1 const char filters_rcs[] = "$Id: filters.c,v 1.61 2006/08/03 02:46:41 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 1.61 2006/08/03 02:46:41 david__schmidt
43 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
45 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
46 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
47 * with what was really the latest development (the v_3_0_branch branch)
49 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
50 * Multiple filter file support
52 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
53 * A few memory leaks plugged (mostly on error paths)
55 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
56 * Add the ability to check jpeg images for invalid
57 * lengths of comment blocks. Defensive strategy
58 * against the exploit:
59 * Microsoft Security Bulletin MS04-028
60 * Buffer Overrun in JPEG Processing (GDI+) Could
61 * Allow Code Execution (833987)
62 * Enabled with +inspect-jpegs in actions files.
64 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
65 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
67 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
68 * Fixed bug #839859: "See why" link URL now gets url-encoded.
70 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
73 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
74 * Added basic support for OPTIONS and TRACE HTTP methods:
75 * New function direct_response which handles OPTIONS and
76 * TRACE requests whose Max-Forwards header field is zero.
78 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
79 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
81 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
82 * - Bugfix: Executing a filters without jobs no longer results in
83 * turing off *all* filters.
84 * - Security fix: Malicious web servers can't cause a seg fault
85 * through bogus chunk sizes anymore
87 * Revision 1.58 2002/04/24 02:11:17 oes
88 * Jon's multiple AF patch: url_actions now evaluates rules
91 * Revision 1.57 2002/04/08 20:38:34 swa
94 * Revision 1.56 2002/04/05 15:51:24 oes
95 * - bugfix: error-pages now get correct request protocol
96 * - fix for invalid HTML in trust info
98 * Revision 1.55 2002/04/02 16:13:51 oes
99 * Fix: No "Go there anyway" for SSL
101 * Revision 1.54 2002/04/02 14:55:56 oes
102 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
104 * Revision 1.53 2002/03/26 22:29:54 swa
105 * we have a new homepage!
107 * Revision 1.52 2002/03/24 16:35:57 jongfoster
110 * Revision 1.51 2002/03/24 15:23:33 jongfoster
113 * Revision 1.50 2002/03/24 13:25:43 swa
114 * name change related issues
116 * Revision 1.49 2002/03/16 20:29:14 oes
119 * Revision 1.48 2002/03/13 20:25:34 oes
120 * Better logging for content filters
122 * Revision 1.47 2002/03/13 00:30:52 jongfoster
124 * Added option of always sending redirect for imageblock,
125 * currently disabled with #if 0.
127 * Revision 1.46 2002/03/12 01:42:49 oes
128 * Introduced modular filters
130 * Revision 1.45 2002/03/08 16:47:50 oes
131 * Added choice beween GIF and PNG built-in images
133 * Revision 1.44 2002/03/07 03:49:31 oes
134 * - Fixed compiler warnings etc
135 * - Changed built-in images from GIF to PNG
136 * (with regard to Unisys patent issue)
137 * - Added a 4x4 pattern PNG which is less intrusive
138 * than the logo but also clearly marks the deleted banners
140 * Revision 1.43 2002/01/22 23:51:59 jongfoster
141 * Replacing strsav() with the safer string_append().
143 * Adding missing html_encode() to error message generators. Where encoded
144 * and unencoded versions of a string were provided, removing the unencoded
147 * Revision 1.42 2002/01/17 21:00:32 jongfoster
148 * Moving all our URL and URL pattern parsing code to urlmatch.c.
150 * Using a single, simple url_match(pattern,url) function - rather than
151 * the 3-line match routine which was repeated all over the place.
153 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
155 * Using parse_http_url() to parse URLs without faking a HTTP
156 * request line for parse_http_request().
158 * Revision 1.41 2001/11/13 00:14:07 jongfoster
159 * Fixing stupid bug now I've figured out what || means.
160 * (It always returns 0 or 1, not one of it's paramaters.)
162 * Revision 1.40 2001/10/26 17:37:55 oes
163 * - Re-enabled Netscape 200/404 bug workaround in block_url():
164 * - Removed OS/2 special case
165 * - Made block_url() independant from sed() having been run
166 * - Made trust_url independant from sed() having been run
167 * - Made is_imageurl independant from sed() having been run.
168 * It now checks User-Agent: and Accept: by itself.
171 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
172 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
173 * threads to call select() simultaneously. So, it's time to do a real, live,
174 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
175 * (native). Both versions will work, but using __OS2__ offers multi-threading.
177 * Revision 1.38 2001/10/23 21:32:33 jongfoster
178 * Adding error-checking to selected functions
180 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
181 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
182 * filters.c. Added a FIXME in front of the offending code. I'll gladly
183 * put in a better/more robust fix for all parties if one is presented...
184 * It seems that just returning 200 instead of 404 would pretty much fix
185 * it for everyone, but I don't know all the history of the problem.
