1 const char filters_rcs[] = "$Id: filters.c,v 1.51 2002/03/24 15:23:33 jongfoster 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://ijbswa.sourceforge.net
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 1.51 2002/03/24 15:23:33 jongfoster
44 * Revision 1.50 2002/03/24 13:25:43 swa
45 * name change related issues
47 * Revision 1.49 2002/03/16 20:29:14 oes
50 * Revision 1.48 2002/03/13 20:25:34 oes
51 * Better logging for content filters
53 * Revision 1.47 2002/03/13 00:30:52 jongfoster
55 * Added option of always sending redirect for imageblock,
56 * currently disabled with #if 0.
58 * Revision 1.46 2002/03/12 01:42:49 oes
59 * Introduced modular filters
61 * Revision 1.45 2002/03/08 16:47:50 oes
62 * Added choice beween GIF and PNG built-in images
64 * Revision 1.44 2002/03/07 03:49:31 oes
65 * - Fixed compiler warnings etc
66 * - Changed built-in images from GIF to PNG
67 * (with regard to Unisys patent issue)
68 * - Added a 4x4 pattern PNG which is less intrusive
69 * than the logo but also clearly marks the deleted banners
71 * Revision 1.43 2002/01/22 23:51:59 jongfoster
72 * Replacing strsav() with the safer string_append().
74 * Adding missing html_encode() to error message generators. Where encoded
75 * and unencoded versions of a string were provided, removing the unencoded
78 * Revision 1.42 2002/01/17 21:00:32 jongfoster
79 * Moving all our URL and URL pattern parsing code to urlmatch.c.
81 * Using a single, simple url_match(pattern,url) function - rather than
82 * the 3-line match routine which was repeated all over the place.
84 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
86 * Using parse_http_url() to parse URLs without faking a HTTP
87 * request line for parse_http_request().
89 * Revision 1.41 2001/11/13 00:14:07 jongfoster
90 * Fixing stupid bug now I've figured out what || means.
91 * (It always returns 0 or 1, not one of it's paramaters.)
93 * Revision 1.40 2001/10/26 17:37:55 oes
94 * - Re-enabled Netscape 200/404 bug workaround in block_url():
95 * - Removed OS/2 special case
96 * - Made block_url() independant from sed() having been run
97 * - Made trust_url independant from sed() having been run
98 * - Made is_imageurl independant from sed() having been run.
99 * It now checks User-Agent: and Accept: by itself.
102 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
103 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
104 * threads to call select() simultaneously. So, it's time to do a real, live,
105 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
106 * (native). Both versions will work, but using __OS2__ offers multi-threading.
108 * Revision 1.38 2001/10/23 21:32:33 jongfoster
109 * Adding error-checking to selected functions
111 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
112 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
113 * filters.c. Added a FIXME in front of the offending code. I'll gladly
114 * put in a better/more robust fix for all parties if one is presented...
115 * It seems that just returning 200 instead of 404 would pretty much fix
116 * it for everyone, but I don't know all the history of the problem.
118 * Revision 1.36 2001/10/10 16:44:16 oes
119 * Added match_portlist function
121 * Revision 1.35 2001/10/07 15:41:23 oes
122 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
124 * New function remove_chunked_transfer_coding that strips chunked
125 * transfer coding to plain and is called by pcrs_filter_response
126 * and gif_deanimate_response if neccessary
128 * Improved handling of zero-change re_filter runs
130 * pcrs_filter_response and gif_deanimate_response now remove
131 * chunked transfer codeing before processing the body.
133 * Revision 1.34 2001/09/20 15:49:36 steudten
135 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
136 * See cgi.c fill_template().
138 * Revision 1.33 2001/09/16 17:05:14 jongfoster
139 * Removing unused #include showarg.h
141 * Revision 1.32 2001/09/16 13:21:27 jongfoster
142 * Changes to use new list functions.
144 * Revision 1.31 2001/09/16 11:38:02 jongfoster
145 * Splitting fill_template() into 2 functions:
146 * template_load() loads the file
147 * template_fill() performs the PCRS regexps.
148 * This is because the CGI edit interface has a "table row"
149 * template which is used many times in the page - this
150 * change means it's only loaded from disk once.
