1 const char filters_rcs[] = "$Id: filters.c,v 1.78 2007/01/28 13:41:18 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, 2004-2006 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 1.78 2007/01/28 13:41:18 fabiankeil
44 * - Add HEAD support to finish_http_response.
45 * - Add error favicon to internal HTML error messages.
47 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
48 * Mark *csp as immutable for is_untrusted_url()
49 * and is_imageurl(). Closes FR 1237736.
51 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
52 * Integrate a modified version of Wil Mahan's
53 * zlib patch (PR #895531).
55 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
56 * Fixed gcc43 conversion warnings,
57 * changed sprintf calls to snprintf.
59 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
60 * Adjust comment in pcrs_filter_response()
61 * to recent pcrs changes. Hohoho.
63 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
64 * Don't crash if pcre returns an error code
65 * that pcrs didn't expect. Fixes BR 1621173.
67 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
68 * Modified is_untrusted_url to complain in case of
69 * write errors and to give a reason when adding new
70 * entries to the trustfile. Closes FR 1097611.
72 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
73 * Skip empty filter files in pcrs_filter_response,
74 * but don't ignore the ones that come afterwards.
75 * Fixes parts of BR 1619208.
77 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
78 * Added some sanity checks for get_last_url().
79 * Fixed possible segfault caused by my last commit.
81 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
82 * Let get_last_url() catch https URLs as well.
84 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
85 * Make sure get_last_url() behaves like advertised
86 * and fast-redirects{} can be combined with redirect{}.
88 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
89 * Implemented +redirect{s@foo@bar@} to generate
90 * a redirect based on a rewritten version of the
93 * Revision 1.66 2006/09/23 13:26:38 roro
94 * Replace TABs by spaces in source code.
96 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
97 * Fix +redirect{}. Didn't work with -fast-redirects.
99 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
100 * Block requests for untrusted URLs with status
101 * code 403 instead of 200.
103 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
104 * Don't free p which is still in use and will be later
105 * freed by free_map(). Don't claim the referrer is unknown
106 * when the client didn't set one.
108 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
109 * Feature request 595948: Re-Filter logging in single line
111 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
112 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
114 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
115 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
116 * with what was really the latest development (the v_3_0_branch branch)
118 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
119 * Multiple filter file support
121 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
122 * A few memory leaks plugged (mostly on error paths)
124 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
125 * Add the ability to check jpeg images for invalid
126 * lengths of comment blocks. Defensive strategy
127 * against the exploit:
128 * Microsoft Security Bulletin MS04-028
129 * Buffer Overrun in JPEG Processing (GDI+) Could
130 * Allow Code Execution (833987)
131 * Enabled with +inspect-jpegs in actions files.
133 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
134 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
136 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
137 * Fixed bug #839859: "See why" link URL now gets url-encoded.
139 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
142 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
143 * Added basic support for OPTIONS and TRACE HTTP methods:
144 * New function direct_response which handles OPTIONS and
145 * TRACE requests whose Max-Forwards header field is zero.
147 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
148 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
150 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
151 * - Bugfix: Executing a filters without jobs no longer results in
152 * turing off *all* filters.
153 * - Security fix: Malicious web servers can't cause a seg fault
154 * through bogus chunk sizes anymore
156 * Revision 1.58 2002/04/24 02:11:17 oes
157 * Jon's multiple AF patch: url_actions now evaluates rules
160 * Revision 1.57 2002/04/08 20:38:34 swa
163 * Revision 1.56 2002/04/05 15:51:24 oes
164 * - bugfix: error-pages now get correct request protocol
165 * - fix for invalid HTML in trust info
167 * Revision 1.55 2002/04/02 16:13:51 oes
168 * Fix: No "Go there anyway" for SSL
170 * Revision 1.54 2002/04/02 14:55:56 oes
171 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
173 * Revision 1.53 2002/03/26 22:29:54 swa
174 * we have a new homepage!
176 * Revision 1.52 2002/03/24 16:35:57 jongfoster
179 * Revision 1.51 2002/03/24 15:23:33 jongfoster
182 * Revision 1.50 2002/03/24 13:25:43 swa
183 * name change related issues
185 * Revision 1.49 2002/03/16 20:29:14 oes
188 * Revision 1.48 2002/03/13 20:25:34 oes
189 * Better logging for content filters
191 * Revision 1.47 2002/03/13 00:30:52 jongfoster
193 * Added option of always sending redirect for imageblock,
194 * currently disabled with #if 0.
196 * Revision 1.46 2002/03/12 01:42:49 oes
197 * Introduced modular filters
199 * Revision 1.45 2002/03/08 16:47:50 oes
200 * Added choice beween GIF and PNG built-in images
202 * Revision 1.44 2002/03/07 03:49:31 oes
203 * - Fixed compiler warnings etc
204 * - Changed built-in images from GIF to PNG
205 * (with regard to Unisys patent issue)
206 * - Added a 4x4 pattern PNG which is less intrusive
207 * than the logo but also clearly marks the deleted banners
209 * Revision 1.43 2002/01/22 23:51:59 jongfoster
210 * Replacing strsav() with the safer string_append().
212 * Adding missing html_encode() to error message generators. Where encoded
213 * and unencoded versions of a string were provided, removing the unencoded
216 * Revision 1.42 2002/01/17 21:00:32 jongfoster
217 * Moving all our URL and URL pattern parsing code to urlmatch.c.
219 * Using a single, simple url_match(pattern,url) function - rather than
220 * the 3-line match routine which was repeated all over the place.
222 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
224 * Using parse_http_url() to parse URLs without faking a HTTP
225 * request line for parse_http_request().
227 * Revision 1.41 2001/11/13 00:14:07 jongfoster
228 * Fixing stupid bug now I've figured out what || means.
229 * (It always returns 0 or 1, not one of it's paramaters.)
