1 const char filters_rcs[] = "$Id: filters.c,v 2.0 2002/06/04 14:34:21 jongfoster Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/src/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 2.0 2002/06/04 14:34:21 jongfoster
42 * Moving source files to src/
44 * Revision 1.58 2002/04/24 02:11:17 oes
45 * Jon's multiple AF patch: url_actions now evaluates rules
48 * Revision 1.57 2002/04/08 20:38:34 swa
51 * Revision 1.56 2002/04/05 15:51:24 oes
52 * - bugfix: error-pages now get correct request protocol
53 * - fix for invalid HTML in trust info
55 * Revision 1.55 2002/04/02 16:13:51 oes
56 * Fix: No "Go there anyway" for SSL
58 * Revision 1.54 2002/04/02 14:55:56 oes
59 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
61 * Revision 1.53 2002/03/26 22:29:54 swa
62 * we have a new homepage!
64 * Revision 1.52 2002/03/24 16:35:57 jongfoster
67 * Revision 1.51 2002/03/24 15:23:33 jongfoster
70 * Revision 1.50 2002/03/24 13:25:43 swa
71 * name change related issues
73 * Revision 1.49 2002/03/16 20:29:14 oes
76 * Revision 1.48 2002/03/13 20:25:34 oes
77 * Better logging for content filters
79 * Revision 1.47 2002/03/13 00:30:52 jongfoster
81 * Added option of always sending redirect for imageblock,
82 * currently disabled with #if 0.
84 * Revision 1.46 2002/03/12 01:42:49 oes
85 * Introduced modular filters
87 * Revision 1.45 2002/03/08 16:47:50 oes
88 * Added choice beween GIF and PNG built-in images
90 * Revision 1.44 2002/03/07 03:49:31 oes
91 * - Fixed compiler warnings etc
92 * - Changed built-in images from GIF to PNG
93 * (with regard to Unisys patent issue)
94 * - Added a 4x4 pattern PNG which is less intrusive
95 * than the logo but also clearly marks the deleted banners
97 * Revision 1.43 2002/01/22 23:51:59 jongfoster
98 * Replacing strsav() with the safer string_append().
100 * Adding missing html_encode() to error message generators. Where encoded
101 * and unencoded versions of a string were provided, removing the unencoded
104 * Revision 1.42 2002/01/17 21:00:32 jongfoster
105 * Moving all our URL and URL pattern parsing code to urlmatch.c.
107 * Using a single, simple url_match(pattern,url) function - rather than
108 * the 3-line match routine which was repeated all over the place.
110 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
112 * Using parse_http_url() to parse URLs without faking a HTTP
113 * request line for parse_http_request().
115 * Revision 1.41 2001/11/13 00:14:07 jongfoster
116 * Fixing stupid bug now I've figured out what || means.
117 * (It always returns 0 or 1, not one of it's paramaters.)
119 * Revision 1.40 2001/10/26 17:37:55 oes
120 * - Re-enabled Netscape 200/404 bug workaround in block_url():
121 * - Removed OS/2 special case
122 * - Made block_url() independant from sed() having been run
123 * - Made trust_url independant from sed() having been run
124 * - Made is_imageurl independant from sed() having been run.
125 * It now checks User-Agent: and Accept: by itself.
128 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
129 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
130 * threads to call select() simultaneously. So, it's time to do a real, live,
131 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
132 * (native). Both versions will work, but using __OS2__ offers multi-threading.
134 * Revision 1.38 2001/10/23 21:32:33 jongfoster
135 * Adding error-checking to selected functions
137 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
138 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
139 * filters.c. Added a FIXME in front of the offending code. I'll gladly
140 * put in a better/more robust fix for all parties if one is presented...
141 * It seems that just returning 200 instead of 404 would pretty much fix
142 * it for everyone, but I don't know all the history of the problem.
144 * Revision 1.36 2001/10/10 16:44:16 oes
145 * Added match_portlist function
147 * Revision 1.35 2001/10/07 15:41:23 oes
148 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
150 * New function remove_chunked_transfer_coding that strips chunked
151 * transfer coding to plain and is called by pcrs_filter_response
152 * and gif_deanimate_response if neccessary
154 * Improved handling of zero-change re_filter runs
156 * pcrs_filter_response and gif_deanimate_response now remove
157 * chunked transfer codeing before processing the body.
