1 const char filters_rcs[] = "$Id: filters.c,v 1.88 2007/06/01 16:41:11 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 1.88 2007/06/01 16:41:11 fabiankeil
44 * Add forward-override{} to change the forwarding settings through
45 * action sections. This is mainly interesting to forward different
46 * clients differently (for example based on User-Agent or request
49 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
50 * Make sure filters with dynamic jobs actually use them.
52 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
53 * - Introduce dynamic pcrs jobs that can resolve variables.
54 * - Don't run redirect functions more than once,
55 * unless they are activated more than once.
57 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
58 * - Log the content size after decompression in decompress_iob()
59 * instead of pcrs_filter_response().
61 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
62 * Use dedicated header filter actions instead of abusing "filter".
63 * Replace "filter-client-headers" and "filter-client-headers"
64 * with "server-header-filter" and "client-header-filter".
66 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
67 * New config option: enforce-blocks.
69 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
70 * - Fix port handling in acl_addr() and use a temporary acl spec
71 * copy so error messages don't contain a truncated version.
72 * - Log size of iob before and after decompression.
74 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
75 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
76 * - Hide the "Go there anyway" link for blocked CONNECT
77 * requests where going there anyway doesn't work anyway.
79 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
80 * - Save the reason for generating http_responses.
81 * - Block (+block) with status code 403 instead of 404.
82 * - Use a different kludge to remember a failed decompression.
84 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
85 * Search for Max-Forwards headers case-insensitive,
86 * don't generate the "501 unsupported" message for invalid
87 * Max-Forwards values and don't increase negative ones.
89 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
90 * - Add HEAD support to finish_http_response.
91 * - Add error favicon to internal HTML error messages.
93 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
94 * Mark *csp as immutable for is_untrusted_url()
95 * and is_imageurl(). Closes FR 1237736.
97 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
98 * Integrate a modified version of Wil Mahan's
99 * zlib patch (PR #895531).
101 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
102 * Fixed gcc43 conversion warnings,
103 * changed sprintf calls to snprintf.
105 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
106 * Adjust comment in pcrs_filter_response()
107 * to recent pcrs changes. Hohoho.
109 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
110 * Don't crash if pcre returns an error code
111 * that pcrs didn't expect. Fixes BR 1621173.
113 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
114 * Modified is_untrusted_url to complain in case of
115 * write errors and to give a reason when adding new
116 * entries to the trustfile. Closes FR 1097611.
118 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
119 * Skip empty filter files in pcrs_filter_response,
120 * but don't ignore the ones that come afterwards.
121 * Fixes parts of BR 1619208.
123 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
124 * Added some sanity checks for get_last_url().
125 * Fixed possible segfault caused by my last commit.
127 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
128 * Let get_last_url() catch https URLs as well.
130 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
131 * Make sure get_last_url() behaves like advertised
132 * and fast-redirects{} can be combined with redirect{}.
134 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
135 * Implemented +redirect{s@foo@bar@} to generate
136 * a redirect based on a rewritten version of the
139 * Revision 1.66 2006/09/23 13:26:38 roro
140 * Replace TABs by spaces in source code.
142 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
143 * Fix +redirect{}. Didn't work with -fast-redirects.
145 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
146 * Block requests for untrusted URLs with status
147 * code 403 instead of 200.
149 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
150 * Don't free p which is still in use and will be later
151 * freed by free_map(). Don't claim the referrer is unknown
152 * when the client didn't set one.
154 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
155 * Feature request 595948: Re-Filter logging in single line
157 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
158 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
160 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
161 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
162 * with what was really the latest development (the v_3_0_branch branch)
164 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
165 * Multiple filter file support
167 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
168 * A few memory leaks plugged (mostly on error paths)
170 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
171 * Add the ability to check jpeg images for invalid
172 * lengths of comment blocks. Defensive strategy
173 * against the exploit:
174 * Microsoft Security Bulletin MS04-028
175 * Buffer Overrun in JPEG Processing (GDI+) Could
176 * Allow Code Execution (833987)
177 * Enabled with +inspect-jpegs in actions files.
179 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
180 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
182 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
183 * Fixed bug #839859: "See why" link URL now gets url-encoded.
185 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
188 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
189 * Added basic support for OPTIONS and TRACE HTTP methods:
190 * New function direct_response which handles OPTIONS and
191 * TRACE requests whose Max-Forwards header field is zero.
193 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
194 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
196 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
197 * - Bugfix: Executing a filters without jobs no longer results in
198 * turing off *all* filters.
199 * - Security fix: Malicious web servers can't cause a seg fault
200 * through bogus chunk sizes anymore
202 * Revision 1.58 2002/04/24 02:11:17 oes
203 * Jon's multiple AF patch: url_actions now evaluates rules
206 * Revision 1.57 2002/04/08 20:38:34 swa
209 * Revision 1.56 2002/04/05 15:51:24 oes
210 * - bugfix: error-pages now get correct request protocol
211 * - fix for invalid HTML in trust info
213 * Revision 1.55 2002/04/02 16:13:51 oes
214 * Fix: No "Go there anyway" for SSL
216 * Revision 1.54 2002/04/02 14:55:56 oes
217 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
219 * Revision 1.53 2002/03/26 22:29:54 swa
220 * we have a new homepage!
222 * Revision 1.52 2002/03/24 16:35:57 jongfoster
225 * Revision 1.51 2002/03/24 15:23:33 jongfoster
228 * Revision 1.50 2002/03/24 13:25:43 swa
229 * name change related issues
231 * Revision 1.49 2002/03/16 20:29:14 oes
234 * Revision 1.48 2002/03/13 20:25:34 oes
235 * Better logging for content filters
237 * Revision 1.47 2002/03/13 00:30:52 jongfoster
239 * Added option of always sending redirect for imageblock,
240 * currently disabled with #if 0.
242 * Revision 1.46 2002/03/12 01:42:49 oes
243 * Introduced modular filters
245 * Revision 1.45 2002/03/08 16:47:50 oes
246 * Added choice beween GIF and PNG built-in images
248 * Revision 1.44 2002/03/07 03:49:31 oes
249 * - Fixed compiler warnings etc
250 * - Changed built-in images from GIF to PNG
251 * (with regard to Unisys patent issue)
252 * - Added a 4x4 pattern PNG which is less intrusive
253 * than the logo but also clearly marks the deleted banners
255 * Revision 1.43 2002/01/22 23:51:59 jongfoster
256 * Replacing strsav() with the safer string_append().
258 * Adding missing html_encode() to error message generators. Where encoded
259 * and unencoded versions of a string were provided, removing the unencoded
262 * Revision 1.42 2002/01/17 21:00:32 jongfoster
263 * Moving all our URL and URL pattern parsing code to urlmatch.c.
265 * Using a single, simple url_match(pattern,url) function - rather than
266 * the 3-line match routine which was repeated all over the place.
