1 const char filters_rcs[] = "$Id: filters.c,v 1.116 2009/04/17 11:34:34 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 1.116 2009/04/17 11:34:34 fabiankeil
44 * Style cosmetics for the IPv6 code.
46 * Revision 1.115 2009/04/17 11:29:18 fabiankeil
47 * Compile fix for BSD libc.
49 * Revision 1.114 2009/04/17 11:27:49 fabiankeil
50 * Petr Pisar's privoxy-3.0.12-ipv6-3.diff.
52 * Revision 1.113 2009/03/08 14:19:23 fabiankeil
53 * Fix justified (but harmless) compiler warnings
54 * on platforms where sizeof(int) < sizeof(long).
56 * Revision 1.112 2009/03/01 18:28:23 fabiankeil
57 * Help clang understand that we aren't dereferencing
60 * Revision 1.111 2008/12/04 18:13:46 fabiankeil
61 * Fix a cparser warning.
63 * Revision 1.110 2008/11/10 16:40:25 fabiankeil
64 * Fix a gcc44 warning.
66 * Revision 1.109 2008/11/08 15:48:41 fabiankeil
67 * Mention actual values when complaining about
68 * the chunk size exceeding the buffer size.
70 * Revision 1.108 2008/05/21 15:35:08 fabiankeil
71 * - Mark csp as immutable for block_acl().
72 * - Remove an obsolete complaint about filter_popups().
74 * Revision 1.107 2008/05/04 17:52:56 fabiankeil
75 * Adjust parse_http_url() call to new prototype.
77 * Revision 1.106 2008/05/03 16:40:44 fabiankeil
78 * Change content_filters_enabled()'s parameter from
79 * csp->action to action so it can be also used in the
80 * CGI code. Don't bother checking if there are filters
81 * loaded, as that's somewhat besides the point.
83 * Revision 1.105 2008/03/28 15:13:39 fabiankeil
84 * Remove inspect-jpegs action.
86 * Revision 1.104 2008/03/27 18:27:24 fabiankeil
87 * Remove kill-popups action.
89 * Revision 1.103 2008/03/06 16:33:45 fabiankeil
90 * If limit-connect isn't used, don't limit CONNECT requests to port 443.
92 * Revision 1.102 2008/03/01 14:00:44 fabiankeil
93 * Let the block action take the reason for the block
94 * as argument and show it on the "blocked" page.
96 * Revision 1.101 2008/02/23 16:57:12 fabiankeil
97 * Rename url_actions() to get_url_actions() and let it
98 * use the standard parameter ordering.
100 * Revision 1.100 2008/02/23 16:33:43 fabiankeil
101 * Let forward_url() use the standard parameter ordering
102 * and mark its second parameter immutable.
104 * Revision 1.99 2008/02/03 13:57:58 fabiankeil
105 * Add SOCKS5 support for forward-override{}.
107 * Revision 1.98 2008/01/04 17:43:45 fabiankeil
108 * Improve the warning messages that get logged if the action files
109 * "enable" filters but no filters of that type have been loaded.
111 * Revision 1.97 2007/11/30 15:37:03 fabiankeil
112 * Use freez instead of free.
114 * Revision 1.96 2007/10/19 16:53:28 fabiankeil
115 * Add helper function to check if any content filters are enabled.
117 * Revision 1.95 2007/10/17 19:31:20 fabiankeil
118 * Omitting the zero chunk that ends the chunk transfer encoding seems
119 * to be the new black. Log the problem and continue filtering anyway.
121 * Revision 1.94 2007/09/29 13:20:20 fabiankeil
122 * Remove two redundant and one useless log messages.
124 * Revision 1.93 2007/09/29 10:21:16 fabiankeil
125 * - Move get_filter_function() from jcc.c to filters.c
126 * so the filter functions can be static.
127 * - Don't bother filtering body-less responses.
129 * Revision 1.92 2007/09/28 16:38:55 fabiankeil
130 * - Execute content filters through execute_content_filter().
131 * - Add prepare_for_filtering() so filter functions don't have to
132 * care about de-chunking and decompression. As a side effect this enables
133 * decompression for gif_deanimate_response() and jpeg_inspect_response().
134 * - Change remove_chunked_transfer_coding()'s return type to jb_err.
135 * Some clowns feel like chunking empty responses in which case
136 * (size == 0) is valid but previously would be interpreted as error.
138 * Revision 1.91 2007/09/02 15:31:20 fabiankeil
139 * Move match_portlist() from filter.c to urlmatch.c.
140 * It's used for url matching, not for filtering.
142 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
143 * Remove newline at the end of a log_error() message.
145 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
146 * #1763173 from Stefan Huehner: declare some more functions static.
148 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
149 * Add forward-override{} to change the forwarding settings through
150 * action sections. This is mainly interesting to forward different
151 * clients differently (for example based on User-Agent or request
154 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
155 * Make sure filters with dynamic jobs actually use them.
157 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
158 * - Introduce dynamic pcrs jobs that can resolve variables.
159 * - Don't run redirect functions more than once,
160 * unless they are activated more than once.
162 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
163 * - Log the content size after decompression in decompress_iob()
164 * instead of pcrs_filter_response().
166 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
167 * Use dedicated header filter actions instead of abusing "filter".
168 * Replace "filter-client-headers" and "filter-client-headers"
169 * with "server-header-filter" and "client-header-filter".
171 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
172 * New config option: enforce-blocks.
174 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
175 * - Fix port handling in acl_addr() and use a temporary acl spec
176 * copy so error messages don't contain a truncated version.
177 * - Log size of iob before and after decompression.
179 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
180 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
181 * - Hide the "Go there anyway" link for blocked CONNECT
182 * requests where going there anyway doesn't work anyway.
184 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
185 * - Save the reason for generating http_responses.
186 * - Block (+block) with status code 403 instead of 404.
187 * - Use a different kludge to remember a failed decompression.
189 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
190 * Search for Max-Forwards headers case-insensitive,
191 * don't generate the "501 unsupported" message for invalid
192 * Max-Forwards values and don't increase negative ones.
194 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
195 * - Add HEAD support to finish_http_response.
196 * - Add error favicon to internal HTML error messages.
198 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
199 * Mark *csp as immutable for is_untrusted_url()
200 * and is_imageurl(). Closes FR 1237736.
202 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
203 * Integrate a modified version of Wil Mahan's
204 * zlib patch (PR #895531).
206 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
207 * Fixed gcc43 conversion warnings,
208 * changed sprintf calls to snprintf.
210 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
211 * Adjust comment in pcrs_filter_response()
212 * to recent pcrs changes. Hohoho.
214 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
215 * Don't crash if pcre returns an error code
216 * that pcrs didn't expect. Fixes BR 1621173.
218 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
219 * Modified is_untrusted_url to complain in case of
220 * write errors and to give a reason when adding new
221 * entries to the trustfile. Closes FR 1097611.
223 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
224 * Skip empty filter files in pcrs_filter_response,
225 * but don't ignore the ones that come afterwards.
226 * Fixes parts of BR 1619208.
228 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
229 * Added some sanity checks for get_last_url().
230 * Fixed possible segfault caused by my last commit.
232 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
233 * Let get_last_url() catch https URLs as well.
235 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
236 * Make sure get_last_url() behaves like advertised
237 * and fast-redirects{} can be combined with redirect{}.
239 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
240 * Implemented +redirect{s@foo@bar@} to generate
241 * a redirect based on a rewritten version of the
244 * Revision 1.66 2006/09/23 13:26:38 roro
245 * Replace TABs by spaces in source code.
247 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
248 * Fix +redirect{}. Didn't work with -fast-redirects.
250 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
251 * Block requests for untrusted URLs with status
252 * code 403 instead of 200.
254 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
255 * Don't free p which is still in use and will be later
256 * freed by free_map(). Don't claim the referrer is unknown
257 * when the client didn't set one.