159 * Revision 1.34 2001/09/20 15:49:36 steudten
161 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
162 * See cgi.c fill_template().
164 * Revision 1.33 2001/09/16 17:05:14 jongfoster
165 * Removing unused #include showarg.h
167 * Revision 1.32 2001/09/16 13:21:27 jongfoster
168 * Changes to use new list functions.
170 * Revision 1.31 2001/09/16 11:38:02 jongfoster
171 * Splitting fill_template() into 2 functions:
172 * template_load() loads the file
173 * template_fill() performs the PCRS regexps.
174 * This is because the CGI edit interface has a "table row"
175 * template which is used many times in the page - this
176 * change means it's only loaded from disk once.
178 * Revision 1.30 2001/09/16 11:00:10 jongfoster
179 * New function alloc_http_response, for symmetry with free_http_response
181 * Revision 1.29 2001/09/13 23:32:40 jongfoster
182 * Moving image data to cgi.c rather than cgi.h
183 * Fixing a GPF under Win32 (and any other OS that protects global
184 * constants from being written to).
186 * Revision 1.28 2001/09/10 10:18:51 oes
187 * Silenced compiler warnings
189 * Revision 1.27 2001/08/05 16:06:20 jongfoster
190 * Modifiying "struct map" so that there are now separate header and
191 * "map_entry" structures. This means that functions which modify a
192 * map no longer need to return a pointer to the modified map.
193 * Also, it no longer reverses the order of the entries (which may be
194 * important with some advanced template substitutions).
196 * Revision 1.26 2001/07/30 22:08:36 jongfoster
197 * Tidying up #defines:
198 * - All feature #defines are now of the form FEATURE_xxx
199 * - Permanently turned off WIN_GUI_EDIT
200 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
202 * Revision 1.25 2001/07/26 10:09:46 oes
203 * Made browser detection a little less naive
205 * Revision 1.24 2001/07/25 17:22:51 oes
206 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
208 * Revision 1.23 2001/07/23 13:40:12 oes
209 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
211 * Revision 1.22 2001/07/18 12:29:34 oes
212 * - Made gif_deanimate_response respect
213 * csp->action->string[ACTION_STRING_DEANIMATE]
214 * - Logging cosmetics
216 * Revision 1.21 2001/07/13 13:59:53 oes
217 * - Introduced gif_deanimate_response which shares the
218 * generic content modification interface of pcrs_filter_response
219 * and acts as a wrapper to deanimate.c:gif_deanimate()
220 * - Renamed re_process_buffer to pcrs_filter_response
221 * - pcrs_filter_response now returns NULL on failure
222 * - Removed all #ifdef PCRS
224 * Revision 1.20 2001/07/01 17:01:04 oes
225 * Added comments and missing return statement in is_untrusted_url()
227 * Revision 1.19 2001/06/29 21:45:41 oes
228 * Indentation, CRLF->LF, Tab-> Space
230 * Revision 1.18 2001/06/29 13:27:38 oes
231 * - Cleaned up, renamed and reorderd functions
232 * and improved comments
235 * - Ported to CGI platform. Now delivers
236 * http_response or NULL
237 * - Unified HTML and GIF generation (moved image detection
238 * and GIF generation here from jcc.c:chat())
239 * - Fixed HTTP status to:
240 * - 403 (Forbidden) for the "blocked" HTML message
241 * - 200 (OK) for GIF answers
242 * - 302 (Redirect) for redirect to GIF
245 * - Ported to CGI platform. Now delivers
246 * http_response or NULL
247 * - Separated detection of untrusted URL into
248 * (bool)is_untrusted_url
249 * - Added enforcement of untrusted requests
251 * - Moved redirect_url() from cgi.c to here
252 * and ported it to the CGI platform
254 * - Removed logentry from cancelled commit
256 * Revision 1.17 2001/06/09 10:55:28 jongfoster
257 * Changing BUFSIZ ==> BUFFER_SIZE
259 * Revision 1.16 2001/06/07 23:10:26 jongfoster
260 * Allowing unanchored domain patterns to back off and retry
261 * if they partially match. Optimized right-anchored patterns.
262 * Moving ACL and forward files into config file.
