1 const char filters_rcs[] = "$Id: filters.c,v 1.57 2002/04/08 20:38:34 swa 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', 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 1.57 2002/04/08 20:38:34 swa
44 * Revision 1.56 2002/04/05 15:51:24 oes
45 * - bugfix: error-pages now get correct request protocol
46 * - fix for invalid HTML in trust info
48 * Revision 1.55 2002/04/02 16:13:51 oes
49 * Fix: No "Go there anyway" for SSL
51 * Revision 1.54 2002/04/02 14:55:56 oes
52 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
54 * Revision 1.53 2002/03/26 22:29:54 swa
55 * we have a new homepage!
57 * Revision 1.52 2002/03/24 16:35:57 jongfoster
60 * Revision 1.51 2002/03/24 15:23:33 jongfoster
63 * Revision 1.50 2002/03/24 13:25:43 swa
64 * name change related issues
66 * Revision 1.49 2002/03/16 20:29:14 oes
69 * Revision 1.48 2002/03/13 20:25:34 oes
70 * Better logging for content filters
72 * Revision 1.47 2002/03/13 00:30:52 jongfoster
74 * Added option of always sending redirect for imageblock,
75 * currently disabled with #if 0.
77 * Revision 1.46 2002/03/12 01:42:49 oes
78 * Introduced modular filters
80 * Revision 1.45 2002/03/08 16:47:50 oes
81 * Added choice beween GIF and PNG built-in images
83 * Revision 1.44 2002/03/07 03:49:31 oes
84 * - Fixed compiler warnings etc
85 * - Changed built-in images from GIF to PNG
86 * (with regard to Unisys patent issue)
87 * - Added a 4x4 pattern PNG which is less intrusive
88 * than the logo but also clearly marks the deleted banners
90 * Revision 1.43 2002/01/22 23:51:59 jongfoster
91 * Replacing strsav() with the safer string_append().
93 * Adding missing html_encode() to error message generators. Where encoded
94 * and unencoded versions of a string were provided, removing the unencoded
97 * Revision 1.42 2002/01/17 21:00:32 jongfoster
98 * Moving all our URL and URL pattern parsing code to urlmatch.c.
100 * Using a single, simple url_match(pattern,url) function - rather than
101 * the 3-line match routine which was repeated all over the place.
103 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
105 * Using parse_http_url() to parse URLs without faking a HTTP
106 * request line for parse_http_request().
108 * Revision 1.41 2001/11/13 00:14:07 jongfoster
109 * Fixing stupid bug now I've figured out what || means.
110 * (It always returns 0 or 1, not one of it's paramaters.)
112 * Revision 1.40 2001/10/26 17:37:55 oes
113 * - Re-enabled Netscape 200/404 bug workaround in block_url():
114 * - Removed OS/2 special case
115 * - Made block_url() independant from sed() having been run
116 * - Made trust_url independant from sed() having been run
117 * - Made is_imageurl independant from sed() having been run.
118 * It now checks User-Agent: and Accept: by itself.
121 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
122 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
123 * threads to call select() simultaneously. So, it's time to do a real, live,
124 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
125 * (native). Both versions will work, but using __OS2__ offers multi-threading.
127 * Revision 1.38 2001/10/23 21:32:33 jongfoster
128 * Adding error-checking to selected functions
130 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
131 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
132 * filters.c. Added a FIXME in front of the offending code. I'll gladly
133 * put in a better/more robust fix for all parties if one is presented...
134 * It seems that just returning 200 instead of 404 would pretty much fix
135 * it for everyone, but I don't know all the history of the problem.
137 * Revision 1.36 2001/10/10 16:44:16 oes
138 * Added match_portlist function
140 * Revision 1.35 2001/10/07 15:41:23 oes
141 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
143 * New function remove_chunked_transfer_coding that strips chunked
144 * transfer coding to plain and is called by pcrs_filter_response
145 * and gif_deanimate_response if neccessary
147 * Improved handling of zero-change re_filter runs
149 * pcrs_filter_response and gif_deanimate_response now remove
150 * chunked transfer codeing before processing the body.
152 * Revision 1.34 2001/09/20 15:49:36 steudten
154 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
155 * See cgi.c fill_template().