259 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
260 * Feature request 595948: Re-Filter logging in single line
262 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
263 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
265 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
266 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
267 * with what was really the latest development (the v_3_0_branch branch)
269 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
270 * Multiple filter file support
272 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
273 * A few memory leaks plugged (mostly on error paths)
275 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
276 * Add the ability to check jpeg images for invalid
277 * lengths of comment blocks. Defensive strategy
278 * against the exploit:
279 * Microsoft Security Bulletin MS04-028
280 * Buffer Overrun in JPEG Processing (GDI+) Could
281 * Allow Code Execution (833987)
282 * Enabled with +inspect-jpegs in actions files.
284 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
285 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
287 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
288 * Fixed bug #839859: "See why" link URL now gets url-encoded.
290 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
293 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
294 * Added basic support for OPTIONS and TRACE HTTP methods:
295 * New function direct_response which handles OPTIONS and
296 * TRACE requests whose Max-Forwards header field is zero.
298 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
299 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
301 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
302 * - Bugfix: Executing a filters without jobs no longer results in
303 * turing off *all* filters.
304 * - Security fix: Malicious web servers can't cause a seg fault
305 * through bogus chunk sizes anymore
307 * Revision 1.58 2002/04/24 02:11:17 oes
308 * Jon's multiple AF patch: url_actions now evaluates rules
311 * Revision 1.57 2002/04/08 20:38:34 swa
314 * Revision 1.56 2002/04/05 15:51:24 oes
315 * - bugfix: error-pages now get correct request protocol
316 * - fix for invalid HTML in trust info
318 * Revision 1.55 2002/04/02 16:13:51 oes
319 * Fix: No "Go there anyway" for SSL
321 * Revision 1.54 2002/04/02 14:55:56 oes
322 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
324 * Revision 1.53 2002/03/26 22:29:54 swa
325 * we have a new homepage!
327 * Revision 1.52 2002/03/24 16:35:57 jongfoster
330 * Revision 1.51 2002/03/24 15:23:33 jongfoster
333 * Revision 1.50 2002/03/24 13:25:43 swa
334 * name change related issues
336 * Revision 1.49 2002/03/16 20:29:14 oes
339 * Revision 1.48 2002/03/13 20:25:34 oes
340 * Better logging for content filters
342 * Revision 1.47 2002/03/13 00:30:52 jongfoster
344 * Added option of always sending redirect for imageblock,
345 * currently disabled with #if 0.
347 * Revision 1.46 2002/03/12 01:42:49 oes
348 * Introduced modular filters
350 * Revision 1.45 2002/03/08 16:47:50 oes
351 * Added choice beween GIF and PNG built-in images
353 * Revision 1.44 2002/03/07 03:49:31 oes
354 * - Fixed compiler warnings etc
355 * - Changed built-in images from GIF to PNG
356 * (with regard to Unisys patent issue)
357 * - Added a 4x4 pattern PNG which is less intrusive
358 * than the logo but also clearly marks the deleted banners
360 * Revision 1.43 2002/01/22 23:51:59 jongfoster
361 * Replacing strsav() with the safer string_append().
363 * Adding missing html_encode() to error message generators. Where encoded
364 * and unencoded versions of a string were provided, removing the unencoded
367 * Revision 1.42 2002/01/17 21:00:32 jongfoster
368 * Moving all our URL and URL pattern parsing code to urlmatch.c.
370 * Using a single, simple url_match(pattern,url) function - rather than
371 * the 3-line match routine which was repeated all over the place.
373 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
375 * Using parse_http_url() to parse URLs without faking a HTTP
376 * request line for parse_http_request().
378 * Revision 1.41 2001/11/13 00:14:07 jongfoster
379 * Fixing stupid bug now I've figured out what || means.
380 * (It always returns 0 or 1, not one of it's paramaters.)
382 * Revision 1.40 2001/10/26 17:37:55 oes
383 * - Re-enabled Netscape 200/404 bug workaround in block_url():
384 * - Removed OS/2 special case
385 * - Made block_url() independant from sed() having been run
386 * - Made trust_url independant from sed() having been run
387 * - Made is_imageurl independant from sed() having been run.
388 * It now checks User-Agent: and Accept: by itself.
391 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
392 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
393 * threads to call select() simultaneously. So, it's time to do a real, live,
394 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
395 * (native). Both versions will work, but using __OS2__ offers multi-threading.
397 * Revision 1.38 2001/10/23 21:32:33 jongfoster
398 * Adding error-checking to selected functions
400 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
401 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
402 * filters.c. Added a FIXME in front of the offending code. I'll gladly
403 * put in a better/more robust fix for all parties if one is presented...
404 * It seems that just returning 200 instead of 404 would pretty much fix
405 * it for everyone, but I don't know all the history of the problem.
407 * Revision 1.36 2001/10/10 16:44:16 oes
408 * Added match_portlist function
410 * Revision 1.35 2001/10/07 15:41:23 oes
411 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
413 * New function remove_chunked_transfer_coding that strips chunked
414 * transfer coding to plain and is called by pcrs_filter_response
415 * and gif_deanimate_response if neccessary
417 * Improved handling of zero-change re_filter runs
419 * pcrs_filter_response and gif_deanimate_response now remove
420 * chunked transfer codeing before processing the body.
422 * Revision 1.34 2001/09/20 15:49:36 steudten
424 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
425 * See cgi.c fill_template().
427 * Revision 1.33 2001/09/16 17:05:14 jongfoster
428 * Removing unused #include showarg.h
430 * Revision 1.32 2001/09/16 13:21:27 jongfoster
431 * Changes to use new list functions.
433 * Revision 1.31 2001/09/16 11:38:02 jongfoster
434 * Splitting fill_template() into 2 functions:
435 * template_load() loads the file
436 * template_fill() performs the PCRS regexps.
437 * This is because the CGI edit interface has a "table row"
438 * template which is used many times in the page - this
439 * change means it's only loaded from disk once.
441 * Revision 1.30 2001/09/16 11:00:10 jongfoster
442 * New function alloc_http_response, for symmetry with free_http_response
444 * Revision 1.29 2001/09/13 23:32:40 jongfoster
445 * Moving image data to cgi.c rather than cgi.h
446 * Fixing a GPF under Win32 (and any other OS that protects global
447 * constants from being written to).
449 * Revision 1.28 2001/09/10 10:18:51 oes
450 * Silenced compiler warnings
452 * Revision 1.27 2001/08/05 16:06:20 jongfoster
453 * Modifiying "struct map" so that there are now separate header and
454 * "map_entry" structures. This means that functions which modify a
455 * map no longer need to return a pointer to the modified map.
456 * Also, it no longer reverses the order of the entries (which may be
457 * important with some advanced template substitutions).
459 * Revision 1.26 2001/07/30 22:08:36 jongfoster
460 * Tidying up #defines:
461 * - All feature #defines are now of the form FEATURE_xxx
462 * - Permanently turned off WIN_GUI_EDIT
463 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
465 * Revision 1.25 2001/07/26 10:09:46 oes
466 * Made browser detection a little less naive
468 * Revision 1.24 2001/07/25 17:22:51 oes
469 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
471 * Revision 1.23 2001/07/23 13:40:12 oes
472 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
474 * Revision 1.22 2001/07/18 12:29:34 oes
475 * - Made gif_deanimate_response respect
476 * csp->action->string[ACTION_STRING_DEANIMATE]
477 * - Logging cosmetics
479 * Revision 1.21 2001/07/13 13:59:53 oes
480 * - Introduced gif_deanimate_response which shares the
481 * generic content modification interface of pcrs_filter_response
482 * and acts as a wrapper to deanimate.c:gif_deanimate()
483 * - Renamed re_process_buffer to pcrs_filter_response
484 * - pcrs_filter_response now returns NULL on failiure
485 * - Removed all #ifdef PCRS
487 * Revision 1.20 2001/07/01 17:01:04 oes
488 * Added comments and missing return statement in is_untrusted_url()
490 * Revision 1.19 2001/06/29 21:45:41 oes
491 * Indentation, CRLF->LF, Tab-> Space
493 * Revision 1.18 2001/06/29 13:27:38 oes
494 * - Cleaned up, renamed and reorderd functions
495 * and improved comments
498 * - Ported to CGI platform. Now delivers
499 * http_response or NULL
500 * - Unified HTML and GIF generation (moved image detection
501 * and GIF generation here from jcc.c:chat())
502 * - Fixed HTTP status to:
503 * - 403 (Forbidden) for the "blocked" HTML message
504 * - 200 (OK) for GIF answers
505 * - 302 (Redirect) for redirect to GIF
508 * - Ported to CGI platform. Now delivers
509 * http_response or NULL
510 * - Separated detection of untrusted URL into
511 * (bool)is_untrusted_url
512 * - Added enforcement of untrusted requests
514 * - Moved redirect_url() from cgi.c to here
515 * and ported it to the CGI platform
517 * - Removed logentry from cancelled commit
519 * Revision 1.17 2001/06/09 10:55:28 jongfoster
520 * Changing BUFSIZ ==> BUFFER_SIZE
522 * Revision 1.16 2001/06/07 23:10:26 jongfoster
523 * Allowing unanchored domain patterns to back off and retry
524 * if they partially match. Optimized right-anchored patterns.
