1 const char filters_rcs[] = "$Id: filters.c,v 1.58.2.2 2002/08/01 17:18:28 oes Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa//current/Attic/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', and `trust_url'
14 * Copyright : Written by and Copyright (C) 2001 the SourceForge
15 * Privoxy team. http://www.privoxy.org/
17 * Based on the Internet Junkbuster originally written
18 * by and Copyright (C) 1997 Anonymous Coders and
19 * Junkbusters Corporation. http://www.junkbusters.com
21 * This program is free software; you can redistribute it
22 * and/or modify it under the terms of the GNU General
23 * Public License as published by the Free Software
24 * Foundation; either version 2 of the License, or (at
25 * your option) any later version.
27 * This program is distributed in the hope that it will
28 * be useful, but WITHOUT ANY WARRANTY; without even the
29 * implied warranty of MERCHANTABILITY or FITNESS FOR A
30 * PARTICULAR PURPOSE. See the GNU General Public
31 * License for more details.
33 * The GNU General Public License should be included with
34 * this file. If not, you can view it at
35 * http://www.gnu.org/copyleft/gpl.html
36 * or write to the Free Software Foundation, Inc., 59
37 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
41 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
42 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
44 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
45 * - Bugfix: Executing a filters without jobs no longer results in
46 * turing off *all* filters.
47 * - Security fix: Malicious web servers can't cause a seg fault
48 * through bogus chunk sizes anymore
50 * Revision 1.58 2002/04/24 02:11:17 oes
51 * Jon's multiple AF patch: url_actions now evaluates rules
54 * Revision 1.57 2002/04/08 20:38:34 swa
57 * Revision 1.56 2002/04/05 15:51:24 oes
58 * - bugfix: error-pages now get correct request protocol
59 * - fix for invalid HTML in trust info
61 * Revision 1.55 2002/04/02 16:13:51 oes
62 * Fix: No "Go there anyway" for SSL
64 * Revision 1.54 2002/04/02 14:55:56 oes
65 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
67 * Revision 1.53 2002/03/26 22:29:54 swa
68 * we have a new homepage!
70 * Revision 1.52 2002/03/24 16:35:57 jongfoster
73 * Revision 1.51 2002/03/24 15:23:33 jongfoster
76 * Revision 1.50 2002/03/24 13:25:43 swa
77 * name change related issues
79 * Revision 1.49 2002/03/16 20:29:14 oes
82 * Revision 1.48 2002/03/13 20:25:34 oes
83 * Better logging for content filters
85 * Revision 1.47 2002/03/13 00:30:52 jongfoster
87 * Added option of always sending redirect for imageblock,
88 * currently disabled with #if 0.
90 * Revision 1.46 2002/03/12 01:42:49 oes
91 * Introduced modular filters
93 * Revision 1.45 2002/03/08 16:47:50 oes
94 * Added choice beween GIF and PNG built-in images
96 * Revision 1.44 2002/03/07 03:49:31 oes
97 * - Fixed compiler warnings etc
98 * - Changed built-in images from GIF to PNG
99 * (with regard to Unisys patent issue)
100 * - Added a 4x4 pattern PNG which is less intrusive
101 * than the logo but also clearly marks the deleted banners
103 * Revision 1.43 2002/01/22 23:51:59 jongfoster
104 * Replacing strsav() with the safer string_append().
106 * Adding missing html_encode() to error message generators. Where encoded
107 * and unencoded versions of a string were provided, removing the unencoded
110 * Revision 1.42 2002/01/17 21:00:32 jongfoster
111 * Moving all our URL and URL pattern parsing code to urlmatch.c.
113 * Using a single, simple url_match(pattern,url) function - rather than
114 * the 3-line match routine which was repeated all over the place.
116 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
118 * Using parse_http_url() to parse URLs without faking a HTTP
119 * request line for parse_http_request().
121 * Revision 1.41 2001/11/13 00:14:07 jongfoster
122 * Fixing stupid bug now I've figured out what || means.
123 * (It always returns 0 or 1, not one of it's paramaters.)
125 * Revision 1.40 2001/10/26 17:37:55 oes
126 * - Re-enabled Netscape 200/404 bug workaround in block_url():
127 * - Removed OS/2 special case
128 * - Made block_url() independant from sed() having been run
129 * - Made trust_url independant from sed() having been run
130 * - Made is_imageurl independant from sed() having been run.
131 * It now checks User-Agent: and Accept: by itself.
134 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
135 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
136 * threads to call select() simultaneously. So, it's time to do a real, live,
137 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
138 * (native). Both versions will work, but using __OS2__ offers multi-threading.
