1 const char filters_rcs[] = "$Id: filters.c,v 1.91 2007/09/02 15:31:20 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 1.91 2007/09/02 15:31:20 fabiankeil
44 * Move match_portlist() from filter.c to urlmatch.c.
45 * It's used for url matching, not for filtering.
47 * Revision 1.90 2007/09/02 12:44:17 fabiankeil
48 * Remove newline at the end of a log_error() message.
50 * Revision 1.89 2007/08/05 13:42:23 fabiankeil
51 * #1763173 from Stefan Huehner: declare some more functions static.
53 * Revision 1.88 2007/06/01 16:41:11 fabiankeil
54 * Add forward-override{} to change the forwarding settings through
55 * action sections. This is mainly interesting to forward different
56 * clients differently (for example based on User-Agent or request
59 * Revision 1.87 2007/04/30 15:53:10 fabiankeil
60 * Make sure filters with dynamic jobs actually use them.
62 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
63 * - Introduce dynamic pcrs jobs that can resolve variables.
64 * - Don't run redirect functions more than once,
65 * unless they are activated more than once.
67 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
68 * - Log the content size after decompression in decompress_iob()
69 * instead of pcrs_filter_response().
71 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
72 * Use dedicated header filter actions instead of abusing "filter".
73 * Replace "filter-client-headers" and "filter-client-headers"
74 * with "server-header-filter" and "client-header-filter".
76 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
77 * New config option: enforce-blocks.
79 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
80 * - Fix port handling in acl_addr() and use a temporary acl spec
81 * copy so error messages don't contain a truncated version.
82 * - Log size of iob before and after decompression.
84 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
85 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
86 * - Hide the "Go there anyway" link for blocked CONNECT
87 * requests where going there anyway doesn't work anyway.
89 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
90 * - Save the reason for generating http_responses.
91 * - Block (+block) with status code 403 instead of 404.
92 * - Use a different kludge to remember a failed decompression.
94 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
95 * Search for Max-Forwards headers case-insensitive,
96 * don't generate the "501 unsupported" message for invalid
97 * Max-Forwards values and don't increase negative ones.
99 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
100 * - Add HEAD support to finish_http_response.
101 * - Add error favicon to internal HTML error messages.
103 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
104 * Mark *csp as immutable for is_untrusted_url()
105 * and is_imageurl(). Closes FR 1237736.
107 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
108 * Integrate a modified version of Wil Mahan's
109 * zlib patch (PR #895531).
111 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
112 * Fixed gcc43 conversion warnings,
113 * changed sprintf calls to snprintf.
115 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
116 * Adjust comment in pcrs_filter_response()
117 * to recent pcrs changes. Hohoho.
119 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
120 * Don't crash if pcre returns an error code
121 * that pcrs didn't expect. Fixes BR 1621173.
123 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
124 * Modified is_untrusted_url to complain in case of
125 * write errors and to give a reason when adding new
126 * entries to the trustfile. Closes FR 1097611.
128 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
129 * Skip empty filter files in pcrs_filter_response,
130 * but don't ignore the ones that come afterwards.
131 * Fixes parts of BR 1619208.
133 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
134 * Added some sanity checks for get_last_url().
135 * Fixed possible segfault caused by my last commit.
137 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
138 * Let get_last_url() catch https URLs as well.
140 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
141 * Make sure get_last_url() behaves like advertised
142 * and fast-redirects{} can be combined with redirect{}.
144 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
145 * Implemented +redirect{s@foo@bar@} to generate
146 * a redirect based on a rewritten version of the
149 * Revision 1.66 2006/09/23 13:26:38 roro
150 * Replace TABs by spaces in source code.
152 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
153 * Fix +redirect{}. Didn't work with -fast-redirects.
155 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
156 * Block requests for untrusted URLs with status
157 * code 403 instead of 200.
159 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
160 * Don't free p which is still in use and will be later
161 * freed by free_map(). Don't claim the referrer is unknown
162 * when the client didn't set one.
164 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
165 * Feature request 595948: Re-Filter logging in single line
167 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
168 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
170 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
171 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
172 * with what was really the latest development (the v_3_0_branch branch)
174 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
175 * Multiple filter file support
177 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
178 * A few memory leaks plugged (mostly on error paths)
180 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
181 * Add the ability to check jpeg images for invalid
182 * lengths of comment blocks. Defensive strategy
183 * against the exploit:
184 * Microsoft Security Bulletin MS04-028
185 * Buffer Overrun in JPEG Processing (GDI+) Could
186 * Allow Code Execution (833987)
187 * Enabled with +inspect-jpegs in actions files.
189 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
190 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
192 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
193 * Fixed bug #839859: "See why" link URL now gets url-encoded.
195 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
198 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
199 * Added basic support for OPTIONS and TRACE HTTP methods:
200 * New function direct_response which handles OPTIONS and
201 * TRACE requests whose Max-Forwards header field is zero.
203 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
204 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
206 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
207 * - Bugfix: Executing a filters without jobs no longer results in
208 * turing off *all* filters.
209 * - Security fix: Malicious web servers can't cause a seg fault
210 * through bogus chunk sizes anymore
212 * Revision 1.58 2002/04/24 02:11:17 oes
213 * Jon's multiple AF patch: url_actions now evaluates rules
216 * Revision 1.57 2002/04/08 20:38:34 swa
219 * Revision 1.56 2002/04/05 15:51:24 oes
220 * - bugfix: error-pages now get correct request protocol
221 * - fix for invalid HTML in trust info
223 * Revision 1.55 2002/04/02 16:13:51 oes
224 * Fix: No "Go there anyway" for SSL
226 * Revision 1.54 2002/04/02 14:55:56 oes
227 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
229 * Revision 1.53 2002/03/26 22:29:54 swa
230 * we have a new homepage!
232 * Revision 1.52 2002/03/24 16:35:57 jongfoster
235 * Revision 1.51 2002/03/24 15:23:33 jongfoster
238 * Revision 1.50 2002/03/24 13:25:43 swa
239 * name change related issues
241 * Revision 1.49 2002/03/16 20:29:14 oes
244 * Revision 1.48 2002/03/13 20:25:34 oes
245 * Better logging for content filters
247 * Revision 1.47 2002/03/13 00:30:52 jongfoster
249 * Added option of always sending redirect for imageblock,
250 * currently disabled with #if 0.
252 * Revision 1.46 2002/03/12 01:42:49 oes
253 * Introduced modular filters
255 * Revision 1.45 2002/03/08 16:47:50 oes
256 * Added choice beween GIF and PNG built-in images
258 * Revision 1.44 2002/03/07 03:49:31 oes
259 * - Fixed compiler warnings etc
260 * - Changed built-in images from GIF to PNG
261 * (with regard to Unisys patent issue)
262 * - Added a 4x4 pattern PNG which is less intrusive
263 * than the logo but also clearly marks the deleted banners
265 * Revision 1.43 2002/01/22 23:51:59 jongfoster
266 * Replacing strsav() with the safer string_append().
