1 const char filters_rcs[] = "$Id: filters.c,v 1.82 2007/03/13 11:28:43 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 1.82 2007/03/13 11:28:43 fabiankeil
44 * - Fix port handling in acl_addr() and use a temporary acl spec
45 * copy so error messages don't contain a truncated version.
46 * - Log size of iob before and after decompression.
48 * Revision 1.81 2007/03/05 14:40:53 fabiankeil
49 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
50 * - Hide the "Go there anyway" link for blocked CONNECT
51 * requests where going there anyway doesn't work anyway.
53 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
54 * - Save the reason for generating http_responses.
55 * - Block (+block) with status code 403 instead of 404.
56 * - Use a different kludge to remember a failed decompression.
58 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
59 * Search for Max-Forwards headers case-insensitive,
60 * don't generate the "501 unsupported" message for invalid
61 * Max-Forwards values and don't increase negative ones.
63 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
64 * - Add HEAD support to finish_http_response.
65 * - Add error favicon to internal HTML error messages.
67 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
68 * Mark *csp as immutable for is_untrusted_url()
69 * and is_imageurl(). Closes FR 1237736.
71 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
72 * Integrate a modified version of Wil Mahan's
73 * zlib patch (PR #895531).
75 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
76 * Fixed gcc43 conversion warnings,
77 * changed sprintf calls to snprintf.
79 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
80 * Adjust comment in pcrs_filter_response()
81 * to recent pcrs changes. Hohoho.
83 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
84 * Don't crash if pcre returns an error code
85 * that pcrs didn't expect. Fixes BR 1621173.
87 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
88 * Modified is_untrusted_url to complain in case of
89 * write errors and to give a reason when adding new
90 * entries to the trustfile. Closes FR 1097611.
92 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
93 * Skip empty filter files in pcrs_filter_response,
94 * but don't ignore the ones that come afterwards.
95 * Fixes parts of BR 1619208.
97 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
98 * Added some sanity checks for get_last_url().
99 * Fixed possible segfault caused by my last commit.
101 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
102 * Let get_last_url() catch https URLs as well.
104 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
105 * Make sure get_last_url() behaves like advertised
106 * and fast-redirects{} can be combined with redirect{}.
108 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
109 * Implemented +redirect{s@foo@bar@} to generate
110 * a redirect based on a rewritten version of the
113 * Revision 1.66 2006/09/23 13:26:38 roro
114 * Replace TABs by spaces in source code.
116 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
117 * Fix +redirect{}. Didn't work with -fast-redirects.
119 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
120 * Block requests for untrusted URLs with status
121 * code 403 instead of 200.
123 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
124 * Don't free p which is still in use and will be later
125 * freed by free_map(). Don't claim the referrer is unknown
126 * when the client didn't set one.
128 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
129 * Feature request 595948: Re-Filter logging in single line
131 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
132 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
134 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
135 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
136 * with what was really the latest development (the v_3_0_branch branch)
138 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
139 * Multiple filter file support
141 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
142 * A few memory leaks plugged (mostly on error paths)
144 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
145 * Add the ability to check jpeg images for invalid
146 * lengths of comment blocks. Defensive strategy
147 * against the exploit:
148 * Microsoft Security Bulletin MS04-028
149 * Buffer Overrun in JPEG Processing (GDI+) Could
150 * Allow Code Execution (833987)
151 * Enabled with +inspect-jpegs in actions files.
153 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
154 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
156 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
157 * Fixed bug #839859: "See why" link URL now gets url-encoded.
159 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
162 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
163 * Added basic support for OPTIONS and TRACE HTTP methods:
164 * New function direct_response which handles OPTIONS and
165 * TRACE requests whose Max-Forwards header field is zero.
167 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
168 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
170 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
171 * - Bugfix: Executing a filters without jobs no longer results in
172 * turing off *all* filters.
173 * - Security fix: Malicious web servers can't cause a seg fault
174 * through bogus chunk sizes anymore
176 * Revision 1.58 2002/04/24 02:11:17 oes
177 * Jon's multiple AF patch: url_actions now evaluates rules
180 * Revision 1.57 2002/04/08 20:38:34 swa
183 * Revision 1.56 2002/04/05 15:51:24 oes
184 * - bugfix: error-pages now get correct request protocol
185 * - fix for invalid HTML in trust info
187 * Revision 1.55 2002/04/02 16:13:51 oes
188 * Fix: No "Go there anyway" for SSL
190 * Revision 1.54 2002/04/02 14:55:56 oes
191 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
193 * Revision 1.53 2002/03/26 22:29:54 swa
194 * we have a new homepage!
196 * Revision 1.52 2002/03/24 16:35:57 jongfoster
199 * Revision 1.51 2002/03/24 15:23:33 jongfoster
202 * Revision 1.50 2002/03/24 13:25:43 swa
203 * name change related issues
205 * Revision 1.49 2002/03/16 20:29:14 oes
208 * Revision 1.48 2002/03/13 20:25:34 oes
209 * Better logging for content filters
211 * Revision 1.47 2002/03/13 00:30:52 jongfoster
213 * Added option of always sending redirect for imageblock,
214 * currently disabled with #if 0.
216 * Revision 1.46 2002/03/12 01:42:49 oes
217 * Introduced modular filters
219 * Revision 1.45 2002/03/08 16:47:50 oes
220 * Added choice beween GIF and PNG built-in images
222 * Revision 1.44 2002/03/07 03:49:31 oes
223 * - Fixed compiler warnings etc
224 * - Changed built-in images from GIF to PNG
225 * (with regard to Unisys patent issue)
226 * - Added a 4x4 pattern PNG which is less intrusive
227 * than the logo but also clearly marks the deleted banners
229 * Revision 1.43 2002/01/22 23:51:59 jongfoster
230 * Replacing strsav() with the safer string_append().
232 * Adding missing html_encode() to error message generators. Where encoded
233 * and unencoded versions of a string were provided, removing the unencoded
236 * Revision 1.42 2002/01/17 21:00:32 jongfoster
237 * Moving all our URL and URL pattern parsing code to urlmatch.c.
239 * Using a single, simple url_match(pattern,url) function - rather than
240 * the 3-line match routine which was repeated all over the place.
