1 const char filters_rcs[] = "$Id: filters.c,v 1.86 2007/04/30 15:03:28 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 1.86 2007/04/30 15:03:28 fabiankeil
44 * - Introduce dynamic pcrs jobs that can resolve variables.
45 * - Don't run redirect functions more than once,
46 * unless they are activated more than once.
48 * Revision 1.85 2007/03/21 12:24:47 fabiankeil
49 * - Log the content size after decompression in decompress_iob()
50 * instead of pcrs_filter_response().
52 * Revision 1.84 2007/03/20 15:16:34 fabiankeil
53 * Use dedicated header filter actions instead of abusing "filter".
54 * Replace "filter-client-headers" and "filter-client-headers"
55 * with "server-header-filter" and "client-header-filter".
57 * Revision 1.83 2007/03/17 15:20:05 fabiankeil
58 * New config option: enforce-blocks.
60 * Revision 1.82 2007/03/13 11:28:43 fabiankeil
61 * - Fix port handling in acl_addr() and use a temporary acl spec
62 * copy so error messages don't contain a truncated version.
63 * - Log size of iob before and after decompression.
65 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
66 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
67 * - Hide the "Go there anyway" link for blocked CONNECT
68 * requests where going there anyway doesn't work anyway.
70 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
71 * - Save the reason for generating http_responses.
72 * - Block (+block) with status code 403 instead of 404.
73 * - Use a different kludge to remember a failed decompression.
75 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
76 * Search for Max-Forwards headers case-insensitive,
77 * don't generate the "501 unsupported" message for invalid
78 * Max-Forwards values and don't increase negative ones.
80 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
81 * - Add HEAD support to finish_http_response.
82 * - Add error favicon to internal HTML error messages.
84 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
85 * Mark *csp as immutable for is_untrusted_url()
86 * and is_imageurl(). Closes FR 1237736.
88 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
89 * Integrate a modified version of Wil Mahan's
90 * zlib patch (PR #895531).
92 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
93 * Fixed gcc43 conversion warnings,
94 * changed sprintf calls to snprintf.
96 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
97 * Adjust comment in pcrs_filter_response()
98 * to recent pcrs changes. Hohoho.
100 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
101 * Don't crash if pcre returns an error code
102 * that pcrs didn't expect. Fixes BR 1621173.
104 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
105 * Modified is_untrusted_url to complain in case of
106 * write errors and to give a reason when adding new
107 * entries to the trustfile. Closes FR 1097611.
109 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
110 * Skip empty filter files in pcrs_filter_response,
111 * but don't ignore the ones that come afterwards.
112 * Fixes parts of BR 1619208.
114 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
115 * Added some sanity checks for get_last_url().
116 * Fixed possible segfault caused by my last commit.
118 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
119 * Let get_last_url() catch https URLs as well.
121 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
122 * Make sure get_last_url() behaves like advertised
123 * and fast-redirects{} can be combined with redirect{}.
125 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
126 * Implemented +redirect{s@foo@bar@} to generate
127 * a redirect based on a rewritten version of the
130 * Revision 1.66 2006/09/23 13:26:38 roro
131 * Replace TABs by spaces in source code.
133 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
134 * Fix +redirect{}. Didn't work with -fast-redirects.
136 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
137 * Block requests for untrusted URLs with status
138 * code 403 instead of 200.
140 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
141 * Don't free p which is still in use and will be later
142 * freed by free_map(). Don't claim the referrer is unknown
143 * when the client didn't set one.
145 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
146 * Feature request 595948: Re-Filter logging in single line
148 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
149 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
151 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
152 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
153 * with what was really the latest development (the v_3_0_branch branch)
155 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
156 * Multiple filter file support
158 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
159 * A few memory leaks plugged (mostly on error paths)
161 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
162 * Add the ability to check jpeg images for invalid
163 * lengths of comment blocks. Defensive strategy
164 * against the exploit:
165 * Microsoft Security Bulletin MS04-028
166 * Buffer Overrun in JPEG Processing (GDI+) Could
167 * Allow Code Execution (833987)
168 * Enabled with +inspect-jpegs in actions files.
170 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
171 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
173 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
174 * Fixed bug #839859: "See why" link URL now gets url-encoded.
176 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
179 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
180 * Added basic support for OPTIONS and TRACE HTTP methods:
181 * New function direct_response which handles OPTIONS and
182 * TRACE requests whose Max-Forwards header field is zero.
184 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
185 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
187 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
188 * - Bugfix: Executing a filters without jobs no longer results in
189 * turing off *all* filters.
190 * - Security fix: Malicious web servers can't cause a seg fault
191 * through bogus chunk sizes anymore
193 * Revision 1.58 2002/04/24 02:11:17 oes
194 * Jon's multiple AF patch: url_actions now evaluates rules
197 * Revision 1.57 2002/04/08 20:38:34 swa
200 * Revision 1.56 2002/04/05 15:51:24 oes
201 * - bugfix: error-pages now get correct request protocol
202 * - fix for invalid HTML in trust info
204 * Revision 1.55 2002/04/02 16:13:51 oes
205 * Fix: No "Go there anyway" for SSL
207 * Revision 1.54 2002/04/02 14:55:56 oes
208 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
210 * Revision 1.53 2002/03/26 22:29:54 swa
211 * we have a new homepage!
213 * Revision 1.52 2002/03/24 16:35:57 jongfoster
216 * Revision 1.51 2002/03/24 15:23:33 jongfoster
219 * Revision 1.50 2002/03/24 13:25:43 swa
220 * name change related issues
222 * Revision 1.49 2002/03/16 20:29:14 oes
225 * Revision 1.48 2002/03/13 20:25:34 oes
226 * Better logging for content filters
228 * Revision 1.47 2002/03/13 00:30:52 jongfoster
230 * Added option of always sending redirect for imageblock,
231 * currently disabled with #if 0.
233 * Revision 1.46 2002/03/12 01:42:49 oes
234 * Introduced modular filters
236 * Revision 1.45 2002/03/08 16:47:50 oes
237 * Added choice beween GIF and PNG built-in images
239 * Revision 1.44 2002/03/07 03:49:31 oes
240 * - Fixed compiler warnings etc
241 * - Changed built-in images from GIF to PNG
242 * (with regard to Unisys patent issue)
243 * - Added a 4x4 pattern PNG which is less intrusive
244 * than the logo but also clearly marks the deleted banners
246 * Revision 1.43 2002/01/22 23:51:59 jongfoster
247 * Replacing strsav() with the safer string_append().
