1 const char filters_rcs[] = "$Id: filters.c,v 1.117 2009/04/17 11:35:28 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 1.117 2009/04/17 11:35:28 fabiankeil
44 * Fix compiler warning.
46 * Revision 1.116 2009/04/17 11:34:34 fabiankeil
47 * Style cosmetics for the IPv6 code.
49 * Revision 1.115 2009/04/17 11:29:18 fabiankeil
50 * Compile fix for BSD libc.
52 * Revision 1.114 2009/04/17 11:27:49 fabiankeil
53 * Petr Pisar's privoxy-3.0.12-ipv6-3.diff.
55 * Revision 1.113 2009/03/08 14:19:23 fabiankeil
56 * Fix justified (but harmless) compiler warnings
57 * on platforms where sizeof(int) < sizeof(long).
59 * Revision 1.112 2009/03/01 18:28:23 fabiankeil
60 * Help clang understand that we aren't dereferencing
63 * Revision 1.111 2008/12/04 18:13:46 fabiankeil
64 * Fix a cparser warning.
66 * Revision 1.110 2008/11/10 16:40:25 fabiankeil
67 * Fix a gcc44 warning.
69 * Revision 1.109 2008/11/08 15:48:41 fabiankeil
70 * Mention actual values when complaining about
71 * the chunk size exceeding the buffer size.
73 * Revision 1.108 2008/05/21 15:35:08 fabiankeil
74 * - Mark csp as immutable for block_acl().
75 * - Remove an obsolete complaint about filter_popups().
77 * Revision 1.107 2008/05/04 17:52:56 fabiankeil
78 * Adjust parse_http_url() call to new prototype.
80 * Revision 1.106 2008/05/03 16:40:44 fabiankeil
81 * Change content_filters_enabled()'s parameter from
82 * csp->action to action so it can be also used in the
83 * CGI code. Don't bother checking if there are filters
84 * loaded, as that's somewhat besides the point.
86 * Revision 1.105 2008/03/28 15:13:39 fabiankeil
87 * Remove inspect-jpegs action.
89 * Revision 1.104 2008/03/27 18:27:24 fabiankeil
90 * Remove kill-popups action.
92 * Revision 1.103 2008/03/06 16:33:45 fabiankeil
93 * If limit-connect isn't used, don't limit CONNECT requests to port 443.
95 * Revision 1.102 2008/03/01 14:00:44 fabiankeil
96 * Let the block action take the reason for the block
97 * as argument and show it on the "blocked" page.
99 * Revision 1.101 2008/02/23 16:57:12 fabiankeil
100 * Rename url_actions() to get_url_actions() and let it
101 * use the standard parameter ordering.
103 * Revision 1.100 2008/02/23 16:33:43 fabiankeil
104 * Let forward_url() use the standard parameter ordering
105 * and mark its second parameter immutable.
107 * Revision 1.99 2008/02/03 13:57:58 fabiankeil
108 * Add SOCKS5 support for forward-override{}.
110 * Revision 1.98 2008/01/04 17:43:45 fabiankeil
111 * Improve the warning messages that get logged if the action files
112 * "enable" filters but no filters of that type have been loaded.
114 * Revision 1.97 2007/11/30 15:37:03 fabiankeil
115 * Use freez instead of free.
117 * Revision 1.96 2007/10/19 16:53:28 fabiankeil
118 * Add helper function to check if any content filters are enabled.
120 * Revision 1.95 2007/10/17 19:31:20 fabiankeil
121 * Omitting the zero chunk that ends the chunk transfer encoding seems
122 * to be the new black. Log the problem and continue filtering anyway.
124 * Revision 1.94 2007/09/29 13:20:20 fabiankeil
125 * Remove two redundant and one useless log messages.
127 * Revision 1.93 2007/09/29 10:21:16 fabiankeil
128 * - Move get_filter_function() from jcc.c to filters.c
129 * so the filter functions can be static.
130 * - Don't bother filtering body-less responses.
132 * Revision 1.92 2007/09/28 16:38:55 fabiankeil
133 * - Execute content filters through execute_content_filter().
134 * - Add prepare_for_filtering() so filter functions don't have to
135 * care about de-chunking and decompression. As a side effect this enables
136 * decompression for gif_deanimate_response() and jpeg_inspect_response().
137 * - Change remove_chunked_transfer_coding()'s return type to jb_err.
138 * Some clowns feel like chunking empty responses in which case
139 * (size == 0) is valid but previously would be interpreted as error.
141 * Revision 1.91 2007/09/02 15:31:20 fabiankeil
142 * Move match_portlist() from filter.c to urlmatch.c.
143 * It's used for url matching, not for filtering.
145 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
146 * Remove newline at the end of a log_error() message.
148 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
149 * #1763173 from Stefan Huehner: declare some more functions static.
151 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
152 * Add forward-override{} to change the forwarding settings through
153 * action sections. This is mainly interesting to forward different
154 * clients differently (for example based on User-Agent or request
157 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
158 * Make sure filters with dynamic jobs actually use them.
160 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
161 * - Introduce dynamic pcrs jobs that can resolve variables.
162 * - Don't run redirect functions more than once,
163 * unless they are activated more than once.
165 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
166 * - Log the content size after decompression in decompress_iob()
167 * instead of pcrs_filter_response().
169 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
170 * Use dedicated header filter actions instead of abusing "filter".
171 * Replace "filter-client-headers" and "filter-client-headers"
172 * with "server-header-filter" and "client-header-filter".
174 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
175 * New config option: enforce-blocks.
177 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
178 * - Fix port handling in acl_addr() and use a temporary acl spec
179 * copy so error messages don't contain a truncated version.
180 * - Log size of iob before and after decompression.
182 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
183 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
184 * - Hide the "Go there anyway" link for blocked CONNECT
185 * requests where going there anyway doesn't work anyway.
187 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
188 * - Save the reason for generating http_responses.
189 * - Block (+block) with status code 403 instead of 404.
190 * - Use a different kludge to remember a failed decompression.
192 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
193 * Search for Max-Forwards headers case-insensitive,
194 * don't generate the "501 unsupported" message for invalid
195 * Max-Forwards values and don't increase negative ones.
197 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
198 * - Add HEAD support to finish_http_response.
199 * - Add error favicon to internal HTML error messages.
201 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
202 * Mark *csp as immutable for is_untrusted_url()
203 * and is_imageurl(). Closes FR 1237736.
205 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
206 * Integrate a modified version of Wil Mahan's
207 * zlib patch (PR #895531).
209 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
210 * Fixed gcc43 conversion warnings,
211 * changed sprintf calls to snprintf.
213 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
214 * Adjust comment in pcrs_filter_response()
215 * to recent pcrs changes. Hohoho.
217 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
218 * Don't crash if pcre returns an error code
219 * that pcrs didn't expect. Fixes BR 1621173.
221 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
222 * Modified is_untrusted_url to complain in case of
223 * write errors and to give a reason when adding new
224 * entries to the trustfile. Closes FR 1097611.
226 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
227 * Skip empty filter files in pcrs_filter_response,
228 * but don't ignore the ones that come afterwards.
229 * Fixes parts of BR 1619208.
231 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
232 * Added some sanity checks for get_last_url().
233 * Fixed possible segfault caused by my last commit.
235 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
236 * Let get_last_url() catch https URLs as well.
238 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
239 * Make sure get_last_url() behaves like advertised
240 * and fast-redirects{} can be combined with redirect{}.
242 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
243 * Implemented +redirect{s@foo@bar@} to generate
244 * a redirect based on a rewritten version of the
247 * Revision 1.66 2006/09/23 13:26:38 roro
248 * Replace TABs by spaces in source code.
250 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
251 * Fix +redirect{}. Didn't work with -fast-redirects.
253 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
254 * Block requests for untrusted URLs with status
255 * code 403 instead of 200.
257 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
258 * Don't free p which is still in use and will be later
259 * freed by free_map(). Don't claim the referrer is unknown
260 * when the client didn't set one.