242 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
244 * Using parse_http_url() to parse URLs without faking a HTTP
245 * request line for parse_http_request().
247 * Revision 1.41 2001/11/13 00:14:07 jongfoster
248 * Fixing stupid bug now I've figured out what || means.
249 * (It always returns 0 or 1, not one of it's paramaters.)
251 * Revision 1.40 2001/10/26 17:37:55 oes
252 * - Re-enabled Netscape 200/404 bug workaround in block_url():
253 * - Removed OS/2 special case
254 * - Made block_url() independant from sed() having been run
255 * - Made trust_url independant from sed() having been run
256 * - Made is_imageurl independant from sed() having been run.
257 * It now checks User-Agent: and Accept: by itself.
260 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
261 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
262 * threads to call select() simultaneously. So, it's time to do a real, live,
263 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
264 * (native). Both versions will work, but using __OS2__ offers multi-threading.
266 * Revision 1.38 2001/10/23 21:32:33 jongfoster
267 * Adding error-checking to selected functions
269 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
270 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
271 * filters.c. Added a FIXME in front of the offending code. I'll gladly
272 * put in a better/more robust fix for all parties if one is presented...
273 * It seems that just returning 200 instead of 404 would pretty much fix
274 * it for everyone, but I don't know all the history of the problem.
276 * Revision 1.36 2001/10/10 16:44:16 oes
277 * Added match_portlist function
279 * Revision 1.35 2001/10/07 15:41:23 oes
280 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
282 * New function remove_chunked_transfer_coding that strips chunked
283 * transfer coding to plain and is called by pcrs_filter_response
284 * and gif_deanimate_response if neccessary
286 * Improved handling of zero-change re_filter runs
288 * pcrs_filter_response and gif_deanimate_response now remove
289 * chunked transfer codeing before processing the body.
291 * Revision 1.34 2001/09/20 15:49:36 steudten
293 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
294 * See cgi.c fill_template().
296 * Revision 1.33 2001/09/16 17:05:14 jongfoster
297 * Removing unused #include showarg.h
299 * Revision 1.32 2001/09/16 13:21:27 jongfoster
300 * Changes to use new list functions.
302 * Revision 1.31 2001/09/16 11:38:02 jongfoster
303 * Splitting fill_template() into 2 functions:
304 * template_load() loads the file
305 * template_fill() performs the PCRS regexps.
306 * This is because the CGI edit interface has a "table row"
307 * template which is used many times in the page - this
308 * change means it's only loaded from disk once.
310 * Revision 1.30 2001/09/16 11:00:10 jongfoster
311 * New function alloc_http_response, for symmetry with free_http_response
313 * Revision 1.29 2001/09/13 23:32:40 jongfoster
314 * Moving image data to cgi.c rather than cgi.h
315 * Fixing a GPF under Win32 (and any other OS that protects global
316 * constants from being written to).
318 * Revision 1.28 2001/09/10 10:18:51 oes
319 * Silenced compiler warnings
321 * Revision 1.27 2001/08/05 16:06:20 jongfoster
322 * Modifiying "struct map" so that there are now separate header and
323 * "map_entry" structures. This means that functions which modify a
324 * map no longer need to return a pointer to the modified map.
325 * Also, it no longer reverses the order of the entries (which may be
326 * important with some advanced template substitutions).
328 * Revision 1.26 2001/07/30 22:08:36 jongfoster
329 * Tidying up #defines:
330 * - All feature #defines are now of the form FEATURE_xxx
331 * - Permanently turned off WIN_GUI_EDIT
332 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
334 * Revision 1.25 2001/07/26 10:09:46 oes
335 * Made browser detection a little less naive
337 * Revision 1.24 2001/07/25 17:22:51 oes
338 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
340 * Revision 1.23 2001/07/23 13:40:12 oes
341 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
343 * Revision 1.22 2001/07/18 12:29:34 oes
344 * - Made gif_deanimate_response respect
345 * csp->action->string[ACTION_STRING_DEANIMATE]
346 * - Logging cosmetics
348 * Revision 1.21 2001/07/13 13:59:53 oes
349 * - Introduced gif_deanimate_response which shares the
350 * generic content modification interface of pcrs_filter_response
351 * and acts as a wrapper to deanimate.c:gif_deanimate()
352 * - Renamed re_process_buffer to pcrs_filter_response
353 * - pcrs_filter_response now returns NULL on failiure
354 * - Removed all #ifdef PCRS
356 * Revision 1.20 2001/07/01 17:01:04 oes
357 * Added comments and missing return statement in is_untrusted_url()
359 * Revision 1.19 2001/06/29 21:45:41 oes
360 * Indentation, CRLF->LF, Tab-> Space
362 * Revision 1.18 2001/06/29 13:27:38 oes
363 * - Cleaned up, renamed and reorderd functions
364 * and improved comments
367 * - Ported to CGI platform. Now delivers
368 * http_response or NULL
369 * - Unified HTML and GIF generation (moved image detection
370 * and GIF generation here from jcc.c:chat())
371 * - Fixed HTTP status to:
372 * - 403 (Forbidden) for the "blocked" HTML message
373 * - 200 (OK) for GIF answers
374 * - 302 (Redirect) for redirect to GIF
377 * - Ported to CGI platform. Now delivers
378 * http_response or NULL
379 * - Separated detection of untrusted URL into
380 * (bool)is_untrusted_url
381 * - Added enforcement of untrusted requests
383 * - Moved redirect_url() from cgi.c to here
384 * and ported it to the CGI platform
386 * - Removed logentry from cancelled commit
388 * Revision 1.17 2001/06/09 10:55:28 jongfoster
389 * Changing BUFSIZ ==> BUFFER_SIZE
391 * Revision 1.16 2001/06/07 23:10:26 jongfoster
392 * Allowing unanchored domain patterns to back off and retry
393 * if they partially match. Optimized right-anchored patterns.
394 * Moving ACL and forward files into config file.
