1 const char filters_rcs[] = "$Id: filters.c,v 1.54 2002/04/02 14:55:56 oes 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 1.54 2002/04/02 14:55:56 oes
42 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
44 * Revision 1.53 2002/03/26 22:29:54 swa
45 * we have a new homepage!
47 * Revision 1.52 2002/03/24 16:35:57 jongfoster
50 * Revision 1.51 2002/03/24 15:23:33 jongfoster
53 * Revision 1.50 2002/03/24 13:25:43 swa
54 * name change related issues
56 * Revision 1.49 2002/03/16 20:29:14 oes
59 * Revision 1.48 2002/03/13 20:25:34 oes
60 * Better logging for content filters
62 * Revision 1.47 2002/03/13 00:30:52 jongfoster
64 * Added option of always sending redirect for imageblock,
65 * currently disabled with #if 0.
67 * Revision 1.46 2002/03/12 01:42:49 oes
68 * Introduced modular filters
70 * Revision 1.45 2002/03/08 16:47:50 oes
71 * Added choice beween GIF and PNG built-in images
73 * Revision 1.44 2002/03/07 03:49:31 oes
74 * - Fixed compiler warnings etc
75 * - Changed built-in images from GIF to PNG
76 * (with regard to Unisys patent issue)
77 * - Added a 4x4 pattern PNG which is less intrusive
78 * than the logo but also clearly marks the deleted banners
80 * Revision 1.43 2002/01/22 23:51:59 jongfoster
81 * Replacing strsav() with the safer string_append().
83 * Adding missing html_encode() to error message generators. Where encoded
84 * and unencoded versions of a string were provided, removing the unencoded
87 * Revision 1.42 2002/01/17 21:00:32 jongfoster
88 * Moving all our URL and URL pattern parsing code to urlmatch.c.
90 * Using a single, simple url_match(pattern,url) function - rather than
91 * the 3-line match routine which was repeated all over the place.
93 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
95 * Using parse_http_url() to parse URLs without faking a HTTP
96 * request line for parse_http_request().
98 * Revision 1.41 2001/11/13 00:14:07 jongfoster
99 * Fixing stupid bug now I've figured out what || means.
100 * (It always returns 0 or 1, not one of it's paramaters.)
102 * Revision 1.40 2001/10/26 17:37:55 oes
103 * - Re-enabled Netscape 200/404 bug workaround in block_url():
104 * - Removed OS/2 special case
105 * - Made block_url() independant from sed() having been run
106 * - Made trust_url independant from sed() having been run
107 * - Made is_imageurl independant from sed() having been run.
108 * It now checks User-Agent: and Accept: by itself.
111 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
112 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
113 * threads to call select() simultaneously. So, it's time to do a real, live,
114 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
115 * (native). Both versions will work, but using __OS2__ offers multi-threading.
117 * Revision 1.38 2001/10/23 21:32:33 jongfoster
118 * Adding error-checking to selected functions
120 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
121 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
122 * filters.c. Added a FIXME in front of the offending code. I'll gladly
123 * put in a better/more robust fix for all parties if one is presented...
124 * It seems that just returning 200 instead of 404 would pretty much fix
125 * it for everyone, but I don't know all the history of the problem.
127 * Revision 1.36 2001/10/10 16:44:16 oes
128 * Added match_portlist function
130 * Revision 1.35 2001/10/07 15:41:23 oes
131 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
133 * New function remove_chunked_transfer_coding that strips chunked
134 * transfer coding to plain and is called by pcrs_filter_response
135 * and gif_deanimate_response if neccessary
137 * Improved handling of zero-change re_filter runs
139 * pcrs_filter_response and gif_deanimate_response now remove
140 * chunked transfer codeing before processing the body.
142 * Revision 1.34 2001/09/20 15:49:36 steudten
144 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
145 * See cgi.c fill_template().
147 * Revision 1.33 2001/09/16 17:05:14 jongfoster
148 * Removing unused #include showarg.h
150 * Revision 1.32 2001/09/16 13:21:27 jongfoster
151 * Changes to use new list functions.
