limit CONNECT requests to port 443.
83 * Revision 1.102 2008/03/01 14:00:44 fabiankeil
84 * Let the block action take the reason for the block
85 * as argument and show it on the "blocked" page.
87 * Revision 1.101 2008/02/23 16:57:12 fabiankeil
88 * Rename url_actions() to get_url_actions() and let it
89 * use the standard parameter ordering.
91 * Revision 1.100 2008/02/23 16:33:43 fabiankeil
92 * Let forward_url() use the standard parameter ordering
93 * and mark its second parameter immutable.
95 * Revision 1.99 2008/02/03 13:57:58 fabiankeil
96 * Add SOCKS5 support for forward-override{}.
98 * Revision 1.98 2008/01/04 17:43:45 fabiankeil
99 * Improve the warning messages that get logged if the action files
100 * "enable" filters but no filters of that type have been loaded.
102 * Revision 1.97 2007/11/30 15:37:03 fabiankeil
103 * Use freez instead of free.
105 * Revision 1.96 2007/10/19 16:53:28 fabiankeil
106 * Add helper function to check if any content filters are enabled.
108 * Revision 1.95 2007/10/17 19:31:20 fabiankeil
109 * Omitting the zero chunk that ends the chunk transfer encoding seems
110 * to be the new black. Log the problem and continue filtering anyway.
112 * Revision 1.94 2007/09/29 13:20:20 fabiankeil
113 * Remove two redundant and one useless log messages.
115 * Revision 1.93 2007/09/29 10:21:16 fabiankeil
116 * - Move get_filter_function() from jcc.c to filters.c
117 * so the filter functions can be static.
118 * - Don't bother filtering body-less responses.
120 * Revision 1.92 2007/09/28 16:38:55 fabiankeil
121 * - Execute content filters through execute_content_filter().
122 * - Add prepare_for_filtering() so filter functions don't have to
123 * care about de-chunking and decompression. As a side effect this enables
124 * decompression for gif_deanimate_response() and jpeg_inspect_response().
125 * - Change remove_chunked_transfer_coding()'s return type to jb_err.
126 * Some clowns feel like chunking empty responses in which case
127 * (size == 0) is valid but previously would be interpreted as error.
129 * Revision 1.91 2007/09/02 15:31:20 fabiankeil
130 * Move match_portlist() from filter.c to urlmatch.c.
131 * It's used for url matching, not for filtering.
133 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
134 * Remove newline at the end of a log_error() message.
136 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
137 * #1763173 from Stefan Huehner: declare some more functions static.
139 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
140 * Add forward-override{} to change the forwarding settings through
141 * action sections. This is mainly interesting to forward different
142 * clients differently (for example based on User-Agent or request
145 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
146 * Make sure filters with dynamic jobs actually use them.
148 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
149 * - Introduce dynamic pcrs jobs that can resolve variables.
150 * - Don't run redirect functions more than once,
151 * unless they are activated more than once.
153 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
154 * - Log the content size after decompression in decompress_iob()
155 * instead of pcrs_filter_response().
157 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
158 * Use dedicated header filter actions instead of abusing "filter".
159 * Replace "filter-client-headers" and "filter-client-headers"
160 * with "server-header-filter" and "client-header-filter".
162 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
163 * New config option: enforce-blocks.
165 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
166 * - Fix port handling in acl_addr() and use a temporary acl spec
167 * copy so error messages don't contain a truncated version.
168 * - Log size of iob before and after decompression.
170 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
171 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
172 * - Hide the "Go there anyway" link for blocked CONNECT
173 * requests where going there anyway doesn't work anyway.
175 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
176 * - Save the reason for generating http_responses.
177 * - Block (+block) with status code 403 instead of 404.
178 * - Use a different kludge to remember a failed decompression.
180 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
181 * Search for Max-Forwards headers case-insensitive,
182 * don't generate the "501 unsupported" message for invalid
183 * Max-Forwards values and don't increase negative ones.
185 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
186 * - Add HEAD support to finish_http_response.
187 * - Add error favicon to internal HTML error messages.
189 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
190 * Mark *csp as immutable for is_untrusted_url()
191 * and is_imageurl(). Closes FR 1237736.
193 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
194 * Integrate a modified version of Wil Mahan's
195 * zlib patch (PR #895531).
197 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
198 * Fixed gcc43 conversion warnings,
199 * changed sprintf calls to snprintf.
201 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
202 * Adjust comment in pcrs_filter_response()
203 * to recent pcrs changes. Hohoho.
205 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
206 * Don't crash if pcre returns an error code
207 * that pcrs didn't expect. Fixes BR 1621173.
209 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
210 * Modified is_untrusted_url to complain in case of
211 * write errors and to give a reason when adding new
212 * entries to the trustfile. Closes FR 1097611.
214 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
215 * Skip empty filter files in pcrs_filter_response,
216 * but don't ignore the ones that come afterwards.
217 * Fixes parts of BR 1619208.
219 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
220 * Added some sanity checks for get_last_url().
221 * Fixed possible segfault caused by my last commit.
223 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
224 * Let get_last_url() catch https URLs as well.
226 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
227 * Make sure get_last_url() behaves like advertised
228 * and fast-redirects{} can be combined with redirect{}.
230 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
231 * Implemented +redirect{s@foo@bar@} to generate
232 * a redirect based on a rewritten version of the
235 * Revision 1.66 2006/09/23 13:26:38 roro
236 * Replace TABs by spaces in source code.
238 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
239 * Fix +redirect{}. Didn't work with -fast-redirects.
241 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
242 * Block requests for untrusted URLs with status
243 * code 403 instead of 200.
245 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
246 * Don't free p which is still in use and will be later
247 * freed by free_map(). Don't claim the referrer is unknown
248 * when the client didn't set one.
250 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
251 * Feature request 595948: Re-Filter logging in single line
253 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
254 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
256 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
257 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
258 * with what was really the latest development (the v_3_0_branch branch)
260 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
261 * Multiple filter file support
263 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
264 * A few memory leaks plugged (mostly on error paths)
266 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
267 * Add the ability to check jpeg images for invalid
268 * lengths of comment blocks. Defensive strategy
269 * against the exploit:
270 * Microsoft Security Bulletin MS04-028
271 * Buffer Overrun in JPEG Processing (GDI+) Could
272 * Allow Code Execution (833987)
273 * Enabled with +inspect-jpegs in actions files.
275 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
276 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
278 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
279 * Fixed bug #839859: "See why" link URL now gets url-encoded.
281 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
284 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
285 * Added basic support for OPTIONS and TRACE HTTP methods:
286 * New function direct_response which handles OPTIONS and
287 * TRACE requests whose Max-Forwards header field is zero.
289 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
290 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
292 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
293 * - Bugfix: Executing a filters without jobs no longer results in
294 * turing off *all* filters.
295 * - Security fix: Malicious web servers can't cause a seg fault
296 * through bogus chunk sizes anymore
298 * Revision 1.58 2002/04/24 02:11:17 oes
299 * Jon's multiple AF patch: url_actions now evaluates rules
302 * Revision 1.57 2002/04/08 20:38:34 swa
305 * Revision 1.56 2002/04/05 15:51:24 oes
306 * - bugfix: error-pages now get correct request protocol
307 * - fix for invalid HTML in trust info
309 * Revision 1.55 2002/04/02 16:13:51 oes
310 * Fix: No "Go there anyway" for SSL
312 * Revision 1.54 2002/04/02 14:55:56 oes
313 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
315 * Revision 1.53 2002/03/26 22:29:54 swa
316 * we have a new homepage!
318 * Revision 1.52 2002/03/24 16:35:57 jongfoster
321 * Revision 1.51 2002/03/24 15:23:33 jongfoster
324 * Revision 1.50 2002/03/24 13:25:43 swa
325 * name change related issues
327 * Revision 1.49 2002/03/16 20:29:14 oes
330 * Revision 1.48 2002/03/13 20:25:34 oes
331 * Better logging for content filters
333 * Revision 1.47 2002/03/13 00:30:52 jongfoster
335 * Added option of always sending redirect for imageblock,
336 * currently disabled with #if 0.
