1 const char filters_rcs[] = "$Id: filters.c,v 1.119 2009/04/17 11:38:28 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 1.119 2009/04/17 11:38:28 fabiankeil
44 * Add and use parse_forwarder_address() to reduce code duplication.
46 * Revision 1.118 2009/04/17 11:37:03 fabiankeil
47 * Allow IPv6 addresses in forward-override{}.
49 * Revision 1.117 2009/04/17 11:35:28 fabiankeil
50 * Fix compiler warning.
52 * Revision 1.116 2009/04/17 11:34:34 fabiankeil
53 * Style cosmetics for the IPv6 code.
55 * Revision 1.115 2009/04/17 11:29:18 fabiankeil
56 * Compile fix for BSD libc.
58 * Revision 1.114 2009/04/17 11:27:49 fabiankeil
59 * Petr Pisar's privoxy-3.0.12-ipv6-3.diff.
61 * Revision 1.113 2009/03/08 14:19:23 fabiankeil
62 * Fix justified (but harmless) compiler warnings
63 * on platforms where sizeof(int) < sizeof(long).
65 * Revision 1.112 2009/03/01 18:28:23 fabiankeil
66 * Help clang understand that we aren't dereferencing
69 * Revision 1.111 2008/12/04 18:13:46 fabiankeil
70 * Fix a cparser warning.
72 * Revision 1.110 2008/11/10 16:40:25 fabiankeil
73 * Fix a gcc44 warning.
75 * Revision 1.109 2008/11/08 15:48:41 fabiankeil
76 * Mention actual values when complaining about
77 * the chunk size exceeding the buffer size.
79 * Revision 1.108 2008/05/21 15:35:08 fabiankeil
80 * - Mark csp as immutable for block_acl().
81 * - Remove an obsolete complaint about filter_popups().
83 * Revision 1.107 2008/05/04 17:52:56 fabiankeil
84 * Adjust parse_http_url() call to new prototype.
86 * Revision 1.106 2008/05/03 16:40:44 fabiankeil
87 * Change content_filters_enabled()'s parameter from
88 * csp->action to action so it can be also used in the
89 * CGI code. Don't bother checking if there are filters
90 * loaded, as that's somewhat besides the point.
92 * Revision 1.105 2008/03/28 15:13:39 fabiankeil
93 * Remove inspect-jpegs action.
95 * Revision 1.104 2008/03/27 18:27:24 fabiankeil
96 * Remove kill-popups action.
98 * Revision 1.103 2008/03/06 16:33:45 fabiankeil
99 * If limit-connect isn't used, don't limit CONNECT requests to port 443.
101 * Revision 1.102 2008/03/01 14:00:44 fabiankeil
102 * Let the block action take the reason for the block
103 * as argument and show it on the "blocked" page.
105 * Revision 1.101 2008/02/23 16:57:12 fabiankeil
106 * Rename url_actions() to get_url_actions() and let it
107 * use the standard parameter ordering.
109 * Revision 1.100 2008/02/23 16:33:43 fabiankeil
110 * Let forward_url() use the standard parameter ordering
111 * and mark its second parameter immutable.
113 * Revision 1.99 2008/02/03 13:57:58 fabiankeil
114 * Add SOCKS5 support for forward-override{}.
116 * Revision 1.98 2008/01/04 17:43:45 fabiankeil
117 * Improve the warning messages that get logged if the action files
118 * "enable" filters but no filters of that type have been loaded.
120 * Revision 1.97 2007/11/30 15:37:03 fabiankeil
121 * Use freez instead of free.
123 * Revision 1.96 2007/10/19 16:53:28 fabiankeil
124 * Add helper function to check if any content filters are enabled.
126 * Revision 1.95 2007/10/17 19:31:20 fabiankeil
127 * Omitting the zero chunk that ends the chunk transfer encoding seems
128 * to be the new black. Log the problem and continue filtering anyway.
130 * Revision 1.94 2007/09/29 13:20:20 fabiankeil
131 * Remove two redundant and one useless log messages.
133 * Revision 1.93 2007/09/29 10:21:16 fabiankeil
134 * - Move get_filter_function() from jcc.c to filters.c
135 * so the filter functions can be static.
136 * - Don't bother filtering body-less responses.
138 * Revision 1.92 2007/09/28 16:38:55 fabiankeil
139 * - Execute content filters through execute_content_filter().
140 * - Add prepare_for_filtering() so filter functions don't have to
141 * care about de-chunking and decompression. As a side effect this enables
142 * decompression for gif_deanimate_response() and jpeg_inspect_response().
143 * - Change remove_chunked_transfer_coding()'s return type to jb_err.
144 * Some clowns feel like chunking empty responses in which case
145 * (size == 0) is valid but previously would be interpreted as error.
147 * Revision 1.91 2007/09/02 15:31:20 fabiankeil
148 * Move match_portlist() from filter.c to urlmatch.c.
149 * It's used for url matching, not for filtering.
151 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
152 * Remove newline at the end of a log_error() message.
154 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
155 * #1763173 from Stefan Huehner: declare some more functions static.
157 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
158 * Add forward-override{} to change the forwarding settings through
159 * action sections. This is mainly interesting to forward different
160 * clients differently (for example based on User-Agent or request
163 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
164 * Make sure filters with dynamic jobs actually use them.
166 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
167 * - Introduce dynamic pcrs jobs that can resolve variables.
168 * - Don't run redirect functions more than once,
169 * unless they are activated more than once.
171 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
172 * - Log the content size after decompression in decompress_iob()
173 * instead of pcrs_filter_response().
175 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
176 * Use dedicated header filter actions instead of abusing "filter".
177 * Replace "filter-client-headers" and "filter-client-headers"
178 * with "server-header-filter" and "client-header-filter".
180 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
181 * New config option: enforce-blocks.
183 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
184 * - Fix port handling in acl_addr() and use a temporary acl spec
185 * copy so error messages don't contain a truncated version.
186 * - Log size of iob before and after decompression.
188 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
189 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
190 * - Hide the "Go there anyway" link for blocked CONNECT
191 * requests where going there anyway doesn't work anyway.
193 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
194 * - Save the reason for generating http_responses.
195 * - Block (+block) with status code 403 instead of 404.
196 * - Use a different kludge to remember a failed decompression.
198 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
199 * Search for Max-Forwards headers case-insensitive,
200 * don't generate the "501 unsupported" message for invalid
201 * Max-Forwards values and don't increase negative ones.
203 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
204 * - Add HEAD support to finish_http_response.
205 * - Add error favicon to internal HTML error messages.
207 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
208 * Mark *csp as immutable for is_untrusted_url()
209 * and is_imageurl(). Closes FR 1237736.
211 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
212 * Integrate a modified version of Wil Mahan's
213 * zlib patch (PR #895531).
215 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
216 * Fixed gcc43 conversion warnings,
217 * changed sprintf calls to snprintf.
219 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
220 * Adjust comment in pcrs_filter_response()
221 * to recent pcrs changes. Hohoho.
223 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
224 * Don't crash if pcre returns an error code
225 * that pcrs didn't expect. Fixes BR 1621173.
227 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
228 * Modified is_untrusted_url to complain in case of
229 * write errors and to give a reason when adding new
230 * entries to the trustfile. Closes FR 1097611.
232 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
233 * Skip empty filter files in pcrs_filter_response,
234 * but don't ignore the ones that come afterwards.
235 * Fixes parts of BR 1619208.
237 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
238 * Added some sanity checks for get_last_url().
239 * Fixed possible segfault caused by my last commit.
241 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
242 * Let get_last_url() catch https URLs as well.
244 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
245 * Make sure get_last_url() behaves like advertised
246 * and fast-redirects{} can be combined with redirect{}.
248 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
249 * Implemented +redirect{s@foo@bar@} to generate
250 * a redirect based on a rewritten version of the
253 * Revision 1.66 2006/09/23 13:26:38 roro
254 * Replace TABs by spaces in source code.
256 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
257 * Fix +redirect{}. Didn't work with -fast-redirects.