395 * Replacing struct gateway with struct forward_spec
397 * Revision 1.15 2001/06/03 19:12:00 oes
398 * extracted-CGI relevant stuff
400 * Revision 1.14 2001/06/01 10:30:55 oes
401 * Added optional left-anchoring to domaincmp
403 * Revision 1.13 2001/05/31 21:21:30 jongfoster
404 * Permissionsfile / actions file changes:
405 * - Changed "permission" to "action" throughout
406 * - changes to file format to allow string parameters
407 * - Moved helper functions to actions.c
409 * Revision 1.12 2001/05/31 17:35:20 oes
411 * - Enhanced domain part globbing with infix and prefix asterisk
412 * matching and optional unanchored operation
414 * Revision 1.11 2001/05/29 11:53:23 oes
415 * "See why" link added to "blocked" page
417 * Revision 1.10 2001/05/29 09:50:24 jongfoster
418 * Unified blocklist/imagelist/permissionslist.
419 * File format is still under discussion, but the internal changes
422 * Also modified interceptor behaviour:
423 * - We now intercept all URLs beginning with one of the following
424 * prefixes (and *only* these prefixes):
426 * * http://ijbswa.sf.net/config/
427 * * http://ijbswa.sourceforge.net/config/
428 * - New interceptors "home page" - go to http://i.j.b/ to see it.
429 * - Internal changes so that intercepted and fast redirect pages
430 * are not replaced with an image.
431 * - Interceptors now have the option to send a binary page direct
432 * to the client. (i.e. ijb-send-banner uses this)
433 * - Implemented show-url-info interceptor. (Which is why I needed
434 * the above interceptors changes - a typical URL is
435 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
436 * The previous mechanism would not have intercepted that, and
437 * if it had been intercepted then it then it would have replaced
440 * Revision 1.9 2001/05/27 22:17:04 oes
442 * - re_process_buffer no longer writes the modified buffer
443 * to the client, which was very ugly. It now returns the
444 * buffer, which it is then written by chat.
446 * - content_length now adjusts the Content-Length: header
447 * for modified documents rather than crunch()ing it.
448 * (Length info in csp->content_length, which is 0 for
449 * unmodified documents)
451 * - For this to work, sed() is called twice when filtering.
453 * Revision 1.8 2001/05/26 17:13:28 jongfoster
454 * Filled in a function comment.
456 * Revision 1.7 2001/05/26 15:26:15 jongfoster
457 * ACL feature now provides more security by immediately dropping
458 * connections from untrusted hosts.
460 * Revision 1.6 2001/05/26 00:28:36 jongfoster
461 * Automatic reloading of config file.
462 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
463 * Most of the global variables have been moved to a new
464 * struct configuration_spec, accessed through csp->config->globalname
465 * Most of the globals remaining are used by the Win32 GUI.
467 * Revision 1.5 2001/05/25 22:34:30 jongfoster
470 * Revision 1.4 2001/05/22 18:46:04 oes
472 * - Enabled filtering banners by size rather than URL
473 * by adding patterns that replace all standard banner
474 * sizes with the "Junkbuster" gif to the re_filterfile
476 * - Enabled filtering WebBugs by providing a pattern
477 * which kills all 1x1 images
479 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
480 * which is selected by the (nonstandard and therefore
481 * capital) letter 'U' in the option string.
482 * It causes the quantifiers to be ungreedy by default.
483 * Appending a ? turns back to greedy (!).
485 * - Added a new interceptor ijb-send-banner, which
486 * sends back the "Junkbuster" gif. Without imagelist or
487 * MSIE detection support, or if tinygif = 1, or the
488 * URL isn't recognized as an imageurl, a lame HTML
489 * explanation is sent instead.
491 * - Added new feature, which permits blocking remote
492 * script redirects and firing back a local redirect
494 * The feature is conditionally compiled, i.e. it
495 * can be disabled with --disable-fast-redirects,
496 * plus it must be activated by a "fast-redirects"
497 * line in the config file, has its own log level
498 * and of course wants to be displayed by show-proxy-args
499 * Note: Boy, all the #ifdefs in 1001 locations and
500 * all the fumbling with configure.in and acconfig.h
501 * were *way* more work than the feature itself :-(
503 * - Because a generic redirect template was needed for
504 * this, tinygif = 3 now uses the same.
506 * - Moved GIFs, and other static HTTP response templates
511 * - Removed some >400 CRs again (Jon, you really worked
514 * Revision 1.3 2001/05/20 16:44:47 jongfoster
515 * Removing last hardcoded Junkbusters.com URLs.
517 * Revision 1.2 2001/05/20 01:21:20 jongfoster
518 * Version 2.9.4 checkin.
519 * - Merged popupfile and cookiefile, and added control over PCRS
520 * filtering, in new "permissionsfile".
521 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
522 * file error you now get a message box (in the Win32 GUI) rather
523 * than the program exiting with no explanation.
524 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
526 * - Removed tabs from "config"
527 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
528 * - Bumped up version number.
530 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
531 * Initial import of version 2.9.3 source tree
534 *********************************************************************/
540 #include <sys/types.h>
549 #endif /* ndef __OS2__ */
550 #include <netinet/in.h>
552 #include <winsock2.h>
553 #endif /* ndef _WIN32 */
557 #endif /* def __OS2__ */
565 #include "jbsockets.h"
566 #include "miscutil.h"
570 #include "deanimate.h"
571 #include "urlmatch.h"
577 const char filters_h_rcs[] = FILTERS_H_VERSION;
579 /* Fix a problem with Solaris. There should be no effect on other
581 * Solaris's isspace() is a macro which uses it's argument directly
582 * as an array index. Therefore we need to make sure that high-bit
583 * characters generate +ve values, and ideally we also want to make
584 * the argument match the declared parameter type of "int".
586 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
590 /*********************************************************************
592 * Function : block_acl
594 * Description : Block this request?
595 * Decide yes or no based on ACL file.
598 * 1 : dst = The proxy or gateway address this is going to.
599 * Or NULL to check all possible targets.
600 * 2 : csp = Current client state (buffers, headers, etc...)
601 * Also includes the client IP address.
603 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
605 *********************************************************************/
606 int block_acl(struct access_control_addr *dst, struct client_state *csp)
608 struct access_control_list *acl = csp->config->acl;
610 /* if not using an access control list, then permit the connection */
616 /* search the list */
619 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
623 /* Just want to check if they have any access */
624 if (acl->action == ACL_PERMIT)
629 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
630 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
632 if (acl->action == ACL_PERMIT)
650 /*********************************************************************
652 * Function : acl_addr
654 * Description : Called from `load_config' to parse an ACL address.
