1 const char filters_rcs[] = "$Id: filters.c,v 1.85 2007/03/21 12:24:47 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 1.85 2007/03/21 12:24:47 fabiankeil
44 * - Log the content size after decompression in decompress_iob()
45 * instead of pcrs_filter_response().
47 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
48 * Use dedicated header filter actions instead of abusing "filter".
49 * Replace "filter-client-headers" and "filter-client-headers"
50 * with "server-header-filter" and "client-header-filter".
52 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
53 * New config option: enforce-blocks.
55 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
56 * - Fix port handling in acl_addr() and use a temporary acl spec
57 * copy so error messages don't contain a truncated version.
58 * - Log size of iob before and after decompression.
60 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
61 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
62 * - Hide the "Go there anyway" link for blocked CONNECT
63 * requests where going there anyway doesn't work anyway.
65 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
66 * - Save the reason for generating http_responses.
67 * - Block (+block) with status code 403 instead of 404.
68 * - Use a different kludge to remember a failed decompression.
70 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
71 * Search for Max-Forwards headers case-insensitive,
72 * don't generate the "501 unsupported" message for invalid
73 * Max-Forwards values and don't increase negative ones.
75 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
76 * - Add HEAD support to finish_http_response.
77 * - Add error favicon to internal HTML error messages.
79 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
80 * Mark *csp as immutable for is_untrusted_url()
81 * and is_imageurl(). Closes FR 1237736.
83 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
84 * Integrate a modified version of Wil Mahan's
85 * zlib patch (PR #895531).
87 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
88 * Fixed gcc43 conversion warnings,
89 * changed sprintf calls to snprintf.
91 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
92 * Adjust comment in pcrs_filter_response()
93 * to recent pcrs changes. Hohoho.
95 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
96 * Don't crash if pcre returns an error code
97 * that pcrs didn't expect. Fixes BR 1621173.
99 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
100 * Modified is_untrusted_url to complain in case of
101 * write errors and to give a reason when adding new
102 * entries to the trustfile. Closes FR 1097611.
104 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
105 * Skip empty filter files in pcrs_filter_response,
106 * but don't ignore the ones that come afterwards.
107 * Fixes parts of BR 1619208.
109 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
110 * Added some sanity checks for get_last_url().
111 * Fixed possible segfault caused by my last commit.
113 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
114 * Let get_last_url() catch https URLs as well.
116 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
117 * Make sure get_last_url() behaves like advertised
118 * and fast-redirects{} can be combined with redirect{}.
120 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
121 * Implemented +redirect{s@foo@bar@} to generate
122 * a redirect based on a rewritten version of the
125 * Revision 1.66 2006/09/23 13:26:38 roro
126 * Replace TABs by spaces in source code.
128 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
129 * Fix +redirect{}. Didn't work with -fast-redirects.
131 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
132 * Block requests for untrusted URLs with status
133 * code 403 instead of 200.
135 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
136 * Don't free p which is still in use and will be later
137 * freed by free_map(). Don't claim the referrer is unknown
138 * when the client didn't set one.
140 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
141 * Feature request 595948: Re-Filter logging in single line
143 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
144 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
146 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
147 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
148 * with what was really the latest development (the v_3_0_branch branch)
150 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
151 * Multiple filter file support
153 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
154 * A few memory leaks plugged (mostly on error paths)
156 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
157 * Add the ability to check jpeg images for invalid
158 * lengths of comment blocks. Defensive strategy
159 * against the exploit:
160 * Microsoft Security Bulletin MS04-028
161 * Buffer Overrun in JPEG Processing (GDI+) Could
162 * Allow Code Execution (833987)
163 * Enabled with +inspect-jpegs in actions files.
165 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
166 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
168 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
169 * Fixed bug #839859: "See why" link URL now gets url-encoded.
171 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
174 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
175 * Added basic support for OPTIONS and TRACE HTTP methods:
176 * New function direct_response which handles OPTIONS and
177 * TRACE requests whose Max-Forwards header field is zero.
179 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
180 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
182 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
183 * - Bugfix: Executing a filters without jobs no longer results in
184 * turing off *all* filters.
185 * - Security fix: Malicious web servers can't cause a seg fault
186 * through bogus chunk sizes anymore
188 * Revision 1.58 2002/04/24 02:11:17 oes
189 * Jon's multiple AF patch: url_actions now evaluates rules
192 * Revision 1.57 2002/04/08 20:38:34 swa
195 * Revision 1.56 2002/04/05 15:51:24 oes
196 * - bugfix: error-pages now get correct request protocol
197 * - fix for invalid HTML in trust info
199 * Revision 1.55 2002/04/02 16:13:51 oes
200 * Fix: No "Go there anyway" for SSL
202 * Revision 1.54 2002/04/02 14:55:56 oes
203 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
205 * Revision 1.53 2002/03/26 22:29:54 swa
206 * we have a new homepage!
208 * Revision 1.52 2002/03/24 16:35:57 jongfoster
211 * Revision 1.51 2002/03/24 15:23:33 jongfoster
214 * Revision 1.50 2002/03/24 13:25:43 swa
215 * name change related issues
217 * Revision 1.49 2002/03/16 20:29:14 oes
220 * Revision 1.48 2002/03/13 20:25:34 oes
221 * Better logging for content filters
223 * Revision 1.47 2002/03/13 00:30:52 jongfoster
225 * Added option of always sending redirect for imageblock,
226 * currently disabled with #if 0.
228 * Revision 1.46 2002/03/12 01:42:49 oes
229 * Introduced modular filters
231 * Revision 1.45 2002/03/08 16:47:50 oes
232 * Added choice beween GIF and PNG built-in images
234 * Revision 1.44 2002/03/07 03:49:31 oes
235 * - Fixed compiler warnings etc
236 * - Changed built-in images from GIF to PNG
237 * (with regard to Unisys patent issue)
238 * - Added a 4x4 pattern PNG which is less intrusive
239 * than the logo but also clearly marks the deleted banners
241 * Revision 1.43 2002/01/22 23:51:59 jongfoster
242 * Replacing strsav() with the safer string_append().
244 * Adding missing html_encode() to error message generators. Where encoded
245 * and unencoded versions of a string were provided, removing the unencoded
248 * Revision 1.42 2002/01/17 21:00:32 jongfoster
249 * Moving all our URL and URL pattern parsing code to urlmatch.c.