249 * Adding missing html_encode() to error message generators. Where encoded
250 * and unencoded versions of a string were provided, removing the unencoded
253 * Revision 1.42 2002/01/17 21:00:32 jongfoster
254 * Moving all our URL and URL pattern parsing code to urlmatch.c.
256 * Using a single, simple url_match(pattern,url) function - rather than
257 * the 3-line match routine which was repeated all over the place.
259 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
261 * Using parse_http_url() to parse URLs without faking a HTTP
262 * request line for parse_http_request().
264 * Revision 1.41 2001/11/13 00:14:07 jongfoster
265 * Fixing stupid bug now I've figured out what || means.
266 * (It always returns 0 or 1, not one of it's paramaters.)
268 * Revision 1.40 2001/10/26 17:37:55 oes
269 * - Re-enabled Netscape 200/404 bug workaround in block_url():
270 * - Removed OS/2 special case
271 * - Made block_url() independant from sed() having been run
272 * - Made trust_url independant from sed() having been run
273 * - Made is_imageurl independant from sed() having been run.
274 * It now checks User-Agent: and Accept: by itself.
277 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
278 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
279 * threads to call select() simultaneously. So, it's time to do a real, live,
280 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
281 * (native). Both versions will work, but using __OS2__ offers multi-threading.
283 * Revision 1.38 2001/10/23 21:32:33 jongfoster
284 * Adding error-checking to selected functions
286 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
287 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
288 * filters.c. Added a FIXME in front of the offending code. I'll gladly
289 * put in a better/more robust fix for all parties if one is presented...
290 * It seems that just returning 200 instead of 404 would pretty much fix
291 * it for everyone, but I don't know all the history of the problem.
293 * Revision 1.36 2001/10/10 16:44:16 oes
294 * Added match_portlist function
296 * Revision 1.35 2001/10/07 15:41:23 oes
297 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
299 * New function remove_chunked_transfer_coding that strips chunked
300 * transfer coding to plain and is called by pcrs_filter_response
301 * and gif_deanimate_response if neccessary
303 * Improved handling of zero-change re_filter runs
305 * pcrs_filter_response and gif_deanimate_response now remove
306 * chunked transfer codeing before processing the body.
308 * Revision 1.34 2001/09/20 15:49:36 steudten
310 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
311 * See cgi.c fill_template().
313 * Revision 1.33 2001/09/16 17:05:14 jongfoster
314 * Removing unused #include showarg.h
316 * Revision 1.32 2001/09/16 13:21:27 jongfoster
317 * Changes to use new list functions.
319 * Revision 1.31 2001/09/16 11:38:02 jongfoster
320 * Splitting fill_template() into 2 functions:
321 * template_load() loads the file
322 * template_fill() performs the PCRS regexps.
323 * This is because the CGI edit interface has a "table row"
324 * template which is used many times in the page - this
325 * change means it's only loaded from disk once.
327 * Revision 1.30 2001/09/16 11:00:10 jongfoster
328 * New function alloc_http_response, for symmetry with free_http_response
330 * Revision 1.29 2001/09/13 23:32:40 jongfoster
331 * Moving image data to cgi.c rather than cgi.h
332 * Fixing a GPF under Win32 (and any other OS that protects global
333 * constants from being written to).
335 * Revision 1.28 2001/09/10 10:18:51 oes
336 * Silenced compiler warnings
338 * Revision 1.27 2001/08/05 16:06:20 jongfoster
339 * Modifiying "struct map" so that there are now separate header and
340 * "map_entry" structures. This means that functions which modify a
341 * map no longer need to return a pointer to the modified map.
342 * Also, it no longer reverses the order of the entries (which may be
343 * important with some advanced template substitutions).
345 * Revision 1.26 2001/07/30 22:08:36 jongfoster
346 * Tidying up #defines:
347 * - All feature #defines are now of the form FEATURE_xxx
348 * - Permanently turned off WIN_GUI_EDIT
349 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
351 * Revision 1.25 2001/07/26 10:09:46 oes
352 * Made browser detection a little less naive
354 * Revision 1.24 2001/07/25 17:22:51 oes
355 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
357 * Revision 1.23 2001/07/23 13:40:12 oes
358 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
360 * Revision 1.22 2001/07/18 12:29:34 oes
361 * - Made gif_deanimate_response respect
362 * csp->action->string[ACTION_STRING_DEANIMATE]
363 * - Logging cosmetics
365 * Revision 1.21 2001/07/13 13:59:53 oes
366 * - Introduced gif_deanimate_response which shares the
367 * generic content modification interface of pcrs_filter_response
368 * and acts as a wrapper to deanimate.c:gif_deanimate()
369 * - Renamed re_process_buffer to pcrs_filter_response
370 * - pcrs_filter_response now returns NULL on failiure
371 * - Removed all #ifdef PCRS
373 * Revision 1.20 2001/07/01 17:01:04 oes
374 * Added comments and missing return statement in is_untrusted_url()
376 * Revision 1.19 2001/06/29 21:45:41 oes
377 * Indentation, CRLF->LF, Tab-> Space
379 * Revision 1.18 2001/06/29 13:27:38 oes
380 * - Cleaned up, renamed and reorderd functions
381 * and improved comments
384 * - Ported to CGI platform. Now delivers
385 * http_response or NULL
386 * - Unified HTML and GIF generation (moved image detection
387 * and GIF generation here from jcc.c:chat())
388 * - Fixed HTTP status to:
389 * - 403 (Forbidden) for the "blocked" HTML message
390 * - 200 (OK) for GIF answers
391 * - 302 (Redirect) for redirect to GIF
394 * - Ported to CGI platform. Now delivers
395 * http_response or NULL
396 * - Separated detection of untrusted URL into
397 * (bool)is_untrusted_url
398 * - Added enforcement of untrusted requests
400 * - Moved redirect_url() from cgi.c to here
401 * and ported it to the CGI platform
403 * - Removed logentry from cancelled commit
405 * Revision 1.17 2001/06/09 10:55:28 jongfoster
406 * Changing BUFSIZ ==> BUFFER_SIZE
408 * Revision 1.16 2001/06/07 23:10:26 jongfoster
409 * Allowing unanchored domain patterns to back off and retry
410 * if they partially match. Optimized right-anchored patterns.
411 * Moving ACL and forward files into config file.