657 * 1 : aspec = String specifying ACL address.
658 * 2 : aca = struct access_control_addr to fill in.
660 * Returns : 0 => Ok, everything else is an error.
662 *********************************************************************/
663 int acl_addr(const char *aspec, struct access_control_addr *aca)
668 char *acl_spec = NULL;
674 * Use a temporary acl spec copy so we can log
675 * the unmodified original in case of parse errors.
677 acl_spec = strdup(aspec);
678 if (acl_spec == NULL)
680 /* XXX: This will be logged as parse error. */
684 if ((p = strchr(acl_spec, '/')) != NULL)
687 if (ijb_isdigit(*p) == 0)
692 masklength = atoi(p);
695 if ((masklength < 0) || (masklength > 32))
701 if ((p = strchr(acl_spec, ':')) != NULL)
706 port = strtol(p, &endptr, 10);
708 if (port <= 0 || port > 65535 || *endptr != '\0')
715 aca->port = (unsigned long)port;
717 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
720 if (aca->addr == INADDR_NONE)
722 /* XXX: This will be logged as parse error. */
726 /* build the netmask */
728 for (i=1; i <= masklength ; i++)
730 aca->mask |= (1 << (32 - i));
733 /* now mask off the host portion of the ip address
734 * (i.e. save on the network portion of the address).
736 aca->addr = aca->addr & aca->mask;
741 #endif /* def FEATURE_ACL */
744 /*********************************************************************
746 * Function : match_portlist
748 * Description : Check if a given number is covered by a comma
749 * separated list of numbers and ranges (a,b-c,d,..)
752 * 1 : portlist = String with list
753 * 2 : port = port to check
755 * Returns : 0 => no match
758 *********************************************************************/
759 int match_portlist(const char *portlist, int port)
761 char *min, *max, *next, *portlist_copy;
763 min = next = portlist_copy = strdup(portlist);
766 * Zero-terminate first item and remember offset for next
768 if (NULL != (next = strchr(portlist_copy, (int) ',')))
774 * Loop through all items, checking for match
778 if (NULL == (max = strchr(min, (int) '-')))
781 * No dash, check for equality
783 if (port == atoi(min))
792 * This is a range, so check if between min and max,
793 * or, if max was omitted, between min and 65K
796 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
810 * Zero-terminate next item and remember offset for n+1
812 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
824 /*********************************************************************
826 * Function : block_url
828 * Description : Called from `chat'. Check to see if we need to block this.
831 * 1 : csp = Current client state (buffers, headers, etc...)
833 * Returns : NULL => unblocked, else HTTP block response
835 *********************************************************************/
836 struct http_response *block_url(struct client_state *csp)
838 struct http_response *rsp;
839 const char *new_content_type = NULL;
842 * If it's not blocked, don't block it ;-)
844 if ((csp->action->flags & ACTION_BLOCK) == 0)
848 if (csp->action->flags & ACTION_REDIRECT)
850 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
853 * Else, prepare a response
855 if (NULL == (rsp = alloc_http_response()))
857 return cgi_error_memory();
861 * If it's an image-url, send back an image or redirect
862 * as specified by the relevant +image action
864 #ifdef FEATURE_IMAGE_BLOCKING
865 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
869 /* determine HOW images should be blocked */
870 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
872 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
874 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
876 #if 1 /* Two alternative strategies, use this one for now: */
878 /* and handle accordingly: */
879 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
881 rsp->status = strdup("403 Request blocked by Privoxy");
882 if (rsp->status == NULL)
884 free_http_response(rsp);
885 return cgi_error_memory();
887 rsp->body = bindup(image_pattern_data, image_pattern_length);
888 if (rsp->body == NULL)
890 free_http_response(rsp);
891 return cgi_error_memory();
893 rsp->content_length = image_pattern_length;
895 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
897 free_http_response(rsp);
898 return cgi_error_memory();
902 else if (0 == strcmpic(p, "blank"))
904 rsp->status = strdup("403 Request blocked by Privoxy");
905 if (rsp->status == NULL)
907 free_http_response(rsp);
908 return cgi_error_memory();
910 rsp->body = bindup(image_blank_data, image_blank_length);
911 if (rsp->body == NULL)
913 free_http_response(rsp);
914 return cgi_error_memory();
916 rsp->content_length = image_blank_length;
918 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
920 free_http_response(rsp);
921 return cgi_error_memory();
927 rsp->status = strdup("302 Local Redirect from Privoxy");
928 if (rsp->status == NULL)
930 free_http_response(rsp);
931 return cgi_error_memory();
934 if (enlist_unique_header(rsp->headers, "Location", p))
936 free_http_response(rsp);
937 return cgi_error_memory();
941 #else /* Following code is disabled for now */
943 /* and handle accordingly: */
944 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
946 p = CGI_PREFIX "send-banner?type=pattern";
948 else if (0 == strcmpic(p, "blank"))
950 p = CGI_PREFIX "send-banner?type=blank";
952 rsp->status = strdup("302 Local Redirect from Privoxy");
953 if (rsp->status == NULL)
955 free_http_response(rsp);
956 return cgi_error_memory();
959 if (enlist_unique_header(rsp->headers, "Location", p))
961 free_http_response(rsp);
962 return cgi_error_memory();
964 #endif /* Preceeding code is disabled for now */
966 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
969 * Send empty document.
971 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
974 rsp->body = strdup(" ");
975 rsp->content_length = 1;
977 rsp->status = strdup("403 Request blocked by Privoxy");
978 if (rsp->status == NULL)
980 free_http_response(rsp);
981 return cgi_error_memory();
983 if (new_content_type != 0)
985 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
986 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
988 free_http_response(rsp);
989 return cgi_error_memory();
994 #endif /* def FEATURE_IMAGE_BLOCKING */
997 * Else, generate an HTML "blocked" message:
1001 struct map * exports;
1005 * Workaround for stupid Netscape bug which prevents
1006 * pages from being displayed if loading a referenced
1007 * JavaScript or style sheet fails. So make it appear
1008 * as if it succeeded.