152 * Revision 1.30 2001/09/16 11:00:10 jongfoster
153 * New function alloc_http_response, for symmetry with free_http_response
155 * Revision 1.29 2001/09/13 23:32:40 jongfoster
156 * Moving image data to cgi.c rather than cgi.h
157 * Fixing a GPF under Win32 (and any other OS that protects global
158 * constants from being written to).
160 * Revision 1.28 2001/09/10 10:18:51 oes
161 * Silenced compiler warnings
163 * Revision 1.27 2001/08/05 16:06:20 jongfoster
164 * Modifiying "struct map" so that there are now separate header and
165 * "map_entry" structures. This means that functions which modify a
166 * map no longer need to return a pointer to the modified map.
167 * Also, it no longer reverses the order of the entries (which may be
168 * important with some advanced template substitutions).
170 * Revision 1.26 2001/07/30 22:08:36 jongfoster
171 * Tidying up #defines:
172 * - All feature #defines are now of the form FEATURE_xxx
173 * - Permanently turned off WIN_GUI_EDIT
174 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
176 * Revision 1.25 2001/07/26 10:09:46 oes
177 * Made browser detection a little less naive
179 * Revision 1.24 2001/07/25 17:22:51 oes
180 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
182 * Revision 1.23 2001/07/23 13:40:12 oes
183 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
185 * Revision 1.22 2001/07/18 12:29:34 oes
186 * - Made gif_deanimate_response respect
187 * csp->action->string[ACTION_STRING_DEANIMATE]
188 * - Logging cosmetics
190 * Revision 1.21 2001/07/13 13:59:53 oes
191 * - Introduced gif_deanimate_response which shares the
192 * generic content modification interface of pcrs_filter_response
193 * and acts as a wrapper to deanimate.c:gif_deanimate()
194 * - Renamed re_process_buffer to pcrs_filter_response
195 * - pcrs_filter_response now returns NULL on failiure
196 * - Removed all #ifdef PCRS
198 * Revision 1.20 2001/07/01 17:01:04 oes
199 * Added comments and missing return statement in is_untrusted_url()
201 * Revision 1.19 2001/06/29 21:45:41 oes
202 * Indentation, CRLF->LF, Tab-> Space
204 * Revision 1.18 2001/06/29 13:27:38 oes
205 * - Cleaned up, renamed and reorderd functions
206 * and improved comments
209 * - Ported to CGI platform. Now delivers
210 * http_response or NULL
211 * - Unified HTML and GIF generation (moved image detection
212 * and GIF generation here from jcc.c:chat())
213 * - Fixed HTTP status to:
214 * - 403 (Forbidden) for the "blocked" HTML message
215 * - 200 (OK) for GIF answers
216 * - 302 (Redirect) for redirect to GIF
219 * - Ported to CGI platform. Now delivers
220 * http_response or NULL
221 * - Separated detection of untrusted URL into
222 * (bool)is_untrusted_url
223 * - Added enforcement of untrusted requests
225 * - Moved redirect_url() from cgi.c to here
226 * and ported it to the CGI platform
228 * - Removed logentry from cancelled commit
230 * Revision 1.17 2001/06/09 10:55:28 jongfoster
231 * Changing BUFSIZ ==> BUFFER_SIZE
233 * Revision 1.16 2001/06/07 23:10:26 jongfoster
234 * Allowing unanchored domain patterns to back off and retry
235 * if they partially match. Optimized right-anchored patterns.
236 * Moving ACL and forward files into config file.
237 * Replacing struct gateway with struct forward_spec
239 * Revision 1.15 2001/06/03 19:12:00 oes
240 * extracted-CGI relevant stuff
242 * Revision 1.14 2001/06/01 10:30:55 oes
243 * Added optional left-anchoring to domaincmp
245 * Revision 1.13 2001/05/31 21:21:30 jongfoster
246 * Permissionsfile / actions file changes:
247 * - Changed "permission" to "action" throughout
248 * - changes to file format to allow string parameters
249 * - Moved helper functions to actions.c
251 * Revision 1.12 2001/05/31 17:35:20 oes
253 * - Enhanced domain part globbing with infix and prefix asterisk
254 * matching and optional unanchored operation
256 * Revision 1.11 2001/05/29 11:53:23 oes
257 * "See why" link added to "blocked" page
259 * Revision 1.10 2001/05/29 09:50:24 jongfoster
260 * Unified blocklist/imagelist/permissionslist.