140 * Revision 1.38 2001/10/23 21:32:33 jongfoster
141 * Adding error-checking to selected functions
143 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
144 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
145 * filters.c. Added a FIXME in front of the offending code. I'll gladly
146 * put in a better/more robust fix for all parties if one is presented...
147 * It seems that just returning 200 instead of 404 would pretty much fix
148 * it for everyone, but I don't know all the history of the problem.
150 * Revision 1.36 2001/10/10 16:44:16 oes
151 * Added match_portlist function
153 * Revision 1.35 2001/10/07 15:41:23 oes
154 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
156 * New function remove_chunked_transfer_coding that strips chunked
157 * transfer coding to plain and is called by pcrs_filter_response
158 * and gif_deanimate_response if neccessary
160 * Improved handling of zero-change re_filter runs
162 * pcrs_filter_response and gif_deanimate_response now remove
163 * chunked transfer codeing before processing the body.
165 * Revision 1.34 2001/09/20 15:49:36 steudten
167 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
168 * See cgi.c fill_template().
170 * Revision 1.33 2001/09/16 17:05:14 jongfoster
171 * Removing unused #include showarg.h
173 * Revision 1.32 2001/09/16 13:21:27 jongfoster
174 * Changes to use new list functions.
176 * Revision 1.31 2001/09/16 11:38:02 jongfoster
177 * Splitting fill_template() into 2 functions:
178 * template_load() loads the file
179 * template_fill() performs the PCRS regexps.
180 * This is because the CGI edit interface has a "table row"
181 * template which is used many times in the page - this
182 * change means it's only loaded from disk once.
184 * Revision 1.30 2001/09/16 11:00:10 jongfoster
185 * New function alloc_http_response, for symmetry with free_http_response
187 * Revision 1.29 2001/09/13 23:32:40 jongfoster
188 * Moving image data to cgi.c rather than cgi.h
189 * Fixing a GPF under Win32 (and any other OS that protects global
190 * constants from being written to).
192 * Revision 1.28 2001/09/10 10:18:51 oes
193 * Silenced compiler warnings
195 * Revision 1.27 2001/08/05 16:06:20 jongfoster
196 * Modifiying "struct map" so that there are now separate header and
197 * "map_entry" structures. This means that functions which modify a
198 * map no longer need to return a pointer to the modified map.
199 * Also, it no longer reverses the order of the entries (which may be
200 * important with some advanced template substitutions).
202 * Revision 1.26 2001/07/30 22:08:36 jongfoster
203 * Tidying up #defines:
204 * - All feature #defines are now of the form FEATURE_xxx
205 * - Permanently turned off WIN_GUI_EDIT
206 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
208 * Revision 1.25 2001/07/26 10:09:46 oes
209 * Made browser detection a little less naive
211 * Revision 1.24 2001/07/25 17:22:51 oes
212 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
214 * Revision 1.23 2001/07/23 13:40:12 oes
215 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
217 * Revision 1.22 2001/07/18 12:29:34 oes
218 * - Made gif_deanimate_response respect
219 * csp->action->string[ACTION_STRING_DEANIMATE]
220 * - Logging cosmetics
222 * Revision 1.21 2001/07/13 13:59:53 oes
223 * - Introduced gif_deanimate_response which shares the
224 * generic content modification interface of pcrs_filter_response
225 * and acts as a wrapper to deanimate.c:gif_deanimate()
226 * - Renamed re_process_buffer to pcrs_filter_response
227 * - pcrs_filter_response now returns NULL on failiure
228 * - Removed all #ifdef PCRS
230 * Revision 1.20 2001/07/01 17:01:04 oes
231 * Added comments and missing return statement in is_untrusted_url()
233 * Revision 1.19 2001/06/29 21:45:41 oes
234 * Indentation, CRLF->LF, Tab-> Space
236 * Revision 1.18 2001/06/29 13:27:38 oes
237 * - Cleaned up, renamed and reorderd functions
238 * and improved comments
241 * - Ported to CGI platform. Now delivers
242 * http_response or NULL
243 * - Unified HTML and GIF generation (moved image detection
244 * and GIF generation here from jcc.c:chat())
245 * - Fixed HTTP status to:
246 * - 403 (Forbidden) for the "blocked" HTML message
247 * - 200 (OK) for GIF answers
248 * - 302 (Redirect) for redirect to GIF
251 * - Ported to CGI platform. Now delivers
252 * http_response or NULL
253 * - Separated detection of untrusted URL into
254 * (bool)is_untrusted_url
255 * - Added enforcement of untrusted requests
257 * - Moved redirect_url() from cgi.c to here
258 * and ported it to the CGI platform
260 * - Removed logentry from cancelled commit
262 * Revision 1.17 2001/06/09 10:55:28 jongfoster
263 * Changing BUFSIZ ==> BUFFER_SIZE
265 * Revision 1.16 2001/06/07 23:10:26 jongfoster
266 * Allowing unanchored domain patterns to back off and retry
267 * if they partially match. Optimized right-anchored patterns.