268 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
270 * Using parse_http_url() to parse URLs without faking a HTTP
271 * request line for parse_http_request().
273 * Revision 1.41 2001/11/13 00:14:07 jongfoster
274 * Fixing stupid bug now I've figured out what || means.
275 * (It always returns 0 or 1, not one of it's paramaters.)
277 * Revision 1.40 2001/10/26 17:37:55 oes
278 * - Re-enabled Netscape 200/404 bug workaround in block_url():
279 * - Removed OS/2 special case
280 * - Made block_url() independant from sed() having been run
281 * - Made trust_url independant from sed() having been run
282 * - Made is_imageurl independant from sed() having been run.
283 * It now checks User-Agent: and Accept: by itself.
286 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
287 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
288 * threads to call select() simultaneously. So, it's time to do a real, live,
289 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
290 * (native). Both versions will work, but using __OS2__ offers multi-threading.
292 * Revision 1.38 2001/10/23 21:32:33 jongfoster
293 * Adding error-checking to selected functions
295 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
296 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
297 * filters.c. Added a FIXME in front of the offending code. I'll gladly
298 * put in a better/more robust fix for all parties if one is presented...
299 * It seems that just returning 200 instead of 404 would pretty much fix
300 * it for everyone, but I don't know all the history of the problem.
302 * Revision 1.36 2001/10/10 16:44:16 oes
303 * Added match_portlist function
305 * Revision 1.35 2001/10/07 15:41:23 oes
306 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
308 * New function remove_chunked_transfer_coding that strips chunked
309 * transfer coding to plain and is called by pcrs_filter_response
310 * and gif_deanimate_response if neccessary
312 * Improved handling of zero-change re_filter runs
314 * pcrs_filter_response and gif_deanimate_response now remove
315 * chunked transfer codeing before processing the body.
317 * Revision 1.34 2001/09/20 15:49:36 steudten
319 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
320 * See cgi.c fill_template().
322 * Revision 1.33 2001/09/16 17:05:14 jongfoster
323 * Removing unused #include showarg.h
325 * Revision 1.32 2001/09/16 13:21:27 jongfoster
326 * Changes to use new list functions.
328 * Revision 1.31 2001/09/16 11:38:02 jongfoster
329 * Splitting fill_template() into 2 functions:
330 * template_load() loads the file
331 * template_fill() performs the PCRS regexps.
332 * This is because the CGI edit interface has a "table row"
333 * template which is used many times in the page - this
334 * change means it's only loaded from disk once.
336 * Revision 1.30 2001/09/16 11:00:10 jongfoster
337 * New function alloc_http_response, for symmetry with free_http_response
339 * Revision 1.29 2001/09/13 23:32:40 jongfoster
340 * Moving image data to cgi.c rather than cgi.h
341 * Fixing a GPF under Win32 (and any other OS that protects global
342 * constants from being written to).
344 * Revision 1.28 2001/09/10 10:18:51 oes
345 * Silenced compiler warnings
347 * Revision 1.27 2001/08/05 16:06:20 jongfoster
348 * Modifiying "struct map" so that there are now separate header and
349 * "map_entry" structures. This means that functions which modify a
350 * map no longer need to return a pointer to the modified map.
351 * Also, it no longer reverses the order of the entries (which may be
352 * important with some advanced template substitutions).
354 * Revision 1.26 2001/07/30 22:08:36 jongfoster
355 * Tidying up #defines:
356 * - All feature #defines are now of the form FEATURE_xxx
357 * - Permanently turned off WIN_GUI_EDIT
358 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
360 * Revision 1.25 2001/07/26 10:09:46 oes
361 * Made browser detection a little less naive
363 * Revision 1.24 2001/07/25 17:22:51 oes
364 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
366 * Revision 1.23 2001/07/23 13:40:12 oes
367 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
369 * Revision 1.22 2001/07/18 12:29:34 oes
370 * - Made gif_deanimate_response respect
371 * csp->action->string[ACTION_STRING_DEANIMATE]
372 * - Logging cosmetics
374 * Revision 1.21 2001/07/13 13:59:53 oes
375 * - Introduced gif_deanimate_response which shares the
376 * generic content modification interface of pcrs_filter_response
377 * and acts as a wrapper to deanimate.c:gif_deanimate()
378 * - Renamed re_process_buffer to pcrs_filter_response
379 * - pcrs_filter_response now returns NULL on failiure
380 * - Removed all #ifdef PCRS
382 * Revision 1.20 2001/07/01 17:01:04 oes
383 * Added comments and missing return statement in is_untrusted_url()
385 * Revision 1.19 2001/06/29 21:45:41 oes
386 * Indentation, CRLF->LF, Tab-> Space
388 * Revision 1.18 2001/06/29 13:27:38 oes
389 * - Cleaned up, renamed and reorderd functions
390 * and improved comments
393 * - Ported to CGI platform. Now delivers
394 * http_response or NULL
395 * - Unified HTML and GIF generation (moved image detection
396 * and GIF generation here from jcc.c:chat())
397 * - Fixed HTTP status to:
398 * - 403 (Forbidden) for the "blocked" HTML message
399 * - 200 (OK) for GIF answers
400 * - 302 (Redirect) for redirect to GIF
403 * - Ported to CGI platform. Now delivers
404 * http_response or NULL
405 * - Separated detection of untrusted URL into
406 * (bool)is_untrusted_url
407 * - Added enforcement of untrusted requests
409 * - Moved redirect_url() from cgi.c to here
410 * and ported it to the CGI platform
412 * - Removed logentry from cancelled commit
414 * Revision 1.17 2001/06/09 10:55:28 jongfoster
415 * Changing BUFSIZ ==> BUFFER_SIZE
417 * Revision 1.16 2001/06/07 23:10:26 jongfoster
418 * Allowing unanchored domain patterns to back off and retry
419 * if they partially match. Optimized right-anchored patterns.
420 * Moving ACL and forward files into config file.
421 * Replacing struct gateway with struct forward_spec
423 * Revision 1.15 2001/06/03 19:12:00 oes
424 * extracted-CGI relevant stuff
426 * Revision 1.14 2001/06/01 10:30:55 oes
427 * Added optional left-anchoring to domaincmp
429 * Revision 1.13 2001/05/31 21:21:30 jongfoster
430 * Permissionsfile / actions file changes:
431 * - Changed "permission" to "action" throughout
432 * - changes to file format to allow string parameters
433 * - Moved helper functions to actions.c
435 * Revision 1.12 2001/05/31 17:35:20 oes
437 * - Enhanced domain part globbing with infix and prefix asterisk
438 * matching and optional unanchored operation
440 * Revision 1.11 2001/05/29 11:53:23 oes
441 * "See why" link added to "blocked" page
443 * Revision 1.10 2001/05/29 09:50:24 jongfoster
444 * Unified blocklist/imagelist/permissionslist.
