1 const char filters_rcs[] = "$Id: filters.c,v 1.83 2007/03/17 15:20:05 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 1.83 2007/03/17 15:20:05 fabiankeil
44 * New config option: enforce-blocks.
46 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
47 * - Fix port handling in acl_addr() and use a temporary acl spec
48 * copy so error messages don't contain a truncated version.
49 * - Log size of iob before and after decompression.
51 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
52 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
53 * - Hide the "Go there anyway" link for blocked CONNECT
54 * requests where going there anyway doesn't work anyway.
56 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
57 * - Save the reason for generating http_responses.
58 * - Block (+block) with status code 403 instead of 404.
59 * - Use a different kludge to remember a failed decompression.
61 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
62 * Search for Max-Forwards headers case-insensitive,
63 * don't generate the "501 unsupported" message for invalid
64 * Max-Forwards values and don't increase negative ones.
66 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
67 * - Add HEAD support to finish_http_response.
68 * - Add error favicon to internal HTML error messages.
70 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
71 * Mark *csp as immutable for is_untrusted_url()
72 * and is_imageurl(). Closes FR 1237736.
74 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
75 * Integrate a modified version of Wil Mahan's
76 * zlib patch (PR #895531).
78 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
79 * Fixed gcc43 conversion warnings,
80 * changed sprintf calls to snprintf.
82 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
83 * Adjust comment in pcrs_filter_response()
84 * to recent pcrs changes. Hohoho.
86 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
87 * Don't crash if pcre returns an error code
88 * that pcrs didn't expect. Fixes BR 1621173.
90 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
91 * Modified is_untrusted_url to complain in case of
92 * write errors and to give a reason when adding new
93 * entries to the trustfile. Closes FR 1097611.
95 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
96 * Skip empty filter files in pcrs_filter_response,
97 * but don't ignore the ones that come afterwards.
98 * Fixes parts of BR 1619208.
100 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
101 * Added some sanity checks for get_last_url().
102 * Fixed possible segfault caused by my last commit.
104 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
105 * Let get_last_url() catch https URLs as well.
107 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
108 * Make sure get_last_url() behaves like advertised
109 * and fast-redirects{} can be combined with redirect{}.
111 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
112 * Implemented +redirect{s@foo@bar@} to generate
113 * a redirect based on a rewritten version of the
116 * Revision 1.66 2006/09/23 13:26:38 roro
117 * Replace TABs by spaces in source code.
119 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
120 * Fix +redirect{}. Didn't work with -fast-redirects.
122 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
123 * Block requests for untrusted URLs with status
124 * code 403 instead of 200.
126 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
127 * Don't free p which is still in use and will be later
128 * freed by free_map(). Don't claim the referrer is unknown
129 * when the client didn't set one.
131 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
132 * Feature request 595948: Re-Filter logging in single line
134 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
135 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
137 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
138 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
139 * with what was really the latest development (the v_3_0_branch branch)
141 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
142 * Multiple filter file support
144 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
145 * A few memory leaks plugged (mostly on error paths)
147 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
148 * Add the ability to check jpeg images for invalid
149 * lengths of comment blocks. Defensive strategy
150 * against the exploit:
151 * Microsoft Security Bulletin MS04-028
152 * Buffer Overrun in JPEG Processing (GDI+) Could
153 * Allow Code Execution (833987)
154 * Enabled with +inspect-jpegs in actions files.
156 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
157 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
159 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
160 * Fixed bug #839859: "See why" link URL now gets url-encoded.
162 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
165 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
166 * Added basic support for OPTIONS and TRACE HTTP methods:
167 * New function direct_response which handles OPTIONS and
168 * TRACE requests whose Max-Forwards header field is zero.
170 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
171 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
173 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
174 * - Bugfix: Executing a filters without jobs no longer results in
175 * turing off *all* filters.
176 * - Security fix: Malicious web servers can't cause a seg fault
177 * through bogus chunk sizes anymore
179 * Revision 1.58 2002/04/24 02:11:17 oes
180 * Jon's multiple AF patch: url_actions now evaluates rules
183 * Revision 1.57 2002/04/08 20:38:34 swa
186 * Revision 1.56 2002/04/05 15:51:24 oes
187 * - bugfix: error-pages now get correct request protocol
188 * - fix for invalid HTML in trust info
190 * Revision 1.55 2002/04/02 16:13:51 oes
191 * Fix: No "Go there anyway" for SSL
193 * Revision 1.54 2002/04/02 14:55:56 oes
194 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
196 * Revision 1.53 2002/03/26 22:29:54 swa
197 * we have a new homepage!
199 * Revision 1.52 2002/03/24 16:35:57 jongfoster
202 * Revision 1.51 2002/03/24 15:23:33 jongfoster
205 * Revision 1.50 2002/03/24 13:25:43 swa
206 * name change related issues
208 * Revision 1.49 2002/03/16 20:29:14 oes
211 * Revision 1.48 2002/03/13 20:25:34 oes
212 * Better logging for content filters
214 * Revision 1.47 2002/03/13 00:30:52 jongfoster
216 * Added option of always sending redirect for imageblock,
217 * currently disabled with #if 0.
219 * Revision 1.46 2002/03/12 01:42:49 oes
220 * Introduced modular filters
222 * Revision 1.45 2002/03/08 16:47:50 oes
223 * Added choice beween GIF and PNG built-in images
225 * Revision 1.44 2002/03/07 03:49:31 oes
226 * - Fixed compiler warnings etc
227 * - Changed built-in images from GIF to PNG
228 * (with regard to Unisys patent issue)
229 * - Added a 4x4 pattern PNG which is less intrusive
230 * than the logo but also clearly marks the deleted banners
232 * Revision 1.43 2002/01/22 23:51:59 jongfoster
233 * Replacing strsav() with the safer string_append().
235 * Adding missing html_encode() to error message generators. Where encoded
236 * and unencoded versions of a string were provided, removing the unencoded
239 * Revision 1.42 2002/01/17 21:00:32 jongfoster
240 * Moving all our URL and URL pattern parsing code to urlmatch.c.
242 * Using a single, simple url_match(pattern,url) function - rather than
243 * the 3-line match routine which was repeated all over the place.
245 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
247 * Using parse_http_url() to parse URLs without faking a HTTP
248 * request line for parse_http_request().