157 * Revision 1.33 2001/09/16 17:05:14 jongfoster
158 * Removing unused #include showarg.h
160 * Revision 1.32 2001/09/16 13:21:27 jongfoster
161 * Changes to use new list functions.
163 * Revision 1.31 2001/09/16 11:38:02 jongfoster
164 * Splitting fill_template() into 2 functions:
165 * template_load() loads the file
166 * template_fill() performs the PCRS regexps.
167 * This is because the CGI edit interface has a "table row"
168 * template which is used many times in the page - this
169 * change means it's only loaded from disk once.
171 * Revision 1.30 2001/09/16 11:00:10 jongfoster
172 * New function alloc_http_response, for symmetry with free_http_response
174 * Revision 1.29 2001/09/13 23:32:40 jongfoster
175 * Moving image data to cgi.c rather than cgi.h
176 * Fixing a GPF under Win32 (and any other OS that protects global
177 * constants from being written to).
179 * Revision 1.28 2001/09/10 10:18:51 oes
180 * Silenced compiler warnings
182 * Revision 1.27 2001/08/05 16:06:20 jongfoster
183 * Modifiying "struct map" so that there are now separate header and
184 * "map_entry" structures. This means that functions which modify a
185 * map no longer need to return a pointer to the modified map.
186 * Also, it no longer reverses the order of the entries (which may be
187 * important with some advanced template substitutions).
189 * Revision 1.26 2001/07/30 22:08:36 jongfoster
190 * Tidying up #defines:
191 * - All feature #defines are now of the form FEATURE_xxx
192 * - Permanently turned off WIN_GUI_EDIT
193 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
195 * Revision 1.25 2001/07/26 10:09:46 oes
196 * Made browser detection a little less naive
198 * Revision 1.24 2001/07/25 17:22:51 oes
199 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
201 * Revision 1.23 2001/07/23 13:40:12 oes
202 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
204 * Revision 1.22 2001/07/18 12:29:34 oes
205 * - Made gif_deanimate_response respect
206 * csp->action->string[ACTION_STRING_DEANIMATE]
207 * - Logging cosmetics
209 * Revision 1.21 2001/07/13 13:59:53 oes
210 * - Introduced gif_deanimate_response which shares the
211 * generic content modification interface of pcrs_filter_response
212 * and acts as a wrapper to deanimate.c:gif_deanimate()
213 * - Renamed re_process_buffer to pcrs_filter_response
214 * - pcrs_filter_response now returns NULL on failiure
215 * - Removed all #ifdef PCRS
217 * Revision 1.20 2001/07/01 17:01:04 oes
218 * Added comments and missing return statement in is_untrusted_url()
220 * Revision 1.19 2001/06/29 21:45:41 oes
221 * Indentation, CRLF->LF, Tab-> Space
223 * Revision 1.18 2001/06/29 13:27:38 oes
224 * - Cleaned up, renamed and reorderd functions
225 * and improved comments
228 * - Ported to CGI platform. Now delivers
229 * http_response or NULL
230 * - Unified HTML and GIF generation (moved image detection
231 * and GIF generation here from jcc.c:chat())
232 * - Fixed HTTP status to:
233 * - 403 (Forbidden) for the "blocked" HTML message
234 * - 200 (OK) for GIF answers
235 * - 302 (Redirect) for redirect to GIF
238 * - Ported to CGI platform. Now delivers
239 * http_response or NULL
240 * - Separated detection of untrusted URL into
241 * (bool)is_untrusted_url
242 * - Added enforcement of untrusted requests
244 * - Moved redirect_url() from cgi.c to here
245 * and ported it to the CGI platform
247 * - Removed logentry from cancelled commit
249 * Revision 1.17 2001/06/09 10:55:28 jongfoster
250 * Changing BUFSIZ ==> BUFFER_SIZE
252 * Revision 1.16 2001/06/07 23:10:26 jongfoster
253 * Allowing unanchored domain patterns to back off and retry
254 * if they partially match. Optimized right-anchored patterns.
255 * Moving ACL and forward files into config file.
