1 const char filters_rcs[] = "$Id: filters.c,v 1.87 2007/04/30 15:53:10 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 1.87 2007/04/30 15:53:10 fabiankeil
44 * Make sure filters with dynamic jobs actually use them.
46 * Revision 1.86 2007/04/30 15:03:28 fabiankeil
47 * - Introduce dynamic pcrs jobs that can resolve variables.
48 * - Don't run redirect functions more than once,
49 * unless they are activated more than once.
51 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
52 * - Log the content size after decompression in decompress_iob()
53 * instead of pcrs_filter_response().
55 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
56 * Use dedicated header filter actions instead of abusing "filter".
57 * Replace "filter-client-headers" and "filter-client-headers"
58 * with "server-header-filter" and "client-header-filter".
60 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
61 * New config option: enforce-blocks.
63 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
64 * - Fix port handling in acl_addr() and use a temporary acl spec
65 * copy so error messages don't contain a truncated version.
66 * - Log size of iob before and after decompression.
68 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
69 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
70 * - Hide the "Go there anyway" link for blocked CONNECT
71 * requests where going there anyway doesn't work anyway.
73 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
74 * - Save the reason for generating http_responses.
75 * - Block (+block) with status code 403 instead of 404.
76 * - Use a different kludge to remember a failed decompression.
78 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
79 * Search for Max-Forwards headers case-insensitive,
80 * don't generate the "501 unsupported" message for invalid
81 * Max-Forwards values and don't increase negative ones.
83 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
84 * - Add HEAD support to finish_http_response.
85 * - Add error favicon to internal HTML error messages.
87 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
88 * Mark *csp as immutable for is_untrusted_url()
89 * and is_imageurl(). Closes FR 1237736.
91 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
92 * Integrate a modified version of Wil Mahan's
93 * zlib patch (PR #895531).
95 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
96 * Fixed gcc43 conversion warnings,
97 * changed sprintf calls to snprintf.
99 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
100 * Adjust comment in pcrs_filter_response()
101 * to recent pcrs changes. Hohoho.
103 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
104 * Don't crash if pcre returns an error code
105 * that pcrs didn't expect. Fixes BR 1621173.
107 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
108 * Modified is_untrusted_url to complain in case of
109 * write errors and to give a reason when adding new
110 * entries to the trustfile. Closes FR 1097611.
112 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
113 * Skip empty filter files in pcrs_filter_response,
114 * but don't ignore the ones that come afterwards.
115 * Fixes parts of BR 1619208.
117 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
118 * Added some sanity checks for get_last_url().
119 * Fixed possible segfault caused by my last commit.
121 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
122 * Let get_last_url() catch https URLs as well.
124 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
125 * Make sure get_last_url() behaves like advertised
126 * and fast-redirects{} can be combined with redirect{}.
128 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
129 * Implemented +redirect{s@foo@bar@} to generate
130 * a redirect based on a rewritten version of the
133 * Revision 1.66 2006/09/23 13:26:38 roro
134 * Replace TABs by spaces in source code.
136 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
137 * Fix +redirect{}. Didn't work with -fast-redirects.
139 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
140 * Block requests for untrusted URLs with status
141 * code 403 instead of 200.
143 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
144 * Don't free p which is still in use and will be later
145 * freed by free_map(). Don't claim the referrer is unknown
146 * when the client didn't set one.
148 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
149 * Feature request 595948: Re-Filter logging in single line
151 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
152 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
154 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
155 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
156 * with what was really the latest development (the v_3_0_branch branch)
158 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
159 * Multiple filter file support
161 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
162 * A few memory leaks plugged (mostly on error paths)
164 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
165 * Add the ability to check jpeg images for invalid
166 * lengths of comment blocks. Defensive strategy
167 * against the exploit:
168 * Microsoft Security Bulletin MS04-028
169 * Buffer Overrun in JPEG Processing (GDI+) Could
170 * Allow Code Execution (833987)
171 * Enabled with +inspect-jpegs in actions files.
173 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
174 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
176 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
177 * Fixed bug #839859: "See why" link URL now gets url-encoded.
179 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
182 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
183 * Added basic support for OPTIONS and TRACE HTTP methods:
184 * New function direct_response which handles OPTIONS and
185 * TRACE requests whose Max-Forwards header field is zero.
187 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
188 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
190 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
191 * - Bugfix: Executing a filters without jobs no longer results in
192 * turing off *all* filters.
193 * - Security fix: Malicious web servers can't cause a seg fault
194 * through bogus chunk sizes anymore
196 * Revision 1.58 2002/04/24 02:11:17 oes
197 * Jon's multiple AF patch: url_actions now evaluates rules
200 * Revision 1.57 2002/04/08 20:38:34 swa
203 * Revision 1.56 2002/04/05 15:51:24 oes
204 * - bugfix: error-pages now get correct request protocol
205 * - fix for invalid HTML in trust info
207 * Revision 1.55 2002/04/02 16:13:51 oes
208 * Fix: No "Go there anyway" for SSL
210 * Revision 1.54 2002/04/02 14:55:56 oes
211 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
213 * Revision 1.53 2002/03/26 22:29:54 swa
214 * we have a new homepage!
216 * Revision 1.52 2002/03/24 16:35:57 jongfoster
219 * Revision 1.51 2002/03/24 15:23:33 jongfoster
222 * Revision 1.50 2002/03/24 13:25:43 swa
223 * name change related issues
225 * Revision 1.49 2002/03/16 20:29:14 oes
228 * Revision 1.48 2002/03/13 20:25:34 oes
229 * Better logging for content filters
231 * Revision 1.47 2002/03/13 00:30:52 jongfoster
233 * Added option of always sending redirect for imageblock,
234 * currently disabled with #if 0.
236 * Revision 1.46 2002/03/12 01:42:49 oes
237 * Introduced modular filters
239 * Revision 1.45 2002/03/08 16:47:50 oes
240 * Added choice beween GIF and PNG built-in images
242 * Revision 1.44 2002/03/07 03:49:31 oes
243 * - Fixed compiler warnings etc
244 * - Changed built-in images from GIF to PNG
245 * (with regard to Unisys patent issue)
246 * - Added a 4x4 pattern PNG which is less intrusive
247 * than the logo but also clearly marks the deleted banners
249 * Revision 1.43 2002/01/22 23:51:59 jongfoster
250 * Replacing strsav() with the safer string_append().
252 * Adding missing html_encode() to error message generators. Where encoded
253 * and unencoded versions of a string were provided, removing the unencoded
256 * Revision 1.42 2002/01/17 21:00:32 jongfoster
257 * Moving all our URL and URL pattern parsing code to urlmatch.c.
259 * Using a single, simple url_match(pattern,url) function - rather than
260 * the 3-line match routine which was repeated all over the place.
262 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
264 * Using parse_http_url() to parse URLs without faking a HTTP
265 * request line for parse_http_request().