525 * Moving ACL and forward files into config file.
526 * Replacing struct gateway with struct forward_spec
528 * Revision 1.15 2001/06/03 19:12:00 oes
529 * extracted-CGI relevant stuff
531 * Revision 1.14 2001/06/01 10:30:55 oes
532 * Added optional left-anchoring to domaincmp
534 * Revision 1.13 2001/05/31 21:21:30 jongfoster
535 * Permissionsfile / actions file changes:
536 * - Changed "permission" to "action" throughout
537 * - changes to file format to allow string parameters
538 * - Moved helper functions to actions.c
540 * Revision 1.12 2001/05/31 17:35:20 oes
542 * - Enhanced domain part globbing with infix and prefix asterisk
543 * matching and optional unanchored operation
545 * Revision 1.11 2001/05/29 11:53:23 oes
546 * "See why" link added to "blocked" page
548 * Revision 1.10 2001/05/29 09:50:24 jongfoster
549 * Unified blocklist/imagelist/permissionslist.
550 * File format is still under discussion, but the internal changes
553 * Also modified interceptor behaviour:
554 * - We now intercept all URLs beginning with one of the following
555 * prefixes (and *only* these prefixes):
557 * * http://ijbswa.sf.net/config/
558 * * http://ijbswa.sourceforge.net/config/
559 * - New interceptors "home page" - go to http://i.j.b/ to see it.
560 * - Internal changes so that intercepted and fast redirect pages
561 * are not replaced with an image.
562 * - Interceptors now have the option to send a binary page direct
563 * to the client. (i.e. ijb-send-banner uses this)
564 * - Implemented show-url-info interceptor. (Which is why I needed
565 * the above interceptors changes - a typical URL is
566 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
567 * The previous mechanism would not have intercepted that, and
568 * if it had been intercepted then it then it would have replaced
571 * Revision 1.9 2001/05/27 22:17:04 oes
573 * - re_process_buffer no longer writes the modified buffer
574 * to the client, which was very ugly. It now returns the
575 * buffer, which it is then written by chat.
577 * - content_length now adjusts the Content-Length: header
578 * for modified documents rather than crunch()ing it.
579 * (Length info in csp->content_length, which is 0 for
580 * unmodified documents)
582 * - For this to work, sed() is called twice when filtering.
584 * Revision 1.8 2001/05/26 17:13:28 jongfoster
585 * Filled in a function comment.
587 * Revision 1.7 2001/05/26 15:26:15 jongfoster
588 * ACL feature now provides more security by immediately dropping
589 * connections from untrusted hosts.
591 * Revision 1.6 2001/05/26 00:28:36 jongfoster
592 * Automatic reloading of config file.
593 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
594 * Most of the global variables have been moved to a new
595 * struct configuration_spec, accessed through csp->config->globalname
596 * Most of the globals remaining are used by the Win32 GUI.
598 * Revision 1.5 2001/05/25 22:34:30 jongfoster
601 * Revision 1.4 2001/05/22 18:46:04 oes
603 * - Enabled filtering banners by size rather than URL
604 * by adding patterns that replace all standard banner
605 * sizes with the "Junkbuster" gif to the re_filterfile
607 * - Enabled filtering WebBugs by providing a pattern
608 * which kills all 1x1 images
610 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
611 * which is selected by the (nonstandard and therefore
612 * capital) letter 'U' in the option string.
613 * It causes the quantifiers to be ungreedy by default.
614 * Appending a ? turns back to greedy (!).
616 * - Added a new interceptor ijb-send-banner, which
617 * sends back the "Junkbuster" gif. Without imagelist or
618 * MSIE detection support, or if tinygif = 1, or the
619 * URL isn't recognized as an imageurl, a lame HTML
620 * explanation is sent instead.
622 * - Added new feature, which permits blocking remote
623 * script redirects and firing back a local redirect
625 * The feature is conditionally compiled, i.e. it
626 * can be disabled with --disable-fast-redirects,
627 * plus it must be activated by a "fast-redirects"
628 * line in the config file, has its own log level
629 * and of course wants to be displayed by show-proxy-args
630 * Note: Boy, all the #ifdefs in 1001 locations and
631 * all the fumbling with configure.in and acconfig.h
632 * were *way* more work than the feature itself :-(
634 * - Because a generic redirect template was needed for
635 * this, tinygif = 3 now uses the same.
637 * - Moved GIFs, and other static HTTP response templates
642 * - Removed some >400 CRs again (Jon, you really worked
645 * Revision 1.3 2001/05/20 16:44:47 jongfoster
646 * Removing last hardcoded Junkbusters.com URLs.
648 * Revision 1.2 2001/05/20 01:21:20 jongfoster
649 * Version 2.9.4 checkin.
650 * - Merged popupfile and cookiefile, and added control over PCRS
651 * filtering, in new "permissionsfile".
652 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
653 * file error you now get a message box (in the Win32 GUI) rather
654 * than the program exiting with no explanation.
655 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
657 * - Removed tabs from "config"
658 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
659 * - Bumped up version number.
661 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
662 * Initial import of version 2.9.3 source tree
665 *********************************************************************/
671 #include <sys/types.h>
677 #ifdef HAVE_GETADDRINFO
679 #include <sys/socket.h>
680 #endif /* def HAVE_GETADDRINFO */
685 #endif /* ndef __OS2__ */
686 #include <netinet/in.h>
688 #include <winsock2.h>
689 #endif /* ndef _WIN32 */
693 #endif /* def __OS2__ */
701 #include "jbsockets.h"
702 #include "miscutil.h"
706 #include "deanimate.h"
707 #include "urlmatch.h"
714 const char filters_h_rcs[] = FILTERS_H_VERSION;
716 /* Fix a problem with Solaris. There should be no effect on other
718 * Solaris's isspace() is a macro which uses it's argument directly
719 * as an array index. Therefore we need to make sure that high-bit
720 * characters generate +ve values, and ideally we also want to make
721 * the argument match the declared parameter type of "int".
723 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
725 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
726 static jb_err prepare_for_filtering(struct client_state *csp);
729 #ifdef HAVE_GETADDRINFO
730 /*********************************************************************
732 * Function : sockaddr_storage_to_ip
734 * Description : Access internal structure of sockaddr_storage
737 * 1 : addr = socket address
738 * 2 : ip = IP address as array of octets in network order
739 * (it points into addr)
740 * 3 : len = length of IP address in octets
741 * 4 : port = port number in network order;
743 * Returns : 0 = no errror; -1 otherwise.