256 * Replacing struct gateway with struct forward_spec
258 * Revision 1.15 2001/06/03 19:12:00 oes
259 * extracted-CGI relevant stuff
261 * Revision 1.14 2001/06/01 10:30:55 oes
262 * Added optional left-anchoring to domaincmp
264 * Revision 1.13 2001/05/31 21:21:30 jongfoster
265 * Permissionsfile / actions file changes:
266 * - Changed "permission" to "action" throughout
267 * - changes to file format to allow string parameters
268 * - Moved helper functions to actions.c
270 * Revision 1.12 2001/05/31 17:35:20 oes
272 * - Enhanced domain part globbing with infix and prefix asterisk
273 * matching and optional unanchored operation
275 * Revision 1.11 2001/05/29 11:53:23 oes
276 * "See why" link added to "blocked" page
278 * Revision 1.10 2001/05/29 09:50:24 jongfoster
279 * Unified blocklist/imagelist/permissionslist.
280 * File format is still under discussion, but the internal changes
283 * Also modified interceptor behaviour:
284 * - We now intercept all URLs beginning with one of the following
285 * prefixes (and *only* these prefixes):
287 * * http://ijbswa.sf.net/config/
288 * * http://ijbswa.sourceforge.net/config/
289 * - New interceptors "home page" - go to http://i.j.b/ to see it.
290 * - Internal changes so that intercepted and fast redirect pages
291 * are not replaced with an image.
292 * - Interceptors now have the option to send a binary page direct
293 * to the client. (i.e. ijb-send-banner uses this)
294 * - Implemented show-url-info interceptor. (Which is why I needed
295 * the above interceptors changes - a typical URL is
296 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
297 * The previous mechanism would not have intercepted that, and
298 * if it had been intercepted then it then it would have replaced
301 * Revision 1.9 2001/05/27 22:17:04 oes
303 * - re_process_buffer no longer writes the modified buffer
304 * to the client, which was very ugly. It now returns the
305 * buffer, which it is then written by chat.
307 * - content_length now adjusts the Content-Length: header
308 * for modified documents rather than crunch()ing it.
309 * (Length info in csp->content_length, which is 0 for
310 * unmodified documents)
312 * - For this to work, sed() is called twice when filtering.
314 * Revision 1.8 2001/05/26 17:13:28 jongfoster
315 * Filled in a function comment.
317 * Revision 1.7 2001/05/26 15:26:15 jongfoster
318 * ACL feature now provides more security by immediately dropping
319 * connections from untrusted hosts.
321 * Revision 1.6 2001/05/26 00:28:36 jongfoster
322 * Automatic reloading of config file.
323 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
324 * Most of the global variables have been moved to a new
325 * struct configuration_spec, accessed through csp->config->globalname
326 * Most of the globals remaining are used by the Win32 GUI.
328 * Revision 1.5 2001/05/25 22:34:30 jongfoster
331 * Revision 1.4 2001/05/22 18:46:04 oes
333 * - Enabled filtering banners by size rather than URL
334 * by adding patterns that replace all standard banner
335 * sizes with the "Junkbuster" gif to the re_filterfile
337 * - Enabled filtering WebBugs by providing a pattern
338 * which kills all 1x1 images
340 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
341 * which is selected by the (nonstandard and therefore
342 * capital) letter 'U' in the option string.
343 * It causes the quantifiers to be ungreedy by default.
344 * Appending a ? turns back to greedy (!).
346 * - Added a new interceptor ijb-send-banner, which
347 * sends back the "Junkbuster" gif. Without imagelist or
348 * MSIE detection support, or if tinygif = 1, or the
349 * URL isn't recognized as an imageurl, a lame HTML
350 * explanation is sent instead.
352 * - Added new feature, which permits blocking remote
353 * script redirects and firing back a local redirect
355 * The feature is conditionally compiled, i.e. it
356 * can be disabled with --disable-fast-redirects,
357 * plus it must be activated by a "fast-redirects"
358 * line in the config file, has its own log level
359 * and of course wants to be displayed by show-proxy-args
360 * Note: Boy, all the #ifdefs in 1001 locations and
361 * all the fumbling with configure.in and acconfig.h
362 * were *way* more work than the feature itself :-(
364 * - Because a generic redirect template was needed for
365 * this, tinygif = 3 now uses the same.