412 * Replacing struct gateway with struct forward_spec
414 * Revision 1.15 2001/06/03 19:12:00 oes
415 * extracted-CGI relevant stuff
417 * Revision 1.14 2001/06/01 10:30:55 oes
418 * Added optional left-anchoring to domaincmp
420 * Revision 1.13 2001/05/31 21:21:30 jongfoster
421 * Permissionsfile / actions file changes:
422 * - Changed "permission" to "action" throughout
423 * - changes to file format to allow string parameters
424 * - Moved helper functions to actions.c
426 * Revision 1.12 2001/05/31 17:35:20 oes
428 * - Enhanced domain part globbing with infix and prefix asterisk
429 * matching and optional unanchored operation
431 * Revision 1.11 2001/05/29 11:53:23 oes
432 * "See why" link added to "blocked" page
434 * Revision 1.10 2001/05/29 09:50:24 jongfoster
435 * Unified blocklist/imagelist/permissionslist.
436 * File format is still under discussion, but the internal changes
439 * Also modified interceptor behaviour:
440 * - We now intercept all URLs beginning with one of the following
441 * prefixes (and *only* these prefixes):
443 * * http://ijbswa.sf.net/config/
444 * * http://ijbswa.sourceforge.net/config/
445 * - New interceptors "home page" - go to http://i.j.b/ to see it.
446 * - Internal changes so that intercepted and fast redirect pages
447 * are not replaced with an image.
448 * - Interceptors now have the option to send a binary page direct
449 * to the client. (i.e. ijb-send-banner uses this)
450 * - Implemented show-url-info interceptor. (Which is why I needed
451 * the above interceptors changes - a typical URL is
452 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
453 * The previous mechanism would not have intercepted that, and
454 * if it had been intercepted then it then it would have replaced
457 * Revision 1.9 2001/05/27 22:17:04 oes
459 * - re_process_buffer no longer writes the modified buffer
460 * to the client, which was very ugly. It now returns the
461 * buffer, which it is then written by chat.
463 * - content_length now adjusts the Content-Length: header
464 * for modified documents rather than crunch()ing it.
465 * (Length info in csp->content_length, which is 0 for
466 * unmodified documents)
468 * - For this to work, sed() is called twice when filtering.
470 * Revision 1.8 2001/05/26 17:13:28 jongfoster
471 * Filled in a function comment.
473 * Revision 1.7 2001/05/26 15:26:15 jongfoster
474 * ACL feature now provides more security by immediately dropping
475 * connections from untrusted hosts.
477 * Revision 1.6 2001/05/26 00:28:36 jongfoster
478 * Automatic reloading of config file.
479 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
480 * Most of the global variables have been moved to a new
481 * struct configuration_spec, accessed through csp->config->globalname
482 * Most of the globals remaining are used by the Win32 GUI.
484 * Revision 1.5 2001/05/25 22:34:30 jongfoster
487 * Revision 1.4 2001/05/22 18:46:04 oes
489 * - Enabled filtering banners by size rather than URL
490 * by adding patterns that replace all standard banner
491 * sizes with the "Junkbuster" gif to the re_filterfile
493 * - Enabled filtering WebBugs by providing a pattern
494 * which kills all 1x1 images
496 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
497 * which is selected by the (nonstandard and therefore
498 * capital) letter 'U' in the option string.
499 * It causes the quantifiers to be ungreedy by default.
500 * Appending a ? turns back to greedy (!).
502 * - Added a new interceptor ijb-send-banner, which
503 * sends back the "Junkbuster" gif. Without imagelist or
504 * MSIE detection support, or if tinygif = 1, or the
505 * URL isn't recognized as an imageurl, a lame HTML
506 * explanation is sent instead.
508 * - Added new feature, which permits blocking remote
509 * script redirects and firing back a local redirect
511 * The feature is conditionally compiled, i.e. it
512 * can be disabled with --disable-fast-redirects,
513 * plus it must be activated by a "fast-redirects"
514 * line in the config file, has its own log level
515 * and of course wants to be displayed by show-proxy-args
516 * Note: Boy, all the #ifdefs in 1001 locations and
517 * all the fumbling with configure.in and acconfig.h
518 * were *way* more work than the feature itself :-(
520 * - Because a generic redirect template was needed for
521 * this, tinygif = 3 now uses the same.
523 * - Moved GIFs, and other static HTTP response templates
528 * - Removed some >400 CRs again (Jon, you really worked
531 * Revision 1.3 2001/05/20 16:44:47 jongfoster
532 * Removing last hardcoded Junkbusters.com URLs.
534 * Revision 1.2 2001/05/20 01:21:20 jongfoster
535 * Version 2.9.4 checkin.
536 * - Merged popupfile and cookiefile, and added control over PCRS
537 * filtering, in new "permissionsfile".
538 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
539 * file error you now get a message box (in the Win32 GUI) rather
540 * than the program exiting with no explanation.
541 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
543 * - Removed tabs from "config"
544 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
545 * - Bumped up version number.
547 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
548 * Initial import of version 2.9.3 source tree
551 *********************************************************************/
557 #include <sys/types.h>
566 #endif /* ndef __OS2__ */
567 #include <netinet/in.h>
569 #include <winsock2.h>
570 #endif /* ndef _WIN32 */
574 #endif /* def __OS2__ */
582 #include "jbsockets.h"
583 #include "miscutil.h"
587 #include "deanimate.h"
588 #include "urlmatch.h"
594 const char filters_h_rcs[] = FILTERS_H_VERSION;
596 /* Fix a problem with Solaris. There should be no effect on other
598 * Solaris's isspace() is a macro which uses it's argument directly
599 * as an array index. Therefore we need to make sure that high-bit
600 * characters generate +ve values, and ideally we also want to make
601 * the argument match the declared parameter type of "int".
603 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
607 /*********************************************************************
609 * Function : block_acl
611 * Description : Block this request?
612 * Decide yes or no based on ACL file.
615 * 1 : dst = The proxy or gateway address this is going to.
616 * Or NULL to check all possible targets.
617 * 2 : csp = Current client state (buffers, headers, etc...)
618 * Also includes the client IP address.
620 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
622 *********************************************************************/
623 int block_acl(struct access_control_addr *dst, struct client_state *csp)
625 struct access_control_list *acl = csp->config->acl;
627 /* if not using an access control list, then permit the connection */
633 /* search the list */
636 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
640 /* Just want to check if they have any access */
641 if (acl->action == ACL_PERMIT)
646 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
647 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
649 if (acl->action == ACL_PERMIT)
667 /*********************************************************************
669 * Function : acl_addr
671 * Description : Called from `load_config' to parse an ACL address.
