1 const char filters_rcs[] = "$Id: filters.c,v 1.118 2009/04/17 11:37:03 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 1.118 2009/04/17 11:37:03 fabiankeil
44 * Allow IPv6 addresses in forward-override{}.
46 * Revision 1.117 2009/04/17 11:35:28 fabiankeil
47 * Fix compiler warning.
49 * Revision 1.116 2009/04/17 11:34:34 fabiankeil
50 * Style cosmetics for the IPv6 code.
52 * Revision 1.115 2009/04/17 11:29:18 fabiankeil
53 * Compile fix for BSD libc.
55 * Revision 1.114 2009/04/17 11:27:49 fabiankeil
56 * Petr Pisar's privoxy-3.0.12-ipv6-3.diff.
58 * Revision 1.113 2009/03/08 14:19:23 fabiankeil
59 * Fix justified (but harmless) compiler warnings
60 * on platforms where sizeof(int) < sizeof(long).
62 * Revision 1.112 2009/03/01 18:28:23 fabiankeil
63 * Help clang understand that we aren't dereferencing
66 * Revision 1.111 2008/12/04 18:13:46 fabiankeil
67 * Fix a cparser warning.
69 * Revision 1.110 2008/11/10 16:40:25 fabiankeil
70 * Fix a gcc44 warning.
72 * Revision 1.109 2008/11/08 15:48:41 fabiankeil
73 * Mention actual values when complaining about
74 * the chunk size exceeding the buffer size.
76 * Revision 1.108 2008/05/21 15:35:08 fabiankeil
77 * - Mark csp as immutable for block_acl().
78 * - Remove an obsolete complaint about filter_popups().
80 * Revision 1.107 2008/05/04 17:52:56 fabiankeil
81 * Adjust parse_http_url() call to new prototype.
83 * Revision 1.106 2008/05/03 16:40:44 fabiankeil
84 * Change content_filters_enabled()'s parameter from
85 * csp->action to action so it can be also used in the
86 * CGI code. Don't bother checking if there are filters
87 * loaded, as that's somewhat besides the point.
89 * Revision 1.105 2008/03/28 15:13:39 fabiankeil
90 * Remove inspect-jpegs action.
92 * Revision 1.104 2008/03/27 18:27:24 fabiankeil
93 * Remove kill-popups action.
95 * Revision 1.103 2008/03/06 16:33:45 fabiankeil
96 * If limit-connect isn't used, don't limit CONNECT requests to port 443.
98 * Revision 1.102 2008/03/01 14:00:44 fabiankeil
99 * Let the block action take the reason for the block
100 * as argument and show it on the "blocked" page.
102 * Revision 1.101 2008/02/23 16:57:12 fabiankeil
103 * Rename url_actions() to get_url_actions() and let it
104 * use the standard parameter ordering.
106 * Revision 1.100 2008/02/23 16:33:43 fabiankeil
107 * Let forward_url() use the standard parameter ordering
108 * and mark its second parameter immutable.
110 * Revision 1.99 2008/02/03 13:57:58 fabiankeil
111 * Add SOCKS5 support for forward-override{}.
113 * Revision 1.98 2008/01/04 17:43:45 fabiankeil
114 * Improve the warning messages that get logged if the action files
115 * "enable" filters but no filters of that type have been loaded.
117 * Revision 1.97 2007/11/30 15:37:03 fabiankeil
118 * Use freez instead of free.
120 * Revision 1.96 2007/10/19 16:53:28 fabiankeil
121 * Add helper function to check if any content filters are enabled.
123 * Revision 1.95 2007/10/17 19:31:20 fabiankeil
124 * Omitting the zero chunk that ends the chunk transfer encoding seems
125 * to be the new black. Log the problem and continue filtering anyway.
127 * Revision 1.94 2007/09/29 13:20:20 fabiankeil
128 * Remove two redundant and one useless log messages.
130 * Revision 1.93 2007/09/29 10:21:16 fabiankeil
131 * - Move get_filter_function() from jcc.c to filters.c
132 * so the filter functions can be static.
133 * - Don't bother filtering body-less responses.
135 * Revision 1.92 2007/09/28 16:38:55 fabiankeil
136 * - Execute content filters through execute_content_filter().
137 * - Add prepare_for_filtering() so filter functions don't have to
138 * care about de-chunking and decompression. As a side effect this enables
139 * decompression for gif_deanimate_response() and jpeg_inspect_response().
140 * - Change remove_chunked_transfer_coding()'s return type to jb_err.
141 * Some clowns feel like chunking empty responses in which case
142 * (size == 0) is valid but previously would be interpreted as error.
144 * Revision 1.91 2007/09/02 15:31:20 fabiankeil
145 * Move match_portlist() from filter.c to urlmatch.c.
146 * It's used for url matching, not for filtering.
148 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
149 * Remove newline at the end of a log_error() message.
151 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
152 * #1763173 from Stefan Huehner: declare some more functions static.
154 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
155 * Add forward-override{} to change the forwarding settings through
156 * action sections. This is mainly interesting to forward different
157 * clients differently (for example based on User-Agent or request
160 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
161 * Make sure filters with dynamic jobs actually use them.
163 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
164 * - Introduce dynamic pcrs jobs that can resolve variables.
165 * - Don't run redirect functions more than once,
166 * unless they are activated more than once.
168 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
169 * - Log the content size after decompression in decompress_iob()
170 * instead of pcrs_filter_response().
172 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
173 * Use dedicated header filter actions instead of abusing "filter".
174 * Replace "filter-client-headers" and "filter-client-headers"
175 * with "server-header-filter" and "client-header-filter".
177 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
178 * New config option: enforce-blocks.
180 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
181 * - Fix port handling in acl_addr() and use a temporary acl spec
182 * copy so error messages don't contain a truncated version.
183 * - Log size of iob before and after decompression.
185 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
186 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
187 * - Hide the "Go there anyway" link for blocked CONNECT
188 * requests where going there anyway doesn't work anyway.
190 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
191 * - Save the reason for generating http_responses.
192 * - Block (+block) with status code 403 instead of 404.
193 * - Use a different kludge to remember a failed decompression.
195 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
196 * Search for Max-Forwards headers case-insensitive,
197 * don't generate the "501 unsupported" message for invalid
198 * Max-Forwards values and don't increase negative ones.
200 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
201 * - Add HEAD support to finish_http_response.
202 * - Add error favicon to internal HTML error messages.
204 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
205 * Mark *csp as immutable for is_untrusted_url()
206 * and is_imageurl(). Closes FR 1237736.
208 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
209 * Integrate a modified version of Wil Mahan's
210 * zlib patch (PR #895531).
212 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
213 * Fixed gcc43 conversion warnings,
214 * changed sprintf calls to snprintf.
216 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
217 * Adjust comment in pcrs_filter_response()
218 * to recent pcrs changes. Hohoho.
220 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
221 * Don't crash if pcre returns an error code
222 * that pcrs didn't expect. Fixes BR 1621173.
224 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
225 * Modified is_untrusted_url to complain in case of
226 * write errors and to give a reason when adding new
227 * entries to the trustfile. Closes FR 1097611.
229 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
230 * Skip empty filter files in pcrs_filter_response,
231 * but don't ignore the ones that come afterwards.
232 * Fixes parts of BR 1619208.
234 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
235 * Added some sanity checks for get_last_url().
236 * Fixed possible segfault caused by my last commit.
238 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
239 * Let get_last_url() catch https URLs as well.
241 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
242 * Make sure get_last_url() behaves like advertised
243 * and fast-redirects{} can be combined with redirect{}.
245 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
246 * Implemented +redirect{s@foo@bar@} to generate
247 * a redirect based on a rewritten version of the
250 * Revision 1.66 2006/09/23 13:26:38 roro
251 * Replace TABs by spaces in source code.
253 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
254 * Fix +redirect{}. Didn't work with -fast-redirects.
256 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
257 * Block requests for untrusted URLs with status
258 * code 403 instead of 200.
260 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
261 * Don't free p which is still in use and will be later
262 * freed by free_map(). Don't claim the referrer is unknown
263 * when the client didn't set one.
