1 const char filters_rcs[] = "$Id: filters.c,v 1.84 2007/03/20 15:16:34 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 1.84 2007/03/20 15:16:34 fabiankeil
44 * Use dedicated header filter actions instead of abusing "filter".
45 * Replace "filter-client-headers" and "filter-client-headers"
46 * with "server-header-filter" and "client-header-filter".
48 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
49 * New config option: enforce-blocks.
51 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
52 * - Fix port handling in acl_addr() and use a temporary acl spec
53 * copy so error messages don't contain a truncated version.
54 * - Log size of iob before and after decompression.
56 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
57 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
58 * - Hide the "Go there anyway" link for blocked CONNECT
59 * requests where going there anyway doesn't work anyway.
61 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
62 * - Save the reason for generating http_responses.
63 * - Block (+block) with status code 403 instead of 404.
64 * - Use a different kludge to remember a failed decompression.
66 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
67 * Search for Max-Forwards headers case-insensitive,
68 * don't generate the "501 unsupported" message for invalid
69 * Max-Forwards values and don't increase negative ones.
71 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
72 * - Add HEAD support to finish_http_response.
73 * - Add error favicon to internal HTML error messages.
75 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
76 * Mark *csp as immutable for is_untrusted_url()
77 * and is_imageurl(). Closes FR 1237736.
79 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
80 * Integrate a modified version of Wil Mahan's
81 * zlib patch (PR #895531).
83 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
84 * Fixed gcc43 conversion warnings,
85 * changed sprintf calls to snprintf.
87 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
88 * Adjust comment in pcrs_filter_response()
89 * to recent pcrs changes. Hohoho.
91 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
92 * Don't crash if pcre returns an error code
93 * that pcrs didn't expect. Fixes BR 1621173.
95 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
96 * Modified is_untrusted_url to complain in case of
97 * write errors and to give a reason when adding new
98 * entries to the trustfile. Closes FR 1097611.
100 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
101 * Skip empty filter files in pcrs_filter_response,
102 * but don't ignore the ones that come afterwards.
103 * Fixes parts of BR 1619208.
105 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
106 * Added some sanity checks for get_last_url().
107 * Fixed possible segfault caused by my last commit.
109 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
110 * Let get_last_url() catch https URLs as well.
112 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
113 * Make sure get_last_url() behaves like advertised
114 * and fast-redirects{} can be combined with redirect{}.
116 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
117 * Implemented +redirect{s@foo@bar@} to generate
118 * a redirect based on a rewritten version of the
121 * Revision 1.66 2006/09/23 13:26:38 roro
122 * Replace TABs by spaces in source code.
124 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
125 * Fix +redirect{}. Didn't work with -fast-redirects.
127 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
128 * Block requests for untrusted URLs with status
129 * code 403 instead of 200.
131 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
132 * Don't free p which is still in use and will be later
133 * freed by free_map(). Don't claim the referrer is unknown
134 * when the client didn't set one.
136 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
137 * Feature request 595948: Re-Filter logging in single line
139 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
140 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
142 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
143 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
144 * with what was really the latest development (the v_3_0_branch branch)
146 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
147 * Multiple filter file support
149 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
150 * A few memory leaks plugged (mostly on error paths)
152 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
153 * Add the ability to check jpeg images for invalid
154 * lengths of comment blocks. Defensive strategy
155 * against the exploit:
156 * Microsoft Security Bulletin MS04-028
157 * Buffer Overrun in JPEG Processing (GDI+) Could
158 * Allow Code Execution (833987)
159 * Enabled with +inspect-jpegs in actions files.
161 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
162 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
164 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
165 * Fixed bug #839859: "See why" link URL now gets url-encoded.
167 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
170 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
171 * Added basic support for OPTIONS and TRACE HTTP methods:
172 * New function direct_response which handles OPTIONS and
173 * TRACE requests whose Max-Forwards header field is zero.
175 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
176 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
178 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
179 * - Bugfix: Executing a filters without jobs no longer results in
180 * turing off *all* filters.
181 * - Security fix: Malicious web servers can't cause a seg fault
182 * through bogus chunk sizes anymore
184 * Revision 1.58 2002/04/24 02:11:17 oes
185 * Jon's multiple AF patch: url_actions now evaluates rules
188 * Revision 1.57 2002/04/08 20:38:34 swa
191 * Revision 1.56 2002/04/05 15:51:24 oes
192 * - bugfix: error-pages now get correct request protocol
193 * - fix for invalid HTML in trust info
195 * Revision 1.55 2002/04/02 16:13:51 oes
196 * Fix: No "Go there anyway" for SSL
198 * Revision 1.54 2002/04/02 14:55:56 oes
199 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
201 * Revision 1.53 2002/03/26 22:29:54 swa
202 * we have a new homepage!
204 * Revision 1.52 2002/03/24 16:35:57 jongfoster
207 * Revision 1.51 2002/03/24 15:23:33 jongfoster
210 * Revision 1.50 2002/03/24 13:25:43 swa
211 * name change related issues
213 * Revision 1.49 2002/03/16 20:29:14 oes
216 * Revision 1.48 2002/03/13 20:25:34 oes
217 * Better logging for content filters
219 * Revision 1.47 2002/03/13 00:30:52 jongfoster
221 * Added option of always sending redirect for imageblock,
222 * currently disabled with #if 0.
224 * Revision 1.46 2002/03/12 01:42:49 oes
225 * Introduced modular filters
227 * Revision 1.45 2002/03/08 16:47:50 oes
228 * Added choice beween GIF and PNG built-in images
230 * Revision 1.44 2002/03/07 03:49:31 oes
231 * - Fixed compiler warnings etc
232 * - Changed built-in images from GIF to PNG
233 * (with regard to Unisys patent issue)
234 * - Added a 4x4 pattern PNG which is less intrusive
235 * than the logo but also clearly marks the deleted banners
237 * Revision 1.43 2002/01/22 23:51:59 jongfoster
238 * Replacing strsav() with the safer string_append().
240 * Adding missing html_encode() to error message generators. Where encoded
241 * and unencoded versions of a string were provided, removing the unencoded
244 * Revision 1.42 2002/01/17 21:00:32 jongfoster
245 * Moving all our URL and URL pattern parsing code to urlmatch.c.