261 * File format is still under discussion, but the internal changes
264 * Also modified interceptor behaviour:
265 * - We now intercept all URLs beginning with one of the following
266 * prefixes (and *only* these prefixes):
268 * * http://ijbswa.sf.net/config/
269 * * http://ijbswa.sourceforge.net/config/
270 * - New interceptors "home page" - go to http://i.j.b/ to see it.
271 * - Internal changes so that intercepted and fast redirect pages
272 * are not replaced with an image.
273 * - Interceptors now have the option to send a binary page direct
274 * to the client. (i.e. ijb-send-banner uses this)
275 * - Implemented show-url-info interceptor. (Which is why I needed
276 * the above interceptors changes - a typical URL is
277 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
278 * The previous mechanism would not have intercepted that, and
279 * if it had been intercepted then it then it would have replaced
282 * Revision 1.9 2001/05/27 22:17:04 oes
284 * - re_process_buffer no longer writes the modified buffer
285 * to the client, which was very ugly. It now returns the
286 * buffer, which it is then written by chat.
288 * - content_length now adjusts the Content-Length: header
289 * for modified documents rather than crunch()ing it.
290 * (Length info in csp->content_length, which is 0 for
291 * unmodified documents)
293 * - For this to work, sed() is called twice when filtering.
295 * Revision 1.8 2001/05/26 17:13:28 jongfoster
296 * Filled in a function comment.
298 * Revision 1.7 2001/05/26 15:26:15 jongfoster
299 * ACL feature now provides more security by immediately dropping
300 * connections from untrusted hosts.
302 * Revision 1.6 2001/05/26 00:28:36 jongfoster
303 * Automatic reloading of config file.
304 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
305 * Most of the global variables have been moved to a new
306 * struct configuration_spec, accessed through csp->config->globalname
307 * Most of the globals remaining are used by the Win32 GUI.
309 * Revision 1.5 2001/05/25 22:34:30 jongfoster
312 * Revision 1.4 2001/05/22 18:46:04 oes
314 * - Enabled filtering banners by size rather than URL
315 * by adding patterns that replace all standard banner
316 * sizes with the "Junkbuster" gif to the re_filterfile
318 * - Enabled filtering WebBugs by providing a pattern
319 * which kills all 1x1 images
321 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
322 * which is selected by the (nonstandard and therefore
323 * capital) letter 'U' in the option string.
324 * It causes the quantifiers to be ungreedy by default.
325 * Appending a ? turns back to greedy (!).
327 * - Added a new interceptor ijb-send-banner, which
328 * sends back the "Junkbuster" gif. Without imagelist or
329 * MSIE detection support, or if tinygif = 1, or the
330 * URL isn't recognized as an imageurl, a lame HTML
331 * explanation is sent instead.
333 * - Added new feature, which permits blocking remote
334 * script redirects and firing back a local redirect
336 * The feature is conditionally compiled, i.e. it
337 * can be disabled with --disable-fast-redirects,
338 * plus it must be activated by a "fast-redirects"
339 * line in the config file, has its own log level
340 * and of course wants to be displayed by show-proxy-args
341 * Note: Boy, all the #ifdefs in 1001 locations and
342 * all the fumbling with configure.in and acconfig.h
343 * were *way* more work than the feature itself :-(
345 * - Because a generic redirect template was needed for
346 * this, tinygif = 3 now uses the same.
348 * - Moved GIFs, and other static HTTP response templates
353 * - Removed some >400 CRs again (Jon, you really worked
356 * Revision 1.3 2001/05/20 16:44:47 jongfoster
357 * Removing last hardcoded JunkBusters.com URLs.
359 * Revision 1.2 2001/05/20 01:21:20 jongfoster
360 * Version 2.9.4 checkin.