265 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
266 * Feature request 595948: Re-Filter logging in single line
268 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
269 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
271 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
272 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
273 * with what was really the latest development (the v_3_0_branch branch)
275 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
276 * Multiple filter file support
278 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
279 * A few memory leaks plugged (mostly on error paths)
281 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
282 * Add the ability to check jpeg images for invalid
283 * lengths of comment blocks. Defensive strategy
284 * against the exploit:
285 * Microsoft Security Bulletin MS04-028
286 * Buffer Overrun in JPEG Processing (GDI+) Could
287 * Allow Code Execution (833987)
288 * Enabled with +inspect-jpegs in actions files.
290 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
291 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
293 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
294 * Fixed bug #839859: "See why" link URL now gets url-encoded.
296 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
299 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
300 * Added basic support for OPTIONS and TRACE HTTP methods:
301 * New function direct_response which handles OPTIONS and
302 * TRACE requests whose Max-Forwards header field is zero.
304 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
305 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
307 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
308 * - Bugfix: Executing a filters without jobs no longer results in
309 * turing off *all* filters.
310 * - Security fix: Malicious web servers can't cause a seg fault
311 * through bogus chunk sizes anymore
313 * Revision 1.58 2002/04/24 02:11:17 oes
314 * Jon's multiple AF patch: url_actions now evaluates rules
317 * Revision 1.57 2002/04/08 20:38:34 swa
320 * Revision 1.56 2002/04/05 15:51:24 oes
321 * - bugfix: error-pages now get correct request protocol
322 * - fix for invalid HTML in trust info
324 * Revision 1.55 2002/04/02 16:13:51 oes
325 * Fix: No "Go there anyway" for SSL
327 * Revision 1.54 2002/04/02 14:55:56 oes
328 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
330 * Revision 1.53 2002/03/26 22:29:54 swa
331 * we have a new homepage!
333 * Revision 1.52 2002/03/24 16:35:57 jongfoster
336 * Revision 1.51 2002/03/24 15:23:33 jongfoster
339 * Revision 1.50 2002/03/24 13:25:43 swa
340 * name change related issues
342 * Revision 1.49 2002/03/16 20:29:14 oes
345 * Revision 1.48 2002/03/13 20:25:34 oes
346 * Better logging for content filters
348 * Revision 1.47 2002/03/13 00:30:52 jongfoster
350 * Added option of always sending redirect for imageblock,
351 * currently disabled with #if 0.
353 * Revision 1.46 2002/03/12 01:42:49 oes
354 * Introduced modular filters
356 * Revision 1.45 2002/03/08 16:47:50 oes
357 * Added choice beween GIF and PNG built-in images
359 * Revision 1.44 2002/03/07 03:49:31 oes
360 * - Fixed compiler warnings etc
361 * - Changed built-in images from GIF to PNG
362 * (with regard to Unisys patent issue)
363 * - Added a 4x4 pattern PNG which is less intrusive
364 * than the logo but also clearly marks the deleted banners
366 * Revision 1.43 2002/01/22 23:51:59 jongfoster
367 * Replacing strsav() with the safer string_append().
369 * Adding missing html_encode() to error message generators. Where encoded
370 * and unencoded versions of a string were provided, removing the unencoded
373 * Revision 1.42 2002/01/17 21:00:32 jongfoster
374 * Moving all our URL and URL pattern parsing code to urlmatch.c.
376 * Using a single, simple url_match(pattern,url) function - rather than
377 * the 3-line match routine which was repeated all over the place.
379 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
381 * Using parse_http_url() to parse URLs without faking a HTTP
382 * request line for parse_http_request().
384 * Revision 1.41 2001/11/13 00:14:07 jongfoster
385 * Fixing stupid bug now I've figured out what || means.
386 * (It always returns 0 or 1, not one of it's paramaters.)
388 * Revision 1.40 2001/10/26 17:37:55 oes
389 * - Re-enabled Netscape 200/404 bug workaround in block_url():
390 * - Removed OS/2 special case
391 * - Made block_url() independant from sed() having been run
392 * - Made trust_url independant from sed() having been run
393 * - Made is_imageurl independant from sed() having been run.
394 * It now checks User-Agent: and Accept: by itself.
397 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
398 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
399 * threads to call select() simultaneously. So, it's time to do a real, live,
400 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
401 * (native). Both versions will work, but using __OS2__ offers multi-threading.
403 * Revision 1.38 2001/10/23 21:32:33 jongfoster
404 * Adding error-checking to selected functions
406 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
407 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
408 * filters.c. Added a FIXME in front of the offending code. I'll gladly
409 * put in a better/more robust fix for all parties if one is presented...
410 * It seems that just returning 200 instead of 404 would pretty much fix
411 * it for everyone, but I don't know all the history of the problem.
413 * Revision 1.36 2001/10/10 16:44:16 oes
414 * Added match_portlist function
416 * Revision 1.35 2001/10/07 15:41:23 oes
417 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
419 * New function remove_chunked_transfer_coding that strips chunked
420 * transfer coding to plain and is called by pcrs_filter_response
421 * and gif_deanimate_response if neccessary
423 * Improved handling of zero-change re_filter runs
425 * pcrs_filter_response and gif_deanimate_response now remove
426 * chunked transfer codeing before processing the body.
428 * Revision 1.34 2001/09/20 15:49:36 steudten
430 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
431 * See cgi.c fill_template().
433 * Revision 1.33 2001/09/16 17:05:14 jongfoster
434 * Removing unused #include showarg.h
436 * Revision 1.32 2001/09/16 13:21:27 jongfoster
437 * Changes to use new list functions.
439 * Revision 1.31 2001/09/16 11:38:02 jongfoster
440 * Splitting fill_template() into 2 functions:
441 * template_load() loads the file
442 * template_fill() performs the PCRS regexps.
443 * This is because the CGI edit interface has a "table row"
444 * template which is used many times in the page - this
445 * change means it's only loaded from disk once.
447 * Revision 1.30 2001/09/16 11:00:10 jongfoster
448 * New function alloc_http_response, for symmetry with free_http_response
450 * Revision 1.29 2001/09/13 23:32:40 jongfoster
451 * Moving image data to cgi.c rather than cgi.h
452 * Fixing a GPF under Win32 (and any other OS that protects global
453 * constants from being written to).
455 * Revision 1.28 2001/09/10 10:18:51 oes
456 * Silenced compiler warnings
458 * Revision 1.27 2001/08/05 16:06:20 jongfoster
459 * Modifiying "struct map" so that there are now separate header and
460 * "map_entry" structures. This means that functions which modify a
461 * map no longer need to return a pointer to the modified map.
462 * Also, it no longer reverses the order of the entries (which may be
463 * important with some advanced template substitutions).
465 * Revision 1.26 2001/07/30 22:08:36 jongfoster
466 * Tidying up #defines:
467 * - All feature #defines are now of the form FEATURE_xxx
468 * - Permanently turned off WIN_GUI_EDIT
469 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
471 * Revision 1.25 2001/07/26 10:09:46 oes
472 * Made browser detection a little less naive
474 * Revision 1.24 2001/07/25 17:22:51 oes
475 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
477 * Revision 1.23 2001/07/23 13:40:12 oes
478 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
480 * Revision 1.22 2001/07/18 12:29:34 oes
481 * - Made gif_deanimate_response respect
482 * csp->action->string[ACTION_STRING_DEANIMATE]
483 * - Logging cosmetics
485 * Revision 1.21 2001/07/13 13:59:53 oes
486 * - Introduced gif_deanimate_response which shares the
487 * generic content modification interface of pcrs_filter_response
488 * and acts as a wrapper to deanimate.c:gif_deanimate()
489 * - Renamed re_process_buffer to pcrs_filter_response
490 * - pcrs_filter_response now returns NULL on failiure
491 * - Removed all #ifdef PCRS
493 * Revision 1.20 2001/07/01 17:01:04 oes
494 * Added comments and missing return statement in is_untrusted_url()
496 * Revision 1.19 2001/06/29 21:45:41 oes
497 * Indentation, CRLF->LF, Tab-> Space
499 * Revision 1.18 2001/06/29 13:27:38 oes
500 * - Cleaned up, renamed and reorderd functions
501 * and improved comments
504 * - Ported to CGI platform. Now delivers
505 * http_response or NULL
506 * - Unified HTML and GIF generation (moved image detection
507 * and GIF generation here from jcc.c:chat())
508 * - Fixed HTTP status to:
509 * - 403 (Forbidden) for the "blocked" HTML message
510 * - 200 (OK) for GIF answers
511 * - 302 (Redirect) for redirect to GIF
514 * - Ported to CGI platform. Now delivers
515 * http_response or NULL
516 * - Separated detection of untrusted URL into
517 * (bool)is_untrusted_url
518 * - Added enforcement of untrusted requests
520 * - Moved redirect_url() from cgi.c to here
521 * and ported it to the CGI platform
523 * - Removed logentry from cancelled commit
525 * Revision 1.17 2001/06/09 10:55:28 jongfoster
526 * Changing BUFSIZ ==> BUFFER_SIZE
528 * Revision 1.16 2001/06/07 23:10:26 jongfoster
529 * Allowing unanchored domain patterns to back off and retry
530 * if they partially match. Optimized right-anchored patterns.