259 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
260 * Block requests for untrusted URLs with status
261 * code 403 instead of 200.
263 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
264 * Don't free p which is still in use and will be later
265 * freed by free_map(). Don't claim the referrer is unknown
266 * when the client didn't set one.
268 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
269 * Feature request 595948: Re-Filter logging in single line
271 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
272 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
274 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
275 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
276 * with what was really the latest development (the v_3_0_branch branch)
278 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
279 * Multiple filter file support
281 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
282 * A few memory leaks plugged (mostly on error paths)
284 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
285 * Add the ability to check jpeg images for invalid
286 * lengths of comment blocks. Defensive strategy
287 * against the exploit:
288 * Microsoft Security Bulletin MS04-028
289 * Buffer Overrun in JPEG Processing (GDI+) Could
290 * Allow Code Execution (833987)
291 * Enabled with +inspect-jpegs in actions files.
293 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
294 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
296 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
297 * Fixed bug #839859: "See why" link URL now gets url-encoded.
299 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
302 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
303 * Added basic support for OPTIONS and TRACE HTTP methods:
304 * New function direct_response which handles OPTIONS and
305 * TRACE requests whose Max-Forwards header field is zero.
307 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
308 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
310 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
311 * - Bugfix: Executing a filters without jobs no longer results in
312 * turing off *all* filters.
313 * - Security fix: Malicious web servers can't cause a seg fault
314 * through bogus chunk sizes anymore
316 * Revision 1.58 2002/04/24 02:11:17 oes
317 * Jon's multiple AF patch: url_actions now evaluates rules
320 * Revision 1.57 2002/04/08 20:38:34 swa
323 * Revision 1.56 2002/04/05 15:51:24 oes
324 * - bugfix: error-pages now get correct request protocol
325 * - fix for invalid HTML in trust info
327 * Revision 1.55 2002/04/02 16:13:51 oes
328 * Fix: No "Go there anyway" for SSL
330 * Revision 1.54 2002/04/02 14:55:56 oes
331 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
333 * Revision 1.53 2002/03/26 22:29:54 swa
334 * we have a new homepage!
336 * Revision 1.52 2002/03/24 16:35:57 jongfoster
339 * Revision 1.51 2002/03/24 15:23:33 jongfoster
342 * Revision 1.50 2002/03/24 13:25:43 swa
343 * name change related issues
345 * Revision 1.49 2002/03/16 20:29:14 oes
348 * Revision 1.48 2002/03/13 20:25:34 oes
349 * Better logging for content filters
351 * Revision 1.47 2002/03/13 00:30:52 jongfoster
353 * Added option of always sending redirect for imageblock,
354 * currently disabled with #if 0.
356 * Revision 1.46 2002/03/12 01:42:49 oes
357 * Introduced modular filters
359 * Revision 1.45 2002/03/08 16:47:50 oes
360 * Added choice beween GIF and PNG built-in images
362 * Revision 1.44 2002/03/07 03:49:31 oes
363 * - Fixed compiler warnings etc
364 * - Changed built-in images from GIF to PNG
365 * (with regard to Unisys patent issue)
366 * - Added a 4x4 pattern PNG which is less intrusive
367 * than the logo but also clearly marks the deleted banners
369 * Revision 1.43 2002/01/22 23:51:59 jongfoster
370 * Replacing strsav() with the safer string_append().
372 * Adding missing html_encode() to error message generators. Where encoded
373 * and unencoded versions of a string were provided, removing the unencoded
376 * Revision 1.42 2002/01/17 21:00:32 jongfoster
377 * Moving all our URL and URL pattern parsing code to urlmatch.c.
379 * Using a single, simple url_match(pattern,url) function - rather than
380 * the 3-line match routine which was repeated all over the place.
382 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
384 * Using parse_http_url() to parse URLs without faking a HTTP
385 * request line for parse_http_request().
387 * Revision 1.41 2001/11/13 00:14:07 jongfoster
388 * Fixing stupid bug now I've figured out what || means.
389 * (It always returns 0 or 1, not one of it's paramaters.)
391 * Revision 1.40 2001/10/26 17:37:55 oes
392 * - Re-enabled Netscape 200/404 bug workaround in block_url():
393 * - Removed OS/2 special case
394 * - Made block_url() independant from sed() having been run
395 * - Made trust_url independant from sed() having been run
396 * - Made is_imageurl independant from sed() having been run.
397 * It now checks User-Agent: and Accept: by itself.
400 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
401 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
402 * threads to call select() simultaneously. So, it's time to do a real, live,
403 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
404 * (native). Both versions will work, but using __OS2__ offers multi-threading.
406 * Revision 1.38 2001/10/23 21:32:33 jongfoster
407 * Adding error-checking to selected functions
409 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
410 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
411 * filters.c. Added a FIXME in front of the offending code. I'll gladly
412 * put in a better/more robust fix for all parties if one is presented...
413 * It seems that just returning 200 instead of 404 would pretty much fix
414 * it for everyone, but I don't know all the history of the problem.
416 * Revision 1.36 2001/10/10 16:44:16 oes
417 * Added match_portlist function
419 * Revision 1.35 2001/10/07 15:41:23 oes
420 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
422 * New function remove_chunked_transfer_coding that strips chunked
423 * transfer coding to plain and is called by pcrs_filter_response
424 * and gif_deanimate_response if neccessary
426 * Improved handling of zero-change re_filter runs
428 * pcrs_filter_response and gif_deanimate_response now remove
429 * chunked transfer codeing before processing the body.
431 * Revision 1.34 2001/09/20 15:49:36 steudten
433 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
434 * See cgi.c fill_template().
436 * Revision 1.33 2001/09/16 17:05:14 jongfoster
437 * Removing unused #include showarg.h
439 * Revision 1.32 2001/09/16 13:21:27 jongfoster
440 * Changes to use new list functions.
442 * Revision 1.31 2001/09/16 11:38:02 jongfoster
443 * Splitting fill_template() into 2 functions:
444 * template_load() loads the file
445 * template_fill() performs the PCRS regexps.
446 * This is because the CGI edit interface has a "table row"
447 * template which is used many times in the page - this
448 * change means it's only loaded from disk once.
450 * Revision 1.30 2001/09/16 11:00:10 jongfoster
451 * New function alloc_http_response, for symmetry with free_http_response
453 * Revision 1.29 2001/09/13 23:32:40 jongfoster
454 * Moving image data to cgi.c rather than cgi.h
455 * Fixing a GPF under Win32 (and any other OS that protects global
456 * constants from being written to).
458 * Revision 1.28 2001/09/10 10:18:51 oes
459 * Silenced compiler warnings
461 * Revision 1.27 2001/08/05 16:06:20 jongfoster
462 * Modifiying "struct map" so that there are now separate header and
463 * "map_entry" structures. This means that functions which modify a
464 * map no longer need to return a pointer to the modified map.
465 * Also, it no longer reverses the order of the entries (which may be
466 * important with some advanced template substitutions).
