1 const char filters_rcs[] = "$Id: filters.c,v 1.62 2006/08/14 00:27:47 david__schmidt 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'
15 * Copyright : Written by and Copyright (C) 2001, 2004 the SourceForge
16 * Privoxy team. http://www.privoxy.org/
18 * Based on the Internet Junkbuster originally written
19 * by and Copyright (C) 1997 Anonymous Coders and
20 * Junkbusters Corporation. http://www.junkbusters.com
22 * This program is free software; you can redistribute it
23 * and/or modify it under the terms of the GNU General
24 * Public License as published by the Free Software
25 * Foundation; either version 2 of the License, or (at
26 * your option) any later version.
28 * This program is distributed in the hope that it will
29 * be useful, but WITHOUT ANY WARRANTY; without even the
30 * implied warranty of MERCHANTABILITY or FITNESS FOR A
31 * PARTICULAR PURPOSE. See the GNU General Public
32 * License for more details.
34 * The GNU General Public License should be included with
35 * this file. If not, you can view it at
36 * http://www.gnu.org/copyleft/gpl.html
37 * or write to the Free Software Foundation, Inc., 59
38 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
42 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
43 * Feature request 595948: Re-Filter logging in single line
45 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
46 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
48 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
49 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
50 * with what was really the latest development (the v_3_0_branch branch)
52 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
53 * Multiple filter file support
55 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
56 * A few memory leaks plugged (mostly on error paths)
58 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
59 * Add the ability to check jpeg images for invalid
60 * lengths of comment blocks. Defensive strategy
61 * against the exploit:
62 * Microsoft Security Bulletin MS04-028
63 * Buffer Overrun in JPEG Processing (GDI+) Could
64 * Allow Code Execution (833987)
65 * Enabled with +inspect-jpegs in actions files.
67 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
68 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
70 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
71 * Fixed bug #839859: "See why" link URL now gets url-encoded.
73 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
76 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
77 * Added basic support for OPTIONS and TRACE HTTP methods:
78 * New function direct_response which handles OPTIONS and
79 * TRACE requests whose Max-Forwards header field is zero.
81 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
82 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
84 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
85 * - Bugfix: Executing a filters without jobs no longer results in
86 * turing off *all* filters.
87 * - Security fix: Malicious web servers can't cause a seg fault
88 * through bogus chunk sizes anymore
90 * Revision 1.58 2002/04/24 02:11:17 oes
91 * Jon's multiple AF patch: url_actions now evaluates rules
94 * Revision 1.57 2002/04/08 20:38:34 swa
97 * Revision 1.56 2002/04/05 15:51:24 oes
98 * - bugfix: error-pages now get correct request protocol
99 * - fix for invalid HTML in trust info
101 * Revision 1.55 2002/04/02 16:13:51 oes
102 * Fix: No "Go there anyway" for SSL
104 * Revision 1.54 2002/04/02 14:55:56 oes
105 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
107 * Revision 1.53 2002/03/26 22:29:54 swa
108 * we have a new homepage!
110 * Revision 1.52 2002/03/24 16:35:57 jongfoster
113 * Revision 1.51 2002/03/24 15:23:33 jongfoster
116 * Revision 1.50 2002/03/24 13:25:43 swa
117 * name change related issues
119 * Revision 1.49 2002/03/16 20:29:14 oes
122 * Revision 1.48 2002/03/13 20:25:34 oes
123 * Better logging for content filters
125 * Revision 1.47 2002/03/13 00:30:52 jongfoster
127 * Added option of always sending redirect for imageblock,
128 * currently disabled with #if 0.
130 * Revision 1.46 2002/03/12 01:42:49 oes
131 * Introduced modular filters
133 * Revision 1.45 2002/03/08 16:47:50 oes
134 * Added choice beween GIF and PNG built-in images
136 * Revision 1.44 2002/03/07 03:49:31 oes
137 * - Fixed compiler warnings etc
138 * - Changed built-in images from GIF to PNG
139 * (with regard to Unisys patent issue)
140 * - Added a 4x4 pattern PNG which is less intrusive
141 * than the logo but also clearly marks the deleted banners
143 * Revision 1.43 2002/01/22 23:51:59 jongfoster
144 * Replacing strsav() with the safer string_append().
146 * Adding missing html_encode() to error message generators. Where encoded
147 * and unencoded versions of a string were provided, removing the unencoded
150 * Revision 1.42 2002/01/17 21:00:32 jongfoster
151 * Moving all our URL and URL pattern parsing code to urlmatch.c.
153 * Using a single, simple url_match(pattern,url) function - rather than
154 * the 3-line match routine which was repeated all over the place.
156 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
158 * Using parse_http_url() to parse URLs without faking a HTTP
159 * request line for parse_http_request().
161 * Revision 1.41 2001/11/13 00:14:07 jongfoster
162 * Fixing stupid bug now I've figured out what || means.
163 * (It always returns 0 or 1, not one of it's paramaters.)