251 * Using a single, simple url_match(pattern,url) function - rather than
252 * the 3-line match routine which was repeated all over the place.
254 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
256 * Using parse_http_url() to parse URLs without faking a HTTP
257 * request line for parse_http_request().
259 * Revision 1.41 2001/11/13 00:14:07 jongfoster
260 * Fixing stupid bug now I've figured out what || means.
261 * (It always returns 0 or 1, not one of it's paramaters.)
263 * Revision 1.40 2001/10/26 17:37:55 oes
264 * - Re-enabled Netscape 200/404 bug workaround in block_url():
265 * - Removed OS/2 special case
266 * - Made block_url() independant from sed() having been run
267 * - Made trust_url independant from sed() having been run
268 * - Made is_imageurl independant from sed() having been run.
269 * It now checks User-Agent: and Accept: by itself.
272 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
273 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
274 * threads to call select() simultaneously. So, it's time to do a real, live,
275 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
276 * (native). Both versions will work, but using __OS2__ offers multi-threading.
278 * Revision 1.38 2001/10/23 21:32:33 jongfoster
279 * Adding error-checking to selected functions
281 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
282 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
283 * filters.c. Added a FIXME in front of the offending code. I'll gladly
284 * put in a better/more robust fix for all parties if one is presented...
285 * It seems that just returning 200 instead of 404 would pretty much fix
286 * it for everyone, but I don't know all the history of the problem.
288 * Revision 1.36 2001/10/10 16:44:16 oes
289 * Added match_portlist function
291 * Revision 1.35 2001/10/07 15:41:23 oes
292 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
294 * New function remove_chunked_transfer_coding that strips chunked
295 * transfer coding to plain and is called by pcrs_filter_response
296 * and gif_deanimate_response if neccessary
298 * Improved handling of zero-change re_filter runs
300 * pcrs_filter_response and gif_deanimate_response now remove
301 * chunked transfer codeing before processing the body.
303 * Revision 1.34 2001/09/20 15:49:36 steudten
305 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
306 * See cgi.c fill_template().
308 * Revision 1.33 2001/09/16 17:05:14 jongfoster
309 * Removing unused #include showarg.h
311 * Revision 1.32 2001/09/16 13:21:27 jongfoster
312 * Changes to use new list functions.
314 * Revision 1.31 2001/09/16 11:38:02 jongfoster
315 * Splitting fill_template() into 2 functions:
316 * template_load() loads the file
317 * template_fill() performs the PCRS regexps.
318 * This is because the CGI edit interface has a "table row"
319 * template which is used many times in the page - this
320 * change means it's only loaded from disk once.
322 * Revision 1.30 2001/09/16 11:00:10 jongfoster
323 * New function alloc_http_response, for symmetry with free_http_response
325 * Revision 1.29 2001/09/13 23:32:40 jongfoster
326 * Moving image data to cgi.c rather than cgi.h
327 * Fixing a GPF under Win32 (and any other OS that protects global
328 * constants from being written to).
330 * Revision 1.28 2001/09/10 10:18:51 oes
331 * Silenced compiler warnings
333 * Revision 1.27 2001/08/05 16:06:20 jongfoster
334 * Modifiying "struct map" so that there are now separate header and
335 * "map_entry" structures. This means that functions which modify a
336 * map no longer need to return a pointer to the modified map.
337 * Also, it no longer reverses the order of the entries (which may be
338 * important with some advanced template substitutions).
340 * Revision 1.26 2001/07/30 22:08:36 jongfoster
341 * Tidying up #defines:
342 * - All feature #defines are now of the form FEATURE_xxx
343 * - Permanently turned off WIN_GUI_EDIT
344 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
346 * Revision 1.25 2001/07/26 10:09:46 oes
347 * Made browser detection a little less naive
349 * Revision 1.24 2001/07/25 17:22:51 oes
350 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
352 * Revision 1.23 2001/07/23 13:40:12 oes
353 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
355 * Revision 1.22 2001/07/18 12:29:34 oes
356 * - Made gif_deanimate_response respect
357 * csp->action->string[ACTION_STRING_DEANIMATE]
358 * - Logging cosmetics
360 * Revision 1.21 2001/07/13 13:59:53 oes
361 * - Introduced gif_deanimate_response which shares the
362 * generic content modification interface of pcrs_filter_response
363 * and acts as a wrapper to deanimate.c:gif_deanimate()
364 * - Renamed re_process_buffer to pcrs_filter_response
365 * - pcrs_filter_response now returns NULL on failiure
366 * - Removed all #ifdef PCRS
368 * Revision 1.20 2001/07/01 17:01:04 oes
369 * Added comments and missing return statement in is_untrusted_url()
371 * Revision 1.19 2001/06/29 21:45:41 oes
372 * Indentation, CRLF->LF, Tab-> Space
374 * Revision 1.18 2001/06/29 13:27:38 oes
375 * - Cleaned up, renamed and reorderd functions
376 * and improved comments
379 * - Ported to CGI platform. Now delivers
380 * http_response or NULL
381 * - Unified HTML and GIF generation (moved image detection
382 * and GIF generation here from jcc.c:chat())
383 * - Fixed HTTP status to:
384 * - 403 (Forbidden) for the "blocked" HTML message
385 * - 200 (OK) for GIF answers
386 * - 302 (Redirect) for redirect to GIF
389 * - Ported to CGI platform. Now delivers
390 * http_response or NULL
391 * - Separated detection of untrusted URL into
392 * (bool)is_untrusted_url
393 * - Added enforcement of untrusted requests
395 * - Moved redirect_url() from cgi.c to here
396 * and ported it to the CGI platform
398 * - Removed logentry from cancelled commit
400 * Revision 1.17 2001/06/09 10:55:28 jongfoster
401 * Changing BUFSIZ ==> BUFFER_SIZE
403 * Revision 1.16 2001/06/07 23:10:26 jongfoster
404 * Allowing unanchored domain patterns to back off and retry
405 * if they partially match. Optimized right-anchored patterns.
406 * Moving ACL and forward files into config file.
