1 const char filters_rcs[] = "$Id: filters.c,v 1.89 2007/08/05 13:42:23 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `acl_addr', `add_stats', `block_acl', `block_imageurl',
9 * `block_url', `url_actions', `domain_split',
10 * `filter_popups', `forward_url', 'redirect_url',
11 * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
12 * `ijb_send_banner', `trust_url', `gif_deanimate_response',
13 * `jpeg_inspect_response', `execute_single_pcrs_command',
14 * `rewrite_url', `get_last_url'
16 * Copyright : Written by and Copyright (C) 2001, 2004-2007 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 1.89 2007/08/05 13:42:23 fabiankeil
44 * #1763173 from Stefan Huehner: declare some more functions static.
46 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
47 * Add forward-override{} to change the forwarding settings through
48 * action sections. This is mainly interesting to forward different
49 * clients differently (for example based on User-Agent or request
52 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
53 * Make sure filters with dynamic jobs actually use them.
55 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
56 * - Introduce dynamic pcrs jobs that can resolve variables.
57 * - Don't run redirect functions more than once,
58 * unless they are activated more than once.
60 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
61 * - Log the content size after decompression in decompress_iob()
62 * instead of pcrs_filter_response().
64 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
65 * Use dedicated header filter actions instead of abusing "filter".
66 * Replace "filter-client-headers" and "filter-client-headers"
67 * with "server-header-filter" and "client-header-filter".
69 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
70 * New config option: enforce-blocks.
72 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
73 * - Fix port handling in acl_addr() and use a temporary acl spec
74 * copy so error messages don't contain a truncated version.
75 * - Log size of iob before and after decompression.
77 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
78 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
79 * - Hide the "Go there anyway" link for blocked CONNECT
80 * requests where going there anyway doesn't work anyway.
82 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
83 * - Save the reason for generating http_responses.
84 * - Block (+block) with status code 403 instead of 404.
85 * - Use a different kludge to remember a failed decompression.
87 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
88 * Search for Max-Forwards headers case-insensitive,
89 * don't generate the "501 unsupported" message for invalid
90 * Max-Forwards values and don't increase negative ones.
92 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
93 * - Add HEAD support to finish_http_response.
94 * - Add error favicon to internal HTML error messages.
96 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
97 * Mark *csp as immutable for is_untrusted_url()
98 * and is_imageurl(). Closes FR 1237736.
100 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
101 * Integrate a modified version of Wil Mahan's
102 * zlib patch (PR #895531).
104 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
105 * Fixed gcc43 conversion warnings,
106 * changed sprintf calls to snprintf.
108 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
109 * Adjust comment in pcrs_filter_response()
110 * to recent pcrs changes. Hohoho.
112 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
113 * Don't crash if pcre returns an error code
114 * that pcrs didn't expect. Fixes BR 1621173.
116 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
117 * Modified is_untrusted_url to complain in case of
118 * write errors and to give a reason when adding new
119 * entries to the trustfile. Closes FR 1097611.
121 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
122 * Skip empty filter files in pcrs_filter_response,
123 * but don't ignore the ones that come afterwards.
124 * Fixes parts of BR 1619208.
126 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
127 * Added some sanity checks for get_last_url().
128 * Fixed possible segfault caused by my last commit.
130 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
131 * Let get_last_url() catch https URLs as well.
133 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
134 * Make sure get_last_url() behaves like advertised
135 * and fast-redirects{} can be combined with redirect{}.
137 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
138 * Implemented +redirect{s@foo@bar@} to generate
139 * a redirect based on a rewritten version of the
142 * Revision 1.66 2006/09/23 13:26:38 roro
143 * Replace TABs by spaces in source code.
145 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
146 * Fix +redirect{}. Didn't work with -fast-redirects.
148 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
149 * Block requests for untrusted URLs with status
150 * code 403 instead of 200.
152 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
153 * Don't free p which is still in use and will be later
154 * freed by free_map(). Don't claim the referrer is unknown
155 * when the client didn't set one.
157 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
158 * Feature request 595948: Re-Filter logging in single line
160 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
161 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
163 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
164 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
165 * with what was really the latest development (the v_3_0_branch branch)
167 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
168 * Multiple filter file support
170 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
171 * A few memory leaks plugged (mostly on error paths)
173 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
174 * Add the ability to check jpeg images for invalid
175 * lengths of comment blocks. Defensive strategy
176 * against the exploit:
177 * Microsoft Security Bulletin MS04-028
178 * Buffer Overrun in JPEG Processing (GDI+) Could
179 * Allow Code Execution (833987)
180 * Enabled with +inspect-jpegs in actions files.
182 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
183 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
185 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
186 * Fixed bug #839859: "See why" link URL now gets url-encoded.
188 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
191 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
192 * Added basic support for OPTIONS and TRACE HTTP methods:
193 * New function direct_response which handles OPTIONS and
194 * TRACE requests whose Max-Forwards header field is zero.
196 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
197 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
199 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
200 * - Bugfix: Executing a filters without jobs no longer results in
201 * turing off *all* filters.
202 * - Security fix: Malicious web servers can't cause a seg fault
203 * through bogus chunk sizes anymore
205 * Revision 1.58 2002/04/24 02:11:17 oes
206 * Jon's multiple AF patch: url_actions now evaluates rules
209 * Revision 1.57 2002/04/08 20:38:34 swa
212 * Revision 1.56 2002/04/05 15:51:24 oes
213 * - bugfix: error-pages now get correct request protocol
214 * - fix for invalid HTML in trust info
216 * Revision 1.55 2002/04/02 16:13:51 oes
217 * Fix: No "Go there anyway" for SSL
219 * Revision 1.54 2002/04/02 14:55:56 oes
220 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
222 * Revision 1.53 2002/03/26 22:29:54 swa
223 * we have a new homepage!
225 * Revision 1.52 2002/03/24 16:35:57 jongfoster
228 * Revision 1.51 2002/03/24 15:23:33 jongfoster
231 * Revision 1.50 2002/03/24 13:25:43 swa
232 * name change related issues
234 * Revision 1.49 2002/03/16 20:29:14 oes
237 * Revision 1.48 2002/03/13 20:25:34 oes
238 * Better logging for content filters
240 * Revision 1.47 2002/03/13 00:30:52 jongfoster
242 * Added option of always sending redirect for imageblock,
243 * currently disabled with #if 0.
245 * Revision 1.46 2002/03/12 01:42:49 oes
246 * Introduced modular filters
248 * Revision 1.45 2002/03/08 16:47:50 oes
249 * Added choice beween GIF and PNG built-in images
251 * Revision 1.44 2002/03/07 03:49:31 oes
252 * - Fixed compiler warnings etc
253 * - Changed built-in images from GIF to PNG
254 * (with regard to Unisys patent issue)
255 * - Added a 4x4 pattern PNG which is less intrusive
256 * than the logo but also clearly marks the deleted banners
258 * Revision 1.43 2002/01/22 23:51:59 jongfoster
259 * Replacing strsav() with the safer string_append().