361 * - Merged popupfile and cookiefile, and added control over PCRS
362 * filtering, in new "permissionsfile".
363 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
364 * file error you now get a message box (in the Win32 GUI) rather
365 * than the program exiting with no explanation.
366 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
368 * - Removed tabs from "config"
369 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
370 * - Bumped up version number.
372 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
373 * Initial import of version 2.9.3 source tree
376 *********************************************************************/
382 #include <sys/types.h>
391 #endif /* ndef __OS2__ */
392 #include <netinet/in.h>
394 #include <winsock2.h>
395 #endif /* ndef _WIN32 */
399 #endif /* def __OS2__ */
407 #include "jbsockets.h"
408 #include "miscutil.h"
412 #include "deanimate.h"
413 #include "urlmatch.h"
419 const char filters_h_rcs[] = FILTERS_H_VERSION;
421 /* Fix a problem with Solaris. There should be no effect on other
423 * Solaris's isspace() is a macro which uses it's argument directly
424 * as an array index. Therefore we need to make sure that high-bit
425 * characters generate +ve values, and ideally we also want to make
426 * the argument match the declared parameter type of "int".
428 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
432 /*********************************************************************
434 * Function : block_acl
436 * Description : Block this request?
437 * Decide yes or no based on ACL file.
440 * 1 : dst = The proxy or gateway address this is going to.
441 * Or NULL to check all possible targets.
442 * 2 : csp = Current client state (buffers, headers, etc...)
443 * Also includes the client IP address.
445 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
447 *********************************************************************/
448 int block_acl(struct access_control_addr *dst, struct client_state *csp)
450 struct access_control_list *acl = csp->config->acl;
452 /* if not using an access control list, then permit the connection */
458 /* search the list */
461 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
465 /* Just want to check if they have any access */
466 if (acl->action == ACL_PERMIT)
471 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
472 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
474 if (acl->action == ACL_PERMIT)
492 /*********************************************************************
494 * Function : acl_addr
496 * Description : Called from `load_config' to parse an ACL address.
499 * 1 : aspec = String specifying ACL address.
500 * 2 : aca = struct access_control_addr to fill in.
502 * Returns : 0 => Ok, everything else is an error.
504 *********************************************************************/
505 int acl_addr(char *aspec, struct access_control_addr *aca)
507 int i, masklength, port;
513 if ((p = strchr(aspec, '/')) != NULL)
517 if (ijb_isdigit(*p) == 0)
521 masklength = atoi(p);
524 if ((masklength < 0) || (masklength > 32))
529 if ((p = strchr(aspec, ':')) != NULL)
533 if (ijb_isdigit(*p) == 0)
542 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
544 if (aca->addr == INADDR_NONE)
549 /* build the netmask */
551 for (i=1; i <= masklength ; i++)
553 aca->mask |= (1 << (32 - i));
556 /* now mask off the host portion of the ip address
557 * (i.e. save on the network portion of the address).
559 aca->addr = aca->addr & aca->mask;
564 #endif /* def FEATURE_ACL */
567 /*********************************************************************
569 * Function : match_portlist
571 * Description : Check if a given number is covered by a comma
572 * separated list of numbers and ranges (a,b-c,d,..)
575 * 1 : portlist = String with list
576 * 2 : port = port to check
578 * Returns : 0 => no match
581 *********************************************************************/
582 int match_portlist(const char *portlist, int port)
584 char *min, *max, *next, *portlist_copy;
586 min = next = portlist_copy = strdup(portlist);
589 * Zero-terminate first item and remember offset for next
591 if (NULL != (next = strchr(portlist_copy, (int) ',')))
597 * Loop through all items, checking for match
601 if (NULL == (max = strchr(min, (int) '-')))
604 * No dash, check for equality
606 if (port == atoi(min))
615 * This is a range, so check if between min and max,
616 * or, if max was omitted, between min and 65K
619 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
633 * Zero-terminate next item and remember offset for n+1
635 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
647 /*********************************************************************
649 * Function : block_url
651 * Description : Called from `chat'. Check to see if we need to block this.
654 * 1 : csp = Current client state (buffers, headers, etc...)