1010 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1011 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1012 && !strstr(p, "Gecko") /* save Mozilla, */
1013 && !strstr(p, "compatible") /* MSIE */
1014 && !strstr(p, "Opera")) /* and Opera. */
1016 rsp->status = strdup("200 Request for blocked URL");
1020 rsp->status = strdup("403 Request for blocked URL");
1023 if (rsp->status == NULL)
1025 free_http_response(rsp);
1026 return cgi_error_memory();
1029 exports = default_exports(csp, NULL);
1030 if (exports == NULL)
1032 free_http_response(rsp);
1033 return cgi_error_memory();
1036 #ifdef FEATURE_FORCE_LOAD
1037 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1039 * Export the force conditional block killer if
1041 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1042 * - Privoxy is configured to enforce blocks, or
1043 * - it's a CONNECT request and enforcing wouldn't work anyway.
1045 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1046 || (0 == strcmpic(csp->http->gpc, "connect")))
1047 #endif /* ndef FEATURE_FORCE_LOAD */
1049 err = map_block_killer(exports, "force-support");
1052 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1053 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1054 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1055 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1060 free_http_response(rsp);
1061 return cgi_error_memory();
1064 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1067 free_http_response(rsp);
1068 return cgi_error_memory();
1071 rsp->reason = RSP_REASON_BLOCKED;
1073 return finish_http_response(csp, rsp);
1078 #ifdef FEATURE_TRUST
1079 /*********************************************************************
1081 * Function : trust_url FIXME: I should be called distrust_url
1083 * Description : Calls is_untrusted_url to determine if the URL is trusted
1084 * and if not, returns a HTTP 403 response with a reject message.
1087 * 1 : csp = Current client state (buffers, headers, etc...)
1089 * Returns : NULL => trusted, else http_response.
1091 *********************************************************************/
1092 struct http_response *trust_url(struct client_state *csp)
1094 struct http_response *rsp;
1095 struct map * exports;
1096 char buf[BUFFER_SIZE];
1098 struct url_spec **tl;
1103 * Don't bother to work on trusted URLs
1105 if (!is_untrusted_url(csp))
1111 * Else, prepare a response:
1113 if (NULL == (rsp = alloc_http_response()))
1115 return cgi_error_memory();
1118 rsp->status = strdup("403 Request blocked by Privoxy");
1119 exports = default_exports(csp, NULL);
1120 if (exports == NULL || rsp->status == NULL)
1122 free_http_response(rsp);
1123 return cgi_error_memory();
1127 * Export the protocol, host, port, and referrer information
1129 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1130 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1131 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1133 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1135 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1139 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1145 free_http_response(rsp);
1146 return cgi_error_memory();
1150 * Export the trust list
1153 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1155 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1156 string_append(&p, buf);
1158 err = map(exports, "trusted-referrers", 1, p, 0);
1163 free_http_response(rsp);
1164 return cgi_error_memory();
1168 * Export the trust info, if available
1170 if (csp->config->trust_info->first)
1172 struct list_entry *l;
1175 for (l = csp->config->trust_info->first; l ; l = l->next)
1177 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1178 string_append(&p, buf);
1180 err = map(exports, "trust-info", 1, p, 0);
1184 err = map_block_killer(exports, "have-trust-info");
1190 free_http_response(rsp);
1191 return cgi_error_memory();
1195 * Export the force conditional block killer if
1197 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
1198 * - Privoxy is configured to enforce blocks, or
1199 * - it's a CONNECT request and enforcing wouldn't work anyway.
1201 #ifdef FEATURE_FORCE_LOAD
1202 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
1203 || (0 == strcmpic(csp->http->gpc, "connect")))
1205 err = map_block_killer(exports, "force-support");
1209 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1211 #else /* ifndef FEATURE_FORCE_LOAD */
1212 err = map_block_killer(exports, "force-support");
1213 #endif /* ndef FEATURE_FORCE_LOAD */
1218 free_http_response(rsp);
1219 return cgi_error_memory();
1223 * Build the response
1225 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1228 free_http_response(rsp);
1229 return cgi_error_memory();
1231 rsp->reason = RSP_REASON_UNTRUSTED;
1233 return finish_http_response(csp, rsp);
1235 #endif /* def FEATURE_TRUST */
1237 /*********************************************************************
1239 * Function : execute_single_pcrs_command
1241 * Description : Apply single pcrs command to the subject.
1242 * The subject itself is left untouched, memory for the result
1243 * is malloc()ed and it is the caller's responsibility to free
1244 * the result when it's no longer needed.
1247 * 1 : subject = the subject (== original) string
1248 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1249 * 3 : hits = int* for returning the number of modifications
1251 * Returns : NULL in case of errors, otherwise the
1252 * result of the pcrs command.
1254 *********************************************************************/
1255 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1259 char *result = NULL;
1263 assert(pcrs_command);
1266 size = strlen(subject);
1268 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1270 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1271 pcrs_command, error);
1273 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1275 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1282 job = pcrs_free_job(job);
1289 /*********************************************************************
1291 * Function : rewrite_url
1293 * Description : Rewrites a URL with a single pcrs command
1294 * and returns the result if it differs from the
1295 * original and isn't obviously invalid.
1298 * 1 : old_url = URL to rewrite.
1299 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1302 * Returns : NULL if the pcrs_command didn't change the url, or
1303 * the result of the modification.
1305 *********************************************************************/
1306 char *rewrite_url(char *old_url, const char *pcrs_command)
1308 char *new_url = NULL;
1312 assert(pcrs_command);
1314 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1318 log_error(LOG_LEVEL_REDIRECTS,
1319 "pcrs command \"%s\" didn't change \"%s\".",
1320 pcrs_command, old_url, new_url);
1323 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1325 log_error(LOG_LEVEL_ERROR,
1326 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1327 "but the result doesn't look like a valid URL and will be ignored.",
1328 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1333 log_error(LOG_LEVEL_REDIRECTS,
1334 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1335 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1343 #ifdef FEATURE_FAST_REDIRECTS
1344 /*********************************************************************
1346 * Function : get_last_url
1348 * Description : Search for the last URL inside a string.
1349 * If the string already is a URL, it will
1350 * be the first URL found.
1353 * 1 : subject = the string to check
1354 * 2 : redirect_mode = +fast-redirect{} mode
1356 * Returns : NULL if no URL was found, or
1357 * the last URL found.