262 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
263 * Feature request 595948: Re-Filter logging in single line
265 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
266 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
268 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
269 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
270 * with what was really the latest development (the v_3_0_branch branch)
272 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
273 * Multiple filter file support
275 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
276 * A few memory leaks plugged (mostly on error paths)
278 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
279 * Add the ability to check jpeg images for invalid
280 * lengths of comment blocks. Defensive strategy
281 * against the exploit:
282 * Microsoft Security Bulletin MS04-028
283 * Buffer Overrun in JPEG Processing (GDI+) Could
284 * Allow Code Execution (833987)
285 * Enabled with +inspect-jpegs in actions files.
287 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
288 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
290 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
291 * Fixed bug #839859: "See why" link URL now gets url-encoded.
293 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
296 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
297 * Added basic support for OPTIONS and TRACE HTTP methods:
298 * New function direct_response which handles OPTIONS and
299 * TRACE requests whose Max-Forwards header field is zero.
301 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
302 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
304 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
305 * - Bugfix: Executing a filters without jobs no longer results in
306 * turing off *all* filters.
307 * - Security fix: Malicious web servers can't cause a seg fault
308 * through bogus chunk sizes anymore
310 * Revision 1.58 2002/04/24 02:11:17 oes
311 * Jon's multiple AF patch: url_actions now evaluates rules
314 * Revision 1.57 2002/04/08 20:38:34 swa
317 * Revision 1.56 2002/04/05 15:51:24 oes
318 * - bugfix: error-pages now get correct request protocol
319 * - fix for invalid HTML in trust info
321 * Revision 1.55 2002/04/02 16:13:51 oes
322 * Fix: No "Go there anyway" for SSL
324 * Revision 1.54 2002/04/02 14:55:56 oes
325 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
327 * Revision 1.53 2002/03/26 22:29:54 swa
328 * we have a new homepage!
330 * Revision 1.52 2002/03/24 16:35:57 jongfoster
333 * Revision 1.51 2002/03/24 15:23:33 jongfoster
336 * Revision 1.50 2002/03/24 13:25:43 swa
337 * name change related issues
339 * Revision 1.49 2002/03/16 20:29:14 oes
342 * Revision 1.48 2002/03/13 20:25:34 oes
343 * Better logging for content filters
345 * Revision 1.47 2002/03/13 00:30:52 jongfoster
347 * Added option of always sending redirect for imageblock,
348 * currently disabled with #if 0.
350 * Revision 1.46 2002/03/12 01:42:49 oes
351 * Introduced modular filters
353 * Revision 1.45 2002/03/08 16:47:50 oes
354 * Added choice beween GIF and PNG built-in images
356 * Revision 1.44 2002/03/07 03:49:31 oes
357 * - Fixed compiler warnings etc
358 * - Changed built-in images from GIF to PNG
359 * (with regard to Unisys patent issue)
360 * - Added a 4x4 pattern PNG which is less intrusive
361 * than the logo but also clearly marks the deleted banners
363 * Revision 1.43 2002/01/22 23:51:59 jongfoster
364 * Replacing strsav() with the safer string_append().
366 * Adding missing html_encode() to error message generators. Where encoded
367 * and unencoded versions of a string were provided, removing the unencoded
370 * Revision 1.42 2002/01/17 21:00:32 jongfoster
371 * Moving all our URL and URL pattern parsing code to urlmatch.c.
373 * Using a single, simple url_match(pattern,url) function - rather than
374 * the 3-line match routine which was repeated all over the place.
376 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
378 * Using parse_http_url() to parse URLs without faking a HTTP
379 * request line for parse_http_request().
381 * Revision 1.41 2001/11/13 00:14:07 jongfoster
382 * Fixing stupid bug now I've figured out what || means.
383 * (It always returns 0 or 1, not one of it's paramaters.)
385 * Revision 1.40 2001/10/26 17:37:55 oes
386 * - Re-enabled Netscape 200/404 bug workaround in block_url():
387 * - Removed OS/2 special case
388 * - Made block_url() independant from sed() having been run
389 * - Made trust_url independant from sed() having been run
390 * - Made is_imageurl independant from sed() having been run.
391 * It now checks User-Agent: and Accept: by itself.
394 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
395 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
396 * threads to call select() simultaneously. So, it's time to do a real, live,
397 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
398 * (native). Both versions will work, but using __OS2__ offers multi-threading.
400 * Revision 1.38 2001/10/23 21:32:33 jongfoster
401 * Adding error-checking to selected functions
403 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
404 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
405 * filters.c. Added a FIXME in front of the offending code. I'll gladly
406 * put in a better/more robust fix for all parties if one is presented...
407 * It seems that just returning 200 instead of 404 would pretty much fix
408 * it for everyone, but I don't know all the history of the problem.
410 * Revision 1.36 2001/10/10 16:44:16 oes
411 * Added match_portlist function
413 * Revision 1.35 2001/10/07 15:41:23 oes
414 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
416 * New function remove_chunked_transfer_coding that strips chunked
417 * transfer coding to plain and is called by pcrs_filter_response
418 * and gif_deanimate_response if neccessary
420 * Improved handling of zero-change re_filter runs
422 * pcrs_filter_response and gif_deanimate_response now remove
423 * chunked transfer codeing before processing the body.
425 * Revision 1.34 2001/09/20 15:49:36 steudten
427 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
428 * See cgi.c fill_template().
430 * Revision 1.33 2001/09/16 17:05:14 jongfoster
431 * Removing unused #include showarg.h
433 * Revision 1.32 2001/09/16 13:21:27 jongfoster
434 * Changes to use new list functions.
436 * Revision 1.31 2001/09/16 11:38:02 jongfoster
437 * Splitting fill_template() into 2 functions:
438 * template_load() loads the file
439 * template_fill() performs the PCRS regexps.
440 * This is because the CGI edit interface has a "table row"
441 * template which is used many times in the page - this
442 * change means it's only loaded from disk once.
444 * Revision 1.30 2001/09/16 11:00:10 jongfoster
445 * New function alloc_http_response, for symmetry with free_http_response
447 * Revision 1.29 2001/09/13 23:32:40 jongfoster
448 * Moving image data to cgi.c rather than cgi.h
449 * Fixing a GPF under Win32 (and any other OS that protects global
450 * constants from being written to).
452 * Revision 1.28 2001/09/10 10:18:51 oes
453 * Silenced compiler warnings
455 * Revision 1.27 2001/08/05 16:06:20 jongfoster
456 * Modifiying "struct map" so that there are now separate header and
457 * "map_entry" structures. This means that functions which modify a
458 * map no longer need to return a pointer to the modified map.
459 * Also, it no longer reverses the order of the entries (which may be
460 * important with some advanced template substitutions).
462 * Revision 1.26 2001/07/30 22:08:36 jongfoster
463 * Tidying up #defines:
464 * - All feature #defines are now of the form FEATURE_xxx
465 * - Permanently turned off WIN_GUI_EDIT
466 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
468 * Revision 1.25 2001/07/26 10:09:46 oes
469 * Made browser detection a little less naive
471 * Revision 1.24 2001/07/25 17:22:51 oes
472 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
474 * Revision 1.23 2001/07/23 13:40:12 oes
475 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
477 * Revision 1.22 2001/07/18 12:29:34 oes
478 * - Made gif_deanimate_response respect
479 * csp->action->string[ACTION_STRING_DEANIMATE]
480 * - Logging cosmetics
482 * Revision 1.21 2001/07/13 13:59:53 oes
483 * - Introduced gif_deanimate_response which shares the
484 * generic content modification interface of pcrs_filter_response
485 * and acts as a wrapper to deanimate.c:gif_deanimate()
486 * - Renamed re_process_buffer to pcrs_filter_response
487 * - pcrs_filter_response now returns NULL on failiure
488 * - Removed all #ifdef PCRS
490 * Revision 1.20 2001/07/01 17:01:04 oes
491 * Added comments and missing return statement in is_untrusted_url()
493 * Revision 1.19 2001/06/29 21:45:41 oes
494 * Indentation, CRLF->LF, Tab-> Space
496 * Revision 1.18 2001/06/29 13:27:38 oes
497 * - Cleaned up, renamed and reorderd functions
498 * and improved comments
501 * - Ported to CGI platform. Now delivers
502 * http_response or NULL
503 * - Unified HTML and GIF generation (moved image detection
504 * and GIF generation here from jcc.c:chat())
505 * - Fixed HTTP status to:
506 * - 403 (Forbidden) for the "blocked" HTML message
507 * - 200 (OK) for GIF answers
508 * - 302 (Redirect) for redirect to GIF
511 * - Ported to CGI platform. Now delivers
512 * http_response or NULL
513 * - Separated detection of untrusted URL into
514 * (bool)is_untrusted_url
515 * - Added enforcement of untrusted requests
517 * - Moved redirect_url() from cgi.c to here
518 * and ported it to the CGI platform
520 * - Removed logentry from cancelled commit
522 * Revision 1.17 2001/06/09 10:55:28 jongfoster
523 * Changing BUFSIZ ==> BUFFER_SIZE
525 * Revision 1.16 2001/06/07 23:10:26 jongfoster
526 * Allowing unanchored domain patterns to back off and retry
527 * if they partially match. Optimized right-anchored patterns.