745 *********************************************************************/
746 int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, uint8_t **ip,
747 unsigned int *len, in_port_t **port)
754 switch (addr->ss_family)
764 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
768 *port = &((struct sockaddr_in *)addr)->sin_port;
779 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
783 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
788 /* Unsupported address family */
796 /*********************************************************************
798 * Function : match_sockaddr
800 * Description : Check whether address matches network (IP address and port)
803 * 1 : network = socket address of subnework
804 * 2 : netmask = network mask as socket address
805 * 3 : address = checked socket address against given network
807 * Returns : 0 = doesn't match; 1 = does match
809 *********************************************************************/
810 int match_sockaddr(const struct sockaddr_storage *network,
811 const struct sockaddr_storage *netmask,
812 const struct sockaddr_storage *address)
814 uint8_t *network_addr, *netmask_addr, *address_addr;
815 unsigned int addr_len;
816 in_port_t *network_port, *netmask_port, *address_port;
819 if (network->ss_family != netmask->ss_family)
821 /* This should never happen */
822 log_error(LOG_LEVEL_ERROR,
823 "Internal error at %s:%llu: network and netmask differ in family",
828 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
829 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
830 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
832 /* Check for family */
833 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
834 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
836 /* Map AF_INET6 V4MAPPED address into AF_INET */
840 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
841 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
843 /* Map AF_INET6 V4MAPPED network into AF_INET */
848 else if (network->ss_family != address->ss_family)
853 /* XXX: Port check is signaled in netmask */
854 if (*netmask_port && *network_port != *address_port)
859 /* TODO: Optimize by checking by words insted of octets */
860 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
862 if ((network_addr[i] & netmask_addr[i]) !=
863 (address_addr[i] & netmask_addr[i]))
871 #endif /* def HAVE_GETADDRINFO */
874 /*********************************************************************
876 * Function : block_acl
878 * Description : Block this request?
879 * Decide yes or no based on ACL file.
882 * 1 : dst = The proxy or gateway address this is going to.
883 * Or NULL to check all possible targets.
884 * 2 : csp = Current client state (buffers, headers, etc...)
885 * Also includes the client IP address.
887 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
889 *********************************************************************/
890 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
892 struct access_control_list *acl = csp->config->acl;
894 /* if not using an access control list, then permit the connection */
900 /* search the list */
904 #ifdef HAVE_GETADDRINFO
905 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
907 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
913 /* Just want to check if they have any access */
914 if (acl->action == ACL_PERMIT)
920 #ifdef HAVE_GETADDRINFO
922 * XXX: An undefined acl->dst is full of zeros and should be
923 * considered a wildcard address. sockaddr_storage_to_ip()
924 * fails on such destinations because of unknown sa_familly
925 * (glibc only?). However this test is not portable.
927 * So, we signal the acl->dst is wildcard in wildcard_dst.
930 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
932 ((dst->addr & acl->dst->mask) == acl->dst->addr)
933 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
937 if (acl->action == ACL_PERMIT)
955 /*********************************************************************
957 * Function : acl_addr
959 * Description : Called from `load_config' to parse an ACL address.
962 * 1 : aspec = String specifying ACL address.
963 * 2 : aca = struct access_control_addr to fill in.
965 * Returns : 0 => Ok, everything else is an error.
967 *********************************************************************/
968 int acl_addr(const char *aspec, struct access_control_addr *aca)
971 #ifdef HAVE_GETADDRINFO
972 struct addrinfo hints, *result;
974 in_port_t *mask_port;
975 unsigned int addr_len;
978 #endif /* def HAVE_GETADDRINFO */
980 char *acl_spec = NULL;
982 #ifdef HAVE_GETADDRINFO
983 /* XXX: Depend on ai_family */
991 * Use a temporary acl spec copy so we can log
992 * the unmodified original in case of parse errors.
994 acl_spec = strdup(aspec);
995 if (acl_spec == NULL)
997 /* XXX: This will be logged as parse error. */
1001 if ((p = strchr(acl_spec, '/')) != NULL)
1004 if (ijb_isdigit(*p) == 0)
1009 masklength = atoi(p);
1012 if ((masklength < 0) ||
1013 #ifdef HAVE_GETADDRINFO
1024 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
1027 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
1036 p = strchr(acl_spec, ':');
1039 #ifdef HAVE_GETADDRINFO
1040 memset(&hints, 0, sizeof(struct addrinfo));
1041 hints.ai_family = AF_UNSPEC;
1042 hints.ai_socktype = SOCK_STREAM;
1044 i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result);
1049 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
1050 acl_spec, p, gai_strerror(i));
1054 /* TODO: Allow multihomed hostnames */
1055 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
1056 freeaddrinfo(result);
1063 port = strtol(p, &endptr, 10);
1065 if (port <= 0 || port > 65535 || *endptr != '\0')
1072 aca->port = (unsigned long)port;
1074 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
1077 if (aca->addr == INADDR_NONE)
1079 /* XXX: This will be logged as parse error. */
1082 #endif /* def HAVE_GETADDRINFO */
1084 /* build the netmask */
1085 #ifdef HAVE_GETADDRINFO
1086 /* Clip masklength according to current family. */
1087 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
1092 aca->mask.ss_family = aca->addr.ss_family;
1093 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
1100 /* ACL contains a port number, check ports in the future. */
1105 * XXX: This could be optimized to operate on whole words instead
1106 * of octets (128-bit CPU could do it in one iteration).
1109 * Octets after prefix can be ommitted because of
1110 * previous initialization to zeros.
1112 for (i = 0; (i < addr_len) && masklength; i++)
1114 if (masklength >= 8)
1116 mask_data[i] = 0xFF;
1122 * XXX: This assumes MSB of octet is on the left side.
1123 * This should be true for all architectures or solved
1124 * by the link layer.
1126 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
1133 for (i=1; i <= masklength ; i++)
1135 aca->mask |= (1U << (32 - i));
1138 /* now mask off the host portion of the ip address
1139 * (i.e. save on the network portion of the address).
1141 aca->addr = aca->addr & aca->mask;
1142 #endif /* def HAVE_GETADDRINFO */
1147 #endif /* def FEATURE_ACL */
1150 /*********************************************************************
1152 * Function : connect_port_is_forbidden
1154 * Description : Check to see if CONNECT requests to the destination
1155 * port of this request are forbidden. The check is
1156 * independend of the actual request method.
1159 * 1 : csp = Current client state (buffers, headers, etc...)
1161 * Returns : True if yes, false otherwise.
1163 *********************************************************************/
1164 int connect_port_is_forbidden(const struct client_state *csp)
1166 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
1167 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
1172 /*********************************************************************
1174 * Function : block_url
1176 * Description : Called from `chat'. Check to see if we need to block this.
1179 * 1 : csp = Current client state (buffers, headers, etc...)
1181 * Returns : NULL => unblocked, else HTTP block response
1183 *********************************************************************/
1184 struct http_response *block_url(struct client_state *csp)
1186 struct http_response *rsp;
1187 const char *new_content_type = NULL;
1190 * If it's not blocked, don't block it ;-)
1192 if ((csp->action->flags & ACTION_BLOCK) == 0)
1196 if (csp->action->flags & ACTION_REDIRECT)
1198 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
1201 * Else, prepare a response
1203 if (NULL == (rsp = alloc_http_response()))
1205 return cgi_error_memory();
1209 * If it's an image-url, send back an image or redirect
1210 * as specified by the relevant +image action
1212 #ifdef FEATURE_IMAGE_BLOCKING
1213 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
1214 && is_imageurl(csp))
1217 /* determine HOW images should be blocked */
1218 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
1220 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1222 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
1224 #if 1 /* Two alternative strategies, use this one for now: */
1226 /* and handle accordingly: */
1227 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1229 rsp->status = strdup("403 Request blocked by Privoxy");
1230 if (rsp->status == NULL)
1232 free_http_response(rsp);
1233 return cgi_error_memory();
1235 rsp->body = bindup(image_pattern_data, image_pattern_length);
1236 if (rsp->body == NULL)
1238 free_http_response(rsp);
1239 return cgi_error_memory();
1241 rsp->content_length = image_pattern_length;
1243 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1245 free_http_response(rsp);
1246 return cgi_error_memory();
1250 else if (0 == strcmpic(p, "blank"))
1252 rsp->status = strdup("403 Request blocked by Privoxy");
1253 if (rsp->status == NULL)
1255 free_http_response(rsp);
1256 return cgi_error_memory();
1258 rsp->body = bindup(image_blank_data, image_blank_length);
1259 if (rsp->body == NULL)
1261 free_http_response(rsp);
1262 return cgi_error_memory();
1264 rsp->content_length = image_blank_length;
1266 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1268 free_http_response(rsp);
1269 return cgi_error_memory();
1275 rsp->status = strdup("302 Local Redirect from Privoxy");
1276 if (rsp->status == NULL)
1278 free_http_response(rsp);
1279 return cgi_error_memory();
1282 if (enlist_unique_header(rsp->headers, "Location", p))
1284 free_http_response(rsp);
1285 return cgi_error_memory();
1289 #else /* Following code is disabled for now */
1291 /* and handle accordingly: */
1292 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1294 p = CGI_PREFIX "send-banner?type=pattern";
1296 else if (0 == strcmpic(p, "blank"))
1298 p = CGI_PREFIX "send-banner?type=blank";
1300 rsp->status = strdup("302 Local Redirect from Privoxy");
1301 if (rsp->status == NULL)
1303 free_http_response(rsp);
1304 return cgi_error_memory();
1307 if (enlist_unique_header(rsp->headers, "Location", p))
1309 free_http_response(rsp);
1310 return cgi_error_memory();
1312 #endif /* Preceeding code is disabled for now */
1314 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1317 * Send empty document.