250 * Revision 1.41 2001/11/13 00:14:07 jongfoster
251 * Fixing stupid bug now I've figured out what || means.
252 * (It always returns 0 or 1, not one of it's paramaters.)
254 * Revision 1.40 2001/10/26 17:37:55 oes
255 * - Re-enabled Netscape 200/404 bug workaround in block_url():
256 * - Removed OS/2 special case
257 * - Made block_url() independant from sed() having been run
258 * - Made trust_url independant from sed() having been run
259 * - Made is_imageurl independant from sed() having been run.
260 * It now checks User-Agent: and Accept: by itself.
263 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
264 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
265 * threads to call select() simultaneously. So, it's time to do a real, live,
266 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
267 * (native). Both versions will work, but using __OS2__ offers multi-threading.
269 * Revision 1.38 2001/10/23 21:32:33 jongfoster
270 * Adding error-checking to selected functions
272 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
273 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
274 * filters.c. Added a FIXME in front of the offending code. I'll gladly
275 * put in a better/more robust fix for all parties if one is presented...
276 * It seems that just returning 200 instead of 404 would pretty much fix
277 * it for everyone, but I don't know all the history of the problem.
279 * Revision 1.36 2001/10/10 16:44:16 oes
280 * Added match_portlist function
282 * Revision 1.35 2001/10/07 15:41:23 oes
283 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
285 * New function remove_chunked_transfer_coding that strips chunked
286 * transfer coding to plain and is called by pcrs_filter_response
287 * and gif_deanimate_response if neccessary
289 * Improved handling of zero-change re_filter runs
291 * pcrs_filter_response and gif_deanimate_response now remove
292 * chunked transfer codeing before processing the body.
294 * Revision 1.34 2001/09/20 15:49:36 steudten
296 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
297 * See cgi.c fill_template().
299 * Revision 1.33 2001/09/16 17:05:14 jongfoster
300 * Removing unused #include showarg.h
302 * Revision 1.32 2001/09/16 13:21:27 jongfoster
303 * Changes to use new list functions.
305 * Revision 1.31 2001/09/16 11:38:02 jongfoster
306 * Splitting fill_template() into 2 functions:
307 * template_load() loads the file
308 * template_fill() performs the PCRS regexps.
309 * This is because the CGI edit interface has a "table row"
310 * template which is used many times in the page - this
311 * change means it's only loaded from disk once.
313 * Revision 1.30 2001/09/16 11:00:10 jongfoster
314 * New function alloc_http_response, for symmetry with free_http_response
316 * Revision 1.29 2001/09/13 23:32:40 jongfoster
317 * Moving image data to cgi.c rather than cgi.h
318 * Fixing a GPF under Win32 (and any other OS that protects global
319 * constants from being written to).
321 * Revision 1.28 2001/09/10 10:18:51 oes
322 * Silenced compiler warnings
324 * Revision 1.27 2001/08/05 16:06:20 jongfoster
325 * Modifiying "struct map" so that there are now separate header and
326 * "map_entry" structures. This means that functions which modify a
327 * map no longer need to return a pointer to the modified map.
328 * Also, it no longer reverses the order of the entries (which may be
329 * important with some advanced template substitutions).
331 * Revision 1.26 2001/07/30 22:08:36 jongfoster
332 * Tidying up #defines:
333 * - All feature #defines are now of the form FEATURE_xxx
334 * - Permanently turned off WIN_GUI_EDIT
335 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
337 * Revision 1.25 2001/07/26 10:09:46 oes
338 * Made browser detection a little less naive
340 * Revision 1.24 2001/07/25 17:22:51 oes
341 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
343 * Revision 1.23 2001/07/23 13:40:12 oes
344 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
346 * Revision 1.22 2001/07/18 12:29:34 oes
347 * - Made gif_deanimate_response respect
348 * csp->action->string[ACTION_STRING_DEANIMATE]
349 * - Logging cosmetics
351 * Revision 1.21 2001/07/13 13:59:53 oes
352 * - Introduced gif_deanimate_response which shares the
353 * generic content modification interface of pcrs_filter_response
354 * and acts as a wrapper to deanimate.c:gif_deanimate()
355 * - Renamed re_process_buffer to pcrs_filter_response
356 * - pcrs_filter_response now returns NULL on failiure
357 * - Removed all #ifdef PCRS
359 * Revision 1.20 2001/07/01 17:01:04 oes
360 * Added comments and missing return statement in is_untrusted_url()
362 * Revision 1.19 2001/06/29 21:45:41 oes
363 * Indentation, CRLF->LF, Tab-> Space
365 * Revision 1.18 2001/06/29 13:27:38 oes
366 * - Cleaned up, renamed and reorderd functions
367 * and improved comments
370 * - Ported to CGI platform. Now delivers
371 * http_response or NULL
372 * - Unified HTML and GIF generation (moved image detection
373 * and GIF generation here from jcc.c:chat())
374 * - Fixed HTTP status to:
375 * - 403 (Forbidden) for the "blocked" HTML message
376 * - 200 (OK) for GIF answers
377 * - 302 (Redirect) for redirect to GIF
380 * - Ported to CGI platform. Now delivers
381 * http_response or NULL
382 * - Separated detection of untrusted URL into
383 * (bool)is_untrusted_url
384 * - Added enforcement of untrusted requests
386 * - Moved redirect_url() from cgi.c to here
387 * and ported it to the CGI platform
389 * - Removed logentry from cancelled commit
391 * Revision 1.17 2001/06/09 10:55:28 jongfoster
392 * Changing BUFSIZ ==> BUFFER_SIZE
394 * Revision 1.16 2001/06/07 23:10:26 jongfoster
395 * Allowing unanchored domain patterns to back off and retry
396 * if they partially match. Optimized right-anchored patterns.
397 * Moving ACL and forward files into config file.