445 * File format is still under discussion, but the internal changes
448 * Also modified interceptor behaviour:
449 * - We now intercept all URLs beginning with one of the following
450 * prefixes (and *only* these prefixes):
452 * * http://ijbswa.sf.net/config/
453 * * http://ijbswa.sourceforge.net/config/
454 * - New interceptors "home page" - go to http://i.j.b/ to see it.
455 * - Internal changes so that intercepted and fast redirect pages
456 * are not replaced with an image.
457 * - Interceptors now have the option to send a binary page direct
458 * to the client. (i.e. ijb-send-banner uses this)
459 * - Implemented show-url-info interceptor. (Which is why I needed
460 * the above interceptors changes - a typical URL is
461 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
462 * The previous mechanism would not have intercepted that, and
463 * if it had been intercepted then it then it would have replaced
466 * Revision 1.9 2001/05/27 22:17:04 oes
468 * - re_process_buffer no longer writes the modified buffer
469 * to the client, which was very ugly. It now returns the
470 * buffer, which it is then written by chat.
472 * - content_length now adjusts the Content-Length: header
473 * for modified documents rather than crunch()ing it.
474 * (Length info in csp->content_length, which is 0 for
475 * unmodified documents)
477 * - For this to work, sed() is called twice when filtering.
479 * Revision 1.8 2001/05/26 17:13:28 jongfoster
480 * Filled in a function comment.
482 * Revision 1.7 2001/05/26 15:26:15 jongfoster
483 * ACL feature now provides more security by immediately dropping
484 * connections from untrusted hosts.
486 * Revision 1.6 2001/05/26 00:28:36 jongfoster
487 * Automatic reloading of config file.
488 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
489 * Most of the global variables have been moved to a new
490 * struct configuration_spec, accessed through csp->config->globalname
491 * Most of the globals remaining are used by the Win32 GUI.
493 * Revision 1.5 2001/05/25 22:34:30 jongfoster
496 * Revision 1.4 2001/05/22 18:46:04 oes
498 * - Enabled filtering banners by size rather than URL
499 * by adding patterns that replace all standard banner
500 * sizes with the "Junkbuster" gif to the re_filterfile
502 * - Enabled filtering WebBugs by providing a pattern
503 * which kills all 1x1 images
505 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
506 * which is selected by the (nonstandard and therefore
507 * capital) letter 'U' in the option string.
508 * It causes the quantifiers to be ungreedy by default.
509 * Appending a ? turns back to greedy (!).
511 * - Added a new interceptor ijb-send-banner, which
512 * sends back the "Junkbuster" gif. Without imagelist or
513 * MSIE detection support, or if tinygif = 1, or the
514 * URL isn't recognized as an imageurl, a lame HTML
515 * explanation is sent instead.
517 * - Added new feature, which permits blocking remote
518 * script redirects and firing back a local redirect
520 * The feature is conditionally compiled, i.e. it
521 * can be disabled with --disable-fast-redirects,
522 * plus it must be activated by a "fast-redirects"
523 * line in the config file, has its own log level
524 * and of course wants to be displayed by show-proxy-args
525 * Note: Boy, all the #ifdefs in 1001 locations and
526 * all the fumbling with configure.in and acconfig.h
527 * were *way* more work than the feature itself :-(
529 * - Because a generic redirect template was needed for
530 * this, tinygif = 3 now uses the same.
532 * - Moved GIFs, and other static HTTP response templates
537 * - Removed some >400 CRs again (Jon, you really worked
540 * Revision 1.3 2001/05/20 16:44:47 jongfoster
541 * Removing last hardcoded Junkbusters.com URLs.
543 * Revision 1.2 2001/05/20 01:21:20 jongfoster
544 * Version 2.9.4 checkin.
545 * - Merged popupfile and cookiefile, and added control over PCRS
546 * filtering, in new "permissionsfile".
547 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
548 * file error you now get a message box (in the Win32 GUI) rather
549 * than the program exiting with no explanation.
550 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
552 * - Removed tabs from "config"
553 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
554 * - Bumped up version number.
556 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
557 * Initial import of version 2.9.3 source tree
560 *********************************************************************/
566 #include <sys/types.h>
575 #endif /* ndef __OS2__ */
576 #include <netinet/in.h>
578 #include <winsock2.h>
579 #endif /* ndef _WIN32 */
583 #endif /* def __OS2__ */
591 #include "jbsockets.h"
592 #include "miscutil.h"
596 #include "deanimate.h"
597 #include "urlmatch.h"
604 const char filters_h_rcs[] = FILTERS_H_VERSION;
606 /* Fix a problem with Solaris. There should be no effect on other
608 * Solaris's isspace() is a macro which uses it's argument directly
609 * as an array index. Therefore we need to make sure that high-bit
610 * characters generate +ve values, and ideally we also want to make
611 * the argument match the declared parameter type of "int".
613 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
617 /*********************************************************************
619 * Function : block_acl
621 * Description : Block this request?
622 * Decide yes or no based on ACL file.
625 * 1 : dst = The proxy or gateway address this is going to.
626 * Or NULL to check all possible targets.
627 * 2 : csp = Current client state (buffers, headers, etc...)
628 * Also includes the client IP address.
630 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
632 *********************************************************************/
633 int block_acl(struct access_control_addr *dst, struct client_state *csp)
635 struct access_control_list *acl = csp->config->acl;
637 /* if not using an access control list, then permit the connection */
643 /* search the list */
646 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
650 /* Just want to check if they have any access */
651 if (acl->action == ACL_PERMIT)
656 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
657 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
659 if (acl->action == ACL_PERMIT)
677 /*********************************************************************
679 * Function : acl_addr
681 * Description : Called from `load_config' to parse an ACL address.
684 * 1 : aspec = String specifying ACL address.
685 * 2 : aca = struct access_control_addr to fill in.
687 * Returns : 0 => Ok, everything else is an error.
689 *********************************************************************/
690 int acl_addr(const char *aspec, struct access_control_addr *aca)
695 char *acl_spec = NULL;
701 * Use a temporary acl spec copy so we can log
702 * the unmodified original in case of parse errors.
704 acl_spec = strdup(aspec);
705 if (acl_spec == NULL)
707 /* XXX: This will be logged as parse error. */
711 if ((p = strchr(acl_spec, '/')) != NULL)
714 if (ijb_isdigit(*p) == 0)
719 masklength = atoi(p);
722 if ((masklength < 0) || (masklength > 32))
728 if ((p = strchr(acl_spec, ':')) != NULL)
733 port = strtol(p, &endptr, 10);
735 if (port <= 0 || port > 65535 || *endptr != '\0')
742 aca->port = (unsigned long)port;
744 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
747 if (aca->addr == INADDR_NONE)
749 /* XXX: This will be logged as parse error. */
753 /* build the netmask */
755 for (i=1; i <= masklength ; i++)
757 aca->mask |= (1 << (32 - i));
760 /* now mask off the host portion of the ip address
761 * (i.e. save on the network portion of the address).