165 * Revision 1.40 2001/10/26 17:37:55 oes
166 * - Re-enabled Netscape 200/404 bug workaround in block_url():
167 * - Removed OS/2 special case
168 * - Made block_url() independant from sed() having been run
169 * - Made trust_url independant from sed() having been run
170 * - Made is_imageurl independant from sed() having been run.
171 * It now checks User-Agent: and Accept: by itself.
174 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
175 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
176 * threads to call select() simultaneously. So, it's time to do a real, live,
177 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
178 * (native). Both versions will work, but using __OS2__ offers multi-threading.
180 * Revision 1.38 2001/10/23 21:32:33 jongfoster
181 * Adding error-checking to selected functions
183 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
184 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
185 * filters.c. Added a FIXME in front of the offending code. I'll gladly
186 * put in a better/more robust fix for all parties if one is presented...
187 * It seems that just returning 200 instead of 404 would pretty much fix
188 * it for everyone, but I don't know all the history of the problem.
190 * Revision 1.36 2001/10/10 16:44:16 oes
191 * Added match_portlist function
193 * Revision 1.35 2001/10/07 15:41:23 oes
194 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
196 * New function remove_chunked_transfer_coding that strips chunked
197 * transfer coding to plain and is called by pcrs_filter_response
198 * and gif_deanimate_response if neccessary
200 * Improved handling of zero-change re_filter runs
202 * pcrs_filter_response and gif_deanimate_response now remove
203 * chunked transfer codeing before processing the body.
205 * Revision 1.34 2001/09/20 15:49:36 steudten
207 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
208 * See cgi.c fill_template().
210 * Revision 1.33 2001/09/16 17:05:14 jongfoster
211 * Removing unused #include showarg.h
213 * Revision 1.32 2001/09/16 13:21:27 jongfoster
214 * Changes to use new list functions.
216 * Revision 1.31 2001/09/16 11:38:02 jongfoster
217 * Splitting fill_template() into 2 functions:
218 * template_load() loads the file
219 * template_fill() performs the PCRS regexps.
220 * This is because the CGI edit interface has a "table row"
221 * template which is used many times in the page - this
222 * change means it's only loaded from disk once.
224 * Revision 1.30 2001/09/16 11:00:10 jongfoster
225 * New function alloc_http_response, for symmetry with free_http_response
227 * Revision 1.29 2001/09/13 23:32:40 jongfoster
228 * Moving image data to cgi.c rather than cgi.h
229 * Fixing a GPF under Win32 (and any other OS that protects global
230 * constants from being written to).
232 * Revision 1.28 2001/09/10 10:18:51 oes
233 * Silenced compiler warnings
235 * Revision 1.27 2001/08/05 16:06:20 jongfoster
236 * Modifiying "struct map" so that there are now separate header and
237 * "map_entry" structures. This means that functions which modify a
238 * map no longer need to return a pointer to the modified map.
239 * Also, it no longer reverses the order of the entries (which may be
240 * important with some advanced template substitutions).
242 * Revision 1.26 2001/07/30 22:08:36 jongfoster
243 * Tidying up #defines:
244 * - All feature #defines are now of the form FEATURE_xxx
245 * - Permanently turned off WIN_GUI_EDIT
246 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
248 * Revision 1.25 2001/07/26 10:09:46 oes
249 * Made browser detection a little less naive
251 * Revision 1.24 2001/07/25 17:22:51 oes
252 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
254 * Revision 1.23 2001/07/23 13:40:12 oes
255 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
257 * Revision 1.22 2001/07/18 12:29:34 oes
258 * - Made gif_deanimate_response respect
259 * csp->action->string[ACTION_STRING_DEANIMATE]
260 * - Logging cosmetics
262 * Revision 1.21 2001/07/13 13:59:53 oes
263 * - Introduced gif_deanimate_response which shares the
264 * generic content modification interface of pcrs_filter_response
265 * and acts as a wrapper to deanimate.c:gif_deanimate()
266 * - Renamed re_process_buffer to pcrs_filter_response
267 * - pcrs_filter_response now returns NULL on failiure
268 * - Removed all #ifdef PCRS
270 * Revision 1.20 2001/07/01 17:01:04 oes
271 * Added comments and missing return statement in is_untrusted_url()
273 * Revision 1.19 2001/06/29 21:45:41 oes
274 * Indentation, CRLF->LF, Tab-> Space
276 * Revision 1.18 2001/06/29 13:27:38 oes
277 * - Cleaned up, renamed and reorderd functions
278 * and improved comments
281 * - Ported to CGI platform. Now delivers
282 * http_response or NULL
283 * - Unified HTML and GIF generation (moved image detection
284 * and GIF generation here from jcc.c:chat())
285 * - Fixed HTTP status to:
286 * - 403 (Forbidden) for the "blocked" HTML message
287 * - 200 (OK) for GIF answers
288 * - 302 (Redirect) for redirect to GIF
291 * - Ported to CGI platform. Now delivers
292 * http_response or NULL
293 * - Separated detection of untrusted URL into
294 * (bool)is_untrusted_url
295 * - Added enforcement of untrusted requests
297 * - Moved redirect_url() from cgi.c to here
298 * and ported it to the CGI platform
300 * - Removed logentry from cancelled commit
302 * Revision 1.17 2001/06/09 10:55:28 jongfoster
303 * Changing BUFSIZ ==> BUFFER_SIZE
305 * Revision 1.16 2001/06/07 23:10:26 jongfoster
306 * Allowing unanchored domain patterns to back off and retry
307 * if they partially match. Optimized right-anchored patterns.