263 * Replacing struct gateway with struct forward_spec
265 * Revision 1.15 2001/06/03 19:12:00 oes
266 * extracted-CGI relevant stuff
268 * Revision 1.14 2001/06/01 10:30:55 oes
269 * Added optional left-anchoring to domaincmp
271 * Revision 1.13 2001/05/31 21:21:30 jongfoster
272 * Permissionsfile / actions file changes:
273 * - Changed "permission" to "action" throughout
274 * - changes to file format to allow string parameters
275 * - Moved helper functions to actions.c
277 * Revision 1.12 2001/05/31 17:35:20 oes
279 * - Enhanced domain part globbing with infix and prefix asterisk
280 * matching and optional unanchored operation
282 * Revision 1.11 2001/05/29 11:53:23 oes
283 * "See why" link added to "blocked" page
285 * Revision 1.10 2001/05/29 09:50:24 jongfoster
286 * Unified blocklist/imagelist/permissionslist.
287 * File format is still under discussion, but the internal changes
290 * Also modified interceptor behaviour:
291 * - We now intercept all URLs beginning with one of the following
292 * prefixes (and *only* these prefixes):
294 * * http://ijbswa.sf.net/config/
295 * * http://ijbswa.sourceforge.net/config/
296 * - New interceptors "home page" - go to http://i.j.b/ to see it.
297 * - Internal changes so that intercepted and fast redirect pages
298 * are not replaced with an image.
299 * - Interceptors now have the option to send a binary page direct
300 * to the client. (i.e. ijb-send-banner uses this)
301 * - Implemented show-url-info interceptor. (Which is why I needed
302 * the above interceptors changes - a typical URL is
303 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
304 * The previous mechanism would not have intercepted that, and
305 * if it had been intercepted then it then it would have replaced
308 * Revision 1.9 2001/05/27 22:17:04 oes
310 * - re_process_buffer no longer writes the modified buffer
311 * to the client, which was very ugly. It now returns the
312 * buffer, which it is then written by chat.
314 * - content_length now adjusts the Content-Length: header
315 * for modified documents rather than crunch()ing it.
316 * (Length info in csp->content_length, which is 0 for
317 * unmodified documents)
319 * - For this to work, sed() is called twice when filtering.
321 * Revision 1.8 2001/05/26 17:13:28 jongfoster
322 * Filled in a function comment.
324 * Revision 1.7 2001/05/26 15:26:15 jongfoster
325 * ACL feature now provides more security by immediately dropping
326 * connections from untrusted hosts.
328 * Revision 1.6 2001/05/26 00:28:36 jongfoster
329 * Automatic reloading of config file.
330 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
331 * Most of the global variables have been moved to a new
332 * struct configuration_spec, accessed through csp->config->globalname
333 * Most of the globals remaining are used by the Win32 GUI.
335 * Revision 1.5 2001/05/25 22:34:30 jongfoster
338 * Revision 1.4 2001/05/22 18:46:04 oes
340 * - Enabled filtering banners by size rather than URL
341 * by adding patterns that replace all standard banner
342 * sizes with the "Junkbuster" gif to the re_filterfile
344 * - Enabled filtering WebBugs by providing a pattern
345 * which kills all 1x1 images
347 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
348 * which is selected by the (nonstandard and therefore
349 * capital) letter 'U' in the option string.
350 * It causes the quantifiers to be ungreedy by default.
351 * Appending a ? turns back to greedy (!).
353 * - Added a new interceptor ijb-send-banner, which
354 * sends back the "Junkbuster" gif. Without imagelist or
355 * MSIE detection support, or if tinygif = 1, or the
356 * URL isn't recognized as an imageurl, a lame HTML
357 * explanation is sent instead.
359 * - Added new feature, which permits blocking remote
360 * script redirects and firing back a local redirect
362 * The feature is conditionally compiled, i.e. it
363 * can be disabled with --disable-fast-redirects,
364 * plus it must be activated by a "fast-redirects"
365 * line in the config file, has its own log level
366 * and of course wants to be displayed by show-proxy-args
367 * Note: Boy, all the #ifdefs in 1001 locations and
368 * all the fumbling with configure.in and acconfig.h
369 * were *way* more work than the feature itself :-(
371 * - Because a generic redirect template was needed for
372 * this, tinygif = 3 now uses the same.