468 * Revision 1.26 2001/07/30 22:08:36 jongfoster
469 * Tidying up #defines:
470 * - All feature #defines are now of the form FEATURE_xxx
471 * - Permanently turned off WIN_GUI_EDIT
472 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
474 * Revision 1.25 2001/07/26 10:09:46 oes
475 * Made browser detection a little less naive
477 * Revision 1.24 2001/07/25 17:22:51 oes
478 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
480 * Revision 1.23 2001/07/23 13:40:12 oes
481 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
483 * Revision 1.22 2001/07/18 12:29:34 oes
484 * - Made gif_deanimate_response respect
485 * csp->action->string[ACTION_STRING_DEANIMATE]
486 * - Logging cosmetics
488 * Revision 1.21 2001/07/13 13:59:53 oes
489 * - Introduced gif_deanimate_response which shares the
490 * generic content modification interface of pcrs_filter_response
491 * and acts as a wrapper to deanimate.c:gif_deanimate()
492 * - Renamed re_process_buffer to pcrs_filter_response
493 * - pcrs_filter_response now returns NULL on failiure
494 * - Removed all #ifdef PCRS
496 * Revision 1.20 2001/07/01 17:01:04 oes
497 * Added comments and missing return statement in is_untrusted_url()
499 * Revision 1.19 2001/06/29 21:45:41 oes
500 * Indentation, CRLF->LF, Tab-> Space
502 * Revision 1.18 2001/06/29 13:27:38 oes
503 * - Cleaned up, renamed and reorderd functions
504 * and improved comments
507 * - Ported to CGI platform. Now delivers
508 * http_response or NULL
509 * - Unified HTML and GIF generation (moved image detection
510 * and GIF generation here from jcc.c:chat())
511 * - Fixed HTTP status to:
512 * - 403 (Forbidden) for the "blocked" HTML message
513 * - 200 (OK) for GIF answers
514 * - 302 (Redirect) for redirect to GIF
517 * - Ported to CGI platform. Now delivers
518 * http_response or NULL
519 * - Separated detection of untrusted URL into
520 * (bool)is_untrusted_url
521 * - Added enforcement of untrusted requests
523 * - Moved redirect_url() from cgi.c to here
524 * and ported it to the CGI platform
526 * - Removed logentry from cancelled commit
528 * Revision 1.17 2001/06/09 10:55:28 jongfoster
529 * Changing BUFSIZ ==> BUFFER_SIZE
531 * Revision 1.16 2001/06/07 23:10:26 jongfoster
532 * Allowing unanchored domain patterns to back off and retry
533 * if they partially match. Optimized right-anchored patterns.
534 * Moving ACL and forward files into config file.
535 * Replacing struct gateway with struct forward_spec
537 * Revision 1.15 2001/06/03 19:12:00 oes
538 * extracted-CGI relevant stuff
540 * Revision 1.14 2001/06/01 10:30:55 oes
541 * Added optional left-anchoring to domaincmp
543 * Revision 1.13 2001/05/31 21:21:30 jongfoster
544 * Permissionsfile / actions file changes:
545 * - Changed "permission" to "action" throughout
546 * - changes to file format to allow string parameters
547 * - Moved helper functions to actions.c
549 * Revision 1.12 2001/05/31 17:35:20 oes
551 * - Enhanced domain part globbing with infix and prefix asterisk
552 * matching and optional unanchored operation
554 * Revision 1.11 2001/05/29 11:53:23 oes
555 * "See why" link added to "blocked" page
557 * Revision 1.10 2001/05/29 09:50:24 jongfoster
558 * Unified blocklist/imagelist/permissionslist.
559 * File format is still under discussion, but the internal changes
562 * Also modified interceptor behaviour:
563 * - We now intercept all URLs beginning with one of the following
564 * prefixes (and *only* these prefixes):
566 * * http://ijbswa.sf.net/config/
567 * * http://ijbswa.sourceforge.net/config/
568 * - New interceptors "home page" - go to http://i.j.b/ to see it.
569 * - Internal changes so that intercepted and fast redirect pages
570 * are not replaced with an image.
571 * - Interceptors now have the option to send a binary page direct
572 * to the client. (i.e. ijb-send-banner uses this)
573 * - Implemented show-url-info interceptor. (Which is why I needed
574 * the above interceptors changes - a typical URL is
575 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
576 * The previous mechanism would not have intercepted that, and
577 * if it had been intercepted then it then it would have replaced
580 * Revision 1.9 2001/05/27 22:17:04 oes
582 * - re_process_buffer no longer writes the modified buffer
583 * to the client, which was very ugly. It now returns the
584 * buffer, which it is then written by chat.
586 * - content_length now adjusts the Content-Length: header
587 * for modified documents rather than crunch()ing it.
588 * (Length info in csp->content_length, which is 0 for
589 * unmodified documents)
591 * - For this to work, sed() is called twice when filtering.
593 * Revision 1.8 2001/05/26 17:13:28 jongfoster
594 * Filled in a function comment.
596 * Revision 1.7 2001/05/26 15:26:15 jongfoster
597 * ACL feature now provides more security by immediately dropping
598 * connections from untrusted hosts.
600 * Revision 1.6 2001/05/26 00:28:36 jongfoster
601 * Automatic reloading of config file.
602 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
603 * Most of the global variables have been moved to a new
604 * struct configuration_spec, accessed through csp->config->globalname
605 * Most of the globals remaining are used by the Win32 GUI.
607 * Revision 1.5 2001/05/25 22:34:30 jongfoster
610 * Revision 1.4 2001/05/22 18:46:04 oes
612 * - Enabled filtering banners by size rather than URL
613 * by adding patterns that replace all standard banner
614 * sizes with the "Junkbuster" gif to the re_filterfile
616 * - Enabled filtering WebBugs by providing a pattern
617 * which kills all 1x1 images
619 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
620 * which is selected by the (nonstandard and therefore
621 * capital) letter 'U' in the option string.
622 * It causes the quantifiers to be ungreedy by default.
623 * Appending a ? turns back to greedy (!).
625 * - Added a new interceptor ijb-send-banner, which
626 * sends back the "Junkbuster" gif. Without imagelist or
627 * MSIE detection support, or if tinygif = 1, or the
628 * URL isn't recognized as an imageurl, a lame HTML
629 * explanation is sent instead.
631 * - Added new feature, which permits blocking remote
632 * script redirects and firing back a local redirect
634 * The feature is conditionally compiled, i.e. it
635 * can be disabled with --disable-fast-redirects,
636 * plus it must be activated by a "fast-redirects"
637 * line in the config file, has its own log level
638 * and of course wants to be displayed by show-proxy-args
639 * Note: Boy, all the #ifdefs in 1001 locations and
640 * all the fumbling with configure.in and acconfig.h
641 * were *way* more work than the feature itself :-(
643 * - Because a generic redirect template was needed for
644 * this, tinygif = 3 now uses the same.
646 * - Moved GIFs, and other static HTTP response templates
651 * - Removed some >400 CRs again (Jon, you really worked
654 * Revision 1.3 2001/05/20 16:44:47 jongfoster
655 * Removing last hardcoded Junkbusters.com URLs.
657 * Revision 1.2 2001/05/20 01:21:20 jongfoster
658 * Version 2.9.4 checkin.
659 * - Merged popupfile and cookiefile, and added control over PCRS
660 * filtering, in new "permissionsfile".
661 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
662 * file error you now get a message box (in the Win32 GUI) rather
663 * than the program exiting with no explanation.
664 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
666 * - Removed tabs from "config"
667 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
668 * - Bumped up version number.
670 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
671 * Initial import of version 2.9.3 source tree
674 *********************************************************************/
680 #include <sys/types.h>
688 #include <sys/socket.h>
689 #endif /* def HAVE_RFC2553 */
694 #endif /* ndef __OS2__ */
695 #include <netinet/in.h>
697 #include <winsock2.h>
698 #endif /* ndef _WIN32 */
702 #endif /* def __OS2__ */
710 #include "jbsockets.h"
711 #include "miscutil.h"
715 #include "deanimate.h"
716 #include "urlmatch.h"
723 const char filters_h_rcs[] = FILTERS_H_VERSION;
725 /* Fix a problem with Solaris. There should be no effect on other
727 * Solaris's isspace() is a macro which uses it's argument directly
728 * as an array index. Therefore we need to make sure that high-bit
729 * characters generate +ve values, and ideally we also want to make
730 * the argument match the declared parameter type of "int".
732 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
734 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
735 static jb_err prepare_for_filtering(struct client_state *csp);
739 /*********************************************************************
741 * Function : sockaddr_storage_to_ip
743 * Description : Access internal structure of sockaddr_storage
746 * 1 : addr = socket address
747 * 2 : ip = IP address as array of octets in network order
748 * (it points into addr)
749 * 3 : len = length of IP address in octets
750 * 4 : port = port number in network order;
752 * Returns : 0 = no errror; -1 otherwise.