268 * Moving ACL and forward files into config file.
269 * Replacing struct gateway with struct forward_spec
271 * Revision 1.15 2001/06/03 19:12:00 oes
272 * extracted-CGI relevant stuff
274 * Revision 1.14 2001/06/01 10:30:55 oes
275 * Added optional left-anchoring to domaincmp
277 * Revision 1.13 2001/05/31 21:21:30 jongfoster
278 * Permissionsfile / actions file changes:
279 * - Changed "permission" to "action" throughout
280 * - changes to file format to allow string parameters
281 * - Moved helper functions to actions.c
283 * Revision 1.12 2001/05/31 17:35:20 oes
285 * - Enhanced domain part globbing with infix and prefix asterisk
286 * matching and optional unanchored operation
288 * Revision 1.11 2001/05/29 11:53:23 oes
289 * "See why" link added to "blocked" page
291 * Revision 1.10 2001/05/29 09:50:24 jongfoster
292 * Unified blocklist/imagelist/permissionslist.
293 * File format is still under discussion, but the internal changes
296 * Also modified interceptor behaviour:
297 * - We now intercept all URLs beginning with one of the following
298 * prefixes (and *only* these prefixes):
300 * * http://ijbswa.sf.net/config/
301 * * http://ijbswa.sourceforge.net/config/
302 * - New interceptors "home page" - go to http://i.j.b/ to see it.
303 * - Internal changes so that intercepted and fast redirect pages
304 * are not replaced with an image.
305 * - Interceptors now have the option to send a binary page direct
306 * to the client. (i.e. ijb-send-banner uses this)
307 * - Implemented show-url-info interceptor. (Which is why I needed
308 * the above interceptors changes - a typical URL is
309 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
310 * The previous mechanism would not have intercepted that, and
311 * if it had been intercepted then it then it would have replaced
314 * Revision 1.9 2001/05/27 22:17:04 oes
316 * - re_process_buffer no longer writes the modified buffer
317 * to the client, which was very ugly. It now returns the
318 * buffer, which it is then written by chat.
320 * - content_length now adjusts the Content-Length: header
321 * for modified documents rather than crunch()ing it.
322 * (Length info in csp->content_length, which is 0 for
323 * unmodified documents)
325 * - For this to work, sed() is called twice when filtering.
327 * Revision 1.8 2001/05/26 17:13:28 jongfoster
328 * Filled in a function comment.
330 * Revision 1.7 2001/05/26 15:26:15 jongfoster
331 * ACL feature now provides more security by immediately dropping
332 * connections from untrusted hosts.
334 * Revision 1.6 2001/05/26 00:28:36 jongfoster
335 * Automatic reloading of config file.
336 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
337 * Most of the global variables have been moved to a new
338 * struct configuration_spec, accessed through csp->config->globalname
339 * Most of the globals remaining are used by the Win32 GUI.
341 * Revision 1.5 2001/05/25 22:34:30 jongfoster
344 * Revision 1.4 2001/05/22 18:46:04 oes
346 * - Enabled filtering banners by size rather than URL
347 * by adding patterns that replace all standard banner
348 * sizes with the "Junkbuster" gif to the re_filterfile
350 * - Enabled filtering WebBugs by providing a pattern
351 * which kills all 1x1 images
353 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
354 * which is selected by the (nonstandard and therefore
355 * capital) letter 'U' in the option string.
356 * It causes the quantifiers to be ungreedy by default.
357 * Appending a ? turns back to greedy (!).
359 * - Added a new interceptor ijb-send-banner, which
360 * sends back the "Junkbuster" gif. Without imagelist or
361 * MSIE detection support, or if tinygif = 1, or the
362 * URL isn't recognized as an imageurl, a lame HTML
363 * explanation is sent instead.
365 * - Added new feature, which permits blocking remote
366 * script redirects and firing back a local redirect
368 * The feature is conditionally compiled, i.e. it
369 * can be disabled with --disable-fast-redirects,
370 * plus it must be activated by a "fast-redirects"
371 * line in the config file, has its own log level
372 * and of course wants to be displayed by show-proxy-args
373 * Note: Boy, all the #ifdefs in 1001 locations and
374 * all the fumbling with configure.in and acconfig.h
375 * were *way* more work than the feature itself :-(
377 * - Because a generic redirect template was needed for
378 * this, tinygif = 3 now uses the same.