247 * Using a single, simple url_match(pattern,url) function - rather than
248 * the 3-line match routine which was repeated all over the place.
250 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
252 * Using parse_http_url() to parse URLs without faking a HTTP
253 * request line for parse_http_request().
255 * Revision 1.41 2001/11/13 00:14:07 jongfoster
256 * Fixing stupid bug now I've figured out what || means.
257 * (It always returns 0 or 1, not one of it's paramaters.)
259 * Revision 1.40 2001/10/26 17:37:55 oes
260 * - Re-enabled Netscape 200/404 bug workaround in block_url():
261 * - Removed OS/2 special case
262 * - Made block_url() independant from sed() having been run
263 * - Made trust_url independant from sed() having been run
264 * - Made is_imageurl independant from sed() having been run.
265 * It now checks User-Agent: and Accept: by itself.
268 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
269 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
270 * threads to call select() simultaneously. So, it's time to do a real, live,
271 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
272 * (native). Both versions will work, but using __OS2__ offers multi-threading.
274 * Revision 1.38 2001/10/23 21:32:33 jongfoster
275 * Adding error-checking to selected functions
277 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
278 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
279 * filters.c. Added a FIXME in front of the offending code. I'll gladly
280 * put in a better/more robust fix for all parties if one is presented...
281 * It seems that just returning 200 instead of 404 would pretty much fix
282 * it for everyone, but I don't know all the history of the problem.
284 * Revision 1.36 2001/10/10 16:44:16 oes
285 * Added match_portlist function
287 * Revision 1.35 2001/10/07 15:41:23 oes
288 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
290 * New function remove_chunked_transfer_coding that strips chunked
291 * transfer coding to plain and is called by pcrs_filter_response
292 * and gif_deanimate_response if neccessary
294 * Improved handling of zero-change re_filter runs
296 * pcrs_filter_response and gif_deanimate_response now remove
297 * chunked transfer codeing before processing the body.
299 * Revision 1.34 2001/09/20 15:49:36 steudten
301 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
302 * See cgi.c fill_template().
304 * Revision 1.33 2001/09/16 17:05:14 jongfoster
305 * Removing unused #include showarg.h
307 * Revision 1.32 2001/09/16 13:21:27 jongfoster
308 * Changes to use new list functions.
310 * Revision 1.31 2001/09/16 11:38:02 jongfoster
311 * Splitting fill_template() into 2 functions:
312 * template_load() loads the file
313 * template_fill() performs the PCRS regexps.
314 * This is because the CGI edit interface has a "table row"
315 * template which is used many times in the page - this
316 * change means it's only loaded from disk once.
318 * Revision 1.30 2001/09/16 11:00:10 jongfoster
319 * New function alloc_http_response, for symmetry with free_http_response
321 * Revision 1.29 2001/09/13 23:32:40 jongfoster
322 * Moving image data to cgi.c rather than cgi.h
323 * Fixing a GPF under Win32 (and any other OS that protects global
324 * constants from being written to).
326 * Revision 1.28 2001/09/10 10:18:51 oes
327 * Silenced compiler warnings
329 * Revision 1.27 2001/08/05 16:06:20 jongfoster
330 * Modifiying "struct map" so that there are now separate header and
331 * "map_entry" structures. This means that functions which modify a
332 * map no longer need to return a pointer to the modified map.
333 * Also, it no longer reverses the order of the entries (which may be
334 * important with some advanced template substitutions).
336 * Revision 1.26 2001/07/30 22:08:36 jongfoster
337 * Tidying up #defines:
338 * - All feature #defines are now of the form FEATURE_xxx
339 * - Permanently turned off WIN_GUI_EDIT
340 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
342 * Revision 1.25 2001/07/26 10:09:46 oes
343 * Made browser detection a little less naive
345 * Revision 1.24 2001/07/25 17:22:51 oes
346 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
348 * Revision 1.23 2001/07/23 13:40:12 oes
349 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
351 * Revision 1.22 2001/07/18 12:29:34 oes
352 * - Made gif_deanimate_response respect
353 * csp->action->string[ACTION_STRING_DEANIMATE]
354 * - Logging cosmetics
356 * Revision 1.21 2001/07/13 13:59:53 oes
357 * - Introduced gif_deanimate_response which shares the
358 * generic content modification interface of pcrs_filter_response
359 * and acts as a wrapper to deanimate.c:gif_deanimate()
360 * - Renamed re_process_buffer to pcrs_filter_response
361 * - pcrs_filter_response now returns NULL on failiure
362 * - Removed all #ifdef PCRS
364 * Revision 1.20 2001/07/01 17:01:04 oes
365 * Added comments and missing return statement in is_untrusted_url()
367 * Revision 1.19 2001/06/29 21:45:41 oes
368 * Indentation, CRLF->LF, Tab-> Space
370 * Revision 1.18 2001/06/29 13:27:38 oes
371 * - Cleaned up, renamed and reorderd functions
372 * and improved comments
375 * - Ported to CGI platform. Now delivers
376 * http_response or NULL
377 * - Unified HTML and GIF generation (moved image detection
378 * and GIF generation here from jcc.c:chat())
379 * - Fixed HTTP status to:
380 * - 403 (Forbidden) for the "blocked" HTML message
381 * - 200 (OK) for GIF answers
382 * - 302 (Redirect) for redirect to GIF
385 * - Ported to CGI platform. Now delivers
386 * http_response or NULL
387 * - Separated detection of untrusted URL into
388 * (bool)is_untrusted_url
389 * - Added enforcement of untrusted requests
391 * - Moved redirect_url() from cgi.c to here
392 * and ported it to the CGI platform
394 * - Removed logentry from cancelled commit
396 * Revision 1.17 2001/06/09 10:55:28 jongfoster
397 * Changing BUFSIZ ==> BUFFER_SIZE
399 * Revision 1.16 2001/06/07 23:10:26 jongfoster
400 * Allowing unanchored domain patterns to back off and retry
401 * if they partially match. Optimized right-anchored patterns.
402 * Moving ACL and forward files into config file.