268 * Adding missing html_encode() to error message generators. Where encoded
269 * and unencoded versions of a string were provided, removing the unencoded
272 * Revision 1.42 2002/01/17 21:00:32 jongfoster
273 * Moving all our URL and URL pattern parsing code to urlmatch.c.
275 * Using a single, simple url_match(pattern,url) function - rather than
276 * the 3-line match routine which was repeated all over the place.
278 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
280 * Using parse_http_url() to parse URLs without faking a HTTP
281 * request line for parse_http_request().
283 * Revision 1.41 2001/11/13 00:14:07 jongfoster
284 * Fixing stupid bug now I've figured out what || means.
285 * (It always returns 0 or 1, not one of it's paramaters.)
287 * Revision 1.40 2001/10/26 17:37:55 oes
288 * - Re-enabled Netscape 200/404 bug workaround in block_url():
289 * - Removed OS/2 special case
290 * - Made block_url() independant from sed() having been run
291 * - Made trust_url independant from sed() having been run
292 * - Made is_imageurl independant from sed() having been run.
293 * It now checks User-Agent: and Accept: by itself.
296 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
297 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
298 * threads to call select() simultaneously. So, it's time to do a real, live,
299 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
300 * (native). Both versions will work, but using __OS2__ offers multi-threading.
302 * Revision 1.38 2001/10/23 21:32:33 jongfoster
303 * Adding error-checking to selected functions
305 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
306 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
307 * filters.c. Added a FIXME in front of the offending code. I'll gladly
308 * put in a better/more robust fix for all parties if one is presented...
309 * It seems that just returning 200 instead of 404 would pretty much fix
310 * it for everyone, but I don't know all the history of the problem.
312 * Revision 1.36 2001/10/10 16:44:16 oes
313 * Added match_portlist function
315 * Revision 1.35 2001/10/07 15:41:23 oes
316 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
318 * New function remove_chunked_transfer_coding that strips chunked
319 * transfer coding to plain and is called by pcrs_filter_response
320 * and gif_deanimate_response if neccessary
322 * Improved handling of zero-change re_filter runs
324 * pcrs_filter_response and gif_deanimate_response now remove
325 * chunked transfer codeing before processing the body.
327 * Revision 1.34 2001/09/20 15:49:36 steudten
329 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
330 * See cgi.c fill_template().
332 * Revision 1.33 2001/09/16 17:05:14 jongfoster
333 * Removing unused #include showarg.h
335 * Revision 1.32 2001/09/16 13:21:27 jongfoster
336 * Changes to use new list functions.
338 * Revision 1.31 2001/09/16 11:38:02 jongfoster
339 * Splitting fill_template() into 2 functions:
340 * template_load() loads the file
341 * template_fill() performs the PCRS regexps.
342 * This is because the CGI edit interface has a "table row"
343 * template which is used many times in the page - this
344 * change means it's only loaded from disk once.
346 * Revision 1.30 2001/09/16 11:00:10 jongfoster
347 * New function alloc_http_response, for symmetry with free_http_response
349 * Revision 1.29 2001/09/13 23:32:40 jongfoster
350 * Moving image data to cgi.c rather than cgi.h
351 * Fixing a GPF under Win32 (and any other OS that protects global
352 * constants from being written to).
354 * Revision 1.28 2001/09/10 10:18:51 oes
355 * Silenced compiler warnings
357 * Revision 1.27 2001/08/05 16:06:20 jongfoster
358 * Modifiying "struct map" so that there are now separate header and
359 * "map_entry" structures. This means that functions which modify a
360 * map no longer need to return a pointer to the modified map.
361 * Also, it no longer reverses the order of the entries (which may be
362 * important with some advanced template substitutions).
364 * Revision 1.26 2001/07/30 22:08:36 jongfoster
365 * Tidying up #defines:
366 * - All feature #defines are now of the form FEATURE_xxx
367 * - Permanently turned off WIN_GUI_EDIT
368 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
370 * Revision 1.25 2001/07/26 10:09:46 oes
371 * Made browser detection a little less naive
373 * Revision 1.24 2001/07/25 17:22:51 oes
374 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
376 * Revision 1.23 2001/07/23 13:40:12 oes
377 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
379 * Revision 1.22 2001/07/18 12:29:34 oes
380 * - Made gif_deanimate_response respect
381 * csp->action->string[ACTION_STRING_DEANIMATE]
382 * - Logging cosmetics
384 * Revision 1.21 2001/07/13 13:59:53 oes
385 * - Introduced gif_deanimate_response which shares the
386 * generic content modification interface of pcrs_filter_response
387 * and acts as a wrapper to deanimate.c:gif_deanimate()
388 * - Renamed re_process_buffer to pcrs_filter_response
389 * - pcrs_filter_response now returns NULL on failiure
390 * - Removed all #ifdef PCRS
392 * Revision 1.20 2001/07/01 17:01:04 oes
393 * Added comments and missing return statement in is_untrusted_url()
395 * Revision 1.19 2001/06/29 21:45:41 oes
396 * Indentation, CRLF->LF, Tab-> Space
398 * Revision 1.18 2001/06/29 13:27:38 oes
399 * - Cleaned up, renamed and reorderd functions
400 * and improved comments
403 * - Ported to CGI platform. Now delivers
404 * http_response or NULL
405 * - Unified HTML and GIF generation (moved image detection
406 * and GIF generation here from jcc.c:chat())
407 * - Fixed HTTP status to:
408 * - 403 (Forbidden) for the "blocked" HTML message
409 * - 200 (OK) for GIF answers
410 * - 302 (Redirect) for redirect to GIF
413 * - Ported to CGI platform. Now delivers
414 * http_response or NULL
415 * - Separated detection of untrusted URL into
416 * (bool)is_untrusted_url
417 * - Added enforcement of untrusted requests
419 * - Moved redirect_url() from cgi.c to here
420 * and ported it to the CGI platform
422 * - Removed logentry from cancelled commit
424 * Revision 1.17 2001/06/09 10:55:28 jongfoster
425 * Changing BUFSIZ ==> BUFFER_SIZE
427 * Revision 1.16 2001/06/07 23:10:26 jongfoster
428 * Allowing unanchored domain patterns to back off and retry
429 * if they partially match. Optimized right-anchored patterns.
430 * Moving ACL and forward files into config file.