407 * Replacing struct gateway with struct forward_spec
409 * Revision 1.15 2001/06/03 19:12:00 oes
410 * extracted-CGI relevant stuff
412 * Revision 1.14 2001/06/01 10:30:55 oes
413 * Added optional left-anchoring to domaincmp
415 * Revision 1.13 2001/05/31 21:21:30 jongfoster
416 * Permissionsfile / actions file changes:
417 * - Changed "permission" to "action" throughout
418 * - changes to file format to allow string parameters
419 * - Moved helper functions to actions.c
421 * Revision 1.12 2001/05/31 17:35:20 oes
423 * - Enhanced domain part globbing with infix and prefix asterisk
424 * matching and optional unanchored operation
426 * Revision 1.11 2001/05/29 11:53:23 oes
427 * "See why" link added to "blocked" page
429 * Revision 1.10 2001/05/29 09:50:24 jongfoster
430 * Unified blocklist/imagelist/permissionslist.
431 * File format is still under discussion, but the internal changes
434 * Also modified interceptor behaviour:
435 * - We now intercept all URLs beginning with one of the following
436 * prefixes (and *only* these prefixes):
438 * * http://ijbswa.sf.net/config/
439 * * http://ijbswa.sourceforge.net/config/
440 * - New interceptors "home page" - go to http://i.j.b/ to see it.
441 * - Internal changes so that intercepted and fast redirect pages
442 * are not replaced with an image.
443 * - Interceptors now have the option to send a binary page direct
444 * to the client. (i.e. ijb-send-banner uses this)
445 * - Implemented show-url-info interceptor. (Which is why I needed
446 * the above interceptors changes - a typical URL is
447 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
448 * The previous mechanism would not have intercepted that, and
449 * if it had been intercepted then it then it would have replaced
452 * Revision 1.9 2001/05/27 22:17:04 oes
454 * - re_process_buffer no longer writes the modified buffer
455 * to the client, which was very ugly. It now returns the
456 * buffer, which it is then written by chat.
458 * - content_length now adjusts the Content-Length: header
459 * for modified documents rather than crunch()ing it.
460 * (Length info in csp->content_length, which is 0 for
461 * unmodified documents)
463 * - For this to work, sed() is called twice when filtering.
465 * Revision 1.8 2001/05/26 17:13:28 jongfoster
466 * Filled in a function comment.
468 * Revision 1.7 2001/05/26 15:26:15 jongfoster
469 * ACL feature now provides more security by immediately dropping
470 * connections from untrusted hosts.
472 * Revision 1.6 2001/05/26 00:28:36 jongfoster
473 * Automatic reloading of config file.
474 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
475 * Most of the global variables have been moved to a new
476 * struct configuration_spec, accessed through csp->config->globalname
477 * Most of the globals remaining are used by the Win32 GUI.
479 * Revision 1.5 2001/05/25 22:34:30 jongfoster
482 * Revision 1.4 2001/05/22 18:46:04 oes
484 * - Enabled filtering banners by size rather than URL
485 * by adding patterns that replace all standard banner
486 * sizes with the "Junkbuster" gif to the re_filterfile
488 * - Enabled filtering WebBugs by providing a pattern
489 * which kills all 1x1 images
491 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
492 * which is selected by the (nonstandard and therefore
493 * capital) letter 'U' in the option string.
494 * It causes the quantifiers to be ungreedy by default.
495 * Appending a ? turns back to greedy (!).
497 * - Added a new interceptor ijb-send-banner, which
498 * sends back the "Junkbuster" gif. Without imagelist or
499 * MSIE detection support, or if tinygif = 1, or the
500 * URL isn't recognized as an imageurl, a lame HTML
501 * explanation is sent instead.
503 * - Added new feature, which permits blocking remote
504 * script redirects and firing back a local redirect
506 * The feature is conditionally compiled, i.e. it
507 * can be disabled with --disable-fast-redirects,
508 * plus it must be activated by a "fast-redirects"
509 * line in the config file, has its own log level
510 * and of course wants to be displayed by show-proxy-args
511 * Note: Boy, all the #ifdefs in 1001 locations and
512 * all the fumbling with configure.in and acconfig.h
513 * were *way* more work than the feature itself :-(
515 * - Because a generic redirect template was needed for
516 * this, tinygif = 3 now uses the same.
518 * - Moved GIFs, and other static HTTP response templates
523 * - Removed some >400 CRs again (Jon, you really worked
526 * Revision 1.3 2001/05/20 16:44:47 jongfoster
527 * Removing last hardcoded Junkbusters.com URLs.
529 * Revision 1.2 2001/05/20 01:21:20 jongfoster
530 * Version 2.9.4 checkin.
531 * - Merged popupfile and cookiefile, and added control over PCRS
532 * filtering, in new "permissionsfile".
533 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
534 * file error you now get a message box (in the Win32 GUI) rather
535 * than the program exiting with no explanation.
536 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
538 * - Removed tabs from "config"
539 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
540 * - Bumped up version number.
542 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
543 * Initial import of version 2.9.3 source tree
546 *********************************************************************/
552 #include <sys/types.h>
561 #endif /* ndef __OS2__ */
562 #include <netinet/in.h>
564 #include <winsock2.h>
565 #endif /* ndef _WIN32 */
569 #endif /* def __OS2__ */
577 #include "jbsockets.h"
578 #include "miscutil.h"
582 #include "deanimate.h"
583 #include "urlmatch.h"
589 const char filters_h_rcs[] = FILTERS_H_VERSION;
591 /* Fix a problem with Solaris. There should be no effect on other
593 * Solaris's isspace() is a macro which uses it's argument directly
594 * as an array index. Therefore we need to make sure that high-bit
595 * characters generate +ve values, and ideally we also want to make
596 * the argument match the declared parameter type of "int".
598 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
602 /*********************************************************************
604 * Function : block_acl
606 * Description : Block this request?
607 * Decide yes or no based on ACL file.
610 * 1 : dst = The proxy or gateway address this is going to.
611 * Or NULL to check all possible targets.
612 * 2 : csp = Current client state (buffers, headers, etc...)
613 * Also includes the client IP address.
615 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
617 *********************************************************************/
618 int block_acl(struct access_control_addr *dst, struct client_state *csp)
620 struct access_control_list *acl = csp->config->acl;
622 /* if not using an access control list, then permit the connection */
628 /* search the list */
631 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
635 /* Just want to check if they have any access */
636 if (acl->action == ACL_PERMIT)
641 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
642 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
644 if (acl->action == ACL_PERMIT)
662 /*********************************************************************
664 * Function : acl_addr
666 * Description : Called from `load_config' to parse an ACL address.
669 * 1 : aspec = String specifying ACL address.
670 * 2 : aca = struct access_control_addr to fill in.