398 * Replacing struct gateway with struct forward_spec
400 * Revision 1.15 2001/06/03 19:12:00 oes
401 * extracted-CGI relevant stuff
403 * Revision 1.14 2001/06/01 10:30:55 oes
404 * Added optional left-anchoring to domaincmp
406 * Revision 1.13 2001/05/31 21:21:30 jongfoster
407 * Permissionsfile / actions file changes:
408 * - Changed "permission" to "action" throughout
409 * - changes to file format to allow string parameters
410 * - Moved helper functions to actions.c
412 * Revision 1.12 2001/05/31 17:35:20 oes
414 * - Enhanced domain part globbing with infix and prefix asterisk
415 * matching and optional unanchored operation
417 * Revision 1.11 2001/05/29 11:53:23 oes
418 * "See why" link added to "blocked" page
420 * Revision 1.10 2001/05/29 09:50:24 jongfoster
421 * Unified blocklist/imagelist/permissionslist.
422 * File format is still under discussion, but the internal changes
425 * Also modified interceptor behaviour:
426 * - We now intercept all URLs beginning with one of the following
427 * prefixes (and *only* these prefixes):
429 * * http://ijbswa.sf.net/config/
430 * * http://ijbswa.sourceforge.net/config/
431 * - New interceptors "home page" - go to http://i.j.b/ to see it.
432 * - Internal changes so that intercepted and fast redirect pages
433 * are not replaced with an image.
434 * - Interceptors now have the option to send a binary page direct
435 * to the client. (i.e. ijb-send-banner uses this)
436 * - Implemented show-url-info interceptor. (Which is why I needed
437 * the above interceptors changes - a typical URL is
438 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
439 * The previous mechanism would not have intercepted that, and
440 * if it had been intercepted then it then it would have replaced
443 * Revision 1.9 2001/05/27 22:17:04 oes
445 * - re_process_buffer no longer writes the modified buffer
446 * to the client, which was very ugly. It now returns the
447 * buffer, which it is then written by chat.
449 * - content_length now adjusts the Content-Length: header
450 * for modified documents rather than crunch()ing it.
451 * (Length info in csp->content_length, which is 0 for
452 * unmodified documents)
454 * - For this to work, sed() is called twice when filtering.
456 * Revision 1.8 2001/05/26 17:13:28 jongfoster
457 * Filled in a function comment.
459 * Revision 1.7 2001/05/26 15:26:15 jongfoster
460 * ACL feature now provides more security by immediately dropping
461 * connections from untrusted hosts.
463 * Revision 1.6 2001/05/26 00:28:36 jongfoster
464 * Automatic reloading of config file.
465 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
466 * Most of the global variables have been moved to a new
467 * struct configuration_spec, accessed through csp->config->globalname
468 * Most of the globals remaining are used by the Win32 GUI.
470 * Revision 1.5 2001/05/25 22:34:30 jongfoster
473 * Revision 1.4 2001/05/22 18:46:04 oes
475 * - Enabled filtering banners by size rather than URL
476 * by adding patterns that replace all standard banner
477 * sizes with the "Junkbuster" gif to the re_filterfile
479 * - Enabled filtering WebBugs by providing a pattern
480 * which kills all 1x1 images
482 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
483 * which is selected by the (nonstandard and therefore
484 * capital) letter 'U' in the option string.
485 * It causes the quantifiers to be ungreedy by default.
486 * Appending a ? turns back to greedy (!).
488 * - Added a new interceptor ijb-send-banner, which
489 * sends back the "Junkbuster" gif. Without imagelist or
490 * MSIE detection support, or if tinygif = 1, or the
491 * URL isn't recognized as an imageurl, a lame HTML
492 * explanation is sent instead.
494 * - Added new feature, which permits blocking remote
495 * script redirects and firing back a local redirect
497 * The feature is conditionally compiled, i.e. it
498 * can be disabled with --disable-fast-redirects,
499 * plus it must be activated by a "fast-redirects"
500 * line in the config file, has its own log level
501 * and of course wants to be displayed by show-proxy-args
502 * Note: Boy, all the #ifdefs in 1001 locations and
503 * all the fumbling with configure.in and acconfig.h
504 * were *way* more work than the feature itself :-(
506 * - Because a generic redirect template was needed for
507 * this, tinygif = 3 now uses the same.
509 * - Moved GIFs, and other static HTTP response templates
514 * - Removed some >400 CRs again (Jon, you really worked
517 * Revision 1.3 2001/05/20 16:44:47 jongfoster
518 * Removing last hardcoded Junkbusters.com URLs.
520 * Revision 1.2 2001/05/20 01:21:20 jongfoster
521 * Version 2.9.4 checkin.
522 * - Merged popupfile and cookiefile, and added control over PCRS
523 * filtering, in new "permissionsfile".
524 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
525 * file error you now get a message box (in the Win32 GUI) rather
526 * than the program exiting with no explanation.
527 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
529 * - Removed tabs from "config"
530 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
531 * - Bumped up version number.
533 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
534 * Initial import of version 2.9.3 source tree
537 *********************************************************************/
543 #include <sys/types.h>
552 #endif /* ndef __OS2__ */
553 #include <netinet/in.h>
555 #include <winsock2.h>
556 #endif /* ndef _WIN32 */
560 #endif /* def __OS2__ */
568 #include "jbsockets.h"
569 #include "miscutil.h"
573 #include "deanimate.h"
574 #include "urlmatch.h"
580 const char filters_h_rcs[] = FILTERS_H_VERSION;
582 /* Fix a problem with Solaris. There should be no effect on other
584 * Solaris's isspace() is a macro which uses it's argument directly
585 * as an array index. Therefore we need to make sure that high-bit
586 * characters generate +ve values, and ideally we also want to make
587 * the argument match the declared parameter type of "int".
589 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
593 /*********************************************************************
595 * Function : block_acl
597 * Description : Block this request?
598 * Decide yes or no based on ACL file.
601 * 1 : dst = The proxy or gateway address this is going to.
602 * Or NULL to check all possible targets.
603 * 2 : csp = Current client state (buffers, headers, etc...)
604 * Also includes the client IP address.
606 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
608 *********************************************************************/
609 int block_acl(struct access_control_addr *dst, struct client_state *csp)
611 struct access_control_list *acl = csp->config->acl;
613 /* if not using an access control list, then permit the connection */
619 /* search the list */
622 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
626 /* Just want to check if they have any access */
627 if (acl->action == ACL_PERMIT)
632 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
633 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
635 if (acl->action == ACL_PERMIT)
653 /*********************************************************************
655 * Function : acl_addr
657 * Description : Called from `load_config' to parse an ACL address.