267 * Revision 1.41 2001/11/13 00:14:07 jongfoster
268 * Fixing stupid bug now I've figured out what || means.
269 * (It always returns 0 or 1, not one of it's paramaters.)
271 * Revision 1.40 2001/10/26 17:37:55 oes
272 * - Re-enabled Netscape 200/404 bug workaround in block_url():
273 * - Removed OS/2 special case
274 * - Made block_url() independant from sed() having been run
275 * - Made trust_url independant from sed() having been run
276 * - Made is_imageurl independant from sed() having been run.
277 * It now checks User-Agent: and Accept: by itself.
280 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
281 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
282 * threads to call select() simultaneously. So, it's time to do a real, live,
283 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
284 * (native). Both versions will work, but using __OS2__ offers multi-threading.
286 * Revision 1.38 2001/10/23 21:32:33 jongfoster
287 * Adding error-checking to selected functions
289 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
290 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
291 * filters.c. Added a FIXME in front of the offending code. I'll gladly
292 * put in a better/more robust fix for all parties if one is presented...
293 * It seems that just returning 200 instead of 404 would pretty much fix
294 * it for everyone, but I don't know all the history of the problem.
296 * Revision 1.36 2001/10/10 16:44:16 oes
297 * Added match_portlist function
299 * Revision 1.35 2001/10/07 15:41:23 oes
300 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
302 * New function remove_chunked_transfer_coding that strips chunked
303 * transfer coding to plain and is called by pcrs_filter_response
304 * and gif_deanimate_response if neccessary
306 * Improved handling of zero-change re_filter runs
308 * pcrs_filter_response and gif_deanimate_response now remove
309 * chunked transfer codeing before processing the body.
311 * Revision 1.34 2001/09/20 15:49:36 steudten
313 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
314 * See cgi.c fill_template().
316 * Revision 1.33 2001/09/16 17:05:14 jongfoster
317 * Removing unused #include showarg.h
319 * Revision 1.32 2001/09/16 13:21:27 jongfoster
320 * Changes to use new list functions.
322 * Revision 1.31 2001/09/16 11:38:02 jongfoster
323 * Splitting fill_template() into 2 functions:
324 * template_load() loads the file
325 * template_fill() performs the PCRS regexps.
326 * This is because the CGI edit interface has a "table row"
327 * template which is used many times in the page - this
328 * change means it's only loaded from disk once.
330 * Revision 1.30 2001/09/16 11:00:10 jongfoster
331 * New function alloc_http_response, for symmetry with free_http_response
333 * Revision 1.29 2001/09/13 23:32:40 jongfoster
334 * Moving image data to cgi.c rather than cgi.h
335 * Fixing a GPF under Win32 (and any other OS that protects global
336 * constants from being written to).
338 * Revision 1.28 2001/09/10 10:18:51 oes
339 * Silenced compiler warnings
341 * Revision 1.27 2001/08/05 16:06:20 jongfoster
342 * Modifiying "struct map" so that there are now separate header and
343 * "map_entry" structures. This means that functions which modify a
344 * map no longer need to return a pointer to the modified map.
345 * Also, it no longer reverses the order of the entries (which may be
346 * important with some advanced template substitutions).
348 * Revision 1.26 2001/07/30 22:08:36 jongfoster
349 * Tidying up #defines:
350 * - All feature #defines are now of the form FEATURE_xxx
351 * - Permanently turned off WIN_GUI_EDIT
352 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
354 * Revision 1.25 2001/07/26 10:09:46 oes
355 * Made browser detection a little less naive
357 * Revision 1.24 2001/07/25 17:22:51 oes
358 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
360 * Revision 1.23 2001/07/23 13:40:12 oes
361 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
363 * Revision 1.22 2001/07/18 12:29:34 oes
364 * - Made gif_deanimate_response respect
365 * csp->action->string[ACTION_STRING_DEANIMATE]
366 * - Logging cosmetics
368 * Revision 1.21 2001/07/13 13:59:53 oes
369 * - Introduced gif_deanimate_response which shares the
370 * generic content modification interface of pcrs_filter_response
371 * and acts as a wrapper to deanimate.c:gif_deanimate()
372 * - Renamed re_process_buffer to pcrs_filter_response
373 * - pcrs_filter_response now returns NULL on failiure
374 * - Removed all #ifdef PCRS
376 * Revision 1.20 2001/07/01 17:01:04 oes
377 * Added comments and missing return statement in is_untrusted_url()
379 * Revision 1.19 2001/06/29 21:45:41 oes
380 * Indentation, CRLF->LF, Tab-> Space
382 * Revision 1.18 2001/06/29 13:27:38 oes
383 * - Cleaned up, renamed and reorderd functions
384 * and improved comments
387 * - Ported to CGI platform. Now delivers
388 * http_response or NULL
389 * - Unified HTML and GIF generation (moved image detection
390 * and GIF generation here from jcc.c:chat())
391 * - Fixed HTTP status to:
392 * - 403 (Forbidden) for the "blocked" HTML message
393 * - 200 (OK) for GIF answers
394 * - 302 (Redirect) for redirect to GIF
397 * - Ported to CGI platform. Now delivers
398 * http_response or NULL
399 * - Separated detection of untrusted URL into
400 * (bool)is_untrusted_url
401 * - Added enforcement of untrusted requests
403 * - Moved redirect_url() from cgi.c to here
404 * and ported it to the CGI platform
406 * - Removed logentry from cancelled commit
408 * Revision 1.17 2001/06/09 10:55:28 jongfoster
409 * Changing BUFSIZ ==> BUFFER_SIZE
411 * Revision 1.16 2001/06/07 23:10:26 jongfoster
412 * Allowing unanchored domain patterns to back off and retry
413 * if they partially match. Optimized right-anchored patterns.
414 * Moving ACL and forward files into config file.
415 * Replacing struct gateway with struct forward_spec
417 * Revision 1.15 2001/06/03 19:12:00 oes
418 * extracted-CGI relevant stuff
420 * Revision 1.14 2001/06/01 10:30:55 oes
421 * Added optional left-anchoring to domaincmp
423 * Revision 1.13 2001/05/31 21:21:30 jongfoster
424 * Permissionsfile / actions file changes:
425 * - Changed "permission" to "action" throughout
426 * - changes to file format to allow string parameters
427 * - Moved helper functions to actions.c
429 * Revision 1.12 2001/05/31 17:35:20 oes
431 * - Enhanced domain part globbing with infix and prefix asterisk
432 * matching and optional unanchored operation
434 * Revision 1.11 2001/05/29 11:53:23 oes
435 * "See why" link added to "blocked" page
437 * Revision 1.10 2001/05/29 09:50:24 jongfoster
438 * Unified blocklist/imagelist/permissionslist.