1359 *********************************************************************/
1360 char *get_last_url(char *subject, const char *redirect_mode)
1362 char *new_url = NULL;
1366 assert(redirect_mode);
1368 subject = strdup(subject);
1369 if (subject == NULL)
1371 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1375 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1377 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1378 new_url = url_decode(subject);
1379 if (new_url != NULL)
1386 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1390 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1393 * Find the last URL encoded in the request
1396 while ((tmp = strstr(tmp, "http://")) != NULL)
1400 tmp = (new_url != NULL) ? new_url : subject;
1401 while ((tmp = strstr(tmp, "https://")) != NULL)
1406 if ((new_url != NULL)
1407 && ( (new_url != subject)
1408 || (0 == strncmpic(subject, "http://", 7))
1409 || (0 == strncmpic(subject, "https://", 8))
1413 * Return new URL if we found a redirect
1414 * or if the subject already was a URL.
1416 * The second case makes sure that we can
1417 * chain get_last_url after another redirection check
1418 * (like rewrite_url) without losing earlier redirects.
1420 new_url = strdup(new_url);
1429 #endif /* def FEATURE_FAST_REDIRECTS */
1432 /*********************************************************************
1434 * Function : redirect_url
1436 * Description : Checks if Privoxy should answer the request with
1437 * a HTTP redirect and generates the redirect if
1441 * 1 : csp = Current client state (buffers, headers, etc...)
1443 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1445 *********************************************************************/
1446 struct http_response *redirect_url(struct client_state *csp)
1448 struct http_response *rsp;
1449 #ifdef FEATURE_FAST_REDIRECTS
1451 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1452 * as compile-time option? The user can easily disable
1453 * it in his action file.
1455 char * redirect_mode;
1456 #endif /* def FEATURE_FAST_REDIRECTS */
1457 char *old_url = NULL;
1458 char *new_url = NULL;
1459 char *redirection_string;
1461 if ((csp->action->flags & ACTION_REDIRECT))
1463 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1466 * If the redirection string begins with 's',
1467 * assume it's a pcrs command, otherwise treat it as
1468 * properly formatted URL and use it for the redirection
1471 * According to RFC 2616 section 14.30 the URL
1472 * has to be absolute and if the user tries:
1473 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1474 * she would get undefined results anyway.
1478 if (*redirection_string == 's')
1480 old_url = csp->http->url;
1481 new_url = rewrite_url(old_url, redirection_string);
1485 log_error(LOG_LEVEL_REDIRECTS,
1486 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1487 redirection_string);
1488 new_url = strdup(redirection_string);
1492 #ifdef FEATURE_FAST_REDIRECTS
1493 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1495 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1498 * If it exists, use the previously rewritten URL as input
1499 * otherwise just use the old path.
1501 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1502 new_url = get_last_url(old_url, redirect_mode);
1505 #endif /* def FEATURE_FAST_REDIRECTS */
1507 /* Did any redirect action trigger? */
1510 if (0 == strcmpic(new_url, csp->http->url))
1512 log_error(LOG_LEVEL_ERROR,
1513 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1514 csp->http->url, new_url);
1519 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1521 if (NULL == (rsp = alloc_http_response()))
1524 return cgi_error_memory();
1527 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1528 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1531 free_http_response(rsp);
1532 return cgi_error_memory();
1534 rsp->reason = RSP_REASON_REDIRECTED;
1537 return finish_http_response(csp, rsp);
1541 /* Only reached if no redirect is required */
1547 #ifdef FEATURE_IMAGE_BLOCKING
1548 /*********************************************************************
1550 * Function : is_imageurl
1552 * Description : Given a URL, decide whether it is an image or not,
1553 * using either the info from a previous +image action
1554 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1555 * is MSIE and not on a Mac, tell from the browser's accept
1559 * 1 : csp = Current client state (buffers, headers, etc...)
1561 * Returns : True (nonzero) if URL is an image, false (0)
1564 *********************************************************************/
1565 int is_imageurl(const struct client_state *csp)
1567 #ifdef FEATURE_IMAGE_DETECT_MSIE
1570 tmp = get_header_value(csp->headers, "User-Agent:");
1571 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1573 tmp = get_header_value(csp->headers, "Accept:");
1574 if (tmp && strstr(tmp, "image/gif"))
1576 /* Client will accept HTML. If this seems counterintuitive,
1586 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1588 return ((csp->action->flags & ACTION_IMAGE) != 0);
1591 #endif /* def FEATURE_IMAGE_BLOCKING */
1594 #ifdef FEATURE_TRUST
1595 /*********************************************************************
1597 * Function : is_untrusted_url
1599 * Description : Should we "distrust" this URL (and block it)?
1601 * Yes if it matches a line in the trustfile, or if the
1602 * referrer matches a line starting with "+" in the
1607 * 1 : csp = Current client state (buffers, headers, etc...)
1609 * Returns : 0 => trusted, 1 => untrusted
1611 *********************************************************************/
1612 int is_untrusted_url(const struct client_state *csp)
1614 struct file_list *fl;
1615 struct block_spec *b;
1616 struct url_spec **trusted_url;
1617 struct http_request rhttp[1];
1618 const char * referer;
1622 * If we don't have a trustlist, we trust everybody
1624 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1629 memset(rhttp, '\0', sizeof(*rhttp));
1632 * Do we trust the request URL itself?
1634 for (b = b->next; b ; b = b->next)
1636 if (url_match(b->url, csp->http))
1642 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1644 /* no referrer was supplied */
1650 * If not, do we maybe trust its referrer?
1652 err = parse_http_url(referer, rhttp, csp);
1658 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1660 if (url_match(*trusted_url, rhttp))
1662 /* if the URL's referrer is from a trusted referrer, then
1663 * add the target spec to the trustfile as an unblocked
1664 * domain and return 0 (which means it's OK).
1669 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1673 char * new_entry = strdup("~");
1675 string_append(&new_entry, csp->http->hostport);
1677 path = csp->http->path;
1678 if ( (path[0] == '/')
1680 && ((path_end = strchr(path + 2, '/')) != NULL))
1682 /* since this path points into a user's home space
1683 * be sure to include this spec in the trustfile.