660 * 1 : aspec = String specifying ACL address.
661 * 2 : aca = struct access_control_addr to fill in.
663 * Returns : 0 => Ok, everything else is an error.
665 *********************************************************************/
666 int acl_addr(const char *aspec, struct access_control_addr *aca)
671 char *acl_spec = NULL;
677 * Use a temporary acl spec copy so we can log
678 * the unmodified original in case of parse errors.
680 acl_spec = strdup(aspec);
681 if (acl_spec == NULL)
683 /* XXX: This will be logged as parse error. */
687 if ((p = strchr(acl_spec, '/')) != NULL)
690 if (ijb_isdigit(*p) == 0)
695 masklength = atoi(p);
698 if ((masklength < 0) || (masklength > 32))
704 if ((p = strchr(acl_spec, ':')) != NULL)
709 port = strtol(p, &endptr, 10);
711 if (port <= 0 || port > 65535 || *endptr != '\0')
718 aca->port = (unsigned long)port;
720 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
723 if (aca->addr == INADDR_NONE)
725 /* XXX: This will be logged as parse error. */
729 /* build the netmask */
731 for (i=1; i <= masklength ; i++)
733 aca->mask |= (1 << (32 - i));
736 /* now mask off the host portion of the ip address
737 * (i.e. save on the network portion of the address).
739 aca->addr = aca->addr & aca->mask;
744 #endif /* def FEATURE_ACL */
747 /*********************************************************************
749 * Function : match_portlist
751 * Description : Check if a given number is covered by a comma
752 * separated list of numbers and ranges (a,b-c,d,..)
755 * 1 : portlist = String with list
756 * 2 : port = port to check
758 * Returns : 0 => no match
761 *********************************************************************/
762 int match_portlist(const char *portlist, int port)
764 char *min, *max, *next, *portlist_copy;
766 min = next = portlist_copy = strdup(portlist);
769 * Zero-terminate first item and remember offset for next
771 if (NULL != (next = strchr(portlist_copy, (int) ',')))
777 * Loop through all items, checking for match
781 if (NULL == (max = strchr(min, (int) '-')))
784 * No dash, check for equality
786 if (port == atoi(min))
795 * This is a range, so check if between min and max,
796 * or, if max was omitted, between min and 65K
799 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
813 * Zero-terminate next item and remember offset for n+1
815 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
827 /*********************************************************************
829 * Function : block_url
831 * Description : Called from `chat'. Check to see if we need to block this.
834 * 1 : csp = Current client state (buffers, headers, etc...)
836 * Returns : NULL => unblocked, else HTTP block response
838 *********************************************************************/
839 struct http_response *block_url(struct client_state *csp)
841 struct http_response *rsp;
842 const char *new_content_type = NULL;
845 * If it's not blocked, don't block it ;-)
847 if ((csp->action->flags & ACTION_BLOCK) == 0)
851 if (csp->action->flags & ACTION_REDIRECT)
853 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
856 * Else, prepare a response
858 if (NULL == (rsp = alloc_http_response()))
860 return cgi_error_memory();
864 * If it's an image-url, send back an image or redirect
865 * as specified by the relevant +image action
867 #ifdef FEATURE_IMAGE_BLOCKING
868 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
872 /* determine HOW images should be blocked */
873 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
875 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
877 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
879 #if 1 /* Two alternative strategies, use this one for now: */
881 /* and handle accordingly: */
882 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
884 rsp->status = strdup("403 Request blocked by Privoxy");
885 if (rsp->status == NULL)
887 free_http_response(rsp);
888 return cgi_error_memory();
890 rsp->body = bindup(image_pattern_data, image_pattern_length);
891 if (rsp->body == NULL)
893 free_http_response(rsp);
894 return cgi_error_memory();
896 rsp->content_length = image_pattern_length;
898 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
900 free_http_response(rsp);
901 return cgi_error_memory();
905 else if (0 == strcmpic(p, "blank"))
907 rsp->status = strdup("403 Request blocked by Privoxy");
908 if (rsp->status == NULL)
910 free_http_response(rsp);
911 return cgi_error_memory();
913 rsp->body = bindup(image_blank_data, image_blank_length);
914 if (rsp->body == NULL)
916 free_http_response(rsp);
917 return cgi_error_memory();
919 rsp->content_length = image_blank_length;
921 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
923 free_http_response(rsp);
924 return cgi_error_memory();
930 rsp->status = strdup("302 Local Redirect from Privoxy");
931 if (rsp->status == NULL)
933 free_http_response(rsp);
934 return cgi_error_memory();
937 if (enlist_unique_header(rsp->headers, "Location", p))
939 free_http_response(rsp);
940 return cgi_error_memory();
944 #else /* Following code is disabled for now */
946 /* and handle accordingly: */
947 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
949 p = CGI_PREFIX "send-banner?type=pattern";
951 else if (0 == strcmpic(p, "blank"))
953 p = CGI_PREFIX "send-banner?type=blank";
955 rsp->status = strdup("302 Local Redirect from Privoxy");
956 if (rsp->status == NULL)
958 free_http_response(rsp);
959 return cgi_error_memory();
962 if (enlist_unique_header(rsp->headers, "Location", p))
964 free_http_response(rsp);
965 return cgi_error_memory();
967 #endif /* Preceeding code is disabled for now */
969 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
972 * Send empty document.
974 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
977 rsp->body = strdup(" ");
978 rsp->content_length = 1;
980 rsp->status = strdup("403 Request blocked by Privoxy");
981 if (rsp->status == NULL)
983 free_http_response(rsp);
984 return cgi_error_memory();
986 if (new_content_type != 0)
988 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
989 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
991 free_http_response(rsp);
992 return cgi_error_memory();
997 #endif /* def FEATURE_IMAGE_BLOCKING */
1000 * Else, generate an HTML "blocked" message:
1004 struct map * exports;
1008 * Workaround for stupid Netscape bug which prevents
1009 * pages from being displayed if loading a referenced
1010 * JavaScript or style sheet fails. So make it appear
1011 * as if it succeeded.