656 * Returns : NULL => unblocked, else HTTP block response
658 *********************************************************************/
659 struct http_response *block_url(struct client_state *csp)
661 #ifdef FEATURE_IMAGE_BLOCKING
663 #endif /* def FEATURE_IMAGE_BLOCKING */
664 struct http_response *rsp;
667 * If it's not blocked, don't block it ;-)
669 if ((csp->action->flags & ACTION_BLOCK) == 0)
675 * Else, prepare a response
677 if (NULL == (rsp = alloc_http_response()))
679 return cgi_error_memory();
683 * If it's an image-url, send back an image or redirect
684 * as specified by the relevant +image action
686 #ifdef FEATURE_IMAGE_BLOCKING
687 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
690 /* determine HOW images should be blocked */
691 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
693 #if 1 /* Two alternative strategies, use this one for now: */
695 /* and handle accordingly: */
696 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
698 rsp->body = bindup(image_pattern_data, image_pattern_length);
699 if (rsp->body == NULL)
701 free_http_response(rsp);
702 return cgi_error_memory();
704 rsp->content_length = image_pattern_length;
706 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
708 free_http_response(rsp);
709 return cgi_error_memory();
713 else if (0 == strcmpic(p, "blank"))
715 rsp->body = bindup(image_blank_data, image_blank_length);
716 if (rsp->body == NULL)
718 free_http_response(rsp);
719 return cgi_error_memory();
721 rsp->content_length = image_blank_length;
723 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
725 free_http_response(rsp);
726 return cgi_error_memory();
732 rsp->status = strdup("302 Local Redirect from Privoxy");
733 if (rsp->status == NULL)
735 free_http_response(rsp);
736 return cgi_error_memory();
739 if (enlist_unique_header(rsp->headers, "Location", p))
741 free_http_response(rsp);
742 return cgi_error_memory();
746 #else /* Following code is disabled for now */
748 /* and handle accordingly: */
749 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
751 p = CGI_PREFIX "send-banner?type=pattern";
753 else if (0 == strcmpic(p, "blank"))
755 p = CGI_PREFIX "send-banner?type=blank";
757 rsp->status = strdup("302 Local Redirect from Privoxy");
758 if (rsp->status == NULL)
760 free_http_response(rsp);
761 return cgi_error_memory();
764 if (enlist_unique_header(rsp->headers, "Location", p))
766 free_http_response(rsp);
767 return cgi_error_memory();
769 #endif /* Preceeding code is disabled for now */
772 #endif /* def FEATURE_IMAGE_BLOCKING */
775 * Else, generate an HTML "blocked" message:
779 struct map * exports;
782 * Workaround for stupid Netscape bug which prevents
783 * pages from being displayed if loading a referenced
784 * JavaScript or style sheet fails. So make it appear
785 * as if it succeeded.
787 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
788 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
789 && !strstr(p, "Gecko") /* save Mozilla, */
790 && !strstr(p, "compatible") /* MSIE */
791 && !strstr(p, "Opera")) /* and Opera. */
793 rsp->status = strdup("200 Request for blocked URL");
797 rsp->status = strdup("404 Request for blocked URL");
800 if (rsp->status == NULL)
802 free_http_response(rsp);
803 return cgi_error_memory();
806 exports = default_exports(csp, NULL);
809 free_http_response(rsp);
810 return cgi_error_memory();
813 #ifdef FEATURE_FORCE_LOAD
814 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
815 #else /* ifndef FEATURE_FORCE_LOAD */
816 err = map_block_killer(exports, "force-support");
817 #endif /* ndef FEATURE_FORCE_LOAD */
819 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
820 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
825 free_http_response(rsp);
826 return cgi_error_memory();
829 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
832 free_http_response(rsp);
833 return cgi_error_memory();
837 return finish_http_response(rsp);
843 /*********************************************************************
845 * Function : trust_url FIXME: I should be called distrust_url
847 * Description : Calls is_untrusted_url to determine if the URL is trusted
848 * and if not, returns a HTTP 304 response with a reject message.
851 * 1 : csp = Current client state (buffers, headers, etc...)
853 * Returns : NULL => trusted, else http_response.