187 * Revision 1.36 2001/10/10 16:44:16 oes
188 * Added match_portlist function
190 * Revision 1.35 2001/10/07 15:41:23 oes
191 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
193 * New function remove_chunked_transfer_coding that strips chunked
194 * transfer coding to plain and is called by pcrs_filter_response
195 * and gif_deanimate_response if neccessary
197 * Improved handling of zero-change re_filter runs
199 * pcrs_filter_response and gif_deanimate_response now remove
200 * chunked transfer codeing before processing the body.
202 * Revision 1.34 2001/09/20 15:49:36 steudten
204 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
205 * See cgi.c fill_template().
207 * Revision 1.33 2001/09/16 17:05:14 jongfoster
208 * Removing unused #include showarg.h
210 * Revision 1.32 2001/09/16 13:21:27 jongfoster
211 * Changes to use new list functions.
213 * Revision 1.31 2001/09/16 11:38:02 jongfoster
214 * Splitting fill_template() into 2 functions:
215 * template_load() loads the file
216 * template_fill() performs the PCRS regexps.
217 * This is because the CGI edit interface has a "table row"
218 * template which is used many times in the page - this
219 * change means it's only loaded from disk once.
221 * Revision 1.30 2001/09/16 11:00:10 jongfoster
222 * New function alloc_http_response, for symmetry with free_http_response
224 * Revision 1.29 2001/09/13 23:32:40 jongfoster
225 * Moving image data to cgi.c rather than cgi.h
226 * Fixing a GPF under Win32 (and any other OS that protects global
227 * constants from being written to).
229 * Revision 1.28 2001/09/10 10:18:51 oes
230 * Silenced compiler warnings
232 * Revision 1.27 2001/08/05 16:06:20 jongfoster
233 * Modifiying "struct map" so that there are now separate header and
234 * "map_entry" structures. This means that functions which modify a
235 * map no longer need to return a pointer to the modified map.
236 * Also, it no longer reverses the order of the entries (which may be
237 * important with some advanced template substitutions).
239 * Revision 1.26 2001/07/30 22:08:36 jongfoster
240 * Tidying up #defines:
241 * - All feature #defines are now of the form FEATURE_xxx
242 * - Permanently turned off WIN_GUI_EDIT
243 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
245 * Revision 1.25 2001/07/26 10:09:46 oes
246 * Made browser detection a little less naive
248 * Revision 1.24 2001/07/25 17:22:51 oes
249 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
251 * Revision 1.23 2001/07/23 13:40:12 oes
252 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
254 * Revision 1.22 2001/07/18 12:29:34 oes
255 * - Made gif_deanimate_response respect
256 * csp->action->string[ACTION_STRING_DEANIMATE]
257 * - Logging cosmetics
259 * Revision 1.21 2001/07/13 13:59:53 oes
260 * - Introduced gif_deanimate_response which shares the
261 * generic content modification interface of pcrs_filter_response
262 * and acts as a wrapper to deanimate.c:gif_deanimate()
263 * - Renamed re_process_buffer to pcrs_filter_response
264 * - pcrs_filter_response now returns NULL on failiure
265 * - Removed all #ifdef PCRS
267 * Revision 1.20 2001/07/01 17:01:04 oes
268 * Added comments and missing return statement in is_untrusted_url()
270 * Revision 1.19 2001/06/29 21:45:41 oes
271 * Indentation, CRLF->LF, Tab-> Space
273 * Revision 1.18 2001/06/29 13:27:38 oes
274 * - Cleaned up, renamed and reorderd functions
275 * and improved comments
278 * - Ported to CGI platform. Now delivers
279 * http_response or NULL
280 * - Unified HTML and GIF generation (moved image detection
281 * and GIF generation here from jcc.c:chat())
282 * - Fixed HTTP status to:
283 * - 403 (Forbidden) for the "blocked" HTML message
284 * - 200 (OK) for GIF answers
285 * - 302 (Redirect) for redirect to GIF
288 * - Ported to CGI platform. Now delivers
289 * http_response or NULL
290 * - Separated detection of untrusted URL into
291 * (bool)is_untrusted_url
292 * - Added enforcement of untrusted requests
294 * - Moved redirect_url() from cgi.c to here
295 * and ported it to the CGI platform
297 * - Removed logentry from cancelled commit
299 * Revision 1.17 2001/06/09 10:55:28 jongfoster
300 * Changing BUFSIZ ==> BUFFER_SIZE
302 * Revision 1.16 2001/06/07 23:10:26 jongfoster
303 * Allowing unanchored domain patterns to back off and retry
304 * if they partially match. Optimized right-anchored patterns.
305 * Moving ACL and forward files into config file.