763 aca->addr = aca->addr & aca->mask;
768 #endif /* def FEATURE_ACL */
771 /*********************************************************************
773 * Function : match_portlist
775 * Description : Check if a given number is covered by a comma
776 * separated list of numbers and ranges (a,b-c,d,..)
779 * 1 : portlist = String with list
780 * 2 : port = port to check
782 * Returns : 0 => no match
785 *********************************************************************/
786 int match_portlist(const char *portlist, int port)
788 char *min, *max, *next, *portlist_copy;
790 min = next = portlist_copy = strdup(portlist);
793 * Zero-terminate first item and remember offset for next
795 if (NULL != (next = strchr(portlist_copy, (int) ',')))
801 * Loop through all items, checking for match
805 if (NULL == (max = strchr(min, (int) '-')))
808 * No dash, check for equality
810 if (port == atoi(min))
819 * This is a range, so check if between min and max,
820 * or, if max was omitted, between min and 65K
823 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
837 * Zero-terminate next item and remember offset for n+1
839 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
851 /*********************************************************************
853 * Function : block_url
855 * Description : Called from `chat'. Check to see if we need to block this.
858 * 1 : csp = Current client state (buffers, headers, etc...)
860 * Returns : NULL => unblocked, else HTTP block response
862 *********************************************************************/
863 struct http_response *block_url(struct client_state *csp)
865 struct http_response *rsp;
866 const char *new_content_type = NULL;
869 * If it's not blocked, don't block it ;-)
871 if ((csp->action->flags & ACTION_BLOCK) == 0)
875 if (csp->action->flags & ACTION_REDIRECT)
877 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
880 * Else, prepare a response
882 if (NULL == (rsp = alloc_http_response()))
884 return cgi_error_memory();
888 * If it's an image-url, send back an image or redirect
889 * as specified by the relevant +image action
891 #ifdef FEATURE_IMAGE_BLOCKING
892 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
896 /* determine HOW images should be blocked */
897 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
899 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
901 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
903 #if 1 /* Two alternative strategies, use this one for now: */
905 /* and handle accordingly: */
906 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
908 rsp->status = strdup("403 Request blocked by Privoxy");
909 if (rsp->status == NULL)
911 free_http_response(rsp);
912 return cgi_error_memory();
914 rsp->body = bindup(image_pattern_data, image_pattern_length);
915 if (rsp->body == NULL)
917 free_http_response(rsp);
918 return cgi_error_memory();
920 rsp->content_length = image_pattern_length;
922 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
924 free_http_response(rsp);
925 return cgi_error_memory();
929 else if (0 == strcmpic(p, "blank"))
931 rsp->status = strdup("403 Request blocked by Privoxy");
932 if (rsp->status == NULL)
934 free_http_response(rsp);
935 return cgi_error_memory();
937 rsp->body = bindup(image_blank_data, image_blank_length);
938 if (rsp->body == NULL)
940 free_http_response(rsp);
941 return cgi_error_memory();
943 rsp->content_length = image_blank_length;
945 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
947 free_http_response(rsp);
948 return cgi_error_memory();
954 rsp->status = strdup("302 Local Redirect from Privoxy");
955 if (rsp->status == NULL)
957 free_http_response(rsp);
958 return cgi_error_memory();
961 if (enlist_unique_header(rsp->headers, "Location", p))
963 free_http_response(rsp);
964 return cgi_error_memory();
968 #else /* Following code is disabled for now */
970 /* and handle accordingly: */
971 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
973 p = CGI_PREFIX "send-banner?type=pattern";
975 else if (0 == strcmpic(p, "blank"))
977 p = CGI_PREFIX "send-banner?type=blank";
979 rsp->status = strdup("302 Local Redirect from Privoxy");
980 if (rsp->status == NULL)
982 free_http_response(rsp);
983 return cgi_error_memory();
986 if (enlist_unique_header(rsp->headers, "Location", p))
988 free_http_response(rsp);
989 return cgi_error_memory();
991 #endif /* Preceeding code is disabled for now */
993 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
996 * Send empty document.
998 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
1001 rsp->body = strdup(" ");
1002 rsp->content_length = 1;
1004 rsp->status = strdup("403 Request blocked by Privoxy");
1005 if (rsp->status == NULL)
1007 free_http_response(rsp);
1008 return cgi_error_memory();
1010 if (new_content_type != 0)
1012 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1013 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1015 free_http_response(rsp);
1016 return cgi_error_memory();
1021 #endif /* def FEATURE_IMAGE_BLOCKING */
1024 * Else, generate an HTML "blocked" message:
1028 struct map * exports;
1032 * Workaround for stupid Netscape bug which prevents
1033 * pages from being displayed if loading a referenced
1034 * JavaScript or style sheet fails. So make it appear
1035 * as if it succeeded.
1037 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1038 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1039 && !strstr(p, "Gecko") /* save Mozilla, */
1040 && !strstr(p, "compatible") /* MSIE */
1041 && !strstr(p, "Opera")) /* and Opera. */
1043 rsp->status = strdup("200 Request for blocked URL");
1047 rsp->status = strdup("403 Request for blocked URL");
1050 if (rsp->status == NULL)
1052 free_http_response(rsp);
1053 return cgi_error_memory();
1056 exports = default_exports(csp, NULL);
1057 if (exports == NULL)
1059 free_http_response(rsp);
1060 return cgi_error_memory();
1063 #ifdef FEATURE_FORCE_LOAD
1064 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1066 * Export the force conditional block killer if
1068 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1069 * - Privoxy is configured to enforce blocks, or
1070 * - it's a CONNECT request and enforcing wouldn't work anyway.
1072 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1073 || (0 == strcmpic(csp->http->gpc, "connect")))
1074 #endif /* ndef FEATURE_FORCE_LOAD */
1076 err = map_block_killer(exports, "force-support");
1079 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1080 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1081 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1082 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1087 free_http_response(rsp);
1088 return cgi_error_memory();
1091 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1094 free_http_response(rsp);
1095 return cgi_error_memory();
1098 rsp->reason = RSP_REASON_BLOCKED;
1100 return finish_http_response(csp, rsp);
1105 #ifdef FEATURE_TRUST
1106 /*********************************************************************
1108 * Function : trust_url FIXME: I should be called distrust_url
1110 * Description : Calls is_untrusted_url to determine if the URL is trusted
1111 * and if not, returns a HTTP 403 response with a reject message.
1114 * 1 : csp = Current client state (buffers, headers, etc...)
1116 * Returns : NULL => trusted, else http_response.