431 * Replacing struct gateway with struct forward_spec
433 * Revision 1.15 2001/06/03 19:12:00 oes
434 * extracted-CGI relevant stuff
436 * Revision 1.14 2001/06/01 10:30:55 oes
437 * Added optional left-anchoring to domaincmp
439 * Revision 1.13 2001/05/31 21:21:30 jongfoster
440 * Permissionsfile / actions file changes:
441 * - Changed "permission" to "action" throughout
442 * - changes to file format to allow string parameters
443 * - Moved helper functions to actions.c
445 * Revision 1.12 2001/05/31 17:35:20 oes
447 * - Enhanced domain part globbing with infix and prefix asterisk
448 * matching and optional unanchored operation
450 * Revision 1.11 2001/05/29 11:53:23 oes
451 * "See why" link added to "blocked" page
453 * Revision 1.10 2001/05/29 09:50:24 jongfoster
454 * Unified blocklist/imagelist/permissionslist.
455 * File format is still under discussion, but the internal changes
458 * Also modified interceptor behaviour:
459 * - We now intercept all URLs beginning with one of the following
460 * prefixes (and *only* these prefixes):
462 * * http://ijbswa.sf.net/config/
463 * * http://ijbswa.sourceforge.net/config/
464 * - New interceptors "home page" - go to http://i.j.b/ to see it.
465 * - Internal changes so that intercepted and fast redirect pages
466 * are not replaced with an image.
467 * - Interceptors now have the option to send a binary page direct
468 * to the client. (i.e. ijb-send-banner uses this)
469 * - Implemented show-url-info interceptor. (Which is why I needed
470 * the above interceptors changes - a typical URL is
471 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
472 * The previous mechanism would not have intercepted that, and
473 * if it had been intercepted then it then it would have replaced
476 * Revision 1.9 2001/05/27 22:17:04 oes
478 * - re_process_buffer no longer writes the modified buffer
479 * to the client, which was very ugly. It now returns the
480 * buffer, which it is then written by chat.
482 * - content_length now adjusts the Content-Length: header
483 * for modified documents rather than crunch()ing it.
484 * (Length info in csp->content_length, which is 0 for
485 * unmodified documents)
487 * - For this to work, sed() is called twice when filtering.
489 * Revision 1.8 2001/05/26 17:13:28 jongfoster
490 * Filled in a function comment.
492 * Revision 1.7 2001/05/26 15:26:15 jongfoster
493 * ACL feature now provides more security by immediately dropping
494 * connections from untrusted hosts.
496 * Revision 1.6 2001/05/26 00:28:36 jongfoster
497 * Automatic reloading of config file.
498 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
499 * Most of the global variables have been moved to a new
500 * struct configuration_spec, accessed through csp->config->globalname
501 * Most of the globals remaining are used by the Win32 GUI.
503 * Revision 1.5 2001/05/25 22:34:30 jongfoster
506 * Revision 1.4 2001/05/22 18:46:04 oes
508 * - Enabled filtering banners by size rather than URL
509 * by adding patterns that replace all standard banner
510 * sizes with the "Junkbuster" gif to the re_filterfile
512 * - Enabled filtering WebBugs by providing a pattern
513 * which kills all 1x1 images
515 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
516 * which is selected by the (nonstandard and therefore
517 * capital) letter 'U' in the option string.
518 * It causes the quantifiers to be ungreedy by default.
519 * Appending a ? turns back to greedy (!).
521 * - Added a new interceptor ijb-send-banner, which
522 * sends back the "Junkbuster" gif. Without imagelist or
523 * MSIE detection support, or if tinygif = 1, or the
524 * URL isn't recognized as an imageurl, a lame HTML
525 * explanation is sent instead.
527 * - Added new feature, which permits blocking remote
528 * script redirects and firing back a local redirect
530 * The feature is conditionally compiled, i.e. it
531 * can be disabled with --disable-fast-redirects,
532 * plus it must be activated by a "fast-redirects"
533 * line in the config file, has its own log level
534 * and of course wants to be displayed by show-proxy-args
535 * Note: Boy, all the #ifdefs in 1001 locations and
536 * all the fumbling with configure.in and acconfig.h
537 * were *way* more work than the feature itself :-(
539 * - Because a generic redirect template was needed for
540 * this, tinygif = 3 now uses the same.
542 * - Moved GIFs, and other static HTTP response templates
547 * - Removed some >400 CRs again (Jon, you really worked
550 * Revision 1.3 2001/05/20 16:44:47 jongfoster
551 * Removing last hardcoded Junkbusters.com URLs.
553 * Revision 1.2 2001/05/20 01:21:20 jongfoster
554 * Version 2.9.4 checkin.
555 * - Merged popupfile and cookiefile, and added control over PCRS
556 * filtering, in new "permissionsfile".
557 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
558 * file error you now get a message box (in the Win32 GUI) rather
559 * than the program exiting with no explanation.
560 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
562 * - Removed tabs from "config"
563 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
564 * - Bumped up version number.
566 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
567 * Initial import of version 2.9.3 source tree
570 *********************************************************************/
576 #include <sys/types.h>
585 #endif /* ndef __OS2__ */
586 #include <netinet/in.h>
588 #include <winsock2.h>
589 #endif /* ndef _WIN32 */
593 #endif /* def __OS2__ */
601 #include "jbsockets.h"
602 #include "miscutil.h"
606 #include "deanimate.h"
607 #include "urlmatch.h"
614 const char filters_h_rcs[] = FILTERS_H_VERSION;
616 /* Fix a problem with Solaris. There should be no effect on other
618 * Solaris's isspace() is a macro which uses it's argument directly
619 * as an array index. Therefore we need to make sure that high-bit
620 * characters generate +ve values, and ideally we also want to make
621 * the argument match the declared parameter type of "int".
623 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
625 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
626 static jb_err prepare_for_filtering(struct client_state *csp);
629 /*********************************************************************
631 * Function : block_acl
633 * Description : Block this request?
634 * Decide yes or no based on ACL file.
637 * 1 : dst = The proxy or gateway address this is going to.
638 * Or NULL to check all possible targets.
639 * 2 : csp = Current client state (buffers, headers, etc...)
640 * Also includes the client IP address.
642 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
644 *********************************************************************/
645 int block_acl(struct access_control_addr *dst, struct client_state *csp)
647 struct access_control_list *acl = csp->config->acl;
649 /* if not using an access control list, then permit the connection */
655 /* search the list */
658 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
662 /* Just want to check if they have any access */
663 if (acl->action == ACL_PERMIT)
668 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
669 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
671 if (acl->action == ACL_PERMIT)
689 /*********************************************************************
691 * Function : acl_addr
693 * Description : Called from `load_config' to parse an ACL address.
696 * 1 : aspec = String specifying ACL address.
697 * 2 : aca = struct access_control_addr to fill in.
699 * Returns : 0 => Ok, everything else is an error.
701 *********************************************************************/
702 int acl_addr(const char *aspec, struct access_control_addr *aca)
707 char *acl_spec = NULL;
713 * Use a temporary acl spec copy so we can log
714 * the unmodified original in case of parse errors.