1013 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1014 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1015 && !strstr(p, "Gecko") /* save Mozilla, */
1016 && !strstr(p, "compatible") /* MSIE */
1017 && !strstr(p, "Opera")) /* and Opera. */
1019 rsp->status = strdup("200 Request for blocked URL");
1023 rsp->status = strdup("403 Request for blocked URL");
1026 if (rsp->status == NULL)
1028 free_http_response(rsp);
1029 return cgi_error_memory();
1032 exports = default_exports(csp, NULL);
1033 if (exports == NULL)
1035 free_http_response(rsp);
1036 return cgi_error_memory();
1039 #ifdef FEATURE_FORCE_LOAD
1040 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1042 * Export the force conditional block killer if
1044 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1045 * - Privoxy is configured to enforce blocks, or
1046 * - it's a CONNECT request and enforcing wouldn't work anyway.
1048 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1049 || (0 == strcmpic(csp->http->gpc, "connect")))
1050 #endif /* ndef FEATURE_FORCE_LOAD */
1052 err = map_block_killer(exports, "force-support");
1055 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1056 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1057 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1058 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1063 free_http_response(rsp);
1064 return cgi_error_memory();
1067 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1070 free_http_response(rsp);
1071 return cgi_error_memory();
1074 rsp->reason = RSP_REASON_BLOCKED;
1076 return finish_http_response(csp, rsp);
1081 #ifdef FEATURE_TRUST
1082 /*********************************************************************
1084 * Function : trust_url FIXME: I should be called distrust_url
1086 * Description : Calls is_untrusted_url to determine if the URL is trusted
1087 * and if not, returns a HTTP 403 response with a reject message.
1090 * 1 : csp = Current client state (buffers, headers, etc...)
1092 * Returns : NULL => trusted, else http_response.
1094 *********************************************************************/
1095 struct http_response *trust_url(struct client_state *csp)
1097 struct http_response *rsp;
1098 struct map * exports;
1099 char buf[BUFFER_SIZE];
1101 struct url_spec **tl;
1106 * Don't bother to work on trusted URLs
1108 if (!is_untrusted_url(csp))
1114 * Else, prepare a response:
1116 if (NULL == (rsp = alloc_http_response()))
1118 return cgi_error_memory();
1121 rsp->status = strdup("403 Request blocked by Privoxy");
1122 exports = default_exports(csp, NULL);
1123 if (exports == NULL || rsp->status == NULL)
1125 free_http_response(rsp);
1126 return cgi_error_memory();
1130 * Export the protocol, host, port, and referrer information
1132 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1133 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1134 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1136 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1138 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1142 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1148 free_http_response(rsp);
1149 return cgi_error_memory();
1153 * Export the trust list
1156 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1158 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1159 string_append(&p, buf);
1161 err = map(exports, "trusted-referrers", 1, p, 0);
1166 free_http_response(rsp);
1167 return cgi_error_memory();
1171 * Export the trust info, if available
1173 if (csp->config->trust_info->first)
1175 struct list_entry *l;
1178 for (l = csp->config->trust_info->first; l ; l = l->next)
1180 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1181 string_append(&p, buf);
1183 err = map(exports, "trust-info", 1, p, 0);
1187 err = map_block_killer(exports, "have-trust-info");
1193 free_http_response(rsp);
1194 return cgi_error_memory();
1198 * Export the force conditional block killer if
1200 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1201 * - Privoxy is configured to enforce blocks, or
1202 * - it's a CONNECT request and enforcing wouldn't work anyway.
1204 #ifdef FEATURE_FORCE_LOAD
1205 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1206 || (0 == strcmpic(csp->http->gpc, "connect")))
1208 err = map_block_killer(exports, "force-support");
1212 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1214 #else /* ifndef FEATURE_FORCE_LOAD */
1215 err = map_block_killer(exports, "force-support");
1216 #endif /* ndef FEATURE_FORCE_LOAD */
1221 free_http_response(rsp);
1222 return cgi_error_memory();
1226 * Build the response
1228 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1231 free_http_response(rsp);
1232 return cgi_error_memory();
1234 rsp->reason = RSP_REASON_UNTRUSTED;
1236 return finish_http_response(csp, rsp);
1238 #endif /* def FEATURE_TRUST */
1240 /*********************************************************************
1242 * Function : execute_single_pcrs_command
1244 * Description : Apply single pcrs command to the subject.
1245 * The subject itself is left untouched, memory for the result
1246 * is malloc()ed and it is the caller's responsibility to free
1247 * the result when it's no longer needed.
1250 * 1 : subject = the subject (== original) string
1251 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1252 * 3 : hits = int* for returning the number of modifications
1254 * Returns : NULL in case of errors, otherwise the
1255 * result of the pcrs command.
1257 *********************************************************************/
1258 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1262 char *result = NULL;
1266 assert(pcrs_command);
1269 size = strlen(subject);
1271 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1273 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1274 pcrs_command, error);
1276 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1278 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1285 job = pcrs_free_job(job);
1292 /*********************************************************************
1294 * Function : rewrite_url
1296 * Description : Rewrites a URL with a single pcrs command
1297 * and returns the result if it differs from the
1298 * original and isn't obviously invalid.
1301 * 1 : old_url = URL to rewrite.
1302 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1305 * Returns : NULL if the pcrs_command didn't change the url, or
1306 * the result of the modification.
1308 *********************************************************************/
1309 char *rewrite_url(char *old_url, const char *pcrs_command)
1311 char *new_url = NULL;
1315 assert(pcrs_command);
1317 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1321 log_error(LOG_LEVEL_REDIRECTS,
1322 "pcrs command \"%s\" didn't change \"%s\".",
1323 pcrs_command, old_url, new_url);
1326 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1328 log_error(LOG_LEVEL_ERROR,
1329 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1330 "but the result doesn't look like a valid URL and will be ignored.",
1331 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1336 log_error(LOG_LEVEL_REDIRECTS,
1337 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1338 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1346 #ifdef FEATURE_FAST_REDIRECTS
1347 /*********************************************************************
1349 * Function : get_last_url
1351 * Description : Search for the last URL inside a string.
1352 * If the string already is a URL, it will
1353 * be the first URL found.
1356 * 1 : subject = the string to check
1357 * 2 : redirect_mode = +fast-redirect{} mode
1359 * Returns : NULL if no URL was found, or
1360 * the last URL found.