439 * File format is still under discussion, but the internal changes
442 * Also modified interceptor behaviour:
443 * - We now intercept all URLs beginning with one of the following
444 * prefixes (and *only* these prefixes):
446 * * http://ijbswa.sf.net/config/
447 * * http://ijbswa.sourceforge.net/config/
448 * - New interceptors "home page" - go to http://i.j.b/ to see it.
449 * - Internal changes so that intercepted and fast redirect pages
450 * are not replaced with an image.
451 * - Interceptors now have the option to send a binary page direct
452 * to the client. (i.e. ijb-send-banner uses this)
453 * - Implemented show-url-info interceptor. (Which is why I needed
454 * the above interceptors changes - a typical URL is
455 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
456 * The previous mechanism would not have intercepted that, and
457 * if it had been intercepted then it then it would have replaced
460 * Revision 1.9 2001/05/27 22:17:04 oes
462 * - re_process_buffer no longer writes the modified buffer
463 * to the client, which was very ugly. It now returns the
464 * buffer, which it is then written by chat.
466 * - content_length now adjusts the Content-Length: header
467 * for modified documents rather than crunch()ing it.
468 * (Length info in csp->content_length, which is 0 for
469 * unmodified documents)
471 * - For this to work, sed() is called twice when filtering.
473 * Revision 1.8 2001/05/26 17:13:28 jongfoster
474 * Filled in a function comment.
476 * Revision 1.7 2001/05/26 15:26:15 jongfoster
477 * ACL feature now provides more security by immediately dropping
478 * connections from untrusted hosts.
480 * Revision 1.6 2001/05/26 00:28:36 jongfoster
481 * Automatic reloading of config file.
482 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
483 * Most of the global variables have been moved to a new
484 * struct configuration_spec, accessed through csp->config->globalname
485 * Most of the globals remaining are used by the Win32 GUI.
487 * Revision 1.5 2001/05/25 22:34:30 jongfoster
490 * Revision 1.4 2001/05/22 18:46:04 oes
492 * - Enabled filtering banners by size rather than URL
493 * by adding patterns that replace all standard banner
494 * sizes with the "Junkbuster" gif to the re_filterfile
496 * - Enabled filtering WebBugs by providing a pattern
497 * which kills all 1x1 images
499 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
500 * which is selected by the (nonstandard and therefore
501 * capital) letter 'U' in the option string.
502 * It causes the quantifiers to be ungreedy by default.
503 * Appending a ? turns back to greedy (!).
505 * - Added a new interceptor ijb-send-banner, which
506 * sends back the "Junkbuster" gif. Without imagelist or
507 * MSIE detection support, or if tinygif = 1, or the
508 * URL isn't recognized as an imageurl, a lame HTML
509 * explanation is sent instead.
511 * - Added new feature, which permits blocking remote
512 * script redirects and firing back a local redirect
514 * The feature is conditionally compiled, i.e. it
515 * can be disabled with --disable-fast-redirects,
516 * plus it must be activated by a "fast-redirects"
517 * line in the config file, has its own log level
518 * and of course wants to be displayed by show-proxy-args
519 * Note: Boy, all the #ifdefs in 1001 locations and
520 * all the fumbling with configure.in and acconfig.h
521 * were *way* more work than the feature itself :-(
523 * - Because a generic redirect template was needed for
524 * this, tinygif = 3 now uses the same.
526 * - Moved GIFs, and other static HTTP response templates
531 * - Removed some >400 CRs again (Jon, you really worked
534 * Revision 1.3 2001/05/20 16:44:47 jongfoster
535 * Removing last hardcoded Junkbusters.com URLs.
537 * Revision 1.2 2001/05/20 01:21:20 jongfoster
538 * Version 2.9.4 checkin.
539 * - Merged popupfile and cookiefile, and added control over PCRS
540 * filtering, in new "permissionsfile".
541 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
542 * file error you now get a message box (in the Win32 GUI) rather
543 * than the program exiting with no explanation.
544 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
546 * - Removed tabs from "config"
547 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
548 * - Bumped up version number.
550 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
551 * Initial import of version 2.9.3 source tree
554 *********************************************************************/
560 #include <sys/types.h>
569 #endif /* ndef __OS2__ */
570 #include <netinet/in.h>
572 #include <winsock2.h>
573 #endif /* ndef _WIN32 */
577 #endif /* def __OS2__ */
585 #include "jbsockets.h"
586 #include "miscutil.h"
590 #include "deanimate.h"
591 #include "urlmatch.h"
598 const char filters_h_rcs[] = FILTERS_H_VERSION;
600 /* Fix a problem with Solaris. There should be no effect on other
602 * Solaris's isspace() is a macro which uses it's argument directly
603 * as an array index. Therefore we need to make sure that high-bit
604 * characters generate +ve values, and ideally we also want to make
605 * the argument match the declared parameter type of "int".
607 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
611 /*********************************************************************
613 * Function : block_acl
615 * Description : Block this request?
616 * Decide yes or no based on ACL file.
619 * 1 : dst = The proxy or gateway address this is going to.
620 * Or NULL to check all possible targets.
621 * 2 : csp = Current client state (buffers, headers, etc...)
622 * Also includes the client IP address.
624 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
626 *********************************************************************/
627 int block_acl(struct access_control_addr *dst, struct client_state *csp)
629 struct access_control_list *acl = csp->config->acl;
631 /* if not using an access control list, then permit the connection */
637 /* search the list */
640 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
644 /* Just want to check if they have any access */
645 if (acl->action == ACL_PERMIT)
650 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
651 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
653 if (acl->action == ACL_PERMIT)
671 /*********************************************************************
673 * Function : acl_addr
675 * Description : Called from `load_config' to parse an ACL address.
678 * 1 : aspec = String specifying ACL address.
679 * 2 : aca = struct access_control_addr to fill in.
681 * Returns : 0 => Ok, everything else is an error.
683 *********************************************************************/
684 int acl_addr(const char *aspec, struct access_control_addr *aca)
689 char *acl_spec = NULL;
695 * Use a temporary acl spec copy so we can log
696 * the unmodified original in case of parse errors.
698 acl_spec = strdup(aspec);
699 if (acl_spec == NULL)
701 /* XXX: This will be logged as parse error. */
705 if ((p = strchr(acl_spec, '/')) != NULL)
708 if (ijb_isdigit(*p) == 0)
713 masklength = atoi(p);
716 if ((masklength < 0) || (masklength > 32))
722 if ((p = strchr(acl_spec, ':')) != NULL)
727 port = strtol(p, &endptr, 10);
729 if (port <= 0 || port > 65535 || *endptr != '\0')
736 aca->port = (unsigned long)port;
738 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
741 if (aca->addr == INADDR_NONE)
743 /* XXX: This will be logged as parse error. */
747 /* build the netmask */
749 for (i=1; i <= masklength ; i++)
751 aca->mask |= (1 << (32 - i));
754 /* now mask off the host portion of the ip address
755 * (i.e. save on the network portion of the address).