531 * Moving ACL and forward files into config file.
532 * Replacing struct gateway with struct forward_spec
534 * Revision 1.15 2001/06/03 19:12:00 oes
535 * extracted-CGI relevant stuff
537 * Revision 1.14 2001/06/01 10:30:55 oes
538 * Added optional left-anchoring to domaincmp
540 * Revision 1.13 2001/05/31 21:21:30 jongfoster
541 * Permissionsfile / actions file changes:
542 * - Changed "permission" to "action" throughout
543 * - changes to file format to allow string parameters
544 * - Moved helper functions to actions.c
546 * Revision 1.12 2001/05/31 17:35:20 oes
548 * - Enhanced domain part globbing with infix and prefix asterisk
549 * matching and optional unanchored operation
551 * Revision 1.11 2001/05/29 11:53:23 oes
552 * "See why" link added to "blocked" page
554 * Revision 1.10 2001/05/29 09:50:24 jongfoster
555 * Unified blocklist/imagelist/permissionslist.
556 * File format is still under discussion, but the internal changes
559 * Also modified interceptor behaviour:
560 * - We now intercept all URLs beginning with one of the following
561 * prefixes (and *only* these prefixes):
563 * * http://ijbswa.sf.net/config/
564 * * http://ijbswa.sourceforge.net/config/
565 * - New interceptors "home page" - go to http://i.j.b/ to see it.
566 * - Internal changes so that intercepted and fast redirect pages
567 * are not replaced with an image.
568 * - Interceptors now have the option to send a binary page direct
569 * to the client. (i.e. ijb-send-banner uses this)
570 * - Implemented show-url-info interceptor. (Which is why I needed
571 * the above interceptors changes - a typical URL is
572 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
573 * The previous mechanism would not have intercepted that, and
574 * if it had been intercepted then it then it would have replaced
577 * Revision 1.9 2001/05/27 22:17:04 oes
579 * - re_process_buffer no longer writes the modified buffer
580 * to the client, which was very ugly. It now returns the
581 * buffer, which it is then written by chat.
583 * - content_length now adjusts the Content-Length: header
584 * for modified documents rather than crunch()ing it.
585 * (Length info in csp->content_length, which is 0 for
586 * unmodified documents)
588 * - For this to work, sed() is called twice when filtering.
590 * Revision 1.8 2001/05/26 17:13:28 jongfoster
591 * Filled in a function comment.
593 * Revision 1.7 2001/05/26 15:26:15 jongfoster
594 * ACL feature now provides more security by immediately dropping
595 * connections from untrusted hosts.
597 * Revision 1.6 2001/05/26 00:28:36 jongfoster
598 * Automatic reloading of config file.
599 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
600 * Most of the global variables have been moved to a new
601 * struct configuration_spec, accessed through csp->config->globalname
602 * Most of the globals remaining are used by the Win32 GUI.
604 * Revision 1.5 2001/05/25 22:34:30 jongfoster
607 * Revision 1.4 2001/05/22 18:46:04 oes
609 * - Enabled filtering banners by size rather than URL
610 * by adding patterns that replace all standard banner
611 * sizes with the "Junkbuster" gif to the re_filterfile
613 * - Enabled filtering WebBugs by providing a pattern
614 * which kills all 1x1 images
616 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
617 * which is selected by the (nonstandard and therefore
618 * capital) letter 'U' in the option string.
619 * It causes the quantifiers to be ungreedy by default.
620 * Appending a ? turns back to greedy (!).
622 * - Added a new interceptor ijb-send-banner, which
623 * sends back the "Junkbuster" gif. Without imagelist or
624 * MSIE detection support, or if tinygif = 1, or the
625 * URL isn't recognized as an imageurl, a lame HTML
626 * explanation is sent instead.
628 * - Added new feature, which permits blocking remote
629 * script redirects and firing back a local redirect
631 * The feature is conditionally compiled, i.e. it
632 * can be disabled with --disable-fast-redirects,
633 * plus it must be activated by a "fast-redirects"
634 * line in the config file, has its own log level
635 * and of course wants to be displayed by show-proxy-args
636 * Note: Boy, all the #ifdefs in 1001 locations and
637 * all the fumbling with configure.in and acconfig.h
638 * were *way* more work than the feature itself :-(
640 * - Because a generic redirect template was needed for
641 * this, tinygif = 3 now uses the same.
643 * - Moved GIFs, and other static HTTP response templates
648 * - Removed some >400 CRs again (Jon, you really worked
651 * Revision 1.3 2001/05/20 16:44:47 jongfoster
652 * Removing last hardcoded Junkbusters.com URLs.
654 * Revision 1.2 2001/05/20 01:21:20 jongfoster
655 * Version 2.9.4 checkin.
656 * - Merged popupfile and cookiefile, and added control over PCRS
657 * filtering, in new "permissionsfile".
658 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
659 * file error you now get a message box (in the Win32 GUI) rather
660 * than the program exiting with no explanation.
661 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
663 * - Removed tabs from "config"
664 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
665 * - Bumped up version number.
667 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
668 * Initial import of version 2.9.3 source tree
671 *********************************************************************/
677 #include <sys/types.h>
683 #ifdef HAVE_GETADDRINFO
685 #include <sys/socket.h>
686 #endif /* def HAVE_GETADDRINFO */
691 #endif /* ndef __OS2__ */
692 #include <netinet/in.h>
694 #include <winsock2.h>
695 #endif /* ndef _WIN32 */
699 #endif /* def __OS2__ */
707 #include "jbsockets.h"
708 #include "miscutil.h"
712 #include "deanimate.h"
713 #include "urlmatch.h"
720 const char filters_h_rcs[] = FILTERS_H_VERSION;
722 /* Fix a problem with Solaris. There should be no effect on other
724 * Solaris's isspace() is a macro which uses it's argument directly
725 * as an array index. Therefore we need to make sure that high-bit
726 * characters generate +ve values, and ideally we also want to make
727 * the argument match the declared parameter type of "int".
729 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
731 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
732 static jb_err prepare_for_filtering(struct client_state *csp);
735 #ifdef HAVE_GETADDRINFO
736 /*********************************************************************
738 * Function : sockaddr_storage_to_ip
740 * Description : Access internal structure of sockaddr_storage
743 * 1 : addr = socket address
744 * 2 : ip = IP address as array of octets in network order
745 * (it points into addr)
746 * 3 : len = length of IP address in octets
747 * 4 : port = port number in network order;
749 * Returns : 0 = no errror; -1 otherwise.