367 * - Moved GIFs, and other static HTTP response templates
372 * - Removed some >400 CRs again (Jon, you really worked
375 * Revision 1.3 2001/05/20 16:44:47 jongfoster
376 * Removing last hardcoded Junkbusters.com URLs.
378 * Revision 1.2 2001/05/20 01:21:20 jongfoster
379 * Version 2.9.4 checkin.
380 * - Merged popupfile and cookiefile, and added control over PCRS
381 * filtering, in new "permissionsfile".
382 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
383 * file error you now get a message box (in the Win32 GUI) rather
384 * than the program exiting with no explanation.
385 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
387 * - Removed tabs from "config"
388 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
389 * - Bumped up version number.
391 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
392 * Initial import of version 2.9.3 source tree
395 *********************************************************************/
401 #include <sys/types.h>
410 #endif /* ndef __OS2__ */
411 #include <netinet/in.h>
413 #include <winsock2.h>
414 #endif /* ndef _WIN32 */
418 #endif /* def __OS2__ */
426 #include "jbsockets.h"
427 #include "miscutil.h"
431 #include "deanimate.h"
432 #include "urlmatch.h"
438 const char filters_h_rcs[] = FILTERS_H_VERSION;
440 /* Fix a problem with Solaris. There should be no effect on other
442 * Solaris's isspace() is a macro which uses it's argument directly
443 * as an array index. Therefore we need to make sure that high-bit
444 * characters generate +ve values, and ideally we also want to make
445 * the argument match the declared parameter type of "int".
447 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
451 /*********************************************************************
453 * Function : block_acl
455 * Description : Block this request?
456 * Decide yes or no based on ACL file.
459 * 1 : dst = The proxy or gateway address this is going to.
460 * Or NULL to check all possible targets.
461 * 2 : csp = Current client state (buffers, headers, etc...)
462 * Also includes the client IP address.
464 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
466 *********************************************************************/
467 int block_acl(struct access_control_addr *dst, struct client_state *csp)
469 struct access_control_list *acl = csp->config->acl;
471 /* if not using an access control list, then permit the connection */
477 /* search the list */
480 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
484 /* Just want to check if they have any access */
485 if (acl->action == ACL_PERMIT)
490 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
491 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
493 if (acl->action == ACL_PERMIT)
511 /*********************************************************************
513 * Function : acl_addr
515 * Description : Called from `load_config' to parse an ACL address.
518 * 1 : aspec = String specifying ACL address.
519 * 2 : aca = struct access_control_addr to fill in.
521 * Returns : 0 => Ok, everything else is an error.
523 *********************************************************************/
524 int acl_addr(char *aspec, struct access_control_addr *aca)
526 int i, masklength, port;
532 if ((p = strchr(aspec, '/')) != NULL)
536 if (ijb_isdigit(*p) == 0)
540 masklength = atoi(p);
543 if ((masklength < 0) || (masklength > 32))
548 if ((p = strchr(aspec, ':')) != NULL)
552 if (ijb_isdigit(*p) == 0)
561 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
563 if (aca->addr == INADDR_NONE)
568 /* build the netmask */
570 for (i=1; i <= masklength ; i++)
572 aca->mask |= (1 << (32 - i));
575 /* now mask off the host portion of the ip address
576 * (i.e. save on the network portion of the address).
578 aca->addr = aca->addr & aca->mask;
583 #endif /* def FEATURE_ACL */
586 /*********************************************************************
588 * Function : match_portlist
590 * Description : Check if a given number is covered by a comma
591 * separated list of numbers and ranges (a,b-c,d,..)
594 * 1 : portlist = String with list
595 * 2 : port = port to check
597 * Returns : 0 => no match
600 *********************************************************************/
601 int match_portlist(const char *portlist, int port)
603 char *min, *max, *next, *portlist_copy;
605 min = next = portlist_copy = strdup(portlist);
608 * Zero-terminate first item and remember offset for next
610 if (NULL != (next = strchr(portlist_copy, (int) ',')))
616 * Loop through all items, checking for match
620 if (NULL == (max = strchr(min, (int) '-')))
623 * No dash, check for equality
625 if (port == atoi(min))
634 * This is a range, so check if between min and max,
635 * or, if max was omitted, between min and 65K
638 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
652 * Zero-terminate next item and remember offset for n+1
654 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
666 /*********************************************************************
668 * Function : block_url
670 * Description : Called from `chat'. Check to see if we need to block this.