855 *********************************************************************/
856 struct http_response *trust_url(struct client_state *csp)
858 struct http_response *rsp;
859 struct map * exports;
860 char buf[BUFFER_SIZE];
862 struct url_spec **tl;
867 * Don't bother to work on trusted URLs
869 if (!is_untrusted_url(csp))
875 * Else, prepare a response:
877 if (NULL == (rsp = alloc_http_response()))
879 return cgi_error_memory();
882 exports = default_exports(csp, NULL);
885 free_http_response(rsp);
886 return cgi_error_memory();
890 * Export the host, port, and referrer information
892 err = map(exports, "hostport", 1, csp->http->hostport, 1);
893 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
895 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
897 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
901 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
907 free_http_response(rsp);
908 return cgi_error_memory();
912 * Export the trust list
915 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
917 sprintf(buf, "<li>%s</li>\n", t->spec);
918 string_append(&p, buf);
920 err = map(exports, "trusted-referrers", 1, p, 0);
925 free_http_response(rsp);
926 return cgi_error_memory();
930 * Export the trust info, if available
932 if (csp->config->trust_info->first)
934 struct list_entry *l;
937 for (l = csp->config->trust_info->first; l ; l = l->next)
939 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
940 string_append(&p, buf);
942 err = map(exports, "trust-info", 1, p, 0);
946 err = map_block_killer(exports, "have-trust-info");
952 free_http_response(rsp);
953 return cgi_error_memory();
957 * Export the force prefix or the force conditional block killer
959 #ifdef FEATURE_FORCE_LOAD
960 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
961 #else /* ifndef FEATURE_FORCE_LOAD */
962 err = map_block_killer(exports, "force-support");
963 #endif /* ndef FEATURE_FORCE_LOAD */
968 free_http_response(rsp);
969 return cgi_error_memory();
975 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
978 free_http_response(rsp);
979 return cgi_error_memory();
982 return finish_http_response(rsp);
984 #endif /* def FEATURE_TRUST */
987 #ifdef FEATURE_FAST_REDIRECTS
988 /*********************************************************************
990 * Function : redirect_url
992 * Description : Checks for redirection URLs and returns a HTTP redirect
993 * to the destination URL, if necessary
996 * 1 : csp = Current client state (buffers, headers, etc...)
998 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1000 *********************************************************************/
1001 struct http_response *redirect_url(struct client_state *csp)
1004 struct http_response *rsp;
1006 p = q = csp->http->path;
1007 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1010 * find the last URL encoded in the request
1012 while ((p = strstr(p, "http://")) != NULL)
1018 * if there was any, generate and return a HTTP redirect
1020 if (q != csp->http->path)
1022 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1024 if (NULL == (rsp = alloc_http_response()))
1026 return cgi_error_memory();
1029 if ( enlist_unique_header(rsp->headers, "Location", q)
1030 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1032 free_http_response(rsp);
1033 return cgi_error_memory();
1036 return finish_http_response(rsp);
1044 #endif /* def FEATURE_FAST_REDIRECTS */
1047 #ifdef FEATURE_IMAGE_BLOCKING
1048 /*********************************************************************
1050 * Function : is_imageurl
1052 * Description : Given a URL, decide whether it is an image or not,
1053 * using either the info from a previous +image action
1054 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1055 * the browser's accept header.
1058 * 1 : csp = Current client state (buffers, headers, etc...)
1060 * Returns : True (nonzero) if URL is an image, false (0)
1063 *********************************************************************/
1064 int is_imageurl(struct client_state *csp)
1066 #ifdef FEATURE_IMAGE_DETECT_MSIE
1069 tmp = get_header_value(csp->headers, "User-Agent:");
1070 if (tmp && strstr(tmp, "MSIE"))
1072 tmp = get_header_value(csp->headers, "Accept:");
1073 if (tmp && strstr(tmp, "image/gif"))
1075 /* Client will accept HTML. If this seems counterintuitive,
1085 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1087 return ((csp->action->flags & ACTION_IMAGE) != 0);
1090 #endif /* def FEATURE_IMAGE_BLOCKING */
1093 #ifdef FEATURE_COOKIE_JAR
1094 /*********************************************************************
1096 * Function : is_untrusted_url
1098 * Description : Should we "distrust" this URL (and block it)?