338 * Revision 1.46 2002/03/12 01:42:49 oes
339 * Introduced modular filters
341 * Revision 1.45 2002/03/08 16:47:50 oes
342 * Added choice beween GIF and PNG built-in images
344 * Revision 1.44 2002/03/07 03:49:31 oes
345 * - Fixed compiler warnings etc
346 * - Changed built-in images from GIF to PNG
347 * (with regard to Unisys patent issue)
348 * - Added a 4x4 pattern PNG which is less intrusive
349 * than the logo but also clearly marks the deleted banners
351 * Revision 1.43 2002/01/22 23:51:59 jongfoster
352 * Replacing strsav() with the safer string_append().
354 * Adding missing html_encode() to error message generators. Where encoded
355 * and unencoded versions of a string were provided, removing the unencoded
358 * Revision 1.42 2002/01/17 21:00:32 jongfoster
359 * Moving all our URL and URL pattern parsing code to urlmatch.c.
361 * Using a single, simple url_match(pattern,url) function - rather than
362 * the 3-line match routine which was repeated all over the place.
364 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
366 * Using parse_http_url() to parse URLs without faking a HTTP
367 * request line for parse_http_request().
369 * Revision 1.41 2001/11/13 00:14:07 jongfoster
370 * Fixing stupid bug now I've figured out what || means.
371 * (It always returns 0 or 1, not one of it's paramaters.)
373 * Revision 1.40 2001/10/26 17:37:55 oes
374 * - Re-enabled Netscape 200/404 bug workaround in block_url():
375 * - Removed OS/2 special case
376 * - Made block_url() independant from sed() having been run
377 * - Made trust_url independant from sed() having been run
378 * - Made is_imageurl independant from sed() having been run.
379 * It now checks User-Agent: and Accept: by itself.
382 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
383 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
384 * threads to call select() simultaneously. So, it's time to do a real, live,
385 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
386 * (native). Both versions will work, but using __OS2__ offers multi-threading.
388 * Revision 1.38 2001/10/23 21:32:33 jongfoster
389 * Adding error-checking to selected functions
391 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
392 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
393 * filters.c. Added a FIXME in front of the offending code. I'll gladly
394 * put in a better/more robust fix for all parties if one is presented...
395 * It seems that just returning 200 instead of 404 would pretty much fix
396 * it for everyone, but I don't know all the history of the problem.
398 * Revision 1.36 2001/10/10 16:44:16 oes
399 * Added match_portlist function
401 * Revision 1.35 2001/10/07 15:41:23 oes
402 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
404 * New function remove_chunked_transfer_coding that strips chunked
405 * transfer coding to plain and is called by pcrs_filter_response
406 * and gif_deanimate_response if neccessary
408 * Improved handling of zero-change re_filter runs
410 * pcrs_filter_response and gif_deanimate_response now remove
411 * chunked transfer codeing before processing the body.
413 * Revision 1.34 2001/09/20 15:49:36 steudten
415 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
416 * See cgi.c fill_template().
418 * Revision 1.33 2001/09/16 17:05:14 jongfoster
419 * Removing unused #include showarg.h
421 * Revision 1.32 2001/09/16 13:21:27 jongfoster
422 * Changes to use new list functions.
424 * Revision 1.31 2001/09/16 11:38:02 jongfoster
425 * Splitting fill_template() into 2 functions:
426 * template_load() loads the file
427 * template_fill() performs the PCRS regexps.
428 * This is because the CGI edit interface has a "table row"
429 * template which is used many times in the page - this
430 * change means it's only loaded from disk once.
432 * Revision 1.30 2001/09/16 11:00:10 jongfoster
433 * New function alloc_http_response, for symmetry with free_http_response
435 * Revision 1.29 2001/09/13 23:32:40 jongfoster
436 * Moving image data to cgi.c rather than cgi.h
437 * Fixing a GPF under Win32 (and any other OS that protects global
438 * constants from being written to).
440 * Revision 1.28 2001/09/10 10:18:51 oes
441 * Silenced compiler warnings
443 * Revision 1.27 2001/08/05 16:06:20 jongfoster
444 * Modifiying "struct map" so that there are now separate header and
445 * "map_entry" structures. This means that functions which modify a
446 * map no longer need to return a pointer to the modified map.
447 * Also, it no longer reverses the order of the entries (which may be
448 * important with some advanced template substitutions).
450 * Revision 1.26 2001/07/30 22:08:36 jongfoster
451 * Tidying up #defines:
452 * - All feature #defines are now of the form FEATURE_xxx
453 * - Permanently turned off WIN_GUI_EDIT
454 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
456 * Revision 1.25 2001/07/26 10:09:46 oes
457 * Made browser detection a little less naive
459 * Revision 1.24 2001/07/25 17:22:51 oes
460 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
462 * Revision 1.23 2001/07/23 13:40:12 oes
463 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
465 * Revision 1.22 2001/07/18 12:29:34 oes
466 * - Made gif_deanimate_response respect
467 * csp->action->string[ACTION_STRING_DEANIMATE]
468 * - Logging cosmetics
470 * Revision 1.21 2001/07/13 13:59:53 oes
471 * - Introduced gif_deanimate_response which shares the
472 * generic content modification interface of pcrs_filter_response
473 * and acts as a wrapper to deanimate.c:gif_deanimate()
474 * - Renamed re_process_buffer to pcrs_filter_response
475 * - pcrs_filter_response now returns NULL on failiure
476 * - Removed all #ifdef PCRS
478 * Revision 1.20 2001/07/01 17:01:04 oes
479 * Added comments and missing return statement in is_untrusted_url()
481 * Revision 1.19 2001/06/29 21:45:41 oes
482 * Indentation, CRLF->LF, Tab-> Space
484 * Revision 1.18 2001/06/29 13:27:38 oes
485 * - Cleaned up, renamed and reorderd functions
486 * and improved comments
489 * - Ported to CGI platform. Now delivers
490 * http_response or NULL
491 * - Unified HTML and GIF generation (moved image detection
492 * and GIF generation here from jcc.c:chat())
493 * - Fixed HTTP status to:
494 * - 403 (Forbidden) for the "blocked" HTML message
495 * - 200 (OK) for GIF answers
496 * - 302 (Redirect) for redirect to GIF
499 * - Ported to CGI platform. Now delivers
500 * http_response or NULL
501 * - Separated detection of untrusted URL into
502 * (bool)is_untrusted_url
503 * - Added enforcement of untrusted requests
505 * - Moved redirect_url() from cgi.c to here
506 * and ported it to the CGI platform
508 * - Removed logentry from cancelled commit
510 * Revision 1.17 2001/06/09 10:55:28 jongfoster
511 * Changing BUFSIZ ==> BUFFER_SIZE
513 * Revision 1.16 2001/06/07 23:10:26 jongfoster
514 * Allowing unanchored domain patterns to back off and retry
515 * if they partially match. Optimized right-anchored patterns.
516 * Moving ACL and forward files into config file.
517 * Replacing struct gateway with struct forward_spec
519 * Revision 1.15 2001/06/03 19:12:00 oes
520 * extracted-CGI relevant stuff
522 * Revision 1.14 2001/06/01 10:30:55 oes
523 * Added optional left-anchoring to domaincmp
525 * Revision 1.13 2001/05/31 21:21:30 jongfoster
526 * Permissionsfile / actions file changes:
527 * - Changed "permission" to "action" throughout
528 * - changes to file format to allow string parameters
529 * - Moved helper functions to actions.c
531 * Revision 1.12 2001/05/31 17:35:20 oes
533 * - Enhanced domain part globbing with infix and prefix asterisk
534 * matching and optional unanchored operation
536 * Revision 1.11 2001/05/29 11:53:23 oes
537 * "See why" link added to "blocked" page
539 * Revision 1.10 2001/05/29 09:50:24 jongfoster
540 * Unified blocklist/imagelist/permissionslist.