1685 int path_len = path_end - path; /* save offset */
1686 path = strdup(path); /* Copy string */
1689 path_end = path + path_len; /* regenerate ptr to new buffer */
1690 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1692 string_join(&new_entry, path);
1696 * Give a reason for generating this entry.
1698 string_append(&new_entry, " # Trusted referrer was: ");
1699 string_append(&new_entry, referer);
1701 if (new_entry != NULL)
1703 if (-1 == fprintf(fp, "%s\n", new_entry))
1705 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1706 new_entry, csp->config->trustfile);
1712 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1713 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1720 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1721 csp->http->hostport, csp->config->trustfile);
1729 #endif /* def FEATURE_TRUST */
1732 /*********************************************************************
1734 * Function : pcrs_filter_response
1736 * Description : Execute all text substitutions from all applying
1737 * +filter actions on the text buffer that's been accumulated
1738 * in csp->iob->buf. If this changes the contents, set
1739 * csp->content_length to the modified size and raise the
1740 * CSP_FLAG_MODIFIED flag.
1742 * XXX: Currently pcrs_filter_response is also responsible
1743 * for dechunking and decompressing. Both should be
1744 * done in separate functions so other content modifiers
1745 * profit as well, even if pcrs filtering is disabled.
1748 * 1 : csp = Current client state (buffers, headers, etc...)
1750 * Returns : a pointer to the (newly allocated) modified buffer.
1751 * or NULL if there were no hits or something went wrong
1753 *********************************************************************/
1754 char *pcrs_filter_response(struct client_state *csp)
1757 size_t size, prev_size;
1759 char *old = csp->iob->cur, *new = NULL;
1762 struct file_list *fl;
1763 struct re_filterfile_spec *b;
1764 struct list_entry *filtername;
1766 int i, found_filters = 0;
1771 if (csp->iob->cur >= csp->iob->eod)
1775 size = (size_t)(csp->iob->eod - csp->iob->cur);
1778 * Need to check the set of re_filterfiles...
1780 for (i = 0; i < MAX_AF_FILES; i++)
1793 if (0 == found_filters)
1795 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1800 * If the body has a "chunked" transfer-encoding,
1801 * get rid of it first, adjusting size and iob->eod
1803 if (csp->flags & CSP_FLAG_CHUNKED)
1805 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1806 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1810 csp->iob->eod = csp->iob->cur + size;
1811 csp->flags |= CSP_FLAG_MODIFIED;
1816 * If the body has a compressed transfer-encoding,
1817 * uncompress it first, adjusting size and iob->eod.
1818 * Note that decompression occurs after de-chunking.
1820 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1822 /* Notice that we at least tried to decompress. */
1823 if (JB_ERR_OK != decompress_iob(csp))
1826 * We failed to decompress the data; there's no point
1827 * in continuing since we can't filter.
1829 * XXX: Actually the Accept-Encoding header may
1830 * just be incorrect in which case we could continue
1833 * Unset CT_GZIP and CT_DEFLATE to remember not
1834 * to modify the Content-Encoding header later.
1836 csp->content_type &= ~CT_GZIP;
1837 csp->content_type &= ~CT_DEFLATE;
1840 log_error(LOG_LEVEL_RE_FILTER,
1841 "Decompression successful. Old size: %d, new size: %d.",
1842 size, csp->iob->eod - csp->iob->cur);
1845 * Decompression gives us a completely new iob,
1846 * so we need to update.
1848 size = (size_t)(csp->iob->eod - csp->iob->cur);
1849 old = csp->iob->cur;
1851 csp->flags |= CSP_FLAG_MODIFIED;
1852 csp->content_type &= ~CT_TABOO;
1856 for (i = 0; i < MAX_AF_FILES; i++)
1859 if ((NULL == fl) || (NULL == fl->f))
1862 * Either there are no filter files
1863 * left, or this filter file just
1864 * contains no valid filters.
1866 * Continue to be sure we don't miss
1867 * valid filter files that are chained
1868 * after empty or invalid ones.
1873 * For all applying +filter actions, look if a filter by that
1874 * name exists and if yes, execute it's pcrs_joblist on the
1877 for (b = fl->f; b; b = b->next)
1879 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1880 filtername ; filtername = filtername->next)
1882 if (strcmp(b->name, filtername->str) == 0)
1884 int current_hits = 0; /* Number of hits caused by this filter */
1885 int job_number = 0; /* Which job we're currently executing */
1886 int job_hits = 0; /* How many hits the current job caused */
1888 if ( NULL == b->joblist )
1890 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1895 /* Apply all jobs from the joblist */
1896 for (job = b->joblist; NULL != job; job = job->next)
1899 job_hits = pcrs_execute(job, old, size, &new, &size);
1904 * That went well. Continue filtering
1905 * and use the result of this job as
1906 * input for the next one.
1908 current_hits += job_hits;
1909 if (old != csp->iob->cur)
1918 * This job caused an unexpected error. Inform the user
1919 * and skip the rest of the jobs in this filter. We could
1920 * continue with the next job, but usually the jobs
1921 * depend on each other or are similar enough to
1922 * fail for the same reason.
1924 * At the moment our pcrs expects the error codes of pcre 3.4,
1925 * but newer pcre versions can return additional error codes.
1926 * As a result pcrs_strerror()'s error message might be
1927 * "Unknown error ...", therefore we print the numerical value
1930 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1931 * should we use LOG_LEVEL_RE_FILTER instead?
1933 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1934 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1939 log_error(LOG_LEVEL_RE_FILTER,
1940 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1941 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1943 hits += current_hits;
1950 * If there were no hits, destroy our copy and let
1951 * chat() use the original in csp->iob
1959 csp->flags |= CSP_FLAG_MODIFIED;
1960 csp->content_length = size;
1968 /*********************************************************************
1970 * Function : gif_deanimate_response
1972 * Description : Deanimate the GIF image that has been accumulated in
1973 * csp->iob->buf, set csp->content_length to the modified
1974 * size and raise the CSP_FLAG_MODIFIED flag.
1977 * 1 : csp = Current client state (buffers, headers, etc...)
1979 * Returns : a pointer to the (newly allocated) modified buffer.
1980 * or NULL in case something went wrong.