308 * Moving ACL and forward files into config file.
309 * Replacing struct gateway with struct forward_spec
311 * Revision 1.15 2001/06/03 19:12:00 oes
312 * extracted-CGI relevant stuff
314 * Revision 1.14 2001/06/01 10:30:55 oes
315 * Added optional left-anchoring to domaincmp
317 * Revision 1.13 2001/05/31 21:21:30 jongfoster
318 * Permissionsfile / actions file changes:
319 * - Changed "permission" to "action" throughout
320 * - changes to file format to allow string parameters
321 * - Moved helper functions to actions.c
323 * Revision 1.12 2001/05/31 17:35:20 oes
325 * - Enhanced domain part globbing with infix and prefix asterisk
326 * matching and optional unanchored operation
328 * Revision 1.11 2001/05/29 11:53:23 oes
329 * "See why" link added to "blocked" page
331 * Revision 1.10 2001/05/29 09:50:24 jongfoster
332 * Unified blocklist/imagelist/permissionslist.
333 * File format is still under discussion, but the internal changes
336 * Also modified interceptor behaviour:
337 * - We now intercept all URLs beginning with one of the following
338 * prefixes (and *only* these prefixes):
340 * * http://ijbswa.sf.net/config/
341 * * http://ijbswa.sourceforge.net/config/
342 * - New interceptors "home page" - go to http://i.j.b/ to see it.
343 * - Internal changes so that intercepted and fast redirect pages
344 * are not replaced with an image.
345 * - Interceptors now have the option to send a binary page direct
346 * to the client. (i.e. ijb-send-banner uses this)
347 * - Implemented show-url-info interceptor. (Which is why I needed
348 * the above interceptors changes - a typical URL is
349 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
350 * The previous mechanism would not have intercepted that, and
351 * if it had been intercepted then it then it would have replaced
354 * Revision 1.9 2001/05/27 22:17:04 oes
356 * - re_process_buffer no longer writes the modified buffer
357 * to the client, which was very ugly. It now returns the
358 * buffer, which it is then written by chat.
360 * - content_length now adjusts the Content-Length: header
361 * for modified documents rather than crunch()ing it.
362 * (Length info in csp->content_length, which is 0 for
363 * unmodified documents)
365 * - For this to work, sed() is called twice when filtering.
367 * Revision 1.8 2001/05/26 17:13:28 jongfoster
368 * Filled in a function comment.
370 * Revision 1.7 2001/05/26 15:26:15 jongfoster
371 * ACL feature now provides more security by immediately dropping
372 * connections from untrusted hosts.
374 * Revision 1.6 2001/05/26 00:28:36 jongfoster
375 * Automatic reloading of config file.
376 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
377 * Most of the global variables have been moved to a new
378 * struct configuration_spec, accessed through csp->config->globalname
379 * Most of the globals remaining are used by the Win32 GUI.
381 * Revision 1.5 2001/05/25 22:34:30 jongfoster
384 * Revision 1.4 2001/05/22 18:46:04 oes
386 * - Enabled filtering banners by size rather than URL
387 * by adding patterns that replace all standard banner
388 * sizes with the "Junkbuster" gif to the re_filterfile
390 * - Enabled filtering WebBugs by providing a pattern
391 * which kills all 1x1 images
393 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
394 * which is selected by the (nonstandard and therefore
395 * capital) letter 'U' in the option string.
396 * It causes the quantifiers to be ungreedy by default.
397 * Appending a ? turns back to greedy (!).
399 * - Added a new interceptor ijb-send-banner, which
400 * sends back the "Junkbuster" gif. Without imagelist or
401 * MSIE detection support, or if tinygif = 1, or the
402 * URL isn't recognized as an imageurl, a lame HTML
403 * explanation is sent instead.
405 * - Added new feature, which permits blocking remote
406 * script redirects and firing back a local redirect
408 * The feature is conditionally compiled, i.e. it
409 * can be disabled with --disable-fast-redirects,
410 * plus it must be activated by a "fast-redirects"
411 * line in the config file, has its own log level
412 * and of course wants to be displayed by show-proxy-args
413 * Note: Boy, all the #ifdefs in 1001 locations and
414 * all the fumbling with configure.in and acconfig.h
415 * were *way* more work than the feature itself :-(
417 * - Because a generic redirect template was needed for
418 * this, tinygif = 3 now uses the same.
420 * - Moved GIFs, and other static HTTP response templates
425 * - Removed some >400 CRs again (Jon, you really worked
428 * Revision 1.3 2001/05/20 16:44:47 jongfoster
429 * Removing last hardcoded Junkbusters.com URLs.