374 * - Moved GIFs, and other static HTTP response templates
379 * - Removed some >400 CRs again (Jon, you really worked
382 * Revision 1.3 2001/05/20 16:44:47 jongfoster
383 * Removing last hardcoded Junkbusters.com URLs.
385 * Revision 1.2 2001/05/20 01:21:20 jongfoster
386 * Version 2.9.4 checkin.
387 * - Merged popupfile and cookiefile, and added control over PCRS
388 * filtering, in new "permissionsfile".
389 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
390 * file error you now get a message box (in the Win32 GUI) rather
391 * than the program exiting with no explanation.
392 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
394 * - Removed tabs from "config"
395 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
396 * - Bumped up version number.
398 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
399 * Initial import of version 2.9.3 source tree
402 *********************************************************************/
408 #include <sys/types.h>
417 #endif /* ndef __OS2__ */
418 #include <netinet/in.h>
420 #include <winsock2.h>
421 #endif /* ndef _WIN32 */
425 #endif /* def __OS2__ */
433 #include "jbsockets.h"
434 #include "miscutil.h"
438 #include "deanimate.h"
439 #include "urlmatch.h"
445 const char filters_h_rcs[] = FILTERS_H_VERSION;
447 /* Fix a problem with Solaris. There should be no effect on other
449 * Solaris's isspace() is a macro which uses it's argument directly
450 * as an array index. Therefore we need to make sure that high-bit
451 * characters generate +ve values, and ideally we also want to make
452 * the argument match the declared parameter type of "int".
454 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
458 /*********************************************************************
460 * Function : block_acl
462 * Description : Block this request?
463 * Decide yes or no based on ACL file.
466 * 1 : dst = The proxy or gateway address this is going to.
467 * Or NULL to check all possible targets.
468 * 2 : csp = Current client state (buffers, headers, etc...)
469 * Also includes the client IP address.
471 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
473 *********************************************************************/
474 int block_acl(struct access_control_addr *dst, struct client_state *csp)
476 struct access_control_list *acl = csp->config->acl;
478 /* if not using an access control list, then permit the connection */
484 /* search the list */
487 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
491 /* Just want to check if they have any access */
492 if (acl->action == ACL_PERMIT)
497 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
498 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
500 if (acl->action == ACL_PERMIT)
518 /*********************************************************************
520 * Function : acl_addr
522 * Description : Called from `load_config' to parse an ACL address.
525 * 1 : aspec = String specifying ACL address.
526 * 2 : aca = struct access_control_addr to fill in.
528 * Returns : 0 => Ok, everything else is an error.
530 *********************************************************************/
531 int acl_addr(char *aspec, struct access_control_addr *aca)
533 int i, masklength, port;
539 if ((p = strchr(aspec, '/')) != NULL)
543 if (ijb_isdigit(*p) == 0)
547 masklength = atoi(p);
550 if ((masklength < 0) || (masklength > 32))
555 if ((p = strchr(aspec, ':')) != NULL)
559 if (ijb_isdigit(*p) == 0)
568 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
570 if (aca->addr == INADDR_NONE)
575 /* build the netmask */
577 for (i=1; i <= masklength ; i++)
579 aca->mask |= (1 << (32 - i));
582 /* now mask off the host portion of the ip address
583 * (i.e. save on the network portion of the address).
585 aca->addr = aca->addr & aca->mask;
590 #endif /* def FEATURE_ACL */
593 /*********************************************************************
595 * Function : match_portlist
597 * Description : Check if a given number is covered by a comma
598 * separated list of numbers and ranges (a,b-c,d,..)
601 * 1 : portlist = String with list
602 * 2 : port = port to check
604 * Returns : 0 => no match
607 *********************************************************************/
608 int match_portlist(const char *portlist, int port)
610 char *min, *max, *next, *portlist_copy;
612 min = next = portlist_copy = strdup(portlist);
615 * Zero-terminate first item and remember offset for next
617 if (NULL != (next = strchr(portlist_copy, (int) ',')))
623 * Loop through all items, checking for match
627 if (NULL == (max = strchr(min, (int) '-')))
630 * No dash, check for equality
632 if (port == atoi(min))
641 * This is a range, so check if between min and max,
642 * or, if max was omitted, between min and 65K
645 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
659 * Zero-terminate next item and remember offset for n+1
661 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
673 /*********************************************************************
675 * Function : block_url
677 * Description : Called from `chat'. Check to see if we need to block this.