673 * 1 : csp = Current client state (buffers, headers, etc...)
675 * Returns : NULL => unblocked, else HTTP block response
677 *********************************************************************/
678 struct http_response *block_url(struct client_state *csp)
680 #ifdef FEATURE_IMAGE_BLOCKING
682 #endif /* def FEATURE_IMAGE_BLOCKING */
683 struct http_response *rsp;
686 * If it's not blocked, don't block it ;-)
688 if ((csp->action->flags & ACTION_BLOCK) == 0)
694 * Else, prepare a response
696 if (NULL == (rsp = alloc_http_response()))
698 return cgi_error_memory();
702 * If it's an image-url, send back an image or redirect
703 * as specified by the relevant +image action
705 #ifdef FEATURE_IMAGE_BLOCKING
706 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
709 /* determine HOW images should be blocked */
710 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
712 #if 1 /* Two alternative strategies, use this one for now: */
714 /* and handle accordingly: */
715 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
717 rsp->body = bindup(image_pattern_data, image_pattern_length);
718 if (rsp->body == NULL)
720 free_http_response(rsp);
721 return cgi_error_memory();
723 rsp->content_length = image_pattern_length;
725 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
727 free_http_response(rsp);
728 return cgi_error_memory();
732 else if (0 == strcmpic(p, "blank"))
734 rsp->body = bindup(image_blank_data, image_blank_length);
735 if (rsp->body == NULL)
737 free_http_response(rsp);
738 return cgi_error_memory();
740 rsp->content_length = image_blank_length;
742 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
744 free_http_response(rsp);
745 return cgi_error_memory();
751 rsp->status = strdup("302 Local Redirect from Privoxy");
752 if (rsp->status == NULL)
754 free_http_response(rsp);
755 return cgi_error_memory();
758 if (enlist_unique_header(rsp->headers, "Location", p))
760 free_http_response(rsp);
761 return cgi_error_memory();
765 #else /* Following code is disabled for now */
767 /* and handle accordingly: */
768 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
770 p = CGI_PREFIX "send-banner?type=pattern";
772 else if (0 == strcmpic(p, "blank"))
774 p = CGI_PREFIX "send-banner?type=blank";
776 rsp->status = strdup("302 Local Redirect from Privoxy");
777 if (rsp->status == NULL)
779 free_http_response(rsp);
780 return cgi_error_memory();
783 if (enlist_unique_header(rsp->headers, "Location", p))
785 free_http_response(rsp);
786 return cgi_error_memory();
788 #endif /* Preceeding code is disabled for now */
791 #endif /* def FEATURE_IMAGE_BLOCKING */
794 * Else, generate an HTML "blocked" message:
798 struct map * exports;
801 * Workaround for stupid Netscape bug which prevents
802 * pages from being displayed if loading a referenced
803 * JavaScript or style sheet fails. So make it appear
804 * as if it succeeded.
806 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
807 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
808 && !strstr(p, "Gecko") /* save Mozilla, */
809 && !strstr(p, "compatible") /* MSIE */
810 && !strstr(p, "Opera")) /* and Opera. */
812 rsp->status = strdup("200 Request for blocked URL");
816 rsp->status = strdup("404 Request for blocked URL");
819 if (rsp->status == NULL)
821 free_http_response(rsp);
822 return cgi_error_memory();
825 exports = default_exports(csp, NULL);
828 free_http_response(rsp);
829 return cgi_error_memory();
832 #ifdef FEATURE_FORCE_LOAD
833 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
834 if (csp->http->ssl != 0)
835 #endif /* ndef FEATURE_FORCE_LOAD */
837 err = map_block_killer(exports, "force-support");
840 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
841 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
842 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
847 free_http_response(rsp);
848 return cgi_error_memory();
851 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
854 free_http_response(rsp);
855 return cgi_error_memory();
859 return finish_http_response(rsp);
865 /*********************************************************************
867 * Function : trust_url FIXME: I should be called distrust_url
869 * Description : Calls is_untrusted_url to determine if the URL is trusted
870 * and if not, returns a HTTP 304 response with a reject message.