528 * Moving ACL and forward files into config file.
529 * Replacing struct gateway with struct forward_spec
531 * Revision 1.15 2001/06/03 19:12:00 oes
532 * extracted-CGI relevant stuff
534 * Revision 1.14 2001/06/01 10:30:55 oes
535 * Added optional left-anchoring to domaincmp
537 * Revision 1.13 2001/05/31 21:21:30 jongfoster
538 * Permissionsfile / actions file changes:
539 * - Changed "permission" to "action" throughout
540 * - changes to file format to allow string parameters
541 * - Moved helper functions to actions.c
543 * Revision 1.12 2001/05/31 17:35:20 oes
545 * - Enhanced domain part globbing with infix and prefix asterisk
546 * matching and optional unanchored operation
548 * Revision 1.11 2001/05/29 11:53:23 oes
549 * "See why" link added to "blocked" page
551 * Revision 1.10 2001/05/29 09:50:24 jongfoster
552 * Unified blocklist/imagelist/permissionslist.
553 * File format is still under discussion, but the internal changes
556 * Also modified interceptor behaviour:
557 * - We now intercept all URLs beginning with one of the following
558 * prefixes (and *only* these prefixes):
560 * * http://ijbswa.sf.net/config/
561 * * http://ijbswa.sourceforge.net/config/
562 * - New interceptors "home page" - go to http://i.j.b/ to see it.
563 * - Internal changes so that intercepted and fast redirect pages
564 * are not replaced with an image.
565 * - Interceptors now have the option to send a binary page direct
566 * to the client. (i.e. ijb-send-banner uses this)
567 * - Implemented show-url-info interceptor. (Which is why I needed
568 * the above interceptors changes - a typical URL is
569 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
570 * The previous mechanism would not have intercepted that, and
571 * if it had been intercepted then it then it would have replaced
574 * Revision 1.9 2001/05/27 22:17:04 oes
576 * - re_process_buffer no longer writes the modified buffer
577 * to the client, which was very ugly. It now returns the
578 * buffer, which it is then written by chat.
580 * - content_length now adjusts the Content-Length: header
581 * for modified documents rather than crunch()ing it.
582 * (Length info in csp->content_length, which is 0 for
583 * unmodified documents)
585 * - For this to work, sed() is called twice when filtering.
587 * Revision 1.8 2001/05/26 17:13:28 jongfoster
588 * Filled in a function comment.
590 * Revision 1.7 2001/05/26 15:26:15 jongfoster
591 * ACL feature now provides more security by immediately dropping
592 * connections from untrusted hosts.
594 * Revision 1.6 2001/05/26 00:28:36 jongfoster
595 * Automatic reloading of config file.
596 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
597 * Most of the global variables have been moved to a new
598 * struct configuration_spec, accessed through csp->config->globalname
599 * Most of the globals remaining are used by the Win32 GUI.
601 * Revision 1.5 2001/05/25 22:34:30 jongfoster
604 * Revision 1.4 2001/05/22 18:46:04 oes
606 * - Enabled filtering banners by size rather than URL
607 * by adding patterns that replace all standard banner
608 * sizes with the "Junkbuster" gif to the re_filterfile
610 * - Enabled filtering WebBugs by providing a pattern
611 * which kills all 1x1 images
613 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
614 * which is selected by the (nonstandard and therefore
615 * capital) letter 'U' in the option string.
616 * It causes the quantifiers to be ungreedy by default.
617 * Appending a ? turns back to greedy (!).
619 * - Added a new interceptor ijb-send-banner, which
620 * sends back the "Junkbuster" gif. Without imagelist or
621 * MSIE detection support, or if tinygif = 1, or the
622 * URL isn't recognized as an imageurl, a lame HTML
623 * explanation is sent instead.
625 * - Added new feature, which permits blocking remote
626 * script redirects and firing back a local redirect
628 * The feature is conditionally compiled, i.e. it
629 * can be disabled with --disable-fast-redirects,
630 * plus it must be activated by a "fast-redirects"
631 * line in the config file, has its own log level
632 * and of course wants to be displayed by show-proxy-args
633 * Note: Boy, all the #ifdefs in 1001 locations and
634 * all the fumbling with configure.in and acconfig.h
635 * were *way* more work than the feature itself :-(
637 * - Because a generic redirect template was needed for
638 * this, tinygif = 3 now uses the same.
640 * - Moved GIFs, and other static HTTP response templates
645 * - Removed some >400 CRs again (Jon, you really worked
648 * Revision 1.3 2001/05/20 16:44:47 jongfoster
649 * Removing last hardcoded Junkbusters.com URLs.
651 * Revision 1.2 2001/05/20 01:21:20 jongfoster
652 * Version 2.9.4 checkin.
653 * - Merged popupfile and cookiefile, and added control over PCRS
654 * filtering, in new "permissionsfile".
655 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
656 * file error you now get a message box (in the Win32 GUI) rather
657 * than the program exiting with no explanation.
658 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
660 * - Removed tabs from "config"
661 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
662 * - Bumped up version number.
664 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
665 * Initial import of version 2.9.3 source tree
668 *********************************************************************/
674 #include <sys/types.h>
680 #ifdef HAVE_GETADDRINFO
682 #include <sys/socket.h>
683 #endif /* def HAVE_GETADDRINFO */
688 #endif /* ndef __OS2__ */
689 #include <netinet/in.h>
691 #include <winsock2.h>
692 #endif /* ndef _WIN32 */
696 #endif /* def __OS2__ */
704 #include "jbsockets.h"
705 #include "miscutil.h"
709 #include "deanimate.h"
710 #include "urlmatch.h"
717 const char filters_h_rcs[] = FILTERS_H_VERSION;
719 /* Fix a problem with Solaris. There should be no effect on other
721 * Solaris's isspace() is a macro which uses it's argument directly
722 * as an array index. Therefore we need to make sure that high-bit
723 * characters generate +ve values, and ideally we also want to make
724 * the argument match the declared parameter type of "int".
726 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
728 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
729 static jb_err prepare_for_filtering(struct client_state *csp);
732 #ifdef HAVE_GETADDRINFO
733 /*********************************************************************
735 * Function : sockaddr_storage_to_ip
737 * Description : Access internal structure of sockaddr_storage
740 * 1 : addr = socket address
741 * 2 : ip = IP address as array of octets in network order
742 * (it points into addr)
743 * 3 : len = length of IP address in octets
744 * 4 : port = port number in network order;
746 * Returns : 0 = no errror; -1 otherwise.
748 *********************************************************************/
749 int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, uint8_t **ip,
750 unsigned int *len, in_port_t **port)
757 switch (addr->ss_family)
767 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
771 *port = &((struct sockaddr_in *)addr)->sin_port;
782 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
786 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
791 /* Unsupported address family */
799 /*********************************************************************
801 * Function : match_sockaddr
803 * Description : Check whether address matches network (IP address and port)
806 * 1 : network = socket address of subnework
807 * 2 : netmask = network mask as socket address
808 * 3 : address = checked socket address against given network
810 * Returns : 0 = doesn't match; 1 = does match
812 *********************************************************************/
813 int match_sockaddr(const struct sockaddr_storage *network,
814 const struct sockaddr_storage *netmask,
815 const struct sockaddr_storage *address)
817 uint8_t *network_addr, *netmask_addr, *address_addr;
818 unsigned int addr_len;
819 in_port_t *network_port, *netmask_port, *address_port;
822 if (network->ss_family != netmask->ss_family)
824 /* This should never happen */
825 log_error(LOG_LEVEL_ERROR,
826 "Internal error at %s:%llu: network and netmask differ in family",
831 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
832 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
833 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
835 /* Check for family */
836 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
837 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
839 /* Map AF_INET6 V4MAPPED address into AF_INET */
843 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
844 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
846 /* Map AF_INET6 V4MAPPED network into AF_INET */
851 else if (network->ss_family != address->ss_family)
856 /* XXX: Port check is signaled in netmask */
857 if (*netmask_port && *network_port != *address_port)
862 /* TODO: Optimize by checking by words insted of octets */
863 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
865 if ((network_addr[i] & netmask_addr[i]) !=
866 (address_addr[i] & netmask_addr[i]))
874 #endif /* def HAVE_GETADDRINFO */
877 /*********************************************************************
879 * Function : block_acl
881 * Description : Block this request?