672 * Returns : 0 => Ok, everything else is an error.
674 *********************************************************************/
675 int acl_addr(const char *aspec, struct access_control_addr *aca)
680 char *acl_spec = NULL;
686 * Use a temporary acl spec copy so we can log
687 * the unmodified original in case of parse errors.
689 acl_spec = strdup(aspec);
690 if (acl_spec == NULL)
692 /* XXX: This will be logged as parse error. */
696 if ((p = strchr(acl_spec, '/')) != NULL)
699 if (ijb_isdigit(*p) == 0)
704 masklength = atoi(p);
707 if ((masklength < 0) || (masklength > 32))
713 if ((p = strchr(acl_spec, ':')) != NULL)
718 port = strtol(p, &endptr, 10);
720 if (port <= 0 || port > 65535 || *endptr != '\0')
727 aca->port = (unsigned long)port;
729 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
732 if (aca->addr == INADDR_NONE)
734 /* XXX: This will be logged as parse error. */
738 /* build the netmask */
740 for (i=1; i <= masklength ; i++)
742 aca->mask |= (1 << (32 - i));
745 /* now mask off the host portion of the ip address
746 * (i.e. save on the network portion of the address).
748 aca->addr = aca->addr & aca->mask;
753 #endif /* def FEATURE_ACL */
756 /*********************************************************************
758 * Function : match_portlist
760 * Description : Check if a given number is covered by a comma
761 * separated list of numbers and ranges (a,b-c,d,..)
764 * 1 : portlist = String with list
765 * 2 : port = port to check
767 * Returns : 0 => no match
770 *********************************************************************/
771 int match_portlist(const char *portlist, int port)
773 char *min, *max, *next, *portlist_copy;
775 min = next = portlist_copy = strdup(portlist);
778 * Zero-terminate first item and remember offset for next
780 if (NULL != (next = strchr(portlist_copy, (int) ',')))
786 * Loop through all items, checking for match
790 if (NULL == (max = strchr(min, (int) '-')))
793 * No dash, check for equality
795 if (port == atoi(min))
804 * This is a range, so check if between min and max,
805 * or, if max was omitted, between min and 65K
808 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
822 * Zero-terminate next item and remember offset for n+1
824 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
836 /*********************************************************************
838 * Function : block_url
840 * Description : Called from `chat'. Check to see if we need to block this.
843 * 1 : csp = Current client state (buffers, headers, etc...)
845 * Returns : NULL => unblocked, else HTTP block response
847 *********************************************************************/
848 struct http_response *block_url(struct client_state *csp)
850 struct http_response *rsp;
851 const char *new_content_type = NULL;
854 * If it's not blocked, don't block it ;-)
856 if ((csp->action->flags & ACTION_BLOCK) == 0)
860 if (csp->action->flags & ACTION_REDIRECT)
862 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
865 * Else, prepare a response
867 if (NULL == (rsp = alloc_http_response()))
869 return cgi_error_memory();
873 * If it's an image-url, send back an image or redirect
874 * as specified by the relevant +image action
876 #ifdef FEATURE_IMAGE_BLOCKING
877 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
881 /* determine HOW images should be blocked */
882 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
884 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
886 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
888 #if 1 /* Two alternative strategies, use this one for now: */
890 /* and handle accordingly: */
891 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
893 rsp->status = strdup("403 Request blocked by Privoxy");
894 if (rsp->status == NULL)
896 free_http_response(rsp);
897 return cgi_error_memory();
899 rsp->body = bindup(image_pattern_data, image_pattern_length);
900 if (rsp->body == NULL)
902 free_http_response(rsp);
903 return cgi_error_memory();
905 rsp->content_length = image_pattern_length;
907 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
909 free_http_response(rsp);
910 return cgi_error_memory();
914 else if (0 == strcmpic(p, "blank"))
916 rsp->status = strdup("403 Request blocked by Privoxy");
917 if (rsp->status == NULL)
919 free_http_response(rsp);
920 return cgi_error_memory();
922 rsp->body = bindup(image_blank_data, image_blank_length);
923 if (rsp->body == NULL)
925 free_http_response(rsp);
926 return cgi_error_memory();
928 rsp->content_length = image_blank_length;
930 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
932 free_http_response(rsp);
933 return cgi_error_memory();
939 rsp->status = strdup("302 Local Redirect from Privoxy");
940 if (rsp->status == NULL)
942 free_http_response(rsp);
943 return cgi_error_memory();
946 if (enlist_unique_header(rsp->headers, "Location", p))
948 free_http_response(rsp);
949 return cgi_error_memory();
953 #else /* Following code is disabled for now */
955 /* and handle accordingly: */
956 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
958 p = CGI_PREFIX "send-banner?type=pattern";
960 else if (0 == strcmpic(p, "blank"))
962 p = CGI_PREFIX "send-banner?type=blank";
964 rsp->status = strdup("302 Local Redirect from Privoxy");
965 if (rsp->status == NULL)
967 free_http_response(rsp);
968 return cgi_error_memory();
971 if (enlist_unique_header(rsp->headers, "Location", p))
973 free_http_response(rsp);
974 return cgi_error_memory();
976 #endif /* Preceeding code is disabled for now */
978 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
981 * Send empty document.
983 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
986 rsp->body = strdup(" ");
987 rsp->content_length = 1;
989 rsp->status = strdup("403 Request blocked by Privoxy");
990 if (rsp->status == NULL)
992 free_http_response(rsp);
993 return cgi_error_memory();
995 if (new_content_type != 0)
997 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
998 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1000 free_http_response(rsp);
1001 return cgi_error_memory();
1006 #endif /* def FEATURE_IMAGE_BLOCKING */
1009 * Else, generate an HTML "blocked" message:
1013 struct map * exports;
1017 * Workaround for stupid Netscape bug which prevents
1018 * pages from being displayed if loading a referenced
1019 * JavaScript or style sheet fails. So make it appear
1020 * as if it succeeded.
1022 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1023 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1024 && !strstr(p, "Gecko") /* save Mozilla, */
1025 && !strstr(p, "compatible") /* MSIE */
1026 && !strstr(p, "Opera")) /* and Opera. */
1028 rsp->status = strdup("200 Request for blocked URL");
1032 rsp->status = strdup("403 Request for blocked URL");
1035 if (rsp->status == NULL)
1037 free_http_response(rsp);
1038 return cgi_error_memory();
1041 exports = default_exports(csp, NULL);
1042 if (exports == NULL)
1044 free_http_response(rsp);
1045 return cgi_error_memory();
1048 #ifdef FEATURE_FORCE_LOAD
1049 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1051 * Export the force conditional block killer if
1053 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1054 * - Privoxy is configured to enforce blocks, or
1055 * - it's a CONNECT request and enforcing wouldn't work anyway.