680 * 1 : csp = Current client state (buffers, headers, etc...)
682 * Returns : NULL => unblocked, else HTTP block response
684 *********************************************************************/
685 struct http_response *block_url(struct client_state *csp)
687 #ifdef FEATURE_IMAGE_BLOCKING
689 #endif /* def FEATURE_IMAGE_BLOCKING */
690 struct http_response *rsp;
693 * If it's not blocked, don't block it ;-)
695 if ((csp->action->flags & ACTION_BLOCK) == 0)
701 * Else, prepare a response
703 if (NULL == (rsp = alloc_http_response()))
705 return cgi_error_memory();
709 * If it's an image-url, send back an image or redirect
710 * as specified by the relevant +image action
712 #ifdef FEATURE_IMAGE_BLOCKING
713 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
716 /* determine HOW images should be blocked */
717 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
719 #if 1 /* Two alternative strategies, use this one for now: */
721 /* and handle accordingly: */
722 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
724 rsp->body = bindup(image_pattern_data, image_pattern_length);
725 if (rsp->body == NULL)
727 free_http_response(rsp);
728 return cgi_error_memory();
730 rsp->content_length = image_pattern_length;
732 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
734 free_http_response(rsp);
735 return cgi_error_memory();
739 else if (0 == strcmpic(p, "blank"))
741 rsp->body = bindup(image_blank_data, image_blank_length);
742 if (rsp->body == NULL)
744 free_http_response(rsp);
745 return cgi_error_memory();
747 rsp->content_length = image_blank_length;
749 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
751 free_http_response(rsp);
752 return cgi_error_memory();
758 rsp->status = strdup("302 Local Redirect from Privoxy");
759 if (rsp->status == NULL)
761 free_http_response(rsp);
762 return cgi_error_memory();
765 if (enlist_unique_header(rsp->headers, "Location", p))
767 free_http_response(rsp);
768 return cgi_error_memory();
772 #else /* Following code is disabled for now */
774 /* and handle accordingly: */
775 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
777 p = CGI_PREFIX "send-banner?type=pattern";
779 else if (0 == strcmpic(p, "blank"))
781 p = CGI_PREFIX "send-banner?type=blank";
783 rsp->status = strdup("302 Local Redirect from Privoxy");
784 if (rsp->status == NULL)
786 free_http_response(rsp);
787 return cgi_error_memory();
790 if (enlist_unique_header(rsp->headers, "Location", p))
792 free_http_response(rsp);
793 return cgi_error_memory();
795 #endif /* Preceeding code is disabled for now */
798 #endif /* def FEATURE_IMAGE_BLOCKING */
801 * Else, generate an HTML "blocked" message:
805 struct map * exports;
808 * Workaround for stupid Netscape bug which prevents
809 * pages from being displayed if loading a referenced
810 * JavaScript or style sheet fails. So make it appear
811 * as if it succeeded.
813 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
814 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
815 && !strstr(p, "Gecko") /* save Mozilla, */
816 && !strstr(p, "compatible") /* MSIE */
817 && !strstr(p, "Opera")) /* and Opera. */
819 rsp->status = strdup("200 Request for blocked URL");
823 rsp->status = strdup("404 Request for blocked URL");
826 if (rsp->status == NULL)
828 free_http_response(rsp);
829 return cgi_error_memory();
832 exports = default_exports(csp, NULL);
835 free_http_response(rsp);
836 return cgi_error_memory();
839 #ifdef FEATURE_FORCE_LOAD
840 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
841 if (csp->http->ssl != 0)
842 #endif /* ndef FEATURE_FORCE_LOAD */
844 err = map_block_killer(exports, "force-support");
847 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
848 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
849 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
854 free_http_response(rsp);
855 return cgi_error_memory();
858 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
861 free_http_response(rsp);
862 return cgi_error_memory();
866 return finish_http_response(rsp);
872 /*********************************************************************
874 * Function : trust_url FIXME: I should be called distrust_url
876 * Description : Calls is_untrusted_url to determine if the URL is trusted
877 * and if not, returns a HTTP 304 response with a reject message.