261 * Adding missing html_encode() to error message generators. Where encoded
262 * and unencoded versions of a string were provided, removing the unencoded
265 * Revision 1.42 2002/01/17 21:00:32 jongfoster
266 * Moving all our URL and URL pattern parsing code to urlmatch.c.
268 * Using a single, simple url_match(pattern,url) function - rather than
269 * the 3-line match routine which was repeated all over the place.
271 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
273 * Using parse_http_url() to parse URLs without faking a HTTP
274 * request line for parse_http_request().
276 * Revision 1.41 2001/11/13 00:14:07 jongfoster
277 * Fixing stupid bug now I've figured out what || means.
278 * (It always returns 0 or 1, not one of it's paramaters.)
280 * Revision 1.40 2001/10/26 17:37:55 oes
281 * - Re-enabled Netscape 200/404 bug workaround in block_url():
282 * - Removed OS/2 special case
283 * - Made block_url() independant from sed() having been run
284 * - Made trust_url independant from sed() having been run
285 * - Made is_imageurl independant from sed() having been run.
286 * It now checks User-Agent: and Accept: by itself.
289 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
290 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
291 * threads to call select() simultaneously. So, it's time to do a real, live,
292 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
293 * (native). Both versions will work, but using __OS2__ offers multi-threading.
295 * Revision 1.38 2001/10/23 21:32:33 jongfoster
296 * Adding error-checking to selected functions
298 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
299 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
300 * filters.c. Added a FIXME in front of the offending code. I'll gladly
301 * put in a better/more robust fix for all parties if one is presented...
302 * It seems that just returning 200 instead of 404 would pretty much fix
303 * it for everyone, but I don't know all the history of the problem.
305 * Revision 1.36 2001/10/10 16:44:16 oes
306 * Added match_portlist function
308 * Revision 1.35 2001/10/07 15:41:23 oes
309 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
311 * New function remove_chunked_transfer_coding that strips chunked
312 * transfer coding to plain and is called by pcrs_filter_response
313 * and gif_deanimate_response if neccessary
315 * Improved handling of zero-change re_filter runs
317 * pcrs_filter_response and gif_deanimate_response now remove
318 * chunked transfer codeing before processing the body.
320 * Revision 1.34 2001/09/20 15:49:36 steudten
322 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
323 * See cgi.c fill_template().
325 * Revision 1.33 2001/09/16 17:05:14 jongfoster
326 * Removing unused #include showarg.h
328 * Revision 1.32 2001/09/16 13:21:27 jongfoster
329 * Changes to use new list functions.
331 * Revision 1.31 2001/09/16 11:38:02 jongfoster
332 * Splitting fill_template() into 2 functions:
333 * template_load() loads the file
334 * template_fill() performs the PCRS regexps.
335 * This is because the CGI edit interface has a "table row"
336 * template which is used many times in the page - this
337 * change means it's only loaded from disk once.
339 * Revision 1.30 2001/09/16 11:00:10 jongfoster
340 * New function alloc_http_response, for symmetry with free_http_response
342 * Revision 1.29 2001/09/13 23:32:40 jongfoster
343 * Moving image data to cgi.c rather than cgi.h
344 * Fixing a GPF under Win32 (and any other OS that protects global
345 * constants from being written to).
347 * Revision 1.28 2001/09/10 10:18:51 oes
348 * Silenced compiler warnings
350 * Revision 1.27 2001/08/05 16:06:20 jongfoster
351 * Modifiying "struct map" so that there are now separate header and
352 * "map_entry" structures. This means that functions which modify a
353 * map no longer need to return a pointer to the modified map.
354 * Also, it no longer reverses the order of the entries (which may be
355 * important with some advanced template substitutions).
357 * Revision 1.26 2001/07/30 22:08:36 jongfoster
358 * Tidying up #defines:
359 * - All feature #defines are now of the form FEATURE_xxx
360 * - Permanently turned off WIN_GUI_EDIT
361 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
363 * Revision 1.25 2001/07/26 10:09:46 oes
364 * Made browser detection a little less naive
366 * Revision 1.24 2001/07/25 17:22:51 oes
367 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
369 * Revision 1.23 2001/07/23 13:40:12 oes
370 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
372 * Revision 1.22 2001/07/18 12:29:34 oes
373 * - Made gif_deanimate_response respect
374 * csp->action->string[ACTION_STRING_DEANIMATE]
375 * - Logging cosmetics
377 * Revision 1.21 2001/07/13 13:59:53 oes
378 * - Introduced gif_deanimate_response which shares the
379 * generic content modification interface of pcrs_filter_response
380 * and acts as a wrapper to deanimate.c:gif_deanimate()
381 * - Renamed re_process_buffer to pcrs_filter_response
382 * - pcrs_filter_response now returns NULL on failiure
383 * - Removed all #ifdef PCRS
385 * Revision 1.20 2001/07/01 17:01:04 oes
386 * Added comments and missing return statement in is_untrusted_url()
388 * Revision 1.19 2001/06/29 21:45:41 oes
389 * Indentation, CRLF->LF, Tab-> Space
391 * Revision 1.18 2001/06/29 13:27:38 oes
392 * - Cleaned up, renamed and reorderd functions
393 * and improved comments
396 * - Ported to CGI platform. Now delivers
397 * http_response or NULL
398 * - Unified HTML and GIF generation (moved image detection
399 * and GIF generation here from jcc.c:chat())
400 * - Fixed HTTP status to:
401 * - 403 (Forbidden) for the "blocked" HTML message
402 * - 200 (OK) for GIF answers
403 * - 302 (Redirect) for redirect to GIF
406 * - Ported to CGI platform. Now delivers
407 * http_response or NULL
408 * - Separated detection of untrusted URL into
409 * (bool)is_untrusted_url
410 * - Added enforcement of untrusted requests
412 * - Moved redirect_url() from cgi.c to here
413 * and ported it to the CGI platform
415 * - Removed logentry from cancelled commit
417 * Revision 1.17 2001/06/09 10:55:28 jongfoster
418 * Changing BUFSIZ ==> BUFFER_SIZE
420 * Revision 1.16 2001/06/07 23:10:26 jongfoster
421 * Allowing unanchored domain patterns to back off and retry
422 * if they partially match. Optimized right-anchored patterns.
423 * Moving ACL and forward files into config file.
424 * Replacing struct gateway with struct forward_spec
426 * Revision 1.15 2001/06/03 19:12:00 oes
427 * extracted-CGI relevant stuff
429 * Revision 1.14 2001/06/01 10:30:55 oes
430 * Added optional left-anchoring to domaincmp
432 * Revision 1.13 2001/05/31 21:21:30 jongfoster
433 * Permissionsfile / actions file changes:
434 * - Changed "permission" to "action" throughout
435 * - changes to file format to allow string parameters
436 * - Moved helper functions to actions.c
438 * Revision 1.12 2001/05/31 17:35:20 oes
440 * - Enhanced domain part globbing with infix and prefix asterisk
441 * matching and optional unanchored operation
443 * Revision 1.11 2001/05/29 11:53:23 oes
444 * "See why" link added to "blocked" page
446 * Revision 1.10 2001/05/29 09:50:24 jongfoster
447 * Unified blocklist/imagelist/permissionslist.