716 acl_spec = strdup(aspec);
717 if (acl_spec == NULL)
719 /* XXX: This will be logged as parse error. */
723 if ((p = strchr(acl_spec, '/')) != NULL)
726 if (ijb_isdigit(*p) == 0)
731 masklength = atoi(p);
734 if ((masklength < 0) || (masklength > 32))
740 if ((p = strchr(acl_spec, ':')) != NULL)
745 port = strtol(p, &endptr, 10);
747 if (port <= 0 || port > 65535 || *endptr != '\0')
754 aca->port = (unsigned long)port;
756 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
759 if (aca->addr == INADDR_NONE)
761 /* XXX: This will be logged as parse error. */
765 /* build the netmask */
767 for (i=1; i <= masklength ; i++)
769 aca->mask |= (1 << (32 - i));
772 /* now mask off the host portion of the ip address
773 * (i.e. save on the network portion of the address).
775 aca->addr = aca->addr & aca->mask;
780 #endif /* def FEATURE_ACL */
783 /*********************************************************************
785 * Function : block_url
787 * Description : Called from `chat'. Check to see if we need to block this.
790 * 1 : csp = Current client state (buffers, headers, etc...)
792 * Returns : NULL => unblocked, else HTTP block response
794 *********************************************************************/
795 struct http_response *block_url(struct client_state *csp)
797 struct http_response *rsp;
798 const char *new_content_type = NULL;
801 * If it's not blocked, don't block it ;-)
803 if ((csp->action->flags & ACTION_BLOCK) == 0)
807 if (csp->action->flags & ACTION_REDIRECT)
809 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
812 * Else, prepare a response
814 if (NULL == (rsp = alloc_http_response()))
816 return cgi_error_memory();
820 * If it's an image-url, send back an image or redirect
821 * as specified by the relevant +image action
823 #ifdef FEATURE_IMAGE_BLOCKING
824 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
828 /* determine HOW images should be blocked */
829 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
831 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
833 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
835 #if 1 /* Two alternative strategies, use this one for now: */
837 /* and handle accordingly: */
838 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
840 rsp->status = strdup("403 Request blocked by Privoxy");
841 if (rsp->status == NULL)
843 free_http_response(rsp);
844 return cgi_error_memory();
846 rsp->body = bindup(image_pattern_data, image_pattern_length);
847 if (rsp->body == NULL)
849 free_http_response(rsp);
850 return cgi_error_memory();
852 rsp->content_length = image_pattern_length;
854 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
856 free_http_response(rsp);
857 return cgi_error_memory();
861 else if (0 == strcmpic(p, "blank"))
863 rsp->status = strdup("403 Request blocked by Privoxy");
864 if (rsp->status == NULL)
866 free_http_response(rsp);
867 return cgi_error_memory();
869 rsp->body = bindup(image_blank_data, image_blank_length);
870 if (rsp->body == NULL)
872 free_http_response(rsp);
873 return cgi_error_memory();
875 rsp->content_length = image_blank_length;
877 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
879 free_http_response(rsp);
880 return cgi_error_memory();
886 rsp->status = strdup("302 Local Redirect from Privoxy");
887 if (rsp->status == NULL)
889 free_http_response(rsp);
890 return cgi_error_memory();
893 if (enlist_unique_header(rsp->headers, "Location", p))
895 free_http_response(rsp);
896 return cgi_error_memory();
900 #else /* Following code is disabled for now */
902 /* and handle accordingly: */
903 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
905 p = CGI_PREFIX "send-banner?type=pattern";
907 else if (0 == strcmpic(p, "blank"))
909 p = CGI_PREFIX "send-banner?type=blank";
911 rsp->status = strdup("302 Local Redirect from Privoxy");
912 if (rsp->status == NULL)
914 free_http_response(rsp);
915 return cgi_error_memory();
918 if (enlist_unique_header(rsp->headers, "Location", p))
920 free_http_response(rsp);
921 return cgi_error_memory();
923 #endif /* Preceeding code is disabled for now */
925 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
928 * Send empty document.
930 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
933 rsp->body = strdup(" ");
934 rsp->content_length = 1;
936 rsp->status = strdup("403 Request blocked by Privoxy");
937 if (rsp->status == NULL)
939 free_http_response(rsp);
940 return cgi_error_memory();
942 if (new_content_type != 0)
944 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
945 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
947 free_http_response(rsp);
948 return cgi_error_memory();
953 #endif /* def FEATURE_IMAGE_BLOCKING */
956 * Else, generate an HTML "blocked" message:
960 struct map * exports;
964 * Workaround for stupid Netscape bug which prevents
965 * pages from being displayed if loading a referenced
966 * JavaScript or style sheet fails. So make it appear
967 * as if it succeeded.
969 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
970 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
971 && !strstr(p, "Gecko") /* save Mozilla, */
972 && !strstr(p, "compatible") /* MSIE */
973 && !strstr(p, "Opera")) /* and Opera. */
975 rsp->status = strdup("200 Request for blocked URL");
979 rsp->status = strdup("403 Request for blocked URL");
982 if (rsp->status == NULL)
984 free_http_response(rsp);
985 return cgi_error_memory();
988 exports = default_exports(csp, NULL);
991 free_http_response(rsp);
992 return cgi_error_memory();
995 #ifdef FEATURE_FORCE_LOAD
996 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
998 * Export the force conditional block killer if
1000 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1001 * - Privoxy is configured to enforce blocks, or
1002 * - it's a CONNECT request and enforcing wouldn't work anyway.
1004 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1005 || (0 == strcmpic(csp->http->gpc, "connect")))
1006 #endif /* ndef FEATURE_FORCE_LOAD */
1008 err = map_block_killer(exports, "force-support");
1011 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1012 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1013 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1014 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1019 free_http_response(rsp);
1020 return cgi_error_memory();
1023 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1026 free_http_response(rsp);
1027 return cgi_error_memory();
1030 rsp->reason = RSP_REASON_BLOCKED;
1032 return finish_http_response(csp, rsp);
1037 #ifdef FEATURE_TRUST
1038 /*********************************************************************
1040 * Function : trust_url FIXME: I should be called distrust_url
1042 * Description : Calls is_untrusted_url to determine if the URL is trusted
1043 * and if not, returns a HTTP 403 response with a reject message.
1046 * 1 : csp = Current client state (buffers, headers, etc...)
1048 * Returns : NULL => trusted, else http_response.