1118 *********************************************************************/
1119 struct http_response *trust_url(struct client_state *csp)
1121 struct http_response *rsp;
1122 struct map * exports;
1123 char buf[BUFFER_SIZE];
1125 struct url_spec **tl;
1130 * Don't bother to work on trusted URLs
1132 if (!is_untrusted_url(csp))
1138 * Else, prepare a response:
1140 if (NULL == (rsp = alloc_http_response()))
1142 return cgi_error_memory();
1145 rsp->status = strdup("403 Request blocked by Privoxy");
1146 exports = default_exports(csp, NULL);
1147 if (exports == NULL || rsp->status == NULL)
1149 free_http_response(rsp);
1150 return cgi_error_memory();
1154 * Export the protocol, host, port, and referrer information
1156 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1157 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1158 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1160 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1162 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1166 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1172 free_http_response(rsp);
1173 return cgi_error_memory();
1177 * Export the trust list
1180 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1182 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1183 string_append(&p, buf);
1185 err = map(exports, "trusted-referrers", 1, p, 0);
1190 free_http_response(rsp);
1191 return cgi_error_memory();
1195 * Export the trust info, if available
1197 if (csp->config->trust_info->first)
1199 struct list_entry *l;
1202 for (l = csp->config->trust_info->first; l ; l = l->next)
1204 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1205 string_append(&p, buf);
1207 err = map(exports, "trust-info", 1, p, 0);
1211 err = map_block_killer(exports, "have-trust-info");
1217 free_http_response(rsp);
1218 return cgi_error_memory();
1222 * Export the force conditional block killer if
1224 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1225 * - Privoxy is configured to enforce blocks, or
1226 * - it's a CONNECT request and enforcing wouldn't work anyway.
1228 #ifdef FEATURE_FORCE_LOAD
1229 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1230 || (0 == strcmpic(csp->http->gpc, "connect")))
1232 err = map_block_killer(exports, "force-support");
1236 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1238 #else /* ifndef FEATURE_FORCE_LOAD */
1239 err = map_block_killer(exports, "force-support");
1240 #endif /* ndef FEATURE_FORCE_LOAD */
1245 free_http_response(rsp);
1246 return cgi_error_memory();
1250 * Build the response
1252 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1255 free_http_response(rsp);
1256 return cgi_error_memory();
1258 rsp->reason = RSP_REASON_UNTRUSTED;
1260 return finish_http_response(csp, rsp);
1262 #endif /* def FEATURE_TRUST */
1265 /*********************************************************************
1267 * Function : compile_dynamic_pcrs_job_list
1269 * Description : Compiles a dynamic pcrs job list (one with variables
1270 * resolved at request time)
1273 * 1 : csp = Current client state (buffers, headers, etc...)
1274 * 2 : b = The filter list to compile
1276 * Returns : NULL in case of errors, otherwise the
1279 *********************************************************************/
1280 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1282 struct list_entry *pattern;
1283 pcrs_job *job_list = NULL;
1284 pcrs_job *dummy = NULL;
1285 pcrs_job *lastjob = NULL;
1288 const struct pcrs_variable variables[] =
1290 {"url", csp->http->url, 1},
1291 {"path", csp->http->path, 1},
1292 {"host", csp->http->host, 1},
1293 {"origin", csp->ip_addr_str, 1},
1297 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1299 assert(pattern->str != NULL);
1301 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1305 log_error(LOG_LEVEL_ERROR,
1306 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1307 pattern->str, b->name, pcrs_strerror(error));
1312 if (error == PCRS_WARN_TRUNCATION)
1314 log_error(LOG_LEVEL_ERROR,
1315 "At least one of the variables in \'%s\' had to "
1316 "be truncated before compilation", pattern->str);
1318 if (job_list == NULL)
1324 lastjob->next = dummy;
1334 /*********************************************************************
1336 * Function : rewrite_url
1338 * Description : Rewrites a URL with a single pcrs command
1339 * and returns the result if it differs from the
1340 * original and isn't obviously invalid.
1343 * 1 : old_url = URL to rewrite.
1344 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1347 * Returns : NULL if the pcrs_command didn't change the url, or
1348 * the result of the modification.
1350 *********************************************************************/
1351 char *rewrite_url(char *old_url, const char *pcrs_command)
1353 char *new_url = NULL;
1357 assert(pcrs_command);
1359 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1363 log_error(LOG_LEVEL_REDIRECTS,
1364 "pcrs command \"%s\" didn't change \"%s\".",
1365 pcrs_command, old_url);
1370 log_error(LOG_LEVEL_REDIRECTS,
1371 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1372 pcrs_command, old_url, pcrs_strerror(hits));
1375 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1377 log_error(LOG_LEVEL_ERROR,
1378 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1379 "but the result doesn't look like a valid URL and will be ignored.",
1380 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1385 log_error(LOG_LEVEL_REDIRECTS,
1386 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1387 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1395 #ifdef FEATURE_FAST_REDIRECTS
1396 /*********************************************************************
1398 * Function : get_last_url
1400 * Description : Search for the last URL inside a string.
1401 * If the string already is a URL, it will
1402 * be the first URL found.
1405 * 1 : subject = the string to check
1406 * 2 : redirect_mode = +fast-redirect{} mode
1408 * Returns : NULL if no URL was found, or
1409 * the last URL found.
1411 *********************************************************************/
1412 char *get_last_url(char *subject, const char *redirect_mode)
1414 char *new_url = NULL;
1418 assert(redirect_mode);
1420 subject = strdup(subject);
1421 if (subject == NULL)
1423 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1427 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1429 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1430 new_url = url_decode(subject);
1431 if (new_url != NULL)
1438 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1442 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1445 * Find the last URL encoded in the request
1448 while ((tmp = strstr(tmp, "http://")) != NULL)
1452 tmp = (new_url != NULL) ? new_url : subject;
1453 while ((tmp = strstr(tmp, "https://")) != NULL)
1458 if ((new_url != NULL)
1459 && ( (new_url != subject)
1460 || (0 == strncmpic(subject, "http://", 7))
1461 || (0 == strncmpic(subject, "https://", 8))
1465 * Return new URL if we found a redirect
1466 * or if the subject already was a URL.
1468 * The second case makes sure that we can
1469 * chain get_last_url after another redirection check
1470 * (like rewrite_url) without losing earlier redirects.
1472 new_url = strdup(new_url);
1481 #endif /* def FEATURE_FAST_REDIRECTS */
1484 /*********************************************************************
1486 * Function : redirect_url
1488 * Description : Checks if Privoxy should answer the request with
1489 * a HTTP redirect and generates the redirect if
1493 * 1 : csp = Current client state (buffers, headers, etc...)
1495 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1497 *********************************************************************/
1498 struct http_response *redirect_url(struct client_state *csp)
1500 struct http_response *rsp;
1501 #ifdef FEATURE_FAST_REDIRECTS
1503 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1504 * as compile-time option? The user can easily disable
1505 * it in his action file.