380 * - Moved GIFs, and other static HTTP response templates
385 * - Removed some >400 CRs again (Jon, you really worked
388 * Revision 1.3 2001/05/20 16:44:47 jongfoster
389 * Removing last hardcoded Junkbusters.com URLs.
391 * Revision 1.2 2001/05/20 01:21:20 jongfoster
392 * Version 2.9.4 checkin.
393 * - Merged popupfile and cookiefile, and added control over PCRS
394 * filtering, in new "permissionsfile".
395 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
396 * file error you now get a message box (in the Win32 GUI) rather
397 * than the program exiting with no explanation.
398 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
400 * - Removed tabs from "config"
401 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
402 * - Bumped up version number.
404 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
405 * Initial import of version 2.9.3 source tree
408 *********************************************************************/
414 #include <sys/types.h>
423 #endif /* ndef __OS2__ */
424 #include <netinet/in.h>
426 #include <winsock2.h>
427 #endif /* ndef _WIN32 */
431 #endif /* def __OS2__ */
439 #include "jbsockets.h"
440 #include "miscutil.h"
444 #include "deanimate.h"
445 #include "urlmatch.h"
451 const char filters_h_rcs[] = FILTERS_H_VERSION;
453 /* Fix a problem with Solaris. There should be no effect on other
455 * Solaris's isspace() is a macro which uses it's argument directly
456 * as an array index. Therefore we need to make sure that high-bit
457 * characters generate +ve values, and ideally we also want to make
458 * the argument match the declared parameter type of "int".
460 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
464 /*********************************************************************
466 * Function : block_acl
468 * Description : Block this request?
469 * Decide yes or no based on ACL file.
472 * 1 : dst = The proxy or gateway address this is going to.
473 * Or NULL to check all possible targets.
474 * 2 : csp = Current client state (buffers, headers, etc...)
475 * Also includes the client IP address.
477 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
479 *********************************************************************/
480 int block_acl(struct access_control_addr *dst, struct client_state *csp)
482 struct access_control_list *acl = csp->config->acl;
484 /* if not using an access control list, then permit the connection */
490 /* search the list */
493 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
497 /* Just want to check if they have any access */
498 if (acl->action == ACL_PERMIT)
503 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
504 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
506 if (acl->action == ACL_PERMIT)
524 /*********************************************************************
526 * Function : acl_addr
528 * Description : Called from `load_config' to parse an ACL address.
531 * 1 : aspec = String specifying ACL address.
532 * 2 : aca = struct access_control_addr to fill in.
534 * Returns : 0 => Ok, everything else is an error.
536 *********************************************************************/
537 int acl_addr(char *aspec, struct access_control_addr *aca)
539 int i, masklength, port;
545 if ((p = strchr(aspec, '/')) != NULL)
549 if (ijb_isdigit(*p) == 0)
553 masklength = atoi(p);
556 if ((masklength < 0) || (masklength > 32))
561 if ((p = strchr(aspec, ':')) != NULL)
565 if (ijb_isdigit(*p) == 0)
574 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
576 if (aca->addr == INADDR_NONE)
581 /* build the netmask */
583 for (i=1; i <= masklength ; i++)
585 aca->mask |= (1 << (32 - i));
588 /* now mask off the host portion of the ip address
589 * (i.e. save on the network portion of the address).
591 aca->addr = aca->addr & aca->mask;
596 #endif /* def FEATURE_ACL */
599 /*********************************************************************
601 * Function : match_portlist
603 * Description : Check if a given number is covered by a comma
604 * separated list of numbers and ranges (a,b-c,d,..)
607 * 1 : portlist = String with list
608 * 2 : port = port to check
610 * Returns : 0 => no match
613 *********************************************************************/
614 int match_portlist(const char *portlist, int port)
616 char *min, *max, *next, *portlist_copy;
618 min = next = portlist_copy = strdup(portlist);
621 * Zero-terminate first item and remember offset for next
623 if (NULL != (next = strchr(portlist_copy, (int) ',')))
629 * Loop through all items, checking for match
633 if (NULL == (max = strchr(min, (int) '-')))
636 * No dash, check for equality
638 if (port == atoi(min))
647 * This is a range, so check if between min and max,
648 * or, if max was omitted, between min and 65K
651 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
665 * Zero-terminate next item and remember offset for n+1
667 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
679 /*********************************************************************
681 * Function : block_url
683 * Description : Called from `chat'. Check to see if we need to block this.