751 *********************************************************************/
752 int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, uint8_t **ip,
753 unsigned int *len, in_port_t **port)
760 switch (addr->ss_family)
770 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
774 *port = &((struct sockaddr_in *)addr)->sin_port;
785 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
789 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
794 /* Unsupported address family */
802 /*********************************************************************
804 * Function : match_sockaddr
806 * Description : Check whether address matches network (IP address and port)
809 * 1 : network = socket address of subnework
810 * 2 : netmask = network mask as socket address
811 * 3 : address = checked socket address against given network
813 * Returns : 0 = doesn't match; 1 = does match
815 *********************************************************************/
816 int match_sockaddr(const struct sockaddr_storage *network,
817 const struct sockaddr_storage *netmask,
818 const struct sockaddr_storage *address)
820 uint8_t *network_addr, *netmask_addr, *address_addr;
821 unsigned int addr_len;
822 in_port_t *network_port, *netmask_port, *address_port;
825 if (network->ss_family != netmask->ss_family)
827 /* This should never happen */
828 log_error(LOG_LEVEL_ERROR,
829 "Internal error at %s:%llu: network and netmask differ in family",
834 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
835 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
836 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
838 /* Check for family */
839 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
840 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
842 /* Map AF_INET6 V4MAPPED address into AF_INET */
846 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
847 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
849 /* Map AF_INET6 V4MAPPED network into AF_INET */
854 else if (network->ss_family != address->ss_family)
859 /* XXX: Port check is signaled in netmask */
860 if (*netmask_port && *network_port != *address_port)
865 /* TODO: Optimize by checking by words insted of octets */
866 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
868 if ((network_addr[i] & netmask_addr[i]) !=
869 (address_addr[i] & netmask_addr[i]))
877 #endif /* def HAVE_GETADDRINFO */
880 /*********************************************************************
882 * Function : block_acl
884 * Description : Block this request?
885 * Decide yes or no based on ACL file.
888 * 1 : dst = The proxy or gateway address this is going to.
889 * Or NULL to check all possible targets.
890 * 2 : csp = Current client state (buffers, headers, etc...)
891 * Also includes the client IP address.
893 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
895 *********************************************************************/
896 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
898 struct access_control_list *acl = csp->config->acl;
900 /* if not using an access control list, then permit the connection */
906 /* search the list */
910 #ifdef HAVE_GETADDRINFO
911 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
913 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
919 /* Just want to check if they have any access */
920 if (acl->action == ACL_PERMIT)
926 #ifdef HAVE_GETADDRINFO
928 * XXX: An undefined acl->dst is full of zeros and should be
929 * considered a wildcard address. sockaddr_storage_to_ip()
930 * fails on such destinations because of unknown sa_familly
931 * (glibc only?). However this test is not portable.
933 * So, we signal the acl->dst is wildcard in wildcard_dst.
936 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
938 ((dst->addr & acl->dst->mask) == acl->dst->addr)
939 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
943 if (acl->action == ACL_PERMIT)
961 /*********************************************************************
963 * Function : acl_addr
965 * Description : Called from `load_config' to parse an ACL address.
968 * 1 : aspec = String specifying ACL address.
969 * 2 : aca = struct access_control_addr to fill in.
971 * Returns : 0 => Ok, everything else is an error.
973 *********************************************************************/
974 int acl_addr(const char *aspec, struct access_control_addr *aca)
977 #ifdef HAVE_GETADDRINFO
978 struct addrinfo hints, *result;
980 in_port_t *mask_port;
981 unsigned int addr_len;
984 #endif /* def HAVE_GETADDRINFO */
986 char *acl_spec = NULL;
988 #ifdef HAVE_GETADDRINFO
989 /* XXX: Depend on ai_family */
997 * Use a temporary acl spec copy so we can log
998 * the unmodified original in case of parse errors.
1000 acl_spec = strdup(aspec);
1001 if (acl_spec == NULL)
1003 /* XXX: This will be logged as parse error. */
1007 if ((p = strchr(acl_spec, '/')) != NULL)
1010 if (ijb_isdigit(*p) == 0)
1015 masklength = atoi(p);
1018 if ((masklength < 0) ||
1019 #ifdef HAVE_GETADDRINFO
1030 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
1033 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
1042 p = strchr(acl_spec, ':');
1045 #ifdef HAVE_GETADDRINFO
1046 memset(&hints, 0, sizeof(struct addrinfo));
1047 hints.ai_family = AF_UNSPEC;
1048 hints.ai_socktype = SOCK_STREAM;
1050 i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result);
1055 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
1056 acl_spec, p, gai_strerror(i));
1060 /* TODO: Allow multihomed hostnames */
1061 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
1062 freeaddrinfo(result);
1069 port = strtol(p, &endptr, 10);
1071 if (port <= 0 || port > 65535 || *endptr != '\0')
1078 aca->port = (unsigned long)port;
1080 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
1083 if (aca->addr == INADDR_NONE)
1085 /* XXX: This will be logged as parse error. */
1088 #endif /* def HAVE_GETADDRINFO */
1090 /* build the netmask */
1091 #ifdef HAVE_GETADDRINFO
1092 /* Clip masklength according to current family. */
1093 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
1098 aca->mask.ss_family = aca->addr.ss_family;
1099 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
1106 /* ACL contains a port number, check ports in the future. */
1111 * XXX: This could be optimized to operate on whole words instead
1112 * of octets (128-bit CPU could do it in one iteration).
1115 * Octets after prefix can be ommitted because of
1116 * previous initialization to zeros.
1118 for (i = 0; (i < addr_len) && masklength; i++)
1120 if (masklength >= 8)
1122 mask_data[i] = 0xFF;
1128 * XXX: This assumes MSB of octet is on the left side.
1129 * This should be true for all architectures or solved
1130 * by the link layer.
1132 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
1139 for (i=1; i <= masklength ; i++)
1141 aca->mask |= (1U << (32 - i));
1144 /* now mask off the host portion of the ip address
1145 * (i.e. save on the network portion of the address).
1147 aca->addr = aca->addr & aca->mask;
1148 #endif /* def HAVE_GETADDRINFO */
1153 #endif /* def FEATURE_ACL */
1156 /*********************************************************************
1158 * Function : connect_port_is_forbidden
1160 * Description : Check to see if CONNECT requests to the destination
1161 * port of this request are forbidden. The check is
1162 * independend of the actual request method.
1165 * 1 : csp = Current client state (buffers, headers, etc...)
1167 * Returns : True if yes, false otherwise.
1169 *********************************************************************/
1170 int connect_port_is_forbidden(const struct client_state *csp)
1172 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
1173 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
1178 /*********************************************************************
1180 * Function : block_url
1182 * Description : Called from `chat'. Check to see if we need to block this.
1185 * 1 : csp = Current client state (buffers, headers, etc...)
1187 * Returns : NULL => unblocked, else HTTP block response
1189 *********************************************************************/
1190 struct http_response *block_url(struct client_state *csp)
1192 struct http_response *rsp;
1193 const char *new_content_type = NULL;
1196 * If it's not blocked, don't block it ;-)
1198 if ((csp->action->flags & ACTION_BLOCK) == 0)
1202 if (csp->action->flags & ACTION_REDIRECT)
1204 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
1207 * Else, prepare a response
1209 if (NULL == (rsp = alloc_http_response()))
1211 return cgi_error_memory();
1215 * If it's an image-url, send back an image or redirect
1216 * as specified by the relevant +image action
1218 #ifdef FEATURE_IMAGE_BLOCKING
1219 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
1220 && is_imageurl(csp))
1223 /* determine HOW images should be blocked */
1224 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
1226 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1228 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
1230 #if 1 /* Two alternative strategies, use this one for now: */
1232 /* and handle accordingly: */
1233 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1235 rsp->status = strdup("403 Request blocked by Privoxy");
1236 if (rsp->status == NULL)
1238 free_http_response(rsp);
1239 return cgi_error_memory();
1241 rsp->body = bindup(image_pattern_data, image_pattern_length);
1242 if (rsp->body == NULL)
1244 free_http_response(rsp);
1245 return cgi_error_memory();
1247 rsp->content_length = image_pattern_length;
1249 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1251 free_http_response(rsp);
1252 return cgi_error_memory();
1256 else if (0 == strcmpic(p, "blank"))
1258 rsp->status = strdup("403 Request blocked by Privoxy");
1259 if (rsp->status == NULL)
1261 free_http_response(rsp);
1262 return cgi_error_memory();
1264 rsp->body = bindup(image_blank_data, image_blank_length);
1265 if (rsp->body == NULL)
1267 free_http_response(rsp);
1268 return cgi_error_memory();
1270 rsp->content_length = image_blank_length;
1272 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
1274 free_http_response(rsp);
1275 return cgi_error_memory();
1281 rsp->status = strdup("302 Local Redirect from Privoxy");
1282 if (rsp->status == NULL)
1284 free_http_response(rsp);
1285 return cgi_error_memory();
1288 if (enlist_unique_header(rsp->headers, "Location", p))
1290 free_http_response(rsp);
1291 return cgi_error_memory();
1295 #else /* Following code is disabled for now */
1297 /* and handle accordingly: */
1298 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
1300 p = CGI_PREFIX "send-banner?type=pattern";
1302 else if (0 == strcmpic(p, "blank"))
1304 p = CGI_PREFIX "send-banner?type=blank";
1306 rsp->status = strdup("302 Local Redirect from Privoxy");
1307 if (rsp->status == NULL)
1309 free_http_response(rsp);
1310 return cgi_error_memory();
1313 if (enlist_unique_header(rsp->headers, "Location", p))
1315 free_http_response(rsp);
1316 return cgi_error_memory();
1318 #endif /* Preceeding code is disabled for now */
1320 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
1323 * Send empty document.