541 * File format is still under discussion, but the internal changes
544 * Also modified interceptor behaviour:
545 * - We now intercept all URLs beginning with one of the following
546 * prefixes (and *only* these prefixes):
548 * * http://ijbswa.sf.net/config/
549 * * http://ijbswa.sourceforge.net/config/
550 * - New interceptors "home page" - go to http://i.j.b/ to see it.
551 * - Internal changes so that intercepted and fast redirect pages
552 * are not replaced with an image.
553 * - Interceptors now have the option to send a binary page direct
554 * to the client. (i.e. ijb-send-banner uses this)
555 * - Implemented show-url-info interceptor. (Which is why I needed
556 * the above interceptors changes - a typical URL is
557 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
558 * The previous mechanism would not have intercepted that, and
559 * if it had been intercepted then it then it would have replaced
562 * Revision 1.9 2001/05/27 22:17:04 oes
564 * - re_process_buffer no longer writes the modified buffer
565 * to the client, which was very ugly. It now returns the
566 * buffer, which it is then written by chat.
568 * - content_length now adjusts the Content-Length: header
569 * for modified documents rather than crunch()ing it.
570 * (Length info in csp->content_length, which is 0 for
571 * unmodified documents)
573 * - For this to work, sed() is called twice when filtering.
575 * Revision 1.8 2001/05/26 17:13:28 jongfoster
576 * Filled in a function comment.
578 * Revision 1.7 2001/05/26 15:26:15 jongfoster
579 * ACL feature now provides more security by immediately dropping
580 * connections from untrusted hosts.
582 * Revision 1.6 2001/05/26 00:28:36 jongfoster
583 * Automatic reloading of config file.
584 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
585 * Most of the global variables have been moved to a new
586 * struct configuration_spec, accessed through csp->config->globalname
587 * Most of the globals remaining are used by the Win32 GUI.
589 * Revision 1.5 2001/05/25 22:34:30 jongfoster
592 * Revision 1.4 2001/05/22 18:46:04 oes
594 * - Enabled filtering banners by size rather than URL
595 * by adding patterns that replace all standard banner
596 * sizes with the "Junkbuster" gif to the re_filterfile
598 * - Enabled filtering WebBugs by providing a pattern
599 * which kills all 1x1 images
601 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
602 * which is selected by the (nonstandard and therefore
603 * capital) letter 'U' in the option string.
604 * It causes the quantifiers to be ungreedy by default.
605 * Appending a ? turns back to greedy (!).
607 * - Added a new interceptor ijb-send-banner, which
608 * sends back the "Junkbuster" gif. Without imagelist or
609 * MSIE detection support, or if tinygif = 1, or the
610 * URL isn't recognized as an imageurl, a lame HTML
611 * explanation is sent instead.
613 * - Added new feature, which permits blocking remote
614 * script redirects and firing back a local redirect
616 * The feature is conditionally compiled, i.e. it
617 * can be disabled with --disable-fast-redirects,
618 * plus it must be activated by a "fast-redirects"
619 * line in the config file, has its own log level
620 * and of course wants to be displayed by show-proxy-args
621 * Note: Boy, all the #ifdefs in 1001 locations and
622 * all the fumbling with configure.in and acconfig.h
623 * were *way* more work than the feature itself :-(
625 * - Because a generic redirect template was needed for
626 * this, tinygif = 3 now uses the same.
628 * - Moved GIFs, and other static HTTP response templates
633 * - Removed some >400 CRs again (Jon, you really worked
636 * Revision 1.3 2001/05/20 16:44:47 jongfoster
637 * Removing last hardcoded Junkbusters.com URLs.
639 * Revision 1.2 2001/05/20 01:21:20 jongfoster
640 * Version 2.9.4 checkin.
641 * - Merged popupfile and cookiefile, and added control over PCRS
642 * filtering, in new "permissionsfile".
643 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
644 * file error you now get a message box (in the Win32 GUI) rather
645 * than the program exiting with no explanation.
646 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
648 * - Removed tabs from "config"
649 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
650 * - Bumped up version number.
652 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
653 * Initial import of version 2.9.3 source tree
656 *********************************************************************/
662 #include <sys/types.h>
668 #ifdef HAVE_GETADDRINFO
670 #include <sys/socket.h>
671 #endif /* def HAVE_GETADDRINFO */
676 #endif /* ndef __OS2__ */
677 #include <netinet/in.h>
679 #include <winsock2.h>
680 #endif /* ndef _WIN32 */
684 #endif /* def __OS2__ */
692 #include "jbsockets.h"
693 #include "miscutil.h"
697 #include "deanimate.h"
698 #include "urlmatch.h"
705 const char filters_h_rcs[] = FILTERS_H_VERSION;
707 /* Fix a problem with Solaris. There should be no effect on other
709 * Solaris's isspace() is a macro which uses it's argument directly
710 * as an array index. Therefore we need to make sure that high-bit
711 * characters generate +ve values, and ideally we also want to make
712 * the argument match the declared parameter type of "int".
714 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
716 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
717 static jb_err prepare_for_filtering(struct client_state *csp);
720 #ifdef HAVE_GETADDRINFO
721 /*********************************************************************
723 * Function : sockaddr_storage_to_ip
725 * Description : Access internal structure of sockaddr_storage
728 * 1 : addr = socket address
729 * 2 : ip = IP address as array of octets in network order
730 * (it points into addr)
731 * 3 : len = length of IP address in octets
732 * 4 : port = port number in network order;
734 * Returns : 0 = no errror; otherwise
736 *********************************************************************/
737 int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, uint8_t **ip,
738 unsigned int *len, in_port_t **port)
745 switch (addr->ss_family)
755 &(( (struct sockaddr_in *) addr)->sin_addr.s_addr);
759 *port = &((struct sockaddr_in *) addr)->sin_port;
770 *ip = ( (struct sockaddr_in6 *) addr)->sin6_addr.s6_addr;
774 *port = &((struct sockaddr_in6 *) addr)->sin6_port;
779 /* Unsupported address family */
787 /*********************************************************************
789 * Function : match_sockaddr
791 * Description : Check whether address matches network (IP address and port)
794 * 1 : network = socket address of subnework
795 * 3 : netmask = network mask as socket address
796 * 2 : address = checked socket address against given network
798 * Returns : 0 = doesn't match; 1 = does match
800 *********************************************************************/
801 int match_sockaddr(const struct sockaddr_storage *network,
802 const struct sockaddr_storage *netmask,
803 const struct sockaddr_storage *address)
805 uint8_t *network_addr, *netmask_addr, *address_addr;
806 unsigned int addr_len;
807 in_port_t *network_port, *netmask_port, *address_port;
810 if (network->ss_family != netmask->ss_family)
812 /* This should never happen */
813 log_error(LOG_LEVEL_ERROR,
814 "Internal error at %s:%llu: network and netmask differ in family",
819 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
820 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
821 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
823 /* Check for family */
824 if (network->ss_family == AF_INET && address->ss_family == AF_INET6 &&
825 IN6_IS_ADDR_V4MAPPED(address_addr))
827 /* Map AF_INET6 V4MAPPED address into AF_INET */
831 else if (network->ss_family == AF_INET6 && address->ss_family == AF_INET &&
832 IN6_IS_ADDR_V4MAPPED(network_addr))
834 /* Map AF_INET6 V4MAPPED network into AF_INET */
839 else if (network->ss_family != address->ss_family)
844 /* XXX: Port check is signaled in netmask */
845 if (*netmask_port && *network_port != *address_port)
850 /* TODO: Optimize by checking by words insted of octets */
851 for (i=0; i < addr_len && netmask_addr[i]; i++)
853 if ( (network_addr[i] & netmask_addr[i]) !=
854 (address_addr[i] & netmask_addr[i]) )
862 #endif /* def HAVE_GETADDRINFO */
865 /*********************************************************************
867 * Function : block_acl
869 * Description : Block this request?