1057 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1058 || (0 == strcmpic(csp->http->gpc, "connect")))
1059 #endif /* ndef FEATURE_FORCE_LOAD */
1061 err = map_block_killer(exports, "force-support");
1064 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1065 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1066 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1067 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1072 free_http_response(rsp);
1073 return cgi_error_memory();
1076 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1079 free_http_response(rsp);
1080 return cgi_error_memory();
1083 rsp->reason = RSP_REASON_BLOCKED;
1085 return finish_http_response(csp, rsp);
1090 #ifdef FEATURE_TRUST
1091 /*********************************************************************
1093 * Function : trust_url FIXME: I should be called distrust_url
1095 * Description : Calls is_untrusted_url to determine if the URL is trusted
1096 * and if not, returns a HTTP 403 response with a reject message.
1099 * 1 : csp = Current client state (buffers, headers, etc...)
1101 * Returns : NULL => trusted, else http_response.
1103 *********************************************************************/
1104 struct http_response *trust_url(struct client_state *csp)
1106 struct http_response *rsp;
1107 struct map * exports;
1108 char buf[BUFFER_SIZE];
1110 struct url_spec **tl;
1115 * Don't bother to work on trusted URLs
1117 if (!is_untrusted_url(csp))
1123 * Else, prepare a response:
1125 if (NULL == (rsp = alloc_http_response()))
1127 return cgi_error_memory();
1130 rsp->status = strdup("403 Request blocked by Privoxy");
1131 exports = default_exports(csp, NULL);
1132 if (exports == NULL || rsp->status == NULL)
1134 free_http_response(rsp);
1135 return cgi_error_memory();
1139 * Export the protocol, host, port, and referrer information
1141 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1142 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1143 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1145 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1147 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1151 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1157 free_http_response(rsp);
1158 return cgi_error_memory();
1162 * Export the trust list
1165 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1167 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1168 string_append(&p, buf);
1170 err = map(exports, "trusted-referrers", 1, p, 0);
1175 free_http_response(rsp);
1176 return cgi_error_memory();
1180 * Export the trust info, if available
1182 if (csp->config->trust_info->first)
1184 struct list_entry *l;
1187 for (l = csp->config->trust_info->first; l ; l = l->next)
1189 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1190 string_append(&p, buf);
1192 err = map(exports, "trust-info", 1, p, 0);
1196 err = map_block_killer(exports, "have-trust-info");
1202 free_http_response(rsp);
1203 return cgi_error_memory();
1207 * Export the force conditional block killer if
1209 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1210 * - Privoxy is configured to enforce blocks, or
1211 * - it's a CONNECT request and enforcing wouldn't work anyway.
1213 #ifdef FEATURE_FORCE_LOAD
1214 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1215 || (0 == strcmpic(csp->http->gpc, "connect")))
1217 err = map_block_killer(exports, "force-support");
1221 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1223 #else /* ifndef FEATURE_FORCE_LOAD */
1224 err = map_block_killer(exports, "force-support");
1225 #endif /* ndef FEATURE_FORCE_LOAD */
1230 free_http_response(rsp);
1231 return cgi_error_memory();
1235 * Build the response
1237 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1240 free_http_response(rsp);
1241 return cgi_error_memory();
1243 rsp->reason = RSP_REASON_UNTRUSTED;
1245 return finish_http_response(csp, rsp);
1247 #endif /* def FEATURE_TRUST */
1250 /*********************************************************************
1252 * Function : compile_dynamic_pcrs_job_list
1254 * Description : Compiles a dynamic pcrs job list (one with variables
1255 * resolved at request time)
1258 * 1 : csp = Current client state (buffers, headers, etc...)
1259 * 2 : b = The filter list to compile
1261 * Returns : NULL in case of errors, otherwise the
1264 *********************************************************************/
1265 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1267 struct list_entry *pattern;
1268 pcrs_job *job_list = NULL;
1269 pcrs_job *dummy = NULL;
1270 pcrs_job *lastjob = NULL;
1273 const struct pcrs_variable variables[] =
1275 {"url", csp->http->url, 1},
1276 {"path", csp->http->path, 1},
1277 {"host", csp->http->host, 1},
1278 {"origin", csp->ip_addr_str, 1},
1282 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1284 assert(pattern->str != NULL);
1286 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1290 log_error(LOG_LEVEL_ERROR,
1291 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1292 pattern->str, b->name, pcrs_strerror(error));
1297 if (error == PCRS_WARN_TRUNCATION)
1299 log_error(LOG_LEVEL_ERROR,
1300 "At least one of the variables in \'%s\' had to "
1301 "be truncated before compilation", pattern->str);
1303 if (job_list == NULL)
1309 lastjob->next = dummy;
1319 /*********************************************************************
1321 * Function : rewrite_url
1323 * Description : Rewrites a URL with a single pcrs command
1324 * and returns the result if it differs from the
1325 * original and isn't obviously invalid.
1328 * 1 : old_url = URL to rewrite.
1329 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1332 * Returns : NULL if the pcrs_command didn't change the url, or
1333 * the result of the modification.
1335 *********************************************************************/
1336 char *rewrite_url(char *old_url, const char *pcrs_command)
1338 char *new_url = NULL;
1342 assert(pcrs_command);
1344 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1348 log_error(LOG_LEVEL_REDIRECTS,
1349 "pcrs command \"%s\" didn't change \"%s\".",
1350 pcrs_command, old_url);
1355 log_error(LOG_LEVEL_REDIRECTS,
1356 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1357 pcrs_command, old_url, pcrs_strerror(hits));
1360 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1362 log_error(LOG_LEVEL_ERROR,
1363 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1364 "but the result doesn't look like a valid URL and will be ignored.",
1365 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1370 log_error(LOG_LEVEL_REDIRECTS,
1371 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1372 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1380 #ifdef FEATURE_FAST_REDIRECTS
1381 /*********************************************************************
1383 * Function : get_last_url
1385 * Description : Search for the last URL inside a string.
1386 * If the string already is a URL, it will
1387 * be the first URL found.