1050 *********************************************************************/
1051 struct http_response *trust_url(struct client_state *csp)
1053 struct http_response *rsp;
1054 struct map * exports;
1055 char buf[BUFFER_SIZE];
1057 struct url_spec **tl;
1062 * Don't bother to work on trusted URLs
1064 if (!is_untrusted_url(csp))
1070 * Else, prepare a response:
1072 if (NULL == (rsp = alloc_http_response()))
1074 return cgi_error_memory();
1077 rsp->status = strdup("403 Request blocked by Privoxy");
1078 exports = default_exports(csp, NULL);
1079 if (exports == NULL || rsp->status == NULL)
1081 free_http_response(rsp);
1082 return cgi_error_memory();
1086 * Export the protocol, host, port, and referrer information
1088 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1089 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1090 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1092 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1094 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1098 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1104 free_http_response(rsp);
1105 return cgi_error_memory();
1109 * Export the trust list
1112 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1114 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1115 string_append(&p, buf);
1117 err = map(exports, "trusted-referrers", 1, p, 0);
1122 free_http_response(rsp);
1123 return cgi_error_memory();
1127 * Export the trust info, if available
1129 if (csp->config->trust_info->first)
1131 struct list_entry *l;
1134 for (l = csp->config->trust_info->first; l ; l = l->next)
1136 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1137 string_append(&p, buf);
1139 err = map(exports, "trust-info", 1, p, 0);
1143 err = map_block_killer(exports, "have-trust-info");
1149 free_http_response(rsp);
1150 return cgi_error_memory();
1154 * Export the force conditional block killer if
1156 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1157 * - Privoxy is configured to enforce blocks, or
1158 * - it's a CONNECT request and enforcing wouldn't work anyway.
1160 #ifdef FEATURE_FORCE_LOAD
1161 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1162 || (0 == strcmpic(csp->http->gpc, "connect")))
1164 err = map_block_killer(exports, "force-support");
1168 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1170 #else /* ifndef FEATURE_FORCE_LOAD */
1171 err = map_block_killer(exports, "force-support");
1172 #endif /* ndef FEATURE_FORCE_LOAD */
1177 free_http_response(rsp);
1178 return cgi_error_memory();
1182 * Build the response
1184 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1187 free_http_response(rsp);
1188 return cgi_error_memory();
1190 rsp->reason = RSP_REASON_UNTRUSTED;
1192 return finish_http_response(csp, rsp);
1194 #endif /* def FEATURE_TRUST */
1197 /*********************************************************************
1199 * Function : compile_dynamic_pcrs_job_list
1201 * Description : Compiles a dynamic pcrs job list (one with variables
1202 * resolved at request time)
1205 * 1 : csp = Current client state (buffers, headers, etc...)
1206 * 2 : b = The filter list to compile
1208 * Returns : NULL in case of errors, otherwise the
1211 *********************************************************************/
1212 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1214 struct list_entry *pattern;
1215 pcrs_job *job_list = NULL;
1216 pcrs_job *dummy = NULL;
1217 pcrs_job *lastjob = NULL;
1220 const struct pcrs_variable variables[] =
1222 {"url", csp->http->url, 1},
1223 {"path", csp->http->path, 1},
1224 {"host", csp->http->host, 1},
1225 {"origin", csp->ip_addr_str, 1},
1229 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1231 assert(pattern->str != NULL);
1233 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1237 log_error(LOG_LEVEL_ERROR,
1238 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1239 pattern->str, b->name, pcrs_strerror(error));
1244 if (error == PCRS_WARN_TRUNCATION)
1246 log_error(LOG_LEVEL_ERROR,
1247 "At least one of the variables in \'%s\' had to "
1248 "be truncated before compilation", pattern->str);
1250 if (job_list == NULL)
1256 lastjob->next = dummy;
1266 /*********************************************************************
1268 * Function : rewrite_url
1270 * Description : Rewrites a URL with a single pcrs command
1271 * and returns the result if it differs from the
1272 * original and isn't obviously invalid.
1275 * 1 : old_url = URL to rewrite.
1276 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1279 * Returns : NULL if the pcrs_command didn't change the url, or
1280 * the result of the modification.
1282 *********************************************************************/
1283 char *rewrite_url(char *old_url, const char *pcrs_command)
1285 char *new_url = NULL;
1289 assert(pcrs_command);
1291 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1295 log_error(LOG_LEVEL_REDIRECTS,
1296 "pcrs command \"%s\" didn't change \"%s\".",
1297 pcrs_command, old_url);
1302 log_error(LOG_LEVEL_REDIRECTS,
1303 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1304 pcrs_command, old_url, pcrs_strerror(hits));
1307 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1309 log_error(LOG_LEVEL_ERROR,
1310 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1311 "but the result doesn't look like a valid URL and will be ignored.",
1312 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1317 log_error(LOG_LEVEL_REDIRECTS,
1318 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1319 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1327 #ifdef FEATURE_FAST_REDIRECTS
1328 /*********************************************************************
1330 * Function : get_last_url
1332 * Description : Search for the last URL inside a string.
1333 * If the string already is a URL, it will
1334 * be the first URL found.
1337 * 1 : subject = the string to check
1338 * 2 : redirect_mode = +fast-redirect{} mode
1340 * Returns : NULL if no URL was found, or
1341 * the last URL found.
1343 *********************************************************************/
1344 char *get_last_url(char *subject, const char *redirect_mode)
1346 char *new_url = NULL;
1350 assert(redirect_mode);
1352 subject = strdup(subject);
1353 if (subject == NULL)
1355 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1359 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1361 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1362 new_url = url_decode(subject);
1363 if (new_url != NULL)
1370 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1374 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1377 * Find the last URL encoded in the request
1380 while ((tmp = strstr(tmp, "http://")) != NULL)
1384 tmp = (new_url != NULL) ? new_url : subject;
1385 while ((tmp = strstr(tmp, "https://")) != NULL)
1390 if ((new_url != NULL)
1391 && ( (new_url != subject)
1392 || (0 == strncmpic(subject, "http://", 7))
1393 || (0 == strncmpic(subject, "https://", 8))
1397 * Return new URL if we found a redirect
1398 * or if the subject already was a URL.
1400 * The second case makes sure that we can
1401 * chain get_last_url after another redirection check
1402 * (like rewrite_url) without losing earlier redirects.
1404 new_url = strdup(new_url);
1413 #endif /* def FEATURE_FAST_REDIRECTS */
1416 /*********************************************************************
1418 * Function : redirect_url
1420 * Description : Checks if Privoxy should answer the request with
1421 * a HTTP redirect and generates the redirect if
1425 * 1 : csp = Current client state (buffers, headers, etc...)
1427 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1429 *********************************************************************/
1430 struct http_response *redirect_url(struct client_state *csp)
1432 struct http_response *rsp;
1433 #ifdef FEATURE_FAST_REDIRECTS
1435 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1436 * as compile-time option? The user can easily disable
1437 * it in his action file.
1439 char * redirect_mode;
1440 #endif /* def FEATURE_FAST_REDIRECTS */
1441 char *old_url = NULL;
1442 char *new_url = NULL;
1443 char *redirection_string;
1445 if ((csp->action->flags & ACTION_REDIRECT))
1447 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1450 * If the redirection string begins with 's',
1451 * assume it's a pcrs command, otherwise treat it as
1452 * properly formatted URL and use it for the redirection
1455 * According to RFC 2616 section 14.30 the URL
1456 * has to be absolute and if the user tries:
1457 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1458 * she would get undefined results anyway.