153 * Revision 1.31 2001/09/16 11:38:02 jongfoster
154 * Splitting fill_template() into 2 functions:
155 * template_load() loads the file
156 * template_fill() performs the PCRS regexps.
157 * This is because the CGI edit interface has a "table row"
158 * template which is used many times in the page - this
159 * change means it's only loaded from disk once.
161 * Revision 1.30 2001/09/16 11:00:10 jongfoster
162 * New function alloc_http_response, for symmetry with free_http_response
164 * Revision 1.29 2001/09/13 23:32:40 jongfoster
165 * Moving image data to cgi.c rather than cgi.h
166 * Fixing a GPF under Win32 (and any other OS that protects global
167 * constants from being written to).
169 * Revision 1.28 2001/09/10 10:18:51 oes
170 * Silenced compiler warnings
172 * Revision 1.27 2001/08/05 16:06:20 jongfoster
173 * Modifiying "struct map" so that there are now separate header and
174 * "map_entry" structures. This means that functions which modify a
175 * map no longer need to return a pointer to the modified map.
176 * Also, it no longer reverses the order of the entries (which may be
177 * important with some advanced template substitutions).
179 * Revision 1.26 2001/07/30 22:08:36 jongfoster
180 * Tidying up #defines:
181 * - All feature #defines are now of the form FEATURE_xxx
182 * - Permanently turned off WIN_GUI_EDIT
183 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
185 * Revision 1.25 2001/07/26 10:09:46 oes
186 * Made browser detection a little less naive
188 * Revision 1.24 2001/07/25 17:22:51 oes
189 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
191 * Revision 1.23 2001/07/23 13:40:12 oes
192 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
194 * Revision 1.22 2001/07/18 12:29:34 oes
195 * - Made gif_deanimate_response respect
196 * csp->action->string[ACTION_STRING_DEANIMATE]
197 * - Logging cosmetics
199 * Revision 1.21 2001/07/13 13:59:53 oes
200 * - Introduced gif_deanimate_response which shares the
201 * generic content modification interface of pcrs_filter_response
202 * and acts as a wrapper to deanimate.c:gif_deanimate()
203 * - Renamed re_process_buffer to pcrs_filter_response
204 * - pcrs_filter_response now returns NULL on failiure
205 * - Removed all #ifdef PCRS
207 * Revision 1.20 2001/07/01 17:01:04 oes
208 * Added comments and missing return statement in is_untrusted_url()
210 * Revision 1.19 2001/06/29 21:45:41 oes
211 * Indentation, CRLF->LF, Tab-> Space
213 * Revision 1.18 2001/06/29 13:27:38 oes
214 * - Cleaned up, renamed and reorderd functions
215 * and improved comments
218 * - Ported to CGI platform. Now delivers
219 * http_response or NULL
220 * - Unified HTML and GIF generation (moved image detection
221 * and GIF generation here from jcc.c:chat())
222 * - Fixed HTTP status to:
223 * - 403 (Forbidden) for the "blocked" HTML message
224 * - 200 (OK) for GIF answers
225 * - 302 (Redirect) for redirect to GIF
228 * - Ported to CGI platform. Now delivers
229 * http_response or NULL
230 * - Separated detection of untrusted URL into
231 * (bool)is_untrusted_url
232 * - Added enforcement of untrusted requests
234 * - Moved redirect_url() from cgi.c to here
235 * and ported it to the CGI platform
237 * - Removed logentry from cancelled commit
239 * Revision 1.17 2001/06/09 10:55:28 jongfoster
240 * Changing BUFSIZ ==> BUFFER_SIZE
242 * Revision 1.16 2001/06/07 23:10:26 jongfoster
243 * Allowing unanchored domain patterns to back off and retry
244 * if they partially match. Optimized right-anchored patterns.
245 * Moving ACL and forward files into config file.