754 *********************************************************************/
755 int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, uint8_t **ip,
756 unsigned int *len, in_port_t **port)
763 switch (addr->ss_family)
773 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
777 *port = &((struct sockaddr_in *)addr)->sin_port;
788 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
792 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
797 /* Unsupported address family */
805 /*********************************************************************
807 * Function : match_sockaddr
809 * Description : Check whether address matches network (IP address and port)
812 * 1 : network = socket address of subnework
813 * 2 : netmask = network mask as socket address
814 * 3 : address = checked socket address against given network
816 * Returns : 0 = doesn't match; 1 = does match
818 *********************************************************************/
819 int match_sockaddr(const struct sockaddr_storage *network,
820 const struct sockaddr_storage *netmask,
821 const struct sockaddr_storage *address)
823 uint8_t *network_addr, *netmask_addr, *address_addr;
824 unsigned int addr_len;
825 in_port_t *network_port, *netmask_port, *address_port;
828 if (network->ss_family != netmask->ss_family)
830 /* This should never happen */
831 log_error(LOG_LEVEL_ERROR,
832 "Internal error at %s:%llu: network and netmask differ in family",
837 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
838 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
839 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
841 /* Check for family */
842 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
843 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
845 /* Map AF_INET6 V4MAPPED address into AF_INET */
849 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
850 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
852 /* Map AF_INET6 V4MAPPED network into AF_INET */
857 else if (network->ss_family != address->ss_family)
862 /* XXX: Port check is signaled in netmask */
863 if (*netmask_port && *network_port != *address_port)
868 /* TODO: Optimize by checking by words insted of octets */
869 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
871 if ((network_addr[i] & netmask_addr[i]) !=
872 (address_addr[i] & netmask_addr[i]))
880 #endif /* def HAVE_RFC2553 */
883 /*********************************************************************
885 * Function : block_acl
887 * Description : Block this request?
888 * Decide yes or no based on ACL file.
891 * 1 : dst = The proxy or gateway address this is going to.
892 * Or NULL to check all possible targets.
893 * 2 : csp = Current client state (buffers, headers, etc...)
894 * Also includes the client IP address.
896 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
898 *********************************************************************/
899 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
901 struct access_control_list *acl = csp->config->acl;
903 /* if not using an access control list, then permit the connection */
909 /* search the list */
914 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
916 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
922 /* Just want to check if they have any access */
923 if (acl->action == ACL_PERMIT)
931 * XXX: An undefined acl->dst is full of zeros and should be
932 * considered a wildcard address. sockaddr_storage_to_ip()
933 * fails on such destinations because of unknown sa_familly
934 * (glibc only?). However this test is not portable.
936 * So, we signal the acl->dst is wildcard in wildcard_dst.
939 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
941 ((dst->addr & acl->dst->mask) == acl->dst->addr)
942 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
946 if (acl->action == ACL_PERMIT)
964 /*********************************************************************
966 * Function : acl_addr
968 * Description : Called from `load_config' to parse an ACL address.
971 * 1 : aspec = String specifying ACL address.
972 * 2 : aca = struct access_control_addr to fill in.
974 * Returns : 0 => Ok, everything else is an error.
976 *********************************************************************/
977 int acl_addr(const char *aspec, struct access_control_addr *aca)
981 struct addrinfo hints, *result;
983 in_port_t *mask_port;
984 unsigned int addr_len;
987 #endif /* def HAVE_RFC2553 */
989 char *acl_spec = NULL;
992 /* XXX: Depend on ai_family */
1000 * Use a temporary acl spec copy so we can log
1001 * the unmodified original in case of parse errors.
1003 acl_spec = strdup(aspec);
1004 if (acl_spec == NULL)
1006 /* XXX: This will be logged as parse error. */
1010 if ((p = strchr(acl_spec, '/')) != NULL)
1013 if (ijb_isdigit(*p) == 0)
1018 masklength = atoi(p);
1021 if ((masklength < 0) ||
1033 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
1036 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
1045 p = strchr(acl_spec, ':');
1049 memset(&hints, 0, sizeof(struct addrinfo));
1050 hints.ai_family = AF_UNSPEC;
1051 hints.ai_socktype = SOCK_STREAM;
1053 i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result);
1058 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
1059 acl_spec, p, gai_strerror(i));
1063 /* TODO: Allow multihomed hostnames */
1064 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
1065 freeaddrinfo(result);
1072 port = strtol(p, &endptr, 10);
1074 if (port <= 0 || port > 65535 || *endptr != '\0')
1081 aca->port = (unsigned long)port;
1083 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
1086 if (aca->addr == INADDR_NONE)
1088 /* XXX: This will be logged as parse error. */
1091 #endif /* def HAVE_RFC2553 */
1093 /* build the netmask */
1095 /* Clip masklength according to current family. */
1096 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
1101 aca->mask.ss_family = aca->addr.ss_family;
1102 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
1109 /* ACL contains a port number, check ports in the future. */
1114 * XXX: This could be optimized to operate on whole words instead
1115 * of octets (128-bit CPU could do it in one iteration).
1118 * Octets after prefix can be ommitted because of
1119 * previous initialization to zeros.
1121 for (i = 0; (i < addr_len) && masklength; i++)
1123 if (masklength >= 8)
1125 mask_data[i] = 0xFF;
1131 * XXX: This assumes MSB of octet is on the left side.
1132 * This should be true for all architectures or solved
1133 * by the link layer.
1135 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
1142 for (i=1; i <= masklength ; i++)
1144 aca->mask |= (1U << (32 - i));
1147 /* now mask off the host portion of the ip address
1148 * (i.e. save on the network portion of the address).
1150 aca->addr = aca->addr & aca->mask;
1151 #endif /* def HAVE_RFC2553 */
1156 #endif /* def FEATURE_ACL */
1159 /*********************************************************************
1161 * Function : connect_port_is_forbidden
1163 * Description : Check to see if CONNECT requests to the destination
1164 * port of this request are forbidden. The check is
1165 * independend of the actual request method.
1168 * 1 : csp = Current client state (buffers, headers, etc...)
1170 * Returns : True if yes, false otherwise.
1172 *********************************************************************/
1173 int connect_port_is_forbidden(const struct client_state *csp)
1175 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
1176 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
1181 /*********************************************************************
1183 * Function : block_url
1185 * Description : Called from `chat'. Check to see if we need to block this.
1188 * 1 : csp = Current client state (buffers, headers, etc...)
1190 * Returns : NULL => unblocked, else HTTP block response
1192 *********************************************************************/
1193 struct http_response *block_url(struct client_state *csp)
1195 struct http_response *rsp;
1196 const char *new_content_type = NULL;
1199 * If it's not blocked, don't block it ;-)
1201 if ((csp->action->flags & ACTION_BLOCK) == 0)
1205 if (csp->action->flags & ACTION_REDIRECT)
1207 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
1210 * Else, prepare a response
1212 if (NULL == (rsp = alloc_http_response()))
1214 return cgi_error_memory();
1218 * If it's an image-url, send back an image or redirect
1219 * as specified by the relevant +image action
1221 #ifdef FEATURE_IMAGE_BLOCKING
1222 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
1223 && is_imageurl(csp))
1226 /* determine HOW images should be blocked */
1227 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
1229 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1231 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
1233 #if 1 /* Two alternative strategies, use this one for now: */
1235 /* and handle accordingly: */
1236 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1238 rsp->status = strdup("403 Request blocked by Privoxy");
1239 if (rsp->status == NULL)
1241 free_http_response(rsp);
1242 return cgi_error_memory();
1244 rsp->body = bindup(image_pattern_data, image_pattern_length);
1245 if (rsp->body == NULL)
1247 free_http_response(rsp);
1248 return cgi_error_memory();
1250 rsp->content_length = image_pattern_length;
1252 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1254 free_http_response(rsp);
1255 return cgi_error_memory();
1259 else if (0 == strcmpic(p, "blank"))
1261 rsp->status = strdup("403 Request blocked by Privoxy");
1262 if (rsp->status == NULL)
1264 free_http_response(rsp);
1265 return cgi_error_memory();
1267 rsp->body = bindup(image_blank_data, image_blank_length);
1268 if (rsp->body == NULL)
1270 free_http_response(rsp);
1271 return cgi_error_memory();
1273 rsp->content_length = image_blank_length;
1275 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1277 free_http_response(rsp);
1278 return cgi_error_memory();
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();
1298 #else /* Following code is disabled for now */
1300 /* and handle accordingly: */
1301 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1303 p = CGI_PREFIX "send-banner?type=pattern";
1305 else if (0 == strcmpic(p, "blank"))
1307 p = CGI_PREFIX "send-banner?type=blank";
1309 rsp->status = strdup("302 Local Redirect from Privoxy");
1310 if (rsp->status == NULL)
1312 free_http_response(rsp);
1313 return cgi_error_memory();
1316 if (enlist_unique_header(rsp->headers, "Location", p))
1318 free_http_response(rsp);
1319 return cgi_error_memory();
1321 #endif /* Preceeding code is disabled for now */
1323 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1326 * Send empty document.