870 * Decide yes or no based on ACL file.
873 * 1 : dst = The proxy or gateway address this is going to.
874 * Or NULL to check all possible targets.
875 * 2 : csp = Current client state (buffers, headers, etc...)
876 * Also includes the client IP address.
878 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
880 *********************************************************************/
881 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
883 struct access_control_list *acl = csp->config->acl;
885 /* if not using an access control list, then permit the connection */
891 /* search the list */
895 #ifdef HAVE_GETADDRINFO
896 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
898 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
904 /* Just want to check if they have any access */
905 if (acl->action == ACL_PERMIT)
911 #ifdef HAVE_GETADDRINFO
912 /* XXX: Undefined acl->dst is full of zeros and should be
913 * considered as wildcard address.
914 * sockaddr_storage_to_ip() failes on such dst because of
915 * uknown sa_familly on glibc. However this test is not
918 * So, we signal the acl->dst is wildcard in wildcard_dst.
921 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
923 ((dst->addr & acl->dst->mask) == acl->dst->addr)
924 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
928 if (acl->action == ACL_PERMIT)
946 /*********************************************************************
948 * Function : acl_addr
950 * Description : Called from `load_config' to parse an ACL address.
953 * 1 : aspec = String specifying ACL address.
954 * 2 : aca = struct access_control_addr to fill in.
956 * Returns : 0 => Ok, everything else is an error.
958 *********************************************************************/
959 int acl_addr(const char *aspec, struct access_control_addr *aca)
962 #ifdef HAVE_GETADDRINFO
963 struct addrinfo hints, *result;
965 in_port_t *mask_port;
966 unsigned int addr_len;
969 #endif /* def HAVE_GETADDRINFO */
971 char *acl_spec = NULL;
973 #ifdef HAVE_GETADDRINFO
974 /* FIXME: Depend on ai_family */
982 * Use a temporary acl spec copy so we can log
983 * the unmodified original in case of parse errors.
985 acl_spec = strdup(aspec);
986 if (acl_spec == NULL)
988 /* XXX: This will be logged as parse error. */
992 if ((p = strchr(acl_spec, '/')) != NULL)
995 if (ijb_isdigit(*p) == 0)
1000 masklength = atoi(p);
1003 if ((masklength < 0) ||
1004 #ifdef HAVE_GETADDRINFO
1015 if (*acl_spec == '[' && NULL != (p = strchr(acl_spec, ']')))
1018 memmove(acl_spec, acl_spec + 1, (size_t) (p - acl_spec));
1027 p = strchr(acl_spec, ':');
1030 #ifdef HAVE_GETADDRINFO
1031 memset(&hints, 0, sizeof(struct addrinfo));
1032 hints.ai_family = AF_UNSPEC;
1033 hints.ai_socktype = SOCK_STREAM;
1035 i = getaddrinfo(acl_spec, (p) ? ++p : NULL, &hints, &result);
1040 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s", acl_spec, p,
1045 /* TODO: Allow multihomed hostnames */
1046 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
1047 freeaddrinfo(result);
1054 port = strtol(p, &endptr, 10);
1056 if (port <= 0 || port > 65535 || *endptr != '\0')
1063 aca->port = (unsigned long)port;
1065 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
1068 if (aca->addr == INADDR_NONE)
1070 /* XXX: This will be logged as parse error. */
1073 #endif /* def HAVE_GETADDRINFO */
1075 /* build the netmask */
1076 #ifdef HAVE_GETADDRINFO
1077 /* Clip masklength according current family */
1078 if (aca->addr.ss_family == AF_INET && masklength > 32)
1083 aca->mask.ss_family = aca->addr.ss_family;
1084 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
1091 /* Port number in ACL has been specified, check ports in future */
1095 /* XXX: This could be optimized to operate on whole words instead of octets
1096 * (128-bit CPU could do it in one iteration). */
1097 /* Octets after prefix can be ommitted because of previous initialization
1099 for (i=0; i < addr_len && masklength; i++)
1101 if (masklength >= 8)
1103 mask_data[i] = 0xFF;
1108 /* XXX: This assumes MSB of octet is on the left site. This should be
1109 * true for all architectures or solved on link layer of OSI model. */
1110 mask_data[i] = ~((1 << (8 - masklength)) - 1);
1117 for (i=1; i <= masklength ; i++)
1119 aca->mask |= (1U << (32 - i));
1122 /* now mask off the host portion of the ip address
1123 * (i.e. save on the network portion of the address).
1125 aca->addr = aca->addr & aca->mask;
1126 #endif /* def HAVE_GETADDRINFO */
1131 #endif /* def FEATURE_ACL */
1134 /*********************************************************************
1136 * Function : connect_port_is_forbidden
1138 * Description : Check to see if CONNECT requests to the destination
1139 * port of this request are forbidden. The check is
1140 * independend of the actual request method.
1143 * 1 : csp = Current client state (buffers, headers, etc...)
1145 * Returns : True if yes, false otherwise.
1147 *********************************************************************/
1148 int connect_port_is_forbidden(const struct client_state *csp)
1150 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
1151 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
1156 /*********************************************************************
1158 * Function : block_url
1160 * Description : Called from `chat'. Check to see if we need to block this.
1163 * 1 : csp = Current client state (buffers, headers, etc...)
1165 * Returns : NULL => unblocked, else HTTP block response
1167 *********************************************************************/
1168 struct http_response *block_url(struct client_state *csp)
1170 struct http_response *rsp;
1171 const char *new_content_type = NULL;
1174 * If it's not blocked, don't block it ;-)
1176 if ((csp->action->flags & ACTION_BLOCK) == 0)
1180 if (csp->action->flags & ACTION_REDIRECT)
1182 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
1185 * Else, prepare a response
1187 if (NULL == (rsp = alloc_http_response()))
1189 return cgi_error_memory();
1193 * If it's an image-url, send back an image or redirect
1194 * as specified by the relevant +image action
1196 #ifdef FEATURE_IMAGE_BLOCKING
1197 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
1198 && is_imageurl(csp))
1201 /* determine HOW images should be blocked */
1202 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
1204 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1206 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
1208 #if 1 /* Two alternative strategies, use this one for now: */
1210 /* and handle accordingly: */
1211 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1213 rsp->status = strdup("403 Request blocked by Privoxy");
1214 if (rsp->status == NULL)
1216 free_http_response(rsp);
1217 return cgi_error_memory();
1219 rsp->body = bindup(image_pattern_data, image_pattern_length);
1220 if (rsp->body == NULL)
1222 free_http_response(rsp);
1223 return cgi_error_memory();
1225 rsp->content_length = image_pattern_length;
1227 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1229 free_http_response(rsp);
1230 return cgi_error_memory();
1234 else if (0 == strcmpic(p, "blank"))
1236 rsp->status = strdup("403 Request blocked by Privoxy");
1237 if (rsp->status == NULL)
1239 free_http_response(rsp);
1240 return cgi_error_memory();
1242 rsp->body = bindup(image_blank_data, image_blank_length);
1243 if (rsp->body == NULL)
1245 free_http_response(rsp);
1246 return cgi_error_memory();
1248 rsp->content_length = image_blank_length;
1250 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1252 free_http_response(rsp);
1253 return cgi_error_memory();
1259 rsp->status = strdup("302 Local Redirect from Privoxy");
1260 if (rsp->status == NULL)
1262 free_http_response(rsp);
1263 return cgi_error_memory();
1266 if (enlist_unique_header(rsp->headers, "Location", p))
1268 free_http_response(rsp);
1269 return cgi_error_memory();
1273 #else /* Following code is disabled for now */
1275 /* and handle accordingly: */
1276 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1278 p = CGI_PREFIX "send-banner?type=pattern";
1280 else if (0 == strcmpic(p, "blank"))
1282 p = CGI_PREFIX "send-banner?type=blank";
1284 rsp->status = strdup("302 Local Redirect from Privoxy");
1285 if (rsp->status == NULL)
1287 free_http_response(rsp);
1288 return cgi_error_memory();
1291 if (enlist_unique_header(rsp->headers, "Location", p))
1293 free_http_response(rsp);
1294 return cgi_error_memory();
1296 #endif /* Preceeding code is disabled for now */
1298 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1301 * Send empty document.