1507 char * redirect_mode;
1508 #endif /* def FEATURE_FAST_REDIRECTS */
1509 char *old_url = NULL;
1510 char *new_url = NULL;
1511 char *redirection_string;
1513 if ((csp->action->flags & ACTION_REDIRECT))
1515 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1518 * If the redirection string begins with 's',
1519 * assume it's a pcrs command, otherwise treat it as
1520 * properly formatted URL and use it for the redirection
1523 * According to RFC 2616 section 14.30 the URL
1524 * has to be absolute and if the user tries:
1525 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1526 * she would get undefined results anyway.
1530 if (*redirection_string == 's')
1532 old_url = csp->http->url;
1533 new_url = rewrite_url(old_url, redirection_string);
1537 log_error(LOG_LEVEL_REDIRECTS,
1538 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1539 redirection_string);
1540 new_url = strdup(redirection_string);
1544 #ifdef FEATURE_FAST_REDIRECTS
1545 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1547 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1550 * If it exists, use the previously rewritten URL as input
1551 * otherwise just use the old path.
1553 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1554 new_url = get_last_url(old_url, redirect_mode);
1559 * Disable redirect checkers, so that they
1560 * will be only run more than once if the user
1561 * also enables them through tags.
1563 * From a performance point of view
1564 * it doesn't matter, but the duplicated
1565 * log messages are annoying.
1567 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1568 #endif /* def FEATURE_FAST_REDIRECTS */
1569 csp->action->flags &= ~ACTION_REDIRECT;
1571 /* Did any redirect action trigger? */
1574 if (0 == strcmpic(new_url, csp->http->url))
1576 log_error(LOG_LEVEL_ERROR,
1577 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1578 csp->http->url, new_url);
1583 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1585 if (NULL == (rsp = alloc_http_response()))
1588 return cgi_error_memory();
1591 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1592 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1595 free_http_response(rsp);
1596 return cgi_error_memory();
1598 rsp->reason = RSP_REASON_REDIRECTED;
1601 return finish_http_response(csp, rsp);
1605 /* Only reached if no redirect is required */
1611 #ifdef FEATURE_IMAGE_BLOCKING
1612 /*********************************************************************
1614 * Function : is_imageurl
1616 * Description : Given a URL, decide whether it is an image or not,
1617 * using either the info from a previous +image action
1618 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1619 * is MSIE and not on a Mac, tell from the browser's accept
1623 * 1 : csp = Current client state (buffers, headers, etc...)
1625 * Returns : True (nonzero) if URL is an image, false (0)
1628 *********************************************************************/
1629 int is_imageurl(const struct client_state *csp)
1631 #ifdef FEATURE_IMAGE_DETECT_MSIE
1634 tmp = get_header_value(csp->headers, "User-Agent:");
1635 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1637 tmp = get_header_value(csp->headers, "Accept:");
1638 if (tmp && strstr(tmp, "image/gif"))
1640 /* Client will accept HTML. If this seems counterintuitive,
1650 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1652 return ((csp->action->flags & ACTION_IMAGE) != 0);
1655 #endif /* def FEATURE_IMAGE_BLOCKING */
1658 #ifdef FEATURE_TRUST
1659 /*********************************************************************
1661 * Function : is_untrusted_url
1663 * Description : Should we "distrust" this URL (and block it)?
1665 * Yes if it matches a line in the trustfile, or if the
1666 * referrer matches a line starting with "+" in the
1671 * 1 : csp = Current client state (buffers, headers, etc...)
1673 * Returns : 0 => trusted, 1 => untrusted
1675 *********************************************************************/
1676 int is_untrusted_url(const struct client_state *csp)
1678 struct file_list *fl;
1679 struct block_spec *b;
1680 struct url_spec **trusted_url;
1681 struct http_request rhttp[1];
1682 const char * referer;
1686 * If we don't have a trustlist, we trust everybody
1688 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1693 memset(rhttp, '\0', sizeof(*rhttp));
1696 * Do we trust the request URL itself?
1698 for (b = b->next; b ; b = b->next)
1700 if (url_match(b->url, csp->http))
1706 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1708 /* no referrer was supplied */
1714 * If not, do we maybe trust its referrer?
1716 err = parse_http_url(referer, rhttp, csp);
1722 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1724 if (url_match(*trusted_url, rhttp))
1726 /* if the URL's referrer is from a trusted referrer, then
1727 * add the target spec to the trustfile as an unblocked
1728 * domain and return 0 (which means it's OK).
1733 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1737 char * new_entry = strdup("~");
1739 string_append(&new_entry, csp->http->hostport);
1741 path = csp->http->path;
1742 if ( (path[0] == '/')
1744 && ((path_end = strchr(path + 2, '/')) != NULL))
1746 /* since this path points into a user's home space
1747 * be sure to include this spec in the trustfile.
1749 int path_len = path_end - path; /* save offset */
1750 path = strdup(path); /* Copy string */
1753 path_end = path + path_len; /* regenerate ptr to new buffer */
1754 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1756 string_join(&new_entry, path);
1760 * Give a reason for generating this entry.
1762 string_append(&new_entry, " # Trusted referrer was: ");
1763 string_append(&new_entry, referer);
1765 if (new_entry != NULL)
1767 if (-1 == fprintf(fp, "%s\n", new_entry))
1769 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1770 new_entry, csp->config->trustfile);
1776 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1777 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1784 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1785 csp->http->hostport, csp->config->trustfile);
1793 #endif /* def FEATURE_TRUST */
1796 /*********************************************************************
1798 * Function : pcrs_filter_response
1800 * Description : Execute all text substitutions from all applying
1801 * +filter actions on the text buffer that's been accumulated
1802 * in csp->iob->buf. If this changes the contents, set
1803 * csp->content_length to the modified size and raise the
1804 * CSP_FLAG_MODIFIED flag.
1806 * XXX: Currently pcrs_filter_response is also responsible
1807 * for dechunking and decompressing. Both should be
1808 * done in separate functions so other content modifiers
1809 * profit as well, even if pcrs filtering is disabled.
1812 * 1 : csp = Current client state (buffers, headers, etc...)
1814 * Returns : a pointer to the (newly allocated) modified buffer.
1815 * or NULL if there were no hits or something went wrong
1817 *********************************************************************/
1818 char *pcrs_filter_response(struct client_state *csp)
1821 size_t size, prev_size;
1823 char *old = csp->iob->cur, *new = NULL;
1826 struct file_list *fl;
1827 struct re_filterfile_spec *b;
1828 struct list_entry *filtername;
1830 int i, found_filters = 0;
1835 if (csp->iob->cur >= csp->iob->eod)
1839 size = (size_t)(csp->iob->eod - csp->iob->cur);
1842 * Need to check the set of re_filterfiles...