1328 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1331 rsp->body = strdup(" ");
1332 rsp->content_length = 1;
1334 rsp->status = strdup("403 Request blocked by Privoxy");
1335 if (rsp->status == NULL)
1337 free_http_response(rsp);
1338 return cgi_error_memory();
1340 if (new_content_type != 0)
1342 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1343 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1345 free_http_response(rsp);
1346 return cgi_error_memory();
1351 #endif /* def FEATURE_IMAGE_BLOCKING */
1354 * Else, generate an HTML "blocked" message:
1358 struct map * exports;
1362 * Workaround for stupid Netscape bug which prevents
1363 * pages from being displayed if loading a referenced
1364 * JavaScript or style sheet fails. So make it appear
1365 * as if it succeeded.
1367 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1368 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1369 && !strstr(p, "Gecko") /* save Mozilla, */
1370 && !strstr(p, "compatible") /* MSIE */
1371 && !strstr(p, "Opera")) /* and Opera. */
1373 rsp->status = strdup("200 Request for blocked URL");
1377 rsp->status = strdup("403 Request for blocked URL");
1380 if (rsp->status == NULL)
1382 free_http_response(rsp);
1383 return cgi_error_memory();
1386 exports = default_exports(csp, NULL);
1387 if (exports == NULL)
1389 free_http_response(rsp);
1390 return cgi_error_memory();
1393 #ifdef FEATURE_FORCE_LOAD
1394 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1396 * Export the force conditional block killer if
1398 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1399 * - Privoxy is configured to enforce blocks, or
1400 * - it's a CONNECT request and enforcing wouldn't work anyway.
1402 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1403 || (0 == strcmpic(csp->http->gpc, "connect")))
1404 #endif /* ndef FEATURE_FORCE_LOAD */
1406 err = map_block_killer(exports, "force-support");
1409 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1410 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1411 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1412 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1415 const char *block_reason;
1416 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
1418 block_reason = csp->action->string[ACTION_STRING_BLOCK];
1422 assert(connect_port_is_forbidden(csp));
1423 block_reason = "Forbidden CONNECT port.";
1425 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
1430 free_http_response(rsp);
1431 return cgi_error_memory();
1434 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1437 free_http_response(rsp);
1438 return cgi_error_memory();
1441 rsp->reason = RSP_REASON_BLOCKED;
1443 return finish_http_response(csp, rsp);
1448 #ifdef FEATURE_TRUST
1449 /*********************************************************************
1451 * Function : trust_url FIXME: I should be called distrust_url
1453 * Description : Calls is_untrusted_url to determine if the URL is trusted
1454 * and if not, returns a HTTP 403 response with a reject message.
1457 * 1 : csp = Current client state (buffers, headers, etc...)
1459 * Returns : NULL => trusted, else http_response.
1461 *********************************************************************/
1462 struct http_response *trust_url(struct client_state *csp)
1464 struct http_response *rsp;
1465 struct map * exports;
1466 char buf[BUFFER_SIZE];
1468 struct url_spec **tl;
1473 * Don't bother to work on trusted URLs
1475 if (!is_untrusted_url(csp))
1481 * Else, prepare a response:
1483 if (NULL == (rsp = alloc_http_response()))
1485 return cgi_error_memory();
1488 rsp->status = strdup("403 Request blocked by Privoxy");
1489 exports = default_exports(csp, NULL);
1490 if (exports == NULL || rsp->status == NULL)
1492 free_http_response(rsp);
1493 return cgi_error_memory();
1497 * Export the protocol, host, port, and referrer information
1499 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1500 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1501 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1503 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1505 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1509 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1515 free_http_response(rsp);
1516 return cgi_error_memory();
1520 * Export the trust list
1523 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1525 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1526 string_append(&p, buf);
1528 err = map(exports, "trusted-referrers", 1, p, 0);
1533 free_http_response(rsp);
1534 return cgi_error_memory();
1538 * Export the trust info, if available
1540 if (csp->config->trust_info->first)
1542 struct list_entry *l;
1545 for (l = csp->config->trust_info->first; l ; l = l->next)
1547 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1548 string_append(&p, buf);
1550 err = map(exports, "trust-info", 1, p, 0);
1554 err = map_block_killer(exports, "have-trust-info");
1560 free_http_response(rsp);
1561 return cgi_error_memory();
1565 * Export the force conditional block killer if
1567 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1568 * - Privoxy is configured to enforce blocks, or
1569 * - it's a CONNECT request and enforcing wouldn't work anyway.
1571 #ifdef FEATURE_FORCE_LOAD
1572 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1573 || (0 == strcmpic(csp->http->gpc, "connect")))
1575 err = map_block_killer(exports, "force-support");
1579 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1581 #else /* ifndef FEATURE_FORCE_LOAD */
1582 err = map_block_killer(exports, "force-support");
1583 #endif /* ndef FEATURE_FORCE_LOAD */
1588 free_http_response(rsp);
1589 return cgi_error_memory();
1593 * Build the response
1595 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1598 free_http_response(rsp);
1599 return cgi_error_memory();
1601 rsp->reason = RSP_REASON_UNTRUSTED;
1603 return finish_http_response(csp, rsp);
1605 #endif /* def FEATURE_TRUST */
1608 /*********************************************************************
1610 * Function : compile_dynamic_pcrs_job_list
1612 * Description : Compiles a dynamic pcrs job list (one with variables
1613 * resolved at request time)
1616 * 1 : csp = Current client state (buffers, headers, etc...)
1617 * 2 : b = The filter list to compile
1619 * Returns : NULL in case of errors, otherwise the
1622 *********************************************************************/
1623 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1625 struct list_entry *pattern;
1626 pcrs_job *job_list = NULL;
1627 pcrs_job *dummy = NULL;
1628 pcrs_job *lastjob = NULL;
1631 const struct pcrs_variable variables[] =
1633 {"url", csp->http->url, 1},
1634 {"path", csp->http->path, 1},
1635 {"host", csp->http->host, 1},
1636 {"origin", csp->ip_addr_str, 1},
1640 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1642 assert(pattern->str != NULL);
1644 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1648 log_error(LOG_LEVEL_ERROR,
1649 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1650 pattern->str, b->name, pcrs_strerror(error));
1655 if (error == PCRS_WARN_TRUNCATION)
1657 log_error(LOG_LEVEL_ERROR,
1658 "At least one of the variables in \'%s\' had to "
1659 "be truncated before compilation", pattern->str);
1661 if (job_list == NULL)
1667 lastjob->next = dummy;
1677 /*********************************************************************
1679 * Function : rewrite_url
1681 * Description : Rewrites a URL with a single pcrs command
1682 * and returns the result if it differs from the
1683 * original and isn't obviously invalid.
1686 * 1 : old_url = URL to rewrite.
1687 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1690 * Returns : NULL if the pcrs_command didn't change the url, or
1691 * the result of the modification.