403 * Replacing struct gateway with struct forward_spec
405 * Revision 1.15 2001/06/03 19:12:00 oes
406 * extracted-CGI relevant stuff
408 * Revision 1.14 2001/06/01 10:30:55 oes
409 * Added optional left-anchoring to domaincmp
411 * Revision 1.13 2001/05/31 21:21:30 jongfoster
412 * Permissionsfile / actions file changes:
413 * - Changed "permission" to "action" throughout
414 * - changes to file format to allow string parameters
415 * - Moved helper functions to actions.c
417 * Revision 1.12 2001/05/31 17:35:20 oes
419 * - Enhanced domain part globbing with infix and prefix asterisk
420 * matching and optional unanchored operation
422 * Revision 1.11 2001/05/29 11:53:23 oes
423 * "See why" link added to "blocked" page
425 * Revision 1.10 2001/05/29 09:50:24 jongfoster
426 * Unified blocklist/imagelist/permissionslist.
427 * File format is still under discussion, but the internal changes
430 * Also modified interceptor behaviour:
431 * - We now intercept all URLs beginning with one of the following
432 * prefixes (and *only* these prefixes):
434 * * http://ijbswa.sf.net/config/
435 * * http://ijbswa.sourceforge.net/config/
436 * - New interceptors "home page" - go to http://i.j.b/ to see it.
437 * - Internal changes so that intercepted and fast redirect pages
438 * are not replaced with an image.
439 * - Interceptors now have the option to send a binary page direct
440 * to the client. (i.e. ijb-send-banner uses this)
441 * - Implemented show-url-info interceptor. (Which is why I needed
442 * the above interceptors changes - a typical URL is
443 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
444 * The previous mechanism would not have intercepted that, and
445 * if it had been intercepted then it then it would have replaced
448 * Revision 1.9 2001/05/27 22:17:04 oes
450 * - re_process_buffer no longer writes the modified buffer
451 * to the client, which was very ugly. It now returns the
452 * buffer, which it is then written by chat.
454 * - content_length now adjusts the Content-Length: header
455 * for modified documents rather than crunch()ing it.
456 * (Length info in csp->content_length, which is 0 for
457 * unmodified documents)
459 * - For this to work, sed() is called twice when filtering.
461 * Revision 1.8 2001/05/26 17:13:28 jongfoster
462 * Filled in a function comment.
464 * Revision 1.7 2001/05/26 15:26:15 jongfoster
465 * ACL feature now provides more security by immediately dropping
466 * connections from untrusted hosts.
468 * Revision 1.6 2001/05/26 00:28:36 jongfoster
469 * Automatic reloading of config file.
470 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
471 * Most of the global variables have been moved to a new
472 * struct configuration_spec, accessed through csp->config->globalname
473 * Most of the globals remaining are used by the Win32 GUI.
475 * Revision 1.5 2001/05/25 22:34:30 jongfoster
478 * Revision 1.4 2001/05/22 18:46:04 oes
480 * - Enabled filtering banners by size rather than URL
481 * by adding patterns that replace all standard banner
482 * sizes with the "Junkbuster" gif to the re_filterfile
484 * - Enabled filtering WebBugs by providing a pattern
485 * which kills all 1x1 images
487 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
488 * which is selected by the (nonstandard and therefore
489 * capital) letter 'U' in the option string.
490 * It causes the quantifiers to be ungreedy by default.
491 * Appending a ? turns back to greedy (!).
493 * - Added a new interceptor ijb-send-banner, which
494 * sends back the "Junkbuster" gif. Without imagelist or
495 * MSIE detection support, or if tinygif = 1, or the
496 * URL isn't recognized as an imageurl, a lame HTML
497 * explanation is sent instead.
499 * - Added new feature, which permits blocking remote
500 * script redirects and firing back a local redirect
502 * The feature is conditionally compiled, i.e. it
503 * can be disabled with --disable-fast-redirects,
504 * plus it must be activated by a "fast-redirects"
505 * line in the config file, has its own log level
506 * and of course wants to be displayed by show-proxy-args
507 * Note: Boy, all the #ifdefs in 1001 locations and
508 * all the fumbling with configure.in and acconfig.h
509 * were *way* more work than the feature itself :-(
511 * - Because a generic redirect template was needed for
512 * this, tinygif = 3 now uses the same.
514 * - Moved GIFs, and other static HTTP response templates
519 * - Removed some >400 CRs again (Jon, you really worked
522 * Revision 1.3 2001/05/20 16:44:47 jongfoster
523 * Removing last hardcoded Junkbusters.com URLs.
525 * Revision 1.2 2001/05/20 01:21:20 jongfoster
526 * Version 2.9.4 checkin.
527 * - Merged popupfile and cookiefile, and added control over PCRS
528 * filtering, in new "permissionsfile".
529 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
530 * file error you now get a message box (in the Win32 GUI) rather
531 * than the program exiting with no explanation.
532 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
534 * - Removed tabs from "config"
535 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
536 * - Bumped up version number.
538 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
539 * Initial import of version 2.9.3 source tree
542 *********************************************************************/
548 #include <sys/types.h>
557 #endif /* ndef __OS2__ */
558 #include <netinet/in.h>
560 #include <winsock2.h>
561 #endif /* ndef _WIN32 */
565 #endif /* def __OS2__ */
573 #include "jbsockets.h"
574 #include "miscutil.h"
578 #include "deanimate.h"
579 #include "urlmatch.h"
585 const char filters_h_rcs[] = FILTERS_H_VERSION;
587 /* Fix a problem with Solaris. There should be no effect on other
589 * Solaris's isspace() is a macro which uses it's argument directly
590 * as an array index. Therefore we need to make sure that high-bit
591 * characters generate +ve values, and ideally we also want to make
592 * the argument match the declared parameter type of "int".
594 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
598 /*********************************************************************
600 * Function : block_acl
602 * Description : Block this request?
603 * Decide yes or no based on ACL file.
606 * 1 : dst = The proxy or gateway address this is going to.
607 * Or NULL to check all possible targets.
608 * 2 : csp = Current client state (buffers, headers, etc...)
609 * Also includes the client IP address.
611 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
613 *********************************************************************/
614 int block_acl(struct access_control_addr *dst, struct client_state *csp)
616 struct access_control_list *acl = csp->config->acl;
618 /* if not using an access control list, then permit the connection */
624 /* search the list */
627 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
631 /* Just want to check if they have any access */
632 if (acl->action == ACL_PERMIT)
637 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
638 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
640 if (acl->action == ACL_PERMIT)
658 /*********************************************************************
660 * Function : acl_addr
662 * Description : Called from `load_config' to parse an ACL address.