306 * Replacing struct gateway with struct forward_spec
308 * Revision 1.15 2001/06/03 19:12:00 oes
309 * extracted-CGI relevant stuff
311 * Revision 1.14 2001/06/01 10:30:55 oes
312 * Added optional left-anchoring to domaincmp
314 * Revision 1.13 2001/05/31 21:21:30 jongfoster
315 * Permissionsfile / actions file changes:
316 * - Changed "permission" to "action" throughout
317 * - changes to file format to allow string parameters
318 * - Moved helper functions to actions.c
320 * Revision 1.12 2001/05/31 17:35:20 oes
322 * - Enhanced domain part globbing with infix and prefix asterisk
323 * matching and optional unanchored operation
325 * Revision 1.11 2001/05/29 11:53:23 oes
326 * "See why" link added to "blocked" page
328 * Revision 1.10 2001/05/29 09:50:24 jongfoster
329 * Unified blocklist/imagelist/permissionslist.
330 * File format is still under discussion, but the internal changes
333 * Also modified interceptor behaviour:
334 * - We now intercept all URLs beginning with one of the following
335 * prefixes (and *only* these prefixes):
337 * * http://ijbswa.sf.net/config/
338 * * http://ijbswa.sourceforge.net/config/
339 * - New interceptors "home page" - go to http://i.j.b/ to see it.
340 * - Internal changes so that intercepted and fast redirect pages
341 * are not replaced with an image.
342 * - Interceptors now have the option to send a binary page direct
343 * to the client. (i.e. ijb-send-banner uses this)
344 * - Implemented show-url-info interceptor. (Which is why I needed
345 * the above interceptors changes - a typical URL is
346 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
347 * The previous mechanism would not have intercepted that, and
348 * if it had been intercepted then it then it would have replaced
351 * Revision 1.9 2001/05/27 22:17:04 oes
353 * - re_process_buffer no longer writes the modified buffer
354 * to the client, which was very ugly. It now returns the
355 * buffer, which it is then written by chat.
357 * - content_length now adjusts the Content-Length: header
358 * for modified documents rather than crunch()ing it.
359 * (Length info in csp->content_length, which is 0 for
360 * unmodified documents)
362 * - For this to work, sed() is called twice when filtering.
364 * Revision 1.8 2001/05/26 17:13:28 jongfoster
365 * Filled in a function comment.
367 * Revision 1.7 2001/05/26 15:26:15 jongfoster
368 * ACL feature now provides more security by immediately dropping
369 * connections from untrusted hosts.
371 * Revision 1.6 2001/05/26 00:28:36 jongfoster
372 * Automatic reloading of config file.
373 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
374 * Most of the global variables have been moved to a new
375 * struct configuration_spec, accessed through csp->config->globalname
376 * Most of the globals remaining are used by the Win32 GUI.
378 * Revision 1.5 2001/05/25 22:34:30 jongfoster
381 * Revision 1.4 2001/05/22 18:46:04 oes
383 * - Enabled filtering banners by size rather than URL
384 * by adding patterns that replace all standard banner
385 * sizes with the "Junkbuster" gif to the re_filterfile
387 * - Enabled filtering WebBugs by providing a pattern
388 * which kills all 1x1 images
390 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
391 * which is selected by the (nonstandard and therefore
392 * capital) letter 'U' in the option string.
393 * It causes the quantifiers to be ungreedy by default.
394 * Appending a ? turns back to greedy (!).
396 * - Added a new interceptor ijb-send-banner, which
397 * sends back the "Junkbuster" gif. Without imagelist or
398 * MSIE detection support, or if tinygif = 1, or the
399 * URL isn't recognized as an imageurl, a lame HTML
400 * explanation is sent instead.
402 * - Added new feature, which permits blocking remote
403 * script redirects and firing back a local redirect
405 * The feature is conditionally compiled, i.e. it
406 * can be disabled with --disable-fast-redirects,
407 * plus it must be activated by a "fast-redirects"
408 * line in the config file, has its own log level
409 * and of course wants to be displayed by show-proxy-args
410 * Note: Boy, all the #ifdefs in 1001 locations and
411 * all the fumbling with configure.in and acconfig.h
412 * were *way* more work than the feature itself :-(
414 * - Because a generic redirect template was needed for
415 * this, tinygif = 3 now uses the same.
417 * - Moved GIFs, and other static HTTP response templates
422 * - Removed some >400 CRs again (Jon, you really worked
425 * Revision 1.3 2001/05/20 16:44:47 jongfoster
426 * Removing last hardcoded Junkbusters.com URLs.
428 * Revision 1.2 2001/05/20 01:21:20 jongfoster
429 * Version 2.9.4 checkin.
430 * - Merged popupfile and cookiefile, and added control over PCRS
431 * filtering, in new "permissionsfile".
432 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
433 * file error you now get a message box (in the Win32 GUI) rather
434 * than the program exiting with no explanation.
435 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
437 * - Removed tabs from "config"
438 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
439 * - Bumped up version number.