757 aca->addr = aca->addr & aca->mask;
762 #endif /* def FEATURE_ACL */
765 /*********************************************************************
767 * Function : match_portlist
769 * Description : Check if a given number is covered by a comma
770 * separated list of numbers and ranges (a,b-c,d,..)
773 * 1 : portlist = String with list
774 * 2 : port = port to check
776 * Returns : 0 => no match
779 *********************************************************************/
780 int match_portlist(const char *portlist, int port)
782 char *min, *max, *next, *portlist_copy;
784 min = next = portlist_copy = strdup(portlist);
787 * Zero-terminate first item and remember offset for next
789 if (NULL != (next = strchr(portlist_copy, (int) ',')))
795 * Loop through all items, checking for match
799 if (NULL == (max = strchr(min, (int) '-')))
802 * No dash, check for equality
804 if (port == atoi(min))
813 * This is a range, so check if between min and max,
814 * or, if max was omitted, between min and 65K
817 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
831 * Zero-terminate next item and remember offset for n+1
833 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
845 /*********************************************************************
847 * Function : block_url
849 * Description : Called from `chat'. Check to see if we need to block this.
852 * 1 : csp = Current client state (buffers, headers, etc...)
854 * Returns : NULL => unblocked, else HTTP block response
856 *********************************************************************/
857 struct http_response *block_url(struct client_state *csp)
859 struct http_response *rsp;
860 const char *new_content_type = NULL;
863 * If it's not blocked, don't block it ;-)
865 if ((csp->action->flags & ACTION_BLOCK) == 0)
869 if (csp->action->flags & ACTION_REDIRECT)
871 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
874 * Else, prepare a response
876 if (NULL == (rsp = alloc_http_response()))
878 return cgi_error_memory();
882 * If it's an image-url, send back an image or redirect
883 * as specified by the relevant +image action
885 #ifdef FEATURE_IMAGE_BLOCKING
886 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
890 /* determine HOW images should be blocked */
891 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
893 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
895 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
897 #if 1 /* Two alternative strategies, use this one for now: */
899 /* and handle accordingly: */
900 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
902 rsp->status = strdup("403 Request blocked by Privoxy");
903 if (rsp->status == NULL)
905 free_http_response(rsp);
906 return cgi_error_memory();
908 rsp->body = bindup(image_pattern_data, image_pattern_length);
909 if (rsp->body == NULL)
911 free_http_response(rsp);
912 return cgi_error_memory();
914 rsp->content_length = image_pattern_length;
916 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
918 free_http_response(rsp);
919 return cgi_error_memory();
923 else if (0 == strcmpic(p, "blank"))
925 rsp->status = strdup("403 Request blocked by Privoxy");
926 if (rsp->status == NULL)
928 free_http_response(rsp);
929 return cgi_error_memory();
931 rsp->body = bindup(image_blank_data, image_blank_length);
932 if (rsp->body == NULL)
934 free_http_response(rsp);
935 return cgi_error_memory();
937 rsp->content_length = image_blank_length;
939 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
941 free_http_response(rsp);
942 return cgi_error_memory();
948 rsp->status = strdup("302 Local Redirect from Privoxy");
949 if (rsp->status == NULL)
951 free_http_response(rsp);
952 return cgi_error_memory();
955 if (enlist_unique_header(rsp->headers, "Location", p))
957 free_http_response(rsp);
958 return cgi_error_memory();
962 #else /* Following code is disabled for now */
964 /* and handle accordingly: */
965 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
967 p = CGI_PREFIX "send-banner?type=pattern";
969 else if (0 == strcmpic(p, "blank"))
971 p = CGI_PREFIX "send-banner?type=blank";
973 rsp->status = strdup("302 Local Redirect from Privoxy");
974 if (rsp->status == NULL)
976 free_http_response(rsp);
977 return cgi_error_memory();
980 if (enlist_unique_header(rsp->headers, "Location", p))
982 free_http_response(rsp);
983 return cgi_error_memory();
985 #endif /* Preceeding code is disabled for now */
987 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
990 * Send empty document.
992 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
995 rsp->body = strdup(" ");
996 rsp->content_length = 1;
998 rsp->status = strdup("403 Request blocked by Privoxy");
999 if (rsp->status == NULL)
1001 free_http_response(rsp);
1002 return cgi_error_memory();
1004 if (new_content_type != 0)
1006 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1007 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1009 free_http_response(rsp);
1010 return cgi_error_memory();
1015 #endif /* def FEATURE_IMAGE_BLOCKING */
1018 * Else, generate an HTML "blocked" message:
1022 struct map * exports;
1026 * Workaround for stupid Netscape bug which prevents
1027 * pages from being displayed if loading a referenced
1028 * JavaScript or style sheet fails. So make it appear
1029 * as if it succeeded.
1031 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1032 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1033 && !strstr(p, "Gecko") /* save Mozilla, */
1034 && !strstr(p, "compatible") /* MSIE */
1035 && !strstr(p, "Opera")) /* and Opera. */
1037 rsp->status = strdup("200 Request for blocked URL");
1041 rsp->status = strdup("403 Request for blocked URL");
1044 if (rsp->status == NULL)
1046 free_http_response(rsp);
1047 return cgi_error_memory();
1050 exports = default_exports(csp, NULL);
1051 if (exports == NULL)
1053 free_http_response(rsp);
1054 return cgi_error_memory();
1057 #ifdef FEATURE_FORCE_LOAD
1058 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1060 * Export the force conditional block killer if
1062 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1063 * - Privoxy is configured to enforce blocks, or
1064 * - it's a CONNECT request and enforcing wouldn't work anyway.
1066 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1067 || (0 == strcmpic(csp->http->gpc, "connect")))
1068 #endif /* ndef FEATURE_FORCE_LOAD */
1070 err = map_block_killer(exports, "force-support");
1073 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1074 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1075 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1076 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1081 free_http_response(rsp);
1082 return cgi_error_memory();
1085 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1088 free_http_response(rsp);
1089 return cgi_error_memory();
1092 rsp->reason = RSP_REASON_BLOCKED;
1094 return finish_http_response(csp, rsp);
1099 #ifdef FEATURE_TRUST
1100 /*********************************************************************
1102 * Function : trust_url FIXME: I should be called distrust_url
1104 * Description : Calls is_untrusted_url to determine if the URL is trusted
1105 * and if not, returns a HTTP 403 response with a reject message.
1108 * 1 : csp = Current client state (buffers, headers, etc...)
1110 * Returns : NULL => trusted, else http_response.