1844 for (i = 0; i < MAX_AF_FILES; i++)
1857 if (0 == found_filters)
1859 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1864 * If the body has a "chunked" transfer-encoding,
1865 * get rid of it first, adjusting size and iob->eod
1867 if (csp->flags & CSP_FLAG_CHUNKED)
1869 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1870 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1874 csp->iob->eod = csp->iob->cur + size;
1875 csp->flags |= CSP_FLAG_MODIFIED;
1880 * If the body has a compressed transfer-encoding,
1881 * uncompress it first, adjusting size and iob->eod.
1882 * Note that decompression occurs after de-chunking.
1884 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1886 /* Notice that we at least tried to decompress. */
1887 if (JB_ERR_OK != decompress_iob(csp))
1890 * We failed to decompress the data; there's no point
1891 * in continuing since we can't filter.
1893 * XXX: Actually the Accept-Encoding header may
1894 * just be incorrect in which case we could continue
1897 * Unset CT_GZIP and CT_DEFLATE to remember not
1898 * to modify the Content-Encoding header later.
1900 csp->content_type &= ~CT_GZIP;
1901 csp->content_type &= ~CT_DEFLATE;
1906 * Decompression gives us a completely new iob,
1907 * so we need to update.
1909 size = (size_t)(csp->iob->eod - csp->iob->cur);
1910 old = csp->iob->cur;
1912 csp->flags |= CSP_FLAG_MODIFIED;
1913 csp->content_type &= ~CT_TABOO;
1917 for (i = 0; i < MAX_AF_FILES; i++)
1920 if ((NULL == fl) || (NULL == fl->f))
1923 * Either there are no filter files
1924 * left, or this filter file just
1925 * contains no valid filters.
1927 * Continue to be sure we don't miss
1928 * valid filter files that are chained
1929 * after empty or invalid ones.
1934 * For all applying +filter actions, look if a filter by that
1935 * name exists and if yes, execute it's pcrs_joblist on the
1938 for (b = fl->f; b; b = b->next)
1940 if (b->type != FT_CONTENT_FILTER)
1942 /* Skip header filters */
1946 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1947 filtername ; filtername = filtername->next)
1949 if (strcmp(b->name, filtername->str) == 0)
1951 int current_hits = 0; /* Number of hits caused by this filter */
1952 int job_number = 0; /* Which job we're currently executing */
1953 int job_hits = 0; /* How many hits the current job caused */
1954 pcrs_job *joblist = b->joblist;
1956 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1958 if (NULL == joblist)
1960 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1965 /* Apply all jobs from the joblist */
1966 for (job = joblist; NULL != job; job = job->next)
1969 job_hits = pcrs_execute(job, old, size, &new, &size);
1974 * That went well. Continue filtering
1975 * and use the result of this job as
1976 * input for the next one.
1978 current_hits += job_hits;
1979 if (old != csp->iob->cur)
1988 * This job caused an unexpected error. Inform the user
1989 * and skip the rest of the jobs in this filter. We could
1990 * continue with the next job, but usually the jobs
1991 * depend on each other or are similar enough to
1992 * fail for the same reason.
1994 * At the moment our pcrs expects the error codes of pcre 3.4,
1995 * but newer pcre versions can return additional error codes.
1996 * As a result pcrs_strerror()'s error message might be
1997 * "Unknown error ...", therefore we print the numerical value
2000 * XXX: Is this important enough for LOG_LEVEL_ERROR or
2001 * should we use LOG_LEVEL_RE_FILTER instead?
2003 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
2004 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2009 if (b->dynamic) pcrs_free_joblist(joblist);
2011 log_error(LOG_LEVEL_RE_FILTER,
2012 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2013 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2015 hits += current_hits;
2022 * If there were no hits, destroy our copy and let
2023 * chat() use the original in csp->iob
2031 csp->flags |= CSP_FLAG_MODIFIED;
2032 csp->content_length = size;
2040 /*********************************************************************
2042 * Function : gif_deanimate_response
2044 * Description : Deanimate the GIF image that has been accumulated in
2045 * csp->iob->buf, set csp->content_length to the modified
2046 * size and raise the CSP_FLAG_MODIFIED flag.
2049 * 1 : csp = Current client state (buffers, headers, etc...)
2051 * Returns : a pointer to the (newly allocated) modified buffer.
2052 * or NULL in case something went wrong.
2054 *********************************************************************/
2055 char *gif_deanimate_response(struct client_state *csp)
2057 struct binbuffer *in, *out;
2061 size = (size_t)(csp->iob->eod - csp->iob->cur);
2064 * If the body has a "chunked" transfer-encoding,
2065 * get rid of it first, adjusting size and iob->eod
2067 if (csp->flags & CSP_FLAG_CHUNKED)
2069 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2070 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2074 csp->iob->eod = csp->iob->cur + size;
2075 csp->flags |= CSP_FLAG_MODIFIED;
2078 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2079 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2081 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2085 in->buffer = csp->iob->cur;
2088 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2090 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2097 if ((int)size == out->offset)
2099 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2103 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2105 csp->content_length = out->offset;
2106 csp->flags |= CSP_FLAG_MODIFIED;
2116 /*********************************************************************
2118 * Function : jpeg_inspect_response
2123 * 1 : csp = Current client state (buffers, headers, etc...)
2125 * Returns : a pointer to the (newly allocated) modified buffer
2126 * or NULL in case something went wrong.
2128 *********************************************************************/
2129 char *jpeg_inspect_response(struct client_state *csp)
2131 struct binbuffer *in = NULL;
2132 struct binbuffer *out = NULL;
2136 size = (size_t)(csp->iob->eod - csp->iob->cur);
2139 * If the body has a "chunked" transfer-encoding,
2140 * get rid of it first, adjusting size and iob->eod
2142 if (csp->flags & CSP_FLAG_CHUNKED)
2144 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2145 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2149 csp->iob->eod = csp->iob->cur + size;
2150 csp->flags |= CSP_FLAG_MODIFIED;
2153 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2155 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2159 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2161 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2165 in->buffer = csp->iob->cur;
2169 * Calling jpeg_inspect has the side-effect of creating and
2170 * modifying the image buffer of "out" directly.
2172 if (jpeg_inspect(in, out))
2174 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2182 csp->content_length = out->offset;
2183 csp->flags |= CSP_FLAG_MODIFIED;
2193 /*********************************************************************
2195 * Function : remove_chunked_transfer_coding
2197 * Description : In-situ remove the "chunked" transfer coding as defined
2198 * in rfc2616 from a buffer.