882 * Decide yes or no based on ACL file.
885 * 1 : dst = The proxy or gateway address this is going to.
886 * Or NULL to check all possible targets.
887 * 2 : csp = Current client state (buffers, headers, etc...)
888 * Also includes the client IP address.
890 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
892 *********************************************************************/
893 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
895 struct access_control_list *acl = csp->config->acl;
897 /* if not using an access control list, then permit the connection */
903 /* search the list */
907 #ifdef HAVE_GETADDRINFO
908 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
910 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
916 /* Just want to check if they have any access */
917 if (acl->action == ACL_PERMIT)
923 #ifdef HAVE_GETADDRINFO
925 * XXX: An undefined acl->dst is full of zeros and should be
926 * considered a wildcard address. sockaddr_storage_to_ip()
927 * fails on such destinations because of unknown sa_familly
928 * (glibc only?). However this test is not portable.
930 * So, we signal the acl->dst is wildcard in wildcard_dst.
933 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
935 ((dst->addr & acl->dst->mask) == acl->dst->addr)
936 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
940 if (acl->action == ACL_PERMIT)
958 /*********************************************************************
960 * Function : acl_addr
962 * Description : Called from `load_config' to parse an ACL address.
965 * 1 : aspec = String specifying ACL address.
966 * 2 : aca = struct access_control_addr to fill in.
968 * Returns : 0 => Ok, everything else is an error.
970 *********************************************************************/
971 int acl_addr(const char *aspec, struct access_control_addr *aca)
974 #ifdef HAVE_GETADDRINFO
975 struct addrinfo hints, *result;
977 in_port_t *mask_port;
978 unsigned int addr_len;
981 #endif /* def HAVE_GETADDRINFO */
983 char *acl_spec = NULL;
985 #ifdef HAVE_GETADDRINFO
986 /* XXX: Depend on ai_family */
994 * Use a temporary acl spec copy so we can log
995 * the unmodified original in case of parse errors.
997 acl_spec = strdup(aspec);
998 if (acl_spec == NULL)
1000 /* XXX: This will be logged as parse error. */
1004 if ((p = strchr(acl_spec, '/')) != NULL)
1007 if (ijb_isdigit(*p) == 0)
1012 masklength = atoi(p);
1015 if ((masklength < 0) ||
1016 #ifdef HAVE_GETADDRINFO
1027 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
1030 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
1039 p = strchr(acl_spec, ':');
1042 #ifdef HAVE_GETADDRINFO
1043 memset(&hints, 0, sizeof(struct addrinfo));
1044 hints.ai_family = AF_UNSPEC;
1045 hints.ai_socktype = SOCK_STREAM;
1047 i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result);
1052 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
1053 acl_spec, p, gai_strerror(i));
1057 /* TODO: Allow multihomed hostnames */
1058 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
1059 freeaddrinfo(result);
1066 port = strtol(p, &endptr, 10);
1068 if (port <= 0 || port > 65535 || *endptr != '\0')
1075 aca->port = (unsigned long)port;
1077 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
1080 if (aca->addr == INADDR_NONE)
1082 /* XXX: This will be logged as parse error. */
1085 #endif /* def HAVE_GETADDRINFO */
1087 /* build the netmask */
1088 #ifdef HAVE_GETADDRINFO
1089 /* Clip masklength according to current family. */
1090 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
1095 aca->mask.ss_family = aca->addr.ss_family;
1096 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
1103 /* ACL contains a port number, check ports in the future. */
1108 * XXX: This could be optimized to operate on whole words instead
1109 * of octets (128-bit CPU could do it in one iteration).
1112 * Octets after prefix can be ommitted because of
1113 * previous initialization to zeros.
1115 for (i = 0; (i < addr_len) && masklength; i++)
1117 if (masklength >= 8)
1119 mask_data[i] = 0xFF;
1125 * XXX: This assumes MSB of octet is on the left side.
1126 * This should be true for all architectures or solved
1127 * by the link layer.
1129 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
1136 for (i=1; i <= masklength ; i++)
1138 aca->mask |= (1U << (32 - i));
1141 /* now mask off the host portion of the ip address
1142 * (i.e. save on the network portion of the address).
1144 aca->addr = aca->addr & aca->mask;
1145 #endif /* def HAVE_GETADDRINFO */
1150 #endif /* def FEATURE_ACL */
1153 /*********************************************************************
1155 * Function : connect_port_is_forbidden
1157 * Description : Check to see if CONNECT requests to the destination
1158 * port of this request are forbidden. The check is
1159 * independend of the actual request method.
1162 * 1 : csp = Current client state (buffers, headers, etc...)
1164 * Returns : True if yes, false otherwise.
1166 *********************************************************************/
1167 int connect_port_is_forbidden(const struct client_state *csp)
1169 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
1170 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
1175 /*********************************************************************
1177 * Function : block_url
1179 * Description : Called from `chat'. Check to see if we need to block this.
1182 * 1 : csp = Current client state (buffers, headers, etc...)
1184 * Returns : NULL => unblocked, else HTTP block response
1186 *********************************************************************/
1187 struct http_response *block_url(struct client_state *csp)
1189 struct http_response *rsp;
1190 const char *new_content_type = NULL;
1193 * If it's not blocked, don't block it ;-)
1195 if ((csp->action->flags & ACTION_BLOCK) == 0)
1199 if (csp->action->flags & ACTION_REDIRECT)
1201 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
1204 * Else, prepare a response
1206 if (NULL == (rsp = alloc_http_response()))
1208 return cgi_error_memory();
1212 * If it's an image-url, send back an image or redirect
1213 * as specified by the relevant +image action
1215 #ifdef FEATURE_IMAGE_BLOCKING
1216 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
1217 && is_imageurl(csp))
1220 /* determine HOW images should be blocked */
1221 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
1223 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1225 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
1227 #if 1 /* Two alternative strategies, use this one for now: */
1229 /* and handle accordingly: */
1230 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1232 rsp->status = strdup("403 Request blocked by Privoxy");
1233 if (rsp->status == NULL)
1235 free_http_response(rsp);
1236 return cgi_error_memory();
1238 rsp->body = bindup(image_pattern_data, image_pattern_length);
1239 if (rsp->body == NULL)
1241 free_http_response(rsp);
1242 return cgi_error_memory();
1244 rsp->content_length = image_pattern_length;
1246 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1248 free_http_response(rsp);
1249 return cgi_error_memory();
1253 else if (0 == strcmpic(p, "blank"))
1255 rsp->status = strdup("403 Request blocked by Privoxy");
1256 if (rsp->status == NULL)
1258 free_http_response(rsp);
1259 return cgi_error_memory();
1261 rsp->body = bindup(image_blank_data, image_blank_length);
1262 if (rsp->body == NULL)
1264 free_http_response(rsp);
1265 return cgi_error_memory();
1267 rsp->content_length = image_blank_length;
1269 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1271 free_http_response(rsp);
1272 return cgi_error_memory();
1278 rsp->status = strdup("302 Local Redirect from Privoxy");
1279 if (rsp->status == NULL)
1281 free_http_response(rsp);
1282 return cgi_error_memory();
1285 if (enlist_unique_header(rsp->headers, "Location", p))
1287 free_http_response(rsp);
1288 return cgi_error_memory();
1292 #else /* Following code is disabled for now */
1294 /* and handle accordingly: */
1295 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1297 p = CGI_PREFIX "send-banner?type=pattern";
1299 else if (0 == strcmpic(p, "blank"))
1301 p = CGI_PREFIX "send-banner?type=blank";
1303 rsp->status = strdup("302 Local Redirect from Privoxy");
1304 if (rsp->status == NULL)
1306 free_http_response(rsp);
1307 return cgi_error_memory();
1310 if (enlist_unique_header(rsp->headers, "Location", p))
1312 free_http_response(rsp);
1313 return cgi_error_memory();
1315 #endif /* Preceeding code is disabled for now */
1317 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1320 * Send empty document.