1462 if (*redirection_string == 's')
1464 old_url = csp->http->url;
1465 new_url = rewrite_url(old_url, redirection_string);
1469 log_error(LOG_LEVEL_REDIRECTS,
1470 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1471 redirection_string);
1472 new_url = strdup(redirection_string);
1476 #ifdef FEATURE_FAST_REDIRECTS
1477 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1479 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1482 * If it exists, use the previously rewritten URL as input
1483 * otherwise just use the old path.
1485 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1486 new_url = get_last_url(old_url, redirect_mode);
1491 * Disable redirect checkers, so that they
1492 * will be only run more than once if the user
1493 * also enables them through tags.
1495 * From a performance point of view
1496 * it doesn't matter, but the duplicated
1497 * log messages are annoying.
1499 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1500 #endif /* def FEATURE_FAST_REDIRECTS */
1501 csp->action->flags &= ~ACTION_REDIRECT;
1503 /* Did any redirect action trigger? */
1506 if (0 == strcmpic(new_url, csp->http->url))
1508 log_error(LOG_LEVEL_ERROR,
1509 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1510 csp->http->url, new_url);
1515 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1517 if (NULL == (rsp = alloc_http_response()))
1520 return cgi_error_memory();
1523 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1524 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1527 free_http_response(rsp);
1528 return cgi_error_memory();
1530 rsp->reason = RSP_REASON_REDIRECTED;
1533 return finish_http_response(csp, rsp);
1537 /* Only reached if no redirect is required */
1543 #ifdef FEATURE_IMAGE_BLOCKING
1544 /*********************************************************************
1546 * Function : is_imageurl
1548 * Description : Given a URL, decide whether it is an image or not,
1549 * using either the info from a previous +image action
1550 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1551 * is MSIE and not on a Mac, tell from the browser's accept
1555 * 1 : csp = Current client state (buffers, headers, etc...)
1557 * Returns : True (nonzero) if URL is an image, false (0)
1560 *********************************************************************/
1561 int is_imageurl(const struct client_state *csp)
1563 #ifdef FEATURE_IMAGE_DETECT_MSIE
1566 tmp = get_header_value(csp->headers, "User-Agent:");
1567 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1569 tmp = get_header_value(csp->headers, "Accept:");
1570 if (tmp && strstr(tmp, "image/gif"))
1572 /* Client will accept HTML. If this seems counterintuitive,
1582 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1584 return ((csp->action->flags & ACTION_IMAGE) != 0);
1587 #endif /* def FEATURE_IMAGE_BLOCKING */
1590 #ifdef FEATURE_TRUST
1591 /*********************************************************************
1593 * Function : is_untrusted_url
1595 * Description : Should we "distrust" this URL (and block it)?
1597 * Yes if it matches a line in the trustfile, or if the
1598 * referrer matches a line starting with "+" in the
1603 * 1 : csp = Current client state (buffers, headers, etc...)
1605 * Returns : 0 => trusted, 1 => untrusted
1607 *********************************************************************/
1608 int is_untrusted_url(const struct client_state *csp)
1610 struct file_list *fl;
1611 struct block_spec *b;
1612 struct url_spec **trusted_url;
1613 struct http_request rhttp[1];
1614 const char * referer;
1618 * If we don't have a trustlist, we trust everybody
1620 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1625 memset(rhttp, '\0', sizeof(*rhttp));
1628 * Do we trust the request URL itself?
1630 for (b = b->next; b ; b = b->next)
1632 if (url_match(b->url, csp->http))
1638 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1640 /* no referrer was supplied */
1646 * If not, do we maybe trust its referrer?
1648 err = parse_http_url(referer, rhttp, csp);
1654 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1656 if (url_match(*trusted_url, rhttp))
1658 /* if the URL's referrer is from a trusted referrer, then
1659 * add the target spec to the trustfile as an unblocked
1660 * domain and return 0 (which means it's OK).
1665 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1669 char * new_entry = strdup("~");
1671 string_append(&new_entry, csp->http->hostport);
1673 path = csp->http->path;
1674 if ( (path[0] == '/')
1676 && ((path_end = strchr(path + 2, '/')) != NULL))
1678 /* since this path points into a user's home space
1679 * be sure to include this spec in the trustfile.
1681 int path_len = path_end - path; /* save offset */
1682 path = strdup(path); /* Copy string */
1685 path_end = path + path_len; /* regenerate ptr to new buffer */
1686 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1688 string_join(&new_entry, path);
1692 * Give a reason for generating this entry.
1694 string_append(&new_entry, " # Trusted referrer was: ");
1695 string_append(&new_entry, referer);
1697 if (new_entry != NULL)
1699 if (-1 == fprintf(fp, "%s\n", new_entry))
1701 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1702 new_entry, csp->config->trustfile);
1708 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1709 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1716 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1717 csp->http->hostport, csp->config->trustfile);
1725 #endif /* def FEATURE_TRUST */
1728 /*********************************************************************
1730 * Function : pcrs_filter_response
1732 * Description : Execute all text substitutions from all applying
1733 * +filter actions on the text buffer that's been
1734 * accumulated in csp->iob->buf.
1737 * 1 : csp = Current client state (buffers, headers, etc...)
1739 * Returns : a pointer to the (newly allocated) modified buffer.
1740 * or NULL if there were no hits or something went wrong
1742 *********************************************************************/
1743 char *pcrs_filter_response(struct client_state *csp)
1746 size_t size, prev_size;
1752 struct file_list *fl;
1753 struct re_filterfile_spec *b;
1754 struct list_entry *filtername;
1756 int i, found_filters = 0;
1761 if (csp->iob->cur >= csp->iob->eod)
1767 * Need to check the set of re_filterfiles...
1769 for (i = 0; i < MAX_AF_FILES; i++)
1782 if (0 == found_filters)
1784 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1788 size = (size_t)(csp->iob->eod - csp->iob->cur);
1789 old = csp->iob->cur;
1791 for (i = 0; i < MAX_AF_FILES; i++)
1794 if ((NULL == fl) || (NULL == fl->f))
1797 * Either there are no filter files
1798 * left, or this filter file just
1799 * contains no valid filters.
1801 * Continue to be sure we don't miss
1802 * valid filter files that are chained
1803 * after empty or invalid ones.