686 * 1 : csp = Current client state (buffers, headers, etc...)
688 * Returns : NULL => unblocked, else HTTP block response
690 *********************************************************************/
691 struct http_response *block_url(struct client_state *csp)
693 #ifdef FEATURE_IMAGE_BLOCKING
695 #endif /* def FEATURE_IMAGE_BLOCKING */
696 struct http_response *rsp;
699 * If it's not blocked, don't block it ;-)
701 if ((csp->action->flags & ACTION_BLOCK) == 0)
707 * Else, prepare a response
709 if (NULL == (rsp = alloc_http_response()))
711 return cgi_error_memory();
715 * If it's an image-url, send back an image or redirect
716 * as specified by the relevant +image action
718 #ifdef FEATURE_IMAGE_BLOCKING
719 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
722 /* determine HOW images should be blocked */
723 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
725 #if 1 /* Two alternative strategies, use this one for now: */
727 /* and handle accordingly: */
728 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
730 rsp->body = bindup(image_pattern_data, image_pattern_length);
731 if (rsp->body == NULL)
733 free_http_response(rsp);
734 return cgi_error_memory();
736 rsp->content_length = image_pattern_length;
738 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
740 free_http_response(rsp);
741 return cgi_error_memory();
745 else if (0 == strcmpic(p, "blank"))
747 rsp->body = bindup(image_blank_data, image_blank_length);
748 if (rsp->body == NULL)
750 free_http_response(rsp);
751 return cgi_error_memory();
753 rsp->content_length = image_blank_length;
755 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
757 free_http_response(rsp);
758 return cgi_error_memory();
764 rsp->status = strdup("302 Local Redirect from Privoxy");
765 if (rsp->status == NULL)
767 free_http_response(rsp);
768 return cgi_error_memory();
771 if (enlist_unique_header(rsp->headers, "Location", p))
773 free_http_response(rsp);
774 return cgi_error_memory();
778 #else /* Following code is disabled for now */
780 /* and handle accordingly: */
781 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
783 p = CGI_PREFIX "send-banner?type=pattern";
785 else if (0 == strcmpic(p, "blank"))
787 p = CGI_PREFIX "send-banner?type=blank";
789 rsp->status = strdup("302 Local Redirect from Privoxy");
790 if (rsp->status == NULL)
792 free_http_response(rsp);
793 return cgi_error_memory();
796 if (enlist_unique_header(rsp->headers, "Location", p))
798 free_http_response(rsp);
799 return cgi_error_memory();
801 #endif /* Preceeding code is disabled for now */
804 #endif /* def FEATURE_IMAGE_BLOCKING */
807 * Else, generate an HTML "blocked" message:
811 struct map * exports;
814 * Workaround for stupid Netscape bug which prevents
815 * pages from being displayed if loading a referenced
816 * JavaScript or style sheet fails. So make it appear
817 * as if it succeeded.
819 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
820 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
821 && !strstr(p, "Gecko") /* save Mozilla, */
822 && !strstr(p, "compatible") /* MSIE */
823 && !strstr(p, "Opera")) /* and Opera. */
825 rsp->status = strdup("200 Request for blocked URL");
829 rsp->status = strdup("404 Request for blocked URL");
832 if (rsp->status == NULL)
834 free_http_response(rsp);
835 return cgi_error_memory();
838 exports = default_exports(csp, NULL);
841 free_http_response(rsp);
842 return cgi_error_memory();
845 #ifdef FEATURE_FORCE_LOAD
846 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
847 if (csp->http->ssl != 0)
848 #endif /* ndef FEATURE_FORCE_LOAD */
850 err = map_block_killer(exports, "force-support");
853 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
854 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
855 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
860 free_http_response(rsp);
861 return cgi_error_memory();
864 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
867 free_http_response(rsp);
868 return cgi_error_memory();
872 return finish_http_response(rsp);
878 /*********************************************************************
880 * Function : trust_url FIXME: I should be called distrust_url
882 * Description : Calls is_untrusted_url to determine if the URL is trusted
883 * and if not, returns a HTTP 304 response with a reject message.
886 * 1 : csp = Current client state (buffers, headers, etc...)
888 * Returns : NULL => trusted, else http_response.