1322 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1325 rsp->body = strdup(" ");
1326 rsp->content_length = 1;
1328 rsp->status = strdup("403 Request blocked by Privoxy");
1329 if (rsp->status == NULL)
1331 free_http_response(rsp);
1332 return cgi_error_memory();
1334 if (new_content_type != 0)
1336 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1337 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1339 free_http_response(rsp);
1340 return cgi_error_memory();
1345 #endif /* def FEATURE_IMAGE_BLOCKING */
1348 * Else, generate an HTML "blocked" message:
1352 struct map * exports;
1356 * Workaround for stupid Netscape bug which prevents
1357 * pages from being displayed if loading a referenced
1358 * JavaScript or style sheet fails. So make it appear
1359 * as if it succeeded.
1361 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1362 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1363 && !strstr(p, "Gecko") /* save Mozilla, */
1364 && !strstr(p, "compatible") /* MSIE */
1365 && !strstr(p, "Opera")) /* and Opera. */
1367 rsp->status = strdup("200 Request for blocked URL");
1371 rsp->status = strdup("403 Request for blocked URL");
1374 if (rsp->status == NULL)
1376 free_http_response(rsp);
1377 return cgi_error_memory();
1380 exports = default_exports(csp, NULL);
1381 if (exports == NULL)
1383 free_http_response(rsp);
1384 return cgi_error_memory();
1387 #ifdef FEATURE_FORCE_LOAD
1388 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1390 * Export the force conditional block killer if
1392 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1393 * - Privoxy is configured to enforce blocks, or
1394 * - it's a CONNECT request and enforcing wouldn't work anyway.
1396 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1397 || (0 == strcmpic(csp->http->gpc, "connect")))
1398 #endif /* ndef FEATURE_FORCE_LOAD */
1400 err = map_block_killer(exports, "force-support");
1403 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1404 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1405 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1406 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1409 const char *block_reason;
1410 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
1412 block_reason = csp->action->string[ACTION_STRING_BLOCK];
1416 assert(connect_port_is_forbidden(csp));
1417 block_reason = "Forbidden CONNECT port.";
1419 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
1424 free_http_response(rsp);
1425 return cgi_error_memory();
1428 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1431 free_http_response(rsp);
1432 return cgi_error_memory();
1435 rsp->reason = RSP_REASON_BLOCKED;
1437 return finish_http_response(csp, rsp);
1442 #ifdef FEATURE_TRUST
1443 /*********************************************************************
1445 * Function : trust_url FIXME: I should be called distrust_url
1447 * Description : Calls is_untrusted_url to determine if the URL is trusted
1448 * and if not, returns a HTTP 403 response with a reject message.
1451 * 1 : csp = Current client state (buffers, headers, etc...)
1453 * Returns : NULL => trusted, else http_response.
1455 *********************************************************************/
1456 struct http_response *trust_url(struct client_state *csp)
1458 struct http_response *rsp;
1459 struct map * exports;
1460 char buf[BUFFER_SIZE];
1462 struct url_spec **tl;
1467 * Don't bother to work on trusted URLs
1469 if (!is_untrusted_url(csp))
1475 * Else, prepare a response:
1477 if (NULL == (rsp = alloc_http_response()))
1479 return cgi_error_memory();
1482 rsp->status = strdup("403 Request blocked by Privoxy");
1483 exports = default_exports(csp, NULL);
1484 if (exports == NULL || rsp->status == NULL)
1486 free_http_response(rsp);
1487 return cgi_error_memory();
1491 * Export the protocol, host, port, and referrer information
1493 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1494 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1495 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1497 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1499 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1503 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1509 free_http_response(rsp);
1510 return cgi_error_memory();
1514 * Export the trust list
1517 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1519 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1520 string_append(&p, buf);
1522 err = map(exports, "trusted-referrers", 1, p, 0);
1527 free_http_response(rsp);
1528 return cgi_error_memory();
1532 * Export the trust info, if available
1534 if (csp->config->trust_info->first)
1536 struct list_entry *l;
1539 for (l = csp->config->trust_info->first; l ; l = l->next)
1541 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1542 string_append(&p, buf);
1544 err = map(exports, "trust-info", 1, p, 0);
1548 err = map_block_killer(exports, "have-trust-info");
1554 free_http_response(rsp);
1555 return cgi_error_memory();
1559 * Export the force conditional block killer if
1561 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1562 * - Privoxy is configured to enforce blocks, or
1563 * - it's a CONNECT request and enforcing wouldn't work anyway.
1565 #ifdef FEATURE_FORCE_LOAD
1566 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1567 || (0 == strcmpic(csp->http->gpc, "connect")))
1569 err = map_block_killer(exports, "force-support");
1573 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1575 #else /* ifndef FEATURE_FORCE_LOAD */
1576 err = map_block_killer(exports, "force-support");
1577 #endif /* ndef FEATURE_FORCE_LOAD */
1582 free_http_response(rsp);
1583 return cgi_error_memory();
1587 * Build the response
1589 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1592 free_http_response(rsp);
1593 return cgi_error_memory();
1595 rsp->reason = RSP_REASON_UNTRUSTED;
1597 return finish_http_response(csp, rsp);
1599 #endif /* def FEATURE_TRUST */
1602 /*********************************************************************
1604 * Function : compile_dynamic_pcrs_job_list
1606 * Description : Compiles a dynamic pcrs job list (one with variables
1607 * resolved at request time)
1610 * 1 : csp = Current client state (buffers, headers, etc...)
1611 * 2 : b = The filter list to compile
1613 * Returns : NULL in case of errors, otherwise the
1616 *********************************************************************/
1617 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1619 struct list_entry *pattern;
1620 pcrs_job *job_list = NULL;
1621 pcrs_job *dummy = NULL;
1622 pcrs_job *lastjob = NULL;
1625 const struct pcrs_variable variables[] =
1627 {"url", csp->http->url, 1},
1628 {"path", csp->http->path, 1},
1629 {"host", csp->http->host, 1},
1630 {"origin", csp->ip_addr_str, 1},
1634 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1636 assert(pattern->str != NULL);
1638 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1642 log_error(LOG_LEVEL_ERROR,
1643 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1644 pattern->str, b->name, pcrs_strerror(error));
1649 if (error == PCRS_WARN_TRUNCATION)
1651 log_error(LOG_LEVEL_ERROR,
1652 "At least one of the variables in \'%s\' had to "
1653 "be truncated before compilation", pattern->str);
1655 if (job_list == NULL)
1661 lastjob->next = dummy;
1671 /*********************************************************************
1673 * Function : rewrite_url
1675 * Description : Rewrites a URL with a single pcrs command
1676 * and returns the result if it differs from the
1677 * original and isn't obviously invalid.
1680 * 1 : old_url = URL to rewrite.
1681 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1684 * Returns : NULL if the pcrs_command didn't change the url, or
1685 * the result of the modification.
1687 *********************************************************************/
1688 char *rewrite_url(char *old_url, const char *pcrs_command)
1690 char *new_url = NULL;
1694 assert(pcrs_command);
1696 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1700 log_error(LOG_LEVEL_REDIRECTS,
1701 "pcrs command \"%s\" didn't change \"%s\".",
1702 pcrs_command, old_url);
1707 log_error(LOG_LEVEL_REDIRECTS,
1708 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1709 pcrs_command, old_url, pcrs_strerror(hits));
1712 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1714 log_error(LOG_LEVEL_ERROR,
1715 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1716 "but the result doesn't look like a valid URL and will be ignored.",
1717 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1722 log_error(LOG_LEVEL_REDIRECTS,
1723 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1724 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1732 #ifdef FEATURE_FAST_REDIRECTS
1733 /*********************************************************************
1735 * Function : get_last_url
1737 * Description : Search for the last URL inside a string.