1693 *********************************************************************/
1694 char *rewrite_url(char *old_url, const char *pcrs_command)
1696 char *new_url = NULL;
1700 assert(pcrs_command);
1702 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1706 log_error(LOG_LEVEL_REDIRECTS,
1707 "pcrs command \"%s\" didn't change \"%s\".",
1708 pcrs_command, old_url);
1713 log_error(LOG_LEVEL_REDIRECTS,
1714 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1715 pcrs_command, old_url, pcrs_strerror(hits));
1718 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1720 log_error(LOG_LEVEL_ERROR,
1721 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1722 "but the result doesn't look like a valid URL and will be ignored.",
1723 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1728 log_error(LOG_LEVEL_REDIRECTS,
1729 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1730 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1738 #ifdef FEATURE_FAST_REDIRECTS
1739 /*********************************************************************
1741 * Function : get_last_url
1743 * Description : Search for the last URL inside a string.
1744 * If the string already is a URL, it will
1745 * be the first URL found.
1748 * 1 : subject = the string to check
1749 * 2 : redirect_mode = +fast-redirect{} mode
1751 * Returns : NULL if no URL was found, or
1752 * the last URL found.
1754 *********************************************************************/
1755 char *get_last_url(char *subject, const char *redirect_mode)
1757 char *new_url = NULL;
1761 assert(redirect_mode);
1763 subject = strdup(subject);
1764 if (subject == NULL)
1766 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1770 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1772 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1773 new_url = url_decode(subject);
1774 if (new_url != NULL)
1781 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1785 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1788 * Find the last URL encoded in the request
1791 while ((tmp = strstr(tmp, "http://")) != NULL)
1795 tmp = (new_url != NULL) ? new_url : subject;
1796 while ((tmp = strstr(tmp, "https://")) != NULL)
1801 if ((new_url != NULL)
1802 && ( (new_url != subject)
1803 || (0 == strncmpic(subject, "http://", 7))
1804 || (0 == strncmpic(subject, "https://", 8))
1808 * Return new URL if we found a redirect
1809 * or if the subject already was a URL.
1811 * The second case makes sure that we can
1812 * chain get_last_url after another redirection check
1813 * (like rewrite_url) without losing earlier redirects.
1815 new_url = strdup(new_url);
1824 #endif /* def FEATURE_FAST_REDIRECTS */
1827 /*********************************************************************
1829 * Function : redirect_url
1831 * Description : Checks if Privoxy should answer the request with
1832 * a HTTP redirect and generates the redirect if
1836 * 1 : csp = Current client state (buffers, headers, etc...)
1838 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1840 *********************************************************************/
1841 struct http_response *redirect_url(struct client_state *csp)
1843 struct http_response *rsp;
1844 #ifdef FEATURE_FAST_REDIRECTS
1846 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1847 * as compile-time option? The user can easily disable
1848 * it in his action file.
1850 char * redirect_mode;
1851 #endif /* def FEATURE_FAST_REDIRECTS */
1852 char *old_url = NULL;
1853 char *new_url = NULL;
1854 char *redirection_string;
1856 if ((csp->action->flags & ACTION_REDIRECT))
1858 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1861 * If the redirection string begins with 's',
1862 * assume it's a pcrs command, otherwise treat it as
1863 * properly formatted URL and use it for the redirection
1866 * According to RFC 2616 section 14.30 the URL
1867 * has to be absolute and if the user tries:
1868 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1869 * she would get undefined results anyway.
1873 if (*redirection_string == 's')
1875 old_url = csp->http->url;
1876 new_url = rewrite_url(old_url, redirection_string);
1880 log_error(LOG_LEVEL_REDIRECTS,
1881 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1882 redirection_string);
1883 new_url = strdup(redirection_string);
1887 #ifdef FEATURE_FAST_REDIRECTS
1888 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1890 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1893 * If it exists, use the previously rewritten URL as input
1894 * otherwise just use the old path.
1896 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1897 new_url = get_last_url(old_url, redirect_mode);
1902 * Disable redirect checkers, so that they
1903 * will be only run more than once if the user
1904 * also enables them through tags.
1906 * From a performance point of view
1907 * it doesn't matter, but the duplicated
1908 * log messages are annoying.
1910 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1911 #endif /* def FEATURE_FAST_REDIRECTS */
1912 csp->action->flags &= ~ACTION_REDIRECT;
1914 /* Did any redirect action trigger? */
1917 if (0 == strcmpic(new_url, csp->http->url))
1919 log_error(LOG_LEVEL_ERROR,
1920 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1921 csp->http->url, new_url);
1926 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1928 if (NULL == (rsp = alloc_http_response()))
1931 return cgi_error_memory();
1934 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1935 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1938 free_http_response(rsp);
1939 return cgi_error_memory();
1941 rsp->reason = RSP_REASON_REDIRECTED;
1944 return finish_http_response(csp, rsp);
1948 /* Only reached if no redirect is required */
1954 #ifdef FEATURE_IMAGE_BLOCKING
1955 /*********************************************************************
1957 * Function : is_imageurl
1959 * Description : Given a URL, decide whether it is an image or not,
1960 * using either the info from a previous +image action
1961 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1962 * is MSIE and not on a Mac, tell from the browser's accept
1966 * 1 : csp = Current client state (buffers, headers, etc...)
1968 * Returns : True (nonzero) if URL is an image, false (0)
1971 *********************************************************************/
1972 int is_imageurl(const struct client_state *csp)
1974 #ifdef FEATURE_IMAGE_DETECT_MSIE
1977 tmp = get_header_value(csp->headers, "User-Agent:");
1978 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1980 tmp = get_header_value(csp->headers, "Accept:");
1981 if (tmp && strstr(tmp, "image/gif"))
1983 /* Client will accept HTML. If this seems counterintuitive,
1993 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1995 return ((csp->action->flags & ACTION_IMAGE) != 0);
1998 #endif /* def FEATURE_IMAGE_BLOCKING */
2001 #ifdef FEATURE_TRUST
2002 /*********************************************************************
2004 * Function : is_untrusted_url
2006 * Description : Should we "distrust" this URL (and block it)?
2008 * Yes if it matches a line in the trustfile, or if the
2009 * referrer matches a line starting with "+" in the
2014 * 1 : csp = Current client state (buffers, headers, etc...)
2016 * Returns : 0 => trusted, 1 => untrusted
2018 *********************************************************************/
2019 int is_untrusted_url(const struct client_state *csp)
2021 struct file_list *fl;
2022 struct block_spec *b;
2023 struct url_spec **trusted_url;
2024 struct http_request rhttp[1];
2025 const char * referer;
2029 * If we don't have a trustlist, we trust everybody
2031 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
2036 memset(rhttp, '\0', sizeof(*rhttp));
2039 * Do we trust the request URL itself?
2041 for (b = b->next; b ; b = b->next)
2043 if (url_match(b->url, csp->http))
2049 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
2051 /* no referrer was supplied */
2057 * If not, do we maybe trust its referrer?
2059 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
2065 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
2067 if (url_match(*trusted_url, rhttp))
2069 /* if the URL's referrer is from a trusted referrer, then
2070 * add the target spec to the trustfile as an unblocked
2071 * domain and return 0 (which means it's OK).
2076 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
2080 char * new_entry = strdup("~");
2082 string_append(&new_entry, csp->http->hostport);
2084 path = csp->http->path;
2085 if ( (path[0] == '/')
2087 && ((path_end = strchr(path + 2, '/')) != NULL))
2089 /* since this path points into a user's home space
2090 * be sure to include this spec in the trustfile.
2092 long path_len = path_end - path; /* save offset */
2093 path = strdup(path); /* Copy string */
2096 path_end = path + path_len; /* regenerate ptr to new buffer */
2097 *(path_end + 1) = '\0'; /* Truncate path after '/' */
2099 string_join(&new_entry, path);
2103 * Give a reason for generating this entry.
2105 string_append(&new_entry, " # Trusted referrer was: ");
2106 string_append(&new_entry, referer);
2108 if (new_entry != NULL)
2110 if (-1 == fprintf(fp, "%s\n", new_entry))
2112 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
2113 new_entry, csp->config->trustfile);
2119 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
2120 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
2127 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
2128 csp->http->hostport, csp->config->trustfile);
2136 #endif /* def FEATURE_TRUST */
2139 /*********************************************************************
2141 * Function : pcrs_filter_response
2143 * Description : Execute all text substitutions from all applying
2144 * +filter actions on the text buffer that's been
2145 * accumulated in csp->iob->buf.