2201 * 1 : buffer = Pointer to the text buffer
2202 * 2 : size = Number of bytes to be processed
2204 * Returns : The new size, i.e. the number of bytes from buffer which
2205 * are occupied by the stripped body, or 0 in case something
2208 *********************************************************************/
2209 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2212 unsigned int chunksize = 0;
2213 char *from_p, *to_p;
2216 from_p = to_p = buffer;
2218 if (sscanf(buffer, "%x", &chunksize) != 1)
2220 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2224 while (chunksize > 0)
2226 if (NULL == (from_p = strstr(from_p, "\r\n")))
2228 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2232 if ((newsize += chunksize) >= size)
2234 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2239 memmove(to_p, from_p, (size_t) chunksize);
2240 to_p = buffer + newsize;
2241 from_p += chunksize + 2;
2243 if (sscanf(from_p, "%x", &chunksize) != 1)
2245 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2250 /* FIXME: Should this get its own loglevel? */
2251 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2257 /*********************************************************************
2259 * Function : url_actions
2261 * Description : Gets the actions for this URL.
2264 * 1 : http = http_request request for blocked URLs
2265 * 2 : csp = Current client state (buffers, headers, etc...)
2269 *********************************************************************/
2270 void url_actions(struct http_request *http,
2271 struct client_state *csp)
2273 struct file_list *fl;
2274 struct url_actions *b;
2277 init_current_action(csp->action);
2279 for (i = 0; i < MAX_AF_FILES; i++)
2281 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2286 apply_url_actions(csp->action, http, b);
2293 /*********************************************************************
2295 * Function : apply_url_actions
2297 * Description : Applies a list of URL actions.
2300 * 1 : action = Destination.
2301 * 2 : http = Current URL
2302 * 3 : b = list of URL actions to apply
2306 *********************************************************************/
2307 void apply_url_actions(struct current_action_spec *action,
2308 struct http_request *http,
2309 struct url_actions *b)
2313 /* Should never happen */
2317 for (b = b->next; NULL != b; b = b->next)
2319 if (url_match(b->url, http))
2321 merge_current_action(action, b->action);
2327 /*********************************************************************
2329 * Function : get_forward_override_settings
2331 * Description : Returns forward settings as specified with the
2332 * forward-override{} action. forward-override accepts
2333 * forward lines similar to the one used in the
2334 * configuration file, but without the URL pattern.
2340 * in the configuration file can be replaced with
2341 * the action section:
2343 * {+forward-override{forward .}}
2347 * 1 : csp = Current client state (buffers, headers, etc...)
2349 * Returns : Pointer to forwarding structure in case of success.
2350 * Invalid syntax is fatal.
2352 *********************************************************************/
2353 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2355 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2356 char forward_settings[BUFFER_SIZE];
2357 char *http_parent = NULL;
2358 /* variable names were chosen for consistency reasons. */
2359 struct forward_spec *fwd = NULL;
2363 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2364 /* Should be enforced by load_one_actions_file() */
2365 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2367 /* Create a copy ssplit can modify */
2368 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2370 if (NULL != csp->fwd)
2373 * XXX: Currently necessary to prevent memory
2374 * leaks when the show-url-info cgi page is visited.
2376 unload_forward_spec(csp->fwd);
2380 * allocate a new forward node, valid only for
2381 * the lifetime of this request. Save its location
2382 * in csp as well, so sweep() can free it later on.
2384 fwd = csp->fwd = zalloc(sizeof(*fwd));
2387 log_error(LOG_LEVEL_FATAL,
2388 "can't allocate memory for forward-override{%s}", forward_override_line);
2389 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2392 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2393 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2395 fwd->type = SOCKS_NONE;
2397 /* Parse the parent HTTP proxy host:port */
2398 http_parent = vec[1];
2401 else if (vec_count == 3)
2403 char *socks_proxy = NULL;
2405 if (!strcasecmp(vec[0], "forward-socks4"))
2407 fwd->type = SOCKS_4;
2408 socks_proxy = vec[1];
2410 else if (!strcasecmp(vec[0], "forward-socks4a"))
2412 fwd->type = SOCKS_4A;
2413 socks_proxy = vec[1];
2416 if (NULL != socks_proxy)
2418 /* Parse the SOCKS proxy host[:port] */
2419 fwd->gateway_host = strdup(socks_proxy);
2421 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2423 *socks_proxy++ = '\0';
2424 fwd->gateway_port = strtol(socks_proxy, NULL, 0);
2427 if (fwd->gateway_port <= 0)
2429 fwd->gateway_port = 1080;
2432 http_parent = vec[2];
2436 if (NULL == http_parent)
2438 log_error(LOG_LEVEL_FATAL,
2439 "Invalid forward-override syntax in: %s", forward_override_line);
2440 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2443 /* Parse http forwarding settings */
2444 if (strcmp(http_parent, ".") != 0)
2446 fwd->forward_host = strdup(http_parent);
2448 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2450 *http_parent++ = '\0';
2451 fwd->forward_port = strtol(http_parent, NULL, 0);
2454 if (fwd->forward_port <= 0)
2456 fwd->forward_port = 8000;
2460 assert (NULL != fwd);
2462 log_error(LOG_LEVEL_CONNECT,
2463 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2468 /*********************************************************************
2470 * Function : forward_url
2472 * Description : Should we forward this to another proxy?
2474 * XXX: Should be changed to make use of csp->fwd.
2477 * 1 : http = http_request request for current URL
2478 * 2 : csp = Current client state (buffers, headers, etc...)
2480 * Returns : Pointer to forwarding information.
2482 *********************************************************************/
2483 const struct forward_spec * forward_url(struct http_request *http,
2484 struct client_state *csp)
2486 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2487 struct forward_spec *fwd = csp->config->forward;
2489 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2491 return get_forward_override_settings(csp);
2501 if (url_match(fwd->url, http))
2512 /*********************************************************************
2514 * Function : direct_response
2516 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2517 * request and if so, return a HTTP 501 to the client.
2519 * FIXME: I have a stupid name and I should handle the
2520 * requests properly. Still, what we do here is rfc-
2521 * compliant, whereas ignoring or forwarding are not.
2524 * 1 : csp = Current client state (buffers, headers, etc...)
2526 * Returns : http_response if , NULL if nonmatch or handler fail
2528 *********************************************************************/
2529 struct http_response *direct_response(struct client_state *csp)
2531 struct http_response *rsp;
2532 struct list_entry *p;
2534 if ((0 == strcmpic(csp->http->gpc, "trace"))
2535 || (0 == strcmpic(csp->http->gpc, "options")))
2537 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2539 if (!strncmpic("Max-Forwards:", p->str, 13))
2541 unsigned int max_forwards;
2544 * If it's a Max-Forwards value of zero,
2545 * we have to intercept the request.
2547 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2550 * FIXME: We could handle at least TRACE here,
2551 * but that would require a verbatim copy of
2552 * the request which we don't have anymore
2554 log_error(LOG_LEVEL_HEADER,
2555 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2558 /* Get mem for response or fail*/
2559 if (NULL == (rsp = alloc_http_response()))
2561 return cgi_error_memory();
2564 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2566 free_http_response(rsp);
2567 return cgi_error_memory();
2571 rsp->reason = RSP_REASON_UNSUPPORTED;
2573 return(finish_http_response(csp, rsp));