1738 * If the string already is a URL, it will
1739 * be the first URL found.
1742 * 1 : subject = the string to check
1743 * 2 : redirect_mode = +fast-redirect{} mode
1745 * Returns : NULL if no URL was found, or
1746 * the last URL found.
1748 *********************************************************************/
1749 char *get_last_url(char *subject, const char *redirect_mode)
1751 char *new_url = NULL;
1755 assert(redirect_mode);
1757 subject = strdup(subject);
1758 if (subject == NULL)
1760 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1764 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1766 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1767 new_url = url_decode(subject);
1768 if (new_url != NULL)
1775 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1779 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1782 * Find the last URL encoded in the request
1785 while ((tmp = strstr(tmp, "http://")) != NULL)
1789 tmp = (new_url != NULL) ? new_url : subject;
1790 while ((tmp = strstr(tmp, "https://")) != NULL)
1795 if ((new_url != NULL)
1796 && ( (new_url != subject)
1797 || (0 == strncmpic(subject, "http://", 7))
1798 || (0 == strncmpic(subject, "https://", 8))
1802 * Return new URL if we found a redirect
1803 * or if the subject already was a URL.
1805 * The second case makes sure that we can
1806 * chain get_last_url after another redirection check
1807 * (like rewrite_url) without losing earlier redirects.
1809 new_url = strdup(new_url);
1818 #endif /* def FEATURE_FAST_REDIRECTS */
1821 /*********************************************************************
1823 * Function : redirect_url
1825 * Description : Checks if Privoxy should answer the request with
1826 * a HTTP redirect and generates the redirect if
1830 * 1 : csp = Current client state (buffers, headers, etc...)
1832 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1834 *********************************************************************/
1835 struct http_response *redirect_url(struct client_state *csp)
1837 struct http_response *rsp;
1838 #ifdef FEATURE_FAST_REDIRECTS
1840 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1841 * as compile-time option? The user can easily disable
1842 * it in his action file.
1844 char * redirect_mode;
1845 #endif /* def FEATURE_FAST_REDIRECTS */
1846 char *old_url = NULL;
1847 char *new_url = NULL;
1848 char *redirection_string;
1850 if ((csp->action->flags & ACTION_REDIRECT))
1852 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1855 * If the redirection string begins with 's',
1856 * assume it's a pcrs command, otherwise treat it as
1857 * properly formatted URL and use it for the redirection
1860 * According to RFC 2616 section 14.30 the URL
1861 * has to be absolute and if the user tries:
1862 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1863 * she would get undefined results anyway.
1867 if (*redirection_string == 's')
1869 old_url = csp->http->url;
1870 new_url = rewrite_url(old_url, redirection_string);
1874 log_error(LOG_LEVEL_REDIRECTS,
1875 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1876 redirection_string);
1877 new_url = strdup(redirection_string);
1881 #ifdef FEATURE_FAST_REDIRECTS
1882 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1884 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1887 * If it exists, use the previously rewritten URL as input
1888 * otherwise just use the old path.
1890 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1891 new_url = get_last_url(old_url, redirect_mode);
1896 * Disable redirect checkers, so that they
1897 * will be only run more than once if the user
1898 * also enables them through tags.
1900 * From a performance point of view
1901 * it doesn't matter, but the duplicated
1902 * log messages are annoying.
1904 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1905 #endif /* def FEATURE_FAST_REDIRECTS */
1906 csp->action->flags &= ~ACTION_REDIRECT;
1908 /* Did any redirect action trigger? */
1911 if (0 == strcmpic(new_url, csp->http->url))
1913 log_error(LOG_LEVEL_ERROR,
1914 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1915 csp->http->url, new_url);
1920 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1922 if (NULL == (rsp = alloc_http_response()))
1925 return cgi_error_memory();
1928 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1929 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1932 free_http_response(rsp);
1933 return cgi_error_memory();
1935 rsp->reason = RSP_REASON_REDIRECTED;
1938 return finish_http_response(csp, rsp);
1942 /* Only reached if no redirect is required */
1948 #ifdef FEATURE_IMAGE_BLOCKING
1949 /*********************************************************************
1951 * Function : is_imageurl
1953 * Description : Given a URL, decide whether it is an image or not,
1954 * using either the info from a previous +image action
1955 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1956 * is MSIE and not on a Mac, tell from the browser's accept
1960 * 1 : csp = Current client state (buffers, headers, etc...)
1962 * Returns : True (nonzero) if URL is an image, false (0)
1965 *********************************************************************/
1966 int is_imageurl(const struct client_state *csp)
1968 #ifdef FEATURE_IMAGE_DETECT_MSIE
1971 tmp = get_header_value(csp->headers, "User-Agent:");
1972 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1974 tmp = get_header_value(csp->headers, "Accept:");
1975 if (tmp && strstr(tmp, "image/gif"))
1977 /* Client will accept HTML. If this seems counterintuitive,
1987 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1989 return ((csp->action->flags & ACTION_IMAGE) != 0);
1992 #endif /* def FEATURE_IMAGE_BLOCKING */
1995 #ifdef FEATURE_TRUST
1996 /*********************************************************************
1998 * Function : is_untrusted_url
2000 * Description : Should we "distrust" this URL (and block it)?
2002 * Yes if it matches a line in the trustfile, or if the
2003 * referrer matches a line starting with "+" in the
2008 * 1 : csp = Current client state (buffers, headers, etc...)
2010 * Returns : 0 => trusted, 1 => untrusted
2012 *********************************************************************/
2013 int is_untrusted_url(const struct client_state *csp)
2015 struct file_list *fl;
2016 struct block_spec *b;
2017 struct url_spec **trusted_url;
2018 struct http_request rhttp[1];
2019 const char * referer;
2023 * If we don't have a trustlist, we trust everybody
2025 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
2030 memset(rhttp, '\0', sizeof(*rhttp));
2033 * Do we trust the request URL itself?
2035 for (b = b->next; b ; b = b->next)
2037 if (url_match(b->url, csp->http))
2043 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
2045 /* no referrer was supplied */
2051 * If not, do we maybe trust its referrer?
2053 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
2059 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
2061 if (url_match(*trusted_url, rhttp))
2063 /* if the URL's referrer is from a trusted referrer, then
2064 * add the target spec to the trustfile as an unblocked
2065 * domain and return 0 (which means it's OK).
2070 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
2074 char * new_entry = strdup("~");
2076 string_append(&new_entry, csp->http->hostport);
2078 path = csp->http->path;
2079 if ( (path[0] == '/')
2081 && ((path_end = strchr(path + 2, '/')) != NULL))
2083 /* since this path points into a user's home space
2084 * be sure to include this spec in the trustfile.
2086 long path_len = path_end - path; /* save offset */
2087 path = strdup(path); /* Copy string */
2090 path_end = path + path_len; /* regenerate ptr to new buffer */
2091 *(path_end + 1) = '\0'; /* Truncate path after '/' */
2093 string_join(&new_entry, path);
2097 * Give a reason for generating this entry.
2099 string_append(&new_entry, " # Trusted referrer was: ");
2100 string_append(&new_entry, referer);
2102 if (new_entry != NULL)
2104 if (-1 == fprintf(fp, "%s\n", new_entry))
2106 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
2107 new_entry, csp->config->trustfile);
2113 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
2114 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
2121 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
2122 csp->http->hostport, csp->config->trustfile);
2130 #endif /* def FEATURE_TRUST */
2133 /*********************************************************************
2135 * Function : pcrs_filter_response
2137 * Description : Execute all text substitutions from all applying
2138 * +filter actions on the text buffer that's been
2139 * accumulated in csp->iob->buf.
2142 * 1 : csp = Current client state (buffers, headers, etc...)
2144 * Returns : a pointer to the (newly allocated) modified buffer.
2145 * or NULL if there were no hits or something went wrong
2147 *********************************************************************/
2148 static char *pcrs_filter_response(struct client_state *csp)
2151 size_t size, prev_size;
2157 struct file_list *fl;
2158 struct re_filterfile_spec *b;
2159 struct list_entry *filtername;
2161 int i, found_filters = 0;
2166 if (csp->iob->cur >= csp->iob->eod)
2172 * Need to check the set of re_filterfiles...