1100 * Yes if it matches a line in the trustfile, or if the
1101 * referrer matches a line starting with "+" in the
1106 * 1 : csp = Current client state (buffers, headers, etc...)
1108 * Returns : 0 => trusted, 1 => untrusted
1110 *********************************************************************/
1111 int is_untrusted_url(struct client_state *csp)
1113 struct file_list *fl;
1114 struct block_spec *b;
1115 struct url_spec **trusted_url;
1116 struct http_request rhttp[1];
1117 const char * referer;
1121 * If we don't have a trustlist, we trust everybody
1123 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1128 memset(rhttp, '\0', sizeof(*rhttp));
1131 * Do we trust the request URL itself?
1133 for (b = b->next; b ; b = b->next)
1135 if (url_match(b->url, csp->http))
1141 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1143 /* no referrer was supplied */
1148 * If not, do we maybe trust its referrer?
1153 * Parse the URL from the referrer
1156 err = parse_http_url(referer, rhttp, csp);
1162 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1164 if (url_match(*trusted_url, rhttp))
1166 /* if the URL's referrer is from a trusted referrer, then
1167 * add the target spec to the trustfile as an unblocked
1168 * domain and return NULL (which means it's OK).
1173 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1177 char * new_entry = strdup("~");
1179 string_append(&new_entry, csp->http->hostport);
1181 path = csp->http->path;
1182 if ( (path[0] == '/')
1184 && ((path_end = strchr(path + 2, '/')) != NULL))
1186 /* since this path points into a user's home space
1187 * be sure to include this spec in the trustfile.
1189 int path_len = path_end - path; /* save offset */
1190 path = strdup(path); /* Copy string */
1193 path_end = path + path_len; /* regenerate ptr to new buffer */
1194 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1196 string_join(&new_entry, path);
1199 if (new_entry != NULL)
1201 fprintf(fp, "%s\n", new_entry);
1206 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1207 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1217 #endif /* def FEATURE_COOKIE_JAR */
1220 /*********************************************************************
1222 * Function : pcrs_filter_response
1224 * Description : Ecexute all text substitutions from all applying
1225 * +filter actions on the text buffer that's been accumulated
1226 * in csp->iob->buf. If this changes the contents, set
1227 * csp->content_length to the modified size and raise the
1228 * CSP_FLAG_MODIFIED flag.
1231 * 1 : csp = Current client state (buffers, headers, etc...)
1233 * Returns : a pointer to the (newly allocated) modified buffer.
1234 * or NULL if there were no hits or something went wrong
1236 *********************************************************************/
1237 char *pcrs_filter_response(struct client_state *csp)
1242 char *old = csp->iob->cur, *new = NULL;
1245 struct file_list *fl;
1246 struct re_filterfile_spec *b;
1247 struct list_entry *filtername;
1252 if (csp->iob->cur >= csp->iob->eod)
1256 size = csp->iob->eod - csp->iob->cur;
1258 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1260 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1265 * If the body has a "chunked" transfer-encoding,
1266 * get rid of it first, adjusting size and iob->eod
1268 if (csp->flags & CSP_FLAG_CHUNKED)
1270 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1271 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1275 csp->iob->eod = csp->iob->cur + size;
1276 csp->flags |= CSP_FLAG_MODIFIED;
1280 * For all applying +filter actions, look if a filter by that
1281 * name exists and if yes, execute it's pcrs_joblist on the
1284 for (b = fl->f; b; b = b->next)
1286 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1287 filtername ; filtername = filtername->next)
1289 if (strcmp(b->name, filtername->str) == 0)
1291 int current_hits = 0;
1293 if ( NULL == b->joblist )
1295 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1299 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1300 csp->http->hostport, csp->http->path, size, b->name);
1302 /* Apply all jobs from the joblist */
1303 for (job = b->joblist; NULL != job; job = job->next)
1305 current_hits += pcrs_execute(job, old, size, &new, &size);
1306 if (old != csp->iob->cur) free(old);
1310 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1311 hits += current_hits;
1317 * If there were no hits, destroy our copy and let
1318 * chat() use the original in csp->iob
1326 csp->flags |= CSP_FLAG_MODIFIED;
1327 csp->content_length = size;
1335 /*********************************************************************
1337 * Function : gif_deanimate_response
1339 * Description : Deanimate the GIF image that has been accumulated in
1340 * csp->iob->buf, set csp->content_length to the modified
1341 * size and raise the CSP_FLAG_MODIFIED flag.