1303 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1306 rsp->body = strdup(" ");
1307 rsp->content_length = 1;
1309 rsp->status = strdup("403 Request blocked by Privoxy");
1310 if (rsp->status == NULL)
1312 free_http_response(rsp);
1313 return cgi_error_memory();
1315 if (new_content_type != 0)
1317 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1318 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1320 free_http_response(rsp);
1321 return cgi_error_memory();
1326 #endif /* def FEATURE_IMAGE_BLOCKING */
1329 * Else, generate an HTML "blocked" message:
1333 struct map * exports;
1337 * Workaround for stupid Netscape bug which prevents
1338 * pages from being displayed if loading a referenced
1339 * JavaScript or style sheet fails. So make it appear
1340 * as if it succeeded.
1342 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1343 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1344 && !strstr(p, "Gecko") /* save Mozilla, */
1345 && !strstr(p, "compatible") /* MSIE */
1346 && !strstr(p, "Opera")) /* and Opera. */
1348 rsp->status = strdup("200 Request for blocked URL");
1352 rsp->status = strdup("403 Request for blocked URL");
1355 if (rsp->status == NULL)
1357 free_http_response(rsp);
1358 return cgi_error_memory();
1361 exports = default_exports(csp, NULL);
1362 if (exports == NULL)
1364 free_http_response(rsp);
1365 return cgi_error_memory();
1368 #ifdef FEATURE_FORCE_LOAD
1369 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1371 * Export the force conditional block killer if
1373 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1374 * - Privoxy is configured to enforce blocks, or
1375 * - it's a CONNECT request and enforcing wouldn't work anyway.
1377 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1378 || (0 == strcmpic(csp->http->gpc, "connect")))
1379 #endif /* ndef FEATURE_FORCE_LOAD */
1381 err = map_block_killer(exports, "force-support");
1384 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1385 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1386 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1387 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1390 const char *block_reason;
1391 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
1393 block_reason = csp->action->string[ACTION_STRING_BLOCK];
1397 assert(connect_port_is_forbidden(csp));
1398 block_reason = "Forbidden CONNECT port.";
1400 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
1405 free_http_response(rsp);
1406 return cgi_error_memory();
1409 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1412 free_http_response(rsp);
1413 return cgi_error_memory();
1416 rsp->reason = RSP_REASON_BLOCKED;
1418 return finish_http_response(csp, rsp);
1423 #ifdef FEATURE_TRUST
1424 /*********************************************************************
1426 * Function : trust_url FIXME: I should be called distrust_url
1428 * Description : Calls is_untrusted_url to determine if the URL is trusted
1429 * and if not, returns a HTTP 403 response with a reject message.
1432 * 1 : csp = Current client state (buffers, headers, etc...)
1434 * Returns : NULL => trusted, else http_response.
1436 *********************************************************************/
1437 struct http_response *trust_url(struct client_state *csp)
1439 struct http_response *rsp;
1440 struct map * exports;
1441 char buf[BUFFER_SIZE];
1443 struct url_spec **tl;
1448 * Don't bother to work on trusted URLs
1450 if (!is_untrusted_url(csp))
1456 * Else, prepare a response:
1458 if (NULL == (rsp = alloc_http_response()))
1460 return cgi_error_memory();
1463 rsp->status = strdup("403 Request blocked by Privoxy");
1464 exports = default_exports(csp, NULL);
1465 if (exports == NULL || rsp->status == NULL)
1467 free_http_response(rsp);
1468 return cgi_error_memory();
1472 * Export the protocol, host, port, and referrer information
1474 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1475 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1476 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1478 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1480 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1484 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1490 free_http_response(rsp);
1491 return cgi_error_memory();
1495 * Export the trust list
1498 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1500 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1501 string_append(&p, buf);
1503 err = map(exports, "trusted-referrers", 1, p, 0);
1508 free_http_response(rsp);
1509 return cgi_error_memory();
1513 * Export the trust info, if available
1515 if (csp->config->trust_info->first)
1517 struct list_entry *l;
1520 for (l = csp->config->trust_info->first; l ; l = l->next)
1522 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1523 string_append(&p, buf);
1525 err = map(exports, "trust-info", 1, p, 0);
1529 err = map_block_killer(exports, "have-trust-info");
1535 free_http_response(rsp);
1536 return cgi_error_memory();
1540 * Export the force conditional block killer if
1542 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1543 * - Privoxy is configured to enforce blocks, or
1544 * - it's a CONNECT request and enforcing wouldn't work anyway.
1546 #ifdef FEATURE_FORCE_LOAD
1547 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1548 || (0 == strcmpic(csp->http->gpc, "connect")))
1550 err = map_block_killer(exports, "force-support");
1554 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1556 #else /* ifndef FEATURE_FORCE_LOAD */
1557 err = map_block_killer(exports, "force-support");
1558 #endif /* ndef FEATURE_FORCE_LOAD */
1563 free_http_response(rsp);
1564 return cgi_error_memory();
1568 * Build the response
1570 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1573 free_http_response(rsp);
1574 return cgi_error_memory();
1576 rsp->reason = RSP_REASON_UNTRUSTED;
1578 return finish_http_response(csp, rsp);
1580 #endif /* def FEATURE_TRUST */
1583 /*********************************************************************
1585 * Function : compile_dynamic_pcrs_job_list
1587 * Description : Compiles a dynamic pcrs job list (one with variables
1588 * resolved at request time)
1591 * 1 : csp = Current client state (buffers, headers, etc...)
1592 * 2 : b = The filter list to compile
1594 * Returns : NULL in case of errors, otherwise the
1597 *********************************************************************/
1598 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1600 struct list_entry *pattern;
1601 pcrs_job *job_list = NULL;
1602 pcrs_job *dummy = NULL;
1603 pcrs_job *lastjob = NULL;
1606 const struct pcrs_variable variables[] =
1608 {"url", csp->http->url, 1},
1609 {"path", csp->http->path, 1},
1610 {"host", csp->http->host, 1},
1611 {"origin", csp->ip_addr_str, 1},
1615 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1617 assert(pattern->str != NULL);
1619 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1623 log_error(LOG_LEVEL_ERROR,
1624 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1625 pattern->str, b->name, pcrs_strerror(error));
1630 if (error == PCRS_WARN_TRUNCATION)
1632 log_error(LOG_LEVEL_ERROR,
1633 "At least one of the variables in \'%s\' had to "
1634 "be truncated before compilation", pattern->str);
1636 if (job_list == NULL)
1642 lastjob->next = dummy;
1652 /*********************************************************************
1654 * Function : rewrite_url
1656 * Description : Rewrites a URL with a single pcrs command
1657 * and returns the result if it differs from the
1658 * original and isn't obviously invalid.
1661 * 1 : old_url = URL to rewrite.
1662 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1665 * Returns : NULL if the pcrs_command didn't change the url, or
1666 * the result of the modification.
1668 *********************************************************************/
1669 char *rewrite_url(char *old_url, const char *pcrs_command)
1671 char *new_url = NULL;
1675 assert(pcrs_command);
1677 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1681 log_error(LOG_LEVEL_REDIRECTS,
1682 "pcrs command \"%s\" didn't change \"%s\".",
1683 pcrs_command, old_url);
1688 log_error(LOG_LEVEL_REDIRECTS,
1689 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1690 pcrs_command, old_url, pcrs_strerror(hits));
1693 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1695 log_error(LOG_LEVEL_ERROR,
1696 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1697 "but the result doesn't look like a valid URL and will be ignored.",
1698 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1703 log_error(LOG_LEVEL_REDIRECTS,
1704 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1705 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1713 #ifdef FEATURE_FAST_REDIRECTS
1714 /*********************************************************************
1716 * Function : get_last_url
1718 * Description : Search for the last URL inside a string.