1319 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1322 rsp->body = strdup(" ");
1323 rsp->content_length = 1;
1325 rsp->status = strdup("403 Request blocked by Privoxy");
1326 if (rsp->status == NULL)
1328 free_http_response(rsp);
1329 return cgi_error_memory();
1331 if (new_content_type != 0)
1333 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1334 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1336 free_http_response(rsp);
1337 return cgi_error_memory();
1342 #endif /* def FEATURE_IMAGE_BLOCKING */
1345 * Else, generate an HTML "blocked" message:
1349 struct map * exports;
1353 * Workaround for stupid Netscape bug which prevents
1354 * pages from being displayed if loading a referenced
1355 * JavaScript or style sheet fails. So make it appear
1356 * as if it succeeded.
1358 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1359 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1360 && !strstr(p, "Gecko") /* save Mozilla, */
1361 && !strstr(p, "compatible") /* MSIE */
1362 && !strstr(p, "Opera")) /* and Opera. */
1364 rsp->status = strdup("200 Request for blocked URL");
1368 rsp->status = strdup("403 Request for blocked URL");
1371 if (rsp->status == NULL)
1373 free_http_response(rsp);
1374 return cgi_error_memory();
1377 exports = default_exports(csp, NULL);
1378 if (exports == NULL)
1380 free_http_response(rsp);
1381 return cgi_error_memory();
1384 #ifdef FEATURE_FORCE_LOAD
1385 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1387 * Export the force conditional block killer if
1389 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1390 * - Privoxy is configured to enforce blocks, or
1391 * - it's a CONNECT request and enforcing wouldn't work anyway.
1393 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1394 || (0 == strcmpic(csp->http->gpc, "connect")))
1395 #endif /* ndef FEATURE_FORCE_LOAD */
1397 err = map_block_killer(exports, "force-support");
1400 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1401 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1402 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1403 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1406 const char *block_reason;
1407 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
1409 block_reason = csp->action->string[ACTION_STRING_BLOCK];
1413 assert(connect_port_is_forbidden(csp));
1414 block_reason = "Forbidden CONNECT port.";
1416 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
1421 free_http_response(rsp);
1422 return cgi_error_memory();
1425 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1428 free_http_response(rsp);
1429 return cgi_error_memory();
1432 rsp->reason = RSP_REASON_BLOCKED;
1434 return finish_http_response(csp, rsp);
1439 #ifdef FEATURE_TRUST
1440 /*********************************************************************
1442 * Function : trust_url FIXME: I should be called distrust_url
1444 * Description : Calls is_untrusted_url to determine if the URL is trusted
1445 * and if not, returns a HTTP 403 response with a reject message.
1448 * 1 : csp = Current client state (buffers, headers, etc...)
1450 * Returns : NULL => trusted, else http_response.
1452 *********************************************************************/
1453 struct http_response *trust_url(struct client_state *csp)
1455 struct http_response *rsp;
1456 struct map * exports;
1457 char buf[BUFFER_SIZE];
1459 struct url_spec **tl;
1464 * Don't bother to work on trusted URLs
1466 if (!is_untrusted_url(csp))
1472 * Else, prepare a response:
1474 if (NULL == (rsp = alloc_http_response()))
1476 return cgi_error_memory();
1479 rsp->status = strdup("403 Request blocked by Privoxy");
1480 exports = default_exports(csp, NULL);
1481 if (exports == NULL || rsp->status == NULL)
1483 free_http_response(rsp);
1484 return cgi_error_memory();
1488 * Export the protocol, host, port, and referrer information
1490 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1491 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1492 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1494 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1496 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1500 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1506 free_http_response(rsp);
1507 return cgi_error_memory();
1511 * Export the trust list
1514 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1516 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1517 string_append(&p, buf);
1519 err = map(exports, "trusted-referrers", 1, p, 0);
1524 free_http_response(rsp);
1525 return cgi_error_memory();
1529 * Export the trust info, if available
1531 if (csp->config->trust_info->first)
1533 struct list_entry *l;
1536 for (l = csp->config->trust_info->first; l ; l = l->next)
1538 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1539 string_append(&p, buf);
1541 err = map(exports, "trust-info", 1, p, 0);
1545 err = map_block_killer(exports, "have-trust-info");
1551 free_http_response(rsp);
1552 return cgi_error_memory();
1556 * Export the force conditional block killer if
1558 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1559 * - Privoxy is configured to enforce blocks, or
1560 * - it's a CONNECT request and enforcing wouldn't work anyway.
1562 #ifdef FEATURE_FORCE_LOAD
1563 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1564 || (0 == strcmpic(csp->http->gpc, "connect")))
1566 err = map_block_killer(exports, "force-support");
1570 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1572 #else /* ifndef FEATURE_FORCE_LOAD */
1573 err = map_block_killer(exports, "force-support");
1574 #endif /* ndef FEATURE_FORCE_LOAD */
1579 free_http_response(rsp);
1580 return cgi_error_memory();
1584 * Build the response
1586 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1589 free_http_response(rsp);
1590 return cgi_error_memory();
1592 rsp->reason = RSP_REASON_UNTRUSTED;
1594 return finish_http_response(csp, rsp);
1596 #endif /* def FEATURE_TRUST */
1599 /*********************************************************************
1601 * Function : compile_dynamic_pcrs_job_list
1603 * Description : Compiles a dynamic pcrs job list (one with variables
1604 * resolved at request time)
1607 * 1 : csp = Current client state (buffers, headers, etc...)
1608 * 2 : b = The filter list to compile
1610 * Returns : NULL in case of errors, otherwise the
1613 *********************************************************************/
1614 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1616 struct list_entry *pattern;
1617 pcrs_job *job_list = NULL;
1618 pcrs_job *dummy = NULL;
1619 pcrs_job *lastjob = NULL;
1622 const struct pcrs_variable variables[] =
1624 {"url", csp->http->url, 1},
1625 {"path", csp->http->path, 1},
1626 {"host", csp->http->host, 1},
1627 {"origin", csp->ip_addr_str, 1},
1631 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1633 assert(pattern->str != NULL);
1635 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1639 log_error(LOG_LEVEL_ERROR,
1640 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1641 pattern->str, b->name, pcrs_strerror(error));
1646 if (error == PCRS_WARN_TRUNCATION)
1648 log_error(LOG_LEVEL_ERROR,
1649 "At least one of the variables in \'%s\' had to "
1650 "be truncated before compilation", pattern->str);
1652 if (job_list == NULL)
1658 lastjob->next = dummy;
1668 /*********************************************************************
1670 * Function : rewrite_url
1672 * Description : Rewrites a URL with a single pcrs command
1673 * and returns the result if it differs from the
1674 * original and isn't obviously invalid.
1677 * 1 : old_url = URL to rewrite.
1678 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1681 * Returns : NULL if the pcrs_command didn't change the url, or
1682 * the result of the modification.