1325 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1328 rsp->body = strdup(" ");
1329 rsp->content_length = 1;
1331 rsp->status = strdup("403 Request blocked by Privoxy");
1332 if (rsp->status == NULL)
1334 free_http_response(rsp);
1335 return cgi_error_memory();
1337 if (new_content_type != 0)
1339 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1340 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1342 free_http_response(rsp);
1343 return cgi_error_memory();
1348 #endif /* def FEATURE_IMAGE_BLOCKING */
1351 * Else, generate an HTML "blocked" message:
1355 struct map * exports;
1359 * Workaround for stupid Netscape bug which prevents
1360 * pages from being displayed if loading a referenced
1361 * JavaScript or style sheet fails. So make it appear
1362 * as if it succeeded.
1364 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1365 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1366 && !strstr(p, "Gecko") /* save Mozilla, */
1367 && !strstr(p, "compatible") /* MSIE */
1368 && !strstr(p, "Opera")) /* and Opera. */
1370 rsp->status = strdup("200 Request for blocked URL");
1374 rsp->status = strdup("403 Request for blocked URL");
1377 if (rsp->status == NULL)
1379 free_http_response(rsp);
1380 return cgi_error_memory();
1383 exports = default_exports(csp, NULL);
1384 if (exports == NULL)
1386 free_http_response(rsp);
1387 return cgi_error_memory();
1390 #ifdef FEATURE_FORCE_LOAD
1391 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1393 * Export the force conditional block killer if
1395 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1396 * - Privoxy is configured to enforce blocks, or
1397 * - it's a CONNECT request and enforcing wouldn't work anyway.
1399 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1400 || (0 == strcmpic(csp->http->gpc, "connect")))
1401 #endif /* ndef FEATURE_FORCE_LOAD */
1403 err = map_block_killer(exports, "force-support");
1406 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1407 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1408 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1409 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1412 const char *block_reason;
1413 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
1415 block_reason = csp->action->string[ACTION_STRING_BLOCK];
1419 assert(connect_port_is_forbidden(csp));
1420 block_reason = "Forbidden CONNECT port.";
1422 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
1427 free_http_response(rsp);
1428 return cgi_error_memory();
1431 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1434 free_http_response(rsp);
1435 return cgi_error_memory();
1438 rsp->reason = RSP_REASON_BLOCKED;
1440 return finish_http_response(csp, rsp);
1445 #ifdef FEATURE_TRUST
1446 /*********************************************************************
1448 * Function : trust_url FIXME: I should be called distrust_url
1450 * Description : Calls is_untrusted_url to determine if the URL is trusted
1451 * and if not, returns a HTTP 403 response with a reject message.
1454 * 1 : csp = Current client state (buffers, headers, etc...)
1456 * Returns : NULL => trusted, else http_response.
1458 *********************************************************************/
1459 struct http_response *trust_url(struct client_state *csp)
1461 struct http_response *rsp;
1462 struct map * exports;
1463 char buf[BUFFER_SIZE];
1465 struct url_spec **tl;
1470 * Don't bother to work on trusted URLs
1472 if (!is_untrusted_url(csp))
1478 * Else, prepare a response:
1480 if (NULL == (rsp = alloc_http_response()))
1482 return cgi_error_memory();
1485 rsp->status = strdup("403 Request blocked by Privoxy");
1486 exports = default_exports(csp, NULL);
1487 if (exports == NULL || rsp->status == NULL)
1489 free_http_response(rsp);
1490 return cgi_error_memory();
1494 * Export the protocol, host, port, and referrer information
1496 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1497 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1498 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1500 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1502 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1506 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1512 free_http_response(rsp);
1513 return cgi_error_memory();
1517 * Export the trust list
1520 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1522 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1523 string_append(&p, buf);
1525 err = map(exports, "trusted-referrers", 1, p, 0);
1530 free_http_response(rsp);
1531 return cgi_error_memory();
1535 * Export the trust info, if available
1537 if (csp->config->trust_info->first)
1539 struct list_entry *l;
1542 for (l = csp->config->trust_info->first; l ; l = l->next)
1544 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1545 string_append(&p, buf);
1547 err = map(exports, "trust-info", 1, p, 0);
1551 err = map_block_killer(exports, "have-trust-info");
1557 free_http_response(rsp);
1558 return cgi_error_memory();
1562 * Export the force conditional block killer if
1564 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1565 * - Privoxy is configured to enforce blocks, or
1566 * - it's a CONNECT request and enforcing wouldn't work anyway.
1568 #ifdef FEATURE_FORCE_LOAD
1569 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1570 || (0 == strcmpic(csp->http->gpc, "connect")))
1572 err = map_block_killer(exports, "force-support");
1576 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1578 #else /* ifndef FEATURE_FORCE_LOAD */
1579 err = map_block_killer(exports, "force-support");
1580 #endif /* ndef FEATURE_FORCE_LOAD */
1585 free_http_response(rsp);
1586 return cgi_error_memory();
1590 * Build the response
1592 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1595 free_http_response(rsp);
1596 return cgi_error_memory();
1598 rsp->reason = RSP_REASON_UNTRUSTED;
1600 return finish_http_response(csp, rsp);
1602 #endif /* def FEATURE_TRUST */
1605 /*********************************************************************
1607 * Function : compile_dynamic_pcrs_job_list
1609 * Description : Compiles a dynamic pcrs job list (one with variables
1610 * resolved at request time)
1613 * 1 : csp = Current client state (buffers, headers, etc...)
1614 * 2 : b = The filter list to compile
1616 * Returns : NULL in case of errors, otherwise the
1619 *********************************************************************/
1620 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1622 struct list_entry *pattern;
1623 pcrs_job *job_list = NULL;
1624 pcrs_job *dummy = NULL;
1625 pcrs_job *lastjob = NULL;
1628 const struct pcrs_variable variables[] =
1630 {"url", csp->http->url, 1},
1631 {"path", csp->http->path, 1},
1632 {"host", csp->http->host, 1},
1633 {"origin", csp->ip_addr_str, 1},
1637 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1639 assert(pattern->str != NULL);
1641 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1645 log_error(LOG_LEVEL_ERROR,
1646 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1647 pattern->str, b->name, pcrs_strerror(error));
1652 if (error == PCRS_WARN_TRUNCATION)
1654 log_error(LOG_LEVEL_ERROR,
1655 "At least one of the variables in \'%s\' had to "
1656 "be truncated before compilation", pattern->str);
1658 if (job_list == NULL)
1664 lastjob->next = dummy;
1674 /*********************************************************************
1676 * Function : rewrite_url
1678 * Description : Rewrites a URL with a single pcrs command
1679 * and returns the result if it differs from the
1680 * original and isn't obviously invalid.
1683 * 1 : old_url = URL to rewrite.
1684 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1687 * Returns : NULL if the pcrs_command didn't change the url, or
1688 * the result of the modification.