441 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
442 * Initial import of version 2.9.3 source tree
445 *********************************************************************/
451 #include <sys/types.h>
460 #endif /* ndef __OS2__ */
461 #include <netinet/in.h>
463 #include <winsock2.h>
464 #endif /* ndef _WIN32 */
468 #endif /* def __OS2__ */
476 #include "jbsockets.h"
477 #include "miscutil.h"
481 #include "deanimate.h"
482 #include "urlmatch.h"
488 const char filters_h_rcs[] = FILTERS_H_VERSION;
490 /* Fix a problem with Solaris. There should be no effect on other
492 * Solaris's isspace() is a macro which uses it's argument directly
493 * as an array index. Therefore we need to make sure that high-bit
494 * characters generate +ve values, and ideally we also want to make
495 * the argument match the declared parameter type of "int".
497 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
501 /*********************************************************************
503 * Function : block_acl
505 * Description : Block this request?
506 * Decide yes or no based on ACL file.
509 * 1 : dst = The proxy or gateway address this is going to.
510 * Or NULL to check all possible targets.
511 * 2 : csp = Current client state (buffers, headers, etc...)
512 * Also includes the client IP address.
514 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
516 *********************************************************************/
517 int block_acl(struct access_control_addr *dst, struct client_state *csp)
519 struct access_control_list *acl = csp->config->acl;
521 /* if not using an access control list, then permit the connection */
527 /* search the list */
530 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
534 /* Just want to check if they have any access */
535 if (acl->action == ACL_PERMIT)
540 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
541 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
543 if (acl->action == ACL_PERMIT)
561 /*********************************************************************
563 * Function : acl_addr
565 * Description : Called from `load_config' to parse an ACL address.
568 * 1 : aspec = String specifying ACL address.
569 * 2 : aca = struct access_control_addr to fill in.
571 * Returns : 0 => Ok, everything else is an error.
573 *********************************************************************/
574 int acl_addr(char *aspec, struct access_control_addr *aca)
576 int i, masklength, port;
582 if ((p = strchr(aspec, '/')) != NULL)
586 if (ijb_isdigit(*p) == 0)
590 masklength = atoi(p);
593 if ((masklength < 0) || (masklength > 32))
598 if ((p = strchr(aspec, ':')) != NULL)
602 if (ijb_isdigit(*p) == 0)
611 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
613 if (aca->addr == INADDR_NONE)
618 /* build the netmask */
620 for (i=1; i <= masklength ; i++)
622 aca->mask |= (1 << (32 - i));
625 /* now mask off the host portion of the ip address
626 * (i.e. save on the network portion of the address).
628 aca->addr = aca->addr & aca->mask;
633 #endif /* def FEATURE_ACL */
636 /*********************************************************************
638 * Function : match_portlist
640 * Description : Check if a given number is covered by a comma
641 * separated list of numbers and ranges (a,b-c,d,..)
644 * 1 : portlist = String with list
645 * 2 : port = port to check
647 * Returns : 0 => no match
650 *********************************************************************/
651 int match_portlist(const char *portlist, int port)
653 char *min, *max, *next, *portlist_copy;
655 min = next = portlist_copy = strdup(portlist);
658 * Zero-terminate first item and remember offset for next
660 if (NULL != (next = strchr(portlist_copy, (int) ',')))
666 * Loop through all items, checking for match
670 if (NULL == (max = strchr(min, (int) '-')))
673 * No dash, check for equality
675 if (port == atoi(min))
684 * This is a range, so check if between min and max,
685 * or, if max was omitted, between min and 65K
688 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
702 * Zero-terminate next item and remember offset for n+1
704 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
716 /*********************************************************************
718 * Function : block_url
720 * Description : Called from `chat'. Check to see if we need to block this.
723 * 1 : csp = Current client state (buffers, headers, etc...)