1684 *********************************************************************/
1685 char *rewrite_url(char *old_url, const char *pcrs_command)
1687 char *new_url = NULL;
1691 assert(pcrs_command);
1693 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1697 log_error(LOG_LEVEL_REDIRECTS,
1698 "pcrs command \"%s\" didn't change \"%s\".",
1699 pcrs_command, old_url);
1704 log_error(LOG_LEVEL_REDIRECTS,
1705 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1706 pcrs_command, old_url, pcrs_strerror(hits));
1709 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1711 log_error(LOG_LEVEL_ERROR,
1712 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1713 "but the result doesn't look like a valid URL and will be ignored.",
1714 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1719 log_error(LOG_LEVEL_REDIRECTS,
1720 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1721 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1729 #ifdef FEATURE_FAST_REDIRECTS
1730 /*********************************************************************
1732 * Function : get_last_url
1734 * Description : Search for the last URL inside a string.
1735 * If the string already is a URL, it will
1736 * be the first URL found.
1739 * 1 : subject = the string to check
1740 * 2 : redirect_mode = +fast-redirect{} mode
1742 * Returns : NULL if no URL was found, or
1743 * the last URL found.
1745 *********************************************************************/
1746 char *get_last_url(char *subject, const char *redirect_mode)
1748 char *new_url = NULL;
1752 assert(redirect_mode);
1754 subject = strdup(subject);
1755 if (subject == NULL)
1757 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1761 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1763 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1764 new_url = url_decode(subject);
1765 if (new_url != NULL)
1772 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1776 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1779 * Find the last URL encoded in the request
1782 while ((tmp = strstr(tmp, "http://")) != NULL)
1786 tmp = (new_url != NULL) ? new_url : subject;
1787 while ((tmp = strstr(tmp, "https://")) != NULL)
1792 if ((new_url != NULL)
1793 && ( (new_url != subject)
1794 || (0 == strncmpic(subject, "http://", 7))
1795 || (0 == strncmpic(subject, "https://", 8))
1799 * Return new URL if we found a redirect
1800 * or if the subject already was a URL.
1802 * The second case makes sure that we can
1803 * chain get_last_url after another redirection check
1804 * (like rewrite_url) without losing earlier redirects.
1806 new_url = strdup(new_url);
1815 #endif /* def FEATURE_FAST_REDIRECTS */
1818 /*********************************************************************
1820 * Function : redirect_url
1822 * Description : Checks if Privoxy should answer the request with
1823 * a HTTP redirect and generates the redirect if
1827 * 1 : csp = Current client state (buffers, headers, etc...)
1829 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1831 *********************************************************************/
1832 struct http_response *redirect_url(struct client_state *csp)
1834 struct http_response *rsp;
1835 #ifdef FEATURE_FAST_REDIRECTS
1837 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1838 * as compile-time option? The user can easily disable
1839 * it in his action file.
1841 char * redirect_mode;
1842 #endif /* def FEATURE_FAST_REDIRECTS */
1843 char *old_url = NULL;
1844 char *new_url = NULL;
1845 char *redirection_string;
1847 if ((csp->action->flags & ACTION_REDIRECT))
1849 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1852 * If the redirection string begins with 's',
1853 * assume it's a pcrs command, otherwise treat it as
1854 * properly formatted URL and use it for the redirection
1857 * According to RFC 2616 section 14.30 the URL
1858 * has to be absolute and if the user tries:
1859 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1860 * she would get undefined results anyway.
1864 if (*redirection_string == 's')
1866 old_url = csp->http->url;
1867 new_url = rewrite_url(old_url, redirection_string);
1871 log_error(LOG_LEVEL_REDIRECTS,
1872 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1873 redirection_string);
1874 new_url = strdup(redirection_string);
1878 #ifdef FEATURE_FAST_REDIRECTS
1879 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1881 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1884 * If it exists, use the previously rewritten URL as input
1885 * otherwise just use the old path.
1887 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1888 new_url = get_last_url(old_url, redirect_mode);
1893 * Disable redirect checkers, so that they
1894 * will be only run more than once if the user
1895 * also enables them through tags.
1897 * From a performance point of view
1898 * it doesn't matter, but the duplicated
1899 * log messages are annoying.
1901 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1902 #endif /* def FEATURE_FAST_REDIRECTS */
1903 csp->action->flags &= ~ACTION_REDIRECT;
1905 /* Did any redirect action trigger? */
1908 if (0 == strcmpic(new_url, csp->http->url))
1910 log_error(LOG_LEVEL_ERROR,
1911 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1912 csp->http->url, new_url);
1917 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1919 if (NULL == (rsp = alloc_http_response()))
1922 return cgi_error_memory();
1925 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1926 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1929 free_http_response(rsp);
1930 return cgi_error_memory();
1932 rsp->reason = RSP_REASON_REDIRECTED;
1935 return finish_http_response(csp, rsp);
1939 /* Only reached if no redirect is required */
1945 #ifdef FEATURE_IMAGE_BLOCKING
1946 /*********************************************************************
1948 * Function : is_imageurl
1950 * Description : Given a URL, decide whether it is an image or not,
1951 * using either the info from a previous +image action
1952 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1953 * is MSIE and not on a Mac, tell from the browser's accept
1957 * 1 : csp = Current client state (buffers, headers, etc...)
1959 * Returns : True (nonzero) if URL is an image, false (0)
1962 *********************************************************************/
1963 int is_imageurl(const struct client_state *csp)
1965 #ifdef FEATURE_IMAGE_DETECT_MSIE
1968 tmp = get_header_value(csp->headers, "User-Agent:");
1969 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1971 tmp = get_header_value(csp->headers, "Accept:");
1972 if (tmp && strstr(tmp, "image/gif"))
1974 /* Client will accept HTML. If this seems counterintuitive,
1984 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1986 return ((csp->action->flags & ACTION_IMAGE) != 0);
1989 #endif /* def FEATURE_IMAGE_BLOCKING */
1992 #ifdef FEATURE_TRUST
1993 /*********************************************************************
1995 * Function : is_untrusted_url
1997 * Description : Should we "distrust" this URL (and block it)?
1999 * Yes if it matches a line in the trustfile, or if the
2000 * referrer matches a line starting with "+" in the
2005 * 1 : csp = Current client state (buffers, headers, etc...)
2007 * Returns : 0 => trusted, 1 => untrusted
2009 *********************************************************************/
2010 int is_untrusted_url(const struct client_state *csp)
2012 struct file_list *fl;
2013 struct block_spec *b;
2014 struct url_spec **trusted_url;
2015 struct http_request rhttp[1];
2016 const char * referer;
2020 * If we don't have a trustlist, we trust everybody
2022 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
2027 memset(rhttp, '\0', sizeof(*rhttp));
2030 * Do we trust the request URL itself?
2032 for (b = b->next; b ; b = b->next)
2034 if (url_match(b->url, csp->http))
2040 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
2042 /* no referrer was supplied */
2048 * If not, do we maybe trust its referrer?
2050 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
2056 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
2058 if (url_match(*trusted_url, rhttp))
2060 /* if the URL's referrer is from a trusted referrer, then
2061 * add the target spec to the trustfile as an unblocked
2062 * domain and return 0 (which means it's OK).
2067 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
2071 char * new_entry = strdup("~");
2073 string_append(&new_entry, csp->http->hostport);
2075 path = csp->http->path;
2076 if ( (path[0] == '/')
2078 && ((path_end = strchr(path + 2, '/')) != NULL))
2080 /* since this path points into a user's home space
2081 * be sure to include this spec in the trustfile.
2083 long path_len = path_end - path; /* save offset */
2084 path = strdup(path); /* Copy string */
2087 path_end = path + path_len; /* regenerate ptr to new buffer */
2088 *(path_end + 1) = '\0'; /* Truncate path after '/' */
2090 string_join(&new_entry, path);
2094 * Give a reason for generating this entry.
2096 string_append(&new_entry, " # Trusted referrer was: ");
2097 string_append(&new_entry, referer);
2099 if (new_entry != NULL)
2101 if (-1 == fprintf(fp, "%s\n", new_entry))
2103 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
2104 new_entry, csp->config->trustfile);
2110 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
2111 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
2118 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
2119 csp->http->hostport, csp->config->trustfile);
2127 #endif /* def FEATURE_TRUST */
2130 /*********************************************************************
2132 * Function : pcrs_filter_response
2134 * Description : Execute all text substitutions from all applying
2135 * +filter actions on the text buffer that's been
2136 * accumulated in csp->iob->buf.