448 * File format is still under discussion, but the internal changes
451 * Also modified interceptor behaviour:
452 * - We now intercept all URLs beginning with one of the following
453 * prefixes (and *only* these prefixes):
455 * * http://ijbswa.sf.net/config/
456 * * http://ijbswa.sourceforge.net/config/
457 * - New interceptors "home page" - go to http://i.j.b/ to see it.
458 * - Internal changes so that intercepted and fast redirect pages
459 * are not replaced with an image.
460 * - Interceptors now have the option to send a binary page direct
461 * to the client. (i.e. ijb-send-banner uses this)
462 * - Implemented show-url-info interceptor. (Which is why I needed
463 * the above interceptors changes - a typical URL is
464 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
465 * The previous mechanism would not have intercepted that, and
466 * if it had been intercepted then it then it would have replaced
469 * Revision 1.9 2001/05/27 22:17:04 oes
471 * - re_process_buffer no longer writes the modified buffer
472 * to the client, which was very ugly. It now returns the
473 * buffer, which it is then written by chat.
475 * - content_length now adjusts the Content-Length: header
476 * for modified documents rather than crunch()ing it.
477 * (Length info in csp->content_length, which is 0 for
478 * unmodified documents)
480 * - For this to work, sed() is called twice when filtering.
482 * Revision 1.8 2001/05/26 17:13:28 jongfoster
483 * Filled in a function comment.
485 * Revision 1.7 2001/05/26 15:26:15 jongfoster
486 * ACL feature now provides more security by immediately dropping
487 * connections from untrusted hosts.
489 * Revision 1.6 2001/05/26 00:28:36 jongfoster
490 * Automatic reloading of config file.
491 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
492 * Most of the global variables have been moved to a new
493 * struct configuration_spec, accessed through csp->config->globalname
494 * Most of the globals remaining are used by the Win32 GUI.
496 * Revision 1.5 2001/05/25 22:34:30 jongfoster
499 * Revision 1.4 2001/05/22 18:46:04 oes
501 * - Enabled filtering banners by size rather than URL
502 * by adding patterns that replace all standard banner
503 * sizes with the "Junkbuster" gif to the re_filterfile
505 * - Enabled filtering WebBugs by providing a pattern
506 * which kills all 1x1 images
508 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
509 * which is selected by the (nonstandard and therefore
510 * capital) letter 'U' in the option string.
511 * It causes the quantifiers to be ungreedy by default.
512 * Appending a ? turns back to greedy (!).
514 * - Added a new interceptor ijb-send-banner, which
515 * sends back the "Junkbuster" gif. Without imagelist or
516 * MSIE detection support, or if tinygif = 1, or the
517 * URL isn't recognized as an imageurl, a lame HTML
518 * explanation is sent instead.
520 * - Added new feature, which permits blocking remote
521 * script redirects and firing back a local redirect
523 * The feature is conditionally compiled, i.e. it
524 * can be disabled with --disable-fast-redirects,
525 * plus it must be activated by a "fast-redirects"
526 * line in the config file, has its own log level
527 * and of course wants to be displayed by show-proxy-args
528 * Note: Boy, all the #ifdefs in 1001 locations and
529 * all the fumbling with configure.in and acconfig.h
530 * were *way* more work than the feature itself :-(
532 * - Because a generic redirect template was needed for
533 * this, tinygif = 3 now uses the same.
535 * - Moved GIFs, and other static HTTP response templates
540 * - Removed some >400 CRs again (Jon, you really worked
543 * Revision 1.3 2001/05/20 16:44:47 jongfoster
544 * Removing last hardcoded Junkbusters.com URLs.
546 * Revision 1.2 2001/05/20 01:21:20 jongfoster
547 * Version 2.9.4 checkin.
548 * - Merged popupfile and cookiefile, and added control over PCRS
549 * filtering, in new "permissionsfile".
550 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
551 * file error you now get a message box (in the Win32 GUI) rather
552 * than the program exiting with no explanation.
553 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
555 * - Removed tabs from "config"
556 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
557 * - Bumped up version number.
559 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
560 * Initial import of version 2.9.3 source tree
563 *********************************************************************/
569 #include <sys/types.h>
578 #endif /* ndef __OS2__ */
579 #include <netinet/in.h>
581 #include <winsock2.h>
582 #endif /* ndef _WIN32 */
586 #endif /* def __OS2__ */
594 #include "jbsockets.h"
595 #include "miscutil.h"
599 #include "deanimate.h"
600 #include "urlmatch.h"
607 const char filters_h_rcs[] = FILTERS_H_VERSION;
609 /* Fix a problem with Solaris. There should be no effect on other
611 * Solaris's isspace() is a macro which uses it's argument directly
612 * as an array index. Therefore we need to make sure that high-bit
613 * characters generate +ve values, and ideally we also want to make
614 * the argument match the declared parameter type of "int".
616 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
620 /*********************************************************************
622 * Function : block_acl
624 * Description : Block this request?
625 * Decide yes or no based on ACL file.
628 * 1 : dst = The proxy or gateway address this is going to.
629 * Or NULL to check all possible targets.
630 * 2 : csp = Current client state (buffers, headers, etc...)
631 * Also includes the client IP address.
633 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
635 *********************************************************************/
636 int block_acl(struct access_control_addr *dst, struct client_state *csp)
638 struct access_control_list *acl = csp->config->acl;
640 /* if not using an access control list, then permit the connection */
646 /* search the list */
649 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
653 /* Just want to check if they have any access */
654 if (acl->action == ACL_PERMIT)
659 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
660 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
662 if (acl->action == ACL_PERMIT)
680 /*********************************************************************
682 * Function : acl_addr
684 * Description : Called from `load_config' to parse an ACL address.
687 * 1 : aspec = String specifying ACL address.
688 * 2 : aca = struct access_control_addr to fill in.
690 * Returns : 0 => Ok, everything else is an error.
692 *********************************************************************/
693 int acl_addr(const char *aspec, struct access_control_addr *aca)
698 char *acl_spec = NULL;
704 * Use a temporary acl spec copy so we can log
705 * the unmodified original in case of parse errors.