725 * Returns : NULL => unblocked, else HTTP block response
727 *********************************************************************/
728 struct http_response *block_url(struct client_state *csp)
730 struct http_response *rsp;
731 const char *new_content_type = NULL;
734 * If it's not blocked, don't block it ;-)
736 if ((csp->action->flags & ACTION_BLOCK) == 0)
740 if (csp->action->flags & ACTION_REDIRECT)
742 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
745 * Else, prepare a response
747 if (NULL == (rsp = alloc_http_response()))
749 return cgi_error_memory();
753 * If it's an image-url, send back an image or redirect
754 * as specified by the relevant +image action
756 #ifdef FEATURE_IMAGE_BLOCKING
757 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
761 /* determine HOW images should be blocked */
762 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
764 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
766 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
768 #if 1 /* Two alternative strategies, use this one for now: */
770 /* and handle accordingly: */
771 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
773 rsp->status = strdup("403 Request blocked by Privoxy");
774 if (rsp->status == NULL)
776 free_http_response(rsp);
777 return cgi_error_memory();
779 rsp->body = bindup(image_pattern_data, image_pattern_length);
780 if (rsp->body == NULL)
782 free_http_response(rsp);
783 return cgi_error_memory();
785 rsp->content_length = image_pattern_length;
787 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
789 free_http_response(rsp);
790 return cgi_error_memory();
794 else if (0 == strcmpic(p, "blank"))
796 rsp->status = strdup("403 Request blocked by Privoxy");
797 if (rsp->status == NULL)
799 free_http_response(rsp);
800 return cgi_error_memory();
802 rsp->body = bindup(image_blank_data, image_blank_length);
803 if (rsp->body == NULL)
805 free_http_response(rsp);
806 return cgi_error_memory();
808 rsp->content_length = image_blank_length;
810 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
812 free_http_response(rsp);
813 return cgi_error_memory();
819 rsp->status = strdup("302 Local Redirect from Privoxy");
820 if (rsp->status == NULL)
822 free_http_response(rsp);
823 return cgi_error_memory();
826 if (enlist_unique_header(rsp->headers, "Location", p))
828 free_http_response(rsp);
829 return cgi_error_memory();
833 #else /* Following code is disabled for now */
835 /* and handle accordingly: */
836 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
838 p = CGI_PREFIX "send-banner?type=pattern";
840 else if (0 == strcmpic(p, "blank"))
842 p = CGI_PREFIX "send-banner?type=blank";
844 rsp->status = strdup("302 Local Redirect from Privoxy");
845 if (rsp->status == NULL)
847 free_http_response(rsp);
848 return cgi_error_memory();
851 if (enlist_unique_header(rsp->headers, "Location", p))
853 free_http_response(rsp);
854 return cgi_error_memory();
856 #endif /* Preceeding code is disabled for now */
858 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
861 * Send empty document.
863 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
866 rsp->body = strdup(" ");
867 rsp->content_length = 1;
869 rsp->status = strdup("403 Request blocked by Privoxy");
870 if (rsp->status == NULL)
872 free_http_response(rsp);
873 return cgi_error_memory();
875 if (new_content_type != 0)
877 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
878 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
880 free_http_response(rsp);
881 return cgi_error_memory();
887 #endif /* def FEATURE_IMAGE_BLOCKING */
890 * Else, generate an HTML "blocked" message:
894 struct map * exports;
898 * Workaround for stupid Netscape bug which prevents
899 * pages from being displayed if loading a referenced
900 * JavaScript or style sheet fails. So make it appear
901 * as if it succeeded.
903 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
904 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
905 && !strstr(p, "Gecko") /* save Mozilla, */
906 && !strstr(p, "compatible") /* MSIE */
907 && !strstr(p, "Opera")) /* and Opera. */
909 rsp->status = strdup("200 Request for blocked URL");
913 rsp->status = strdup("404 Request for blocked URL");
916 if (rsp->status == NULL)
918 free_http_response(rsp);
919 return cgi_error_memory();
922 exports = default_exports(csp, NULL);
925 free_http_response(rsp);
926 return cgi_error_memory();
929 #ifdef FEATURE_FORCE_LOAD
930 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
931 if (csp->http->ssl != 0)
932 #endif /* ndef FEATURE_FORCE_LOAD */
934 err = map_block_killer(exports, "force-support");
937 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
938 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
939 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
940 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
945 free_http_response(rsp);
946 return cgi_error_memory();
949 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
952 free_http_response(rsp);
953 return cgi_error_memory();
957 return finish_http_response(rsp);
963 /*********************************************************************
965 * Function : trust_url FIXME: I should be called distrust_url
967 * Description : Calls is_untrusted_url to determine if the URL is trusted
968 * and if not, returns a HTTP 304 response with a reject message.
971 * 1 : csp = Current client state (buffers, headers, etc...)
973 * Returns : NULL => trusted, else http_response.
975 *********************************************************************/
976 struct http_response *trust_url(struct client_state *csp)
978 struct http_response *rsp;
979 struct map * exports;
980 char buf[BUFFER_SIZE];
982 struct url_spec **tl;
987 * Don't bother to work on trusted URLs
989 if (!is_untrusted_url(csp))
995 * Else, prepare a response:
997 if (NULL == (rsp = alloc_http_response()))
999 return cgi_error_memory();
1002 exports = default_exports(csp, NULL);
1003 if (exports == NULL)
1005 free_http_response(rsp);
1006 return cgi_error_memory();
1010 * Export the protocol, host, port, and referrer information
1012 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1013 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1014 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1016 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1018 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1022 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
1028 free_http_response(rsp);
1029 return cgi_error_memory();
1033 * Export the trust list
1036 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1038 sprintf(buf, "<li>%s</li>\n", t->spec);
1039 string_append(&p, buf);
1041 err = map(exports, "trusted-referrers", 1, p, 0);
1047 free_http_response(rsp);
1048 return cgi_error_memory();
1052 * Export the trust info, if available
1054 if (csp->config->trust_info->first)
1056 struct list_entry *l;
1059 for (l = csp->config->trust_info->first; l ; l = l->next)
1061 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1062 string_append(&p, buf);
1064 err = map(exports, "trust-info", 1, p, 0);
1069 err = map_block_killer(exports, "have-trust-info");
1075 free_http_response(rsp);
1076 return cgi_error_memory();
1080 * Export the force prefix or the force conditional block killer
1082 #ifdef FEATURE_FORCE_LOAD
1083 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1084 #else /* ifndef FEATURE_FORCE_LOAD */
1085 err = map_block_killer(exports, "force-support");
1086 #endif /* ndef FEATURE_FORCE_LOAD */
1091 free_http_response(rsp);
1092 return cgi_error_memory();
1096 * Build the response
1098 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1101 free_http_response(rsp);
1102 return cgi_error_memory();
1105 return finish_http_response(rsp);
1107 #endif /* def FEATURE_TRUST */
1110 #ifdef FEATURE_FAST_REDIRECTS
1111 /*********************************************************************
1113 * Function : redirect_url
1115 * Description : Checks for redirection URLs and returns a HTTP redirect
1116 * to the destination URL, if necessary
1119 * 1 : csp = Current client state (buffers, headers, etc...)