431 * Revision 1.2 2001/05/20 01:21:20 jongfoster
432 * Version 2.9.4 checkin.
433 * - Merged popupfile and cookiefile, and added control over PCRS
434 * filtering, in new "permissionsfile".
435 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
436 * file error you now get a message box (in the Win32 GUI) rather
437 * than the program exiting with no explanation.
438 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
440 * - Removed tabs from "config"
441 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
442 * - Bumped up version number.
444 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
445 * Initial import of version 2.9.3 source tree
448 *********************************************************************/
454 #include <sys/types.h>
463 #endif /* ndef __OS2__ */
464 #include <netinet/in.h>
466 #include <winsock2.h>
467 #endif /* ndef _WIN32 */
471 #endif /* def __OS2__ */
479 #include "jbsockets.h"
480 #include "miscutil.h"
484 #include "deanimate.h"
485 #include "urlmatch.h"
491 const char filters_h_rcs[] = FILTERS_H_VERSION;
493 /* Fix a problem with Solaris. There should be no effect on other
495 * Solaris's isspace() is a macro which uses it's argument directly
496 * as an array index. Therefore we need to make sure that high-bit
497 * characters generate +ve values, and ideally we also want to make
498 * the argument match the declared parameter type of "int".
500 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
504 /*********************************************************************
506 * Function : block_acl
508 * Description : Block this request?
509 * Decide yes or no based on ACL file.
512 * 1 : dst = The proxy or gateway address this is going to.
513 * Or NULL to check all possible targets.
514 * 2 : csp = Current client state (buffers, headers, etc...)
515 * Also includes the client IP address.
517 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
519 *********************************************************************/
520 int block_acl(struct access_control_addr *dst, struct client_state *csp)
522 struct access_control_list *acl = csp->config->acl;
524 /* if not using an access control list, then permit the connection */
530 /* search the list */
533 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
537 /* Just want to check if they have any access */
538 if (acl->action == ACL_PERMIT)
543 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
544 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
546 if (acl->action == ACL_PERMIT)
564 /*********************************************************************
566 * Function : acl_addr
568 * Description : Called from `load_config' to parse an ACL address.
571 * 1 : aspec = String specifying ACL address.
572 * 2 : aca = struct access_control_addr to fill in.
574 * Returns : 0 => Ok, everything else is an error.
576 *********************************************************************/
577 int acl_addr(char *aspec, struct access_control_addr *aca)
579 int i, masklength, port;
585 if ((p = strchr(aspec, '/')) != NULL)
589 if (ijb_isdigit(*p) == 0)
593 masklength = atoi(p);
596 if ((masklength < 0) || (masklength > 32))
601 if ((p = strchr(aspec, ':')) != NULL)
605 if (ijb_isdigit(*p) == 0)
614 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
616 if (aca->addr == INADDR_NONE)
621 /* build the netmask */
623 for (i=1; i <= masklength ; i++)
625 aca->mask |= (1 << (32 - i));
628 /* now mask off the host portion of the ip address
629 * (i.e. save on the network portion of the address).
631 aca->addr = aca->addr & aca->mask;
636 #endif /* def FEATURE_ACL */
639 /*********************************************************************
641 * Function : match_portlist
643 * Description : Check if a given number is covered by a comma
644 * separated list of numbers and ranges (a,b-c,d,..)
647 * 1 : portlist = String with list
648 * 2 : port = port to check
650 * Returns : 0 => no match
653 *********************************************************************/
654 int match_portlist(const char *portlist, int port)
656 char *min, *max, *next, *portlist_copy;
658 min = next = portlist_copy = strdup(portlist);
661 * Zero-terminate first item and remember offset for next
663 if (NULL != (next = strchr(portlist_copy, (int) ',')))
669 * Loop through all items, checking for match
673 if (NULL == (max = strchr(min, (int) '-')))
676 * No dash, check for equality
678 if (port == atoi(min))
687 * This is a range, so check if between min and max,
688 * or, if max was omitted, between min and 65K
691 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
705 * Zero-terminate next item and remember offset for n+1
707 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
719 /*********************************************************************
721 * Function : block_url
723 * Description : Called from `chat'. Check to see if we need to block this.
726 * 1 : csp = Current client state (buffers, headers, etc...)