246 * Replacing struct gateway with struct forward_spec
248 * Revision 1.15 2001/06/03 19:12:00 oes
249 * extracted-CGI relevant stuff
251 * Revision 1.14 2001/06/01 10:30:55 oes
252 * Added optional left-anchoring to domaincmp
254 * Revision 1.13 2001/05/31 21:21:30 jongfoster
255 * Permissionsfile / actions file changes:
256 * - Changed "permission" to "action" throughout
257 * - changes to file format to allow string parameters
258 * - Moved helper functions to actions.c
260 * Revision 1.12 2001/05/31 17:35:20 oes
262 * - Enhanced domain part globbing with infix and prefix asterisk
263 * matching and optional unanchored operation
265 * Revision 1.11 2001/05/29 11:53:23 oes
266 * "See why" link added to "blocked" page
268 * Revision 1.10 2001/05/29 09:50:24 jongfoster
269 * Unified blocklist/imagelist/permissionslist.
270 * File format is still under discussion, but the internal changes
273 * Also modified interceptor behaviour:
274 * - We now intercept all URLs beginning with one of the following
275 * prefixes (and *only* these prefixes):
277 * * http://ijbswa.sf.net/config/
278 * * http://ijbswa.sourceforge.net/config/
279 * - New interceptors "home page" - go to http://i.j.b/ to see it.
280 * - Internal changes so that intercepted and fast redirect pages
281 * are not replaced with an image.
282 * - Interceptors now have the option to send a binary page direct
283 * to the client. (i.e. ijb-send-banner uses this)
284 * - Implemented show-url-info interceptor. (Which is why I needed
285 * the above interceptors changes - a typical URL is
286 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
287 * The previous mechanism would not have intercepted that, and
288 * if it had been intercepted then it then it would have replaced
291 * Revision 1.9 2001/05/27 22:17:04 oes
293 * - re_process_buffer no longer writes the modified buffer
294 * to the client, which was very ugly. It now returns the
295 * buffer, which it is then written by chat.
297 * - content_length now adjusts the Content-Length: header
298 * for modified documents rather than crunch()ing it.
299 * (Length info in csp->content_length, which is 0 for
300 * unmodified documents)
302 * - For this to work, sed() is called twice when filtering.
304 * Revision 1.8 2001/05/26 17:13:28 jongfoster
305 * Filled in a function comment.
307 * Revision 1.7 2001/05/26 15:26:15 jongfoster
308 * ACL feature now provides more security by immediately dropping
309 * connections from untrusted hosts.
311 * Revision 1.6 2001/05/26 00:28:36 jongfoster
312 * Automatic reloading of config file.
313 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
314 * Most of the global variables have been moved to a new
315 * struct configuration_spec, accessed through csp->config->globalname
316 * Most of the globals remaining are used by the Win32 GUI.
318 * Revision 1.5 2001/05/25 22:34:30 jongfoster
321 * Revision 1.4 2001/05/22 18:46:04 oes
323 * - Enabled filtering banners by size rather than URL
324 * by adding patterns that replace all standard banner
325 * sizes with the "Junkbuster" gif to the re_filterfile
327 * - Enabled filtering WebBugs by providing a pattern
328 * which kills all 1x1 images
330 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
331 * which is selected by the (nonstandard and therefore
332 * capital) letter 'U' in the option string.
333 * It causes the quantifiers to be ungreedy by default.
334 * Appending a ? turns back to greedy (!).
336 * - Added a new interceptor ijb-send-banner, which
337 * sends back the "Junkbuster" gif. Without imagelist or
338 * MSIE detection support, or if tinygif = 1, or the
339 * URL isn't recognized as an imageurl, a lame HTML
340 * explanation is sent instead.
342 * - Added new feature, which permits blocking remote
343 * script redirects and firing back a local redirect
345 * The feature is conditionally compiled, i.e. it
346 * can be disabled with --disable-fast-redirects,
347 * plus it must be activated by a "fast-redirects"
348 * line in the config file, has its own log level
349 * and of course wants to be displayed by show-proxy-args
350 * Note: Boy, all the #ifdefs in 1001 locations and
351 * all the fumbling with configure.in and acconfig.h
352 * were *way* more work than the feature itself :-(
354 * - Because a generic redirect template was needed for
355 * this, tinygif = 3 now uses the same.
357 * - Moved GIFs, and other static HTTP response templates
362 * - Removed some >400 CRs again (Jon, you really worked
365 * Revision 1.3 2001/05/20 16:44:47 jongfoster
366 * Removing last hardcoded JunkBusters.com URLs.