1390 * 1 : subject = the string to check
1391 * 2 : redirect_mode = +fast-redirect{} mode
1393 * Returns : NULL if no URL was found, or
1394 * the last URL found.
1396 *********************************************************************/
1397 char *get_last_url(char *subject, const char *redirect_mode)
1399 char *new_url = NULL;
1403 assert(redirect_mode);
1405 subject = strdup(subject);
1406 if (subject == NULL)
1408 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1412 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1414 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1415 new_url = url_decode(subject);
1416 if (new_url != NULL)
1423 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1427 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1430 * Find the last URL encoded in the request
1433 while ((tmp = strstr(tmp, "http://")) != NULL)
1437 tmp = (new_url != NULL) ? new_url : subject;
1438 while ((tmp = strstr(tmp, "https://")) != NULL)
1443 if ((new_url != NULL)
1444 && ( (new_url != subject)
1445 || (0 == strncmpic(subject, "http://", 7))
1446 || (0 == strncmpic(subject, "https://", 8))
1450 * Return new URL if we found a redirect
1451 * or if the subject already was a URL.
1453 * The second case makes sure that we can
1454 * chain get_last_url after another redirection check
1455 * (like rewrite_url) without losing earlier redirects.
1457 new_url = strdup(new_url);
1466 #endif /* def FEATURE_FAST_REDIRECTS */
1469 /*********************************************************************
1471 * Function : redirect_url
1473 * Description : Checks if Privoxy should answer the request with
1474 * a HTTP redirect and generates the redirect if
1478 * 1 : csp = Current client state (buffers, headers, etc...)
1480 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1482 *********************************************************************/
1483 struct http_response *redirect_url(struct client_state *csp)
1485 struct http_response *rsp;
1486 #ifdef FEATURE_FAST_REDIRECTS
1488 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1489 * as compile-time option? The user can easily disable
1490 * it in his action file.
1492 char * redirect_mode;
1493 #endif /* def FEATURE_FAST_REDIRECTS */
1494 char *old_url = NULL;
1495 char *new_url = NULL;
1496 char *redirection_string;
1498 if ((csp->action->flags & ACTION_REDIRECT))
1500 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1503 * If the redirection string begins with 's',
1504 * assume it's a pcrs command, otherwise treat it as
1505 * properly formatted URL and use it for the redirection
1508 * According to RFC 2616 section 14.30 the URL
1509 * has to be absolute and if the user tries:
1510 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1511 * she would get undefined results anyway.
1515 if (*redirection_string == 's')
1517 old_url = csp->http->url;
1518 new_url = rewrite_url(old_url, redirection_string);
1522 log_error(LOG_LEVEL_REDIRECTS,
1523 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1524 redirection_string);
1525 new_url = strdup(redirection_string);
1529 #ifdef FEATURE_FAST_REDIRECTS
1530 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1532 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1535 * If it exists, use the previously rewritten URL as input
1536 * otherwise just use the old path.
1538 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1539 new_url = get_last_url(old_url, redirect_mode);
1544 * Disable redirect checkers, so that they
1545 * will be only run more than once if the user
1546 * also enables them through tags.
1548 * From a performance point of view
1549 * it doesn't matter, but the duplicated
1550 * log messages are annoying.
1552 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1553 #endif /* def FEATURE_FAST_REDIRECTS */
1554 csp->action->flags &= ~ACTION_REDIRECT;
1556 /* Did any redirect action trigger? */
1559 if (0 == strcmpic(new_url, csp->http->url))
1561 log_error(LOG_LEVEL_ERROR,
1562 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1563 csp->http->url, new_url);
1568 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1570 if (NULL == (rsp = alloc_http_response()))
1573 return cgi_error_memory();
1576 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1577 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1580 free_http_response(rsp);
1581 return cgi_error_memory();
1583 rsp->reason = RSP_REASON_REDIRECTED;
1586 return finish_http_response(csp, rsp);
1590 /* Only reached if no redirect is required */
1596 #ifdef FEATURE_IMAGE_BLOCKING
1597 /*********************************************************************
1599 * Function : is_imageurl
1601 * Description : Given a URL, decide whether it is an image or not,
1602 * using either the info from a previous +image action
1603 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1604 * is MSIE and not on a Mac, tell from the browser's accept
1608 * 1 : csp = Current client state (buffers, headers, etc...)
1610 * Returns : True (nonzero) if URL is an image, false (0)
1613 *********************************************************************/
1614 int is_imageurl(const struct client_state *csp)
1616 #ifdef FEATURE_IMAGE_DETECT_MSIE
1619 tmp = get_header_value(csp->headers, "User-Agent:");
1620 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1622 tmp = get_header_value(csp->headers, "Accept:");
1623 if (tmp && strstr(tmp, "image/gif"))
1625 /* Client will accept HTML. If this seems counterintuitive,
1635 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1637 return ((csp->action->flags & ACTION_IMAGE) != 0);
1640 #endif /* def FEATURE_IMAGE_BLOCKING */
1643 #ifdef FEATURE_TRUST
1644 /*********************************************************************
1646 * Function : is_untrusted_url
1648 * Description : Should we "distrust" this URL (and block it)?
1650 * Yes if it matches a line in the trustfile, or if the
1651 * referrer matches a line starting with "+" in the
1656 * 1 : csp = Current client state (buffers, headers, etc...)
1658 * Returns : 0 => trusted, 1 => untrusted
1660 *********************************************************************/
1661 int is_untrusted_url(const struct client_state *csp)
1663 struct file_list *fl;
1664 struct block_spec *b;
1665 struct url_spec **trusted_url;
1666 struct http_request rhttp[1];
1667 const char * referer;
1671 * If we don't have a trustlist, we trust everybody
1673 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1678 memset(rhttp, '\0', sizeof(*rhttp));
1681 * Do we trust the request URL itself?
1683 for (b = b->next; b ; b = b->next)
1685 if (url_match(b->url, csp->http))
1691 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1693 /* no referrer was supplied */
1699 * If not, do we maybe trust its referrer?
1701 err = parse_http_url(referer, rhttp, csp);
1707 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1709 if (url_match(*trusted_url, rhttp))
1711 /* if the URL's referrer is from a trusted referrer, then
1712 * add the target spec to the trustfile as an unblocked
1713 * domain and return 0 (which means it's OK).