1808 * For all applying +filter actions, look if a filter by that
1809 * name exists and if yes, execute it's pcrs_joblist on the
1812 for (b = fl->f; b; b = b->next)
1814 if (b->type != FT_CONTENT_FILTER)
1816 /* Skip header filters */
1820 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1821 filtername ; filtername = filtername->next)
1823 if (strcmp(b->name, filtername->str) == 0)
1825 int current_hits = 0; /* Number of hits caused by this filter */
1826 int job_number = 0; /* Which job we're currently executing */
1827 int job_hits = 0; /* How many hits the current job caused */
1828 pcrs_job *joblist = b->joblist;
1830 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1832 if (NULL == joblist)
1834 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1839 /* Apply all jobs from the joblist */
1840 for (job = joblist; NULL != job; job = job->next)
1843 job_hits = pcrs_execute(job, old, size, &new, &size);
1848 * That went well. Continue filtering
1849 * and use the result of this job as
1850 * input for the next one.
1852 current_hits += job_hits;
1853 if (old != csp->iob->cur)
1862 * This job caused an unexpected error. Inform the user
1863 * and skip the rest of the jobs in this filter. We could
1864 * continue with the next job, but usually the jobs
1865 * depend on each other or are similar enough to
1866 * fail for the same reason.
1868 * At the moment our pcrs expects the error codes of pcre 3.4,
1869 * but newer pcre versions can return additional error codes.
1870 * As a result pcrs_strerror()'s error message might be
1871 * "Unknown error ...", therefore we print the numerical value
1874 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1875 * should we use LOG_LEVEL_RE_FILTER instead?
1877 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1878 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1883 if (b->dynamic) pcrs_free_joblist(joblist);
1885 log_error(LOG_LEVEL_RE_FILTER,
1886 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1887 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1889 hits += current_hits;
1896 * If there were no hits, destroy our copy and let
1897 * chat() use the original in csp->iob
1905 csp->flags |= CSP_FLAG_MODIFIED;
1906 csp->content_length = size;
1914 /*********************************************************************
1916 * Function : gif_deanimate_response
1918 * Description : Deanimate the GIF image that has been accumulated in
1919 * csp->iob->buf, set csp->content_length to the modified
1920 * size and raise the CSP_FLAG_MODIFIED flag.
1923 * 1 : csp = Current client state (buffers, headers, etc...)
1925 * Returns : a pointer to the (newly allocated) modified buffer.
1926 * or NULL in case something went wrong.
1928 *********************************************************************/
1929 char *gif_deanimate_response(struct client_state *csp)
1931 struct binbuffer *in, *out;
1935 size = (size_t)(csp->iob->eod - csp->iob->cur);
1937 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1938 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1940 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1944 in->buffer = csp->iob->cur;
1947 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1949 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1956 if ((int)size == out->offset)
1958 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1962 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1964 csp->content_length = out->offset;
1965 csp->flags |= CSP_FLAG_MODIFIED;
1975 /*********************************************************************
1977 * Function : jpeg_inspect_response
1982 * 1 : csp = Current client state (buffers, headers, etc...)
1984 * Returns : a pointer to the (newly allocated) modified buffer
1985 * or NULL in case something went wrong.
1987 *********************************************************************/
1988 char *jpeg_inspect_response(struct client_state *csp)
1990 struct binbuffer *in = NULL;
1991 struct binbuffer *out = NULL;
1995 size = (size_t)(csp->iob->eod - csp->iob->cur);
1997 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1999 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2003 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2005 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2009 in->buffer = csp->iob->cur;
2013 * Calling jpeg_inspect has the side-effect of creating and
2014 * modifying the image buffer of "out" directly.
2016 if (jpeg_inspect(in, out))
2018 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2026 csp->content_length = out->offset;
2027 csp->flags |= CSP_FLAG_MODIFIED;
2037 /*********************************************************************
2039 * Function : remove_chunked_transfer_coding
2041 * Description : In-situ remove the "chunked" transfer coding as defined
2042 * in rfc2616 from a buffer.
2045 * 1 : buffer = Pointer to the text buffer
2046 * 2 : size = In: Number of bytes to be processed,
2047 * Out: Number of bytes after de-chunking.
2048 * (undefined in case of errors)
2050 * Returns : JB_ERR_OK for success,
2051 * JB_ERR_PARSE otherwise
2053 *********************************************************************/
2054 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2057 unsigned int chunksize = 0;
2058 char *from_p, *to_p;
2061 from_p = to_p = buffer;
2063 if (sscanf(buffer, "%x", &chunksize) != 1)
2065 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2066 return JB_ERR_PARSE;
2069 while (chunksize > 0)
2071 if (NULL == (from_p = strstr(from_p, "\r\n")))
2073 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2074 return JB_ERR_PARSE;
2077 if ((newsize += chunksize) >= *size)
2079 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2080 return JB_ERR_PARSE;
2084 memmove(to_p, from_p, (size_t) chunksize);
2085 to_p = buffer + newsize;
2086 from_p += chunksize + 2;
2088 if (sscanf(from_p, "%x", &chunksize) != 1)
2090 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2091 return JB_ERR_PARSE;
2097 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2100 /* XXX: Should get its own loglevel. */
2101 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2110 /*********************************************************************
2112 * Function : prepare_for_filtering
2114 * Description : If necessary, de-chunks and decompresses
2115 * the content so it can get filterd.
2118 * 1 : csp = Current client state (buffers, headers, etc...)
2120 * Returns : JB_ERR_OK for success,
2121 * JB_ERR_PARSE otherwise
2123 *********************************************************************/
2124 static jb_err prepare_for_filtering(struct client_state *csp)
2126 jb_err err = JB_ERR_OK;
2129 * If the body has a "chunked" transfer-encoding,
2130 * get rid of it, adjusting size and iob->eod
2132 if (csp->flags & CSP_FLAG_CHUNKED)
2134 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2136 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2137 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2138 if (JB_ERR_OK == err)
2140 csp->iob->eod = csp->iob->cur + size;
2141 csp->flags |= CSP_FLAG_MODIFIED;
2145 log_error(LOG_LEVEL_ERROR, "Failed to de-chunk content.");
2146 return JB_ERR_PARSE;
2152 * If the body has a supported transfer-encoding,
2153 * decompress it, adjusting size and iob->eod.
2155 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2157 err = decompress_iob(csp);
2159 if (JB_ERR_OK == err)
2161 csp->flags |= CSP_FLAG_MODIFIED;
2162 csp->content_type &= ~CT_TABOO;
2167 * Unset CT_GZIP and CT_DEFLATE to remember not
2168 * to modify the Content-Encoding header later.
2170 csp->content_type &= ~CT_GZIP;
2171 csp->content_type &= ~CT_DEFLATE;
2172 log_error(LOG_LEVEL_ERROR, "Failed to decompress content.");
2181 /*********************************************************************
2183 * Function : execute_content_filter
2185 * Description : Executes a given content filter.
2188 * 1 : csp = Current client state (buffers, headers, etc...)