2139 * 1 : csp = Current client state (buffers, headers, etc...)
2141 * Returns : a pointer to the (newly allocated) modified buffer.
2142 * or NULL if there were no hits or something went wrong
2144 *********************************************************************/
2145 static char *pcrs_filter_response(struct client_state *csp)
2148 size_t size, prev_size;
2154 struct file_list *fl;
2155 struct re_filterfile_spec *b;
2156 struct list_entry *filtername;
2158 int i, found_filters = 0;
2163 if (csp->iob->cur >= csp->iob->eod)
2169 * Need to check the set of re_filterfiles...
2171 for (i = 0; i < MAX_AF_FILES; i++)
2184 if (0 == found_filters)
2186 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2187 "content filtering enabled, but no content filters available.");
2191 size = (size_t)(csp->iob->eod - csp->iob->cur);
2192 old = csp->iob->cur;
2194 for (i = 0; i < MAX_AF_FILES; i++)
2197 if ((NULL == fl) || (NULL == fl->f))
2200 * Either there are no filter files
2201 * left, or this filter file just
2202 * contains no valid filters.
2204 * Continue to be sure we don't miss
2205 * valid filter files that are chained
2206 * after empty or invalid ones.
2211 * For all applying +filter actions, look if a filter by that
2212 * name exists and if yes, execute it's pcrs_joblist on the
2215 for (b = fl->f; b; b = b->next)
2217 if (b->type != FT_CONTENT_FILTER)
2219 /* Skip header filters */
2223 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
2224 filtername ; filtername = filtername->next)
2226 if (strcmp(b->name, filtername->str) == 0)
2228 int current_hits = 0; /* Number of hits caused by this filter */
2229 int job_number = 0; /* Which job we're currently executing */
2230 int job_hits = 0; /* How many hits the current job caused */
2231 pcrs_job *joblist = b->joblist;
2233 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2235 if (NULL == joblist)
2237 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2242 /* Apply all jobs from the joblist */
2243 for (job = joblist; NULL != job; job = job->next)
2246 job_hits = pcrs_execute(job, old, size, &new, &size);
2251 * That went well. Continue filtering
2252 * and use the result of this job as
2253 * input for the next one.
2255 current_hits += job_hits;
2256 if (old != csp->iob->cur)
2265 * This job caused an unexpected error. Inform the user
2266 * and skip the rest of the jobs in this filter. We could
2267 * continue with the next job, but usually the jobs
2268 * depend on each other or are similar enough to
2269 * fail for the same reason.
2271 * At the moment our pcrs expects the error codes of pcre 3.4,
2272 * but newer pcre versions can return additional error codes.
2273 * As a result pcrs_strerror()'s error message might be
2274 * "Unknown error ...", therefore we print the numerical value
2277 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2278 * should we use LOG_LEVEL_RE_FILTER instead?
2280 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2281 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2286 if (b->dynamic) pcrs_free_joblist(joblist);
2288 log_error(LOG_LEVEL_RE_FILTER,
2289 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2290 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2292 hits += current_hits;
2299 * If there were no hits, destroy our copy and let
2300 * chat() use the original in csp->iob
2308 csp->flags |= CSP_FLAG_MODIFIED;
2309 csp->content_length = size;
2317 /*********************************************************************
2319 * Function : gif_deanimate_response
2321 * Description : Deanimate the GIF image that has been accumulated in
2322 * csp->iob->buf, set csp->content_length to the modified
2323 * size and raise the CSP_FLAG_MODIFIED flag.
2326 * 1 : csp = Current client state (buffers, headers, etc...)
2328 * Returns : a pointer to the (newly allocated) modified buffer.
2329 * or NULL in case something went wrong.
2331 *********************************************************************/
2332 static char *gif_deanimate_response(struct client_state *csp)
2334 struct binbuffer *in, *out;
2338 size = (size_t)(csp->iob->eod - csp->iob->cur);
2340 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2341 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2343 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2347 in->buffer = csp->iob->cur;
2350 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2352 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2359 if ((int)size == out->offset)
2361 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2365 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2367 csp->content_length = out->offset;
2368 csp->flags |= CSP_FLAG_MODIFIED;
2378 /*********************************************************************
2380 * Function : get_filter_function
2382 * Description : Decides which content filter function has
2383 * to be applied (if any).
2386 * 1 : csp = Current client state (buffers, headers, etc...)
2388 * Returns : The content filter function to run, or
2389 * NULL if no content filter is active
2391 *********************************************************************/
2392 filter_function_ptr get_filter_function(struct client_state *csp)
2394 filter_function_ptr filter_function = NULL;
2397 * Are we enabling text mode by force?
2399 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2402 * Do we really have to?
2404 if (csp->content_type & CT_TEXT)
2406 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2410 csp->content_type |= CT_TEXT;
2411 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2415 if (!(csp->content_type & CT_DECLARED))
2418 * The server didn't bother to declare a MIME-Type.
2419 * Assume it's text that can be filtered.
2421 * This also regulary happens with 304 responses,
2422 * therefore logging anything here would cause
2425 csp->content_type |= CT_TEXT;
2429 * Choose the applying filter function based on
2430 * the content type and action settings.
2432 if ((csp->content_type & CT_TEXT) &&
2433 (csp->rlist != NULL) &&
2434 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2436 filter_function = pcrs_filter_response;
2438 else if ((csp->content_type & CT_GIF) &&
2439 (csp->action->flags & ACTION_DEANIMATE))
2441 filter_function = gif_deanimate_response;
2444 return filter_function;
2448 /*********************************************************************
2450 * Function : remove_chunked_transfer_coding
2452 * Description : In-situ remove the "chunked" transfer coding as defined
2453 * in rfc2616 from a buffer.
2456 * 1 : buffer = Pointer to the text buffer
2457 * 2 : size = In: Number of bytes to be processed,
2458 * Out: Number of bytes after de-chunking.
2459 * (undefined in case of errors)
2461 * Returns : JB_ERR_OK for success,
2462 * JB_ERR_PARSE otherwise
2464 *********************************************************************/
2465 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2468 unsigned int chunksize = 0;
2469 char *from_p, *to_p;
2472 from_p = to_p = buffer;
2474 if (sscanf(buffer, "%x", &chunksize) != 1)
2476 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2477 return JB_ERR_PARSE;
2480 while (chunksize > 0U)
2482 if (NULL == (from_p = strstr(from_p, "\r\n")))
2484 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2485 return JB_ERR_PARSE;
2488 if ((newsize += chunksize) >= *size)
2490 log_error(LOG_LEVEL_ERROR,
2491 "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding",
2493 return JB_ERR_PARSE;
2497 memmove(to_p, from_p, (size_t) chunksize);
2498 to_p = buffer + newsize;
2499 from_p += chunksize + 2;
2501 if (sscanf(from_p, "%x", &chunksize) != 1)
2503 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2508 /* XXX: Should get its own loglevel. */
2509 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2518 /*********************************************************************
2520 * Function : prepare_for_filtering
2522 * Description : If necessary, de-chunks and decompresses
2523 * the content so it can get filterd.
2526 * 1 : csp = Current client state (buffers, headers, etc...)
2528 * Returns : JB_ERR_OK for success,
2529 * JB_ERR_PARSE otherwise
2531 *********************************************************************/
2532 static jb_err prepare_for_filtering(struct client_state *csp)
2534 jb_err err = JB_ERR_OK;
2537 * If the body has a "chunked" transfer-encoding,
2538 * get rid of it, adjusting size and iob->eod
2540 if (csp->flags & CSP_FLAG_CHUNKED)
2542 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2544 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2545 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2546 if (JB_ERR_OK == err)
2548 csp->iob->eod = csp->iob->cur + size;
2549 csp->flags |= CSP_FLAG_MODIFIED;
2553 return JB_ERR_PARSE;
2559 * If the body has a supported transfer-encoding,
2560 * decompress it, adjusting size and iob->eod.