873 * 1 : csp = Current client state (buffers, headers, etc...)
875 * Returns : NULL => trusted, else http_response.
877 *********************************************************************/
878 struct http_response *trust_url(struct client_state *csp)
880 struct http_response *rsp;
881 struct map * exports;
882 char buf[BUFFER_SIZE];
884 struct url_spec **tl;
889 * Don't bother to work on trusted URLs
891 if (!is_untrusted_url(csp))
897 * Else, prepare a response:
899 if (NULL == (rsp = alloc_http_response()))
901 return cgi_error_memory();
904 exports = default_exports(csp, NULL);
907 free_http_response(rsp);
908 return cgi_error_memory();
912 * Export the protocol, host, port, and referrer information
914 err = map(exports, "hostport", 1, csp->http->hostport, 1);
915 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
916 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
918 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
920 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
924 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
930 free_http_response(rsp);
931 return cgi_error_memory();
935 * Export the trust list
938 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
940 sprintf(buf, "<li>%s</li>\n", t->spec);
941 string_append(&p, buf);
943 err = map(exports, "trusted-referrers", 1, p, 0);
948 free_http_response(rsp);
949 return cgi_error_memory();
953 * Export the trust info, if available
955 if (csp->config->trust_info->first)
957 struct list_entry *l;
960 for (l = csp->config->trust_info->first; l ; l = l->next)
962 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
963 string_append(&p, buf);
965 err = map(exports, "trust-info", 1, p, 0);
969 err = map_block_killer(exports, "have-trust-info");
975 free_http_response(rsp);
976 return cgi_error_memory();
980 * Export the force prefix or the force conditional block killer
982 #ifdef FEATURE_FORCE_LOAD
983 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
984 #else /* ifndef FEATURE_FORCE_LOAD */
985 err = map_block_killer(exports, "force-support");
986 #endif /* ndef FEATURE_FORCE_LOAD */
991 free_http_response(rsp);
992 return cgi_error_memory();
998 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1001 free_http_response(rsp);
1002 return cgi_error_memory();
1005 return finish_http_response(rsp);
1007 #endif /* def FEATURE_TRUST */
1010 #ifdef FEATURE_FAST_REDIRECTS
1011 /*********************************************************************
1013 * Function : redirect_url
1015 * Description : Checks for redirection URLs and returns a HTTP redirect
1016 * to the destination URL, if necessary
1019 * 1 : csp = Current client state (buffers, headers, etc...)
1021 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1023 *********************************************************************/
1024 struct http_response *redirect_url(struct client_state *csp)
1027 struct http_response *rsp;
1029 p = q = csp->http->path;
1030 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1033 * find the last URL encoded in the request
1035 while ((p = strstr(p, "http://")) != NULL)
1041 * if there was any, generate and return a HTTP redirect
1043 if (q != csp->http->path)
1045 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1047 if (NULL == (rsp = alloc_http_response()))
1049 return cgi_error_memory();
1052 if ( enlist_unique_header(rsp->headers, "Location", q)
1053 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1055 free_http_response(rsp);
1056 return cgi_error_memory();
1059 return finish_http_response(rsp);
1067 #endif /* def FEATURE_FAST_REDIRECTS */
1070 #ifdef FEATURE_IMAGE_BLOCKING
1071 /*********************************************************************
1073 * Function : is_imageurl
1075 * Description : Given a URL, decide whether it is an image or not,
1076 * using either the info from a previous +image action
1077 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1078 * the browser's accept header.
1081 * 1 : csp = Current client state (buffers, headers, etc...)
1083 * Returns : True (nonzero) if URL is an image, false (0)
1086 *********************************************************************/
1087 int is_imageurl(struct client_state *csp)
1089 #ifdef FEATURE_IMAGE_DETECT_MSIE
1092 tmp = get_header_value(csp->headers, "User-Agent:");
1093 if (tmp && strstr(tmp, "MSIE"))
1095 tmp = get_header_value(csp->headers, "Accept:");
1096 if (tmp && strstr(tmp, "image/gif"))
1098 /* Client will accept HTML. If this seems counterintuitive,
1108 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1110 return ((csp->action->flags & ACTION_IMAGE) != 0);
1113 #endif /* def FEATURE_IMAGE_BLOCKING */
1116 #ifdef FEATURE_TRUST
1117 /*********************************************************************
1119 * Function : is_untrusted_url
1121 * Description : Should we "distrust" this URL (and block it)?