2174 for (i = 0; i < MAX_AF_FILES; i++)
2187 if (0 == found_filters)
2189 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2190 "content filtering enabled, but no content filters available.");
2194 size = (size_t)(csp->iob->eod - csp->iob->cur);
2195 old = csp->iob->cur;
2197 for (i = 0; i < MAX_AF_FILES; i++)
2200 if ((NULL == fl) || (NULL == fl->f))
2203 * Either there are no filter files
2204 * left, or this filter file just
2205 * contains no valid filters.
2207 * Continue to be sure we don't miss
2208 * valid filter files that are chained
2209 * after empty or invalid ones.
2214 * For all applying +filter actions, look if a filter by that
2215 * name exists and if yes, execute it's pcrs_joblist on the
2218 for (b = fl->f; b; b = b->next)
2220 if (b->type != FT_CONTENT_FILTER)
2222 /* Skip header filters */
2226 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
2227 filtername ; filtername = filtername->next)
2229 if (strcmp(b->name, filtername->str) == 0)
2231 int current_hits = 0; /* Number of hits caused by this filter */
2232 int job_number = 0; /* Which job we're currently executing */
2233 int job_hits = 0; /* How many hits the current job caused */
2234 pcrs_job *joblist = b->joblist;
2236 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2238 if (NULL == joblist)
2240 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2245 /* Apply all jobs from the joblist */
2246 for (job = joblist; NULL != job; job = job->next)
2249 job_hits = pcrs_execute(job, old, size, &new, &size);
2254 * That went well. Continue filtering
2255 * and use the result of this job as
2256 * input for the next one.
2258 current_hits += job_hits;
2259 if (old != csp->iob->cur)
2268 * This job caused an unexpected error. Inform the user
2269 * and skip the rest of the jobs in this filter. We could
2270 * continue with the next job, but usually the jobs
2271 * depend on each other or are similar enough to
2272 * fail for the same reason.
2274 * At the moment our pcrs expects the error codes of pcre 3.4,
2275 * but newer pcre versions can return additional error codes.
2276 * As a result pcrs_strerror()'s error message might be
2277 * "Unknown error ...", therefore we print the numerical value
2280 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2281 * should we use LOG_LEVEL_RE_FILTER instead?
2283 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2284 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2289 if (b->dynamic) pcrs_free_joblist(joblist);
2291 log_error(LOG_LEVEL_RE_FILTER,
2292 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2293 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2295 hits += current_hits;
2302 * If there were no hits, destroy our copy and let
2303 * chat() use the original in csp->iob
2311 csp->flags |= CSP_FLAG_MODIFIED;
2312 csp->content_length = size;
2320 /*********************************************************************
2322 * Function : gif_deanimate_response
2324 * Description : Deanimate the GIF image that has been accumulated in
2325 * csp->iob->buf, set csp->content_length to the modified
2326 * size and raise the CSP_FLAG_MODIFIED flag.
2329 * 1 : csp = Current client state (buffers, headers, etc...)
2331 * Returns : a pointer to the (newly allocated) modified buffer.
2332 * or NULL in case something went wrong.
2334 *********************************************************************/
2335 static char *gif_deanimate_response(struct client_state *csp)
2337 struct binbuffer *in, *out;
2341 size = (size_t)(csp->iob->eod - csp->iob->cur);
2343 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2344 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2346 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2350 in->buffer = csp->iob->cur;
2353 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2355 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2362 if ((int)size == out->offset)
2364 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2368 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2370 csp->content_length = out->offset;
2371 csp->flags |= CSP_FLAG_MODIFIED;
2381 /*********************************************************************
2383 * Function : get_filter_function
2385 * Description : Decides which content filter function has
2386 * to be applied (if any).
2389 * 1 : csp = Current client state (buffers, headers, etc...)
2391 * Returns : The content filter function to run, or
2392 * NULL if no content filter is active
2394 *********************************************************************/
2395 filter_function_ptr get_filter_function(struct client_state *csp)
2397 filter_function_ptr filter_function = NULL;
2400 * Are we enabling text mode by force?
2402 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2405 * Do we really have to?
2407 if (csp->content_type & CT_TEXT)
2409 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2413 csp->content_type |= CT_TEXT;
2414 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2418 if (!(csp->content_type & CT_DECLARED))
2421 * The server didn't bother to declare a MIME-Type.
2422 * Assume it's text that can be filtered.
2424 * This also regulary happens with 304 responses,
2425 * therefore logging anything here would cause
2428 csp->content_type |= CT_TEXT;
2432 * Choose the applying filter function based on
2433 * the content type and action settings.
2435 if ((csp->content_type & CT_TEXT) &&
2436 (csp->rlist != NULL) &&
2437 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2439 filter_function = pcrs_filter_response;
2441 else if ((csp->content_type & CT_GIF) &&
2442 (csp->action->flags & ACTION_DEANIMATE))
2444 filter_function = gif_deanimate_response;
2447 return filter_function;
2451 /*********************************************************************
2453 * Function : remove_chunked_transfer_coding
2455 * Description : In-situ remove the "chunked" transfer coding as defined
2456 * in rfc2616 from a buffer.
2459 * 1 : buffer = Pointer to the text buffer
2460 * 2 : size = In: Number of bytes to be processed,
2461 * Out: Number of bytes after de-chunking.
2462 * (undefined in case of errors)
2464 * Returns : JB_ERR_OK for success,
2465 * JB_ERR_PARSE otherwise
2467 *********************************************************************/
2468 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2471 unsigned int chunksize = 0;
2472 char *from_p, *to_p;
2475 from_p = to_p = buffer;
2477 if (sscanf(buffer, "%x", &chunksize) != 1)
2479 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2480 return JB_ERR_PARSE;
2483 while (chunksize > 0U)
2485 if (NULL == (from_p = strstr(from_p, "\r\n")))
2487 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2488 return JB_ERR_PARSE;
2491 if ((newsize += chunksize) >= *size)
2493 log_error(LOG_LEVEL_ERROR,
2494 "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding",
2496 return JB_ERR_PARSE;
2500 memmove(to_p, from_p, (size_t) chunksize);
2501 to_p = buffer + newsize;
2502 from_p += chunksize + 2;
2504 if (sscanf(from_p, "%x", &chunksize) != 1)
2506 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2511 /* XXX: Should get its own loglevel. */
2512 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2521 /*********************************************************************
2523 * Function : prepare_for_filtering
2525 * Description : If necessary, de-chunks and decompresses
2526 * the content so it can get filterd.
2529 * 1 : csp = Current client state (buffers, headers, etc...)
2531 * Returns : JB_ERR_OK for success,
2532 * JB_ERR_PARSE otherwise
2534 *********************************************************************/
2535 static jb_err prepare_for_filtering(struct client_state *csp)
2537 jb_err err = JB_ERR_OK;
2540 * If the body has a "chunked" transfer-encoding,
2541 * get rid of it, adjusting size and iob->eod
2543 if (csp->flags & CSP_FLAG_CHUNKED)
2545 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2547 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2548 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2549 if (JB_ERR_OK == err)
2551 csp->iob->eod = csp->iob->cur + size;
2552 csp->flags |= CSP_FLAG_MODIFIED;
2556 return JB_ERR_PARSE;
2562 * If the body has a supported transfer-encoding,
2563 * decompress it, adjusting size and iob->eod.
2565 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2567 if (0 == csp->iob->eod - csp->iob->cur)
2569 /* Nothing left after de-chunking. */
2573 err = decompress_iob(csp);
2575 if (JB_ERR_OK == err)
2577 csp->flags |= CSP_FLAG_MODIFIED;
2578 csp->content_type &= ~CT_TABOO;
2583 * Unset CT_GZIP and CT_DEFLATE to remember not
2584 * to modify the Content-Encoding header later.
2586 csp->content_type &= ~CT_GZIP;
2587 csp->content_type &= ~CT_DEFLATE;
2596 /*********************************************************************
2598 * Function : execute_content_filter
2600 * Description : Executes a given content filter.
2603 * 1 : csp = Current client state (buffers, headers, etc...)
2604 * 2 : content_filter = The filter function to execute
2606 * Returns : Pointer to the modified buffer, or
2607 * NULL if filtering failed or wasn't necessary.