1362 *********************************************************************/
1363 char *get_last_url(char *subject, const char *redirect_mode)
1365 char *new_url = NULL;
1369 assert(redirect_mode);
1371 subject = strdup(subject);
1372 if (subject == NULL)
1374 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1378 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1380 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1381 new_url = url_decode(subject);
1382 if (new_url != NULL)
1389 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1393 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1396 * Find the last URL encoded in the request
1399 while ((tmp = strstr(tmp, "http://")) != NULL)
1403 tmp = (new_url != NULL) ? new_url : subject;
1404 while ((tmp = strstr(tmp, "https://")) != NULL)
1409 if ((new_url != NULL)
1410 && ( (new_url != subject)
1411 || (0 == strncmpic(subject, "http://", 7))
1412 || (0 == strncmpic(subject, "https://", 8))
1416 * Return new URL if we found a redirect
1417 * or if the subject already was a URL.
1419 * The second case makes sure that we can
1420 * chain get_last_url after another redirection check
1421 * (like rewrite_url) without losing earlier redirects.
1423 new_url = strdup(new_url);
1432 #endif /* def FEATURE_FAST_REDIRECTS */
1435 /*********************************************************************
1437 * Function : redirect_url
1439 * Description : Checks if Privoxy should answer the request with
1440 * a HTTP redirect and generates the redirect if
1444 * 1 : csp = Current client state (buffers, headers, etc...)
1446 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1448 *********************************************************************/
1449 struct http_response *redirect_url(struct client_state *csp)
1451 struct http_response *rsp;
1452 #ifdef FEATURE_FAST_REDIRECTS
1454 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1455 * as compile-time option? The user can easily disable
1456 * it in his action file.
1458 char * redirect_mode;
1459 #endif /* def FEATURE_FAST_REDIRECTS */
1460 char *old_url = NULL;
1461 char *new_url = NULL;
1462 char *redirection_string;
1464 if ((csp->action->flags & ACTION_REDIRECT))
1466 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1469 * If the redirection string begins with 's',
1470 * assume it's a pcrs command, otherwise treat it as
1471 * properly formatted URL and use it for the redirection
1474 * According to RFC 2616 section 14.30 the URL
1475 * has to be absolute and if the user tries:
1476 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1477 * she would get undefined results anyway.
1481 if (*redirection_string == 's')
1483 old_url = csp->http->url;
1484 new_url = rewrite_url(old_url, redirection_string);
1488 log_error(LOG_LEVEL_REDIRECTS,
1489 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1490 redirection_string);
1491 new_url = strdup(redirection_string);
1495 #ifdef FEATURE_FAST_REDIRECTS
1496 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1498 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1501 * If it exists, use the previously rewritten URL as input
1502 * otherwise just use the old path.
1504 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1505 new_url = get_last_url(old_url, redirect_mode);
1508 #endif /* def FEATURE_FAST_REDIRECTS */
1510 /* Did any redirect action trigger? */
1513 if (0 == strcmpic(new_url, csp->http->url))
1515 log_error(LOG_LEVEL_ERROR,
1516 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1517 csp->http->url, new_url);
1522 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1524 if (NULL == (rsp = alloc_http_response()))
1527 return cgi_error_memory();
1530 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1531 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1534 free_http_response(rsp);
1535 return cgi_error_memory();
1537 rsp->reason = RSP_REASON_REDIRECTED;
1540 return finish_http_response(csp, rsp);
1544 /* Only reached if no redirect is required */
1550 #ifdef FEATURE_IMAGE_BLOCKING
1551 /*********************************************************************
1553 * Function : is_imageurl
1555 * Description : Given a URL, decide whether it is an image or not,
1556 * using either the info from a previous +image action
1557 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1558 * is MSIE and not on a Mac, tell from the browser's accept
1562 * 1 : csp = Current client state (buffers, headers, etc...)
1564 * Returns : True (nonzero) if URL is an image, false (0)
1567 *********************************************************************/
1568 int is_imageurl(const struct client_state *csp)
1570 #ifdef FEATURE_IMAGE_DETECT_MSIE
1573 tmp = get_header_value(csp->headers, "User-Agent:");
1574 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1576 tmp = get_header_value(csp->headers, "Accept:");
1577 if (tmp && strstr(tmp, "image/gif"))
1579 /* Client will accept HTML. If this seems counterintuitive,
1589 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1591 return ((csp->action->flags & ACTION_IMAGE) != 0);
1594 #endif /* def FEATURE_IMAGE_BLOCKING */
1597 #ifdef FEATURE_TRUST
1598 /*********************************************************************
1600 * Function : is_untrusted_url
1602 * Description : Should we "distrust" this URL (and block it)?
1604 * Yes if it matches a line in the trustfile, or if the
1605 * referrer matches a line starting with "+" in the
1610 * 1 : csp = Current client state (buffers, headers, etc...)
1612 * Returns : 0 => trusted, 1 => untrusted
1614 *********************************************************************/
1615 int is_untrusted_url(const struct client_state *csp)
1617 struct file_list *fl;
1618 struct block_spec *b;
1619 struct url_spec **trusted_url;
1620 struct http_request rhttp[1];
1621 const char * referer;
1625 * If we don't have a trustlist, we trust everybody
1627 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1632 memset(rhttp, '\0', sizeof(*rhttp));
1635 * Do we trust the request URL itself?
1637 for (b = b->next; b ; b = b->next)
1639 if (url_match(b->url, csp->http))
1645 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1647 /* no referrer was supplied */
1653 * If not, do we maybe trust its referrer?
1655 err = parse_http_url(referer, rhttp, csp);
1661 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1663 if (url_match(*trusted_url, rhttp))
1665 /* if the URL's referrer is from a trusted referrer, then
1666 * add the target spec to the trustfile as an unblocked
1667 * domain and return 0 (which means it's OK).
1672 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1676 char * new_entry = strdup("~");
1678 string_append(&new_entry, csp->http->hostport);
1680 path = csp->http->path;
1681 if ( (path[0] == '/')
1683 && ((path_end = strchr(path + 2, '/')) != NULL))
1685 /* since this path points into a user's home space
1686 * be sure to include this spec in the trustfile.
1688 int path_len = path_end - path; /* save offset */
1689 path = strdup(path); /* Copy string */
1692 path_end = path + path_len; /* regenerate ptr to new buffer */
1693 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1695 string_join(&new_entry, path);
1699 * Give a reason for generating this entry.