2562 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2564 if (0 == csp->iob->eod - csp->iob->cur)
2566 /* Nothing left after de-chunking. */
2570 err = decompress_iob(csp);
2572 if (JB_ERR_OK == err)
2574 csp->flags |= CSP_FLAG_MODIFIED;
2575 csp->content_type &= ~CT_TABOO;
2580 * Unset CT_GZIP and CT_DEFLATE to remember not
2581 * to modify the Content-Encoding header later.
2583 csp->content_type &= ~CT_GZIP;
2584 csp->content_type &= ~CT_DEFLATE;
2593 /*********************************************************************
2595 * Function : execute_content_filter
2597 * Description : Executes a given content filter.
2600 * 1 : csp = Current client state (buffers, headers, etc...)
2601 * 2 : content_filter = The filter function to execute
2603 * Returns : Pointer to the modified buffer, or
2604 * NULL if filtering failed or wasn't necessary.
2606 *********************************************************************/
2607 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2609 if (0 == csp->iob->eod - csp->iob->cur)
2612 * No content (probably status code 301, 302 ...),
2613 * no filtering necessary.
2618 if (JB_ERR_OK != prepare_for_filtering(csp))
2621 * failed to de-chunk or decompress.
2626 if (0 == csp->iob->eod - csp->iob->cur)
2629 * Clown alarm: chunked and/or compressed nothing delivered.
2634 return ((*content_filter)(csp));
2638 /*********************************************************************
2640 * Function : get_url_actions
2642 * Description : Gets the actions for this URL.
2645 * 1 : csp = Current client state (buffers, headers, etc...)
2646 * 2 : http = http_request request for blocked URLs
2650 *********************************************************************/
2651 void get_url_actions(struct client_state *csp, struct http_request *http)
2653 struct file_list *fl;
2654 struct url_actions *b;
2657 init_current_action(csp->action);
2659 for (i = 0; i < MAX_AF_FILES; i++)
2661 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2666 apply_url_actions(csp->action, http, b);
2673 /*********************************************************************
2675 * Function : apply_url_actions
2677 * Description : Applies a list of URL actions.
2680 * 1 : action = Destination.
2681 * 2 : http = Current URL
2682 * 3 : b = list of URL actions to apply
2686 *********************************************************************/
2687 void apply_url_actions(struct current_action_spec *action,
2688 struct http_request *http,
2689 struct url_actions *b)
2693 /* Should never happen */
2697 for (b = b->next; NULL != b; b = b->next)
2699 if (url_match(b->url, http))
2701 merge_current_action(action, b->action);
2707 /*********************************************************************
2709 * Function : get_forward_override_settings
2711 * Description : Returns forward settings as specified with the
2712 * forward-override{} action. forward-override accepts
2713 * forward lines similar to the one used in the
2714 * configuration file, but without the URL pattern.
2720 * in the configuration file can be replaced with
2721 * the action section:
2723 * {+forward-override{forward .}}
2727 * 1 : csp = Current client state (buffers, headers, etc...)
2729 * Returns : Pointer to forwarding structure in case of success.
2730 * Invalid syntax is fatal.
2732 *********************************************************************/
2733 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2735 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2736 char forward_settings[BUFFER_SIZE];
2737 char *http_parent = NULL;
2738 /* variable names were chosen for consistency reasons. */
2739 struct forward_spec *fwd = NULL;
2743 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2744 /* Should be enforced by load_one_actions_file() */
2745 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2747 /* Create a copy ssplit can modify */
2748 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2750 if (NULL != csp->fwd)
2753 * XXX: Currently necessary to prevent memory
2754 * leaks when the show-url-info cgi page is visited.
2756 unload_forward_spec(csp->fwd);
2760 * allocate a new forward node, valid only for
2761 * the lifetime of this request. Save its location
2762 * in csp as well, so sweep() can free it later on.
2764 fwd = csp->fwd = zalloc(sizeof(*fwd));
2767 log_error(LOG_LEVEL_FATAL,
2768 "can't allocate memory for forward-override{%s}", forward_override_line);
2769 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2773 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2774 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2776 fwd->type = SOCKS_NONE;
2778 /* Parse the parent HTTP proxy host:port */
2779 http_parent = vec[1];
2782 else if (vec_count == 3)
2784 char *socks_proxy = NULL;
2786 if (!strcasecmp(vec[0], "forward-socks4"))
2788 fwd->type = SOCKS_4;
2789 socks_proxy = vec[1];
2791 else if (!strcasecmp(vec[0], "forward-socks4a"))
2793 fwd->type = SOCKS_4A;
2794 socks_proxy = vec[1];
2796 else if (!strcasecmp(vec[0], "forward-socks5"))
2798 fwd->type = SOCKS_5;
2799 socks_proxy = vec[1];
2802 if (NULL != socks_proxy)
2804 /* Parse the SOCKS proxy host[:port] */
2805 fwd->gateway_host = strdup(socks_proxy);
2807 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2809 *socks_proxy++ = '\0';
2810 fwd->gateway_port = (int)strtol(socks_proxy, NULL, 0);
2813 if (fwd->gateway_port <= 0)
2815 fwd->gateway_port = 1080;
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_host = strdup(http_parent);
2834 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2836 *http_parent++ = '\0';
2837 fwd->forward_port = (int)strtol(http_parent, NULL, 0);
2840 if (fwd->forward_port <= 0)
2842 fwd->forward_port = 8000;
2846 assert (NULL != fwd);
2848 log_error(LOG_LEVEL_CONNECT,
2849 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2855 /*********************************************************************
2857 * Function : forward_url
2859 * Description : Should we forward this to another proxy?
2862 * 1 : csp = Current client state (buffers, headers, etc...)
2863 * 2 : http = http_request request for current URL
2865 * Returns : Pointer to forwarding information.
2867 *********************************************************************/
2868 const struct forward_spec *forward_url(struct client_state *csp,
2869 const struct http_request *http)
2871 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2872 struct forward_spec *fwd = csp->config->forward;
2874 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2876 return get_forward_override_settings(csp);
2886 if (url_match(fwd->url, http))
2897 /*********************************************************************
2899 * Function : direct_response
2901 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2902 * request and if so, return a HTTP 501 to the client.
2904 * FIXME: I have a stupid name and I should handle the
2905 * requests properly. Still, what we do here is rfc-
2906 * compliant, whereas ignoring or forwarding are not.
2909 * 1 : csp = Current client state (buffers, headers, etc...)
2911 * Returns : http_response if , NULL if nonmatch or handler fail
2913 *********************************************************************/
2914 struct http_response *direct_response(struct client_state *csp)
2916 struct http_response *rsp;
2917 struct list_entry *p;
2919 if ((0 == strcmpic(csp->http->gpc, "trace"))
2920 || (0 == strcmpic(csp->http->gpc, "options")))
2922 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2924 if (!strncmpic("Max-Forwards:", p->str, 13))
2926 unsigned int max_forwards;
2929 * If it's a Max-Forwards value of zero,
2930 * we have to intercept the request.
2932 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2935 * FIXME: We could handle at least TRACE here,
2936 * but that would require a verbatim copy of
2937 * the request which we don't have anymore
2939 log_error(LOG_LEVEL_HEADER,
2940 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2943 /* Get mem for response or fail*/
2944 if (NULL == (rsp = alloc_http_response()))
2946 return cgi_error_memory();
2949 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2951 free_http_response(rsp);
2952 return cgi_error_memory();
2956 rsp->reason = RSP_REASON_UNSUPPORTED;
2958 return(finish_http_response(csp, rsp));
2967 /*********************************************************************
2969 * Function : content_filters_enabled
2971 * Description : Checks whether there are any content filters
2972 * enabled for the current request.
2975 * 1 : action = Action spec to check.
2977 * Returns : TRUE for yes, FALSE otherwise
2979 *********************************************************************/
2980 int content_filters_enabled(const struct current_action_spec *action)
2982 return ((action->flags & ACTION_DEANIMATE) ||
2983 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));