1344 * 1 : csp = Current client state (buffers, headers, etc...)
1346 * Returns : a pointer to the (newly allocated) modified buffer.
1347 * or NULL in case something went wrong.
1349 *********************************************************************/
1350 char *gif_deanimate_response(struct client_state *csp)
1352 struct binbuffer *in, *out;
1354 size_t size = csp->iob->eod - csp->iob->cur;
1357 * If the body has a "chunked" transfer-encoding,
1358 * get rid of it first, adjusting size and iob->eod
1360 if (csp->flags & CSP_FLAG_CHUNKED)
1362 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1363 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1367 csp->iob->eod = csp->iob->cur + size;
1368 csp->flags |= CSP_FLAG_MODIFIED;
1371 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1372 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1374 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1378 in->buffer = csp->iob->cur;
1381 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1383 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1390 if ((int)size == out->offset)
1392 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1396 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1398 csp->content_length = out->offset;
1399 csp->flags |= CSP_FLAG_MODIFIED;
1409 /*********************************************************************
1411 * Function : remove_chunked_transfer_coding
1413 * Description : In-situ remove the "chunked" transfer coding as defined
1414 * in rfc2616 from a buffer.
1417 * 1 : buffer = Pointer to the text buffer
1418 * 2 : size = Number of bytes to be processed
1420 * Returns : The new size, i.e. the number of bytes from buffer which
1421 * are occupied by the stripped body, or 0 in case something
1424 *********************************************************************/
1425 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1428 unsigned int chunksize = 0;
1429 char *from_p, *to_p;
1432 from_p = to_p = buffer;
1434 if (sscanf(buffer, "%x", &chunksize) != 1)
1436 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1440 while (chunksize > 0)
1442 if (NULL == (from_p = strstr(from_p, "\r\n")))
1444 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1447 newsize += chunksize;
1450 memmove(to_p, from_p, (size_t) chunksize);
1451 to_p = buffer + newsize;
1452 from_p += chunksize + 2;
1454 if (sscanf(from_p, "%x", &chunksize) != 1)
1456 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1461 /* FIXME: Should this get its own loglevel? */
1462 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1468 /*********************************************************************
1470 * Function : url_actions
1472 * Description : Gets the actions for this URL.
1475 * 1 : http = http_request request for blocked URLs
1476 * 2 : csp = Current client state (buffers, headers, etc...)
1480 *********************************************************************/
1481 void url_actions(struct http_request *http,
1482 struct client_state *csp)
1484 struct file_list *fl;
1485 struct url_actions *b;
1487 init_current_action(csp->action);
1489 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1494 apply_url_actions(csp->action, http, b);
1499 /*********************************************************************
1501 * Function : apply_url_actions
1503 * Description : Applies a list of URL actions.
1506 * 1 : action = Destination.
1507 * 2 : http = Current URL
1508 * 3 : b = list of URL actions to apply
1512 *********************************************************************/
1513 void apply_url_actions(struct current_action_spec *action,
1514 struct http_request *http,
1515 struct url_actions *b)
1519 /* Should never happen */
1523 for (b = b->next; NULL != b; b = b->next)
1525 if (url_match(b->url, http))
1527 merge_current_action(action, b->action);
1533 /*********************************************************************
1535 * Function : forward_url
1537 * Description : Should we forward this to another proxy?
1540 * 1 : http = http_request request for current URL
1541 * 2 : csp = Current client state (buffers, headers, etc...)
1543 * Returns : Pointer to forwarding information.
1545 *********************************************************************/
1546 const struct forward_spec * forward_url(struct http_request *http,
1547 struct client_state *csp)
1549 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1550 struct forward_spec *fwd = csp->config->forward;
1559 if (url_match(fwd->url, http))