1701 string_append(&new_entry, " # Trusted referrer was: ");
1702 string_append(&new_entry, referer);
1704 if (new_entry != NULL)
1706 if (-1 == fprintf(fp, "%s\n", new_entry))
1708 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1709 new_entry, csp->config->trustfile);
1715 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1716 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1723 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1724 csp->http->hostport, csp->config->trustfile);
1732 #endif /* def FEATURE_TRUST */
1735 /*********************************************************************
1737 * Function : pcrs_filter_response
1739 * Description : Execute all text substitutions from all applying
1740 * +filter actions on the text buffer that's been accumulated
1741 * in csp->iob->buf. If this changes the contents, set
1742 * csp->content_length to the modified size and raise the
1743 * CSP_FLAG_MODIFIED flag.
1745 * XXX: Currently pcrs_filter_response is also responsible
1746 * for dechunking and decompressing. Both should be
1747 * done in separate functions so other content modifiers
1748 * profit as well, even if pcrs filtering is disabled.
1751 * 1 : csp = Current client state (buffers, headers, etc...)
1753 * Returns : a pointer to the (newly allocated) modified buffer.
1754 * or NULL if there were no hits or something went wrong
1756 *********************************************************************/
1757 char *pcrs_filter_response(struct client_state *csp)
1760 size_t size, prev_size;
1762 char *old = csp->iob->cur, *new = NULL;
1765 struct file_list *fl;
1766 struct re_filterfile_spec *b;
1767 struct list_entry *filtername;
1769 int i, found_filters = 0;
1774 if (csp->iob->cur >= csp->iob->eod)
1778 size = (size_t)(csp->iob->eod - csp->iob->cur);
1781 * Need to check the set of re_filterfiles...
1783 for (i = 0; i < MAX_AF_FILES; i++)
1796 if (0 == found_filters)
1798 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1803 * If the body has a "chunked" transfer-encoding,
1804 * get rid of it first, adjusting size and iob->eod
1806 if (csp->flags & CSP_FLAG_CHUNKED)
1808 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1809 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1813 csp->iob->eod = csp->iob->cur + size;
1814 csp->flags |= CSP_FLAG_MODIFIED;
1819 * If the body has a compressed transfer-encoding,
1820 * uncompress it first, adjusting size and iob->eod.
1821 * Note that decompression occurs after de-chunking.
1823 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1825 /* Notice that we at least tried to decompress. */
1826 if (JB_ERR_OK != decompress_iob(csp))
1829 * We failed to decompress the data; there's no point
1830 * in continuing since we can't filter.
1832 * XXX: Actually the Accept-Encoding header may
1833 * just be incorrect in which case we could continue
1836 * Unset CT_GZIP and CT_DEFLATE to remember not
1837 * to modify the Content-Encoding header later.
1839 csp->content_type &= ~CT_GZIP;
1840 csp->content_type &= ~CT_DEFLATE;
1843 log_error(LOG_LEVEL_RE_FILTER,
1844 "Decompression successful. Old size: %d, new size: %d.",
1845 size, csp->iob->eod - csp->iob->cur);
1848 * Decompression gives us a completely new iob,
1849 * so we need to update.
1851 size = (size_t)(csp->iob->eod - csp->iob->cur);
1852 old = csp->iob->cur;
1854 csp->flags |= CSP_FLAG_MODIFIED;
1855 csp->content_type &= ~CT_TABOO;
1859 for (i = 0; i < MAX_AF_FILES; i++)
1862 if ((NULL == fl) || (NULL == fl->f))
1865 * Either there are no filter files
1866 * left, or this filter file just
1867 * contains no valid filters.
1869 * Continue to be sure we don't miss
1870 * valid filter files that are chained
1871 * after empty or invalid ones.
1876 * For all applying +filter actions, look if a filter by that
1877 * name exists and if yes, execute it's pcrs_joblist on the
1880 for (b = fl->f; b; b = b->next)
1882 if (b->type != FT_CONTENT_FILTER)
1884 /* Skip header filters */
1888 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1889 filtername ; filtername = filtername->next)
1891 if (strcmp(b->name, filtername->str) == 0)
1893 int current_hits = 0; /* Number of hits caused by this filter */
1894 int job_number = 0; /* Which job we're currently executing */
1895 int job_hits = 0; /* How many hits the current job caused */
1897 if ( NULL == b->joblist )
1899 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1904 /* Apply all jobs from the joblist */
1905 for (job = b->joblist; NULL != job; job = job->next)
1908 job_hits = pcrs_execute(job, old, size, &new, &size);
1913 * That went well. Continue filtering
1914 * and use the result of this job as
1915 * input for the next one.
1917 current_hits += job_hits;
1918 if (old != csp->iob->cur)
1927 * This job caused an unexpected error. Inform the user
1928 * and skip the rest of the jobs in this filter. We could
1929 * continue with the next job, but usually the jobs
1930 * depend on each other or are similar enough to
1931 * fail for the same reason.
1933 * At the moment our pcrs expects the error codes of pcre 3.4,
1934 * but newer pcre versions can return additional error codes.
1935 * As a result pcrs_strerror()'s error message might be
1936 * "Unknown error ...", therefore we print the numerical value
1939 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1940 * should we use LOG_LEVEL_RE_FILTER instead?
1942 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1943 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1948 log_error(LOG_LEVEL_RE_FILTER,
1949 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1950 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1952 hits += current_hits;
1959 * If there were no hits, destroy our copy and let
1960 * chat() use the original in csp->iob
1968 csp->flags |= CSP_FLAG_MODIFIED;
1969 csp->content_length = size;
1977 /*********************************************************************
1979 * Function : gif_deanimate_response
1981 * Description : Deanimate the GIF image that has been accumulated in
1982 * csp->iob->buf, set csp->content_length to the modified
1983 * size and raise the CSP_FLAG_MODIFIED flag.
1986 * 1 : csp = Current client state (buffers, headers, etc...)
1988 * Returns : a pointer to the (newly allocated) modified buffer.
1989 * or NULL in case something went wrong.