890 *********************************************************************/
891 struct http_response *trust_url(struct client_state *csp)
893 struct http_response *rsp;
894 struct map * exports;
895 char buf[BUFFER_SIZE];
897 struct url_spec **tl;
902 * Don't bother to work on trusted URLs
904 if (!is_untrusted_url(csp))
910 * Else, prepare a response:
912 if (NULL == (rsp = alloc_http_response()))
914 return cgi_error_memory();
917 exports = default_exports(csp, NULL);
920 free_http_response(rsp);
921 return cgi_error_memory();
925 * Export the protocol, host, port, and referrer information
927 err = map(exports, "hostport", 1, csp->http->hostport, 1);
928 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
929 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
931 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
933 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
937 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
943 free_http_response(rsp);
944 return cgi_error_memory();
948 * Export the trust list
951 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
953 sprintf(buf, "<li>%s</li>\n", t->spec);
954 string_append(&p, buf);
956 err = map(exports, "trusted-referrers", 1, p, 0);
961 free_http_response(rsp);
962 return cgi_error_memory();
966 * Export the trust info, if available
968 if (csp->config->trust_info->first)
970 struct list_entry *l;
973 for (l = csp->config->trust_info->first; l ; l = l->next)
975 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
976 string_append(&p, buf);
978 err = map(exports, "trust-info", 1, p, 0);
982 err = map_block_killer(exports, "have-trust-info");
988 free_http_response(rsp);
989 return cgi_error_memory();
993 * Export the force prefix or the force conditional block killer
995 #ifdef FEATURE_FORCE_LOAD
996 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
997 #else /* ifndef FEATURE_FORCE_LOAD */
998 err = map_block_killer(exports, "force-support");
999 #endif /* ndef FEATURE_FORCE_LOAD */
1004 free_http_response(rsp);
1005 return cgi_error_memory();
1009 * Build the response
1011 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1014 free_http_response(rsp);
1015 return cgi_error_memory();
1018 return finish_http_response(rsp);
1020 #endif /* def FEATURE_TRUST */
1023 #ifdef FEATURE_FAST_REDIRECTS
1024 /*********************************************************************
1026 * Function : redirect_url
1028 * Description : Checks for redirection URLs and returns a HTTP redirect
1029 * to the destination URL, if necessary
1032 * 1 : csp = Current client state (buffers, headers, etc...)
1034 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1036 *********************************************************************/
1037 struct http_response *redirect_url(struct client_state *csp)
1040 struct http_response *rsp;
1042 p = q = csp->http->path;
1043 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1046 * find the last URL encoded in the request
1048 while ((p = strstr(p, "http://")) != NULL)
1054 * if there was any, generate and return a HTTP redirect
1056 if (q != csp->http->path)
1058 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1060 if (NULL == (rsp = alloc_http_response()))
1062 return cgi_error_memory();
1065 if ( enlist_unique_header(rsp->headers, "Location", q)
1066 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1068 free_http_response(rsp);
1069 return cgi_error_memory();
1072 return finish_http_response(rsp);
1080 #endif /* def FEATURE_FAST_REDIRECTS */
1083 #ifdef FEATURE_IMAGE_BLOCKING
1084 /*********************************************************************
1086 * Function : is_imageurl
1088 * Description : Given a URL, decide whether it is an image or not,
1089 * using either the info from a previous +image action
1090 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1091 * is MSIE and not on a Mac, tell from the browser's accept
1095 * 1 : csp = Current client state (buffers, headers, etc...)
1097 * Returns : True (nonzero) if URL is an image, false (0)
1100 *********************************************************************/
1101 int is_imageurl(struct client_state *csp)
1103 #ifdef FEATURE_IMAGE_DETECT_MSIE
1106 tmp = get_header_value(csp->headers, "User-Agent:");
1107 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1109 tmp = get_header_value(csp->headers, "Accept:");
1110 if (tmp && strstr(tmp, "image/gif"))
1112 /* Client will accept HTML. If this seems counterintuitive,
1122 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1124 return ((csp->action->flags & ACTION_IMAGE) != 0);
1127 #endif /* def FEATURE_IMAGE_BLOCKING */
1130 #ifdef FEATURE_TRUST
1131 /*********************************************************************
1133 * Function : is_untrusted_url
1135 * Description : Should we "distrust" this URL (and block it)?
1137 * Yes if it matches a line in the trustfile, or if the
1138 * referrer matches a line starting with "+" in the
1143 * 1 : csp = Current client state (buffers, headers, etc...)
1145 * Returns : 0 => trusted, 1 => untrusted
1147 *********************************************************************/
1148 int is_untrusted_url(struct client_state *csp)
1150 struct file_list *fl;
1151 struct block_spec *b;
1152 struct url_spec **trusted_url;
1153 struct http_request rhttp[1];
1154 const char * referer;
1158 * If we don't have a trustlist, we trust everybody
1160 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1165 memset(rhttp, '\0', sizeof(*rhttp));
1168 * Do we trust the request URL itself?