2148 * 1 : csp = Current client state (buffers, headers, etc...)
2150 * Returns : a pointer to the (newly allocated) modified buffer.
2151 * or NULL if there were no hits or something went wrong
2153 *********************************************************************/
2154 static char *pcrs_filter_response(struct client_state *csp)
2157 size_t size, prev_size;
2163 struct file_list *fl;
2164 struct re_filterfile_spec *b;
2165 struct list_entry *filtername;
2167 int i, found_filters = 0;
2172 if (csp->iob->cur >= csp->iob->eod)
2178 * Need to check the set of re_filterfiles...
2180 for (i = 0; i < MAX_AF_FILES; i++)
2193 if (0 == found_filters)
2195 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2196 "content filtering enabled, but no content filters available.");
2200 size = (size_t)(csp->iob->eod - csp->iob->cur);
2201 old = csp->iob->cur;
2203 for (i = 0; i < MAX_AF_FILES; i++)
2206 if ((NULL == fl) || (NULL == fl->f))
2209 * Either there are no filter files
2210 * left, or this filter file just
2211 * contains no valid filters.
2213 * Continue to be sure we don't miss
2214 * valid filter files that are chained
2215 * after empty or invalid ones.
2220 * For all applying +filter actions, look if a filter by that
2221 * name exists and if yes, execute it's pcrs_joblist on the
2224 for (b = fl->f; b; b = b->next)
2226 if (b->type != FT_CONTENT_FILTER)
2228 /* Skip header filters */
2232 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
2233 filtername ; filtername = filtername->next)
2235 if (strcmp(b->name, filtername->str) == 0)
2237 int current_hits = 0; /* Number of hits caused by this filter */
2238 int job_number = 0; /* Which job we're currently executing */
2239 int job_hits = 0; /* How many hits the current job caused */
2240 pcrs_job *joblist = b->joblist;
2242 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2244 if (NULL == joblist)
2246 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2251 /* Apply all jobs from the joblist */
2252 for (job = joblist; NULL != job; job = job->next)
2255 job_hits = pcrs_execute(job, old, size, &new, &size);
2260 * That went well. Continue filtering
2261 * and use the result of this job as
2262 * input for the next one.
2264 current_hits += job_hits;
2265 if (old != csp->iob->cur)
2274 * This job caused an unexpected error. Inform the user
2275 * and skip the rest of the jobs in this filter. We could
2276 * continue with the next job, but usually the jobs
2277 * depend on each other or are similar enough to
2278 * fail for the same reason.
2280 * At the moment our pcrs expects the error codes of pcre 3.4,
2281 * but newer pcre versions can return additional error codes.
2282 * As a result pcrs_strerror()'s error message might be
2283 * "Unknown error ...", therefore we print the numerical value
2286 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2287 * should we use LOG_LEVEL_RE_FILTER instead?
2289 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2290 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2295 if (b->dynamic) pcrs_free_joblist(joblist);
2297 log_error(LOG_LEVEL_RE_FILTER,
2298 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2299 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2301 hits += current_hits;
2308 * If there were no hits, destroy our copy and let
2309 * chat() use the original in csp->iob
2317 csp->flags |= CSP_FLAG_MODIFIED;
2318 csp->content_length = size;
2326 /*********************************************************************
2328 * Function : gif_deanimate_response
2330 * Description : Deanimate the GIF image that has been accumulated in
2331 * csp->iob->buf, set csp->content_length to the modified
2332 * size and raise the CSP_FLAG_MODIFIED flag.
2335 * 1 : csp = Current client state (buffers, headers, etc...)
2337 * Returns : a pointer to the (newly allocated) modified buffer.
2338 * or NULL in case something went wrong.
2340 *********************************************************************/
2341 static char *gif_deanimate_response(struct client_state *csp)
2343 struct binbuffer *in, *out;
2347 size = (size_t)(csp->iob->eod - csp->iob->cur);
2349 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2350 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2352 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2356 in->buffer = csp->iob->cur;
2359 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2361 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2368 if ((int)size == out->offset)
2370 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2374 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2376 csp->content_length = out->offset;
2377 csp->flags |= CSP_FLAG_MODIFIED;
2387 /*********************************************************************
2389 * Function : get_filter_function
2391 * Description : Decides which content filter function has
2392 * to be applied (if any).
2395 * 1 : csp = Current client state (buffers, headers, etc...)
2397 * Returns : The content filter function to run, or
2398 * NULL if no content filter is active
2400 *********************************************************************/
2401 filter_function_ptr get_filter_function(struct client_state *csp)
2403 filter_function_ptr filter_function = NULL;
2406 * Are we enabling text mode by force?
2408 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2411 * Do we really have to?
2413 if (csp->content_type & CT_TEXT)
2415 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2419 csp->content_type |= CT_TEXT;
2420 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2424 if (!(csp->content_type & CT_DECLARED))
2427 * The server didn't bother to declare a MIME-Type.
2428 * Assume it's text that can be filtered.
2430 * This also regulary happens with 304 responses,
2431 * therefore logging anything here would cause
2434 csp->content_type |= CT_TEXT;
2438 * Choose the applying filter function based on
2439 * the content type and action settings.
2441 if ((csp->content_type & CT_TEXT) &&
2442 (csp->rlist != NULL) &&
2443 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2445 filter_function = pcrs_filter_response;
2447 else if ((csp->content_type & CT_GIF) &&
2448 (csp->action->flags & ACTION_DEANIMATE))
2450 filter_function = gif_deanimate_response;
2453 return filter_function;
2457 /*********************************************************************
2459 * Function : remove_chunked_transfer_coding
2461 * Description : In-situ remove the "chunked" transfer coding as defined
2462 * in rfc2616 from a buffer.
2465 * 1 : buffer = Pointer to the text buffer
2466 * 2 : size = In: Number of bytes to be processed,
2467 * Out: Number of bytes after de-chunking.
2468 * (undefined in case of errors)
2470 * Returns : JB_ERR_OK for success,
2471 * JB_ERR_PARSE otherwise
2473 *********************************************************************/
2474 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2477 unsigned int chunksize = 0;
2478 char *from_p, *to_p;
2481 from_p = to_p = buffer;
2483 if (sscanf(buffer, "%x", &chunksize) != 1)
2485 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2486 return JB_ERR_PARSE;
2489 while (chunksize > 0U)
2491 if (NULL == (from_p = strstr(from_p, "\r\n")))
2493 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2494 return JB_ERR_PARSE;
2497 if ((newsize += chunksize) >= *size)
2499 log_error(LOG_LEVEL_ERROR,
2500 "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding",
2502 return JB_ERR_PARSE;
2506 memmove(to_p, from_p, (size_t) chunksize);
2507 to_p = buffer + newsize;
2508 from_p += chunksize + 2;
2510 if (sscanf(from_p, "%x", &chunksize) != 1)
2512 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2517 /* XXX: Should get its own loglevel. */
2518 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2527 /*********************************************************************
2529 * Function : prepare_for_filtering
2531 * Description : If necessary, de-chunks and decompresses
2532 * the content so it can get filterd.
2535 * 1 : csp = Current client state (buffers, headers, etc...)
2537 * Returns : JB_ERR_OK for success,
2538 * JB_ERR_PARSE otherwise
2540 *********************************************************************/
2541 static jb_err prepare_for_filtering(struct client_state *csp)
2543 jb_err err = JB_ERR_OK;
2546 * If the body has a "chunked" transfer-encoding,
2547 * get rid of it, adjusting size and iob->eod
2549 if (csp->flags & CSP_FLAG_CHUNKED)
2551 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2553 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2554 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2555 if (JB_ERR_OK == err)
2557 csp->iob->eod = csp->iob->cur + size;
2558 csp->flags |= CSP_FLAG_MODIFIED;
2562 return JB_ERR_PARSE;
2568 * If the body has a supported transfer-encoding,
2569 * decompress it, adjusting size and iob->eod.