707 acl_spec = strdup(aspec);
708 if (acl_spec == NULL)
710 /* XXX: This will be logged as parse error. */
714 if ((p = strchr(acl_spec, '/')) != NULL)
717 if (ijb_isdigit(*p) == 0)
722 masklength = atoi(p);
725 if ((masklength < 0) || (masklength > 32))
731 if ((p = strchr(acl_spec, ':')) != NULL)
736 port = strtol(p, &endptr, 10);
738 if (port <= 0 || port > 65535 || *endptr != '\0')
745 aca->port = (unsigned long)port;
747 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
750 if (aca->addr == INADDR_NONE)
752 /* XXX: This will be logged as parse error. */
756 /* build the netmask */
758 for (i=1; i <= masklength ; i++)
760 aca->mask |= (1 << (32 - i));
763 /* now mask off the host portion of the ip address
764 * (i.e. save on the network portion of the address).
766 aca->addr = aca->addr & aca->mask;
771 #endif /* def FEATURE_ACL */
774 /*********************************************************************
776 * Function : match_portlist
778 * Description : Check if a given number is covered by a comma
779 * separated list of numbers and ranges (a,b-c,d,..)
782 * 1 : portlist = String with list
783 * 2 : port = port to check
785 * Returns : 0 => no match
788 *********************************************************************/
789 int match_portlist(const char *portlist, int port)
791 char *min, *max, *next, *portlist_copy;
793 min = next = portlist_copy = strdup(portlist);
796 * Zero-terminate first item and remember offset for next
798 if (NULL != (next = strchr(portlist_copy, (int) ',')))
804 * Loop through all items, checking for match
808 if (NULL == (max = strchr(min, (int) '-')))
811 * No dash, check for equality
813 if (port == atoi(min))
822 * This is a range, so check if between min and max,
823 * or, if max was omitted, between min and 65K
826 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
840 * Zero-terminate next item and remember offset for n+1
842 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
854 /*********************************************************************
856 * Function : block_url
858 * Description : Called from `chat'. Check to see if we need to block this.
861 * 1 : csp = Current client state (buffers, headers, etc...)
863 * Returns : NULL => unblocked, else HTTP block response
865 *********************************************************************/
866 struct http_response *block_url(struct client_state *csp)
868 struct http_response *rsp;
869 const char *new_content_type = NULL;
872 * If it's not blocked, don't block it ;-)
874 if ((csp->action->flags & ACTION_BLOCK) == 0)
878 if (csp->action->flags & ACTION_REDIRECT)
880 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
883 * Else, prepare a response
885 if (NULL == (rsp = alloc_http_response()))
887 return cgi_error_memory();
891 * If it's an image-url, send back an image or redirect
892 * as specified by the relevant +image action
894 #ifdef FEATURE_IMAGE_BLOCKING
895 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
899 /* determine HOW images should be blocked */
900 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
902 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
904 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
906 #if 1 /* Two alternative strategies, use this one for now: */
908 /* and handle accordingly: */
909 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
911 rsp->status = strdup("403 Request blocked by Privoxy");
912 if (rsp->status == NULL)
914 free_http_response(rsp);
915 return cgi_error_memory();
917 rsp->body = bindup(image_pattern_data, image_pattern_length);
918 if (rsp->body == NULL)
920 free_http_response(rsp);
921 return cgi_error_memory();
923 rsp->content_length = image_pattern_length;
925 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
927 free_http_response(rsp);
928 return cgi_error_memory();
932 else if (0 == strcmpic(p, "blank"))
934 rsp->status = strdup("403 Request blocked by Privoxy");
935 if (rsp->status == NULL)
937 free_http_response(rsp);
938 return cgi_error_memory();
940 rsp->body = bindup(image_blank_data, image_blank_length);
941 if (rsp->body == NULL)
943 free_http_response(rsp);
944 return cgi_error_memory();
946 rsp->content_length = image_blank_length;
948 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
950 free_http_response(rsp);
951 return cgi_error_memory();
957 rsp->status = strdup("302 Local Redirect from Privoxy");
958 if (rsp->status == NULL)
960 free_http_response(rsp);
961 return cgi_error_memory();
964 if (enlist_unique_header(rsp->headers, "Location", p))
966 free_http_response(rsp);
967 return cgi_error_memory();
971 #else /* Following code is disabled for now */
973 /* and handle accordingly: */
974 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
976 p = CGI_PREFIX "send-banner?type=pattern";
978 else if (0 == strcmpic(p, "blank"))
980 p = CGI_PREFIX "send-banner?type=blank";
982 rsp->status = strdup("302 Local Redirect from Privoxy");
983 if (rsp->status == NULL)
985 free_http_response(rsp);
986 return cgi_error_memory();
989 if (enlist_unique_header(rsp->headers, "Location", p))
991 free_http_response(rsp);
992 return cgi_error_memory();
994 #endif /* Preceeding code is disabled for now */
996 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
999 * Send empty document.
1001 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1004 rsp->body = strdup(" ");
1005 rsp->content_length = 1;
1007 rsp->status = strdup("403 Request blocked by Privoxy");
1008 if (rsp->status == NULL)
1010 free_http_response(rsp);
1011 return cgi_error_memory();
1013 if (new_content_type != 0)
1015 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1016 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1018 free_http_response(rsp);
1019 return cgi_error_memory();
1024 #endif /* def FEATURE_IMAGE_BLOCKING */
1027 * Else, generate an HTML "blocked" message:
1031 struct map * exports;
1035 * Workaround for stupid Netscape bug which prevents
1036 * pages from being displayed if loading a referenced
1037 * JavaScript or style sheet fails. So make it appear
1038 * as if it succeeded.
1040 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1041 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1042 && !strstr(p, "Gecko") /* save Mozilla, */
1043 && !strstr(p, "compatible") /* MSIE */
1044 && !strstr(p, "Opera")) /* and Opera. */
1046 rsp->status = strdup("200 Request for blocked URL");
1050 rsp->status = strdup("403 Request for blocked URL");
1053 if (rsp->status == NULL)
1055 free_http_response(rsp);
1056 return cgi_error_memory();
1059 exports = default_exports(csp, NULL);
1060 if (exports == NULL)
1062 free_http_response(rsp);
1063 return cgi_error_memory();
1066 #ifdef FEATURE_FORCE_LOAD
1067 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1069 * Export the force conditional block killer if
1071 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1072 * - Privoxy is configured to enforce blocks, or
1073 * - it's a CONNECT request and enforcing wouldn't work anyway.
1075 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1076 || (0 == strcmpic(csp->http->gpc, "connect")))
1077 #endif /* ndef FEATURE_FORCE_LOAD */
1079 err = map_block_killer(exports, "force-support");
1082 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1083 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1084 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1085 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1090 free_http_response(rsp);
1091 return cgi_error_memory();
1094 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1097 free_http_response(rsp);
1098 return cgi_error_memory();
1101 rsp->reason = RSP_REASON_BLOCKED;
1103 return finish_http_response(csp, rsp);
1108 #ifdef FEATURE_TRUST
1109 /*********************************************************************
1111 * Function : trust_url FIXME: I should be called distrust_url
1113 * Description : Calls is_untrusted_url to determine if the URL is trusted
1114 * and if not, returns a HTTP 403 response with a reject message.