1112 *********************************************************************/
1113 struct http_response *trust_url(struct client_state *csp)
1115 struct http_response *rsp;
1116 struct map * exports;
1117 char buf[BUFFER_SIZE];
1119 struct url_spec **tl;
1124 * Don't bother to work on trusted URLs
1126 if (!is_untrusted_url(csp))
1132 * Else, prepare a response:
1134 if (NULL == (rsp = alloc_http_response()))
1136 return cgi_error_memory();
1139 rsp->status = strdup("403 Request blocked by Privoxy");
1140 exports = default_exports(csp, NULL);
1141 if (exports == NULL || rsp->status == NULL)
1143 free_http_response(rsp);
1144 return cgi_error_memory();
1148 * Export the protocol, host, port, and referrer information
1150 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1151 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1152 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1154 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1156 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1160 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1166 free_http_response(rsp);
1167 return cgi_error_memory();
1171 * Export the trust list
1174 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1176 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1177 string_append(&p, buf);
1179 err = map(exports, "trusted-referrers", 1, p, 0);
1184 free_http_response(rsp);
1185 return cgi_error_memory();
1189 * Export the trust info, if available
1191 if (csp->config->trust_info->first)
1193 struct list_entry *l;
1196 for (l = csp->config->trust_info->first; l ; l = l->next)
1198 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1199 string_append(&p, buf);
1201 err = map(exports, "trust-info", 1, p, 0);
1205 err = map_block_killer(exports, "have-trust-info");
1211 free_http_response(rsp);
1212 return cgi_error_memory();
1216 * Export the force conditional block killer if
1218 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1219 * - Privoxy is configured to enforce blocks, or
1220 * - it's a CONNECT request and enforcing wouldn't work anyway.
1222 #ifdef FEATURE_FORCE_LOAD
1223 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1224 || (0 == strcmpic(csp->http->gpc, "connect")))
1226 err = map_block_killer(exports, "force-support");
1230 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1232 #else /* ifndef FEATURE_FORCE_LOAD */
1233 err = map_block_killer(exports, "force-support");
1234 #endif /* ndef FEATURE_FORCE_LOAD */
1239 free_http_response(rsp);
1240 return cgi_error_memory();
1244 * Build the response
1246 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1249 free_http_response(rsp);
1250 return cgi_error_memory();
1252 rsp->reason = RSP_REASON_UNTRUSTED;
1254 return finish_http_response(csp, rsp);
1256 #endif /* def FEATURE_TRUST */
1259 /*********************************************************************
1261 * Function : compile_dynamic_pcrs_job_list
1263 * Description : Compiles a dynamic pcrs job list (one with variables
1264 * resolved at request time)
1267 * 1 : csp = Current client state (buffers, headers, etc...)
1268 * 2 : b = The filter list to compile
1270 * Returns : NULL in case of errors, otherwise the
1273 *********************************************************************/
1274 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1276 struct list_entry *pattern;
1277 pcrs_job *job_list = NULL;
1278 pcrs_job *dummy = NULL;
1279 pcrs_job *lastjob = NULL;
1282 const struct pcrs_variable variables[] =
1284 {"url", csp->http->url, 1},
1285 {"path", csp->http->path, 1},
1286 {"host", csp->http->host, 1},
1287 {"origin", csp->ip_addr_str, 1},
1291 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1293 assert(pattern->str != NULL);
1295 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1299 log_error(LOG_LEVEL_ERROR,
1300 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1301 pattern->str, b->name, pcrs_strerror(error));
1306 if (error == PCRS_WARN_TRUNCATION)
1308 log_error(LOG_LEVEL_ERROR,
1309 "At least one of the variables in \'%s\' had to "
1310 "be truncated before compilation", pattern->str);
1312 if (job_list == NULL)
1318 lastjob->next = dummy;
1328 /*********************************************************************
1330 * Function : rewrite_url
1332 * Description : Rewrites a URL with a single pcrs command
1333 * and returns the result if it differs from the
1334 * original and isn't obviously invalid.
1337 * 1 : old_url = URL to rewrite.
1338 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1341 * Returns : NULL if the pcrs_command didn't change the url, or
1342 * the result of the modification.
1344 *********************************************************************/
1345 char *rewrite_url(char *old_url, const char *pcrs_command)
1347 char *new_url = NULL;
1351 assert(pcrs_command);
1353 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1357 log_error(LOG_LEVEL_REDIRECTS,
1358 "pcrs command \"%s\" didn't change \"%s\".",
1359 pcrs_command, old_url);
1364 log_error(LOG_LEVEL_REDIRECTS,
1365 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1366 pcrs_command, old_url, pcrs_strerror(hits));
1369 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1371 log_error(LOG_LEVEL_ERROR,
1372 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1373 "but the result doesn't look like a valid URL and will be ignored.",
1374 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1379 log_error(LOG_LEVEL_REDIRECTS,
1380 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1381 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1389 #ifdef FEATURE_FAST_REDIRECTS
1390 /*********************************************************************
1392 * Function : get_last_url
1394 * Description : Search for the last URL inside a string.
1395 * If the string already is a URL, it will
1396 * be the first URL found.
1399 * 1 : subject = the string to check
1400 * 2 : redirect_mode = +fast-redirect{} mode
1402 * Returns : NULL if no URL was found, or
1403 * the last URL found.
1405 *********************************************************************/
1406 char *get_last_url(char *subject, const char *redirect_mode)
1408 char *new_url = NULL;
1412 assert(redirect_mode);
1414 subject = strdup(subject);
1415 if (subject == NULL)
1417 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1421 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1423 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1424 new_url = url_decode(subject);
1425 if (new_url != NULL)
1432 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1436 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1439 * Find the last URL encoded in the request
1442 while ((tmp = strstr(tmp, "http://")) != NULL)
1446 tmp = (new_url != NULL) ? new_url : subject;
1447 while ((tmp = strstr(tmp, "https://")) != NULL)
1452 if ((new_url != NULL)
1453 && ( (new_url != subject)
1454 || (0 == strncmpic(subject, "http://", 7))
1455 || (0 == strncmpic(subject, "https://", 8))
1459 * Return new URL if we found a redirect
1460 * or if the subject already was a URL.
1462 * The second case makes sure that we can
1463 * chain get_last_url after another redirection check
1464 * (like rewrite_url) without losing earlier redirects.
1466 new_url = strdup(new_url);
1475 #endif /* def FEATURE_FAST_REDIRECTS */
1478 /*********************************************************************
1480 * Function : redirect_url
1482 * Description : Checks if Privoxy should answer the request with
1483 * a HTTP redirect and generates the redirect if
1487 * 1 : csp = Current client state (buffers, headers, etc...)
1489 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1491 *********************************************************************/
1492 struct http_response *redirect_url(struct client_state *csp)
1494 struct http_response *rsp;
1495 #ifdef FEATURE_FAST_REDIRECTS
1497 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1498 * as compile-time option? The user can easily disable
1499 * it in his action file.