1719 * If the string already is a URL, it will
1720 * be the first URL found.
1723 * 1 : subject = the string to check
1724 * 2 : redirect_mode = +fast-redirect{} mode
1726 * Returns : NULL if no URL was found, or
1727 * the last URL found.
1729 *********************************************************************/
1730 char *get_last_url(char *subject, const char *redirect_mode)
1732 char *new_url = NULL;
1736 assert(redirect_mode);
1738 subject = strdup(subject);
1739 if (subject == NULL)
1741 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1745 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1747 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1748 new_url = url_decode(subject);
1749 if (new_url != NULL)
1756 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1760 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1763 * Find the last URL encoded in the request
1766 while ((tmp = strstr(tmp, "http://")) != NULL)
1770 tmp = (new_url != NULL) ? new_url : subject;
1771 while ((tmp = strstr(tmp, "https://")) != NULL)
1776 if ((new_url != NULL)
1777 && ( (new_url != subject)
1778 || (0 == strncmpic(subject, "http://", 7))
1779 || (0 == strncmpic(subject, "https://", 8))
1783 * Return new URL if we found a redirect
1784 * or if the subject already was a URL.
1786 * The second case makes sure that we can
1787 * chain get_last_url after another redirection check
1788 * (like rewrite_url) without losing earlier redirects.
1790 new_url = strdup(new_url);
1799 #endif /* def FEATURE_FAST_REDIRECTS */
1802 /*********************************************************************
1804 * Function : redirect_url
1806 * Description : Checks if Privoxy should answer the request with
1807 * a HTTP redirect and generates the redirect if
1811 * 1 : csp = Current client state (buffers, headers, etc...)
1813 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1815 *********************************************************************/
1816 struct http_response *redirect_url(struct client_state *csp)
1818 struct http_response *rsp;
1819 #ifdef FEATURE_FAST_REDIRECTS
1821 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1822 * as compile-time option? The user can easily disable
1823 * it in his action file.
1825 char * redirect_mode;
1826 #endif /* def FEATURE_FAST_REDIRECTS */
1827 char *old_url = NULL;
1828 char *new_url = NULL;
1829 char *redirection_string;
1831 if ((csp->action->flags & ACTION_REDIRECT))
1833 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1836 * If the redirection string begins with 's',
1837 * assume it's a pcrs command, otherwise treat it as
1838 * properly formatted URL and use it for the redirection
1841 * According to RFC 2616 section 14.30 the URL
1842 * has to be absolute and if the user tries:
1843 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1844 * she would get undefined results anyway.
1848 if (*redirection_string == 's')
1850 old_url = csp->http->url;
1851 new_url = rewrite_url(old_url, redirection_string);
1855 log_error(LOG_LEVEL_REDIRECTS,
1856 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1857 redirection_string);
1858 new_url = strdup(redirection_string);
1862 #ifdef FEATURE_FAST_REDIRECTS
1863 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1865 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1868 * If it exists, use the previously rewritten URL as input
1869 * otherwise just use the old path.
1871 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1872 new_url = get_last_url(old_url, redirect_mode);
1877 * Disable redirect checkers, so that they
1878 * will be only run more than once if the user
1879 * also enables them through tags.
1881 * From a performance point of view
1882 * it doesn't matter, but the duplicated
1883 * log messages are annoying.
1885 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1886 #endif /* def FEATURE_FAST_REDIRECTS */
1887 csp->action->flags &= ~ACTION_REDIRECT;
1889 /* Did any redirect action trigger? */
1892 if (0 == strcmpic(new_url, csp->http->url))
1894 log_error(LOG_LEVEL_ERROR,
1895 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1896 csp->http->url, new_url);
1901 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1903 if (NULL == (rsp = alloc_http_response()))
1906 return cgi_error_memory();
1909 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1910 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1913 free_http_response(rsp);
1914 return cgi_error_memory();
1916 rsp->reason = RSP_REASON_REDIRECTED;
1919 return finish_http_response(csp, rsp);
1923 /* Only reached if no redirect is required */
1929 #ifdef FEATURE_IMAGE_BLOCKING
1930 /*********************************************************************
1932 * Function : is_imageurl
1934 * Description : Given a URL, decide whether it is an image or not,
1935 * using either the info from a previous +image action
1936 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1937 * is MSIE and not on a Mac, tell from the browser's accept
1941 * 1 : csp = Current client state (buffers, headers, etc...)
1943 * Returns : True (nonzero) if URL is an image, false (0)
1946 *********************************************************************/
1947 int is_imageurl(const struct client_state *csp)
1949 #ifdef FEATURE_IMAGE_DETECT_MSIE
1952 tmp = get_header_value(csp->headers, "User-Agent:");
1953 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1955 tmp = get_header_value(csp->headers, "Accept:");
1956 if (tmp && strstr(tmp, "image/gif"))
1958 /* Client will accept HTML. If this seems counterintuitive,
1968 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1970 return ((csp->action->flags & ACTION_IMAGE) != 0);
1973 #endif /* def FEATURE_IMAGE_BLOCKING */
1976 #ifdef FEATURE_TRUST
1977 /*********************************************************************
1979 * Function : is_untrusted_url
1981 * Description : Should we "distrust" this URL (and block it)?
1983 * Yes if it matches a line in the trustfile, or if the
1984 * referrer matches a line starting with "+" in the
1989 * 1 : csp = Current client state (buffers, headers, etc...)
1991 * Returns : 0 => trusted, 1 => untrusted
1993 *********************************************************************/
1994 int is_untrusted_url(const struct client_state *csp)
1996 struct file_list *fl;
1997 struct block_spec *b;
1998 struct url_spec **trusted_url;
1999 struct http_request rhttp[1];
2000 const char * referer;
2004 * If we don't have a trustlist, we trust everybody
2006 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
2011 memset(rhttp, '\0', sizeof(*rhttp));
2014 * Do we trust the request URL itself?
2016 for (b = b->next; b ; b = b->next)
2018 if (url_match(b->url, csp->http))
2024 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
2026 /* no referrer was supplied */
2032 * If not, do we maybe trust its referrer?
2034 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
2040 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
2042 if (url_match(*trusted_url, rhttp))
2044 /* if the URL's referrer is from a trusted referrer, then
2045 * add the target spec to the trustfile as an unblocked
2046 * domain and return 0 (which means it's OK).
2051 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
2055 char * new_entry = strdup("~");
2057 string_append(&new_entry, csp->http->hostport);
2059 path = csp->http->path;
2060 if ( (path[0] == '/')
2062 && ((path_end = strchr(path + 2, '/')) != NULL))
2064 /* since this path points into a user's home space
2065 * be sure to include this spec in the trustfile.
2067 long path_len = path_end - path; /* save offset */
2068 path = strdup(path); /* Copy string */
2071 path_end = path + path_len; /* regenerate ptr to new buffer */
2072 *(path_end + 1) = '\0'; /* Truncate path after '/' */
2074 string_join(&new_entry, path);
2078 * Give a reason for generating this entry.
2080 string_append(&new_entry, " # Trusted referrer was: ");
2081 string_append(&new_entry, referer);
2083 if (new_entry != NULL)
2085 if (-1 == fprintf(fp, "%s\n", new_entry))
2087 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
2088 new_entry, csp->config->trustfile);
2094 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
2095 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
2102 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
2103 csp->http->hostport, csp->config->trustfile);
2111 #endif /* def FEATURE_TRUST */
2114 /*********************************************************************
2116 * Function : pcrs_filter_response
2118 * Description : Execute all text substitutions from all applying
2119 * +filter actions on the text buffer that's been
2120 * accumulated in csp->iob->buf.
2123 * 1 : csp = Current client state (buffers, headers, etc...)
2125 * Returns : a pointer to the (newly allocated) modified buffer.