1117 * 1 : csp = Current client state (buffers, headers, etc...)
1119 * Returns : NULL => trusted, else http_response.
1121 *********************************************************************/
1122 struct http_response *trust_url(struct client_state *csp)
1124 struct http_response *rsp;
1125 struct map * exports;
1126 char buf[BUFFER_SIZE];
1128 struct url_spec **tl;
1133 * Don't bother to work on trusted URLs
1135 if (!is_untrusted_url(csp))
1141 * Else, prepare a response:
1143 if (NULL == (rsp = alloc_http_response()))
1145 return cgi_error_memory();
1148 rsp->status = strdup("403 Request blocked by Privoxy");
1149 exports = default_exports(csp, NULL);
1150 if (exports == NULL || rsp->status == NULL)
1152 free_http_response(rsp);
1153 return cgi_error_memory();
1157 * Export the protocol, host, port, and referrer information
1159 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1160 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1161 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1163 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1165 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1169 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1175 free_http_response(rsp);
1176 return cgi_error_memory();
1180 * Export the trust list
1183 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1185 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1186 string_append(&p, buf);
1188 err = map(exports, "trusted-referrers", 1, p, 0);
1193 free_http_response(rsp);
1194 return cgi_error_memory();
1198 * Export the trust info, if available
1200 if (csp->config->trust_info->first)
1202 struct list_entry *l;
1205 for (l = csp->config->trust_info->first; l ; l = l->next)
1207 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1208 string_append(&p, buf);
1210 err = map(exports, "trust-info", 1, p, 0);
1214 err = map_block_killer(exports, "have-trust-info");
1220 free_http_response(rsp);
1221 return cgi_error_memory();
1225 * Export the force conditional block killer if
1227 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1228 * - Privoxy is configured to enforce blocks, or
1229 * - it's a CONNECT request and enforcing wouldn't work anyway.
1231 #ifdef FEATURE_FORCE_LOAD
1232 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1233 || (0 == strcmpic(csp->http->gpc, "connect")))
1235 err = map_block_killer(exports, "force-support");
1239 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1241 #else /* ifndef FEATURE_FORCE_LOAD */
1242 err = map_block_killer(exports, "force-support");
1243 #endif /* ndef FEATURE_FORCE_LOAD */
1248 free_http_response(rsp);
1249 return cgi_error_memory();
1253 * Build the response
1255 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1258 free_http_response(rsp);
1259 return cgi_error_memory();
1261 rsp->reason = RSP_REASON_UNTRUSTED;
1263 return finish_http_response(csp, rsp);
1265 #endif /* def FEATURE_TRUST */
1268 /*********************************************************************
1270 * Function : compile_dynamic_pcrs_job_list
1272 * Description : Compiles a dynamic pcrs job list (one with variables
1273 * resolved at request time)
1276 * 1 : csp = Current client state (buffers, headers, etc...)
1277 * 2 : b = The filter list to compile
1279 * Returns : NULL in case of errors, otherwise the
1282 *********************************************************************/
1283 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1285 struct list_entry *pattern;
1286 pcrs_job *job_list = NULL;
1287 pcrs_job *dummy = NULL;
1288 pcrs_job *lastjob = NULL;
1291 const struct pcrs_variable variables[] =
1293 {"url", csp->http->url, 1},
1294 {"path", csp->http->path, 1},
1295 {"host", csp->http->host, 1},
1296 {"origin", csp->ip_addr_str, 1},
1300 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1302 assert(pattern->str != NULL);
1304 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1308 log_error(LOG_LEVEL_ERROR,
1309 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1310 pattern->str, b->name, pcrs_strerror(error));
1315 if (error == PCRS_WARN_TRUNCATION)
1317 log_error(LOG_LEVEL_ERROR,
1318 "At least one of the variables in \'%s\' had to "
1319 "be truncated before compilation", pattern->str);
1321 if (job_list == NULL)
1327 lastjob->next = dummy;
1337 /*********************************************************************
1339 * Function : rewrite_url
1341 * Description : Rewrites a URL with a single pcrs command
1342 * and returns the result if it differs from the
1343 * original and isn't obviously invalid.
1346 * 1 : old_url = URL to rewrite.
1347 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1350 * Returns : NULL if the pcrs_command didn't change the url, or
1351 * the result of the modification.
1353 *********************************************************************/
1354 char *rewrite_url(char *old_url, const char *pcrs_command)
1356 char *new_url = NULL;
1360 assert(pcrs_command);
1362 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1366 log_error(LOG_LEVEL_REDIRECTS,
1367 "pcrs command \"%s\" didn't change \"%s\".",
1368 pcrs_command, old_url);
1373 log_error(LOG_LEVEL_REDIRECTS,
1374 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1375 pcrs_command, old_url, pcrs_strerror(hits));
1378 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1380 log_error(LOG_LEVEL_ERROR,
1381 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1382 "but the result doesn't look like a valid URL and will be ignored.",
1383 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1388 log_error(LOG_LEVEL_REDIRECTS,
1389 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1390 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1398 #ifdef FEATURE_FAST_REDIRECTS
1399 /*********************************************************************
1401 * Function : get_last_url
1403 * Description : Search for the last URL inside a string.
1404 * If the string already is a URL, it will
1405 * be the first URL found.
1408 * 1 : subject = the string to check
1409 * 2 : redirect_mode = +fast-redirect{} mode
1411 * Returns : NULL if no URL was found, or
1412 * the last URL found.
1414 *********************************************************************/
1415 char *get_last_url(char *subject, const char *redirect_mode)
1417 char *new_url = NULL;
1421 assert(redirect_mode);
1423 subject = strdup(subject);
1424 if (subject == NULL)
1426 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1430 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1432 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1433 new_url = url_decode(subject);
1434 if (new_url != NULL)
1441 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1445 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1448 * Find the last URL encoded in the request
1451 while ((tmp = strstr(tmp, "http://")) != NULL)
1455 tmp = (new_url != NULL) ? new_url : subject;
1456 while ((tmp = strstr(tmp, "https://")) != NULL)
1461 if ((new_url != NULL)
1462 && ( (new_url != subject)
1463 || (0 == strncmpic(subject, "http://", 7))
1464 || (0 == strncmpic(subject, "https://", 8))
1468 * Return new URL if we found a redirect
1469 * or if the subject already was a URL.
1471 * The second case makes sure that we can
1472 * chain get_last_url after another redirection check
1473 * (like rewrite_url) without losing earlier redirects.
1475 new_url = strdup(new_url);
1484 #endif /* def FEATURE_FAST_REDIRECTS */
1487 /*********************************************************************
1489 * Function : redirect_url
1491 * Description : Checks if Privoxy should answer the request with
1492 * a HTTP redirect and generates the redirect if
1496 * 1 : csp = Current client state (buffers, headers, etc...)