368 * Revision 1.2 2001/05/20 01:21:20 jongfoster
369 * Version 2.9.4 checkin.
370 * - Merged popupfile and cookiefile, and added control over PCRS
371 * filtering, in new "permissionsfile".
372 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
373 * file error you now get a message box (in the Win32 GUI) rather
374 * than the program exiting with no explanation.
375 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
377 * - Removed tabs from "config"
378 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
379 * - Bumped up version number.
381 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
382 * Initial import of version 2.9.3 source tree
385 *********************************************************************/
391 #include <sys/types.h>
400 #endif /* ndef __OS2__ */
401 #include <netinet/in.h>
403 #include <winsock2.h>
404 #endif /* ndef _WIN32 */
408 #endif /* def __OS2__ */
416 #include "jbsockets.h"
417 #include "miscutil.h"
421 #include "deanimate.h"
422 #include "urlmatch.h"
428 const char filters_h_rcs[] = FILTERS_H_VERSION;
430 /* Fix a problem with Solaris. There should be no effect on other
432 * Solaris's isspace() is a macro which uses it's argument directly
433 * as an array index. Therefore we need to make sure that high-bit
434 * characters generate +ve values, and ideally we also want to make
435 * the argument match the declared parameter type of "int".
437 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
441 /*********************************************************************
443 * Function : block_acl
445 * Description : Block this request?
446 * Decide yes or no based on ACL file.
449 * 1 : dst = The proxy or gateway address this is going to.
450 * Or NULL to check all possible targets.
451 * 2 : csp = Current client state (buffers, headers, etc...)
452 * Also includes the client IP address.
454 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
456 *********************************************************************/
457 int block_acl(struct access_control_addr *dst, struct client_state *csp)
459 struct access_control_list *acl = csp->config->acl;
461 /* if not using an access control list, then permit the connection */
467 /* search the list */
470 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
474 /* Just want to check if they have any access */
475 if (acl->action == ACL_PERMIT)
480 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
481 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
483 if (acl->action == ACL_PERMIT)
501 /*********************************************************************
503 * Function : acl_addr
505 * Description : Called from `load_config' to parse an ACL address.
508 * 1 : aspec = String specifying ACL address.
509 * 2 : aca = struct access_control_addr to fill in.
511 * Returns : 0 => Ok, everything else is an error.
513 *********************************************************************/
514 int acl_addr(char *aspec, struct access_control_addr *aca)
516 int i, masklength, port;
522 if ((p = strchr(aspec, '/')) != NULL)
526 if (ijb_isdigit(*p) == 0)
530 masklength = atoi(p);
533 if ((masklength < 0) || (masklength > 32))
538 if ((p = strchr(aspec, ':')) != NULL)
542 if (ijb_isdigit(*p) == 0)
551 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
553 if (aca->addr == INADDR_NONE)
558 /* build the netmask */
560 for (i=1; i <= masklength ; i++)
562 aca->mask |= (1 << (32 - i));
565 /* now mask off the host portion of the ip address
566 * (i.e. save on the network portion of the address).
568 aca->addr = aca->addr & aca->mask;
573 #endif /* def FEATURE_ACL */
576 /*********************************************************************
578 * Function : match_portlist
580 * Description : Check if a given number is covered by a comma
581 * separated list of numbers and ranges (a,b-c,d,..)
584 * 1 : portlist = String with list
585 * 2 : port = port to check
587 * Returns : 0 => no match
590 *********************************************************************/
591 int match_portlist(const char *portlist, int port)
593 char *min, *max, *next, *portlist_copy;
595 min = next = portlist_copy = strdup(portlist);
598 * Zero-terminate first item and remember offset for next
600 if (NULL != (next = strchr(portlist_copy, (int) ',')))
606 * Loop through all items, checking for match
610 if (NULL == (max = strchr(min, (int) '-')))
613 * No dash, check for equality
615 if (port == atoi(min))
624 * This is a range, so check if between min and max,
625 * or, if max was omitted, between min and 65K
628 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
642 * Zero-terminate next item and remember offset for n+1
644 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
656 /*********************************************************************
658 * Function : block_url
660 * Description : Called from `chat'. Check to see if we need to block this.