1123 * Yes if it matches a line in the trustfile, or if the
1124 * referrer matches a line starting with "+" in the
1129 * 1 : csp = Current client state (buffers, headers, etc...)
1131 * Returns : 0 => trusted, 1 => untrusted
1133 *********************************************************************/
1134 int is_untrusted_url(struct client_state *csp)
1136 struct file_list *fl;
1137 struct block_spec *b;
1138 struct url_spec **trusted_url;
1139 struct http_request rhttp[1];
1140 const char * referer;
1144 * If we don't have a trustlist, we trust everybody
1146 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1151 memset(rhttp, '\0', sizeof(*rhttp));
1154 * Do we trust the request URL itself?
1156 for (b = b->next; b ; b = b->next)
1158 if (url_match(b->url, csp->http))
1164 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1166 /* no referrer was supplied */
1172 * If not, do we maybe trust its referrer?
1174 err = parse_http_url(referer, rhttp, csp);
1180 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1182 if (url_match(*trusted_url, rhttp))
1184 /* if the URL's referrer is from a trusted referrer, then
1185 * add the target spec to the trustfile as an unblocked
1186 * domain and return NULL (which means it's OK).
1191 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1195 char * new_entry = strdup("~");
1197 string_append(&new_entry, csp->http->hostport);
1199 path = csp->http->path;
1200 if ( (path[0] == '/')
1202 && ((path_end = strchr(path + 2, '/')) != NULL))
1204 /* since this path points into a user's home space
1205 * be sure to include this spec in the trustfile.
1207 int path_len = path_end - path; /* save offset */
1208 path = strdup(path); /* Copy string */
1211 path_end = path + path_len; /* regenerate ptr to new buffer */
1212 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1214 string_join(&new_entry, path);
1217 if (new_entry != NULL)
1219 fprintf(fp, "%s\n", new_entry);
1224 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1225 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1235 #endif /* def FEATURE_TRUST */
1238 /*********************************************************************
1240 * Function : pcrs_filter_response
1242 * Description : Ecexute all text substitutions from all applying
1243 * +filter actions on the text buffer that's been accumulated
1244 * in csp->iob->buf. If this changes the contents, set
1245 * csp->content_length to the modified size and raise the
1246 * CSP_FLAG_MODIFIED flag.
1249 * 1 : csp = Current client state (buffers, headers, etc...)
1251 * Returns : a pointer to the (newly allocated) modified buffer.
1252 * or NULL if there were no hits or something went wrong
1254 *********************************************************************/
1255 char *pcrs_filter_response(struct client_state *csp)
1260 char *old = csp->iob->cur, *new = NULL;
1263 struct file_list *fl;
1264 struct re_filterfile_spec *b;
1265 struct list_entry *filtername;
1270 if (csp->iob->cur >= csp->iob->eod)
1274 size = csp->iob->eod - csp->iob->cur;
1276 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1278 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1283 * If the body has a "chunked" transfer-encoding,
1284 * get rid of it first, adjusting size and iob->eod
1286 if (csp->flags & CSP_FLAG_CHUNKED)
1288 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1289 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1293 csp->iob->eod = csp->iob->cur + size;
1294 csp->flags |= CSP_FLAG_MODIFIED;
1298 * For all applying +filter actions, look if a filter by that
1299 * name exists and if yes, execute it's pcrs_joblist on the
1302 for (b = fl->f; b; b = b->next)
1304 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1305 filtername ; filtername = filtername->next)
1307 if (strcmp(b->name, filtername->str) == 0)
1309 int current_hits = 0;
1311 if ( NULL == b->joblist )
1313 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1317 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1318 csp->http->hostport, csp->http->path, size, b->name);
1320 /* Apply all jobs from the joblist */
1321 for (job = b->joblist; NULL != job; job = job->next)
1323 current_hits += pcrs_execute(job, old, size, &new, &size);
1324 if (old != csp->iob->cur) free(old);
1328 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1329 hits += current_hits;
1335 * If there were no hits, destroy our copy and let
1336 * chat() use the original in csp->iob
1344 csp->flags |= CSP_FLAG_MODIFIED;
1345 csp->content_length = size;
1353 /*********************************************************************
1355 * Function : gif_deanimate_response
1357 * Description : Deanimate the GIF image that has been accumulated in
1358 * csp->iob->buf, set csp->content_length to the modified
1359 * size and raise the CSP_FLAG_MODIFIED flag.