674 * 1 : aspec = String specifying ACL address.
675 * 2 : aca = struct access_control_addr to fill in.
677 * Returns : 0 => Ok, everything else is an error.
679 *********************************************************************/
680 int acl_addr(const char *aspec, struct access_control_addr *aca)
685 char *acl_spec = NULL;
691 * Use a temporary acl spec copy so we can log
692 * the unmodified original in case of parse errors.
694 acl_spec = strdup(aspec);
695 if (acl_spec == NULL)
697 /* XXX: This will be logged as parse error. */
701 if ((p = strchr(acl_spec, '/')) != NULL)
704 if (ijb_isdigit(*p) == 0)
709 masklength = atoi(p);
712 if ((masklength < 0) || (masklength > 32))
718 if ((p = strchr(acl_spec, ':')) != NULL)
723 port = strtol(p, &endptr, 10);
725 if (port <= 0 || port > 65535 || *endptr != '\0')
732 aca->port = (unsigned long)port;
734 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
737 if (aca->addr == INADDR_NONE)
739 /* XXX: This will be logged as parse error. */
743 /* build the netmask */
745 for (i=1; i <= masklength ; i++)
747 aca->mask |= (1 << (32 - i));
750 /* now mask off the host portion of the ip address
751 * (i.e. save on the network portion of the address).
753 aca->addr = aca->addr & aca->mask;
758 #endif /* def FEATURE_ACL */
761 /*********************************************************************
763 * Function : match_portlist
765 * Description : Check if a given number is covered by a comma
766 * separated list of numbers and ranges (a,b-c,d,..)
769 * 1 : portlist = String with list
770 * 2 : port = port to check
772 * Returns : 0 => no match
775 *********************************************************************/
776 int match_portlist(const char *portlist, int port)
778 char *min, *max, *next, *portlist_copy;
780 min = next = portlist_copy = strdup(portlist);
783 * Zero-terminate first item and remember offset for next
785 if (NULL != (next = strchr(portlist_copy, (int) ',')))
791 * Loop through all items, checking for match
795 if (NULL == (max = strchr(min, (int) '-')))
798 * No dash, check for equality
800 if (port == atoi(min))
809 * This is a range, so check if between min and max,
810 * or, if max was omitted, between min and 65K
813 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
827 * Zero-terminate next item and remember offset for n+1
829 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
841 /*********************************************************************
843 * Function : block_url
845 * Description : Called from `chat'. Check to see if we need to block this.
848 * 1 : csp = Current client state (buffers, headers, etc...)
850 * Returns : NULL => unblocked, else HTTP block response
852 *********************************************************************/
853 struct http_response *block_url(struct client_state *csp)
855 struct http_response *rsp;
856 const char *new_content_type = NULL;
859 * If it's not blocked, don't block it ;-)
861 if ((csp->action->flags & ACTION_BLOCK) == 0)
865 if (csp->action->flags & ACTION_REDIRECT)
867 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
870 * Else, prepare a response
872 if (NULL == (rsp = alloc_http_response()))
874 return cgi_error_memory();
878 * If it's an image-url, send back an image or redirect
879 * as specified by the relevant +image action
881 #ifdef FEATURE_IMAGE_BLOCKING
882 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
886 /* determine HOW images should be blocked */
887 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
889 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
891 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
893 #if 1 /* Two alternative strategies, use this one for now: */
895 /* and handle accordingly: */
896 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
898 rsp->status = strdup("403 Request blocked by Privoxy");
899 if (rsp->status == NULL)
901 free_http_response(rsp);
902 return cgi_error_memory();
904 rsp->body = bindup(image_pattern_data, image_pattern_length);
905 if (rsp->body == NULL)
907 free_http_response(rsp);
908 return cgi_error_memory();
910 rsp->content_length = image_pattern_length;
912 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
914 free_http_response(rsp);
915 return cgi_error_memory();
919 else if (0 == strcmpic(p, "blank"))
921 rsp->status = strdup("403 Request blocked by Privoxy");
922 if (rsp->status == NULL)
924 free_http_response(rsp);
925 return cgi_error_memory();
927 rsp->body = bindup(image_blank_data, image_blank_length);
928 if (rsp->body == NULL)
930 free_http_response(rsp);
931 return cgi_error_memory();
933 rsp->content_length = image_blank_length;
935 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
937 free_http_response(rsp);
938 return cgi_error_memory();
944 rsp->status = strdup("302 Local Redirect from Privoxy");
945 if (rsp->status == NULL)
947 free_http_response(rsp);
948 return cgi_error_memory();
951 if (enlist_unique_header(rsp->headers, "Location", p))
953 free_http_response(rsp);
954 return cgi_error_memory();
958 #else /* Following code is disabled for now */
960 /* and handle accordingly: */
961 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
963 p = CGI_PREFIX "send-banner?type=pattern";
965 else if (0 == strcmpic(p, "blank"))
967 p = CGI_PREFIX "send-banner?type=blank";
969 rsp->status = strdup("302 Local Redirect from Privoxy");
970 if (rsp->status == NULL)
972 free_http_response(rsp);
973 return cgi_error_memory();
976 if (enlist_unique_header(rsp->headers, "Location", p))
978 free_http_response(rsp);
979 return cgi_error_memory();
981 #endif /* Preceeding code is disabled for now */
983 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
986 * Send empty document.
988 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
991 rsp->body = strdup(" ");
992 rsp->content_length = 1;
994 rsp->status = strdup("403 Request blocked by Privoxy");
995 if (rsp->status == NULL)
997 free_http_response(rsp);
998 return cgi_error_memory();
1000 if (new_content_type != 0)
1002 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
1003 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
1005 free_http_response(rsp);
1006 return cgi_error_memory();
1011 #endif /* def FEATURE_IMAGE_BLOCKING */
1014 * Else, generate an HTML "blocked" message:
1018 struct map * exports;
1022 * Workaround for stupid Netscape bug which prevents
1023 * pages from being displayed if loading a referenced
1024 * JavaScript or style sheet fails. So make it appear
1025 * as if it succeeded.