1498 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1500 *********************************************************************/
1501 struct http_response *redirect_url(struct client_state *csp)
1503 struct http_response *rsp;
1504 #ifdef FEATURE_FAST_REDIRECTS
1506 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1507 * as compile-time option? The user can easily disable
1508 * it in his action file.
1510 char * redirect_mode;
1511 #endif /* def FEATURE_FAST_REDIRECTS */
1512 char *old_url = NULL;
1513 char *new_url = NULL;
1514 char *redirection_string;
1516 if ((csp->action->flags & ACTION_REDIRECT))
1518 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1521 * If the redirection string begins with 's',
1522 * assume it's a pcrs command, otherwise treat it as
1523 * properly formatted URL and use it for the redirection
1526 * According to RFC 2616 section 14.30 the URL
1527 * has to be absolute and if the user tries:
1528 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1529 * she would get undefined results anyway.
1533 if (*redirection_string == 's')
1535 old_url = csp->http->url;
1536 new_url = rewrite_url(old_url, redirection_string);
1540 log_error(LOG_LEVEL_REDIRECTS,
1541 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1542 redirection_string);
1543 new_url = strdup(redirection_string);
1547 #ifdef FEATURE_FAST_REDIRECTS
1548 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1550 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1553 * If it exists, use the previously rewritten URL as input
1554 * otherwise just use the old path.
1556 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1557 new_url = get_last_url(old_url, redirect_mode);
1562 * Disable redirect checkers, so that they
1563 * will be only run more than once if the user
1564 * also enables them through tags.
1566 * From a performance point of view
1567 * it doesn't matter, but the duplicated
1568 * log messages are annoying.
1570 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1571 #endif /* def FEATURE_FAST_REDIRECTS */
1572 csp->action->flags &= ~ACTION_REDIRECT;
1574 /* Did any redirect action trigger? */
1577 if (0 == strcmpic(new_url, csp->http->url))
1579 log_error(LOG_LEVEL_ERROR,
1580 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1581 csp->http->url, new_url);
1586 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1588 if (NULL == (rsp = alloc_http_response()))
1591 return cgi_error_memory();
1594 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1595 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1598 free_http_response(rsp);
1599 return cgi_error_memory();
1601 rsp->reason = RSP_REASON_REDIRECTED;
1604 return finish_http_response(csp, rsp);
1608 /* Only reached if no redirect is required */
1614 #ifdef FEATURE_IMAGE_BLOCKING
1615 /*********************************************************************
1617 * Function : is_imageurl
1619 * Description : Given a URL, decide whether it is an image or not,
1620 * using either the info from a previous +image action
1621 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1622 * is MSIE and not on a Mac, tell from the browser's accept
1626 * 1 : csp = Current client state (buffers, headers, etc...)
1628 * Returns : True (nonzero) if URL is an image, false (0)
1631 *********************************************************************/
1632 int is_imageurl(const struct client_state *csp)
1634 #ifdef FEATURE_IMAGE_DETECT_MSIE
1637 tmp = get_header_value(csp->headers, "User-Agent:");
1638 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1640 tmp = get_header_value(csp->headers, "Accept:");
1641 if (tmp && strstr(tmp, "image/gif"))
1643 /* Client will accept HTML. If this seems counterintuitive,
1653 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1655 return ((csp->action->flags & ACTION_IMAGE) != 0);
1658 #endif /* def FEATURE_IMAGE_BLOCKING */
1661 #ifdef FEATURE_TRUST
1662 /*********************************************************************
1664 * Function : is_untrusted_url
1666 * Description : Should we "distrust" this URL (and block it)?
1668 * Yes if it matches a line in the trustfile, or if the
1669 * referrer matches a line starting with "+" in the
1674 * 1 : csp = Current client state (buffers, headers, etc...)
1676 * Returns : 0 => trusted, 1 => untrusted
1678 *********************************************************************/
1679 int is_untrusted_url(const struct client_state *csp)
1681 struct file_list *fl;
1682 struct block_spec *b;
1683 struct url_spec **trusted_url;
1684 struct http_request rhttp[1];
1685 const char * referer;
1689 * If we don't have a trustlist, we trust everybody
1691 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1696 memset(rhttp, '\0', sizeof(*rhttp));
1699 * Do we trust the request URL itself?
1701 for (b = b->next; b ; b = b->next)
1703 if (url_match(b->url, csp->http))
1709 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1711 /* no referrer was supplied */
1717 * If not, do we maybe trust its referrer?
1719 err = parse_http_url(referer, rhttp, csp);
1725 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1727 if (url_match(*trusted_url, rhttp))
1729 /* if the URL's referrer is from a trusted referrer, then
1730 * add the target spec to the trustfile as an unblocked
1731 * domain and return 0 (which means it's OK).
1736 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1740 char * new_entry = strdup("~");
1742 string_append(&new_entry, csp->http->hostport);
1744 path = csp->http->path;
1745 if ( (path[0] == '/')
1747 && ((path_end = strchr(path + 2, '/')) != NULL))
1749 /* since this path points into a user's home space
1750 * be sure to include this spec in the trustfile.
1752 int path_len = path_end - path; /* save offset */
1753 path = strdup(path); /* Copy string */
1756 path_end = path + path_len; /* regenerate ptr to new buffer */
1757 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1759 string_join(&new_entry, path);
1763 * Give a reason for generating this entry.
1765 string_append(&new_entry, " # Trusted referrer was: ");
1766 string_append(&new_entry, referer);
1768 if (new_entry != NULL)
1770 if (-1 == fprintf(fp, "%s\n", new_entry))
1772 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1773 new_entry, csp->config->trustfile);
1779 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1780 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1787 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1788 csp->http->hostport, csp->config->trustfile);
1796 #endif /* def FEATURE_TRUST */
1799 /*********************************************************************
1801 * Function : pcrs_filter_response
1803 * Description : Execute all text substitutions from all applying
1804 * +filter actions on the text buffer that's been accumulated
1805 * in csp->iob->buf. If this changes the contents, set
1806 * csp->content_length to the modified size and raise the
1807 * CSP_FLAG_MODIFIED flag.
1809 * XXX: Currently pcrs_filter_response is also responsible
1810 * for dechunking and decompressing. Both should be
1811 * done in separate functions so other content modifiers
1812 * profit as well, even if pcrs filtering is disabled.
1815 * 1 : csp = Current client state (buffers, headers, etc...)
1817 * Returns : a pointer to the (newly allocated) modified buffer.
1818 * or NULL if there were no hits or something went wrong
1820 *********************************************************************/
1821 char *pcrs_filter_response(struct client_state *csp)
1824 size_t size, prev_size;
1826 char *old = csp->iob->cur, *new = NULL;
1829 struct file_list *fl;
1830 struct re_filterfile_spec *b;
1831 struct list_entry *filtername;
1833 int i, found_filters = 0;
1838 if (csp->iob->cur >= csp->iob->eod)
1842 size = (size_t)(csp->iob->eod - csp->iob->cur);
1845 * Need to check the set of re_filterfiles...