1362 * 1 : csp = Current client state (buffers, headers, etc...)
1364 * Returns : a pointer to the (newly allocated) modified buffer.
1365 * or NULL in case something went wrong.
1367 *********************************************************************/
1368 char *gif_deanimate_response(struct client_state *csp)
1370 struct binbuffer *in, *out;
1372 size_t size = csp->iob->eod - csp->iob->cur;
1375 * If the body has a "chunked" transfer-encoding,
1376 * get rid of it first, adjusting size and iob->eod
1378 if (csp->flags & CSP_FLAG_CHUNKED)
1380 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1381 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1385 csp->iob->eod = csp->iob->cur + size;
1386 csp->flags |= CSP_FLAG_MODIFIED;
1389 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1390 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1392 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1396 in->buffer = csp->iob->cur;
1399 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1401 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1408 if ((int)size == out->offset)
1410 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1414 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1416 csp->content_length = out->offset;
1417 csp->flags |= CSP_FLAG_MODIFIED;
1427 /*********************************************************************
1429 * Function : remove_chunked_transfer_coding
1431 * Description : In-situ remove the "chunked" transfer coding as defined
1432 * in rfc2616 from a buffer.
1435 * 1 : buffer = Pointer to the text buffer
1436 * 2 : size = Number of bytes to be processed
1438 * Returns : The new size, i.e. the number of bytes from buffer which
1439 * are occupied by the stripped body, or 0 in case something
1442 *********************************************************************/
1443 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1446 unsigned int chunksize = 0;
1447 char *from_p, *to_p;
1450 from_p = to_p = buffer;
1452 if (sscanf(buffer, "%x", &chunksize) != 1)
1454 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1458 while (chunksize > 0)
1460 if (NULL == (from_p = strstr(from_p, "\r\n")))
1462 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1465 newsize += chunksize;
1468 memmove(to_p, from_p, (size_t) chunksize);
1469 to_p = buffer + newsize;
1470 from_p += chunksize + 2;
1472 if (sscanf(from_p, "%x", &chunksize) != 1)
1474 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1479 /* FIXME: Should this get its own loglevel? */
1480 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1486 /*********************************************************************
1488 * Function : url_actions
1490 * Description : Gets the actions for this URL.
1493 * 1 : http = http_request request for blocked URLs
1494 * 2 : csp = Current client state (buffers, headers, etc...)
1498 *********************************************************************/
1499 void url_actions(struct http_request *http,
1500 struct client_state *csp)
1502 struct file_list *fl;
1503 struct url_actions *b;
1506 init_current_action(csp->action);
1508 for (i = 0; i < MAX_ACTION_FILES; i++)
1510 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1515 apply_url_actions(csp->action, http, b);
1522 /*********************************************************************
1524 * Function : apply_url_actions
1526 * Description : Applies a list of URL actions.
1529 * 1 : action = Destination.
1530 * 2 : http = Current URL
1531 * 3 : b = list of URL actions to apply
1535 *********************************************************************/
1536 void apply_url_actions(struct current_action_spec *action,
1537 struct http_request *http,
1538 struct url_actions *b)
1542 /* Should never happen */
1546 for (b = b->next; NULL != b; b = b->next)
1548 if (url_match(b->url, http))
1550 merge_current_action(action, b->action);
1556 /*********************************************************************
1558 * Function : forward_url
1560 * Description : Should we forward this to another proxy?
1563 * 1 : http = http_request request for current URL
1564 * 2 : csp = Current client state (buffers, headers, etc...)
1566 * Returns : Pointer to forwarding information.
1568 *********************************************************************/
1569 const struct forward_spec * forward_url(struct http_request *http,
1570 struct client_state *csp)
1572 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1573 struct forward_spec *fwd = csp->config->forward;
1582 if (url_match(fwd->url, http))