665 * 1 : aspec = String specifying ACL address.
666 * 2 : aca = struct access_control_addr to fill in.
668 * Returns : 0 => Ok, everything else is an error.
670 *********************************************************************/
671 int acl_addr(const char *aspec, struct access_control_addr *aca)
676 char *acl_spec = NULL;
682 * Use a temporary acl spec copy so we can log
683 * the unmodified original in case of parse errors.
685 acl_spec = strdup(aspec);
686 if (acl_spec == NULL)
688 /* XXX: This will be logged as parse error. */
692 if ((p = strchr(acl_spec, '/')) != NULL)
695 if (ijb_isdigit(*p) == 0)
700 masklength = atoi(p);
703 if ((masklength < 0) || (masklength > 32))
709 if ((p = strchr(acl_spec, ':')) != NULL)
714 port = strtol(p, &endptr, 10);
716 if (port <= 0 || port > 65535 || *endptr != '\0')
723 aca->port = (unsigned long)port;
725 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
728 if (aca->addr == INADDR_NONE)
730 /* XXX: This will be logged as parse error. */
734 /* build the netmask */
736 for (i=1; i <= masklength ; i++)
738 aca->mask |= (1 << (32 - i));
741 /* now mask off the host portion of the ip address
742 * (i.e. save on the network portion of the address).
744 aca->addr = aca->addr & aca->mask;
749 #endif /* def FEATURE_ACL */
752 /*********************************************************************
754 * Function : match_portlist
756 * Description : Check if a given number is covered by a comma
757 * separated list of numbers and ranges (a,b-c,d,..)
760 * 1 : portlist = String with list
761 * 2 : port = port to check
763 * Returns : 0 => no match
766 *********************************************************************/
767 int match_portlist(const char *portlist, int port)
769 char *min, *max, *next, *portlist_copy;
771 min = next = portlist_copy = strdup(portlist);
774 * Zero-terminate first item and remember offset for next
776 if (NULL != (next = strchr(portlist_copy, (int) ',')))
782 * Loop through all items, checking for match
786 if (NULL == (max = strchr(min, (int) '-')))
789 * No dash, check for equality
791 if (port == atoi(min))
800 * This is a range, so check if between min and max,
801 * or, if max was omitted, between min and 65K
804 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
818 * Zero-terminate next item and remember offset for n+1
820 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
832 /*********************************************************************
834 * Function : block_url
836 * Description : Called from `chat'. Check to see if we need to block this.
839 * 1 : csp = Current client state (buffers, headers, etc...)
841 * Returns : NULL => unblocked, else HTTP block response
843 *********************************************************************/
844 struct http_response *block_url(struct client_state *csp)
846 struct http_response *rsp;
847 const char *new_content_type = NULL;
850 * If it's not blocked, don't block it ;-)
852 if ((csp->action->flags & ACTION_BLOCK) == 0)
856 if (csp->action->flags & ACTION_REDIRECT)
858 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
861 * Else, prepare a response
863 if (NULL == (rsp = alloc_http_response()))
865 return cgi_error_memory();
869 * If it's an image-url, send back an image or redirect
870 * as specified by the relevant +image action
872 #ifdef FEATURE_IMAGE_BLOCKING
873 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
877 /* determine HOW images should be blocked */
878 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
880 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
882 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
884 #if 1 /* Two alternative strategies, use this one for now: */
886 /* and handle accordingly: */
887 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
889 rsp->status = strdup("403 Request blocked by Privoxy");
890 if (rsp->status == NULL)
892 free_http_response(rsp);
893 return cgi_error_memory();
895 rsp->body = bindup(image_pattern_data, image_pattern_length);
896 if (rsp->body == NULL)
898 free_http_response(rsp);
899 return cgi_error_memory();
901 rsp->content_length = image_pattern_length;
903 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
905 free_http_response(rsp);
906 return cgi_error_memory();
910 else if (0 == strcmpic(p, "blank"))
912 rsp->status = strdup("403 Request blocked by Privoxy");
913 if (rsp->status == NULL)
915 free_http_response(rsp);
916 return cgi_error_memory();
918 rsp->body = bindup(image_blank_data, image_blank_length);
919 if (rsp->body == NULL)
921 free_http_response(rsp);
922 return cgi_error_memory();
924 rsp->content_length = image_blank_length;
926 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
928 free_http_response(rsp);
929 return cgi_error_memory();
935 rsp->status = strdup("302 Local Redirect from Privoxy");
936 if (rsp->status == NULL)
938 free_http_response(rsp);
939 return cgi_error_memory();
942 if (enlist_unique_header(rsp->headers, "Location", p))
944 free_http_response(rsp);
945 return cgi_error_memory();
949 #else /* Following code is disabled for now */
951 /* and handle accordingly: */
952 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
954 p = CGI_PREFIX "send-banner?type=pattern";
956 else if (0 == strcmpic(p, "blank"))
958 p = CGI_PREFIX "send-banner?type=blank";
960 rsp->status = strdup("302 Local Redirect from Privoxy");
961 if (rsp->status == NULL)
963 free_http_response(rsp);
964 return cgi_error_memory();
967 if (enlist_unique_header(rsp->headers, "Location", p))
969 free_http_response(rsp);
970 return cgi_error_memory();
972 #endif /* Preceeding code is disabled for now */
974 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
977 * Send empty document.
979 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
982 rsp->body = strdup(" ");
983 rsp->content_length = 1;
985 rsp->status = strdup("403 Request blocked by Privoxy");
986 if (rsp->status == NULL)
988 free_http_response(rsp);
989 return cgi_error_memory();
991 if (new_content_type != 0)
993 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
994 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
996 free_http_response(rsp);
997 return cgi_error_memory();
1002 #endif /* def FEATURE_IMAGE_BLOCKING */
1005 * Else, generate an HTML "blocked" message:
1009 struct map * exports;
1013 * Workaround for stupid Netscape bug which prevents
1014 * pages from being displayed if loading a referenced
1015 * JavaScript or style sheet fails. So make it appear
1016 * as if it succeeded.