1847 for (i = 0; i < MAX_AF_FILES; i++)
1860 if (0 == found_filters)
1862 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1867 * If the body has a "chunked" transfer-encoding,
1868 * get rid of it first, adjusting size and iob->eod
1870 if (csp->flags & CSP_FLAG_CHUNKED)
1872 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1873 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1877 csp->iob->eod = csp->iob->cur + size;
1878 csp->flags |= CSP_FLAG_MODIFIED;
1883 * If the body has a compressed transfer-encoding,
1884 * uncompress it first, adjusting size and iob->eod.
1885 * Note that decompression occurs after de-chunking.
1887 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1889 /* Notice that we at least tried to decompress. */
1890 if (JB_ERR_OK != decompress_iob(csp))
1893 * We failed to decompress the data; there's no point
1894 * in continuing since we can't filter.
1896 * XXX: Actually the Accept-Encoding header may
1897 * just be incorrect in which case we could continue
1900 * Unset CT_GZIP and CT_DEFLATE to remember not
1901 * to modify the Content-Encoding header later.
1903 csp->content_type &= ~CT_GZIP;
1904 csp->content_type &= ~CT_DEFLATE;
1909 * Decompression gives us a completely new iob,
1910 * so we need to update.
1912 size = (size_t)(csp->iob->eod - csp->iob->cur);
1913 old = csp->iob->cur;
1915 csp->flags |= CSP_FLAG_MODIFIED;
1916 csp->content_type &= ~CT_TABOO;
1920 for (i = 0; i < MAX_AF_FILES; i++)
1923 if ((NULL == fl) || (NULL == fl->f))
1926 * Either there are no filter files
1927 * left, or this filter file just
1928 * contains no valid filters.
1930 * Continue to be sure we don't miss
1931 * valid filter files that are chained
1932 * after empty or invalid ones.
1937 * For all applying +filter actions, look if a filter by that
1938 * name exists and if yes, execute it's pcrs_joblist on the
1941 for (b = fl->f; b; b = b->next)
1943 if (b->type != FT_CONTENT_FILTER)
1945 /* Skip header filters */
1949 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1950 filtername ; filtername = filtername->next)
1952 if (strcmp(b->name, filtername->str) == 0)
1954 int current_hits = 0; /* Number of hits caused by this filter */
1955 int job_number = 0; /* Which job we're currently executing */
1956 int job_hits = 0; /* How many hits the current job caused */
1957 pcrs_job *joblist = b->joblist;
1959 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1961 if (NULL == joblist)
1963 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1968 /* Apply all jobs from the joblist */
1969 for (job = joblist; NULL != job; job = job->next)
1972 job_hits = pcrs_execute(job, old, size, &new, &size);
1977 * That went well. Continue filtering
1978 * and use the result of this job as
1979 * input for the next one.
1981 current_hits += job_hits;
1982 if (old != csp->iob->cur)
1991 * This job caused an unexpected error. Inform the user
1992 * and skip the rest of the jobs in this filter. We could
1993 * continue with the next job, but usually the jobs
1994 * depend on each other or are similar enough to
1995 * fail for the same reason.
1997 * At the moment our pcrs expects the error codes of pcre 3.4,
1998 * but newer pcre versions can return additional error codes.
1999 * As a result pcrs_strerror()'s error message might be
2000 * "Unknown error ...", therefore we print the numerical value
2003 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2004 * should we use LOG_LEVEL_RE_FILTER instead?
2006 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2007 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2012 if (b->dynamic) pcrs_free_joblist(joblist);
2014 log_error(LOG_LEVEL_RE_FILTER,
2015 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2016 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2018 hits += current_hits;
2025 * If there were no hits, destroy our copy and let
2026 * chat() use the original in csp->iob
2034 csp->flags |= CSP_FLAG_MODIFIED;
2035 csp->content_length = size;
2043 /*********************************************************************
2045 * Function : gif_deanimate_response
2047 * Description : Deanimate the GIF image that has been accumulated in
2048 * csp->iob->buf, set csp->content_length to the modified
2049 * size and raise the CSP_FLAG_MODIFIED flag.
2052 * 1 : csp = Current client state (buffers, headers, etc...)
2054 * Returns : a pointer to the (newly allocated) modified buffer.
2055 * or NULL in case something went wrong.
2057 *********************************************************************/
2058 char *gif_deanimate_response(struct client_state *csp)
2060 struct binbuffer *in, *out;
2064 size = (size_t)(csp->iob->eod - csp->iob->cur);
2067 * If the body has a "chunked" transfer-encoding,
2068 * get rid of it first, adjusting size and iob->eod
2070 if (csp->flags & CSP_FLAG_CHUNKED)
2072 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2073 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2077 csp->iob->eod = csp->iob->cur + size;
2078 csp->flags |= CSP_FLAG_MODIFIED;
2081 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2082 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2084 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2088 in->buffer = csp->iob->cur;
2091 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2093 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2100 if ((int)size == out->offset)
2102 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2106 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2108 csp->content_length = out->offset;
2109 csp->flags |= CSP_FLAG_MODIFIED;
2119 /*********************************************************************
2121 * Function : jpeg_inspect_response
2126 * 1 : csp = Current client state (buffers, headers, etc...)
2128 * Returns : a pointer to the (newly allocated) modified buffer
2129 * or NULL in case something went wrong.
2131 *********************************************************************/
2132 char *jpeg_inspect_response(struct client_state *csp)
2134 struct binbuffer *in = NULL;
2135 struct binbuffer *out = NULL;
2139 size = (size_t)(csp->iob->eod - csp->iob->cur);
2142 * If the body has a "chunked" transfer-encoding,
2143 * get rid of it first, adjusting size and iob->eod
2145 if (csp->flags & CSP_FLAG_CHUNKED)
2147 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2148 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2152 csp->iob->eod = csp->iob->cur + size;
2153 csp->flags |= CSP_FLAG_MODIFIED;
2156 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2158 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2162 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2164 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2168 in->buffer = csp->iob->cur;
2172 * Calling jpeg_inspect has the side-effect of creating and
2173 * modifying the image buffer of "out" directly.
2175 if (jpeg_inspect(in, out))
2177 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2185 csp->content_length = out->offset;
2186 csp->flags |= CSP_FLAG_MODIFIED;
2196 /*********************************************************************
2198 * Function : remove_chunked_transfer_coding
2200 * Description : In-situ remove the "chunked" transfer coding as defined
2201 * in rfc2616 from a buffer.