2189 * 2 : content_filter = The filter function to execute
2191 * Returns : JB_ERR_OK for success,
2192 * JB_ERR_PARSE otherwise
2194 *********************************************************************/
2195 char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
2197 if (JB_ERR_OK != prepare_for_filtering(csp))
2200 * failed to de-chunk or decompress.
2201 * XXX: if possible, we should continue anyway.
2206 if (0 == csp->iob->eod - csp->iob->cur)
2208 /* Empty buffer, nothing to do. */
2212 return ((*content_filter)(csp));
2216 /*********************************************************************
2218 * Function : url_actions
2220 * Description : Gets the actions for this URL.
2223 * 1 : http = http_request request for blocked URLs
2224 * 2 : csp = Current client state (buffers, headers, etc...)
2228 *********************************************************************/
2229 void url_actions(struct http_request *http,
2230 struct client_state *csp)
2232 struct file_list *fl;
2233 struct url_actions *b;
2236 init_current_action(csp->action);
2238 for (i = 0; i < MAX_AF_FILES; i++)
2240 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2245 apply_url_actions(csp->action, http, b);
2252 /*********************************************************************
2254 * Function : apply_url_actions
2256 * Description : Applies a list of URL actions.
2259 * 1 : action = Destination.
2260 * 2 : http = Current URL
2261 * 3 : b = list of URL actions to apply
2265 *********************************************************************/
2266 void apply_url_actions(struct current_action_spec *action,
2267 struct http_request *http,
2268 struct url_actions *b)
2272 /* Should never happen */
2276 for (b = b->next; NULL != b; b = b->next)
2278 if (url_match(b->url, http))
2280 merge_current_action(action, b->action);
2286 /*********************************************************************
2288 * Function : get_forward_override_settings
2290 * Description : Returns forward settings as specified with the
2291 * forward-override{} action. forward-override accepts
2292 * forward lines similar to the one used in the
2293 * configuration file, but without the URL pattern.
2299 * in the configuration file can be replaced with
2300 * the action section:
2302 * {+forward-override{forward .}}
2306 * 1 : csp = Current client state (buffers, headers, etc...)
2308 * Returns : Pointer to forwarding structure in case of success.
2309 * Invalid syntax is fatal.
2311 *********************************************************************/
2312 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2314 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2315 char forward_settings[BUFFER_SIZE];
2316 char *http_parent = NULL;
2317 /* variable names were chosen for consistency reasons. */
2318 struct forward_spec *fwd = NULL;
2322 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2323 /* Should be enforced by load_one_actions_file() */
2324 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2326 /* Create a copy ssplit can modify */
2327 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2329 if (NULL != csp->fwd)
2332 * XXX: Currently necessary to prevent memory
2333 * leaks when the show-url-info cgi page is visited.
2335 unload_forward_spec(csp->fwd);
2339 * allocate a new forward node, valid only for
2340 * the lifetime of this request. Save its location
2341 * in csp as well, so sweep() can free it later on.
2343 fwd = csp->fwd = zalloc(sizeof(*fwd));
2346 log_error(LOG_LEVEL_FATAL,
2347 "can't allocate memory for forward-override{%s}", forward_override_line);
2348 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2351 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2352 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2354 fwd->type = SOCKS_NONE;
2356 /* Parse the parent HTTP proxy host:port */
2357 http_parent = vec[1];
2360 else if (vec_count == 3)
2362 char *socks_proxy = NULL;
2364 if (!strcasecmp(vec[0], "forward-socks4"))
2366 fwd->type = SOCKS_4;
2367 socks_proxy = vec[1];
2369 else if (!strcasecmp(vec[0], "forward-socks4a"))
2371 fwd->type = SOCKS_4A;
2372 socks_proxy = vec[1];
2375 if (NULL != socks_proxy)
2377 /* Parse the SOCKS proxy host[:port] */
2378 fwd->gateway_host = strdup(socks_proxy);
2380 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2382 *socks_proxy++ = '\0';
2383 fwd->gateway_port = strtol(socks_proxy, NULL, 0);
2386 if (fwd->gateway_port <= 0)
2388 fwd->gateway_port = 1080;
2391 http_parent = vec[2];
2395 if (NULL == http_parent)
2397 log_error(LOG_LEVEL_FATAL,
2398 "Invalid forward-override syntax in: %s", forward_override_line);
2399 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2402 /* Parse http forwarding settings */
2403 if (strcmp(http_parent, ".") != 0)
2405 fwd->forward_host = strdup(http_parent);
2407 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2409 *http_parent++ = '\0';
2410 fwd->forward_port = strtol(http_parent, NULL, 0);
2413 if (fwd->forward_port <= 0)
2415 fwd->forward_port = 8000;
2419 assert (NULL != fwd);
2421 log_error(LOG_LEVEL_CONNECT,
2422 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2428 /*********************************************************************
2430 * Function : forward_url
2432 * Description : Should we forward this to another proxy?
2434 * XXX: Should be changed to make use of csp->fwd.
2437 * 1 : http = http_request request for current URL
2438 * 2 : csp = Current client state (buffers, headers, etc...)
2440 * Returns : Pointer to forwarding information.
2442 *********************************************************************/
2443 const struct forward_spec * forward_url(struct http_request *http,
2444 struct client_state *csp)
2446 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2447 struct forward_spec *fwd = csp->config->forward;
2449 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2451 return get_forward_override_settings(csp);
2461 if (url_match(fwd->url, http))
2472 /*********************************************************************
2474 * Function : direct_response
2476 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2477 * request and if so, return a HTTP 501 to the client.
2479 * FIXME: I have a stupid name and I should handle the
2480 * requests properly. Still, what we do here is rfc-
2481 * compliant, whereas ignoring or forwarding are not.
2484 * 1 : csp = Current client state (buffers, headers, etc...)
2486 * Returns : http_response if , NULL if nonmatch or handler fail
2488 *********************************************************************/
2489 struct http_response *direct_response(struct client_state *csp)
2491 struct http_response *rsp;
2492 struct list_entry *p;
2494 if ((0 == strcmpic(csp->http->gpc, "trace"))
2495 || (0 == strcmpic(csp->http->gpc, "options")))
2497 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2499 if (!strncmpic("Max-Forwards:", p->str, 13))
2501 unsigned int max_forwards;
2504 * If it's a Max-Forwards value of zero,
2505 * we have to intercept the request.
2507 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2510 * FIXME: We could handle at least TRACE here,
2511 * but that would require a verbatim copy of
2512 * the request which we don't have anymore
2514 log_error(LOG_LEVEL_HEADER,
2515 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2518 /* Get mem for response or fail*/
2519 if (NULL == (rsp = alloc_http_response()))
2521 return cgi_error_memory();
2524 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2526 free_http_response(rsp);
2527 return cgi_error_memory();
2531 rsp->reason = RSP_REASON_UNSUPPORTED;
2533 return(finish_http_response(csp, rsp));