1121 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1123 *********************************************************************/
1124 struct http_response *redirect_url(struct client_state *csp)
1127 struct http_response *rsp;
1128 char *redirect_mode = NULL;
1131 if ((csp->action->flags & ACTION_REDIRECT))
1133 q = csp->action->string[ACTION_STRING_REDIRECT];
1137 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1138 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1140 p = q = csp->http->path;
1141 log_error(LOG_LEVEL_REDIRECTS, "Decoding path: %s if necessary.", p);
1144 if (*p == '%') /* Escape sequence? */
1146 /* Yes, translate from hexadecimal to decimal */
1160 /* No, forward character. */
1166 p = q = csp->http->path;
1167 log_error(LOG_LEVEL_REDIRECTS, "Checking path for redirects: %s", p);
1170 * find the last URL encoded in the request
1172 while ((p = strstr(p, "http://")) != NULL)
1178 * if there was any, generate and return a HTTP redirect
1180 if (q != csp->http->path)
1182 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1184 if (NULL == (rsp = alloc_http_response()))
1186 return cgi_error_memory();
1189 if ( enlist_unique_header(rsp->headers, "Location", q)
1190 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1192 free_http_response(rsp);
1193 return cgi_error_memory();
1196 return finish_http_response(rsp);
1204 #endif /* def FEATURE_FAST_REDIRECTS */
1207 #ifdef FEATURE_IMAGE_BLOCKING
1208 /*********************************************************************
1210 * Function : is_imageurl
1212 * Description : Given a URL, decide whether it is an image or not,
1213 * using either the info from a previous +image action
1214 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1215 * is MSIE and not on a Mac, tell from the browser's accept
1219 * 1 : csp = Current client state (buffers, headers, etc...)
1221 * Returns : True (nonzero) if URL is an image, false (0)
1224 *********************************************************************/
1225 int is_imageurl(struct client_state *csp)
1227 #ifdef FEATURE_IMAGE_DETECT_MSIE
1230 tmp = get_header_value(csp->headers, "User-Agent:");
1231 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1233 tmp = get_header_value(csp->headers, "Accept:");
1234 if (tmp && strstr(tmp, "image/gif"))
1236 /* Client will accept HTML. If this seems counterintuitive,
1246 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1248 return ((csp->action->flags & ACTION_IMAGE) != 0);
1251 #endif /* def FEATURE_IMAGE_BLOCKING */
1254 #ifdef FEATURE_TRUST
1255 /*********************************************************************
1257 * Function : is_untrusted_url
1259 * Description : Should we "distrust" this URL (and block it)?
1261 * Yes if it matches a line in the trustfile, or if the
1262 * referrer matches a line starting with "+" in the
1267 * 1 : csp = Current client state (buffers, headers, etc...)
1269 * Returns : 0 => trusted, 1 => untrusted
1271 *********************************************************************/
1272 int is_untrusted_url(struct client_state *csp)
1274 struct file_list *fl;
1275 struct block_spec *b;
1276 struct url_spec **trusted_url;
1277 struct http_request rhttp[1];
1278 const char * referer;
1282 * If we don't have a trustlist, we trust everybody
1284 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1289 memset(rhttp, '\0', sizeof(*rhttp));
1292 * Do we trust the request URL itself?
1294 for (b = b->next; b ; b = b->next)
1296 if (url_match(b->url, csp->http))
1302 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1304 /* no referrer was supplied */
1310 * If not, do we maybe trust its referrer?
1312 err = parse_http_url(referer, rhttp, csp);
1318 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1320 if (url_match(*trusted_url, rhttp))
1322 /* if the URL's referrer is from a trusted referrer, then
1323 * add the target spec to the trustfile as an unblocked
1324 * domain and return NULL (which means it's OK).