1027 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1028 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1029 && !strstr(p, "Gecko") /* save Mozilla, */
1030 && !strstr(p, "compatible") /* MSIE */
1031 && !strstr(p, "Opera")) /* and Opera. */
1033 rsp->status = strdup("200 Request for blocked URL");
1037 rsp->status = strdup("403 Request for blocked URL");
1040 if (rsp->status == NULL)
1042 free_http_response(rsp);
1043 return cgi_error_memory();
1046 exports = default_exports(csp, NULL);
1047 if (exports == NULL)
1049 free_http_response(rsp);
1050 return cgi_error_memory();
1053 #ifdef FEATURE_FORCE_LOAD
1054 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1056 * Export the force conditional block killer if
1058 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1059 * - Privoxy is configured to enforce blocks, or
1060 * - it's a CONNECT request and enforcing wouldn't work anyway.
1062 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1063 || (0 == strcmpic(csp->http->gpc, "connect")))
1064 #endif /* ndef FEATURE_FORCE_LOAD */
1066 err = map_block_killer(exports, "force-support");
1069 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1070 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1071 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1072 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1077 free_http_response(rsp);
1078 return cgi_error_memory();
1081 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1084 free_http_response(rsp);
1085 return cgi_error_memory();
1088 rsp->reason = RSP_REASON_BLOCKED;
1090 return finish_http_response(csp, rsp);
1095 #ifdef FEATURE_TRUST
1096 /*********************************************************************
1098 * Function : trust_url FIXME: I should be called distrust_url
1100 * Description : Calls is_untrusted_url to determine if the URL is trusted
1101 * and if not, returns a HTTP 403 response with a reject message.
1104 * 1 : csp = Current client state (buffers, headers, etc...)
1106 * Returns : NULL => trusted, else http_response.
1108 *********************************************************************/
1109 struct http_response *trust_url(struct client_state *csp)
1111 struct http_response *rsp;
1112 struct map * exports;
1113 char buf[BUFFER_SIZE];
1115 struct url_spec **tl;
1120 * Don't bother to work on trusted URLs
1122 if (!is_untrusted_url(csp))
1128 * Else, prepare a response:
1130 if (NULL == (rsp = alloc_http_response()))
1132 return cgi_error_memory();
1135 rsp->status = strdup("403 Request blocked by Privoxy");
1136 exports = default_exports(csp, NULL);
1137 if (exports == NULL || rsp->status == NULL)
1139 free_http_response(rsp);
1140 return cgi_error_memory();
1144 * Export the protocol, host, port, and referrer information
1146 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1147 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1148 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1150 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1152 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1156 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1162 free_http_response(rsp);
1163 return cgi_error_memory();
1167 * Export the trust list
1170 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1172 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1173 string_append(&p, buf);
1175 err = map(exports, "trusted-referrers", 1, p, 0);
1180 free_http_response(rsp);
1181 return cgi_error_memory();
1185 * Export the trust info, if available
1187 if (csp->config->trust_info->first)
1189 struct list_entry *l;
1192 for (l = csp->config->trust_info->first; l ; l = l->next)
1194 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1195 string_append(&p, buf);
1197 err = map(exports, "trust-info", 1, p, 0);
1201 err = map_block_killer(exports, "have-trust-info");
1207 free_http_response(rsp);
1208 return cgi_error_memory();
1212 * Export the force conditional block killer if
1214 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1215 * - Privoxy is configured to enforce blocks, or
1216 * - it's a CONNECT request and enforcing wouldn't work anyway.
1218 #ifdef FEATURE_FORCE_LOAD
1219 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1220 || (0 == strcmpic(csp->http->gpc, "connect")))
1222 err = map_block_killer(exports, "force-support");
1226 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1228 #else /* ifndef FEATURE_FORCE_LOAD */
1229 err = map_block_killer(exports, "force-support");
1230 #endif /* ndef FEATURE_FORCE_LOAD */
1235 free_http_response(rsp);
1236 return cgi_error_memory();
1240 * Build the response
1242 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1245 free_http_response(rsp);
1246 return cgi_error_memory();
1248 rsp->reason = RSP_REASON_UNTRUSTED;
1250 return finish_http_response(csp, rsp);
1252 #endif /* def FEATURE_TRUST */
1255 /*********************************************************************
1257 * Function : compile_dynamic_pcrs_job_list
1259 * Description : Compiles a dynamic pcrs job list (one with variables
1260 * resolved at request time)
1263 * 1 : csp = Current client state (buffers, headers, etc...)
1264 * 2 : b = The filter list to compile
1266 * Returns : NULL in case of errors, otherwise the
1269 *********************************************************************/
1270 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
1272 struct list_entry *pattern;
1273 pcrs_job *job_list = NULL;
1274 pcrs_job *dummy = NULL;
1275 pcrs_job *lastjob = NULL;
1278 const struct pcrs_variable variables[] =
1280 {"url", csp->http->url, 1},
1281 {"path", csp->http->path, 1},
1282 {"host", csp->http->host, 1},
1283 {"origin", csp->ip_addr_str, 1},
1287 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
1289 assert(pattern->str != NULL);
1291 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
1295 log_error(LOG_LEVEL_ERROR,
1296 "Adding filter job \'%s\' to dynamic filter %s failed: %s",
1297 pattern->str, b->name, pcrs_strerror(error));
1302 if (error == PCRS_WARN_TRUNCATION)
1304 log_error(LOG_LEVEL_ERROR,
1305 "At least one of the variables in \'%s\' had to "
1306 "be truncated before compilation", pattern->str);
1308 if (job_list == NULL)
1314 lastjob->next = dummy;
1324 /*********************************************************************
1326 * Function : rewrite_url
1328 * Description : Rewrites a URL with a single pcrs command
1329 * and returns the result if it differs from the
1330 * original and isn't obviously invalid.
1333 * 1 : old_url = URL to rewrite.
1334 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1337 * Returns : NULL if the pcrs_command didn't change the url, or
1338 * the result of the modification.
1340 *********************************************************************/
1341 char *rewrite_url(char *old_url, const char *pcrs_command)
1343 char *new_url = NULL;
1347 assert(pcrs_command);
1349 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1353 log_error(LOG_LEVEL_REDIRECTS,
1354 "pcrs command \"%s\" didn't change \"%s\".",
1355 pcrs_command, old_url);
1360 log_error(LOG_LEVEL_REDIRECTS,
1361 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1362 pcrs_command, old_url, pcrs_strerror(hits));
1365 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1367 log_error(LOG_LEVEL_ERROR,
1368 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1369 "but the result doesn't look like a valid URL and will be ignored.",
1370 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1375 log_error(LOG_LEVEL_REDIRECTS,
1376 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1377 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1385 #ifdef FEATURE_FAST_REDIRECTS
1386 /*********************************************************************
1388 * Function : get_last_url
1390 * Description : Search for the last URL inside a string.