2126 * or NULL if there were no hits or something went wrong
2128 *********************************************************************/
2129 static char *pcrs_filter_response(struct client_state *csp)
2132 size_t size, prev_size;
2138 struct file_list *fl;
2139 struct re_filterfile_spec *b;
2140 struct list_entry *filtername;
2142 int i, found_filters = 0;
2147 if (csp->iob->cur >= csp->iob->eod)
2153 * Need to check the set of re_filterfiles...
2155 for (i = 0; i < MAX_AF_FILES; i++)
2168 if (0 == found_filters)
2170 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2171 "content filtering enabled, but no content filters available.");
2175 size = (size_t)(csp->iob->eod - csp->iob->cur);
2176 old = csp->iob->cur;
2178 for (i = 0; i < MAX_AF_FILES; i++)
2181 if ((NULL == fl) || (NULL == fl->f))
2184 * Either there are no filter files
2185 * left, or this filter file just
2186 * contains no valid filters.
2188 * Continue to be sure we don't miss
2189 * valid filter files that are chained
2190 * after empty or invalid ones.
2195 * For all applying +filter actions, look if a filter by that
2196 * name exists and if yes, execute it's pcrs_joblist on the
2199 for (b = fl->f; b; b = b->next)
2201 if (b->type != FT_CONTENT_FILTER)
2203 /* Skip header filters */
2207 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
2208 filtername ; filtername = filtername->next)
2210 if (strcmp(b->name, filtername->str) == 0)
2212 int current_hits = 0; /* Number of hits caused by this filter */
2213 int job_number = 0; /* Which job we're currently executing */
2214 int job_hits = 0; /* How many hits the current job caused */
2215 pcrs_job *joblist = b->joblist;
2217 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2219 if (NULL == joblist)
2221 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2226 /* Apply all jobs from the joblist */
2227 for (job = joblist; NULL != job; job = job->next)
2230 job_hits = pcrs_execute(job, old, size, &new, &size);
2235 * That went well. Continue filtering
2236 * and use the result of this job as
2237 * input for the next one.
2239 current_hits += job_hits;
2240 if (old != csp->iob->cur)
2249 * This job caused an unexpected error. Inform the user
2250 * and skip the rest of the jobs in this filter. We could
2251 * continue with the next job, but usually the jobs
2252 * depend on each other or are similar enough to
2253 * fail for the same reason.
2255 * At the moment our pcrs expects the error codes of pcre 3.4,
2256 * but newer pcre versions can return additional error codes.
2257 * As a result pcrs_strerror()'s error message might be
2258 * "Unknown error ...", therefore we print the numerical value
2261 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2262 * should we use LOG_LEVEL_RE_FILTER instead?
2264 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2265 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2270 if (b->dynamic) pcrs_free_joblist(joblist);
2272 log_error(LOG_LEVEL_RE_FILTER,
2273 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2274 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2276 hits += current_hits;
2283 * If there were no hits, destroy our copy and let
2284 * chat() use the original in csp->iob
2292 csp->flags |= CSP_FLAG_MODIFIED;
2293 csp->content_length = size;
2301 /*********************************************************************
2303 * Function : gif_deanimate_response
2305 * Description : Deanimate the GIF image that has been accumulated in
2306 * csp->iob->buf, set csp->content_length to the modified
2307 * size and raise the CSP_FLAG_MODIFIED flag.
2310 * 1 : csp = Current client state (buffers, headers, etc...)
2312 * Returns : a pointer to the (newly allocated) modified buffer.
2313 * or NULL in case something went wrong.
2315 *********************************************************************/
2316 static char *gif_deanimate_response(struct client_state *csp)
2318 struct binbuffer *in, *out;
2322 size = (size_t)(csp->iob->eod - csp->iob->cur);
2324 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2325 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2327 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2331 in->buffer = csp->iob->cur;
2334 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2336 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2343 if ((int)size == out->offset)
2345 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2349 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2351 csp->content_length = out->offset;
2352 csp->flags |= CSP_FLAG_MODIFIED;
2362 /*********************************************************************
2364 * Function : get_filter_function
2366 * Description : Decides which content filter function has
2367 * to be applied (if any).
2370 * 1 : csp = Current client state (buffers, headers, etc...)
2372 * Returns : The content filter function to run, or
2373 * NULL if no content filter is active
2375 *********************************************************************/
2376 filter_function_ptr get_filter_function(struct client_state *csp)
2378 filter_function_ptr filter_function = NULL;
2381 * Are we enabling text mode by force?
2383 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2386 * Do we really have to?
2388 if (csp->content_type & CT_TEXT)
2390 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2394 csp->content_type |= CT_TEXT;
2395 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2399 if (!(csp->content_type & CT_DECLARED))
2402 * The server didn't bother to declare a MIME-Type.
2403 * Assume it's text that can be filtered.
2405 * This also regulary happens with 304 responses,
2406 * therefore logging anything here would cause
2409 csp->content_type |= CT_TEXT;
2413 * Choose the applying filter function based on
2414 * the content type and action settings.
2416 if ((csp->content_type & CT_TEXT) &&
2417 (csp->rlist != NULL) &&
2418 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2420 filter_function = pcrs_filter_response;
2422 else if ((csp->content_type & CT_GIF) &&
2423 (csp->action->flags & ACTION_DEANIMATE))
2425 filter_function = gif_deanimate_response;
2428 return filter_function;
2432 /*********************************************************************
2434 * Function : remove_chunked_transfer_coding
2436 * Description : In-situ remove the "chunked" transfer coding as defined
2437 * in rfc2616 from a buffer.
2440 * 1 : buffer = Pointer to the text buffer
2441 * 2 : size = In: Number of bytes to be processed,
2442 * Out: Number of bytes after de-chunking.
2443 * (undefined in case of errors)
2445 * Returns : JB_ERR_OK for success,
2446 * JB_ERR_PARSE otherwise
2448 *********************************************************************/
2449 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2452 unsigned int chunksize = 0;
2453 char *from_p, *to_p;
2456 from_p = to_p = buffer;
2458 if (sscanf(buffer, "%x", &chunksize) != 1)
2460 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2461 return JB_ERR_PARSE;
2464 while (chunksize > 0U)
2466 if (NULL == (from_p = strstr(from_p, "\r\n")))
2468 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2469 return JB_ERR_PARSE;
2472 if ((newsize += chunksize) >= *size)
2474 log_error(LOG_LEVEL_ERROR,
2475 "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding",
2477 return JB_ERR_PARSE;
2481 memmove(to_p, from_p, (size_t) chunksize);
2482 to_p = buffer + newsize;
2483 from_p += chunksize + 2;
2485 if (sscanf(from_p, "%x", &chunksize) != 1)
2487 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2492 /* XXX: Should get its own loglevel. */
2493 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2502 /*********************************************************************
2504 * Function : prepare_for_filtering
2506 * Description : If necessary, de-chunks and decompresses
2507 * the content so it can get filterd.
2510 * 1 : csp = Current client state (buffers, headers, etc...)
2512 * Returns : JB_ERR_OK for success,
2513 * JB_ERR_PARSE otherwise
2515 *********************************************************************/
2516 static jb_err prepare_for_filtering(struct client_state *csp)
2518 jb_err err = JB_ERR_OK;
2521 * If the body has a "chunked" transfer-encoding,
2522 * get rid of it, adjusting size and iob->eod
2524 if (csp->flags & CSP_FLAG_CHUNKED)
2526 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2528 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2529 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2530 if (JB_ERR_OK == err)
2532 csp->iob->eod = csp->iob->cur + size;
2533 csp->flags |= CSP_FLAG_MODIFIED;
2537 return JB_ERR_PARSE;
2543 * If the body has a supported transfer-encoding,
2544 * decompress it, adjusting size and iob->eod.
2546 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2548 if (0 == csp->iob->eod - csp->iob->cur)
2550 /* Nothing left after de-chunking. */
2554 err = decompress_iob(csp);
2556 if (JB_ERR_OK == err)
2558 csp->flags |= CSP_FLAG_MODIFIED;
2559 csp->content_type &= ~CT_TABOO;
2564 * Unset CT_GZIP and CT_DEFLATE to remember not
2565 * to modify the Content-Encoding header later.