1718 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1722 char * new_entry = strdup("~");
1724 string_append(&new_entry, csp->http->hostport);
1726 path = csp->http->path;
1727 if ( (path[0] == '/')
1729 && ((path_end = strchr(path + 2, '/')) != NULL))
1731 /* since this path points into a user's home space
1732 * be sure to include this spec in the trustfile.
1734 int path_len = path_end - path; /* save offset */
1735 path = strdup(path); /* Copy string */
1738 path_end = path + path_len; /* regenerate ptr to new buffer */
1739 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1741 string_join(&new_entry, path);
1745 * Give a reason for generating this entry.
1747 string_append(&new_entry, " # Trusted referrer was: ");
1748 string_append(&new_entry, referer);
1750 if (new_entry != NULL)
1752 if (-1 == fprintf(fp, "%s\n", new_entry))
1754 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1755 new_entry, csp->config->trustfile);
1761 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1762 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1769 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1770 csp->http->hostport, csp->config->trustfile);
1778 #endif /* def FEATURE_TRUST */
1781 /*********************************************************************
1783 * Function : pcrs_filter_response
1785 * Description : Execute all text substitutions from all applying
1786 * +filter actions on the text buffer that's been accumulated
1787 * in csp->iob->buf. If this changes the contents, set
1788 * csp->content_length to the modified size and raise the
1789 * CSP_FLAG_MODIFIED flag.
1791 * XXX: Currently pcrs_filter_response is also responsible
1792 * for dechunking and decompressing. Both should be
1793 * done in separate functions so other content modifiers
1794 * profit as well, even if pcrs filtering is disabled.
1797 * 1 : csp = Current client state (buffers, headers, etc...)
1799 * Returns : a pointer to the (newly allocated) modified buffer.
1800 * or NULL if there were no hits or something went wrong
1802 *********************************************************************/
1803 char *pcrs_filter_response(struct client_state *csp)
1806 size_t size, prev_size;
1808 char *old = csp->iob->cur, *new = NULL;
1811 struct file_list *fl;
1812 struct re_filterfile_spec *b;
1813 struct list_entry *filtername;
1815 int i, found_filters = 0;
1820 if (csp->iob->cur >= csp->iob->eod)
1824 size = (size_t)(csp->iob->eod - csp->iob->cur);
1827 * Need to check the set of re_filterfiles...
1829 for (i = 0; i < MAX_AF_FILES; i++)
1842 if (0 == found_filters)
1844 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1849 * If the body has a "chunked" transfer-encoding,
1850 * get rid of it first, adjusting size and iob->eod
1852 if (csp->flags & CSP_FLAG_CHUNKED)
1854 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1855 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1859 csp->iob->eod = csp->iob->cur + size;
1860 csp->flags |= CSP_FLAG_MODIFIED;
1865 * If the body has a compressed transfer-encoding,
1866 * uncompress it first, adjusting size and iob->eod.
1867 * Note that decompression occurs after de-chunking.
1869 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1871 /* Notice that we at least tried to decompress. */
1872 if (JB_ERR_OK != decompress_iob(csp))
1875 * We failed to decompress the data; there's no point
1876 * in continuing since we can't filter.
1878 * XXX: Actually the Accept-Encoding header may
1879 * just be incorrect in which case we could continue
1882 * Unset CT_GZIP and CT_DEFLATE to remember not
1883 * to modify the Content-Encoding header later.
1885 csp->content_type &= ~CT_GZIP;
1886 csp->content_type &= ~CT_DEFLATE;
1891 * Decompression gives us a completely new iob,
1892 * so we need to update.
1894 size = (size_t)(csp->iob->eod - csp->iob->cur);
1895 old = csp->iob->cur;
1897 csp->flags |= CSP_FLAG_MODIFIED;
1898 csp->content_type &= ~CT_TABOO;
1902 for (i = 0; i < MAX_AF_FILES; i++)
1905 if ((NULL == fl) || (NULL == fl->f))
1908 * Either there are no filter files
1909 * left, or this filter file just
1910 * contains no valid filters.
1912 * Continue to be sure we don't miss
1913 * valid filter files that are chained
1914 * after empty or invalid ones.
1919 * For all applying +filter actions, look if a filter by that
1920 * name exists and if yes, execute it's pcrs_joblist on the
1923 for (b = fl->f; b; b = b->next)
1925 if (b->type != FT_CONTENT_FILTER)
1927 /* Skip header filters */
1931 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1932 filtername ; filtername = filtername->next)
1934 if (strcmp(b->name, filtername->str) == 0)
1936 int current_hits = 0; /* Number of hits caused by this filter */
1937 int job_number = 0; /* Which job we're currently executing */
1938 int job_hits = 0; /* How many hits the current job caused */
1939 pcrs_job *joblist = b->joblist;
1941 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1943 if (NULL == joblist)
1945 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1950 /* Apply all jobs from the joblist */
1951 for (job = b->joblist; NULL != job; job = job->next)
1954 job_hits = pcrs_execute(job, old, size, &new, &size);
1959 * That went well. Continue filtering
1960 * and use the result of this job as
1961 * input for the next one.
1963 current_hits += job_hits;
1964 if (old != csp->iob->cur)
1973 * This job caused an unexpected error. Inform the user
1974 * and skip the rest of the jobs in this filter. We could
1975 * continue with the next job, but usually the jobs
1976 * depend on each other or are similar enough to
1977 * fail for the same reason.
1979 * At the moment our pcrs expects the error codes of pcre 3.4,
1980 * but newer pcre versions can return additional error codes.
1981 * As a result pcrs_strerror()'s error message might be
1982 * "Unknown error ...", therefore we print the numerical value
1985 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1986 * should we use LOG_LEVEL_RE_FILTER instead?
1988 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1989 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1994 if (b->dynamic) pcrs_free_joblist(joblist);
1996 log_error(LOG_LEVEL_RE_FILTER,
1997 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1998 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2000 hits += current_hits;
2007 * If there were no hits, destroy our copy and let
2008 * chat() use the original in csp->iob
2016 csp->flags |= CSP_FLAG_MODIFIED;
2017 csp->content_length = size;
2025 /*********************************************************************
2027 * Function : gif_deanimate_response
2029 * Description : Deanimate the GIF image that has been accumulated in
2030 * csp->iob->buf, set csp->content_length to the modified
2031 * size and raise the CSP_FLAG_MODIFIED flag.
2034 * 1 : csp = Current client state (buffers, headers, etc...)
2036 * Returns : a pointer to the (newly allocated) modified buffer.