1391 * If the string already is a URL, it will
1392 * be the first URL found.
1395 * 1 : subject = the string to check
1396 * 2 : redirect_mode = +fast-redirect{} mode
1398 * Returns : NULL if no URL was found, or
1399 * the last URL found.
1401 *********************************************************************/
1402 char *get_last_url(char *subject, const char *redirect_mode)
1404 char *new_url = NULL;
1408 assert(redirect_mode);
1410 subject = strdup(subject);
1411 if (subject == NULL)
1413 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1417 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1419 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1420 new_url = url_decode(subject);
1421 if (new_url != NULL)
1428 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1432 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1435 * Find the last URL encoded in the request
1438 while ((tmp = strstr(tmp, "http://")) != NULL)
1442 tmp = (new_url != NULL) ? new_url : subject;
1443 while ((tmp = strstr(tmp, "https://")) != NULL)
1448 if ((new_url != NULL)
1449 && ( (new_url != subject)
1450 || (0 == strncmpic(subject, "http://", 7))
1451 || (0 == strncmpic(subject, "https://", 8))
1455 * Return new URL if we found a redirect
1456 * or if the subject already was a URL.
1458 * The second case makes sure that we can
1459 * chain get_last_url after another redirection check
1460 * (like rewrite_url) without losing earlier redirects.
1462 new_url = strdup(new_url);
1471 #endif /* def FEATURE_FAST_REDIRECTS */
1474 /*********************************************************************
1476 * Function : redirect_url
1478 * Description : Checks if Privoxy should answer the request with
1479 * a HTTP redirect and generates the redirect if
1483 * 1 : csp = Current client state (buffers, headers, etc...)
1485 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1487 *********************************************************************/
1488 struct http_response *redirect_url(struct client_state *csp)
1490 struct http_response *rsp;
1491 #ifdef FEATURE_FAST_REDIRECTS
1493 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1494 * as compile-time option? The user can easily disable
1495 * it in his action file.
1497 char * redirect_mode;
1498 #endif /* def FEATURE_FAST_REDIRECTS */
1499 char *old_url = NULL;
1500 char *new_url = NULL;
1501 char *redirection_string;
1503 if ((csp->action->flags & ACTION_REDIRECT))
1505 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1508 * If the redirection string begins with 's',
1509 * assume it's a pcrs command, otherwise treat it as
1510 * properly formatted URL and use it for the redirection
1513 * According to RFC 2616 section 14.30 the URL
1514 * has to be absolute and if the user tries:
1515 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1516 * she would get undefined results anyway.
1520 if (*redirection_string == 's')
1522 old_url = csp->http->url;
1523 new_url = rewrite_url(old_url, redirection_string);
1527 log_error(LOG_LEVEL_REDIRECTS,
1528 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1529 redirection_string);
1530 new_url = strdup(redirection_string);
1534 #ifdef FEATURE_FAST_REDIRECTS
1535 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1537 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1540 * If it exists, use the previously rewritten URL as input
1541 * otherwise just use the old path.
1543 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1544 new_url = get_last_url(old_url, redirect_mode);
1549 * Disable redirect checkers, so that they
1550 * will be only run more than once if the user
1551 * also enables them through tags.
1553 * From a performance point of view
1554 * it doesn't matter, but the duplicated
1555 * log messages are annoying.
1557 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1558 #endif /* def FEATURE_FAST_REDIRECTS */
1559 csp->action->flags &= ~ACTION_REDIRECT;
1561 /* Did any redirect action trigger? */
1564 if (0 == strcmpic(new_url, csp->http->url))
1566 log_error(LOG_LEVEL_ERROR,
1567 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1568 csp->http->url, new_url);
1573 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1575 if (NULL == (rsp = alloc_http_response()))
1578 return cgi_error_memory();
1581 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1582 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1585 free_http_response(rsp);
1586 return cgi_error_memory();
1588 rsp->reason = RSP_REASON_REDIRECTED;
1591 return finish_http_response(csp, rsp);
1595 /* Only reached if no redirect is required */
1601 #ifdef FEATURE_IMAGE_BLOCKING
1602 /*********************************************************************
1604 * Function : is_imageurl
1606 * Description : Given a URL, decide whether it is an image or not,
1607 * using either the info from a previous +image action
1608 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1609 * is MSIE and not on a Mac, tell from the browser's accept
1613 * 1 : csp = Current client state (buffers, headers, etc...)
1615 * Returns : True (nonzero) if URL is an image, false (0)
1618 *********************************************************************/
1619 int is_imageurl(const struct client_state *csp)
1621 #ifdef FEATURE_IMAGE_DETECT_MSIE
1624 tmp = get_header_value(csp->headers, "User-Agent:");
1625 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1627 tmp = get_header_value(csp->headers, "Accept:");
1628 if (tmp && strstr(tmp, "image/gif"))
1630 /* Client will accept HTML. If this seems counterintuitive,
1640 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1642 return ((csp->action->flags & ACTION_IMAGE) != 0);
1645 #endif /* def FEATURE_IMAGE_BLOCKING */
1648 #ifdef FEATURE_TRUST
1649 /*********************************************************************
1651 * Function : is_untrusted_url
1653 * Description : Should we "distrust" this URL (and block it)?
1655 * Yes if it matches a line in the trustfile, or if the
1656 * referrer matches a line starting with "+" in the
1661 * 1 : csp = Current client state (buffers, headers, etc...)
1663 * Returns : 0 => trusted, 1 => untrusted
1665 *********************************************************************/
1666 int is_untrusted_url(const struct client_state *csp)
1668 struct file_list *fl;
1669 struct block_spec *b;
1670 struct url_spec **trusted_url;
1671 struct http_request rhttp[1];
1672 const char * referer;
1676 * If we don't have a trustlist, we trust everybody
1678 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1683 memset(rhttp, '\0', sizeof(*rhttp));
1686 * Do we trust the request URL itself?
1688 for (b = b->next; b ; b = b->next)
1690 if (url_match(b->url, csp->http))
1696 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1698 /* no referrer was supplied */
1704 * If not, do we maybe trust its referrer?
1706 err = parse_http_url(referer, rhttp, csp);
1712 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1714 if (url_match(*trusted_url, rhttp))
1716 /* if the URL's referrer is from a trusted referrer, then
1717 * add the target spec to the trustfile as an unblocked
1718 * domain and return 0 (which means it's OK).