1018 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1019 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1020 && !strstr(p, "Gecko") /* save Mozilla, */
1021 && !strstr(p, "compatible") /* MSIE */
1022 && !strstr(p, "Opera")) /* and Opera. */
1024 rsp->status = strdup("200 Request for blocked URL");
1028 rsp->status = strdup("403 Request for blocked URL");
1031 if (rsp->status == NULL)
1033 free_http_response(rsp);
1034 return cgi_error_memory();
1037 exports = default_exports(csp, NULL);
1038 if (exports == NULL)
1040 free_http_response(rsp);
1041 return cgi_error_memory();
1044 #ifdef FEATURE_FORCE_LOAD
1045 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1047 * Export the force conditional block killer if
1049 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1050 * - Privoxy is configured to enforce blocks, or
1051 * - it's a CONNECT request and enforcing wouldn't work anyway.
1053 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1054 || (0 == strcmpic(csp->http->gpc, "connect")))
1055 #endif /* ndef FEATURE_FORCE_LOAD */
1057 err = map_block_killer(exports, "force-support");
1060 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1061 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1062 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1063 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1068 free_http_response(rsp);
1069 return cgi_error_memory();
1072 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1075 free_http_response(rsp);
1076 return cgi_error_memory();
1079 rsp->reason = RSP_REASON_BLOCKED;
1081 return finish_http_response(csp, rsp);
1086 #ifdef FEATURE_TRUST
1087 /*********************************************************************
1089 * Function : trust_url FIXME: I should be called distrust_url
1091 * Description : Calls is_untrusted_url to determine if the URL is trusted
1092 * and if not, returns a HTTP 403 response with a reject message.
1095 * 1 : csp = Current client state (buffers, headers, etc...)
1097 * Returns : NULL => trusted, else http_response.
1099 *********************************************************************/
1100 struct http_response *trust_url(struct client_state *csp)
1102 struct http_response *rsp;
1103 struct map * exports;
1104 char buf[BUFFER_SIZE];
1106 struct url_spec **tl;
1111 * Don't bother to work on trusted URLs
1113 if (!is_untrusted_url(csp))
1119 * Else, prepare a response:
1121 if (NULL == (rsp = alloc_http_response()))
1123 return cgi_error_memory();
1126 rsp->status = strdup("403 Request blocked by Privoxy");
1127 exports = default_exports(csp, NULL);
1128 if (exports == NULL || rsp->status == NULL)
1130 free_http_response(rsp);
1131 return cgi_error_memory();
1135 * Export the protocol, host, port, and referrer information
1137 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1138 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1139 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1141 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1143 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1147 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1153 free_http_response(rsp);
1154 return cgi_error_memory();
1158 * Export the trust list
1161 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1163 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1164 string_append(&p, buf);
1166 err = map(exports, "trusted-referrers", 1, p, 0);
1171 free_http_response(rsp);
1172 return cgi_error_memory();
1176 * Export the trust info, if available
1178 if (csp->config->trust_info->first)
1180 struct list_entry *l;
1183 for (l = csp->config->trust_info->first; l ; l = l->next)
1185 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1186 string_append(&p, buf);
1188 err = map(exports, "trust-info", 1, p, 0);
1192 err = map_block_killer(exports, "have-trust-info");
1198 free_http_response(rsp);
1199 return cgi_error_memory();
1203 * Export the force conditional block killer if
1205 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1206 * - Privoxy is configured to enforce blocks, or
1207 * - it's a CONNECT request and enforcing wouldn't work anyway.
1209 #ifdef FEATURE_FORCE_LOAD
1210 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1211 || (0 == strcmpic(csp->http->gpc, "connect")))
1213 err = map_block_killer(exports, "force-support");
1217 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1219 #else /* ifndef FEATURE_FORCE_LOAD */
1220 err = map_block_killer(exports, "force-support");
1221 #endif /* ndef FEATURE_FORCE_LOAD */
1226 free_http_response(rsp);
1227 return cgi_error_memory();
1231 * Build the response
1233 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1236 free_http_response(rsp);
1237 return cgi_error_memory();
1239 rsp->reason = RSP_REASON_UNTRUSTED;
1241 return finish_http_response(csp, rsp);
1243 #endif /* def FEATURE_TRUST */
1245 /*********************************************************************
1247 * Function : execute_single_pcrs_command
1249 * Description : Apply single pcrs command to the subject.
1250 * The subject itself is left untouched, memory for the result
1251 * is malloc()ed and it is the caller's responsibility to free
1252 * the result when it's no longer needed.
1255 * 1 : subject = the subject (== original) string
1256 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1257 * 3 : hits = int* for returning the number of modifications
1259 * Returns : NULL in case of errors, otherwise the
1260 * result of the pcrs command.
1262 *********************************************************************/
1263 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1267 char *result = NULL;
1271 assert(pcrs_command);
1274 size = strlen(subject);
1276 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1278 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1279 pcrs_command, error);
1281 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1283 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1290 job = pcrs_free_job(job);
1297 /*********************************************************************
1299 * Function : rewrite_url
1301 * Description : Rewrites a URL with a single pcrs command
1302 * and returns the result if it differs from the
1303 * original and isn't obviously invalid.
1306 * 1 : old_url = URL to rewrite.
1307 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1310 * Returns : NULL if the pcrs_command didn't change the url, or
1311 * the result of the modification.
1313 *********************************************************************/
1314 char *rewrite_url(char *old_url, const char *pcrs_command)
1316 char *new_url = NULL;
1320 assert(pcrs_command);
1322 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1326 log_error(LOG_LEVEL_REDIRECTS,
1327 "pcrs command \"%s\" didn't change \"%s\".",
1328 pcrs_command, old_url, new_url);
1331 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1333 log_error(LOG_LEVEL_ERROR,
1334 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1335 "but the result doesn't look like a valid URL and will be ignored.",
1336 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1341 log_error(LOG_LEVEL_REDIRECTS,
1342 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1343 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1351 #ifdef FEATURE_FAST_REDIRECTS
1352 /*********************************************************************
1354 * Function : get_last_url
1356 * Description : Search for the last URL inside a string.
1357 * If the string already is a URL, it will
1358 * be the first URL found.