231 * Revision 1.40 2001/10/26 17:37:55 oes
232 * - Re-enabled Netscape 200/404 bug workaround in block_url():
233 * - Removed OS/2 special case
234 * - Made block_url() independant from sed() having been run
235 * - Made trust_url independant from sed() having been run
236 * - Made is_imageurl independant from sed() having been run.
237 * It now checks User-Agent: and Accept: by itself.
240 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
241 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
242 * threads to call select() simultaneously. So, it's time to do a real, live,
243 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
244 * (native). Both versions will work, but using __OS2__ offers multi-threading.
246 * Revision 1.38 2001/10/23 21:32:33 jongfoster
247 * Adding error-checking to selected functions
249 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
250 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
251 * filters.c. Added a FIXME in front of the offending code. I'll gladly
252 * put in a better/more robust fix for all parties if one is presented...
253 * It seems that just returning 200 instead of 404 would pretty much fix
254 * it for everyone, but I don't know all the history of the problem.
256 * Revision 1.36 2001/10/10 16:44:16 oes
257 * Added match_portlist function
259 * Revision 1.35 2001/10/07 15:41:23 oes
260 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
262 * New function remove_chunked_transfer_coding that strips chunked
263 * transfer coding to plain and is called by pcrs_filter_response
264 * and gif_deanimate_response if neccessary
266 * Improved handling of zero-change re_filter runs
268 * pcrs_filter_response and gif_deanimate_response now remove
269 * chunked transfer codeing before processing the body.
271 * Revision 1.34 2001/09/20 15:49:36 steudten
273 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
274 * See cgi.c fill_template().
276 * Revision 1.33 2001/09/16 17:05:14 jongfoster
277 * Removing unused #include showarg.h
279 * Revision 1.32 2001/09/16 13:21:27 jongfoster
280 * Changes to use new list functions.
282 * Revision 1.31 2001/09/16 11:38:02 jongfoster
283 * Splitting fill_template() into 2 functions:
284 * template_load() loads the file
285 * template_fill() performs the PCRS regexps.
286 * This is because the CGI edit interface has a "table row"
287 * template which is used many times in the page - this
288 * change means it's only loaded from disk once.
290 * Revision 1.30 2001/09/16 11:00:10 jongfoster
291 * New function alloc_http_response, for symmetry with free_http_response
293 * Revision 1.29 2001/09/13 23:32:40 jongfoster
294 * Moving image data to cgi.c rather than cgi.h
295 * Fixing a GPF under Win32 (and any other OS that protects global
296 * constants from being written to).
298 * Revision 1.28 2001/09/10 10:18:51 oes
299 * Silenced compiler warnings
301 * Revision 1.27 2001/08/05 16:06:20 jongfoster
302 * Modifiying "struct map" so that there are now separate header and
303 * "map_entry" structures. This means that functions which modify a
304 * map no longer need to return a pointer to the modified map.
305 * Also, it no longer reverses the order of the entries (which may be
306 * important with some advanced template substitutions).
308 * Revision 1.26 2001/07/30 22:08:36 jongfoster
309 * Tidying up #defines:
310 * - All feature #defines are now of the form FEATURE_xxx
311 * - Permanently turned off WIN_GUI_EDIT
312 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
314 * Revision 1.25 2001/07/26 10:09:46 oes
315 * Made browser detection a little less naive
317 * Revision 1.24 2001/07/25 17:22:51 oes
318 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
320 * Revision 1.23 2001/07/23 13:40:12 oes
321 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
323 * Revision 1.22 2001/07/18 12:29:34 oes
324 * - Made gif_deanimate_response respect
325 * csp->action->string[ACTION_STRING_DEANIMATE]
326 * - Logging cosmetics
328 * Revision 1.21 2001/07/13 13:59:53 oes
329 * - Introduced gif_deanimate_response which shares the
330 * generic content modification interface of pcrs_filter_response
331 * and acts as a wrapper to deanimate.c:gif_deanimate()
332 * - Renamed re_process_buffer to pcrs_filter_response
333 * - pcrs_filter_response now returns NULL on failiure
334 * - Removed all #ifdef PCRS
336 * Revision 1.20 2001/07/01 17:01:04 oes
337 * Added comments and missing return statement in is_untrusted_url()
339 * Revision 1.19 2001/06/29 21:45:41 oes
340 * Indentation, CRLF->LF, Tab-> Space
342 * Revision 1.18 2001/06/29 13:27:38 oes
343 * - Cleaned up, renamed and reorderd functions
344 * and improved comments
347 * - Ported to CGI platform. Now delivers
348 * http_response or NULL
349 * - Unified HTML and GIF generation (moved image detection
350 * and GIF generation here from jcc.c:chat())
351 * - Fixed HTTP status to:
352 * - 403 (Forbidden) for the "blocked" HTML message
353 * - 200 (OK) for GIF answers
354 * - 302 (Redirect) for redirect to GIF
357 * - Ported to CGI platform. Now delivers
358 * http_response or NULL
359 * - Separated detection of untrusted URL into
360 * (bool)is_untrusted_url
361 * - Added enforcement of untrusted requests
363 * - Moved redirect_url() from cgi.c to here
364 * and ported it to the CGI platform
366 * - Removed logentry from cancelled commit
368 * Revision 1.17 2001/06/09 10:55:28 jongfoster
369 * Changing BUFSIZ ==> BUFFER_SIZE
371 * Revision 1.16 2001/06/07 23:10:26 jongfoster
372 * Allowing unanchored domain patterns to back off and retry
373 * if they partially match. Optimized right-anchored patterns.
374 * Moving ACL and forward files into config file.
375 * Replacing struct gateway with struct forward_spec
377 * Revision 1.15 2001/06/03 19:12:00 oes
378 * extracted-CGI relevant stuff
380 * Revision 1.14 2001/06/01 10:30:55 oes
381 * Added optional left-anchoring to domaincmp
383 * Revision 1.13 2001/05/31 21:21:30 jongfoster
384 * Permissionsfile / actions file changes:
385 * - Changed "permission" to "action" throughout
386 * - changes to file format to allow string parameters
387 * - Moved helper functions to actions.c
389 * Revision 1.12 2001/05/31 17:35:20 oes
391 * - Enhanced domain part globbing with infix and prefix asterisk
392 * matching and optional unanchored operation
394 * Revision 1.11 2001/05/29 11:53:23 oes
395 * "See why" link added to "blocked" page
397 * Revision 1.10 2001/05/29 09:50:24 jongfoster
398 * Unified blocklist/imagelist/permissionslist.