880 * 1 : csp = Current client state (buffers, headers, etc...)
882 * Returns : NULL => trusted, else http_response.
884 *********************************************************************/
885 struct http_response *trust_url(struct client_state *csp)
887 struct http_response *rsp;
888 struct map * exports;
889 char buf[BUFFER_SIZE];
891 struct url_spec **tl;
896 * Don't bother to work on trusted URLs
898 if (!is_untrusted_url(csp))
904 * Else, prepare a response:
906 if (NULL == (rsp = alloc_http_response()))
908 return cgi_error_memory();
911 exports = default_exports(csp, NULL);
914 free_http_response(rsp);
915 return cgi_error_memory();
919 * Export the protocol, host, port, and referrer information
921 err = map(exports, "hostport", 1, csp->http->hostport, 1);
922 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
923 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
925 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
927 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
931 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
937 free_http_response(rsp);
938 return cgi_error_memory();
942 * Export the trust list
945 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
947 sprintf(buf, "<li>%s</li>\n", t->spec);
948 string_append(&p, buf);
950 err = map(exports, "trusted-referrers", 1, p, 0);
955 free_http_response(rsp);
956 return cgi_error_memory();
960 * Export the trust info, if available
962 if (csp->config->trust_info->first)
964 struct list_entry *l;
967 for (l = csp->config->trust_info->first; l ; l = l->next)
969 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
970 string_append(&p, buf);
972 err = map(exports, "trust-info", 1, p, 0);
976 err = map_block_killer(exports, "have-trust-info");
982 free_http_response(rsp);
983 return cgi_error_memory();
987 * Export the force prefix or the force conditional block killer
989 #ifdef FEATURE_FORCE_LOAD
990 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
991 #else /* ifndef FEATURE_FORCE_LOAD */
992 err = map_block_killer(exports, "force-support");
993 #endif /* ndef FEATURE_FORCE_LOAD */
998 free_http_response(rsp);
999 return cgi_error_memory();
1003 * Build the response
1005 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1008 free_http_response(rsp);
1009 return cgi_error_memory();
1012 return finish_http_response(rsp);
1014 #endif /* def FEATURE_TRUST */
1017 #ifdef FEATURE_FAST_REDIRECTS
1018 /*********************************************************************
1020 * Function : redirect_url
1022 * Description : Checks for redirection URLs and returns a HTTP redirect
1023 * to the destination URL, if necessary
1026 * 1 : csp = Current client state (buffers, headers, etc...)
1028 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1030 *********************************************************************/
1031 struct http_response *redirect_url(struct client_state *csp)
1034 struct http_response *rsp;
1036 p = q = csp->http->path;
1037 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1040 * find the last URL encoded in the request
1042 while ((p = strstr(p, "http://")) != NULL)
1048 * if there was any, generate and return a HTTP redirect
1050 if (q != csp->http->path)
1052 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1054 if (NULL == (rsp = alloc_http_response()))
1056 return cgi_error_memory();
1059 if ( enlist_unique_header(rsp->headers, "Location", q)
1060 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1062 free_http_response(rsp);
1063 return cgi_error_memory();
1066 return finish_http_response(rsp);
1074 #endif /* def FEATURE_FAST_REDIRECTS */
1077 #ifdef FEATURE_IMAGE_BLOCKING
1078 /*********************************************************************
1080 * Function : is_imageurl
1082 * Description : Given a URL, decide whether it is an image or not,
1083 * using either the info from a previous +image action
1084 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1085 * the browser's accept header.
1088 * 1 : csp = Current client state (buffers, headers, etc...)
1090 * Returns : True (nonzero) if URL is an image, false (0)
1093 *********************************************************************/
1094 int is_imageurl(struct client_state *csp)
1096 #ifdef FEATURE_IMAGE_DETECT_MSIE
1099 tmp = get_header_value(csp->headers, "User-Agent:");
1100 if (tmp && strstr(tmp, "MSIE"))
1102 tmp = get_header_value(csp->headers, "Accept:");
1103 if (tmp && strstr(tmp, "image/gif"))
1105 /* Client will accept HTML. If this seems counterintuitive,
1115 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1117 return ((csp->action->flags & ACTION_IMAGE) != 0);
1120 #endif /* def FEATURE_IMAGE_BLOCKING */
1123 #ifdef FEATURE_TRUST
1124 /*********************************************************************
1126 * Function : is_untrusted_url
1128 * Description : Should we "distrust" this URL (and block it)?