728 * Returns : NULL => unblocked, else HTTP block response
730 *********************************************************************/
731 struct http_response *block_url(struct client_state *csp)
733 struct http_response *rsp;
734 const char *new_content_type = NULL;
737 * If it's not blocked, don't block it ;-)
739 if ((csp->action->flags & ACTION_BLOCK) == 0)
743 if (csp->action->flags & ACTION_REDIRECT)
745 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
748 * Else, prepare a response
750 if (NULL == (rsp = alloc_http_response()))
752 return cgi_error_memory();
756 * If it's an image-url, send back an image or redirect
757 * as specified by the relevant +image action
759 #ifdef FEATURE_IMAGE_BLOCKING
760 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
764 /* determine HOW images should be blocked */
765 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
767 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
769 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
771 #if 1 /* Two alternative strategies, use this one for now: */
773 /* and handle accordingly: */
774 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
776 rsp->status = strdup("403 Request blocked by Privoxy");
777 if (rsp->status == NULL)
779 free_http_response(rsp);
780 return cgi_error_memory();
782 rsp->body = bindup(image_pattern_data, image_pattern_length);
783 if (rsp->body == NULL)
785 free_http_response(rsp);
786 return cgi_error_memory();
788 rsp->content_length = image_pattern_length;
790 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
792 free_http_response(rsp);
793 return cgi_error_memory();
797 else if (0 == strcmpic(p, "blank"))
799 rsp->status = strdup("403 Request blocked by Privoxy");
800 if (rsp->status == NULL)
802 free_http_response(rsp);
803 return cgi_error_memory();
805 rsp->body = bindup(image_blank_data, image_blank_length);
806 if (rsp->body == NULL)
808 free_http_response(rsp);
809 return cgi_error_memory();
811 rsp->content_length = image_blank_length;
813 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
815 free_http_response(rsp);
816 return cgi_error_memory();
822 rsp->status = strdup("302 Local Redirect from Privoxy");
823 if (rsp->status == NULL)
825 free_http_response(rsp);
826 return cgi_error_memory();
829 if (enlist_unique_header(rsp->headers, "Location", p))
831 free_http_response(rsp);
832 return cgi_error_memory();
836 #else /* Following code is disabled for now */
838 /* and handle accordingly: */
839 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
841 p = CGI_PREFIX "send-banner?type=pattern";
843 else if (0 == strcmpic(p, "blank"))
845 p = CGI_PREFIX "send-banner?type=blank";
847 rsp->status = strdup("302 Local Redirect from Privoxy");
848 if (rsp->status == NULL)
850 free_http_response(rsp);
851 return cgi_error_memory();
854 if (enlist_unique_header(rsp->headers, "Location", p))
856 free_http_response(rsp);
857 return cgi_error_memory();
859 #endif /* Preceeding code is disabled for now */
861 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
864 * Send empty document.
866 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
869 rsp->body = strdup(" ");
870 rsp->content_length = 1;
872 rsp->status = strdup("403 Request blocked by Privoxy");
873 if (rsp->status == NULL)
875 free_http_response(rsp);
876 return cgi_error_memory();
878 if (new_content_type != 0)
880 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
881 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
883 free_http_response(rsp);
884 return cgi_error_memory();
890 #endif /* def FEATURE_IMAGE_BLOCKING */
893 * Else, generate an HTML "blocked" message:
897 struct map * exports;
901 * Workaround for stupid Netscape bug which prevents
902 * pages from being displayed if loading a referenced
903 * JavaScript or style sheet fails. So make it appear
904 * as if it succeeded.
906 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
907 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
908 && !strstr(p, "Gecko") /* save Mozilla, */
909 && !strstr(p, "compatible") /* MSIE */
910 && !strstr(p, "Opera")) /* and Opera. */
912 rsp->status = strdup("200 Request for blocked URL");
916 rsp->status = strdup("404 Request for blocked URL");
919 if (rsp->status == NULL)
921 free_http_response(rsp);
922 return cgi_error_memory();
925 exports = default_exports(csp, NULL);
928 free_http_response(rsp);
929 return cgi_error_memory();
932 #ifdef FEATURE_FORCE_LOAD
933 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
934 if (csp->http->ssl != 0)
935 #endif /* ndef FEATURE_FORCE_LOAD */
937 err = map_block_killer(exports, "force-support");
940 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
941 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
942 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
943 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
948 free_http_response(rsp);
949 return cgi_error_memory();
952 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
955 free_http_response(rsp);
956 return cgi_error_memory();
960 return finish_http_response(rsp);
966 /*********************************************************************
968 * Function : trust_url FIXME: I should be called distrust_url
970 * Description : Calls is_untrusted_url to determine if the URL is trusted
971 * and if not, returns a HTTP 304 response with a reject message.
974 * 1 : csp = Current client state (buffers, headers, etc...)
976 * Returns : NULL => trusted, else http_response.