1501 char * redirect_mode;
1502 #endif /* def FEATURE_FAST_REDIRECTS */
1503 char *old_url = NULL;
1504 char *new_url = NULL;
1505 char *redirection_string;
1507 if ((csp->action->flags & ACTION_REDIRECT))
1509 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1512 * If the redirection string begins with 's',
1513 * assume it's a pcrs command, otherwise treat it as
1514 * properly formatted URL and use it for the redirection
1517 * According to RFC 2616 section 14.30 the URL
1518 * has to be absolute and if the user tries:
1519 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1520 * she would get undefined results anyway.
1524 if (*redirection_string == 's')
1526 old_url = csp->http->url;
1527 new_url = rewrite_url(old_url, redirection_string);
1531 log_error(LOG_LEVEL_REDIRECTS,
1532 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1533 redirection_string);
1534 new_url = strdup(redirection_string);
1538 #ifdef FEATURE_FAST_REDIRECTS
1539 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1541 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1544 * If it exists, use the previously rewritten URL as input
1545 * otherwise just use the old path.
1547 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1548 new_url = get_last_url(old_url, redirect_mode);
1553 * Disable redirect checkers, so that they
1554 * will be only run more than once if the user
1555 * also enables them through tags.
1557 * From a performance point of view
1558 * it doesn't matter, but the duplicated
1559 * log messages are annoying.
1561 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1562 #endif /* def FEATURE_FAST_REDIRECTS */
1563 csp->action->flags &= ~ACTION_REDIRECT;
1565 /* Did any redirect action trigger? */
1568 if (0 == strcmpic(new_url, csp->http->url))
1570 log_error(LOG_LEVEL_ERROR,
1571 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1572 csp->http->url, new_url);
1577 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1579 if (NULL == (rsp = alloc_http_response()))
1582 return cgi_error_memory();
1585 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1586 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1589 free_http_response(rsp);
1590 return cgi_error_memory();
1592 rsp->reason = RSP_REASON_REDIRECTED;
1595 return finish_http_response(csp, rsp);
1599 /* Only reached if no redirect is required */
1605 #ifdef FEATURE_IMAGE_BLOCKING
1606 /*********************************************************************
1608 * Function : is_imageurl
1610 * Description : Given a URL, decide whether it is an image or not,
1611 * using either the info from a previous +image action
1612 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1613 * is MSIE and not on a Mac, tell from the browser's accept
1617 * 1 : csp = Current client state (buffers, headers, etc...)
1619 * Returns : True (nonzero) if URL is an image, false (0)
1622 *********************************************************************/
1623 int is_imageurl(const struct client_state *csp)
1625 #ifdef FEATURE_IMAGE_DETECT_MSIE
1628 tmp = get_header_value(csp->headers, "User-Agent:");
1629 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1631 tmp = get_header_value(csp->headers, "Accept:");
1632 if (tmp && strstr(tmp, "image/gif"))
1634 /* Client will accept HTML. If this seems counterintuitive,
1644 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1646 return ((csp->action->flags & ACTION_IMAGE) != 0);
1649 #endif /* def FEATURE_IMAGE_BLOCKING */
1652 #ifdef FEATURE_TRUST
1653 /*********************************************************************
1655 * Function : is_untrusted_url
1657 * Description : Should we "distrust" this URL (and block it)?
1659 * Yes if it matches a line in the trustfile, or if the
1660 * referrer matches a line starting with "+" in the
1665 * 1 : csp = Current client state (buffers, headers, etc...)
1667 * Returns : 0 => trusted, 1 => untrusted
1669 *********************************************************************/
1670 int is_untrusted_url(const struct client_state *csp)
1672 struct file_list *fl;
1673 struct block_spec *b;
1674 struct url_spec **trusted_url;
1675 struct http_request rhttp[1];
1676 const char * referer;
1680 * If we don't have a trustlist, we trust everybody
1682 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1687 memset(rhttp, '\0', sizeof(*rhttp));
1690 * Do we trust the request URL itself?
1692 for (b = b->next; b ; b = b->next)
1694 if (url_match(b->url, csp->http))
1700 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1702 /* no referrer was supplied */
1708 * If not, do we maybe trust its referrer?
1710 err = parse_http_url(referer, rhttp, csp);
1716 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1718 if (url_match(*trusted_url, rhttp))
1720 /* if the URL's referrer is from a trusted referrer, then
1721 * add the target spec to the trustfile as an unblocked
1722 * domain and return 0 (which means it's OK).
1727 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1731 char * new_entry = strdup("~");
1733 string_append(&new_entry, csp->http->hostport);
1735 path = csp->http->path;
1736 if ( (path[0] == '/')
1738 && ((path_end = strchr(path + 2, '/')) != NULL))
1740 /* since this path points into a user's home space
1741 * be sure to include this spec in the trustfile.
1743 int path_len = path_end - path; /* save offset */
1744 path = strdup(path); /* Copy string */
1747 path_end = path + path_len; /* regenerate ptr to new buffer */
1748 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1750 string_join(&new_entry, path);
1754 * Give a reason for generating this entry.
1756 string_append(&new_entry, " # Trusted referrer was: ");
1757 string_append(&new_entry, referer);
1759 if (new_entry != NULL)
1761 if (-1 == fprintf(fp, "%s\n", new_entry))
1763 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1764 new_entry, csp->config->trustfile);
1770 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1771 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1778 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1779 csp->http->hostport, csp->config->trustfile);
1787 #endif /* def FEATURE_TRUST */
1790 /*********************************************************************
1792 * Function : pcrs_filter_response
1794 * Description : Execute all text substitutions from all applying
1795 * +filter actions on the text buffer that's been accumulated
1796 * in csp->iob->buf. If this changes the contents, set
1797 * csp->content_length to the modified size and raise the
1798 * CSP_FLAG_MODIFIED flag.
1800 * XXX: Currently pcrs_filter_response is also responsible
1801 * for dechunking and decompressing. Both should be
1802 * done in separate functions so other content modifiers
1803 * profit as well, even if pcrs filtering is disabled.
1806 * 1 : csp = Current client state (buffers, headers, etc...)
1808 * Returns : a pointer to the (newly allocated) modified buffer.
1809 * or NULL if there were no hits or something went wrong
1811 *********************************************************************/
1812 char *pcrs_filter_response(struct client_state *csp)
1815 size_t size, prev_size;
1817 char *old = csp->iob->cur, *new = NULL;
1820 struct file_list *fl;
1821 struct re_filterfile_spec *b;
1822 struct list_entry *filtername;
1824 int i, found_filters = 0;
1829 if (csp->iob->cur >= csp->iob->eod)
1833 size = (size_t)(csp->iob->eod - csp->iob->cur);
1836 * Need to check the set of re_filterfiles...