2204 * 1 : buffer = Pointer to the text buffer
2205 * 2 : size = Number of bytes to be processed
2207 * Returns : The new size, i.e. the number of bytes from buffer which
2208 * are occupied by the stripped body, or 0 in case something
2211 *********************************************************************/
2212 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2215 unsigned int chunksize = 0;
2216 char *from_p, *to_p;
2219 from_p = to_p = buffer;
2221 if (sscanf(buffer, "%x", &chunksize) != 1)
2223 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2227 while (chunksize > 0)
2229 if (NULL == (from_p = strstr(from_p, "\r\n")))
2231 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2235 if ((newsize += chunksize) >= size)
2237 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2242 memmove(to_p, from_p, (size_t) chunksize);
2243 to_p = buffer + newsize;
2244 from_p += chunksize + 2;
2246 if (sscanf(from_p, "%x", &chunksize) != 1)
2248 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2253 /* FIXME: Should this get its own loglevel? */
2254 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", size, newsize);
2260 /*********************************************************************
2262 * Function : url_actions
2264 * Description : Gets the actions for this URL.
2267 * 1 : http = http_request request for blocked URLs
2268 * 2 : csp = Current client state (buffers, headers, etc...)
2272 *********************************************************************/
2273 void url_actions(struct http_request *http,
2274 struct client_state *csp)
2276 struct file_list *fl;
2277 struct url_actions *b;
2280 init_current_action(csp->action);
2282 for (i = 0; i < MAX_AF_FILES; i++)
2284 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2289 apply_url_actions(csp->action, http, b);
2296 /*********************************************************************
2298 * Function : apply_url_actions
2300 * Description : Applies a list of URL actions.
2303 * 1 : action = Destination.
2304 * 2 : http = Current URL
2305 * 3 : b = list of URL actions to apply
2309 *********************************************************************/
2310 void apply_url_actions(struct current_action_spec *action,
2311 struct http_request *http,
2312 struct url_actions *b)
2316 /* Should never happen */
2320 for (b = b->next; NULL != b; b = b->next)
2322 if (url_match(b->url, http))
2324 merge_current_action(action, b->action);
2330 /*********************************************************************
2332 * Function : get_forward_override_settings
2334 * Description : Returns forward settings as specified with the
2335 * forward-override{} action. forward-override accepts
2336 * forward lines similar to the one used in the
2337 * configuration file, but without the URL pattern.
2343 * in the configuration file can be replaced with
2344 * the action section:
2346 * {+forward-override{forward .}}
2350 * 1 : csp = Current client state (buffers, headers, etc...)
2352 * Returns : Pointer to forwarding structure in case of success.
2353 * Invalid syntax is fatal.
2355 *********************************************************************/
2356 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2358 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2359 char forward_settings[BUFFER_SIZE];
2360 char *http_parent = NULL;
2361 /* variable names were chosen for consistency reasons. */
2362 struct forward_spec *fwd = NULL;
2366 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2367 /* Should be enforced by load_one_actions_file() */
2368 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2370 /* Create a copy ssplit can modify */
2371 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2373 if (NULL != csp->fwd)
2376 * XXX: Currently necessary to prevent memory
2377 * leaks when the show-url-info cgi page is visited.
2379 unload_forward_spec(csp->fwd);
2383 * allocate a new forward node, valid only for
2384 * the lifetime of this request. Save its location
2385 * in csp as well, so sweep() can free it later on.
2387 fwd = csp->fwd = zalloc(sizeof(*fwd));
2390 log_error(LOG_LEVEL_FATAL,
2391 "can't allocate memory for forward-override{%s}", forward_override_line);
2392 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2395 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2396 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2398 fwd->type = SOCKS_NONE;
2400 /* Parse the parent HTTP proxy host:port */
2401 http_parent = vec[1];
2404 else if (vec_count == 3)
2406 char *socks_proxy = NULL;
2408 if (!strcasecmp(vec[0], "forward-socks4"))
2410 fwd->type = SOCKS_4;
2411 socks_proxy = vec[1];
2413 else if (!strcasecmp(vec[0], "forward-socks4a"))
2415 fwd->type = SOCKS_4A;
2416 socks_proxy = vec[1];
2419 if (NULL != socks_proxy)
2421 /* Parse the SOCKS proxy host[:port] */
2422 fwd->gateway_host = strdup(socks_proxy);
2424 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2426 *socks_proxy++ = '\0';
2427 fwd->gateway_port = strtol(socks_proxy, NULL, 0);
2430 if (fwd->gateway_port <= 0)
2432 fwd->gateway_port = 1080;
2435 http_parent = vec[2];
2439 if (NULL == http_parent)
2441 log_error(LOG_LEVEL_FATAL,
2442 "Invalid forward-override syntax in: %s", forward_override_line);
2443 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2446 /* Parse http forwarding settings */
2447 if (strcmp(http_parent, ".") != 0)
2449 fwd->forward_host = strdup(http_parent);
2451 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2453 *http_parent++ = '\0';
2454 fwd->forward_port = strtol(http_parent, NULL, 0);
2457 if (fwd->forward_port <= 0)
2459 fwd->forward_port = 8000;
2463 assert (NULL != fwd);
2465 log_error(LOG_LEVEL_CONNECT,
2466 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2472 /*********************************************************************
2474 * Function : forward_url
2476 * Description : Should we forward this to another proxy?
2478 * XXX: Should be changed to make use of csp->fwd.
2481 * 1 : http = http_request request for current URL
2482 * 2 : csp = Current client state (buffers, headers, etc...)
2484 * Returns : Pointer to forwarding information.
2486 *********************************************************************/
2487 const struct forward_spec * forward_url(struct http_request *http,
2488 struct client_state *csp)
2490 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2491 struct forward_spec *fwd = csp->config->forward;
2493 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2495 return get_forward_override_settings(csp);
2505 if (url_match(fwd->url, http))
2516 /*********************************************************************
2518 * Function : direct_response
2520 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2521 * request and if so, return a HTTP 501 to the client.
2523 * FIXME: I have a stupid name and I should handle the
2524 * requests properly. Still, what we do here is rfc-
2525 * compliant, whereas ignoring or forwarding are not.
2528 * 1 : csp = Current client state (buffers, headers, etc...)
2530 * Returns : http_response if , NULL if nonmatch or handler fail
2532 *********************************************************************/
2533 struct http_response *direct_response(struct client_state *csp)
2535 struct http_response *rsp;
2536 struct list_entry *p;
2538 if ((0 == strcmpic(csp->http->gpc, "trace"))
2539 || (0 == strcmpic(csp->http->gpc, "options")))
2541 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2543 if (!strncmpic("Max-Forwards:", p->str, 13))
2545 unsigned int max_forwards;
2548 * If it's a Max-Forwards value of zero,
2549 * we have to intercept the request.
2551 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2554 * FIXME: We could handle at least TRACE here,
2555 * but that would require a verbatim copy of
2556 * the request which we don't have anymore
2558 log_error(LOG_LEVEL_HEADER,
2559 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2562 /* Get mem for response or fail*/
2563 if (NULL == (rsp = alloc_http_response()))
2565 return cgi_error_memory();
2568 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2570 free_http_response(rsp);
2571 return cgi_error_memory();
2575 rsp->reason = RSP_REASON_UNSUPPORTED;
2577 return(finish_http_response(csp, rsp));