978 *********************************************************************/
979 struct http_response *trust_url(struct client_state *csp)
981 struct http_response *rsp;
982 struct map * exports;
983 char buf[BUFFER_SIZE];
985 struct url_spec **tl;
990 * Don't bother to work on trusted URLs
992 if (!is_untrusted_url(csp))
998 * Else, prepare a response:
1000 if (NULL == (rsp = alloc_http_response()))
1002 return cgi_error_memory();
1005 exports = default_exports(csp, NULL);
1006 if (exports == NULL)
1008 free_http_response(rsp);
1009 return cgi_error_memory();
1013 * Export the protocol, host, port, and referrer information
1015 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1016 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1017 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1019 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1021 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1025 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1031 free_http_response(rsp);
1032 return cgi_error_memory();
1036 * Export the trust list
1039 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1041 sprintf(buf, "<li>%s</li>\n", t->spec);
1042 string_append(&p, buf);
1044 err = map(exports, "trusted-referrers", 1, p, 0);
1049 free_http_response(rsp);
1050 return cgi_error_memory();
1054 * Export the trust info, if available
1056 if (csp->config->trust_info->first)
1058 struct list_entry *l;
1061 for (l = csp->config->trust_info->first; l ; l = l->next)
1063 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1064 string_append(&p, buf);
1066 err = map(exports, "trust-info", 1, p, 0);
1070 err = map_block_killer(exports, "have-trust-info");
1076 free_http_response(rsp);
1077 return cgi_error_memory();
1081 * Export the force prefix or the force conditional block killer
1083 #ifdef FEATURE_FORCE_LOAD
1084 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1085 #else /* ifndef FEATURE_FORCE_LOAD */
1086 err = map_block_killer(exports, "force-support");
1087 #endif /* ndef FEATURE_FORCE_LOAD */
1092 free_http_response(rsp);
1093 return cgi_error_memory();
1097 * Build the response
1099 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1102 free_http_response(rsp);
1103 return cgi_error_memory();
1106 return finish_http_response(rsp);
1108 #endif /* def FEATURE_TRUST */
1111 #ifdef FEATURE_FAST_REDIRECTS
1112 /*********************************************************************
1114 * Function : redirect_url
1116 * Description : Checks for redirection URLs and returns a HTTP redirect
1117 * to the destination URL, if necessary
1120 * 1 : csp = Current client state (buffers, headers, etc...)
1122 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1124 *********************************************************************/
1125 struct http_response *redirect_url(struct client_state *csp)
1128 struct http_response *rsp;
1129 char *redirect_mode = NULL;
1132 if ((csp->action->flags & ACTION_REDIRECT))
1134 q = csp->action->string[ACTION_STRING_REDIRECT];
1138 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1139 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1141 p = q = csp->http->path;
1142 log_error(LOG_LEVEL_REDIRECTS, "Decoding path: %s if necessary.", p);
1145 if (*p == '%') /* Escape sequence? */
1147 /* Yes, translate from hexadecimal to decimal */
1161 /* No, forward character. */
1167 p = q = csp->http->path;
1168 log_error(LOG_LEVEL_REDIRECTS, "Checking path for redirects: %s", p);
1171 * find the last URL encoded in the request
1173 while ((p = strstr(p, "http://")) != NULL)
1179 * if there was any, generate and return a HTTP redirect
1181 if (q != csp->http->path)
1183 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1185 if (NULL == (rsp = alloc_http_response()))
1187 return cgi_error_memory();
1190 if ( enlist_unique_header(rsp->headers, "Location", q)
1191 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1193 free_http_response(rsp);
1194 return cgi_error_memory();
1197 return finish_http_response(rsp);
1205 #endif /* def FEATURE_FAST_REDIRECTS */
1208 #ifdef FEATURE_IMAGE_BLOCKING
1209 /*********************************************************************
1211 * Function : is_imageurl
1213 * Description : Given a URL, decide whether it is an image or not,
1214 * using either the info from a previous +image action
1215 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1216 * is MSIE and not on a Mac, tell from the browser's accept
1220 * 1 : csp = Current client state (buffers, headers, etc...)
1222 * Returns : True (nonzero) if URL is an image, false (0)
1225 *********************************************************************/
1226 int is_imageurl(struct client_state *csp)
1228 #ifdef FEATURE_IMAGE_DETECT_MSIE
1231 tmp = get_header_value(csp->headers, "User-Agent:");
1232 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1234 tmp = get_header_value(csp->headers, "Accept:");
1235 if (tmp && strstr(tmp, "image/gif"))
1237 /* Client will accept HTML. If this seems counterintuitive,
1247 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1249 return ((csp->action->flags & ACTION_IMAGE) != 0);
1252 #endif /* def FEATURE_IMAGE_BLOCKING */
1255 #ifdef FEATURE_TRUST
1256 /*********************************************************************
1258 * Function : is_untrusted_url
1260 * Description : Should we "distrust" this URL (and block it)?
1262 * Yes if it matches a line in the trustfile, or if the
1263 * referrer matches a line starting with "+" in the
1268 * 1 : csp = Current client state (buffers, headers, etc...)
1270 * Returns : 0 => trusted, 1 => untrusted
1272 *********************************************************************/
1273 int is_untrusted_url(struct client_state *csp)
1275 struct file_list *fl;
1276 struct block_spec *b;
1277 struct url_spec **trusted_url;
1278 struct http_request rhttp[1];
1279 const char * referer;
1283 * If we don't have a trustlist, we trust everybody
1285 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1290 memset(rhttp, '\0', sizeof(*rhttp));
1293 * Do we trust the request URL itself?
1295 for (b = b->next; b ; b = b->next)
1297 if (url_match(b->url, csp->http))
1303 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1305 /* no referrer was supplied */
1311 * If not, do we maybe trust its referrer?