2609 *********************************************************************/
2610 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2612 if (0 == csp->iob->eod - csp->iob->cur)
2615 * No content (probably status code 301, 302 ...),
2616 * no filtering necessary.
2621 if (JB_ERR_OK != prepare_for_filtering(csp))
2624 * failed to de-chunk or decompress.
2629 if (0 == csp->iob->eod - csp->iob->cur)
2632 * Clown alarm: chunked and/or compressed nothing delivered.
2637 return ((*content_filter)(csp));
2641 /*********************************************************************
2643 * Function : get_url_actions
2645 * Description : Gets the actions for this URL.
2648 * 1 : csp = Current client state (buffers, headers, etc...)
2649 * 2 : http = http_request request for blocked URLs
2653 *********************************************************************/
2654 void get_url_actions(struct client_state *csp, struct http_request *http)
2656 struct file_list *fl;
2657 struct url_actions *b;
2660 init_current_action(csp->action);
2662 for (i = 0; i < MAX_AF_FILES; i++)
2664 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2669 apply_url_actions(csp->action, http, b);
2676 /*********************************************************************
2678 * Function : apply_url_actions
2680 * Description : Applies a list of URL actions.
2683 * 1 : action = Destination.
2684 * 2 : http = Current URL
2685 * 3 : b = list of URL actions to apply
2689 *********************************************************************/
2690 void apply_url_actions(struct current_action_spec *action,
2691 struct http_request *http,
2692 struct url_actions *b)
2696 /* Should never happen */
2700 for (b = b->next; NULL != b; b = b->next)
2702 if (url_match(b->url, http))
2704 merge_current_action(action, b->action);
2710 /*********************************************************************
2712 * Function : get_forward_override_settings
2714 * Description : Returns forward settings as specified with the
2715 * forward-override{} action. forward-override accepts
2716 * forward lines similar to the one used in the
2717 * configuration file, but without the URL pattern.
2723 * in the configuration file can be replaced with
2724 * the action section:
2726 * {+forward-override{forward .}}
2730 * 1 : csp = Current client state (buffers, headers, etc...)
2732 * Returns : Pointer to forwarding structure in case of success.
2733 * Invalid syntax is fatal.
2735 *********************************************************************/
2736 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2738 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2739 char forward_settings[BUFFER_SIZE];
2740 char *http_parent = NULL;
2741 /* variable names were chosen for consistency reasons. */
2742 struct forward_spec *fwd = NULL;
2746 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2747 /* Should be enforced by load_one_actions_file() */
2748 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2750 /* Create a copy ssplit can modify */
2751 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2753 if (NULL != csp->fwd)
2756 * XXX: Currently necessary to prevent memory
2757 * leaks when the show-url-info cgi page is visited.
2759 unload_forward_spec(csp->fwd);
2763 * allocate a new forward node, valid only for
2764 * the lifetime of this request. Save its location
2765 * in csp as well, so sweep() can free it later on.
2767 fwd = csp->fwd = zalloc(sizeof(*fwd));
2770 log_error(LOG_LEVEL_FATAL,
2771 "can't allocate memory for forward-override{%s}", forward_override_line);
2772 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2776 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2777 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2779 fwd->type = SOCKS_NONE;
2781 /* Parse the parent HTTP proxy host:port */
2782 http_parent = vec[1];
2785 else if (vec_count == 3)
2787 char *socks_proxy = NULL;
2789 if (!strcasecmp(vec[0], "forward-socks4"))
2791 fwd->type = SOCKS_4;
2792 socks_proxy = vec[1];
2794 else if (!strcasecmp(vec[0], "forward-socks4a"))
2796 fwd->type = SOCKS_4A;
2797 socks_proxy = vec[1];
2799 else if (!strcasecmp(vec[0], "forward-socks5"))
2801 fwd->type = SOCKS_5;
2802 socks_proxy = vec[1];
2805 if (NULL != socks_proxy)
2807 /* Parse the SOCKS proxy host[:port] */
2808 char *p = socks_proxy;
2809 fwd->gateway_host = strdup(socks_proxy);
2811 if ((*fwd->gateway_host == '[')
2812 && (NULL != (p = strchr(fwd->gateway_host, ']'))))
2815 memmove(fwd->gateway_host, fwd->gateway_host + 1,
2816 (size_t)(p - fwd->gateway_host));
2819 fwd->gateway_port = (int)strtol(++p, NULL, 0);
2822 else if (NULL != (p = strchr(fwd->gateway_host, ':')))
2825 fwd->gateway_port = (int)strtol(p, NULL, 0);
2828 if (fwd->gateway_port <= 0)
2830 fwd->gateway_port = 1080;
2833 http_parent = vec[2];
2837 if (NULL == http_parent)
2839 log_error(LOG_LEVEL_FATAL,
2840 "Invalid forward-override syntax in: %s", forward_override_line);
2841 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2844 /* Parse http forwarding settings */
2845 if (strcmp(http_parent, ".") != 0)
2847 char *p = http_parent;
2848 fwd->forward_host = strdup(http_parent);
2850 if ((*fwd->forward_host == '[')
2851 && (NULL != (p = strchr(fwd->forward_host, ']'))))
2854 memmove(fwd->forward_host, fwd->forward_host + 1,
2855 (size_t)(p - fwd->forward_host));
2858 fwd->forward_port = (int)strtol(++p, NULL, 0);
2861 else if (NULL != (p = strchr(fwd->forward_host, ':')))
2864 fwd->forward_port = (int)strtol(p, NULL, 0);
2867 if (fwd->forward_port <= 0)
2869 fwd->forward_port = 8000;
2873 assert (NULL != fwd);
2875 log_error(LOG_LEVEL_CONNECT,
2876 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2882 /*********************************************************************
2884 * Function : forward_url
2886 * Description : Should we forward this to another proxy?
2889 * 1 : csp = Current client state (buffers, headers, etc...)
2890 * 2 : http = http_request request for current URL
2892 * Returns : Pointer to forwarding information.
2894 *********************************************************************/
2895 const struct forward_spec *forward_url(struct client_state *csp,
2896 const struct http_request *http)
2898 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2899 struct forward_spec *fwd = csp->config->forward;
2901 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2903 return get_forward_override_settings(csp);
2913 if (url_match(fwd->url, http))
2924 /*********************************************************************
2926 * Function : direct_response
2928 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2929 * request and if so, return a HTTP 501 to the client.
2931 * FIXME: I have a stupid name and I should handle the
2932 * requests properly. Still, what we do here is rfc-
2933 * compliant, whereas ignoring or forwarding are not.
2936 * 1 : csp = Current client state (buffers, headers, etc...)
2938 * Returns : http_response if , NULL if nonmatch or handler fail
2940 *********************************************************************/
2941 struct http_response *direct_response(struct client_state *csp)
2943 struct http_response *rsp;
2944 struct list_entry *p;
2946 if ((0 == strcmpic(csp->http->gpc, "trace"))
2947 || (0 == strcmpic(csp->http->gpc, "options")))
2949 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2951 if (!strncmpic("Max-Forwards:", p->str, 13))
2953 unsigned int max_forwards;
2956 * If it's a Max-Forwards value of zero,
2957 * we have to intercept the request.
2959 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2962 * FIXME: We could handle at least TRACE here,
2963 * but that would require a verbatim copy of
2964 * the request which we don't have anymore
2966 log_error(LOG_LEVEL_HEADER,
2967 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2970 /* Get mem for response or fail*/
2971 if (NULL == (rsp = alloc_http_response()))
2973 return cgi_error_memory();
2976 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2978 free_http_response(rsp);
2979 return cgi_error_memory();
2983 rsp->reason = RSP_REASON_UNSUPPORTED;
2985 return(finish_http_response(csp, rsp));
2994 /*********************************************************************
2996 * Function : content_filters_enabled
2998 * Description : Checks whether there are any content filters
2999 * enabled for the current request.
3002 * 1 : action = Action spec to check.
3004 * Returns : TRUE for yes, FALSE otherwise
3006 *********************************************************************/
3007 int content_filters_enabled(const struct current_action_spec *action)
3009 return ((action->flags & ACTION_DEANIMATE) ||
3010 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));