663 * 1 : csp = Current client state (buffers, headers, etc...)
665 * Returns : NULL => unblocked, else HTTP block response
667 *********************************************************************/
668 struct http_response *block_url(struct client_state *csp)
670 #ifdef FEATURE_IMAGE_BLOCKING
672 #endif /* def FEATURE_IMAGE_BLOCKING */
673 struct http_response *rsp;
676 * If it's not blocked, don't block it ;-)
678 if ((csp->action->flags & ACTION_BLOCK) == 0)
684 * Else, prepare a response
686 if (NULL == (rsp = alloc_http_response()))
688 return cgi_error_memory();
692 * If it's an image-url, send back an image or redirect
693 * as specified by the relevant +image action
695 #ifdef FEATURE_IMAGE_BLOCKING
696 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
699 /* determine HOW images should be blocked */
700 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
702 #if 1 /* Two alternative strategies, use this one for now: */
704 /* and handle accordingly: */
705 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
707 rsp->body = bindup(image_pattern_data, image_pattern_length);
708 if (rsp->body == NULL)
710 free_http_response(rsp);
711 return cgi_error_memory();
713 rsp->content_length = image_pattern_length;
715 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
717 free_http_response(rsp);
718 return cgi_error_memory();
722 else if (0 == strcmpic(p, "blank"))
724 rsp->body = bindup(image_blank_data, image_blank_length);
725 if (rsp->body == NULL)
727 free_http_response(rsp);
728 return cgi_error_memory();
730 rsp->content_length = image_blank_length;
732 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
734 free_http_response(rsp);
735 return cgi_error_memory();
741 rsp->status = strdup("302 Local Redirect from Privoxy");
742 if (rsp->status == NULL)
744 free_http_response(rsp);
745 return cgi_error_memory();
748 if (enlist_unique_header(rsp->headers, "Location", p))
750 free_http_response(rsp);
751 return cgi_error_memory();
755 #else /* Following code is disabled for now */
757 /* and handle accordingly: */
758 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
760 p = CGI_PREFIX "send-banner?type=pattern";
762 else if (0 == strcmpic(p, "blank"))
764 p = CGI_PREFIX "send-banner?type=blank";
766 rsp->status = strdup("302 Local Redirect from Privoxy");
767 if (rsp->status == NULL)
769 free_http_response(rsp);
770 return cgi_error_memory();
773 if (enlist_unique_header(rsp->headers, "Location", p))
775 free_http_response(rsp);
776 return cgi_error_memory();
778 #endif /* Preceeding code is disabled for now */
781 #endif /* def FEATURE_IMAGE_BLOCKING */
784 * Else, generate an HTML "blocked" message:
788 struct map * exports;
791 * Workaround for stupid Netscape bug which prevents
792 * pages from being displayed if loading a referenced
793 * JavaScript or style sheet fails. So make it appear
794 * as if it succeeded.
796 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
797 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
798 && !strstr(p, "Gecko") /* save Mozilla, */
799 && !strstr(p, "compatible") /* MSIE */
800 && !strstr(p, "Opera")) /* and Opera. */
802 rsp->status = strdup("200 Request for blocked URL");
806 rsp->status = strdup("404 Request for blocked URL");
809 if (rsp->status == NULL)
811 free_http_response(rsp);
812 return cgi_error_memory();
815 exports = default_exports(csp, NULL);
818 free_http_response(rsp);
819 return cgi_error_memory();
822 #ifdef FEATURE_FORCE_LOAD
823 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
824 if (csp->http->ssl != 0)
825 #endif /* ndef FEATURE_FORCE_LOAD */
827 err = map_block_killer(exports, "force-support");
830 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
831 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
836 free_http_response(rsp);
837 return cgi_error_memory();
840 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
843 free_http_response(rsp);
844 return cgi_error_memory();
848 return finish_http_response(rsp);
854 /*********************************************************************
856 * Function : trust_url FIXME: I should be called distrust_url
858 * Description : Calls is_untrusted_url to determine if the URL is trusted
859 * and if not, returns a HTTP 304 response with a reject message.
862 * 1 : csp = Current client state (buffers, headers, etc...)