1313 err = parse_http_url(referer, rhttp, csp);
1319 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1321 if (url_match(*trusted_url, rhttp))
1323 /* if the URL's referrer is from a trusted referrer, then
1324 * add the target spec to the trustfile as an unblocked
1325 * domain and return NULL (which means it's OK).
1330 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1334 char * new_entry = strdup("~");
1336 string_append(&new_entry, csp->http->hostport);
1338 path = csp->http->path;
1339 if ( (path[0] == '/')
1341 && ((path_end = strchr(path + 2, '/')) != NULL))
1343 /* since this path points into a user's home space
1344 * be sure to include this spec in the trustfile.
1346 int path_len = path_end - path; /* save offset */
1347 path = strdup(path); /* Copy string */
1350 path_end = path + path_len; /* regenerate ptr to new buffer */
1351 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1353 string_join(&new_entry, path);
1356 if (new_entry != NULL)
1358 fprintf(fp, "%s\n", new_entry);
1363 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1364 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1374 #endif /* def FEATURE_TRUST */
1377 /*********************************************************************
1379 * Function : pcrs_filter_response
1381 * Description : Execute all text substitutions from all applying
1382 * +filter actions on the text buffer that's been accumulated
1383 * in csp->iob->buf. If this changes the contents, set
1384 * csp->content_length to the modified size and raise the
1385 * CSP_FLAG_MODIFIED flag.
1388 * 1 : csp = Current client state (buffers, headers, etc...)
1390 * Returns : a pointer to the (newly allocated) modified buffer.
1391 * or NULL if there were no hits or something went wrong
1393 *********************************************************************/
1394 char *pcrs_filter_response(struct client_state *csp)
1397 size_t size, prev_size;
1399 char *old = csp->iob->cur, *new = NULL;
1402 struct file_list *fl;
1403 struct re_filterfile_spec *b;
1404 struct list_entry *filtername;
1406 int i, found_filters = 0;
1411 if (csp->iob->cur >= csp->iob->eod)
1415 size = csp->iob->eod - csp->iob->cur;
1418 * Need to check the set of re_filterfiles...
1420 for (i = 0; i < MAX_AF_FILES; i++)
1433 if (0 == found_filters)
1435 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1440 * If the body has a "chunked" transfer-encoding,
1441 * get rid of it first, adjusting size and iob->eod
1443 if (csp->flags & CSP_FLAG_CHUNKED)
1445 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1446 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1450 csp->iob->eod = csp->iob->cur + size;
1451 csp->flags |= CSP_FLAG_MODIFIED;
1454 for (i = 0; i < MAX_AF_FILES; i++)
1457 if ((NULL == fl) || (NULL == fl->f))
1460 * For all applying +filter actions, look if a filter by that
1461 * name exists and if yes, execute it's pcrs_joblist on the
1464 for (b = fl->f; b; b = b->next)
1466 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1467 filtername ; filtername = filtername->next)
1469 if (strcmp(b->name, filtername->str) == 0)
1471 int current_hits = 0;
1473 if ( NULL == b->joblist )
1475 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1480 /* Apply all jobs from the joblist */
1481 for (job = b->joblist; NULL != job; job = job->next)
1483 current_hits += pcrs_execute(job, old, size, &new, &size);
1484 if (old != csp->iob->cur) free(old);
1488 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1489 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1491 hits += current_hits;
1498 * If there were no hits, destroy our copy and let
1499 * chat() use the original in csp->iob
1507 csp->flags |= CSP_FLAG_MODIFIED;
1508 csp->content_length = size;
1516 /*********************************************************************
1518 * Function : gif_deanimate_response
1520 * Description : Deanimate the GIF image that has been accumulated in
1521 * csp->iob->buf, set csp->content_length to the modified
1522 * size and raise the CSP_FLAG_MODIFIED flag.
1525 * 1 : csp = Current client state (buffers, headers, etc...)
1527 * Returns : a pointer to the (newly allocated) modified buffer.
1528 * or NULL in case something went wrong.
1530 *********************************************************************/
1531 char *gif_deanimate_response(struct client_state *csp)
1533 struct binbuffer *in, *out;
1535 size_t size = csp->iob->eod - csp->iob->cur;
1538 * If the body has a "chunked" transfer-encoding,
1539 * get rid of it first, adjusting size and iob->eod
1541 if (csp->flags & CSP_FLAG_CHUNKED)
1543 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1544 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1548 csp->iob->eod = csp->iob->cur + size;
1549 csp->flags |= CSP_FLAG_MODIFIED;
1552 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1553 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1555 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1559 in->buffer = csp->iob->cur;
1562 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1564 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1571 if ((int)size == out->offset)
1573 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1577 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1579 csp->content_length = out->offset;
1580 csp->flags |= CSP_FLAG_MODIFIED;
1590 /*********************************************************************
1592 * Function : jpeg_inspect_response
1597 * 1 : csp = Current client state (buffers, headers, etc...)
1599 * Returns : a pointer to the (newly allocated) modified buffer
1600 * or NULL in case something went wrong.