1690 *********************************************************************/
1691 char *rewrite_url(char *old_url, const char *pcrs_command)
1693 char *new_url = NULL;
1697 assert(pcrs_command);
1699 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1703 log_error(LOG_LEVEL_REDIRECTS,
1704 "pcrs command \"%s\" didn't change \"%s\".",
1705 pcrs_command, old_url);
1710 log_error(LOG_LEVEL_REDIRECTS,
1711 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1712 pcrs_command, old_url, pcrs_strerror(hits));
1715 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1717 log_error(LOG_LEVEL_ERROR,
1718 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1719 "but the result doesn't look like a valid URL and will be ignored.",
1720 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1725 log_error(LOG_LEVEL_REDIRECTS,
1726 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1727 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1735 #ifdef FEATURE_FAST_REDIRECTS
1736 /*********************************************************************
1738 * Function : get_last_url
1740 * Description : Search for the last URL inside a string.
1741 * If the string already is a URL, it will
1742 * be the first URL found.
1745 * 1 : subject = the string to check
1746 * 2 : redirect_mode = +fast-redirect{} mode
1748 * Returns : NULL if no URL was found, or
1749 * the last URL found.
1751 *********************************************************************/
1752 char *get_last_url(char *subject, const char *redirect_mode)
1754 char *new_url = NULL;
1758 assert(redirect_mode);
1760 subject = strdup(subject);
1761 if (subject == NULL)
1763 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1767 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1769 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1770 new_url = url_decode(subject);
1771 if (new_url != NULL)
1778 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1782 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1785 * Find the last URL encoded in the request
1788 while ((tmp = strstr(tmp, "http://")) != NULL)
1792 tmp = (new_url != NULL) ? new_url : subject;
1793 while ((tmp = strstr(tmp, "https://")) != NULL)
1798 if ((new_url != NULL)
1799 && ( (new_url != subject)
1800 || (0 == strncmpic(subject, "http://", 7))
1801 || (0 == strncmpic(subject, "https://", 8))
1805 * Return new URL if we found a redirect
1806 * or if the subject already was a URL.
1808 * The second case makes sure that we can
1809 * chain get_last_url after another redirection check
1810 * (like rewrite_url) without losing earlier redirects.
1812 new_url = strdup(new_url);
1821 #endif /* def FEATURE_FAST_REDIRECTS */
1824 /*********************************************************************
1826 * Function : redirect_url
1828 * Description : Checks if Privoxy should answer the request with
1829 * a HTTP redirect and generates the redirect if
1833 * 1 : csp = Current client state (buffers, headers, etc...)
1835 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1837 *********************************************************************/
1838 struct http_response *redirect_url(struct client_state *csp)
1840 struct http_response *rsp;
1841 #ifdef FEATURE_FAST_REDIRECTS
1843 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1844 * as compile-time option? The user can easily disable
1845 * it in his action file.
1847 char * redirect_mode;
1848 #endif /* def FEATURE_FAST_REDIRECTS */
1849 char *old_url = NULL;
1850 char *new_url = NULL;
1851 char *redirection_string;
1853 if ((csp->action->flags & ACTION_REDIRECT))
1855 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1858 * If the redirection string begins with 's',
1859 * assume it's a pcrs command, otherwise treat it as
1860 * properly formatted URL and use it for the redirection
1863 * According to RFC 2616 section 14.30 the URL
1864 * has to be absolute and if the user tries:
1865 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1866 * she would get undefined results anyway.
1870 if (*redirection_string == 's')
1872 old_url = csp->http->url;
1873 new_url = rewrite_url(old_url, redirection_string);
1877 log_error(LOG_LEVEL_REDIRECTS,
1878 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1879 redirection_string);
1880 new_url = strdup(redirection_string);
1884 #ifdef FEATURE_FAST_REDIRECTS
1885 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1887 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1890 * If it exists, use the previously rewritten URL as input
1891 * otherwise just use the old path.
1893 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1894 new_url = get_last_url(old_url, redirect_mode);
1899 * Disable redirect checkers, so that they
1900 * will be only run more than once if the user
1901 * also enables them through tags.
1903 * From a performance point of view
1904 * it doesn't matter, but the duplicated
1905 * log messages are annoying.
1907 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1908 #endif /* def FEATURE_FAST_REDIRECTS */
1909 csp->action->flags &= ~ACTION_REDIRECT;
1911 /* Did any redirect action trigger? */
1914 if (0 == strcmpic(new_url, csp->http->url))
1916 log_error(LOG_LEVEL_ERROR,
1917 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1918 csp->http->url, new_url);
1923 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1925 if (NULL == (rsp = alloc_http_response()))
1928 return cgi_error_memory();
1931 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1932 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1935 free_http_response(rsp);
1936 return cgi_error_memory();
1938 rsp->reason = RSP_REASON_REDIRECTED;
1941 return finish_http_response(csp, rsp);
1945 /* Only reached if no redirect is required */
1951 #ifdef FEATURE_IMAGE_BLOCKING
1952 /*********************************************************************
1954 * Function : is_imageurl
1956 * Description : Given a URL, decide whether it is an image or not,
1957 * using either the info from a previous +image action
1958 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1959 * is MSIE and not on a Mac, tell from the browser's accept
1963 * 1 : csp = Current client state (buffers, headers, etc...)
1965 * Returns : True (nonzero) if URL is an image, false (0)
1968 *********************************************************************/
1969 int is_imageurl(const struct client_state *csp)
1971 #ifdef FEATURE_IMAGE_DETECT_MSIE
1974 tmp = get_header_value(csp->headers, "User-Agent:");
1975 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1977 tmp = get_header_value(csp->headers, "Accept:");
1978 if (tmp && strstr(tmp, "image/gif"))
1980 /* Client will accept HTML. If this seems counterintuitive,
1990 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1992 return ((csp->action->flags & ACTION_IMAGE) != 0);
1995 #endif /* def FEATURE_IMAGE_BLOCKING */
1998 #ifdef FEATURE_TRUST
1999 /*********************************************************************
2001 * Function : is_untrusted_url
2003 * Description : Should we "distrust" this URL (and block it)?
2005 * Yes if it matches a line in the trustfile, or if the
2006 * referrer matches a line starting with "+" in the
2011 * 1 : csp = Current client state (buffers, headers, etc...)
2013 * Returns : 0 => trusted, 1 => untrusted
2015 *********************************************************************/
2016 int is_untrusted_url(const struct client_state *csp)
2018 struct file_list *fl;
2019 struct block_spec *b;
2020 struct url_spec **trusted_url;
2021 struct http_request rhttp[1];
2022 const char * referer;
2026 * If we don't have a trustlist, we trust everybody
2028 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
2033 memset(rhttp, '\0', sizeof(*rhttp));
2036 * Do we trust the request URL itself?
2038 for (b = b->next; b ; b = b->next)
2040 if (url_match(b->url, csp->http))
2046 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
2048 /* no referrer was supplied */
2054 * If not, do we maybe trust its referrer?
2056 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
2062 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
2064 if (url_match(*trusted_url, rhttp))
2066 /* if the URL's referrer is from a trusted referrer, then
2067 * add the target spec to the trustfile as an unblocked
2068 * domain and return 0 (which means it's OK).
2073 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
2077 char * new_entry = strdup("~");
2079 string_append(&new_entry, csp->http->hostport);
2081 path = csp->http->path;
2082 if ( (path[0] == '/')
2084 && ((path_end = strchr(path + 2, '/')) != NULL))
2086 /* since this path points into a user's home space
2087 * be sure to include this spec in the trustfile.
2089 long path_len = path_end - path; /* save offset */
2090 path = strdup(path); /* Copy string */
2093 path_end = path + path_len; /* regenerate ptr to new buffer */
2094 *(path_end + 1) = '\0'; /* Truncate path after '/' */
2096 string_join(&new_entry, path);
2100 * Give a reason for generating this entry.
2102 string_append(&new_entry, " # Trusted referrer was: ");
2103 string_append(&new_entry, referer);
2105 if (new_entry != NULL)
2107 if (-1 == fprintf(fp, "%s\n", new_entry))
2109 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
2110 new_entry, csp->config->trustfile);
2116 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
2117 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
2124 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
2125 csp->http->hostport, csp->config->trustfile);
2133 #endif /* def FEATURE_TRUST */
2136 /*********************************************************************
2138 * Function : pcrs_filter_response
2140 * Description : Execute all text substitutions from all applying
2141 * +filter actions on the text buffer that's been
2142 * accumulated in csp->iob->buf.