399 * File format is still under discussion, but the internal changes
402 * Also modified interceptor behaviour:
403 * - We now intercept all URLs beginning with one of the following
404 * prefixes (and *only* these prefixes):
406 * * http://ijbswa.sf.net/config/
407 * * http://ijbswa.sourceforge.net/config/
408 * - New interceptors "home page" - go to http://i.j.b/ to see it.
409 * - Internal changes so that intercepted and fast redirect pages
410 * are not replaced with an image.
411 * - Interceptors now have the option to send a binary page direct
412 * to the client. (i.e. ijb-send-banner uses this)
413 * - Implemented show-url-info interceptor. (Which is why I needed
414 * the above interceptors changes - a typical URL is
415 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
416 * The previous mechanism would not have intercepted that, and
417 * if it had been intercepted then it then it would have replaced
420 * Revision 1.9 2001/05/27 22:17:04 oes
422 * - re_process_buffer no longer writes the modified buffer
423 * to the client, which was very ugly. It now returns the
424 * buffer, which it is then written by chat.
426 * - content_length now adjusts the Content-Length: header
427 * for modified documents rather than crunch()ing it.
428 * (Length info in csp->content_length, which is 0 for
429 * unmodified documents)
431 * - For this to work, sed() is called twice when filtering.
433 * Revision 1.8 2001/05/26 17:13:28 jongfoster
434 * Filled in a function comment.
436 * Revision 1.7 2001/05/26 15:26:15 jongfoster
437 * ACL feature now provides more security by immediately dropping
438 * connections from untrusted hosts.
440 * Revision 1.6 2001/05/26 00:28:36 jongfoster
441 * Automatic reloading of config file.
442 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
443 * Most of the global variables have been moved to a new
444 * struct configuration_spec, accessed through csp->config->globalname
445 * Most of the globals remaining are used by the Win32 GUI.
447 * Revision 1.5 2001/05/25 22:34:30 jongfoster
450 * Revision 1.4 2001/05/22 18:46:04 oes
452 * - Enabled filtering banners by size rather than URL
453 * by adding patterns that replace all standard banner
454 * sizes with the "Junkbuster" gif to the re_filterfile
456 * - Enabled filtering WebBugs by providing a pattern
457 * which kills all 1x1 images
459 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
460 * which is selected by the (nonstandard and therefore
461 * capital) letter 'U' in the option string.
462 * It causes the quantifiers to be ungreedy by default.
463 * Appending a ? turns back to greedy (!).
465 * - Added a new interceptor ijb-send-banner, which
466 * sends back the "Junkbuster" gif. Without imagelist or
467 * MSIE detection support, or if tinygif = 1, or the
468 * URL isn't recognized as an imageurl, a lame HTML
469 * explanation is sent instead.
471 * - Added new feature, which permits blocking remote
472 * script redirects and firing back a local redirect
474 * The feature is conditionally compiled, i.e. it
475 * can be disabled with --disable-fast-redirects,
476 * plus it must be activated by a "fast-redirects"
477 * line in the config file, has its own log level
478 * and of course wants to be displayed by show-proxy-args
479 * Note: Boy, all the #ifdefs in 1001 locations and
480 * all the fumbling with configure.in and acconfig.h
481 * were *way* more work than the feature itself :-(
483 * - Because a generic redirect template was needed for
484 * this, tinygif = 3 now uses the same.
486 * - Moved GIFs, and other static HTTP response templates
491 * - Removed some >400 CRs again (Jon, you really worked
494 * Revision 1.3 2001/05/20 16:44:47 jongfoster
495 * Removing last hardcoded Junkbusters.com URLs.
497 * Revision 1.2 2001/05/20 01:21:20 jongfoster
498 * Version 2.9.4 checkin.
499 * - Merged popupfile and cookiefile, and added control over PCRS
500 * filtering, in new "permissionsfile".
501 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
502 * file error you now get a message box (in the Win32 GUI) rather
503 * than the program exiting with no explanation.
504 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
506 * - Removed tabs from "config"
507 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
508 * - Bumped up version number.
510 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
511 * Initial import of version 2.9.3 source tree
514 *********************************************************************/
520 #include <sys/types.h>
529 #endif /* ndef __OS2__ */
530 #include <netinet/in.h>
532 #include <winsock2.h>
533 #endif /* ndef _WIN32 */
537 #endif /* def __OS2__ */
545 #include "jbsockets.h"
546 #include "miscutil.h"
550 #include "deanimate.h"
551 #include "urlmatch.h"
557 const char filters_h_rcs[] = FILTERS_H_VERSION;
559 /* Fix a problem with Solaris. There should be no effect on other
561 * Solaris's isspace() is a macro which uses it's argument directly
562 * as an array index. Therefore we need to make sure that high-bit
563 * characters generate +ve values, and ideally we also want to make
564 * the argument match the declared parameter type of "int".
566 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
570 /*********************************************************************
572 * Function : block_acl
574 * Description : Block this request?
575 * Decide yes or no based on ACL file.
578 * 1 : dst = The proxy or gateway address this is going to.
579 * Or NULL to check all possible targets.
580 * 2 : csp = Current client state (buffers, headers, etc...)
581 * Also includes the client IP address.
583 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
585 *********************************************************************/
586 int block_acl(struct access_control_addr *dst, struct client_state *csp)
588 struct access_control_list *acl = csp->config->acl;
590 /* if not using an access control list, then permit the connection */
596 /* search the list */
599 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
603 /* Just want to check if they have any access */
604 if (acl->action == ACL_PERMIT)
609 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
610 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
612 if (acl->action == ACL_PERMIT)
630 /*********************************************************************
632 * Function : acl_addr
634 * Description : Called from `load_config' to parse an ACL address.