1170 for (b = b->next; b ; b = b->next)
1172 if (url_match(b->url, csp->http))
1178 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1180 /* no referrer was supplied */
1186 * If not, do we maybe trust its referrer?
1188 err = parse_http_url(referer, rhttp, csp);
1194 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1196 if (url_match(*trusted_url, rhttp))
1198 /* if the URL's referrer is from a trusted referrer, then
1199 * add the target spec to the trustfile as an unblocked
1200 * domain and return NULL (which means it's OK).
1205 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1209 char * new_entry = strdup("~");
1211 string_append(&new_entry, csp->http->hostport);
1213 path = csp->http->path;
1214 if ( (path[0] == '/')
1216 && ((path_end = strchr(path + 2, '/')) != NULL))
1218 /* since this path points into a user's home space
1219 * be sure to include this spec in the trustfile.
1221 int path_len = path_end - path; /* save offset */
1222 path = strdup(path); /* Copy string */
1225 path_end = path + path_len; /* regenerate ptr to new buffer */
1226 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1228 string_join(&new_entry, path);
1231 if (new_entry != NULL)
1233 fprintf(fp, "%s\n", new_entry);
1238 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1239 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1249 #endif /* def FEATURE_TRUST */
1252 /*********************************************************************
1254 * Function : pcrs_filter_response
1256 * Description : Ecexute all text substitutions from all applying
1257 * +filter actions on the text buffer that's been accumulated
1258 * in csp->iob->buf. If this changes the contents, set
1259 * csp->content_length to the modified size and raise the
1260 * CSP_FLAG_MODIFIED flag.
1263 * 1 : csp = Current client state (buffers, headers, etc...)
1265 * Returns : a pointer to the (newly allocated) modified buffer.
1266 * or NULL if there were no hits or something went wrong
1268 *********************************************************************/
1269 char *pcrs_filter_response(struct client_state *csp)
1274 char *old = csp->iob->cur, *new = NULL;
1277 struct file_list *fl;
1278 struct re_filterfile_spec *b;
1279 struct list_entry *filtername;
1284 if (csp->iob->cur >= csp->iob->eod)
1288 size = csp->iob->eod - csp->iob->cur;
1290 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1292 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1297 * If the body has a "chunked" transfer-encoding,
1298 * get rid of it first, adjusting size and iob->eod
1300 if (csp->flags & CSP_FLAG_CHUNKED)
1302 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1303 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1307 csp->iob->eod = csp->iob->cur + size;
1308 csp->flags |= CSP_FLAG_MODIFIED;
1312 * For all applying +filter actions, look if a filter by that
1313 * name exists and if yes, execute it's pcrs_joblist on the
1316 for (b = fl->f; b; b = b->next)
1318 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1319 filtername ; filtername = filtername->next)
1321 if (strcmp(b->name, filtername->str) == 0)
1323 int current_hits = 0;
1325 if ( NULL == b->joblist )
1327 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1331 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1332 csp->http->hostport, csp->http->path, size, b->name);
1334 /* Apply all jobs from the joblist */
1335 for (job = b->joblist; NULL != job; job = job->next)
1337 current_hits += pcrs_execute(job, old, size, &new, &size);
1338 if (old != csp->iob->cur) free(old);
1342 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1343 hits += current_hits;
1349 * If there were no hits, destroy our copy and let
1350 * chat() use the original in csp->iob
1358 csp->flags |= CSP_FLAG_MODIFIED;
1359 csp->content_length = size;
1367 /*********************************************************************
1369 * Function : gif_deanimate_response
1371 * Description : Deanimate the GIF image that has been accumulated in
1372 * csp->iob->buf, set csp->content_length to the modified
1373 * size and raise the CSP_FLAG_MODIFIED flag.
1376 * 1 : csp = Current client state (buffers, headers, etc...)
1378 * Returns : a pointer to the (newly allocated) modified buffer.
1379 * or NULL in case something went wrong.