1982 *********************************************************************/
1983 char *gif_deanimate_response(struct client_state *csp)
1985 struct binbuffer *in, *out;
1989 size = (size_t)(csp->iob->eod - csp->iob->cur);
1992 * If the body has a "chunked" transfer-encoding,
1993 * get rid of it first, adjusting size and iob->eod
1995 if (csp->flags & CSP_FLAG_CHUNKED)
1997 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1998 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2002 csp->iob->eod = csp->iob->cur + size;
2003 csp->flags |= CSP_FLAG_MODIFIED;
2006 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2007 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
2009 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
2013 in->buffer = csp->iob->cur;
2016 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2018 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2025 if ((int)size == out->offset)
2027 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2031 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2033 csp->content_length = out->offset;
2034 csp->flags |= CSP_FLAG_MODIFIED;
2044 /*********************************************************************
2046 * Function : jpeg_inspect_response
2051 * 1 : csp = Current client state (buffers, headers, etc...)
2053 * Returns : a pointer to the (newly allocated) modified buffer
2054 * or NULL in case something went wrong.
2056 *********************************************************************/
2057 char *jpeg_inspect_response(struct client_state *csp)
2059 struct binbuffer *in = NULL;
2060 struct binbuffer *out = NULL;
2064 size = (size_t)(csp->iob->eod - csp->iob->cur);
2067 * If the body has a "chunked" transfer-encoding,
2068 * get rid of it first, adjusting size and iob->eod
2070 if (csp->flags & CSP_FLAG_CHUNKED)
2072 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2073 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2077 csp->iob->eod = csp->iob->cur + size;
2078 csp->flags |= CSP_FLAG_MODIFIED;
2081 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2083 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2087 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2089 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2093 in->buffer = csp->iob->cur;
2097 * Calling jpeg_inspect has the side-effect of creating and
2098 * modifying the image buffer of "out" directly.
2100 if (jpeg_inspect(in, out))
2102 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2110 csp->content_length = out->offset;
2111 csp->flags |= CSP_FLAG_MODIFIED;
2121 /*********************************************************************
2123 * Function : remove_chunked_transfer_coding
2125 * Description : In-situ remove the "chunked" transfer coding as defined
2126 * in rfc2616 from a buffer.
2129 * 1 : buffer = Pointer to the text buffer
2130 * 2 : size = Number of bytes to be processed
2132 * Returns : The new size, i.e. the number of bytes from buffer which
2133 * are occupied by the stripped body, or 0 in case something
2136 *********************************************************************/
2137 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2140 unsigned int chunksize = 0;
2141 char *from_p, *to_p;
2144 from_p = to_p = buffer;
2146 if (sscanf(buffer, "%x", &chunksize) != 1)
2148 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2152 while (chunksize > 0)
2154 if (NULL == (from_p = strstr(from_p, "\r\n")))
2156 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2160 if ((newsize += chunksize) >= size)
2162 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2167 memmove(to_p, from_p, (size_t) chunksize);
2168 to_p = buffer + newsize;
2169 from_p += chunksize + 2;
2171 if (sscanf(from_p, "%x", &chunksize) != 1)
2173 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2178 /* FIXME: Should this get its own loglevel? */
2179 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2185 /*********************************************************************
2187 * Function : url_actions
2189 * Description : Gets the actions for this URL.
2192 * 1 : http = http_request request for blocked URLs
2193 * 2 : csp = Current client state (buffers, headers, etc...)
2197 *********************************************************************/
2198 void url_actions(struct http_request *http,
2199 struct client_state *csp)
2201 struct file_list *fl;
2202 struct url_actions *b;
2205 init_current_action(csp->action);
2207 for (i = 0; i < MAX_AF_FILES; i++)
2209 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2214 apply_url_actions(csp->action, http, b);
2221 /*********************************************************************
2223 * Function : apply_url_actions
2225 * Description : Applies a list of URL actions.
2228 * 1 : action = Destination.
2229 * 2 : http = Current URL
2230 * 3 : b = list of URL actions to apply
2234 *********************************************************************/
2235 void apply_url_actions(struct current_action_spec *action,
2236 struct http_request *http,
2237 struct url_actions *b)
2241 /* Should never happen */
2245 for (b = b->next; NULL != b; b = b->next)
2247 if (url_match(b->url, http))
2249 merge_current_action(action, b->action);
2255 /*********************************************************************
2257 * Function : forward_url
2259 * Description : Should we forward this to another proxy?
2262 * 1 : http = http_request request for current URL
2263 * 2 : csp = Current client state (buffers, headers, etc...)
2265 * Returns : Pointer to forwarding information.
2267 *********************************************************************/
2268 const struct forward_spec * forward_url(struct http_request *http,
2269 struct client_state *csp)
2271 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2272 struct forward_spec *fwd = csp->config->forward;
2281 if (url_match(fwd->url, http))
2292 /*********************************************************************
2294 * Function : direct_response
2296 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2297 * request and if so, return a HTTP 501 to the client.
2299 * FIXME: I have a stupid name and I should handle the
2300 * requests properly. Still, what we do here is rfc-
2301 * compliant, whereas ignoring or forwarding are not.
2304 * 1 : csp = Current client state (buffers, headers, etc...)
2306 * Returns : http_response if , NULL if nonmatch or handler fail
2308 *********************************************************************/
2309 struct http_response *direct_response(struct client_state *csp)
2311 struct http_response *rsp;
2312 struct list_entry *p;
2314 if ((0 == strcmpic(csp->http->gpc, "trace"))
2315 || (0 == strcmpic(csp->http->gpc, "options")))
2317 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2319 if (!strncmpic("Max-Forwards:", p->str, 13))
2321 unsigned int max_forwards;
2324 * If it's a Max-Forwards value of zero,
2325 * we have to intercept the request.
2327 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2330 * FIXME: We could handle at least TRACE here,
2331 * but that would require a verbatim copy of
2332 * the request which we don't have anymore
2334 log_error(LOG_LEVEL_HEADER,
2335 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2338 /* Get mem for response or fail*/
2339 if (NULL == (rsp = alloc_http_response()))
2341 return cgi_error_memory();
2344 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2346 free_http_response(rsp);
2347 return cgi_error_memory();
2351 rsp->reason = RSP_REASON_UNSUPPORTED;
2353 return(finish_http_response(csp, rsp));