637 * 1 : aspec = String specifying ACL address.
638 * 2 : aca = struct access_control_addr to fill in.
640 * Returns : 0 => Ok, everything else is an error.
642 *********************************************************************/
643 int acl_addr(char *aspec, struct access_control_addr *aca)
645 int i, masklength, port;
651 if ((p = strchr(aspec, '/')) != NULL)
655 if (ijb_isdigit(*p) == 0)
659 masklength = atoi(p);
662 if ((masklength < 0) || (masklength > 32))
667 if ((p = strchr(aspec, ':')) != NULL)
671 if (ijb_isdigit(*p) == 0)
680 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
682 if (aca->addr == INADDR_NONE)
687 /* build the netmask */
689 for (i=1; i <= masklength ; i++)
691 aca->mask |= (1 << (32 - i));
694 /* now mask off the host portion of the ip address
695 * (i.e. save on the network portion of the address).
697 aca->addr = aca->addr & aca->mask;
702 #endif /* def FEATURE_ACL */
705 /*********************************************************************
707 * Function : match_portlist
709 * Description : Check if a given number is covered by a comma
710 * separated list of numbers and ranges (a,b-c,d,..)
713 * 1 : portlist = String with list
714 * 2 : port = port to check
716 * Returns : 0 => no match
719 *********************************************************************/
720 int match_portlist(const char *portlist, int port)
722 char *min, *max, *next, *portlist_copy;
724 min = next = portlist_copy = strdup(portlist);
727 * Zero-terminate first item and remember offset for next
729 if (NULL != (next = strchr(portlist_copy, (int) ',')))
735 * Loop through all items, checking for match
739 if (NULL == (max = strchr(min, (int) '-')))
742 * No dash, check for equality
744 if (port == atoi(min))
753 * This is a range, so check if between min and max,
754 * or, if max was omitted, between min and 65K
757 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
771 * Zero-terminate next item and remember offset for n+1
773 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
785 /*********************************************************************
787 * Function : block_url
789 * Description : Called from `chat'. Check to see if we need to block this.
792 * 1 : csp = Current client state (buffers, headers, etc...)
794 * Returns : NULL => unblocked, else HTTP block response
796 *********************************************************************/
797 struct http_response *block_url(struct client_state *csp)
799 struct http_response *rsp;
800 const char *new_content_type = NULL;
803 * If it's not blocked, don't block it ;-)
805 if ((csp->action->flags & ACTION_BLOCK) == 0)
809 if (csp->action->flags & ACTION_REDIRECT)
811 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
814 * Else, prepare a response
816 if (NULL == (rsp = alloc_http_response()))
818 return cgi_error_memory();
822 * If it's an image-url, send back an image or redirect
823 * as specified by the relevant +image action
825 #ifdef FEATURE_IMAGE_BLOCKING
826 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
830 /* determine HOW images should be blocked */
831 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
833 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
835 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
837 #if 1 /* Two alternative strategies, use this one for now: */
839 /* and handle accordingly: */
840 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
842 rsp->status = strdup("403 Request blocked by Privoxy");
843 if (rsp->status == NULL)
845 free_http_response(rsp);
846 return cgi_error_memory();
848 rsp->body = bindup(image_pattern_data, image_pattern_length);
849 if (rsp->body == NULL)
851 free_http_response(rsp);
852 return cgi_error_memory();
854 rsp->content_length = image_pattern_length;
856 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
858 free_http_response(rsp);
859 return cgi_error_memory();
863 else if (0 == strcmpic(p, "blank"))
865 rsp->status = strdup("403 Request blocked by Privoxy");
866 if (rsp->status == NULL)
868 free_http_response(rsp);
869 return cgi_error_memory();
871 rsp->body = bindup(image_blank_data, image_blank_length);
872 if (rsp->body == NULL)
874 free_http_response(rsp);
875 return cgi_error_memory();
877 rsp->content_length = image_blank_length;
879 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
881 free_http_response(rsp);
882 return cgi_error_memory();
888 rsp->status = strdup("302 Local Redirect from Privoxy");
889 if (rsp->status == NULL)
891 free_http_response(rsp);
892 return cgi_error_memory();
895 if (enlist_unique_header(rsp->headers, "Location", p))
897 free_http_response(rsp);
898 return cgi_error_memory();
902 #else /* Following code is disabled for now */
904 /* and handle accordingly: */
905 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
907 p = CGI_PREFIX "send-banner?type=pattern";
909 else if (0 == strcmpic(p, "blank"))
911 p = CGI_PREFIX "send-banner?type=blank";
913 rsp->status = strdup("302 Local Redirect from Privoxy");
914 if (rsp->status == NULL)
916 free_http_response(rsp);
917 return cgi_error_memory();
920 if (enlist_unique_header(rsp->headers, "Location", p))
922 free_http_response(rsp);
923 return cgi_error_memory();
925 #endif /* Preceeding code is disabled for now */
927 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
930 * Send empty document.
932 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
935 rsp->body = strdup(" ");
936 rsp->content_length = 1;
938 rsp->status = strdup("403 Request blocked by Privoxy");
939 if (rsp->status == NULL)
941 free_http_response(rsp);
942 return cgi_error_memory();
944 if (new_content_type != 0)
946 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
947 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
949 free_http_response(rsp);
950 return cgi_error_memory();
955 #endif /* def FEATURE_IMAGE_BLOCKING */
958 * Else, generate an HTML "blocked" message:
962 struct map * exports;
966 * Workaround for stupid Netscape bug which prevents
967 * pages from being displayed if loading a referenced
968 * JavaScript or style sheet fails. So make it appear
969 * as if it succeeded.