1838 for (i = 0; i < MAX_AF_FILES; i++)
1851 if (0 == found_filters)
1853 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1858 * If the body has a "chunked" transfer-encoding,
1859 * get rid of it first, adjusting size and iob->eod
1861 if (csp->flags & CSP_FLAG_CHUNKED)
1863 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1864 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1868 csp->iob->eod = csp->iob->cur + size;
1869 csp->flags |= CSP_FLAG_MODIFIED;
1874 * If the body has a compressed transfer-encoding,
1875 * uncompress it first, adjusting size and iob->eod.
1876 * Note that decompression occurs after de-chunking.
1878 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1880 /* Notice that we at least tried to decompress. */
1881 if (JB_ERR_OK != decompress_iob(csp))
1884 * We failed to decompress the data; there's no point
1885 * in continuing since we can't filter.
1887 * XXX: Actually the Accept-Encoding header may
1888 * just be incorrect in which case we could continue
1891 * Unset CT_GZIP and CT_DEFLATE to remember not
1892 * to modify the Content-Encoding header later.
1894 csp->content_type &= ~CT_GZIP;
1895 csp->content_type &= ~CT_DEFLATE;
1900 * Decompression gives us a completely new iob,
1901 * so we need to update.
1903 size = (size_t)(csp->iob->eod - csp->iob->cur);
1904 old = csp->iob->cur;
1906 csp->flags |= CSP_FLAG_MODIFIED;
1907 csp->content_type &= ~CT_TABOO;
1911 for (i = 0; i < MAX_AF_FILES; i++)
1914 if ((NULL == fl) || (NULL == fl->f))
1917 * Either there are no filter files
1918 * left, or this filter file just
1919 * contains no valid filters.
1921 * Continue to be sure we don't miss
1922 * valid filter files that are chained
1923 * after empty or invalid ones.
1928 * For all applying +filter actions, look if a filter by that
1929 * name exists and if yes, execute it's pcrs_joblist on the
1932 for (b = fl->f; b; b = b->next)
1934 if (b->type != FT_CONTENT_FILTER)
1936 /* Skip header filters */
1940 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1941 filtername ; filtername = filtername->next)
1943 if (strcmp(b->name, filtername->str) == 0)
1945 int current_hits = 0; /* Number of hits caused by this filter */
1946 int job_number = 0; /* Which job we're currently executing */
1947 int job_hits = 0; /* How many hits the current job caused */
1948 pcrs_job *joblist = b->joblist;
1950 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1952 if (NULL == joblist)
1954 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1959 /* Apply all jobs from the joblist */
1960 for (job = joblist; NULL != job; job = job->next)
1963 job_hits = pcrs_execute(job, old, size, &new, &size);
1968 * That went well. Continue filtering
1969 * and use the result of this job as
1970 * input for the next one.
1972 current_hits += job_hits;
1973 if (old != csp->iob->cur)
1982 * This job caused an unexpected error. Inform the user
1983 * and skip the rest of the jobs in this filter. We could
1984 * continue with the next job, but usually the jobs
1985 * depend on each other or are similar enough to
1986 * fail for the same reason.
1988 * At the moment our pcrs expects the error codes of pcre 3.4,
1989 * but newer pcre versions can return additional error codes.
1990 * As a result pcrs_strerror()'s error message might be
1991 * "Unknown error ...", therefore we print the numerical value
1994 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1995 * should we use LOG_LEVEL_RE_FILTER instead?
1997 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1998 b->name, job_number, pcrs_strerror(job_hits), job_hits);
2003 if (b->dynamic) pcrs_free_joblist(joblist);
2005 log_error(LOG_LEVEL_RE_FILTER,
2006 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2007 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2009 hits += current_hits;
2016 * If there were no hits, destroy our copy and let
2017 * chat() use the original in csp->iob
2025 csp->flags |= CSP_FLAG_MODIFIED;
2026 csp->content_length = size;
2034 /*********************************************************************
2036 * Function : gif_deanimate_response
2038 * Description : Deanimate the GIF image that has been accumulated in
2039 * csp->iob->buf, set csp->content_length to the modified
2040 * size and raise the CSP_FLAG_MODIFIED flag.
2043 * 1 : csp = Current client state (buffers, headers, etc...)
2045 * Returns : a pointer to the (newly allocated) modified buffer.
2046 * or NULL in case something went wrong.
2048 *********************************************************************/
2049 char *gif_deanimate_response(struct client_state *csp)
2051 struct binbuffer *in, *out;
2055 size = (size_t)(csp->iob->eod - csp->iob->cur);
2058 * If the body has a "chunked" transfer-encoding,
2059 * get rid of it first, adjusting size and iob->eod
2061 if (csp->flags & CSP_FLAG_CHUNKED)
2063 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2064 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2068 csp->iob->eod = csp->iob->cur + size;
2069 csp->flags |= CSP_FLAG_MODIFIED;
2072 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2073 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2075 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2079 in->buffer = csp->iob->cur;
2082 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2084 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2091 if ((int)size == out->offset)
2093 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2097 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2099 csp->content_length = out->offset;
2100 csp->flags |= CSP_FLAG_MODIFIED;
2110 /*********************************************************************
2112 * Function : jpeg_inspect_response
2117 * 1 : csp = Current client state (buffers, headers, etc...)
2119 * Returns : a pointer to the (newly allocated) modified buffer
2120 * or NULL in case something went wrong.
2122 *********************************************************************/
2123 char *jpeg_inspect_response(struct client_state *csp)
2125 struct binbuffer *in = NULL;
2126 struct binbuffer *out = NULL;
2130 size = (size_t)(csp->iob->eod - csp->iob->cur);
2133 * If the body has a "chunked" transfer-encoding,
2134 * get rid of it first, adjusting size and iob->eod
2136 if (csp->flags & CSP_FLAG_CHUNKED)
2138 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2139 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2143 csp->iob->eod = csp->iob->cur + size;
2144 csp->flags |= CSP_FLAG_MODIFIED;
2147 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2149 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2153 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2155 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2159 in->buffer = csp->iob->cur;
2163 * Calling jpeg_inspect has the side-effect of creating and
2164 * modifying the image buffer of "out" directly.
2166 if (jpeg_inspect(in, out))
2168 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2176 csp->content_length = out->offset;
2177 csp->flags |= CSP_FLAG_MODIFIED;
2187 /*********************************************************************
2189 * Function : remove_chunked_transfer_coding
2191 * Description : In-situ remove the "chunked" transfer coding as defined
2192 * in rfc2616 from a buffer.