1381 *********************************************************************/
1382 char *gif_deanimate_response(struct client_state *csp)
1384 struct binbuffer *in, *out;
1386 size_t size = csp->iob->eod - csp->iob->cur;
1389 * If the body has a "chunked" transfer-encoding,
1390 * get rid of it first, adjusting size and iob->eod
1392 if (csp->flags & CSP_FLAG_CHUNKED)
1394 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1395 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1399 csp->iob->eod = csp->iob->cur + size;
1400 csp->flags |= CSP_FLAG_MODIFIED;
1403 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1404 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1406 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1410 in->buffer = csp->iob->cur;
1413 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1415 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1422 if ((int)size == out->offset)
1424 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1428 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1430 csp->content_length = out->offset;
1431 csp->flags |= CSP_FLAG_MODIFIED;
1441 /*********************************************************************
1443 * Function : remove_chunked_transfer_coding
1445 * Description : In-situ remove the "chunked" transfer coding as defined
1446 * in rfc2616 from a buffer.
1449 * 1 : buffer = Pointer to the text buffer
1450 * 2 : size = Number of bytes to be processed
1452 * Returns : The new size, i.e. the number of bytes from buffer which
1453 * are occupied by the stripped body, or 0 in case something
1456 *********************************************************************/
1457 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1460 unsigned int chunksize = 0;
1461 char *from_p, *to_p;
1464 from_p = to_p = buffer;
1466 if (sscanf(buffer, "%x", &chunksize) != 1)
1468 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1472 while (chunksize > 0)
1474 if (NULL == (from_p = strstr(from_p, "\r\n")))
1476 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1480 if ((newsize += chunksize) >= size)
1482 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1487 memmove(to_p, from_p, (size_t) chunksize);
1488 to_p = buffer + newsize;
1489 from_p += chunksize + 2;
1491 if (sscanf(from_p, "%x", &chunksize) != 1)
1493 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1498 /* FIXME: Should this get its own loglevel? */
1499 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1505 /*********************************************************************
1507 * Function : url_actions
1509 * Description : Gets the actions for this URL.
1512 * 1 : http = http_request request for blocked URLs
1513 * 2 : csp = Current client state (buffers, headers, etc...)
1517 *********************************************************************/
1518 void url_actions(struct http_request *http,
1519 struct client_state *csp)
1521 struct file_list *fl;
1522 struct url_actions *b;
1525 init_current_action(csp->action);
1527 for (i = 0; i < MAX_ACTION_FILES; i++)
1529 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1534 apply_url_actions(csp->action, http, b);
1541 /*********************************************************************
1543 * Function : apply_url_actions
1545 * Description : Applies a list of URL actions.
1548 * 1 : action = Destination.
1549 * 2 : http = Current URL
1550 * 3 : b = list of URL actions to apply
1554 *********************************************************************/
1555 void apply_url_actions(struct current_action_spec *action,
1556 struct http_request *http,
1557 struct url_actions *b)
1561 /* Should never happen */
1565 for (b = b->next; NULL != b; b = b->next)
1567 if (url_match(b->url, http))
1569 merge_current_action(action, b->action);
1575 /*********************************************************************
1577 * Function : forward_url
1579 * Description : Should we forward this to another proxy?
1582 * 1 : http = http_request request for current URL
1583 * 2 : csp = Current client state (buffers, headers, etc...)
1585 * Returns : Pointer to forwarding information.
1587 *********************************************************************/
1588 const struct forward_spec * forward_url(struct http_request *http,
1589 struct client_state *csp)
1591 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1592 struct forward_spec *fwd = csp->config->forward;
1601 if (url_match(fwd->url, http))
1612 /*********************************************************************
1614 * Function : direct_response
1616 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
1617 * request and if so, return a HTTP 501 to the client.
1619 * FIXME: I have a stupid name and I should handle the
1620 * requests properly. Still, what we do here is rfc-
1621 * compliant, whereas ignoring or forwarding are not.
1624 * 1 : csp = Current client state (buffers, headers, etc...)
1626 * Returns : http_response if , NULL if nonmatch or handler fail
1628 *********************************************************************/
1629 struct http_response *direct_response(struct client_state *csp)
1631 struct http_response *rsp;
1632 struct list_entry *p;
1634 if ((0 == strcmpic(csp->http->gpc, "trace"))
1635 || (0 == strcmpic(csp->http->gpc, "options")))
1637 for (p = csp->headers->first; (p != NULL) ; p = p->next)
1639 if (!strncmp("Max-Forwards:", p->str, 13)
1640 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
1642 /* FIXME: We could handle at least TRACE here,
1643 but that would require a verbatim copy of
1644 the request which we don't have anymore */
1646 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
1648 /* Get mem for response or fail*/
1649 if (NULL == (rsp = alloc_http_response()))
1651 return cgi_error_memory();
1654 if (NULL == (rsp->status = strdup("501 Not Implemented")))
1656 free_http_response(rsp);
1657 return cgi_error_memory();
1661 return(finish_http_response(rsp));