1723 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1727 char * new_entry = strdup("~");
1729 string_append(&new_entry, csp->http->hostport);
1731 path = csp->http->path;
1732 if ( (path[0] == '/')
1734 && ((path_end = strchr(path + 2, '/')) != NULL))
1736 /* since this path points into a user's home space
1737 * be sure to include this spec in the trustfile.
1739 int path_len = path_end - path; /* save offset */
1740 path = strdup(path); /* Copy string */
1743 path_end = path + path_len; /* regenerate ptr to new buffer */
1744 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1746 string_join(&new_entry, path);
1750 * Give a reason for generating this entry.
1752 string_append(&new_entry, " # Trusted referrer was: ");
1753 string_append(&new_entry, referer);
1755 if (new_entry != NULL)
1757 if (-1 == fprintf(fp, "%s\n", new_entry))
1759 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1760 new_entry, csp->config->trustfile);
1766 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1767 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1774 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1775 csp->http->hostport, csp->config->trustfile);
1783 #endif /* def FEATURE_TRUST */
1786 /*********************************************************************
1788 * Function : pcrs_filter_response
1790 * Description : Execute all text substitutions from all applying
1791 * +filter actions on the text buffer that's been accumulated
1792 * in csp->iob->buf. If this changes the contents, set
1793 * csp->content_length to the modified size and raise the
1794 * CSP_FLAG_MODIFIED flag.
1796 * XXX: Currently pcrs_filter_response is also responsible
1797 * for dechunking and decompressing. Both should be
1798 * done in separate functions so other content modifiers
1799 * profit as well, even if pcrs filtering is disabled.
1802 * 1 : csp = Current client state (buffers, headers, etc...)
1804 * Returns : a pointer to the (newly allocated) modified buffer.
1805 * or NULL if there were no hits or something went wrong
1807 *********************************************************************/
1808 char *pcrs_filter_response(struct client_state *csp)
1811 size_t size, prev_size;
1813 char *old = csp->iob->cur, *new = NULL;
1816 struct file_list *fl;
1817 struct re_filterfile_spec *b;
1818 struct list_entry *filtername;
1820 int i, found_filters = 0;
1825 if (csp->iob->cur >= csp->iob->eod)
1829 size = (size_t)(csp->iob->eod - csp->iob->cur);
1832 * Need to check the set of re_filterfiles...
1834 for (i = 0; i < MAX_AF_FILES; i++)
1847 if (0 == found_filters)
1849 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1854 * If the body has a "chunked" transfer-encoding,
1855 * get rid of it first, adjusting size and iob->eod
1857 if (csp->flags & CSP_FLAG_CHUNKED)
1859 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1860 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1864 csp->iob->eod = csp->iob->cur + size;
1865 csp->flags |= CSP_FLAG_MODIFIED;
1870 * If the body has a compressed transfer-encoding,
1871 * uncompress it first, adjusting size and iob->eod.
1872 * Note that decompression occurs after de-chunking.
1874 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1876 /* Notice that we at least tried to decompress. */
1877 if (JB_ERR_OK != decompress_iob(csp))
1880 * We failed to decompress the data; there's no point
1881 * in continuing since we can't filter.
1883 * XXX: Actually the Accept-Encoding header may
1884 * just be incorrect in which case we could continue
1887 * Unset CT_GZIP and CT_DEFLATE to remember not
1888 * to modify the Content-Encoding header later.
1890 csp->content_type &= ~CT_GZIP;
1891 csp->content_type &= ~CT_DEFLATE;
1896 * Decompression gives us a completely new iob,
1897 * so we need to update.
1899 size = (size_t)(csp->iob->eod - csp->iob->cur);
1900 old = csp->iob->cur;
1902 csp->flags |= CSP_FLAG_MODIFIED;
1903 csp->content_type &= ~CT_TABOO;
1907 for (i = 0; i < MAX_AF_FILES; i++)
1910 if ((NULL == fl) || (NULL == fl->f))
1913 * Either there are no filter files
1914 * left, or this filter file just
1915 * contains no valid filters.
1917 * Continue to be sure we don't miss
1918 * valid filter files that are chained
1919 * after empty or invalid ones.
1924 * For all applying +filter actions, look if a filter by that
1925 * name exists and if yes, execute it's pcrs_joblist on the
1928 for (b = fl->f; b; b = b->next)
1930 if (b->type != FT_CONTENT_FILTER)
1932 /* Skip header filters */
1936 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1937 filtername ; filtername = filtername->next)
1939 if (strcmp(b->name, filtername->str) == 0)
1941 int current_hits = 0; /* Number of hits caused by this filter */
1942 int job_number = 0; /* Which job we're currently executing */
1943 int job_hits = 0; /* How many hits the current job caused */
1944 pcrs_job *joblist = b->joblist;
1946 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1948 if (NULL == joblist)
1950 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1955 /* Apply all jobs from the joblist */
1956 for (job = joblist; NULL != job; job = job->next)
1959 job_hits = pcrs_execute(job, old, size, &new, &size);
1964 * That went well. Continue filtering
1965 * and use the result of this job as
1966 * input for the next one.
1968 current_hits += job_hits;
1969 if (old != csp->iob->cur)
1978 * This job caused an unexpected error. Inform the user
1979 * and skip the rest of the jobs in this filter. We could
1980 * continue with the next job, but usually the jobs
1981 * depend on each other or are similar enough to
1982 * fail for the same reason.
1984 * At the moment our pcrs expects the error codes of pcre 3.4,
1985 * but newer pcre versions can return additional error codes.
1986 * As a result pcrs_strerror()'s error message might be
1987 * "Unknown error ...", therefore we print the numerical value
1990 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1991 * should we use LOG_LEVEL_RE_FILTER instead?
1993 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1994 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1999 if (b->dynamic) pcrs_free_joblist(joblist);
2001 log_error(LOG_LEVEL_RE_FILTER,
2002 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
2003 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
2005 hits += current_hits;
2012 * If there were no hits, destroy our copy and let
2013 * chat() use the original in csp->iob
2021 csp->flags |= CSP_FLAG_MODIFIED;
2022 csp->content_length = size;
2030 /*********************************************************************
2032 * Function : gif_deanimate_response
2034 * Description : Deanimate the GIF image that has been accumulated in
2035 * csp->iob->buf, set csp->content_length to the modified
2036 * size and raise the CSP_FLAG_MODIFIED flag.
2039 * 1 : csp = Current client state (buffers, headers, etc...)