1130 * Yes if it matches a line in the trustfile, or if the
1131 * referrer matches a line starting with "+" in the
1136 * 1 : csp = Current client state (buffers, headers, etc...)
1138 * Returns : 0 => trusted, 1 => untrusted
1140 *********************************************************************/
1141 int is_untrusted_url(struct client_state *csp)
1143 struct file_list *fl;
1144 struct block_spec *b;
1145 struct url_spec **trusted_url;
1146 struct http_request rhttp[1];
1147 const char * referer;
1151 * If we don't have a trustlist, we trust everybody
1153 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1158 memset(rhttp, '\0', sizeof(*rhttp));
1161 * Do we trust the request URL itself?
1163 for (b = b->next; b ; b = b->next)
1165 if (url_match(b->url, csp->http))
1171 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1173 /* no referrer was supplied */
1179 * If not, do we maybe trust its referrer?
1181 err = parse_http_url(referer, rhttp, csp);
1187 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1189 if (url_match(*trusted_url, rhttp))
1191 /* if the URL's referrer is from a trusted referrer, then
1192 * add the target spec to the trustfile as an unblocked
1193 * domain and return NULL (which means it's OK).
1198 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1202 char * new_entry = strdup("~");
1204 string_append(&new_entry, csp->http->hostport);
1206 path = csp->http->path;
1207 if ( (path[0] == '/')
1209 && ((path_end = strchr(path + 2, '/')) != NULL))
1211 /* since this path points into a user's home space
1212 * be sure to include this spec in the trustfile.
1214 int path_len = path_end - path; /* save offset */
1215 path = strdup(path); /* Copy string */
1218 path_end = path + path_len; /* regenerate ptr to new buffer */
1219 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1221 string_join(&new_entry, path);
1224 if (new_entry != NULL)
1226 fprintf(fp, "%s\n", new_entry);
1231 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1232 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1242 #endif /* def FEATURE_TRUST */
1245 /*********************************************************************
1247 * Function : pcrs_filter_response
1249 * Description : Ecexute all text substitutions from all applying
1250 * +filter actions on the text buffer that's been accumulated
1251 * in csp->iob->buf. If this changes the contents, set
1252 * csp->content_length to the modified size and raise the
1253 * CSP_FLAG_MODIFIED flag.
1256 * 1 : csp = Current client state (buffers, headers, etc...)
1258 * Returns : a pointer to the (newly allocated) modified buffer.
1259 * or NULL if there were no hits or something went wrong
1261 *********************************************************************/
1262 char *pcrs_filter_response(struct client_state *csp)
1267 char *old = csp->iob->cur, *new = NULL;
1270 struct file_list *fl;
1271 struct re_filterfile_spec *b;
1272 struct list_entry *filtername;
1277 if (csp->iob->cur >= csp->iob->eod)
1281 size = csp->iob->eod - csp->iob->cur;
1283 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1285 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1290 * If the body has a "chunked" transfer-encoding,
1291 * get rid of it first, adjusting size and iob->eod
1293 if (csp->flags & CSP_FLAG_CHUNKED)
1295 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1296 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1300 csp->iob->eod = csp->iob->cur + size;
1301 csp->flags |= CSP_FLAG_MODIFIED;
1305 * For all applying +filter actions, look if a filter by that
1306 * name exists and if yes, execute it's pcrs_joblist on the
1309 for (b = fl->f; b; b = b->next)
1311 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1312 filtername ; filtername = filtername->next)
1314 if (strcmp(b->name, filtername->str) == 0)
1316 int current_hits = 0;
1318 if ( NULL == b->joblist )
1320 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1324 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1325 csp->http->hostport, csp->http->path, size, b->name);
1327 /* Apply all jobs from the joblist */
1328 for (job = b->joblist; NULL != job; job = job->next)
1330 current_hits += pcrs_execute(job, old, size, &new, &size);
1331 if (old != csp->iob->cur) free(old);
1335 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1336 hits += current_hits;
1342 * If there were no hits, destroy our copy and let
1343 * chat() use the original in csp->iob
1351 csp->flags |= CSP_FLAG_MODIFIED;
1352 csp->content_length = size;
1360 /*********************************************************************
1362 * Function : gif_deanimate_response
1364 * Description : Deanimate the GIF image that has been accumulated in
1365 * csp->iob->buf, set csp->content_length to the modified
1366 * size and raise the CSP_FLAG_MODIFIED flag.