1602 *********************************************************************/
1603 char *jpeg_inspect_response(struct client_state *csp)
1605 struct binbuffer *in = NULL, *out = NULL;
1607 size_t size = csp->iob->eod - csp->iob->cur;
1610 * If the body has a "chunked" transfer-encoding,
1611 * get rid of it first, adjusting size and iob->eod
1613 if (csp->flags & CSP_FLAG_CHUNKED)
1615 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1616 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1620 csp->iob->eod = csp->iob->cur + size;
1621 csp->flags |= CSP_FLAG_MODIFIED;
1624 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1626 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1630 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1632 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1636 in->buffer = csp->iob->cur;
1640 * Calling jpeg_inspect has the side-effect of creating and
1641 * modifying the image buffer of "out" directly.
1643 if (jpeg_inspect(in, out))
1645 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1653 csp->content_length = out->offset;
1654 csp->flags |= CSP_FLAG_MODIFIED;
1664 /*********************************************************************
1666 * Function : remove_chunked_transfer_coding
1668 * Description : In-situ remove the "chunked" transfer coding as defined
1669 * in rfc2616 from a buffer.
1672 * 1 : buffer = Pointer to the text buffer
1673 * 2 : size = Number of bytes to be processed
1675 * Returns : The new size, i.e. the number of bytes from buffer which
1676 * are occupied by the stripped body, or 0 in case something
1679 *********************************************************************/
1680 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1683 unsigned int chunksize = 0;
1684 char *from_p, *to_p;
1687 from_p = to_p = buffer;
1689 if (sscanf(buffer, "%x", &chunksize) != 1)
1691 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1695 while (chunksize > 0)
1697 if (NULL == (from_p = strstr(from_p, "\r\n")))
1699 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1703 if ((newsize += chunksize) >= size)
1705 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1710 memmove(to_p, from_p, (size_t) chunksize);
1711 to_p = buffer + newsize;
1712 from_p += chunksize + 2;
1714 if (sscanf(from_p, "%x", &chunksize) != 1)
1716 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1721 /* FIXME: Should this get its own loglevel? */
1722 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1728 /*********************************************************************
1730 * Function : url_actions
1732 * Description : Gets the actions for this URL.
1735 * 1 : http = http_request request for blocked URLs
1736 * 2 : csp = Current client state (buffers, headers, etc...)
1740 *********************************************************************/
1741 void url_actions(struct http_request *http,
1742 struct client_state *csp)
1744 struct file_list *fl;
1745 struct url_actions *b;
1748 init_current_action(csp->action);
1750 for (i = 0; i < MAX_AF_FILES; i++)
1752 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1757 apply_url_actions(csp->action, http, b);
1764 /*********************************************************************
1766 * Function : apply_url_actions
1768 * Description : Applies a list of URL actions.
1771 * 1 : action = Destination.
1772 * 2 : http = Current URL
1773 * 3 : b = list of URL actions to apply
1777 *********************************************************************/
1778 void apply_url_actions(struct current_action_spec *action,
1779 struct http_request *http,
1780 struct url_actions *b)
1784 /* Should never happen */
1788 for (b = b->next; NULL != b; b = b->next)
1790 if (url_match(b->url, http))
1792 merge_current_action(action, b->action);
1798 /*********************************************************************
1800 * Function : forward_url
1802 * Description : Should we forward this to another proxy?
1805 * 1 : http = http_request request for current URL
1806 * 2 : csp = Current client state (buffers, headers, etc...)
1808 * Returns : Pointer to forwarding information.
1810 *********************************************************************/
1811 const struct forward_spec * forward_url(struct http_request *http,
1812 struct client_state *csp)
1814 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1815 struct forward_spec *fwd = csp->config->forward;
1824 if (url_match(fwd->url, http))
1835 /*********************************************************************
1837 * Function : direct_response
1839 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
1840 * request and if so, return a HTTP 501 to the client.
1842 * FIXME: I have a stupid name and I should handle the
1843 * requests properly. Still, what we do here is rfc-
1844 * compliant, whereas ignoring or forwarding are not.
1847 * 1 : csp = Current client state (buffers, headers, etc...)
1849 * Returns : http_response if , NULL if nonmatch or handler fail
1851 *********************************************************************/
1852 struct http_response *direct_response(struct client_state *csp)
1854 struct http_response *rsp;
1855 struct list_entry *p;
1857 if ((0 == strcmpic(csp->http->gpc, "trace"))
1858 || (0 == strcmpic(csp->http->gpc, "options")))
1860 for (p = csp->headers->first; (p != NULL) ; p = p->next)
1862 if (!strncmp("Max-Forwards:", p->str, 13)
1863 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
1865 /* FIXME: We could handle at least TRACE here,
1866 but that would require a verbatim copy of
1867 the request which we don't have anymore */
1869 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
1871 /* Get mem for response or fail*/
1872 if (NULL == (rsp = alloc_http_response()))
1874 return cgi_error_memory();
1877 if (NULL == (rsp->status = strdup("501 Not Implemented")))
1879 free_http_response(rsp);
1880 return cgi_error_memory();
1884 return(finish_http_response(rsp));