2041 * Returns : a pointer to the (newly allocated) modified buffer.
2042 * or NULL in case something went wrong.
2044 *********************************************************************/
2045 char *gif_deanimate_response(struct client_state *csp)
2047 struct binbuffer *in, *out;
2051 size = (size_t)(csp->iob->eod - csp->iob->cur);
2054 * If the body has a "chunked" transfer-encoding,
2055 * get rid of it first, adjusting size and iob->eod
2057 if (csp->flags & CSP_FLAG_CHUNKED)
2059 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2060 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2064 csp->iob->eod = csp->iob->cur + size;
2065 csp->flags |= CSP_FLAG_MODIFIED;
2068 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2069 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2071 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2075 in->buffer = csp->iob->cur;
2078 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2080 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2087 if ((int)size == out->offset)
2089 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2093 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2095 csp->content_length = out->offset;
2096 csp->flags |= CSP_FLAG_MODIFIED;
2106 /*********************************************************************
2108 * Function : jpeg_inspect_response
2113 * 1 : csp = Current client state (buffers, headers, etc...)
2115 * Returns : a pointer to the (newly allocated) modified buffer
2116 * or NULL in case something went wrong.
2118 *********************************************************************/
2119 char *jpeg_inspect_response(struct client_state *csp)
2121 struct binbuffer *in = NULL;
2122 struct binbuffer *out = NULL;
2126 size = (size_t)(csp->iob->eod - csp->iob->cur);
2129 * If the body has a "chunked" transfer-encoding,
2130 * get rid of it first, adjusting size and iob->eod
2132 if (csp->flags & CSP_FLAG_CHUNKED)
2134 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2135 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2139 csp->iob->eod = csp->iob->cur + size;
2140 csp->flags |= CSP_FLAG_MODIFIED;
2143 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2145 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2149 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2151 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2155 in->buffer = csp->iob->cur;
2159 * Calling jpeg_inspect has the side-effect of creating and
2160 * modifying the image buffer of "out" directly.
2162 if (jpeg_inspect(in, out))
2164 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2172 csp->content_length = out->offset;
2173 csp->flags |= CSP_FLAG_MODIFIED;
2183 /*********************************************************************
2185 * Function : remove_chunked_transfer_coding
2187 * Description : In-situ remove the "chunked" transfer coding as defined
2188 * in rfc2616 from a buffer.
2191 * 1 : buffer = Pointer to the text buffer
2192 * 2 : size = Number of bytes to be processed
2194 * Returns : The new size, i.e. the number of bytes from buffer which
2195 * are occupied by the stripped body, or 0 in case something
2198 *********************************************************************/
2199 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2202 unsigned int chunksize = 0;
2203 char *from_p, *to_p;
2206 from_p = to_p = buffer;
2208 if (sscanf(buffer, "%x", &chunksize) != 1)
2210 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2214 while (chunksize > 0)
2216 if (NULL == (from_p = strstr(from_p, "\r\n")))
2218 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2222 if ((newsize += chunksize) >= size)
2224 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2229 memmove(to_p, from_p, (size_t) chunksize);
2230 to_p = buffer + newsize;
2231 from_p += chunksize + 2;
2233 if (sscanf(from_p, "%x", &chunksize) != 1)
2235 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2240 /* FIXME: Should this get its own loglevel? */
2241 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2247 /*********************************************************************
2249 * Function : url_actions
2251 * Description : Gets the actions for this URL.
2254 * 1 : http = http_request request for blocked URLs
2255 * 2 : csp = Current client state (buffers, headers, etc...)
2259 *********************************************************************/
2260 void url_actions(struct http_request *http,
2261 struct client_state *csp)
2263 struct file_list *fl;
2264 struct url_actions *b;
2267 init_current_action(csp->action);
2269 for (i = 0; i < MAX_AF_FILES; i++)
2271 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2276 apply_url_actions(csp->action, http, b);
2283 /*********************************************************************
2285 * Function : apply_url_actions
2287 * Description : Applies a list of URL actions.
2290 * 1 : action = Destination.
2291 * 2 : http = Current URL
2292 * 3 : b = list of URL actions to apply
2296 *********************************************************************/
2297 void apply_url_actions(struct current_action_spec *action,
2298 struct http_request *http,
2299 struct url_actions *b)
2303 /* Should never happen */
2307 for (b = b->next; NULL != b; b = b->next)
2309 if (url_match(b->url, http))
2311 merge_current_action(action, b->action);
2317 /*********************************************************************
2319 * Function : forward_url
2321 * Description : Should we forward this to another proxy?
2324 * 1 : http = http_request request for current URL
2325 * 2 : csp = Current client state (buffers, headers, etc...)
2327 * Returns : Pointer to forwarding information.
2329 *********************************************************************/
2330 const struct forward_spec * forward_url(struct http_request *http,
2331 struct client_state *csp)
2333 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2334 struct forward_spec *fwd = csp->config->forward;
2343 if (url_match(fwd->url, http))
2354 /*********************************************************************
2356 * Function : direct_response
2358 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2359 * request and if so, return a HTTP 501 to the client.
2361 * FIXME: I have a stupid name and I should handle the
2362 * requests properly. Still, what we do here is rfc-
2363 * compliant, whereas ignoring or forwarding are not.
2366 * 1 : csp = Current client state (buffers, headers, etc...)
2368 * Returns : http_response if , NULL if nonmatch or handler fail
2370 *********************************************************************/
2371 struct http_response *direct_response(struct client_state *csp)
2373 struct http_response *rsp;
2374 struct list_entry *p;
2376 if ((0 == strcmpic(csp->http->gpc, "trace"))
2377 || (0 == strcmpic(csp->http->gpc, "options")))
2379 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2381 if (!strncmpic("Max-Forwards:", p->str, 13))
2383 unsigned int max_forwards;
2386 * If it's a Max-Forwards value of zero,
2387 * we have to intercept the request.
2389 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2392 * FIXME: We could handle at least TRACE here,
2393 * but that would require a verbatim copy of
2394 * the request which we don't have anymore
2396 log_error(LOG_LEVEL_HEADER,
2397 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2400 /* Get mem for response or fail*/
2401 if (NULL == (rsp = alloc_http_response()))
2403 return cgi_error_memory();
2406 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2408 free_http_response(rsp);
2409 return cgi_error_memory();
2413 rsp->reason = RSP_REASON_UNSUPPORTED;
2415 return(finish_http_response(csp, rsp));