2145 * 1 : csp = Current client state (buffers, headers, etc...)
2147 * Returns : a pointer to the (newly allocated) modified buffer.
2148 * or NULL if there were no hits or something went wrong
2150 *********************************************************************/
2151 static char *pcrs_filter_response(struct client_state *csp)
2154 size_t size, prev_size;
2160 struct file_list *fl;
2161 struct re_filterfile_spec *b;
2162 struct list_entry *filtername;
2164 int i, found_filters = 0;
2169 if (csp->iob->cur >= csp->iob->eod)
2175 * Need to check the set of re_filterfiles...
2177 for (i = 0; i < MAX_AF_FILES; i++)
2190 if (0 == found_filters)
2192 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2193 "content filtering enabled, but no content filters available.");
2197 size = (size_t)(csp->iob->eod - csp->iob->cur);
2198 old = csp->iob->cur;
2200 for (i = 0; i < MAX_AF_FILES; i++)
2203 if ((NULL == fl) || (NULL == fl->f))
2206 * Either there are no filter files
2207 * left, or this filter file just
2208 * contains no valid filters.
2210 * Continue to be sure we don't miss
2211 * valid filter files that are chained
2212 * after empty or invalid ones.
2217 * For all applying +filter actions, look if a filter by that
2218 * name exists and if yes, execute it's pcrs_joblist on the
2221 for (b = fl->f; b; b = b->next)
2223 if (b->type != FT_CONTENT_FILTER)
2225 /* Skip header filters */
2229 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
2230 filtername ; filtername = filtername->next)
2232 if (strcmp(b->name, filtername->str) == 0)
2234 int current_hits = 0; /* Number of hits caused by this filter */
2235 int job_number = 0; /* Which job we're currently executing */
2236 int job_hits = 0; /* How many hits the current job caused */
2237 pcrs_job *joblist = b->joblist;
2239 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2241 if (NULL == joblist)
2243 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2248 /* Apply all jobs from the joblist */
2249 for (job = joblist; NULL != job; job = job->next)
2252 job_hits = pcrs_execute(job, old, size, &new, &size);
2257 * That went well. Continue filtering
2258 * and use the result of this job as
2259 * input for the next one.
2261 current_hits += job_hits;
2262 if (old != csp->iob->cur)
2271 * This job caused an unexpected error. Inform the user
2272 * and skip the rest of the jobs in this filter. We could
2273 * continue with the next job, but usually the jobs
2274 * depend on each other or are similar enough to
2275 * fail for the same reason.
2277 * At the moment our pcrs expects the error codes of pcre 3.4,
2278 * but newer pcre versions can return additional error codes.
2279 * As a result pcrs_strerror()'s error message might be
2280 * "Unknown error ...", therefore we print the numerical value
2283 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2284 * should we use LOG_LEVEL_RE_FILTER instead?
2286 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2287 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2292 if (b->dynamic) pcrs_free_joblist(joblist);
2294 log_error(LOG_LEVEL_RE_FILTER,
2295 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2296 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2298 hits += current_hits;
2305 * If there were no hits, destroy our copy and let
2306 * chat() use the original in csp->iob
2314 csp->flags |= CSP_FLAG_MODIFIED;
2315 csp->content_length = size;
2323 /*********************************************************************
2325 * Function : gif_deanimate_response
2327 * Description : Deanimate the GIF image that has been accumulated in
2328 * csp->iob->buf, set csp->content_length to the modified
2329 * size and raise the CSP_FLAG_MODIFIED flag.
2332 * 1 : csp = Current client state (buffers, headers, etc...)
2334 * Returns : a pointer to the (newly allocated) modified buffer.
2335 * or NULL in case something went wrong.
2337 *********************************************************************/
2338 static char *gif_deanimate_response(struct client_state *csp)
2340 struct binbuffer *in, *out;
2344 size = (size_t)(csp->iob->eod - csp->iob->cur);
2346 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2347 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2349 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2353 in->buffer = csp->iob->cur;
2356 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2358 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2365 if ((int)size == out->offset)
2367 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2371 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2373 csp->content_length = out->offset;
2374 csp->flags |= CSP_FLAG_MODIFIED;
2384 /*********************************************************************
2386 * Function : get_filter_function
2388 * Description : Decides which content filter function has
2389 * to be applied (if any).
2392 * 1 : csp = Current client state (buffers, headers, etc...)
2394 * Returns : The content filter function to run, or
2395 * NULL if no content filter is active
2397 *********************************************************************/
2398 filter_function_ptr get_filter_function(struct client_state *csp)
2400 filter_function_ptr filter_function = NULL;
2403 * Are we enabling text mode by force?
2405 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2408 * Do we really have to?
2410 if (csp->content_type & CT_TEXT)
2412 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2416 csp->content_type |= CT_TEXT;
2417 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2421 if (!(csp->content_type & CT_DECLARED))
2424 * The server didn't bother to declare a MIME-Type.
2425 * Assume it's text that can be filtered.
2427 * This also regulary happens with 304 responses,
2428 * therefore logging anything here would cause
2431 csp->content_type |= CT_TEXT;
2435 * Choose the applying filter function based on
2436 * the content type and action settings.
2438 if ((csp->content_type & CT_TEXT) &&
2439 (csp->rlist != NULL) &&
2440 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2442 filter_function = pcrs_filter_response;
2444 else if ((csp->content_type & CT_GIF) &&
2445 (csp->action->flags & ACTION_DEANIMATE))
2447 filter_function = gif_deanimate_response;
2450 return filter_function;
2454 /*********************************************************************
2456 * Function : remove_chunked_transfer_coding
2458 * Description : In-situ remove the "chunked" transfer coding as defined
2459 * in rfc2616 from a buffer.
2462 * 1 : buffer = Pointer to the text buffer
2463 * 2 : size = In: Number of bytes to be processed,
2464 * Out: Number of bytes after de-chunking.
2465 * (undefined in case of errors)
2467 * Returns : JB_ERR_OK for success,
2468 * JB_ERR_PARSE otherwise
2470 *********************************************************************/
2471 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2474 unsigned int chunksize = 0;
2475 char *from_p, *to_p;
2478 from_p = to_p = buffer;
2480 if (sscanf(buffer, "%x", &chunksize) != 1)
2482 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2483 return JB_ERR_PARSE;
2486 while (chunksize > 0U)
2488 if (NULL == (from_p = strstr(from_p, "\r\n")))
2490 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2491 return JB_ERR_PARSE;
2494 if ((newsize += chunksize) >= *size)
2496 log_error(LOG_LEVEL_ERROR,
2497 "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding",
2499 return JB_ERR_PARSE;
2503 memmove(to_p, from_p, (size_t) chunksize);
2504 to_p = buffer + newsize;
2505 from_p += chunksize + 2;
2507 if (sscanf(from_p, "%x", &chunksize) != 1)
2509 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2514 /* XXX: Should get its own loglevel. */
2515 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2524 /*********************************************************************
2526 * Function : prepare_for_filtering
2528 * Description : If necessary, de-chunks and decompresses
2529 * the content so it can get filterd.
2532 * 1 : csp = Current client state (buffers, headers, etc...)
2534 * Returns : JB_ERR_OK for success,
2535 * JB_ERR_PARSE otherwise
2537 *********************************************************************/
2538 static jb_err prepare_for_filtering(struct client_state *csp)
2540 jb_err err = JB_ERR_OK;
2543 * If the body has a "chunked" transfer-encoding,
2544 * get rid of it, adjusting size and iob->eod
2546 if (csp->flags & CSP_FLAG_CHUNKED)
2548 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2550 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2551 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2552 if (JB_ERR_OK == err)
2554 csp->iob->eod = csp->iob->cur + size;
2555 csp->flags |= CSP_FLAG_MODIFIED;
2559 return JB_ERR_PARSE;
2565 * If the body has a supported transfer-encoding,
2566 * decompress it, adjusting size and iob->eod.