2195 * 1 : buffer = Pointer to the text buffer
2196 * 2 : size = Number of bytes to be processed
2198 * Returns : The new size, i.e. the number of bytes from buffer which
2199 * are occupied by the stripped body, or 0 in case something
2202 *********************************************************************/
2203 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2206 unsigned int chunksize = 0;
2207 char *from_p, *to_p;
2210 from_p = to_p = buffer;
2212 if (sscanf(buffer, "%x", &chunksize) != 1)
2214 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2218 while (chunksize > 0)
2220 if (NULL == (from_p = strstr(from_p, "\r\n")))
2222 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2226 if ((newsize += chunksize) >= size)
2228 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2233 memmove(to_p, from_p, (size_t) chunksize);
2234 to_p = buffer + newsize;
2235 from_p += chunksize + 2;
2237 if (sscanf(from_p, "%x", &chunksize) != 1)
2239 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2244 /* FIXME: Should this get its own loglevel? */
2245 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2251 /*********************************************************************
2253 * Function : url_actions
2255 * Description : Gets the actions for this URL.
2258 * 1 : http = http_request request for blocked URLs
2259 * 2 : csp = Current client state (buffers, headers, etc...)
2263 *********************************************************************/
2264 void url_actions(struct http_request *http,
2265 struct client_state *csp)
2267 struct file_list *fl;
2268 struct url_actions *b;
2271 init_current_action(csp->action);
2273 for (i = 0; i < MAX_AF_FILES; i++)
2275 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2280 apply_url_actions(csp->action, http, b);
2287 /*********************************************************************
2289 * Function : apply_url_actions
2291 * Description : Applies a list of URL actions.
2294 * 1 : action = Destination.
2295 * 2 : http = Current URL
2296 * 3 : b = list of URL actions to apply
2300 *********************************************************************/
2301 void apply_url_actions(struct current_action_spec *action,
2302 struct http_request *http,
2303 struct url_actions *b)
2307 /* Should never happen */
2311 for (b = b->next; NULL != b; b = b->next)
2313 if (url_match(b->url, http))
2315 merge_current_action(action, b->action);
2321 /*********************************************************************
2323 * Function : get_forward_override_settings
2325 * Description : Returns forward settings as specified with the
2326 * forward-override{} action. forward-override accepts
2327 * forward lines similar to the one used in the
2328 * configuration file, but without the URL pattern.
2334 * in the configuration file can be replaced with
2335 * the action section:
2337 * {+forward-override{forward .}}
2341 * 1 : csp = Current client state (buffers, headers, etc...)
2343 * Returns : Pointer to forwarding structure in case of success.
2344 * Invalid syntax is fatal.
2346 *********************************************************************/
2347 const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2349 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2350 char forward_settings[BUFFER_SIZE];
2351 char *http_parent = NULL;
2352 /* variable names were chosen for consistency reasons. */
2353 struct forward_spec *fwd = NULL;
2357 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2358 /* Should be enforced by load_one_actions_file() */
2359 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2361 /* Create a copy ssplit can modify */
2362 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2364 if (NULL != csp->fwd)
2367 * XXX: Currently necessary to prevent memory
2368 * leaks when the show-url-info cgi page is visited.
2370 unload_forward_spec(csp->fwd);
2374 * allocate a new forward node, valid only for
2375 * the lifetime of this request. Save its location
2376 * in csp as well, so sweep() can free it later on.
2378 fwd = csp->fwd = zalloc(sizeof(*fwd));
2381 log_error(LOG_LEVEL_FATAL,
2382 "can't allocate memory for forward-override{%s}", forward_override_line);
2383 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2386 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2387 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2389 fwd->type = SOCKS_NONE;
2391 /* Parse the parent HTTP proxy host:port */
2392 http_parent = vec[1];
2395 else if (vec_count == 3)
2397 char *socks_proxy = NULL;
2399 if (!strcasecmp(vec[0], "forward-socks4"))
2401 fwd->type = SOCKS_4;
2402 socks_proxy = vec[1];
2404 else if (!strcasecmp(vec[0], "forward-socks4a"))
2406 fwd->type = SOCKS_4A;
2407 socks_proxy = vec[1];
2410 if (NULL != socks_proxy)
2412 /* Parse the SOCKS proxy host[:port] */
2413 fwd->gateway_host = strdup(socks_proxy);
2415 if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
2417 *socks_proxy++ = '\0';
2418 fwd->gateway_port = strtol(socks_proxy, NULL, 0);
2421 if (fwd->gateway_port <= 0)
2423 fwd->gateway_port = 1080;
2426 http_parent = vec[2];
2430 if (NULL == http_parent)
2432 log_error(LOG_LEVEL_FATAL,
2433 "Invalid forward-override syntax in: %s", forward_override_line);
2434 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2437 /* Parse http forwarding settings */
2438 if (strcmp(http_parent, ".") != 0)
2440 fwd->forward_host = strdup(http_parent);
2442 if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
2444 *http_parent++ = '\0';
2445 fwd->forward_port = strtol(http_parent, NULL, 0);
2448 if (fwd->forward_port <= 0)
2450 fwd->forward_port = 8000;
2454 assert (NULL != fwd);
2456 log_error(LOG_LEVEL_CONNECT,
2457 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2462 /*********************************************************************
2464 * Function : forward_url
2466 * Description : Should we forward this to another proxy?
2468 * XXX: Should be changed to make use of csp->fwd.
2471 * 1 : http = http_request request for current URL
2472 * 2 : csp = Current client state (buffers, headers, etc...)
2474 * Returns : Pointer to forwarding information.
2476 *********************************************************************/
2477 const struct forward_spec * forward_url(struct http_request *http,
2478 struct client_state *csp)
2480 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2481 struct forward_spec *fwd = csp->config->forward;
2483 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2485 return get_forward_override_settings(csp);
2495 if (url_match(fwd->url, http))
2506 /*********************************************************************
2508 * Function : direct_response
2510 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2511 * request and if so, return a HTTP 501 to the client.
2513 * FIXME: I have a stupid name and I should handle the
2514 * requests properly. Still, what we do here is rfc-
2515 * compliant, whereas ignoring or forwarding are not.
2518 * 1 : csp = Current client state (buffers, headers, etc...)
2520 * Returns : http_response if , NULL if nonmatch or handler fail
2522 *********************************************************************/
2523 struct http_response *direct_response(struct client_state *csp)
2525 struct http_response *rsp;
2526 struct list_entry *p;
2528 if ((0 == strcmpic(csp->http->gpc, "trace"))
2529 || (0 == strcmpic(csp->http->gpc, "options")))
2531 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2533 if (!strncmpic("Max-Forwards:", p->str, 13))
2535 unsigned int max_forwards;
2538 * If it's a Max-Forwards value of zero,
2539 * we have to intercept the request.
2541 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2544 * FIXME: We could handle at least TRACE here,
2545 * but that would require a verbatim copy of
2546 * the request which we don't have anymore
2548 log_error(LOG_LEVEL_HEADER,
2549 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2552 /* Get mem for response or fail*/
2553 if (NULL == (rsp = alloc_http_response()))
2555 return cgi_error_memory();
2558 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2560 free_http_response(rsp);
2561 return cgi_error_memory();
2565 rsp->reason = RSP_REASON_UNSUPPORTED;
2567 return(finish_http_response(csp, rsp));