1361 * 1 : subject = the string to check
1362 * 2 : redirect_mode = +fast-redirect{} mode
1364 * Returns : NULL if no URL was found, or
1365 * the last URL found.
1367 *********************************************************************/
1368 char *get_last_url(char *subject, const char *redirect_mode)
1370 char *new_url = NULL;
1374 assert(redirect_mode);
1376 subject = strdup(subject);
1377 if (subject == NULL)
1379 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1383 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1385 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1386 new_url = url_decode(subject);
1387 if (new_url != NULL)
1394 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1398 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1401 * Find the last URL encoded in the request
1404 while ((tmp = strstr(tmp, "http://")) != NULL)
1408 tmp = (new_url != NULL) ? new_url : subject;
1409 while ((tmp = strstr(tmp, "https://")) != NULL)
1414 if ((new_url != NULL)
1415 && ( (new_url != subject)
1416 || (0 == strncmpic(subject, "http://", 7))
1417 || (0 == strncmpic(subject, "https://", 8))
1421 * Return new URL if we found a redirect
1422 * or if the subject already was a URL.
1424 * The second case makes sure that we can
1425 * chain get_last_url after another redirection check
1426 * (like rewrite_url) without losing earlier redirects.
1428 new_url = strdup(new_url);
1437 #endif /* def FEATURE_FAST_REDIRECTS */
1440 /*********************************************************************
1442 * Function : redirect_url
1444 * Description : Checks if Privoxy should answer the request with
1445 * a HTTP redirect and generates the redirect if
1449 * 1 : csp = Current client state (buffers, headers, etc...)
1451 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1453 *********************************************************************/
1454 struct http_response *redirect_url(struct client_state *csp)
1456 struct http_response *rsp;
1457 #ifdef FEATURE_FAST_REDIRECTS
1459 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1460 * as compile-time option? The user can easily disable
1461 * it in his action file.
1463 char * redirect_mode;
1464 #endif /* def FEATURE_FAST_REDIRECTS */
1465 char *old_url = NULL;
1466 char *new_url = NULL;
1467 char *redirection_string;
1469 if ((csp->action->flags & ACTION_REDIRECT))
1471 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1474 * If the redirection string begins with 's',
1475 * assume it's a pcrs command, otherwise treat it as
1476 * properly formatted URL and use it for the redirection
1479 * According to RFC 2616 section 14.30 the URL
1480 * has to be absolute and if the user tries:
1481 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1482 * she would get undefined results anyway.
1486 if (*redirection_string == 's')
1488 old_url = csp->http->url;
1489 new_url = rewrite_url(old_url, redirection_string);
1493 log_error(LOG_LEVEL_REDIRECTS,
1494 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1495 redirection_string);
1496 new_url = strdup(redirection_string);
1500 #ifdef FEATURE_FAST_REDIRECTS
1501 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1503 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1506 * If it exists, use the previously rewritten URL as input
1507 * otherwise just use the old path.
1509 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1510 new_url = get_last_url(old_url, redirect_mode);
1513 #endif /* def FEATURE_FAST_REDIRECTS */
1515 /* Did any redirect action trigger? */
1518 if (0 == strcmpic(new_url, csp->http->url))
1520 log_error(LOG_LEVEL_ERROR,
1521 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1522 csp->http->url, new_url);
1527 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1529 if (NULL == (rsp = alloc_http_response()))
1532 return cgi_error_memory();
1535 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1536 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1539 free_http_response(rsp);
1540 return cgi_error_memory();
1542 rsp->reason = RSP_REASON_REDIRECTED;
1545 return finish_http_response(csp, rsp);
1549 /* Only reached if no redirect is required */
1555 #ifdef FEATURE_IMAGE_BLOCKING
1556 /*********************************************************************
1558 * Function : is_imageurl
1560 * Description : Given a URL, decide whether it is an image or not,
1561 * using either the info from a previous +image action
1562 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1563 * is MSIE and not on a Mac, tell from the browser's accept
1567 * 1 : csp = Current client state (buffers, headers, etc...)
1569 * Returns : True (nonzero) if URL is an image, false (0)
1572 *********************************************************************/
1573 int is_imageurl(const struct client_state *csp)
1575 #ifdef FEATURE_IMAGE_DETECT_MSIE
1578 tmp = get_header_value(csp->headers, "User-Agent:");
1579 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1581 tmp = get_header_value(csp->headers, "Accept:");
1582 if (tmp && strstr(tmp, "image/gif"))
1584 /* Client will accept HTML. If this seems counterintuitive,
1594 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1596 return ((csp->action->flags & ACTION_IMAGE) != 0);
1599 #endif /* def FEATURE_IMAGE_BLOCKING */
1602 #ifdef FEATURE_TRUST
1603 /*********************************************************************
1605 * Function : is_untrusted_url
1607 * Description : Should we "distrust" this URL (and block it)?
1609 * Yes if it matches a line in the trustfile, or if the
1610 * referrer matches a line starting with "+" in the
1615 * 1 : csp = Current client state (buffers, headers, etc...)
1617 * Returns : 0 => trusted, 1 => untrusted
1619 *********************************************************************/
1620 int is_untrusted_url(const struct client_state *csp)
1622 struct file_list *fl;
1623 struct block_spec *b;
1624 struct url_spec **trusted_url;
1625 struct http_request rhttp[1];
1626 const char * referer;
1630 * If we don't have a trustlist, we trust everybody
1632 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1637 memset(rhttp, '\0', sizeof(*rhttp));
1640 * Do we trust the request URL itself?
1642 for (b = b->next; b ; b = b->next)
1644 if (url_match(b->url, csp->http))
1650 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1652 /* no referrer was supplied */
1658 * If not, do we maybe trust its referrer?
1660 err = parse_http_url(referer, rhttp, csp);
1666 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1668 if (url_match(*trusted_url, rhttp))
1670 /* if the URL's referrer is from a trusted referrer, then
1671 * add the target spec to the trustfile as an unblocked
1672 * domain and return 0 (which means it's OK).
1677 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1681 char * new_entry = strdup("~");
1683 string_append(&new_entry, csp->http->hostport);
1685 path = csp->http->path;
1686 if ( (path[0] == '/')
1688 && ((path_end = strchr(path + 2, '/')) != NULL))
1690 /* since this path points into a user's home space
1691 * be sure to include this spec in the trustfile.