971 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
972 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
973 && !strstr(p, "Gecko") /* save Mozilla, */
974 && !strstr(p, "compatible") /* MSIE */
975 && !strstr(p, "Opera")) /* and Opera. */
977 rsp->status = strdup("200 Request for blocked URL");
981 rsp->status = strdup("404 Request for blocked URL");
984 if (rsp->status == NULL)
986 free_http_response(rsp);
987 return cgi_error_memory();
990 exports = default_exports(csp, NULL);
993 free_http_response(rsp);
994 return cgi_error_memory();
997 #ifdef FEATURE_FORCE_LOAD
998 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
999 if (csp->http->ssl != 0)
1000 #endif /* ndef FEATURE_FORCE_LOAD */
1002 err = map_block_killer(exports, "force-support");
1005 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1006 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1007 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1008 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1013 free_http_response(rsp);
1014 return cgi_error_memory();
1017 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1020 free_http_response(rsp);
1021 return cgi_error_memory();
1025 return finish_http_response(csp, rsp);
1030 #ifdef FEATURE_TRUST
1031 /*********************************************************************
1033 * Function : trust_url FIXME: I should be called distrust_url
1035 * Description : Calls is_untrusted_url to determine if the URL is trusted
1036 * and if not, returns a HTTP 403 response with a reject message.
1039 * 1 : csp = Current client state (buffers, headers, etc...)
1041 * Returns : NULL => trusted, else http_response.
1043 *********************************************************************/
1044 struct http_response *trust_url(struct client_state *csp)
1046 struct http_response *rsp;
1047 struct map * exports;
1048 char buf[BUFFER_SIZE];
1050 struct url_spec **tl;
1055 * Don't bother to work on trusted URLs
1057 if (!is_untrusted_url(csp))
1063 * Else, prepare a response:
1065 if (NULL == (rsp = alloc_http_response()))
1067 return cgi_error_memory();
1070 rsp->status = strdup("403 Request blocked by Privoxy");
1071 exports = default_exports(csp, NULL);
1072 if (exports == NULL || rsp->status == NULL)
1074 free_http_response(rsp);
1075 return cgi_error_memory();
1079 * Export the protocol, host, port, and referrer information
1081 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1082 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1083 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1085 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1087 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1091 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1097 free_http_response(rsp);
1098 return cgi_error_memory();
1102 * Export the trust list
1105 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1107 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1108 string_append(&p, buf);
1110 err = map(exports, "trusted-referrers", 1, p, 0);
1115 free_http_response(rsp);
1116 return cgi_error_memory();
1120 * Export the trust info, if available
1122 if (csp->config->trust_info->first)
1124 struct list_entry *l;
1127 for (l = csp->config->trust_info->first; l ; l = l->next)
1129 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1130 string_append(&p, buf);
1132 err = map(exports, "trust-info", 1, p, 0);
1136 err = map_block_killer(exports, "have-trust-info");
1142 free_http_response(rsp);
1143 return cgi_error_memory();
1147 * Export the force prefix or the force conditional block killer
1149 #ifdef FEATURE_FORCE_LOAD
1150 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1151 #else /* ifndef FEATURE_FORCE_LOAD */
1152 err = map_block_killer(exports, "force-support");
1153 #endif /* ndef FEATURE_FORCE_LOAD */
1158 free_http_response(rsp);
1159 return cgi_error_memory();
1163 * Build the response
1165 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1168 free_http_response(rsp);
1169 return cgi_error_memory();
1172 return finish_http_response(csp, rsp);
1174 #endif /* def FEATURE_TRUST */
1176 /*********************************************************************
1178 * Function : execute_single_pcrs_command
1180 * Description : Apply single pcrs command to the subject.
1181 * The subject itself is left untouched, memory for the result
1182 * is malloc()ed and it is the caller's responsibility to free
1183 * the result when it's no longer needed.
1186 * 1 : subject = the subject (== original) string
1187 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1188 * 3 : hits = int* for returning the number of modifications
1190 * Returns : NULL in case of errors, otherwise the
1191 * result of the pcrs command.
1193 *********************************************************************/
1194 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1198 char *result = NULL;
1202 assert(pcrs_command);
1205 size = strlen(subject);
1207 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1209 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1210 pcrs_command, error);
1212 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1214 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1221 job = pcrs_free_job(job);
1228 /*********************************************************************
1230 * Function : rewrite_url
1232 * Description : Rewrites a URL with a single pcrs command
1233 * and returns the result if it differs from the
1234 * original and isn't obviously invalid.
1237 * 1 : old_url = URL to rewrite.
1238 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1241 * Returns : NULL if the pcrs_command didn't change the url, or
1242 * the result of the modification.
1244 *********************************************************************/
1245 char *rewrite_url(char *old_url, const char *pcrs_command)
1247 char *new_url = NULL;
1251 assert(pcrs_command);
1253 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1257 log_error(LOG_LEVEL_REDIRECTS,
1258 "pcrs command \"%s\" didn't change \"%s\".",
1259 pcrs_command, old_url, new_url);
1262 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1264 log_error(LOG_LEVEL_ERROR,
1265 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1266 "but the result doesn't look like a valid URL and will be ignored.",
1267 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1272 log_error(LOG_LEVEL_REDIRECTS,
1273 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1274 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1282 #ifdef FEATURE_FAST_REDIRECTS
1283 /*********************************************************************
1285 * Function : get_last_url
1287 * Description : Search for the last URL inside a string.
1288 * If the string already is a URL, it will
1289 * be the first URL found.
1292 * 1 : subject = the string to check
1293 * 2 : redirect_mode = +fast-redirect{} mode
1295 * Returns : NULL if no URL was found, or
1296 * the last URL found.
1298 *********************************************************************/
1299 char *get_last_url(char *subject, const char *redirect_mode)
1301 char *new_url = NULL;
1305 assert(redirect_mode);
1307 subject = strdup(subject);
1308 if (subject == NULL)
1310 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1314 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1316 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1317 new_url = url_decode(subject);
1318 if (new_url != NULL)
1325 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1329 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1332 * Find the last URL encoded in the request
1335 while ((tmp = strstr(tmp, "http://")) != NULL)
1339 tmp = (new_url != NULL) ? new_url : subject;
1340 while ((tmp = strstr(tmp, "https://")) != NULL)
1345 if ((new_url != NULL)
1346 && ( (new_url != subject)
1347 || (0 == strncmpic(subject, "http://", 7))
1348 || (0 == strncmpic(subject, "https://", 8))
1352 * Return new URL if we found a redirect
1353 * or if the subject already was a URL.