864 * Returns : NULL => trusted, else http_response.
866 *********************************************************************/
867 struct http_response *trust_url(struct client_state *csp)
869 struct http_response *rsp;
870 struct map * exports;
871 char buf[BUFFER_SIZE];
873 struct url_spec **tl;
878 * Don't bother to work on trusted URLs
880 if (!is_untrusted_url(csp))
886 * Else, prepare a response:
888 if (NULL == (rsp = alloc_http_response()))
890 return cgi_error_memory();
893 exports = default_exports(csp, NULL);
896 free_http_response(rsp);
897 return cgi_error_memory();
901 * Export the host, port, and referrer information
903 err = map(exports, "hostport", 1, csp->http->hostport, 1);
904 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
906 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
908 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
912 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
918 free_http_response(rsp);
919 return cgi_error_memory();
923 * Export the trust list
926 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
928 sprintf(buf, "<li>%s</li>\n", t->spec);
929 string_append(&p, buf);
931 err = map(exports, "trusted-referrers", 1, p, 0);
936 free_http_response(rsp);
937 return cgi_error_memory();
941 * Export the trust info, if available
943 if (csp->config->trust_info->first)
945 struct list_entry *l;
948 for (l = csp->config->trust_info->first; l ; l = l->next)
950 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
951 string_append(&p, buf);
953 err = map(exports, "trust-info", 1, p, 0);
957 err = map_block_killer(exports, "have-trust-info");
963 free_http_response(rsp);
964 return cgi_error_memory();
968 * Export the force prefix or the force conditional block killer
970 #ifdef FEATURE_FORCE_LOAD
971 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
972 #else /* ifndef FEATURE_FORCE_LOAD */
973 err = map_block_killer(exports, "force-support");
974 #endif /* ndef FEATURE_FORCE_LOAD */
979 free_http_response(rsp);
980 return cgi_error_memory();
986 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
989 free_http_response(rsp);
990 return cgi_error_memory();
993 return finish_http_response(rsp);
995 #endif /* def FEATURE_TRUST */
998 #ifdef FEATURE_FAST_REDIRECTS
999 /*********************************************************************
1001 * Function : redirect_url
1003 * Description : Checks for redirection URLs and returns a HTTP redirect
1004 * to the destination URL, if necessary
1007 * 1 : csp = Current client state (buffers, headers, etc...)
1009 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1011 *********************************************************************/
1012 struct http_response *redirect_url(struct client_state *csp)
1015 struct http_response *rsp;
1017 p = q = csp->http->path;
1018 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1021 * find the last URL encoded in the request
1023 while ((p = strstr(p, "http://")) != NULL)
1029 * if there was any, generate and return a HTTP redirect
1031 if (q != csp->http->path)
1033 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1035 if (NULL == (rsp = alloc_http_response()))
1037 return cgi_error_memory();
1040 if ( enlist_unique_header(rsp->headers, "Location", q)
1041 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1043 free_http_response(rsp);
1044 return cgi_error_memory();
1047 return finish_http_response(rsp);
1055 #endif /* def FEATURE_FAST_REDIRECTS */
1058 #ifdef FEATURE_IMAGE_BLOCKING
1059 /*********************************************************************
1061 * Function : is_imageurl
1063 * Description : Given a URL, decide whether it is an image or not,
1064 * using either the info from a previous +image action
1065 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1066 * the browser's accept header.
1069 * 1 : csp = Current client state (buffers, headers, etc...)
1071 * Returns : True (nonzero) if URL is an image, false (0)
1074 *********************************************************************/
1075 int is_imageurl(struct client_state *csp)
1077 #ifdef FEATURE_IMAGE_DETECT_MSIE
1080 tmp = get_header_value(csp->headers, "User-Agent:");
1081 if (tmp && strstr(tmp, "MSIE"))
1083 tmp = get_header_value(csp->headers, "Accept:");
1084 if (tmp && strstr(tmp, "image/gif"))
1086 /* Client will accept HTML. If this seems counterintuitive,
1096 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1098 return ((csp->action->flags & ACTION_IMAGE) != 0);
1101 #endif /* def FEATURE_IMAGE_BLOCKING */
1104 #ifdef FEATURE_TRUST
1105 /*********************************************************************
1107 * Function : is_untrusted_url
1109 * Description : Should we "distrust" this URL (and block it)?