2567 csp->content_type &= ~CT_GZIP;
2568 csp->content_type &= ~CT_DEFLATE;
2577 /*********************************************************************
2579 * Function : execute_content_filter
2581 * Description : Executes a given content filter.
2584 * 1 : csp = Current client state (buffers, headers, etc...)
2585 * 2 : content_filter = The filter function to execute
2587 * Returns : Pointer to the modified buffer, or
2588 * NULL if filtering failed or wasn't necessary.
2590 *********************************************************************/
2591 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2593 if (0 == csp->iob->eod - csp->iob->cur)
2596 * No content (probably status code 301, 302 ...),
2597 * no filtering necessary.
2602 if (JB_ERR_OK != prepare_for_filtering(csp))
2605 * failed to de-chunk or decompress.
2610 if (0 == csp->iob->eod - csp->iob->cur)
2613 * Clown alarm: chunked and/or compressed nothing delivered.
2618 return ((*content_filter)(csp));
2622 /*********************************************************************
2624 * Function : get_url_actions
2626 * Description : Gets the actions for this URL.
2629 * 1 : csp = Current client state (buffers, headers, etc...)
2630 * 2 : http = http_request request for blocked URLs
2634 *********************************************************************/
2635 void get_url_actions(struct client_state *csp, struct http_request *http)
2637 struct file_list *fl;
2638 struct url_actions *b;
2641 init_current_action(csp->action);
2643 for (i = 0; i < MAX_AF_FILES; i++)
2645 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2650 apply_url_actions(csp->action, http, b);
2657 /*********************************************************************
2659 * Function : apply_url_actions
2661 * Description : Applies a list of URL actions.
2664 * 1 : action = Destination.
2665 * 2 : http = Current URL
2666 * 3 : b = list of URL actions to apply
2670 *********************************************************************/
2671 void apply_url_actions(struct current_action_spec *action,
2672 struct http_request *http,
2673 struct url_actions *b)
2677 /* Should never happen */
2681 for (b = b->next; NULL != b; b = b->next)
2683 if (url_match(b->url, http))
2685 merge_current_action(action, b->action);
2691 /*********************************************************************
2693 * Function : get_forward_override_settings
2695 * Description : Returns forward settings as specified with the
2696 * forward-override{} action. forward-override accepts
2697 * forward lines similar to the one used in the
2698 * configuration file, but without the URL pattern.
2704 * in the configuration file can be replaced with
2705 * the action section:
2707 * {+forward-override{forward .}}
2711 * 1 : csp = Current client state (buffers, headers, etc...)
2713 * Returns : Pointer to forwarding structure in case of success.
2714 * Invalid syntax is fatal.
2716 *********************************************************************/
2717 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2719 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2720 char forward_settings[BUFFER_SIZE];
2721 char *http_parent = NULL;
2722 /* variable names were chosen for consistency reasons. */
2723 struct forward_spec *fwd = NULL;
2727 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2728 /* Should be enforced by load_one_actions_file() */
2729 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2731 /* Create a copy ssplit can modify */
2732 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2734 if (NULL != csp->fwd)
2737 * XXX: Currently necessary to prevent memory
2738 * leaks when the show-url-info cgi page is visited.
2740 unload_forward_spec(csp->fwd);
2744 * allocate a new forward node, valid only for
2745 * the lifetime of this request. Save its location
2746 * in csp as well, so sweep() can free it later on.
2748 fwd = csp->fwd = zalloc(sizeof(*fwd));
2751 log_error(LOG_LEVEL_FATAL,
2752 "can't allocate memory for forward-override{%s}", forward_override_line);
2753 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2757 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2758 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2760 fwd->type = SOCKS_NONE;
2762 /* Parse the parent HTTP proxy host:port */
2763 http_parent = vec[1];
2766 else if (vec_count == 3)
2768 char *socks_proxy = NULL;
2770 if (!strcasecmp(vec[0], "forward-socks4"))
2772 fwd->type = SOCKS_4;
2773 socks_proxy = vec[1];
2775 else if (!strcasecmp(vec[0], "forward-socks4a"))
2777 fwd->type = SOCKS_4A;
2778 socks_proxy = vec[1];
2780 else if (!strcasecmp(vec[0], "forward-socks5"))
2782 fwd->type = SOCKS_5;
2783 socks_proxy = vec[1];
2786 if (NULL != socks_proxy)
2788 /* Parse the SOCKS proxy host[:port] */
2789 fwd->gateway_host = strdup(socks_proxy);
2791 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2793 *socks_proxy++ = '\0';
2794 fwd->gateway_port = (int)strtol(socks_proxy, NULL, 0);
2797 if (fwd->gateway_port <= 0)
2799 fwd->gateway_port = 1080;
2802 http_parent = vec[2];
2806 if (NULL == http_parent)
2808 log_error(LOG_LEVEL_FATAL,
2809 "Invalid forward-override syntax in: %s", forward_override_line);
2810 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2813 /* Parse http forwarding settings */
2814 if (strcmp(http_parent, ".") != 0)
2816 fwd->forward_host = strdup(http_parent);
2818 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2820 *http_parent++ = '\0';
2821 fwd->forward_port = (int)strtol(http_parent, NULL, 0);
2824 if (fwd->forward_port <= 0)
2826 fwd->forward_port = 8000;
2830 assert (NULL != fwd);
2832 log_error(LOG_LEVEL_CONNECT,
2833 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2839 /*********************************************************************
2841 * Function : forward_url
2843 * Description : Should we forward this to another proxy?
2846 * 1 : csp = Current client state (buffers, headers, etc...)
2847 * 2 : http = http_request request for current URL
2849 * Returns : Pointer to forwarding information.
2851 *********************************************************************/
2852 const struct forward_spec *forward_url(struct client_state *csp,
2853 const struct http_request *http)
2855 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2856 struct forward_spec *fwd = csp->config->forward;
2858 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2860 return get_forward_override_settings(csp);
2870 if (url_match(fwd->url, http))
2881 /*********************************************************************
2883 * Function : direct_response
2885 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2886 * request and if so, return a HTTP 501 to the client.
2888 * FIXME: I have a stupid name and I should handle the
2889 * requests properly. Still, what we do here is rfc-
2890 * compliant, whereas ignoring or forwarding are not.
2893 * 1 : csp = Current client state (buffers, headers, etc...)
2895 * Returns : http_response if , NULL if nonmatch or handler fail
2897 *********************************************************************/
2898 struct http_response *direct_response(struct client_state *csp)
2900 struct http_response *rsp;
2901 struct list_entry *p;
2903 if ((0 == strcmpic(csp->http->gpc, "trace"))
2904 || (0 == strcmpic(csp->http->gpc, "options")))
2906 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2908 if (!strncmpic("Max-Forwards:", p->str, 13))
2910 unsigned int max_forwards;
2913 * If it's a Max-Forwards value of zero,
2914 * we have to intercept the request.
2916 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2919 * FIXME: We could handle at least TRACE here,
2920 * but that would require a verbatim copy of
2921 * the request which we don't have anymore
2923 log_error(LOG_LEVEL_HEADER,
2924 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2927 /* Get mem for response or fail*/
2928 if (NULL == (rsp = alloc_http_response()))
2930 return cgi_error_memory();
2933 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2935 free_http_response(rsp);
2936 return cgi_error_memory();
2940 rsp->reason = RSP_REASON_UNSUPPORTED;
2942 return(finish_http_response(csp, rsp));
2951 /*********************************************************************
2953 * Function : content_filters_enabled
2955 * Description : Checks whether there are any content filters
2956 * enabled for the current request.
2959 * 1 : action = Action spec to check.
2961 * Returns : TRUE for yes, FALSE otherwise
2963 *********************************************************************/
2964 int content_filters_enabled(const struct current_action_spec *action)
2966 return ((action->flags & ACTION_DEANIMATE) ||
2967 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));
2975 vim:softtabstop=3 shiftwidth=3