1329 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1333 char * new_entry = strdup("~");
1335 string_append(&new_entry, csp->http->hostport);
1337 path = csp->http->path;
1338 if ( (path[0] == '/')
1340 && ((path_end = strchr(path + 2, '/')) != NULL))
1342 /* since this path points into a user's home space
1343 * be sure to include this spec in the trustfile.
1345 int path_len = path_end - path; /* save offset */
1346 path = strdup(path); /* Copy string */
1349 path_end = path + path_len; /* regenerate ptr to new buffer */
1350 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1352 string_join(&new_entry, path);
1355 if (new_entry != NULL)
1357 fprintf(fp, "%s\n", new_entry);
1362 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1363 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1373 #endif /* def FEATURE_TRUST */
1376 /*********************************************************************
1378 * Function : pcrs_filter_response
1380 * Description : Execute all text substitutions from all applying
1381 * +filter actions on the text buffer that's been accumulated
1382 * in csp->iob->buf. If this changes the contents, set
1383 * csp->content_length to the modified size and raise the
1384 * CSP_FLAG_MODIFIED flag.
1387 * 1 : csp = Current client state (buffers, headers, etc...)
1389 * Returns : a pointer to the (newly allocated) modified buffer.
1390 * or NULL if there were no hits or something went wrong
1392 *********************************************************************/
1393 char *pcrs_filter_response(struct client_state *csp)
1396 size_t size, prev_size;
1398 char *old = csp->iob->cur, *new = NULL;
1401 struct file_list *fl;
1402 struct re_filterfile_spec *b;
1403 struct list_entry *filtername;
1405 int i, found_filters = 0;
1410 if (csp->iob->cur >= csp->iob->eod)
1414 size = csp->iob->eod - csp->iob->cur;
1417 * Need to check the set of re_filterfiles...
1419 for (i = 0; i < MAX_AF_FILES; i++)
1432 if (0 == found_filters)
1434 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1439 * If the body has a "chunked" transfer-encoding,
1440 * get rid of it first, adjusting size and iob->eod
1442 if (csp->flags & CSP_FLAG_CHUNKED)
1444 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1445 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1449 csp->iob->eod = csp->iob->cur + size;
1450 csp->flags |= CSP_FLAG_MODIFIED;
1453 for (i = 0; i < MAX_AF_FILES; i++)
1456 if ((NULL == fl) || (NULL == fl->f))
1459 * For all applying +filter actions, look if a filter by that
1460 * name exists and if yes, execute it's pcrs_joblist on the
1463 for (b = fl->f; b; b = b->next)
1465 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1466 filtername ; filtername = filtername->next)
1468 if (strcmp(b->name, filtername->str) == 0)
1470 int current_hits = 0;
1472 if ( NULL == b->joblist )
1474 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1479 /* Apply all jobs from the joblist */
1480 for (job = b->joblist; NULL != job; job = job->next)
1482 current_hits += pcrs_execute(job, old, size, &new, &size);
1483 if (old != csp->iob->cur) free(old);
1487 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1488 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1490 hits += current_hits;
1497 * If there were no hits, destroy our copy and let
1498 * chat() use the original in csp->iob
1506 csp->flags |= CSP_FLAG_MODIFIED;
1507 csp->content_length = size;
1515 /*********************************************************************
1517 * Function : gif_deanimate_response
1519 * Description : Deanimate the GIF image that has been accumulated in
1520 * csp->iob->buf, set csp->content_length to the modified
1521 * size and raise the CSP_FLAG_MODIFIED flag.
1524 * 1 : csp = Current client state (buffers, headers, etc...)
1526 * Returns : a pointer to the (newly allocated) modified buffer.
1527 * or NULL in case something went wrong.
1529 *********************************************************************/
1530 char *gif_deanimate_response(struct client_state *csp)
1532 struct binbuffer *in, *out;
1534 size_t size = csp->iob->eod - csp->iob->cur;
1537 * If the body has a "chunked" transfer-encoding,
1538 * get rid of it first, adjusting size and iob->eod
1540 if (csp->flags & CSP_FLAG_CHUNKED)
1542 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1543 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1547 csp->iob->eod = csp->iob->cur + size;
1548 csp->flags |= CSP_FLAG_MODIFIED;
1551 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1552 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1554 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1558 in->buffer = csp->iob->cur;
1561 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1563 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1570 if ((int)size == out->offset)
1572 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1576 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1578 csp->content_length = out->offset;
1579 csp->flags |= CSP_FLAG_MODIFIED;
1589 /*********************************************************************
1591 * Function : jpeg_inspect_response
1596 * 1 : csp = Current client state (buffers, headers, etc...)
1598 * Returns : a pointer to the (newly allocated) modified buffer
1599 * or NULL in case something went wrong.