1355 * The second case makes sure that we can
1356 * chain get_last_url after another redirection check
1357 * (like rewrite_url) without losing earlier redirects.
1359 new_url = strdup(new_url);
1368 #endif /* def FEATURE_FAST_REDIRECTS */
1371 /*********************************************************************
1373 * Function : redirect_url
1375 * Description : Checks if Privoxy should answer the request with
1376 * a HTTP redirect and generates the redirect if
1380 * 1 : csp = Current client state (buffers, headers, etc...)
1382 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1384 *********************************************************************/
1385 struct http_response *redirect_url(struct client_state *csp)
1387 struct http_response *rsp;
1388 #ifdef FEATURE_FAST_REDIRECTS
1390 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1391 * as compile-time option? The user can easily disable
1392 * it in his action file.
1394 char * redirect_mode;
1395 #endif /* def FEATURE_FAST_REDIRECTS */
1396 char *old_url = NULL;
1397 char *new_url = NULL;
1398 char *redirection_string;
1400 if ((csp->action->flags & ACTION_REDIRECT))
1402 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1405 * If the redirection string begins with 's',
1406 * assume it's a pcrs command, otherwise treat it as
1407 * properly formatted URL and use it for the redirection
1410 * According to RFC 2616 section 14.30 the URL
1411 * has to be absolute and if the user tries:
1412 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1413 * she would get undefined results anyway.
1417 if (*redirection_string == 's')
1419 old_url = csp->http->url;
1420 new_url = rewrite_url(old_url, redirection_string);
1424 log_error(LOG_LEVEL_REDIRECTS,
1425 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1426 redirection_string);
1427 new_url = strdup(redirection_string);
1431 #ifdef FEATURE_FAST_REDIRECTS
1432 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1434 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1437 * If it exists, use the previously rewritten URL as input
1438 * otherwise just use the old path.
1440 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1441 new_url = get_last_url(old_url, redirect_mode);
1444 #endif /* def FEATURE_FAST_REDIRECTS */
1446 /* Did any redirect action trigger? */
1449 if (0 == strcmpic(new_url, csp->http->url))
1451 log_error(LOG_LEVEL_ERROR,
1452 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1453 csp->http->url, new_url);
1458 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1460 if (NULL == (rsp = alloc_http_response()))
1463 return cgi_error_memory();
1466 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1467 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1470 free_http_response(rsp);
1471 return cgi_error_memory();
1474 return finish_http_response(csp, rsp);
1478 /* Only reached if no redirect is required */
1484 #ifdef FEATURE_IMAGE_BLOCKING
1485 /*********************************************************************
1487 * Function : is_imageurl
1489 * Description : Given a URL, decide whether it is an image or not,
1490 * using either the info from a previous +image action
1491 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1492 * is MSIE and not on a Mac, tell from the browser's accept
1496 * 1 : csp = Current client state (buffers, headers, etc...)
1498 * Returns : True (nonzero) if URL is an image, false (0)
1501 *********************************************************************/
1502 int is_imageurl(const struct client_state *csp)
1504 #ifdef FEATURE_IMAGE_DETECT_MSIE
1507 tmp = get_header_value(csp->headers, "User-Agent:");
1508 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1510 tmp = get_header_value(csp->headers, "Accept:");
1511 if (tmp && strstr(tmp, "image/gif"))
1513 /* Client will accept HTML. If this seems counterintuitive,
1523 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1525 return ((csp->action->flags & ACTION_IMAGE) != 0);
1528 #endif /* def FEATURE_IMAGE_BLOCKING */
1531 #ifdef FEATURE_TRUST
1532 /*********************************************************************
1534 * Function : is_untrusted_url
1536 * Description : Should we "distrust" this URL (and block it)?
1538 * Yes if it matches a line in the trustfile, or if the
1539 * referrer matches a line starting with "+" in the
1544 * 1 : csp = Current client state (buffers, headers, etc...)
1546 * Returns : 0 => trusted, 1 => untrusted
1548 *********************************************************************/
1549 int is_untrusted_url(const struct client_state *csp)
1551 struct file_list *fl;
1552 struct block_spec *b;
1553 struct url_spec **trusted_url;
1554 struct http_request rhttp[1];
1555 const char * referer;
1559 * If we don't have a trustlist, we trust everybody
1561 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1566 memset(rhttp, '\0', sizeof(*rhttp));
1569 * Do we trust the request URL itself?
1571 for (b = b->next; b ; b = b->next)
1573 if (url_match(b->url, csp->http))
1579 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1581 /* no referrer was supplied */
1587 * If not, do we maybe trust its referrer?
1589 err = parse_http_url(referer, rhttp, csp);
1595 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1597 if (url_match(*trusted_url, rhttp))
1599 /* if the URL's referrer is from a trusted referrer, then
1600 * add the target spec to the trustfile as an unblocked
1601 * domain and return 0 (which means it's OK).
1606 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1610 char * new_entry = strdup("~");
1612 string_append(&new_entry, csp->http->hostport);
1614 path = csp->http->path;
1615 if ( (path[0] == '/')
1617 && ((path_end = strchr(path + 2, '/')) != NULL))
1619 /* since this path points into a user's home space
1620 * be sure to include this spec in the trustfile.
1622 int path_len = path_end - path; /* save offset */
1623 path = strdup(path); /* Copy string */
1626 path_end = path + path_len; /* regenerate ptr to new buffer */
1627 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1629 string_join(&new_entry, path);
1633 * Give a reason for generating this entry.