1693 int path_len = path_end - path; /* save offset */
1694 path = strdup(path); /* Copy string */
1697 path_end = path + path_len; /* regenerate ptr to new buffer */
1698 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1700 string_join(&new_entry, path);
1704 * Give a reason for generating this entry.
1706 string_append(&new_entry, " # Trusted referrer was: ");
1707 string_append(&new_entry, referer);
1709 if (new_entry != NULL)
1711 if (-1 == fprintf(fp, "%s\n", new_entry))
1713 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1714 new_entry, csp->config->trustfile);
1720 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1721 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1728 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1729 csp->http->hostport, csp->config->trustfile);
1737 #endif /* def FEATURE_TRUST */
1740 /*********************************************************************
1742 * Function : pcrs_filter_response
1744 * Description : Execute all text substitutions from all applying
1745 * +filter actions on the text buffer that's been accumulated
1746 * in csp->iob->buf. If this changes the contents, set
1747 * csp->content_length to the modified size and raise the
1748 * CSP_FLAG_MODIFIED flag.
1750 * XXX: Currently pcrs_filter_response is also responsible
1751 * for dechunking and decompressing. Both should be
1752 * done in separate functions so other content modifiers
1753 * profit as well, even if pcrs filtering is disabled.
1756 * 1 : csp = Current client state (buffers, headers, etc...)
1758 * Returns : a pointer to the (newly allocated) modified buffer.
1759 * or NULL if there were no hits or something went wrong
1761 *********************************************************************/
1762 char *pcrs_filter_response(struct client_state *csp)
1765 size_t size, prev_size;
1767 char *old = csp->iob->cur, *new = NULL;
1770 struct file_list *fl;
1771 struct re_filterfile_spec *b;
1772 struct list_entry *filtername;
1774 int i, found_filters = 0;
1779 if (csp->iob->cur >= csp->iob->eod)
1783 size = (size_t)(csp->iob->eod - csp->iob->cur);
1786 * Need to check the set of re_filterfiles...
1788 for (i = 0; i < MAX_AF_FILES; i++)
1801 if (0 == found_filters)
1803 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1808 * If the body has a "chunked" transfer-encoding,
1809 * get rid of it first, adjusting size and iob->eod
1811 if (csp->flags & CSP_FLAG_CHUNKED)
1813 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1814 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1818 csp->iob->eod = csp->iob->cur + size;
1819 csp->flags |= CSP_FLAG_MODIFIED;
1824 * If the body has a compressed transfer-encoding,
1825 * uncompress it first, adjusting size and iob->eod.
1826 * Note that decompression occurs after de-chunking.
1828 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1830 /* Notice that we at least tried to decompress. */
1831 if (JB_ERR_OK != decompress_iob(csp))
1834 * We failed to decompress the data; there's no point
1835 * in continuing since we can't filter.
1837 * XXX: Actually the Accept-Encoding header may
1838 * just be incorrect in which case we could continue
1841 * Unset CT_GZIP and CT_DEFLATE to remember not
1842 * to modify the Content-Encoding header later.
1844 csp->content_type &= ~CT_GZIP;
1845 csp->content_type &= ~CT_DEFLATE;
1850 * Decompression gives us a completely new iob,
1851 * so we need to update.
1853 size = (size_t)(csp->iob->eod - csp->iob->cur);
1854 old = csp->iob->cur;
1856 csp->flags |= CSP_FLAG_MODIFIED;
1857 csp->content_type &= ~CT_TABOO;
1861 for (i = 0; i < MAX_AF_FILES; i++)
1864 if ((NULL == fl) || (NULL == fl->f))
1867 * Either there are no filter files
1868 * left, or this filter file just
1869 * contains no valid filters.
1871 * Continue to be sure we don't miss
1872 * valid filter files that are chained
1873 * after empty or invalid ones.
1878 * For all applying +filter actions, look if a filter by that
1879 * name exists and if yes, execute it's pcrs_joblist on the
1882 for (b = fl->f; b; b = b->next)
1884 if (b->type != FT_CONTENT_FILTER)
1886 /* Skip header filters */
1890 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1891 filtername ; filtername = filtername->next)
1893 if (strcmp(b->name, filtername->str) == 0)
1895 int current_hits = 0; /* Number of hits caused by this filter */
1896 int job_number = 0; /* Which job we're currently executing */
1897 int job_hits = 0; /* How many hits the current job caused */
1899 if ( NULL == b->joblist )
1901 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1906 /* Apply all jobs from the joblist */
1907 for (job = b->joblist; NULL != job; job = job->next)
1910 job_hits = pcrs_execute(job, old, size, &new, &size);
1915 * That went well. Continue filtering
1916 * and use the result of this job as
1917 * input for the next one.
1919 current_hits += job_hits;
1920 if (old != csp->iob->cur)
1929 * This job caused an unexpected error. Inform the user
1930 * and skip the rest of the jobs in this filter. We could
1931 * continue with the next job, but usually the jobs
1932 * depend on each other or are similar enough to
1933 * fail for the same reason.
1935 * At the moment our pcrs expects the error codes of pcre 3.4,
1936 * but newer pcre versions can return additional error codes.
1937 * As a result pcrs_strerror()'s error message might be
1938 * "Unknown error ...", therefore we print the numerical value
1941 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1942 * should we use LOG_LEVEL_RE_FILTER instead?
1944 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1945 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1950 log_error(LOG_LEVEL_RE_FILTER,
1951 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1952 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1954 hits += current_hits;
1961 * If there were no hits, destroy our copy and let
1962 * chat() use the original in csp->iob
1970 csp->flags |= CSP_FLAG_MODIFIED;
1971 csp->content_length = size;
1979 /*********************************************************************
1981 * Function : gif_deanimate_response
1983 * Description : Deanimate the GIF image that has been accumulated in
1984 * csp->iob->buf, set csp->content_length to the modified
1985 * size and raise the CSP_FLAG_MODIFIED flag.
1988 * 1 : csp = Current client state (buffers, headers, etc...)
1990 * Returns : a pointer to the (newly allocated) modified buffer.
1991 * or NULL in case something went wrong.