1601 *********************************************************************/
1602 char *jpeg_inspect_response(struct client_state *csp)
1604 struct binbuffer *in = NULL, *out = NULL;
1606 size_t size = csp->iob->eod - csp->iob->cur;
1609 * If the body has a "chunked" transfer-encoding,
1610 * get rid of it first, adjusting size and iob->eod
1612 if (csp->flags & CSP_FLAG_CHUNKED)
1614 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1615 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1619 csp->iob->eod = csp->iob->cur + size;
1620 csp->flags |= CSP_FLAG_MODIFIED;
1623 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1625 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1629 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1631 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1635 in->buffer = csp->iob->cur;
1639 * Calling jpeg_inspect has the side-effect of creating and
1640 * modifying the image buffer of "out" directly.
1642 if (jpeg_inspect(in, out))
1644 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1652 csp->content_length = out->offset;
1653 csp->flags |= CSP_FLAG_MODIFIED;
1663 /*********************************************************************
1665 * Function : remove_chunked_transfer_coding
1667 * Description : In-situ remove the "chunked" transfer coding as defined
1668 * in rfc2616 from a buffer.
1671 * 1 : buffer = Pointer to the text buffer
1672 * 2 : size = Number of bytes to be processed
1674 * Returns : The new size, i.e. the number of bytes from buffer which
1675 * are occupied by the stripped body, or 0 in case something
1678 *********************************************************************/
1679 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1682 unsigned int chunksize = 0;
1683 char *from_p, *to_p;
1686 from_p = to_p = buffer;
1688 if (sscanf(buffer, "%x", &chunksize) != 1)
1690 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1694 while (chunksize > 0)
1696 if (NULL == (from_p = strstr(from_p, "\r\n")))
1698 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1702 if ((newsize += chunksize) >= size)
1704 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1709 memmove(to_p, from_p, (size_t) chunksize);
1710 to_p = buffer + newsize;
1711 from_p += chunksize + 2;
1713 if (sscanf(from_p, "%x", &chunksize) != 1)
1715 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1720 /* FIXME: Should this get its own loglevel? */
1721 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1727 /*********************************************************************
1729 * Function : url_actions
1731 * Description : Gets the actions for this URL.
1734 * 1 : http = http_request request for blocked URLs
1735 * 2 : csp = Current client state (buffers, headers, etc...)
1739 *********************************************************************/
1740 void url_actions(struct http_request *http,
1741 struct client_state *csp)
1743 struct file_list *fl;
1744 struct url_actions *b;
1747 init_current_action(csp->action);
1749 for (i = 0; i < MAX_AF_FILES; i++)
1751 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1756 apply_url_actions(csp->action, http, b);
1763 /*********************************************************************
1765 * Function : apply_url_actions
1767 * Description : Applies a list of URL actions.
1770 * 1 : action = Destination.
1771 * 2 : http = Current URL
1772 * 3 : b = list of URL actions to apply
1776 *********************************************************************/
1777 void apply_url_actions(struct current_action_spec *action,
1778 struct http_request *http,
1779 struct url_actions *b)
1783 /* Should never happen */
1787 for (b = b->next; NULL != b; b = b->next)
1789 if (url_match(b->url, http))
1791 merge_current_action(action, b->action);
1797 /*********************************************************************
1799 * Function : forward_url
1801 * Description : Should we forward this to another proxy?
1804 * 1 : http = http_request request for current URL
1805 * 2 : csp = Current client state (buffers, headers, etc...)
1807 * Returns : Pointer to forwarding information.
1809 *********************************************************************/
1810 const struct forward_spec * forward_url(struct http_request *http,
1811 struct client_state *csp)
1813 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1814 struct forward_spec *fwd = csp->config->forward;
1823 if (url_match(fwd->url, http))
1834 /*********************************************************************
1836 * Function : direct_response
1838 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
1839 * request and if so, return a HTTP 501 to the client.
1841 * FIXME: I have a stupid name and I should handle the
1842 * requests properly. Still, what we do here is rfc-
1843 * compliant, whereas ignoring or forwarding are not.
1846 * 1 : csp = Current client state (buffers, headers, etc...)
1848 * Returns : http_response if , NULL if nonmatch or handler fail
1850 *********************************************************************/
1851 struct http_response *direct_response(struct client_state *csp)
1853 struct http_response *rsp;
1854 struct list_entry *p;
1856 if ((0 == strcmpic(csp->http->gpc, "trace"))
1857 || (0 == strcmpic(csp->http->gpc, "options")))
1859 for (p = csp->headers->first; (p != NULL) ; p = p->next)
1861 if (!strncmp("Max-Forwards:", p->str, 13)
1862 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
1864 /* FIXME: We could handle at least TRACE here,
1865 but that would require a verbatim copy of
1866 the request which we don't have anymore */
1868 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
1870 /* Get mem for response or fail*/
1871 if (NULL == (rsp = alloc_http_response()))
1873 return cgi_error_memory();
1876 if (NULL == (rsp->status = strdup("501 Not Implemented")))
1878 free_http_response(rsp);
1879 return cgi_error_memory();
1883 return(finish_http_response(rsp));