1991 *********************************************************************/
1992 char *gif_deanimate_response(struct client_state *csp)
1994 struct binbuffer *in, *out;
1998 size = (size_t)(csp->iob->eod - csp->iob->cur);
2001 * If the body has a "chunked" transfer-encoding,
2002 * get rid of it first, adjusting size and iob->eod
2004 if (csp->flags & CSP_FLAG_CHUNKED)
2006 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2007 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2011 csp->iob->eod = csp->iob->cur + size;
2012 csp->flags |= CSP_FLAG_MODIFIED;
2015 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2016 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2018 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2022 in->buffer = csp->iob->cur;
2025 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2027 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2034 if ((int)size == out->offset)
2036 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2040 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2042 csp->content_length = out->offset;
2043 csp->flags |= CSP_FLAG_MODIFIED;
2053 /*********************************************************************
2055 * Function : jpeg_inspect_response
2060 * 1 : csp = Current client state (buffers, headers, etc...)
2062 * Returns : a pointer to the (newly allocated) modified buffer
2063 * or NULL in case something went wrong.
2065 *********************************************************************/
2066 char *jpeg_inspect_response(struct client_state *csp)
2068 struct binbuffer *in = NULL;
2069 struct binbuffer *out = NULL;
2073 size = (size_t)(csp->iob->eod - csp->iob->cur);
2076 * If the body has a "chunked" transfer-encoding,
2077 * get rid of it first, adjusting size and iob->eod
2079 if (csp->flags & CSP_FLAG_CHUNKED)
2081 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2082 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2086 csp->iob->eod = csp->iob->cur + size;
2087 csp->flags |= CSP_FLAG_MODIFIED;
2090 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2092 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2096 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2098 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2102 in->buffer = csp->iob->cur;
2106 * Calling jpeg_inspect has the side-effect of creating and
2107 * modifying the image buffer of "out" directly.
2109 if (jpeg_inspect(in, out))
2111 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2119 csp->content_length = out->offset;
2120 csp->flags |= CSP_FLAG_MODIFIED;
2130 /*********************************************************************
2132 * Function : remove_chunked_transfer_coding
2134 * Description : In-situ remove the "chunked" transfer coding as defined
2135 * in rfc2616 from a buffer.
2138 * 1 : buffer = Pointer to the text buffer
2139 * 2 : size = Number of bytes to be processed
2141 * Returns : The new size, i.e. the number of bytes from buffer which
2142 * are occupied by the stripped body, or 0 in case something
2145 *********************************************************************/
2146 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2149 unsigned int chunksize = 0;
2150 char *from_p, *to_p;
2153 from_p = to_p = buffer;
2155 if (sscanf(buffer, "%x", &chunksize) != 1)
2157 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2161 while (chunksize > 0)
2163 if (NULL == (from_p = strstr(from_p, "\r\n")))
2165 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2169 if ((newsize += chunksize) >= size)
2171 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2176 memmove(to_p, from_p, (size_t) chunksize);
2177 to_p = buffer + newsize;
2178 from_p += chunksize + 2;
2180 if (sscanf(from_p, "%x", &chunksize) != 1)
2182 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2187 /* FIXME: Should this get its own loglevel? */
2188 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2194 /*********************************************************************
2196 * Function : url_actions
2198 * Description : Gets the actions for this URL.
2201 * 1 : http = http_request request for blocked URLs
2202 * 2 : csp = Current client state (buffers, headers, etc...)
2206 *********************************************************************/
2207 void url_actions(struct http_request *http,
2208 struct client_state *csp)
2210 struct file_list *fl;
2211 struct url_actions *b;
2214 init_current_action(csp->action);
2216 for (i = 0; i < MAX_AF_FILES; i++)
2218 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2223 apply_url_actions(csp->action, http, b);
2230 /*********************************************************************
2232 * Function : apply_url_actions
2234 * Description : Applies a list of URL actions.
2237 * 1 : action = Destination.
2238 * 2 : http = Current URL
2239 * 3 : b = list of URL actions to apply
2243 *********************************************************************/
2244 void apply_url_actions(struct current_action_spec *action,
2245 struct http_request *http,
2246 struct url_actions *b)
2250 /* Should never happen */
2254 for (b = b->next; NULL != b; b = b->next)
2256 if (url_match(b->url, http))
2258 merge_current_action(action, b->action);
2264 /*********************************************************************
2266 * Function : forward_url
2268 * Description : Should we forward this to another proxy?
2271 * 1 : http = http_request request for current URL
2272 * 2 : csp = Current client state (buffers, headers, etc...)
2274 * Returns : Pointer to forwarding information.
2276 *********************************************************************/
2277 const struct forward_spec * forward_url(struct http_request *http,
2278 struct client_state *csp)
2280 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2281 struct forward_spec *fwd = csp->config->forward;
2290 if (url_match(fwd->url, http))
2301 /*********************************************************************
2303 * Function : direct_response
2305 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2306 * request and if so, return a HTTP 501 to the client.
2308 * FIXME: I have a stupid name and I should handle the
2309 * requests properly. Still, what we do here is rfc-
2310 * compliant, whereas ignoring or forwarding are not.
2313 * 1 : csp = Current client state (buffers, headers, etc...)
2315 * Returns : http_response if , NULL if nonmatch or handler fail
2317 *********************************************************************/
2318 struct http_response *direct_response(struct client_state *csp)
2320 struct http_response *rsp;
2321 struct list_entry *p;
2323 if ((0 == strcmpic(csp->http->gpc, "trace"))
2324 || (0 == strcmpic(csp->http->gpc, "options")))
2326 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2328 if (!strncmpic("Max-Forwards:", p->str, 13))
2330 unsigned int max_forwards;
2333 * If it's a Max-Forwards value of zero,
2334 * we have to intercept the request.
2336 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2339 * FIXME: We could handle at least TRACE here,
2340 * but that would require a verbatim copy of
2341 * the request which we don't have anymore
2343 log_error(LOG_LEVEL_HEADER,
2344 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2347 /* Get mem for response or fail*/
2348 if (NULL == (rsp = alloc_http_response()))
2350 return cgi_error_memory();
2353 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2355 free_http_response(rsp);
2356 return cgi_error_memory();
2360 rsp->reason = RSP_REASON_UNSUPPORTED;
2362 return(finish_http_response(csp, rsp));