1635 string_append(&new_entry, " # Trusted referrer was: ");
1636 string_append(&new_entry, referer);
1638 if (new_entry != NULL)
1640 if (-1 == fprintf(fp, "%s\n", new_entry))
1642 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1643 new_entry, csp->config->trustfile);
1649 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1650 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1657 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1658 csp->http->hostport, csp->config->trustfile);
1665 #endif /* def FEATURE_TRUST */
1668 /*********************************************************************
1670 * Function : pcrs_filter_response
1672 * Description : Execute all text substitutions from all applying
1673 * +filter actions on the text buffer that's been accumulated
1674 * in csp->iob->buf. If this changes the contents, set
1675 * csp->content_length to the modified size and raise the
1676 * CSP_FLAG_MODIFIED flag.
1678 * XXX: Currently pcrs_filter_response is also responsible
1679 * for dechunking and decompressing. Both should be
1680 * done in separate functions so other content modifiers
1681 * profit as well, even if pcrs filtering is disabled.
1684 * 1 : csp = Current client state (buffers, headers, etc...)
1686 * Returns : a pointer to the (newly allocated) modified buffer.
1687 * or NULL if there were no hits or something went wrong
1689 *********************************************************************/
1690 char *pcrs_filter_response(struct client_state *csp)
1693 size_t size, prev_size;
1695 char *old = csp->iob->cur, *new = NULL;
1698 struct file_list *fl;
1699 struct re_filterfile_spec *b;
1700 struct list_entry *filtername;
1702 int i, found_filters = 0;
1707 if (csp->iob->cur >= csp->iob->eod)
1711 size = (size_t)(csp->iob->eod - csp->iob->cur);
1714 * Need to check the set of re_filterfiles...
1716 for (i = 0; i < MAX_AF_FILES; i++)
1729 if (0 == found_filters)
1731 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1736 * If the body has a "chunked" transfer-encoding,
1737 * get rid of it first, adjusting size and iob->eod
1739 if (csp->flags & CSP_FLAG_CHUNKED)
1741 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1742 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1746 csp->iob->eod = csp->iob->cur + size;
1747 csp->flags |= CSP_FLAG_MODIFIED;
1752 * If the body has a compressed transfer-encoding,
1753 * uncompress it first, adjusting size and iob->eod.
1754 * Note that decompression occurs after de-chunking.
1756 if (csp->content_type & CT_GZIP || csp->content_type & CT_DEFLATE)
1758 /* Notice that we at least tried to decompress. */
1759 if (JB_ERR_OK != decompress_iob(csp))
1762 * We failed to decompress the data; there's no point
1763 * in continuing since we can't filter. This is
1764 * slightly tricky because we need to remember not to
1765 * modify the Content-Encoding header later; using
1766 * CT_TABOO flag is a kludge for this purpose.
1768 csp->content_type |= CT_TABOO;
1771 log_error(LOG_LEVEL_RE_FILTER, "Decompressing successful");
1774 * Decompression gives us a completely new iob,
1775 * so we need to update.
1777 size = (size_t)(csp->iob->eod - csp->iob->cur);
1778 old = csp->iob->cur;
1780 csp->flags |= CSP_FLAG_MODIFIED;
1781 csp->content_type &= ~CT_TABOO;
1785 for (i = 0; i < MAX_AF_FILES; i++)
1788 if ((NULL == fl) || (NULL == fl->f))
1791 * Either there are no filter files
1792 * left, or this filter file just
1793 * contains no valid filters.
1795 * Continue to be sure we don't miss
1796 * valid filter files that are chained
1797 * after empty or invalid ones.
1802 * For all applying +filter actions, look if a filter by that
1803 * name exists and if yes, execute it's pcrs_joblist on the
1806 for (b = fl->f; b; b = b->next)
1808 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1809 filtername ; filtername = filtername->next)
1811 if (strcmp(b->name, filtername->str) == 0)
1813 int current_hits = 0; /* Number of hits caused by this filter */
1814 int job_number = 0; /* Which job we're currently executing */
1815 int job_hits = 0; /* How many hits the current job caused */
1817 if ( NULL == b->joblist )
1819 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1824 /* Apply all jobs from the joblist */
1825 for (job = b->joblist; NULL != job; job = job->next)
1828 job_hits = pcrs_execute(job, old, size, &new, &size);
1833 * That went well. Continue filtering
1834 * and use the result of this job as
1835 * input for the next one.
1837 current_hits += job_hits;
1838 if (old != csp->iob->cur)
1847 * This job caused an unexpected error. Inform the user
1848 * and skip the rest of the jobs in this filter. We could
1849 * continue with the next job, but usually the jobs
1850 * depend on each other or are similar enough to
1851 * fail for the same reason.
1853 * At the moment our pcrs expects the error codes of pcre 3.4,
1854 * but newer pcre versions can return additional error codes.
1855 * As a result pcrs_strerror()'s error message might be
1856 * "Unknown error ...", therefore we print the numerical value
1859 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1860 * should we use LOG_LEVEL_RE_FILTER instead?
1862 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1863 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1868 log_error(LOG_LEVEL_RE_FILTER,
1869 "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1870 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1872 hits += current_hits;
1879 * If there were no hits, destroy our copy and let
1880 * chat() use the original in csp->iob
1888 csp->flags |= CSP_FLAG_MODIFIED;
1889 csp->content_length = size;
1897 /*********************************************************************
1899 * Function : gif_deanimate_response
1901 * Description : Deanimate the GIF image that has been accumulated in
1902 * csp->iob->buf, set csp->content_length to the modified
1903 * size and raise the CSP_FLAG_MODIFIED flag.
1906 * 1 : csp = Current client state (buffers, headers, etc...)
1908 * Returns : a pointer to the (newly allocated) modified buffer.
1909 * or NULL in case something went wrong.