1111 * Yes if it matches a line in the trustfile, or if the
1112 * referrer matches a line starting with "+" in the
1117 * 1 : csp = Current client state (buffers, headers, etc...)
1119 * Returns : 0 => trusted, 1 => untrusted
1121 *********************************************************************/
1122 int is_untrusted_url(struct client_state *csp)
1124 struct file_list *fl;
1125 struct block_spec *b;
1126 struct url_spec **trusted_url;
1127 struct http_request rhttp[1];
1128 const char * referer;
1132 * If we don't have a trustlist, we trust everybody
1134 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1139 memset(rhttp, '\0', sizeof(*rhttp));
1142 * Do we trust the request URL itself?
1144 for (b = b->next; b ; b = b->next)
1146 if (url_match(b->url, csp->http))
1152 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1154 /* no referrer was supplied */
1160 * If not, do we maybe trust its referrer?
1162 err = parse_http_url(referer, rhttp, csp);
1168 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1170 if (url_match(*trusted_url, rhttp))
1172 /* if the URL's referrer is from a trusted referrer, then
1173 * add the target spec to the trustfile as an unblocked
1174 * domain and return NULL (which means it's OK).
1179 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1183 char * new_entry = strdup("~");
1185 string_append(&new_entry, csp->http->hostport);
1187 path = csp->http->path;
1188 if ( (path[0] == '/')
1190 && ((path_end = strchr(path + 2, '/')) != NULL))
1192 /* since this path points into a user's home space
1193 * be sure to include this spec in the trustfile.
1195 int path_len = path_end - path; /* save offset */
1196 path = strdup(path); /* Copy string */
1199 path_end = path + path_len; /* regenerate ptr to new buffer */
1200 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1202 string_join(&new_entry, path);
1205 if (new_entry != NULL)
1207 fprintf(fp, "%s\n", new_entry);
1212 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1213 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1223 #endif /* def FEATURE_TRUST */
1226 /*********************************************************************
1228 * Function : pcrs_filter_response
1230 * Description : Ecexute all text substitutions from all applying
1231 * +filter actions on the text buffer that's been accumulated
1232 * in csp->iob->buf. If this changes the contents, set
1233 * csp->content_length to the modified size and raise the
1234 * CSP_FLAG_MODIFIED flag.
1237 * 1 : csp = Current client state (buffers, headers, etc...)
1239 * Returns : a pointer to the (newly allocated) modified buffer.
1240 * or NULL if there were no hits or something went wrong
1242 *********************************************************************/
1243 char *pcrs_filter_response(struct client_state *csp)
1248 char *old = csp->iob->cur, *new = NULL;
1251 struct file_list *fl;
1252 struct re_filterfile_spec *b;
1253 struct list_entry *filtername;
1258 if (csp->iob->cur >= csp->iob->eod)
1262 size = csp->iob->eod - csp->iob->cur;
1264 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1266 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1271 * If the body has a "chunked" transfer-encoding,
1272 * get rid of it first, adjusting size and iob->eod
1274 if (csp->flags & CSP_FLAG_CHUNKED)
1276 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1277 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1281 csp->iob->eod = csp->iob->cur + size;
1282 csp->flags |= CSP_FLAG_MODIFIED;
1286 * For all applying +filter actions, look if a filter by that
1287 * name exists and if yes, execute it's pcrs_joblist on the
1290 for (b = fl->f; b; b = b->next)
1292 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1293 filtername ; filtername = filtername->next)
1295 if (strcmp(b->name, filtername->str) == 0)
1297 int current_hits = 0;
1299 if ( NULL == b->joblist )
1301 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1305 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1306 csp->http->hostport, csp->http->path, size, b->name);
1308 /* Apply all jobs from the joblist */
1309 for (job = b->joblist; NULL != job; job = job->next)
1311 current_hits += pcrs_execute(job, old, size, &new, &size);
1312 if (old != csp->iob->cur) free(old);
1316 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1317 hits += current_hits;
1323 * If there were no hits, destroy our copy and let
1324 * chat() use the original in csp->iob
1332 csp->flags |= CSP_FLAG_MODIFIED;
1333 csp->content_length = size;
1341 /*********************************************************************
1343 * Function : gif_deanimate_response
1345 * Description : Deanimate the GIF image that has been accumulated in
1346 * csp->iob->buf, set csp->content_length to the modified
1347 * size and raise the CSP_FLAG_MODIFIED flag.