1369 * 1 : csp = Current client state (buffers, headers, etc...)
1371 * Returns : a pointer to the (newly allocated) modified buffer.
1372 * or NULL in case something went wrong.
1374 *********************************************************************/
1375 char *gif_deanimate_response(struct client_state *csp)
1377 struct binbuffer *in, *out;
1379 size_t size = csp->iob->eod - csp->iob->cur;
1382 * If the body has a "chunked" transfer-encoding,
1383 * get rid of it first, adjusting size and iob->eod
1385 if (csp->flags & CSP_FLAG_CHUNKED)
1387 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1388 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1392 csp->iob->eod = csp->iob->cur + size;
1393 csp->flags |= CSP_FLAG_MODIFIED;
1396 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1397 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1399 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1403 in->buffer = csp->iob->cur;
1406 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1408 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1415 if ((int)size == out->offset)
1417 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1421 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1423 csp->content_length = out->offset;
1424 csp->flags |= CSP_FLAG_MODIFIED;
1434 /*********************************************************************
1436 * Function : remove_chunked_transfer_coding
1438 * Description : In-situ remove the "chunked" transfer coding as defined
1439 * in rfc2616 from a buffer.
1442 * 1 : buffer = Pointer to the text buffer
1443 * 2 : size = Number of bytes to be processed
1445 * Returns : The new size, i.e. the number of bytes from buffer which
1446 * are occupied by the stripped body, or 0 in case something
1449 *********************************************************************/
1450 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1453 unsigned int chunksize = 0;
1454 char *from_p, *to_p;
1457 from_p = to_p = buffer;
1459 if (sscanf(buffer, "%x", &chunksize) != 1)
1461 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1465 while (chunksize > 0)
1467 if (NULL == (from_p = strstr(from_p, "\r\n")))
1469 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1472 newsize += chunksize;
1475 memmove(to_p, from_p, (size_t) chunksize);
1476 to_p = buffer + newsize;
1477 from_p += chunksize + 2;
1479 if (sscanf(from_p, "%x", &chunksize) != 1)
1481 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1486 /* FIXME: Should this get its own loglevel? */
1487 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1493 /*********************************************************************
1495 * Function : url_actions
1497 * Description : Gets the actions for this URL.
1500 * 1 : http = http_request request for blocked URLs
1501 * 2 : csp = Current client state (buffers, headers, etc...)
1505 *********************************************************************/
1506 void url_actions(struct http_request *http,
1507 struct client_state *csp)
1509 struct file_list *fl;
1510 struct url_actions *b;
1513 init_current_action(csp->action);
1515 for (i = 0; i < MAX_ACTION_FILES; i++)
1517 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1522 apply_url_actions(csp->action, http, b);
1529 /*********************************************************************
1531 * Function : apply_url_actions
1533 * Description : Applies a list of URL actions.
1536 * 1 : action = Destination.
1537 * 2 : http = Current URL
1538 * 3 : b = list of URL actions to apply
1542 *********************************************************************/
1543 void apply_url_actions(struct current_action_spec *action,
1544 struct http_request *http,
1545 struct url_actions *b)
1549 /* Should never happen */
1553 for (b = b->next; NULL != b; b = b->next)
1555 if (url_match(b->url, http))
1557 merge_current_action(action, b->action);
1563 /*********************************************************************
1565 * Function : forward_url
1567 * Description : Should we forward this to another proxy?
1570 * 1 : http = http_request request for current URL
1571 * 2 : csp = Current client state (buffers, headers, etc...)
1573 * Returns : Pointer to forwarding information.
1575 *********************************************************************/
1576 const struct forward_spec * forward_url(struct http_request *http,
1577 struct client_state *csp)
1579 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1580 struct forward_spec *fwd = csp->config->forward;
1589 if (url_match(fwd->url, http))