2568 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2570 if (0 == csp->iob->eod - csp->iob->cur)
2572 /* Nothing left after de-chunking. */
2576 err = decompress_iob(csp);
2578 if (JB_ERR_OK == err)
2580 csp->flags |= CSP_FLAG_MODIFIED;
2581 csp->content_type &= ~CT_TABOO;
2586 * Unset CT_GZIP and CT_DEFLATE to remember not
2587 * to modify the Content-Encoding header later.
2589 csp->content_type &= ~CT_GZIP;
2590 csp->content_type &= ~CT_DEFLATE;
2599 /*********************************************************************
2601 * Function : execute_content_filter
2603 * Description : Executes a given content filter.
2606 * 1 : csp = Current client state (buffers, headers, etc...)
2607 * 2 : content_filter = The filter function to execute
2609 * Returns : Pointer to the modified buffer, or
2610 * NULL if filtering failed or wasn't necessary.
2612 *********************************************************************/
2613 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2615 if (0 == csp->iob->eod - csp->iob->cur)
2618 * No content (probably status code 301, 302 ...),
2619 * no filtering necessary.
2624 if (JB_ERR_OK != prepare_for_filtering(csp))
2627 * failed to de-chunk or decompress.
2632 if (0 == csp->iob->eod - csp->iob->cur)
2635 * Clown alarm: chunked and/or compressed nothing delivered.
2640 return ((*content_filter)(csp));
2644 /*********************************************************************
2646 * Function : get_url_actions
2648 * Description : Gets the actions for this URL.
2651 * 1 : csp = Current client state (buffers, headers, etc...)
2652 * 2 : http = http_request request for blocked URLs
2656 *********************************************************************/
2657 void get_url_actions(struct client_state *csp, struct http_request *http)
2659 struct file_list *fl;
2660 struct url_actions *b;
2663 init_current_action(csp->action);
2665 for (i = 0; i < MAX_AF_FILES; i++)
2667 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2672 apply_url_actions(csp->action, http, b);
2679 /*********************************************************************
2681 * Function : apply_url_actions
2683 * Description : Applies a list of URL actions.
2686 * 1 : action = Destination.
2687 * 2 : http = Current URL
2688 * 3 : b = list of URL actions to apply
2692 *********************************************************************/
2693 void apply_url_actions(struct current_action_spec *action,
2694 struct http_request *http,
2695 struct url_actions *b)
2699 /* Should never happen */
2703 for (b = b->next; NULL != b; b = b->next)
2705 if (url_match(b->url, http))
2707 merge_current_action(action, b->action);
2713 /*********************************************************************
2715 * Function : get_forward_override_settings
2717 * Description : Returns forward settings as specified with the
2718 * forward-override{} action. forward-override accepts
2719 * forward lines similar to the one used in the
2720 * configuration file, but without the URL pattern.
2726 * in the configuration file can be replaced with
2727 * the action section:
2729 * {+forward-override{forward .}}
2733 * 1 : csp = Current client state (buffers, headers, etc...)
2735 * Returns : Pointer to forwarding structure in case of success.
2736 * Invalid syntax is fatal.
2738 *********************************************************************/
2739 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2741 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2742 char forward_settings[BUFFER_SIZE];
2743 char *http_parent = NULL;
2744 /* variable names were chosen for consistency reasons. */
2745 struct forward_spec *fwd = NULL;
2749 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2750 /* Should be enforced by load_one_actions_file() */
2751 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2753 /* Create a copy ssplit can modify */
2754 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2756 if (NULL != csp->fwd)
2759 * XXX: Currently necessary to prevent memory
2760 * leaks when the show-url-info cgi page is visited.
2762 unload_forward_spec(csp->fwd);
2766 * allocate a new forward node, valid only for
2767 * the lifetime of this request. Save its location
2768 * in csp as well, so sweep() can free it later on.
2770 fwd = csp->fwd = zalloc(sizeof(*fwd));
2773 log_error(LOG_LEVEL_FATAL,
2774 "can't allocate memory for forward-override{%s}", forward_override_line);
2775 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2779 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2780 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2782 fwd->type = SOCKS_NONE;
2784 /* Parse the parent HTTP proxy host:port */
2785 http_parent = vec[1];
2788 else if (vec_count == 3)
2790 char *socks_proxy = NULL;
2792 if (!strcasecmp(vec[0], "forward-socks4"))
2794 fwd->type = SOCKS_4;
2795 socks_proxy = vec[1];
2797 else if (!strcasecmp(vec[0], "forward-socks4a"))
2799 fwd->type = SOCKS_4A;
2800 socks_proxy = vec[1];
2802 else if (!strcasecmp(vec[0], "forward-socks5"))
2804 fwd->type = SOCKS_5;
2805 socks_proxy = vec[1];
2808 if (NULL != socks_proxy)
2810 /* Parse the SOCKS proxy host[:port] */
2811 fwd->gateway_port = 1080;
2812 parse_forwarder_address(socks_proxy,
2813 &fwd->gateway_host, &fwd->gateway_port);
2815 http_parent = vec[2];
2819 if (NULL == http_parent)
2821 log_error(LOG_LEVEL_FATAL,
2822 "Invalid forward-override syntax in: %s", forward_override_line);
2823 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2826 /* Parse http forwarding settings */
2827 if (strcmp(http_parent, ".") != 0)
2829 fwd->forward_port = 8000;
2830 parse_forwarder_address(http_parent,
2831 &fwd->forward_host, &fwd->forward_port);
2834 assert (NULL != fwd);
2836 log_error(LOG_LEVEL_CONNECT,
2837 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2843 /*********************************************************************
2845 * Function : forward_url
2847 * Description : Should we forward this to another proxy?
2850 * 1 : csp = Current client state (buffers, headers, etc...)
2851 * 2 : http = http_request request for current URL
2853 * Returns : Pointer to forwarding information.
2855 *********************************************************************/
2856 const struct forward_spec *forward_url(struct client_state *csp,
2857 const struct http_request *http)
2859 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2860 struct forward_spec *fwd = csp->config->forward;
2862 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2864 return get_forward_override_settings(csp);
2874 if (url_match(fwd->url, http))
2885 /*********************************************************************
2887 * Function : direct_response
2889 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2890 * request and if so, return a HTTP 501 to the client.
2892 * FIXME: I have a stupid name and I should handle the
2893 * requests properly. Still, what we do here is rfc-
2894 * compliant, whereas ignoring or forwarding are not.
2897 * 1 : csp = Current client state (buffers, headers, etc...)
2899 * Returns : http_response if , NULL if nonmatch or handler fail
2901 *********************************************************************/
2902 struct http_response *direct_response(struct client_state *csp)
2904 struct http_response *rsp;
2905 struct list_entry *p;
2907 if ((0 == strcmpic(csp->http->gpc, "trace"))
2908 || (0 == strcmpic(csp->http->gpc, "options")))
2910 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2912 if (!strncmpic("Max-Forwards:", p->str, 13))
2914 unsigned int max_forwards;
2917 * If it's a Max-Forwards value of zero,
2918 * we have to intercept the request.
2920 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2923 * FIXME: We could handle at least TRACE here,
2924 * but that would require a verbatim copy of
2925 * the request which we don't have anymore
2927 log_error(LOG_LEVEL_HEADER,
2928 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2931 /* Get mem for response or fail*/
2932 if (NULL == (rsp = alloc_http_response()))
2934 return cgi_error_memory();
2937 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2939 free_http_response(rsp);
2940 return cgi_error_memory();
2944 rsp->reason = RSP_REASON_UNSUPPORTED;
2946 return(finish_http_response(csp, rsp));
2955 /*********************************************************************
2957 * Function : content_filters_enabled
2959 * Description : Checks whether there are any content filters
2960 * enabled for the current request.
2963 * 1 : action = Action spec to check.
2965 * Returns : TRUE for yes, FALSE otherwise
2967 *********************************************************************/
2968 int content_filters_enabled(const struct current_action_spec *action)
2970 return ((action->flags & ACTION_DEANIMATE) ||
2971 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));