1911 *********************************************************************/
1912 char *gif_deanimate_response(struct client_state *csp)
1914 struct binbuffer *in, *out;
1918 size = (size_t)(csp->iob->eod - csp->iob->cur);
1921 * If the body has a "chunked" transfer-encoding,
1922 * get rid of it first, adjusting size and iob->eod
1924 if (csp->flags & CSP_FLAG_CHUNKED)
1926 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1927 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1931 csp->iob->eod = csp->iob->cur + size;
1932 csp->flags |= CSP_FLAG_MODIFIED;
1935 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1936 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1938 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1942 in->buffer = csp->iob->cur;
1945 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1947 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1954 if ((int)size == out->offset)
1956 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1960 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1962 csp->content_length = out->offset;
1963 csp->flags |= CSP_FLAG_MODIFIED;
1973 /*********************************************************************
1975 * Function : jpeg_inspect_response
1980 * 1 : csp = Current client state (buffers, headers, etc...)
1982 * Returns : a pointer to the (newly allocated) modified buffer
1983 * or NULL in case something went wrong.
1985 *********************************************************************/
1986 char *jpeg_inspect_response(struct client_state *csp)
1988 struct binbuffer *in = NULL;
1989 struct binbuffer *out = NULL;
1993 size = (size_t)(csp->iob->eod - csp->iob->cur);
1996 * If the body has a "chunked" transfer-encoding,
1997 * get rid of it first, adjusting size and iob->eod
1999 if (csp->flags & CSP_FLAG_CHUNKED)
2001 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2002 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2006 csp->iob->eod = csp->iob->cur + size;
2007 csp->flags |= CSP_FLAG_MODIFIED;
2010 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2012 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2016 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2018 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2022 in->buffer = csp->iob->cur;
2026 * Calling jpeg_inspect has the side-effect of creating and
2027 * modifying the image buffer of "out" directly.
2029 if (jpeg_inspect(in, out))
2031 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2039 csp->content_length = out->offset;
2040 csp->flags |= CSP_FLAG_MODIFIED;
2050 /*********************************************************************
2052 * Function : remove_chunked_transfer_coding
2054 * Description : In-situ remove the "chunked" transfer coding as defined
2055 * in rfc2616 from a buffer.
2058 * 1 : buffer = Pointer to the text buffer
2059 * 2 : size = Number of bytes to be processed
2061 * Returns : The new size, i.e. the number of bytes from buffer which
2062 * are occupied by the stripped body, or 0 in case something
2065 *********************************************************************/
2066 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2069 unsigned int chunksize = 0;
2070 char *from_p, *to_p;
2073 from_p = to_p = buffer;
2075 if (sscanf(buffer, "%x", &chunksize) != 1)
2077 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2081 while (chunksize > 0)
2083 if (NULL == (from_p = strstr(from_p, "\r\n")))
2085 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2089 if ((newsize += chunksize) >= size)
2091 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2096 memmove(to_p, from_p, (size_t) chunksize);
2097 to_p = buffer + newsize;
2098 from_p += chunksize + 2;
2100 if (sscanf(from_p, "%x", &chunksize) != 1)
2102 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2107 /* FIXME: Should this get its own loglevel? */
2108 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2114 /*********************************************************************
2116 * Function : url_actions
2118 * Description : Gets the actions for this URL.
2121 * 1 : http = http_request request for blocked URLs
2122 * 2 : csp = Current client state (buffers, headers, etc...)
2126 *********************************************************************/
2127 void url_actions(struct http_request *http,
2128 struct client_state *csp)
2130 struct file_list *fl;
2131 struct url_actions *b;
2134 init_current_action(csp->action);
2136 for (i = 0; i < MAX_AF_FILES; i++)
2138 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2143 apply_url_actions(csp->action, http, b);
2150 /*********************************************************************
2152 * Function : apply_url_actions
2154 * Description : Applies a list of URL actions.
2157 * 1 : action = Destination.
2158 * 2 : http = Current URL
2159 * 3 : b = list of URL actions to apply
2163 *********************************************************************/
2164 void apply_url_actions(struct current_action_spec *action,
2165 struct http_request *http,
2166 struct url_actions *b)
2170 /* Should never happen */
2174 for (b = b->next; NULL != b; b = b->next)
2176 if (url_match(b->url, http))
2178 merge_current_action(action, b->action);
2184 /*********************************************************************
2186 * Function : forward_url
2188 * Description : Should we forward this to another proxy?
2191 * 1 : http = http_request request for current URL
2192 * 2 : csp = Current client state (buffers, headers, etc...)
2194 * Returns : Pointer to forwarding information.
2196 *********************************************************************/
2197 const struct forward_spec * forward_url(struct http_request *http,
2198 struct client_state *csp)
2200 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2201 struct forward_spec *fwd = csp->config->forward;
2210 if (url_match(fwd->url, http))
2221 /*********************************************************************
2223 * Function : direct_response
2225 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2226 * request and if so, return a HTTP 501 to the client.
2228 * FIXME: I have a stupid name and I should handle the
2229 * requests properly. Still, what we do here is rfc-
2230 * compliant, whereas ignoring or forwarding are not.
2233 * 1 : csp = Current client state (buffers, headers, etc...)
2235 * Returns : http_response if , NULL if nonmatch or handler fail
2237 *********************************************************************/
2238 struct http_response *direct_response(struct client_state *csp)
2240 struct http_response *rsp;
2241 struct list_entry *p;
2243 if ((0 == strcmpic(csp->http->gpc, "trace"))
2244 || (0 == strcmpic(csp->http->gpc, "options")))
2246 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2248 if (!strncmpic("Max-Forwards:", p->str, 13))
2250 unsigned int max_forwards;
2253 * If it's a Max-Forwards value of zero,
2254 * we have to intercept the request.
2256 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2259 * FIXME: We could handle at least TRACE here,
2260 * but that would require a verbatim copy of
2261 * the request which we don't have anymore
2263 log_error(LOG_LEVEL_HEADER,
2264 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2267 /* Get mem for response or fail*/
2268 if (NULL == (rsp = alloc_http_response()))
2270 return cgi_error_memory();
2273 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2275 free_http_response(rsp);
2276 return cgi_error_memory();
2280 return(finish_http_response(csp, rsp));