2037 * or NULL in case something went wrong.
2039 *********************************************************************/
2040 char *gif_deanimate_response(struct client_state *csp)
2042 struct binbuffer *in, *out;
2046 size = (size_t)(csp->iob->eod - csp->iob->cur);
2049 * If the body has a "chunked" transfer-encoding,
2050 * get rid of it first, adjusting size and iob->eod
2052 if (csp->flags & CSP_FLAG_CHUNKED)
2054 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2055 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2059 csp->iob->eod = csp->iob->cur + size;
2060 csp->flags |= CSP_FLAG_MODIFIED;
2063 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2064 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2066 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2070 in->buffer = csp->iob->cur;
2073 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2075 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2082 if ((int)size == out->offset)
2084 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2088 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2090 csp->content_length = out->offset;
2091 csp->flags |= CSP_FLAG_MODIFIED;
2101 /*********************************************************************
2103 * Function : jpeg_inspect_response
2108 * 1 : csp = Current client state (buffers, headers, etc...)
2110 * Returns : a pointer to the (newly allocated) modified buffer
2111 * or NULL in case something went wrong.
2113 *********************************************************************/
2114 char *jpeg_inspect_response(struct client_state *csp)
2116 struct binbuffer *in = NULL;
2117 struct binbuffer *out = NULL;
2121 size = (size_t)(csp->iob->eod - csp->iob->cur);
2124 * If the body has a "chunked" transfer-encoding,
2125 * get rid of it first, adjusting size and iob->eod
2127 if (csp->flags & CSP_FLAG_CHUNKED)
2129 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2130 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2134 csp->iob->eod = csp->iob->cur + size;
2135 csp->flags |= CSP_FLAG_MODIFIED;
2138 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2140 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2144 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2146 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2150 in->buffer = csp->iob->cur;
2154 * Calling jpeg_inspect has the side-effect of creating and
2155 * modifying the image buffer of "out" directly.
2157 if (jpeg_inspect(in, out))
2159 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2167 csp->content_length = out->offset;
2168 csp->flags |= CSP_FLAG_MODIFIED;
2178 /*********************************************************************
2180 * Function : remove_chunked_transfer_coding
2182 * Description : In-situ remove the "chunked" transfer coding as defined
2183 * in rfc2616 from a buffer.
2186 * 1 : buffer = Pointer to the text buffer
2187 * 2 : size = Number of bytes to be processed
2189 * Returns : The new size, i.e. the number of bytes from buffer which
2190 * are occupied by the stripped body, or 0 in case something
2193 *********************************************************************/
2194 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2197 unsigned int chunksize = 0;
2198 char *from_p, *to_p;
2201 from_p = to_p = buffer;
2203 if (sscanf(buffer, "%x", &chunksize) != 1)
2205 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2209 while (chunksize > 0)
2211 if (NULL == (from_p = strstr(from_p, "\r\n")))
2213 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2217 if ((newsize += chunksize) >= size)
2219 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2224 memmove(to_p, from_p, (size_t) chunksize);
2225 to_p = buffer + newsize;
2226 from_p += chunksize + 2;
2228 if (sscanf(from_p, "%x", &chunksize) != 1)
2230 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2235 /* FIXME: Should this get its own loglevel? */
2236 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2242 /*********************************************************************
2244 * Function : url_actions
2246 * Description : Gets the actions for this URL.
2249 * 1 : http = http_request request for blocked URLs
2250 * 2 : csp = Current client state (buffers, headers, etc...)
2254 *********************************************************************/
2255 void url_actions(struct http_request *http,
2256 struct client_state *csp)
2258 struct file_list *fl;
2259 struct url_actions *b;
2262 init_current_action(csp->action);
2264 for (i = 0; i < MAX_AF_FILES; i++)
2266 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2271 apply_url_actions(csp->action, http, b);
2278 /*********************************************************************
2280 * Function : apply_url_actions
2282 * Description : Applies a list of URL actions.
2285 * 1 : action = Destination.
2286 * 2 : http = Current URL
2287 * 3 : b = list of URL actions to apply
2291 *********************************************************************/
2292 void apply_url_actions(struct current_action_spec *action,
2293 struct http_request *http,
2294 struct url_actions *b)
2298 /* Should never happen */
2302 for (b = b->next; NULL != b; b = b->next)
2304 if (url_match(b->url, http))
2306 merge_current_action(action, b->action);
2312 /*********************************************************************
2314 * Function : forward_url
2316 * Description : Should we forward this to another proxy?
2319 * 1 : http = http_request request for current URL
2320 * 2 : csp = Current client state (buffers, headers, etc...)
2322 * Returns : Pointer to forwarding information.
2324 *********************************************************************/
2325 const struct forward_spec * forward_url(struct http_request *http,
2326 struct client_state *csp)
2328 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2329 struct forward_spec *fwd = csp->config->forward;
2338 if (url_match(fwd->url, http))
2349 /*********************************************************************
2351 * Function : direct_response
2353 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2354 * request and if so, return a HTTP 501 to the client.
2356 * FIXME: I have a stupid name and I should handle the
2357 * requests properly. Still, what we do here is rfc-
2358 * compliant, whereas ignoring or forwarding are not.
2361 * 1 : csp = Current client state (buffers, headers, etc...)
2363 * Returns : http_response if , NULL if nonmatch or handler fail
2365 *********************************************************************/
2366 struct http_response *direct_response(struct client_state *csp)
2368 struct http_response *rsp;
2369 struct list_entry *p;
2371 if ((0 == strcmpic(csp->http->gpc, "trace"))
2372 || (0 == strcmpic(csp->http->gpc, "options")))
2374 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2376 if (!strncmpic("Max-Forwards:", p->str, 13))
2378 unsigned int max_forwards;
2381 * If it's a Max-Forwards value of zero,
2382 * we have to intercept the request.
2384 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2387 * FIXME: We could handle at least TRACE here,
2388 * but that would require a verbatim copy of
2389 * the request which we don't have anymore
2391 log_error(LOG_LEVEL_HEADER,
2392 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2395 /* Get mem for response or fail*/
2396 if (NULL == (rsp = alloc_http_response()))
2398 return cgi_error_memory();
2401 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2403 free_http_response(rsp);
2404 return cgi_error_memory();
2408 rsp->reason = RSP_REASON_UNSUPPORTED;
2410 return(finish_http_response(csp, rsp));