2571 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2573 if (0 == csp->iob->eod - csp->iob->cur)
2575 /* Nothing left after de-chunking. */
2579 err = decompress_iob(csp);
2581 if (JB_ERR_OK == err)
2583 csp->flags |= CSP_FLAG_MODIFIED;
2584 csp->content_type &= ~CT_TABOO;
2589 * Unset CT_GZIP and CT_DEFLATE to remember not
2590 * to modify the Content-Encoding header later.
2592 csp->content_type &= ~CT_GZIP;
2593 csp->content_type &= ~CT_DEFLATE;
2602 /*********************************************************************
2604 * Function : execute_content_filter
2606 * Description : Executes a given content filter.
2609 * 1 : csp = Current client state (buffers, headers, etc...)
2610 * 2 : content_filter = The filter function to execute
2612 * Returns : Pointer to the modified buffer, or
2613 * NULL if filtering failed or wasn't necessary.
2615 *********************************************************************/
2616 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2618 if (0 == csp->iob->eod - csp->iob->cur)
2621 * No content (probably status code 301, 302 ...),
2622 * no filtering necessary.
2627 if (JB_ERR_OK != prepare_for_filtering(csp))
2630 * failed to de-chunk or decompress.
2635 if (0 == csp->iob->eod - csp->iob->cur)
2638 * Clown alarm: chunked and/or compressed nothing delivered.
2643 return ((*content_filter)(csp));
2647 /*********************************************************************
2649 * Function : get_url_actions
2651 * Description : Gets the actions for this URL.
2654 * 1 : csp = Current client state (buffers, headers, etc...)
2655 * 2 : http = http_request request for blocked URLs
2659 *********************************************************************/
2660 void get_url_actions(struct client_state *csp, struct http_request *http)
2662 struct file_list *fl;
2663 struct url_actions *b;
2666 init_current_action(csp->action);
2668 for (i = 0; i < MAX_AF_FILES; i++)
2670 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2675 apply_url_actions(csp->action, http, b);
2682 /*********************************************************************
2684 * Function : apply_url_actions
2686 * Description : Applies a list of URL actions.
2689 * 1 : action = Destination.
2690 * 2 : http = Current URL
2691 * 3 : b = list of URL actions to apply
2695 *********************************************************************/
2696 void apply_url_actions(struct current_action_spec *action,
2697 struct http_request *http,
2698 struct url_actions *b)
2702 /* Should never happen */
2706 for (b = b->next; NULL != b; b = b->next)
2708 if (url_match(b->url, http))
2710 merge_current_action(action, b->action);
2716 /*********************************************************************
2718 * Function : get_forward_override_settings
2720 * Description : Returns forward settings as specified with the
2721 * forward-override{} action. forward-override accepts
2722 * forward lines similar to the one used in the
2723 * configuration file, but without the URL pattern.
2729 * in the configuration file can be replaced with
2730 * the action section:
2732 * {+forward-override{forward .}}
2736 * 1 : csp = Current client state (buffers, headers, etc...)
2738 * Returns : Pointer to forwarding structure in case of success.
2739 * Invalid syntax is fatal.
2741 *********************************************************************/
2742 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2744 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2745 char forward_settings[BUFFER_SIZE];
2746 char *http_parent = NULL;
2747 /* variable names were chosen for consistency reasons. */
2748 struct forward_spec *fwd = NULL;
2752 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2753 /* Should be enforced by load_one_actions_file() */
2754 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2756 /* Create a copy ssplit can modify */
2757 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2759 if (NULL != csp->fwd)
2762 * XXX: Currently necessary to prevent memory
2763 * leaks when the show-url-info cgi page is visited.
2765 unload_forward_spec(csp->fwd);
2769 * allocate a new forward node, valid only for
2770 * the lifetime of this request. Save its location
2771 * in csp as well, so sweep() can free it later on.
2773 fwd = csp->fwd = zalloc(sizeof(*fwd));
2776 log_error(LOG_LEVEL_FATAL,
2777 "can't allocate memory for forward-override{%s}", forward_override_line);
2778 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2782 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2783 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2785 fwd->type = SOCKS_NONE;
2787 /* Parse the parent HTTP proxy host:port */
2788 http_parent = vec[1];
2791 else if (vec_count == 3)
2793 char *socks_proxy = NULL;
2795 if (!strcasecmp(vec[0], "forward-socks4"))
2797 fwd->type = SOCKS_4;
2798 socks_proxy = vec[1];
2800 else if (!strcasecmp(vec[0], "forward-socks4a"))
2802 fwd->type = SOCKS_4A;
2803 socks_proxy = vec[1];
2805 else if (!strcasecmp(vec[0], "forward-socks5"))
2807 fwd->type = SOCKS_5;
2808 socks_proxy = vec[1];
2811 if (NULL != socks_proxy)
2813 /* Parse the SOCKS proxy host[:port] */
2814 fwd->gateway_port = 1080;
2815 parse_forwarder_address(socks_proxy,
2816 &fwd->gateway_host, &fwd->gateway_port);
2818 http_parent = vec[2];
2822 if (NULL == http_parent)
2824 log_error(LOG_LEVEL_FATAL,
2825 "Invalid forward-override syntax in: %s", forward_override_line);
2826 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2829 /* Parse http forwarding settings */
2830 if (strcmp(http_parent, ".") != 0)
2832 fwd->forward_port = 8000;
2833 parse_forwarder_address(http_parent,
2834 &fwd->forward_host, &fwd->forward_port);
2837 assert (NULL != fwd);
2839 log_error(LOG_LEVEL_CONNECT,
2840 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2846 /*********************************************************************
2848 * Function : forward_url
2850 * Description : Should we forward this to another proxy?
2853 * 1 : csp = Current client state (buffers, headers, etc...)
2854 * 2 : http = http_request request for current URL
2856 * Returns : Pointer to forwarding information.
2858 *********************************************************************/
2859 const struct forward_spec *forward_url(struct client_state *csp,
2860 const struct http_request *http)
2862 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2863 struct forward_spec *fwd = csp->config->forward;
2865 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2867 return get_forward_override_settings(csp);
2877 if (url_match(fwd->url, http))
2888 /*********************************************************************
2890 * Function : direct_response
2892 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2893 * request and if so, return a HTTP 501 to the client.
2895 * FIXME: I have a stupid name and I should handle the
2896 * requests properly. Still, what we do here is rfc-
2897 * compliant, whereas ignoring or forwarding are not.
2900 * 1 : csp = Current client state (buffers, headers, etc...)
2902 * Returns : http_response if , NULL if nonmatch or handler fail
2904 *********************************************************************/
2905 struct http_response *direct_response(struct client_state *csp)
2907 struct http_response *rsp;
2908 struct list_entry *p;
2910 if ((0 == strcmpic(csp->http->gpc, "trace"))
2911 || (0 == strcmpic(csp->http->gpc, "options")))
2913 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2915 if (!strncmpic("Max-Forwards:", p->str, 13))
2917 unsigned int max_forwards;
2920 * If it's a Max-Forwards value of zero,
2921 * we have to intercept the request.
2923 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2926 * FIXME: We could handle at least TRACE here,
2927 * but that would require a verbatim copy of
2928 * the request which we don't have anymore
2930 log_error(LOG_LEVEL_HEADER,
2931 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2934 /* Get mem for response or fail*/
2935 if (NULL == (rsp = alloc_http_response()))
2937 return cgi_error_memory();
2940 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2942 free_http_response(rsp);
2943 return cgi_error_memory();
2947 rsp->reason = RSP_REASON_UNSUPPORTED;
2949 return(finish_http_response(csp, rsp));
2958 /*********************************************************************
2960 * Function : content_filters_enabled
2962 * Description : Checks whether there are any content filters
2963 * enabled for the current request.
2966 * 1 : action = Action spec to check.
2968 * Returns : TRUE for yes, FALSE otherwise
2970 *********************************************************************/
2971 int content_filters_enabled(const struct current_action_spec *action)
2973 return ((action->flags & ACTION_DEANIMATE) ||
2974 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));