1350 * 1 : csp = Current client state (buffers, headers, etc...)
1352 * Returns : a pointer to the (newly allocated) modified buffer.
1353 * or NULL in case something went wrong.
1355 *********************************************************************/
1356 char *gif_deanimate_response(struct client_state *csp)
1358 struct binbuffer *in, *out;
1360 size_t size = csp->iob->eod - csp->iob->cur;
1363 * If the body has a "chunked" transfer-encoding,
1364 * get rid of it first, adjusting size and iob->eod
1366 if (csp->flags & CSP_FLAG_CHUNKED)
1368 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1369 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1373 csp->iob->eod = csp->iob->cur + size;
1374 csp->flags |= CSP_FLAG_MODIFIED;
1377 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1378 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1380 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1384 in->buffer = csp->iob->cur;
1387 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1389 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1396 if ((int)size == out->offset)
1398 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1402 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1404 csp->content_length = out->offset;
1405 csp->flags |= CSP_FLAG_MODIFIED;
1415 /*********************************************************************
1417 * Function : remove_chunked_transfer_coding
1419 * Description : In-situ remove the "chunked" transfer coding as defined
1420 * in rfc2616 from a buffer.
1423 * 1 : buffer = Pointer to the text buffer
1424 * 2 : size = Number of bytes to be processed
1426 * Returns : The new size, i.e. the number of bytes from buffer which
1427 * are occupied by the stripped body, or 0 in case something
1430 *********************************************************************/
1431 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1434 unsigned int chunksize = 0;
1435 char *from_p, *to_p;
1438 from_p = to_p = buffer;
1440 if (sscanf(buffer, "%x", &chunksize) != 1)
1442 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1446 while (chunksize > 0)
1448 if (NULL == (from_p = strstr(from_p, "\r\n")))
1450 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1453 newsize += chunksize;
1456 memmove(to_p, from_p, (size_t) chunksize);
1457 to_p = buffer + newsize;
1458 from_p += chunksize + 2;
1460 if (sscanf(from_p, "%x", &chunksize) != 1)
1462 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1467 /* FIXME: Should this get its own loglevel? */
1468 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1474 /*********************************************************************
1476 * Function : url_actions
1478 * Description : Gets the actions for this URL.
1481 * 1 : http = http_request request for blocked URLs
1482 * 2 : csp = Current client state (buffers, headers, etc...)
1486 *********************************************************************/
1487 void url_actions(struct http_request *http,
1488 struct client_state *csp)
1490 struct file_list *fl;
1491 struct url_actions *b;
1493 init_current_action(csp->action);
1495 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1500 apply_url_actions(csp->action, http, b);
1505 /*********************************************************************
1507 * Function : apply_url_actions
1509 * Description : Applies a list of URL actions.
1512 * 1 : action = Destination.
1513 * 2 : http = Current URL
1514 * 3 : b = list of URL actions to apply
1518 *********************************************************************/
1519 void apply_url_actions(struct current_action_spec *action,
1520 struct http_request *http,
1521 struct url_actions *b)
1525 /* Should never happen */
1529 for (b = b->next; NULL != b; b = b->next)
1531 if (url_match(b->url, http))
1533 merge_current_action(action, b->action);
1539 /*********************************************************************
1541 * Function : forward_url
1543 * Description : Should we forward this to another proxy?
1546 * 1 : http = http_request request for current URL
1547 * 2 : csp = Current client state (buffers, headers, etc...)
1549 * Returns : Pointer to forwarding information.
1551 *********************************************************************/
1552 const struct forward_spec * forward_url(struct http_request *http,
1553 struct client_state *csp)
1555 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1556 struct forward_spec *fwd = csp->config->forward;
1565 if (url_match(fwd->url, http))