1993 *********************************************************************/
1994 char *gif_deanimate_response(struct client_state *csp)
1996 struct binbuffer *in, *out;
2000 size = (size_t)(csp->iob->eod - csp->iob->cur);
2003 * If the body has a "chunked" transfer-encoding,
2004 * get rid of it first, adjusting size and iob->eod
2006 if (csp->flags & CSP_FLAG_CHUNKED)
2008 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2009 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2013 csp->iob->eod = csp->iob->cur + size;
2014 csp->flags |= CSP_FLAG_MODIFIED;
2017 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2018 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2020 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2024 in->buffer = csp->iob->cur;
2027 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2029 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2036 if ((int)size == out->offset)
2038 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2042 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2044 csp->content_length = out->offset;
2045 csp->flags |= CSP_FLAG_MODIFIED;
2055 /*********************************************************************
2057 * Function : jpeg_inspect_response
2062 * 1 : csp = Current client state (buffers, headers, etc...)
2064 * Returns : a pointer to the (newly allocated) modified buffer
2065 * or NULL in case something went wrong.
2067 *********************************************************************/
2068 char *jpeg_inspect_response(struct client_state *csp)
2070 struct binbuffer *in = NULL;
2071 struct binbuffer *out = NULL;
2075 size = (size_t)(csp->iob->eod - csp->iob->cur);
2078 * If the body has a "chunked" transfer-encoding,
2079 * get rid of it first, adjusting size and iob->eod
2081 if (csp->flags & CSP_FLAG_CHUNKED)
2083 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2084 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2088 csp->iob->eod = csp->iob->cur + size;
2089 csp->flags |= CSP_FLAG_MODIFIED;
2092 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2094 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2098 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2100 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2104 in->buffer = csp->iob->cur;
2108 * Calling jpeg_inspect has the side-effect of creating and
2109 * modifying the image buffer of "out" directly.
2111 if (jpeg_inspect(in, out))
2113 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2121 csp->content_length = out->offset;
2122 csp->flags |= CSP_FLAG_MODIFIED;
2132 /*********************************************************************
2134 * Function : remove_chunked_transfer_coding
2136 * Description : In-situ remove the "chunked" transfer coding as defined
2137 * in rfc2616 from a buffer.
2140 * 1 : buffer = Pointer to the text buffer
2141 * 2 : size = Number of bytes to be processed
2143 * Returns : The new size, i.e. the number of bytes from buffer which
2144 * are occupied by the stripped body, or 0 in case something
2147 *********************************************************************/
2148 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2151 unsigned int chunksize = 0;
2152 char *from_p, *to_p;
2155 from_p = to_p = buffer;
2157 if (sscanf(buffer, "%x", &chunksize) != 1)
2159 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2163 while (chunksize > 0)
2165 if (NULL == (from_p = strstr(from_p, "\r\n")))
2167 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2171 if ((newsize += chunksize) >= size)
2173 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2178 memmove(to_p, from_p, (size_t) chunksize);
2179 to_p = buffer + newsize;
2180 from_p += chunksize + 2;
2182 if (sscanf(from_p, "%x", &chunksize) != 1)
2184 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2189 /* FIXME: Should this get its own loglevel? */
2190 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2196 /*********************************************************************
2198 * Function : url_actions
2200 * Description : Gets the actions for this URL.
2203 * 1 : http = http_request request for blocked URLs
2204 * 2 : csp = Current client state (buffers, headers, etc...)
2208 *********************************************************************/
2209 void url_actions(struct http_request *http,
2210 struct client_state *csp)
2212 struct file_list *fl;
2213 struct url_actions *b;
2216 init_current_action(csp->action);
2218 for (i = 0; i < MAX_AF_FILES; i++)
2220 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2225 apply_url_actions(csp->action, http, b);
2232 /*********************************************************************
2234 * Function : apply_url_actions
2236 * Description : Applies a list of URL actions.
2239 * 1 : action = Destination.
2240 * 2 : http = Current URL
2241 * 3 : b = list of URL actions to apply
2245 *********************************************************************/
2246 void apply_url_actions(struct current_action_spec *action,
2247 struct http_request *http,
2248 struct url_actions *b)
2252 /* Should never happen */
2256 for (b = b->next; NULL != b; b = b->next)
2258 if (url_match(b->url, http))
2260 merge_current_action(action, b->action);
2266 /*********************************************************************
2268 * Function : forward_url
2270 * Description : Should we forward this to another proxy?
2273 * 1 : http = http_request request for current URL
2274 * 2 : csp = Current client state (buffers, headers, etc...)
2276 * Returns : Pointer to forwarding information.
2278 *********************************************************************/
2279 const struct forward_spec * forward_url(struct http_request *http,
2280 struct client_state *csp)
2282 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2283 struct forward_spec *fwd = csp->config->forward;
2292 if (url_match(fwd->url, http))
2303 /*********************************************************************
2305 * Function : direct_response
2307 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2308 * request and if so, return a HTTP 501 to the client.
2310 * FIXME: I have a stupid name and I should handle the
2311 * requests properly. Still, what we do here is rfc-
2312 * compliant, whereas ignoring or forwarding are not.
2315 * 1 : csp = Current client state (buffers, headers, etc...)
2317 * Returns : http_response if , NULL if nonmatch or handler fail
2319 *********************************************************************/
2320 struct http_response *direct_response(struct client_state *csp)
2322 struct http_response *rsp;
2323 struct list_entry *p;
2325 if ((0 == strcmpic(csp->http->gpc, "trace"))
2326 || (0 == strcmpic(csp->http->gpc, "options")))
2328 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2330 if (!strncmpic("Max-Forwards:", p->str, 13))
2332 unsigned int max_forwards;
2335 * If it's a Max-Forwards value of zero,
2336 * we have to intercept the request.
2338 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2341 * FIXME: We could handle at least TRACE here,
2342 * but that would require a verbatim copy of
2343 * the request which we don't have anymore
2345 log_error(LOG_LEVEL_HEADER,
2346 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2349 /* Get mem for response or fail*/
2350 if (NULL == (rsp = alloc_http_response()))
2352 return cgi_error_memory();
2355 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2357 free_http_response(rsp);
2358 return cgi_error_memory();
2362 rsp->reason = RSP_REASON_UNSUPPORTED;
2364 return(finish_http_response(csp, rsp));