1 const char filters_rcs[] = "$Id: filters.c,v 1.50 2002/03/24 13:25:43 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://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 1.50 2002/03/24 13:25:43 swa
42 * name change related issues
44 * Revision 1.49 2002/03/16 20:29:14 oes
47 * Revision 1.48 2002/03/13 20:25:34 oes
48 * Better logging for content filters
50 * Revision 1.47 2002/03/13 00:30:52 jongfoster
52 * Added option of always sending redirect for imageblock,
53 * currently disabled with #if 0.
55 * Revision 1.46 2002/03/12 01:42:49 oes
56 * Introduced modular filters
58 * Revision 1.45 2002/03/08 16:47:50 oes
59 * Added choice beween GIF and PNG built-in images
61 * Revision 1.44 2002/03/07 03:49:31 oes
62 * - Fixed compiler warnings etc
63 * - Changed built-in images from GIF to PNG
64 * (with regard to Unisys patent issue)
65 * - Added a 4x4 pattern PNG which is less intrusive
66 * than the logo but also clearly marks the deleted banners
68 * Revision 1.43 2002/01/22 23:51:59 jongfoster
69 * Replacing strsav() with the safer string_append().
71 * Adding missing html_encode() to error message generators. Where encoded
72 * and unencoded versions of a string were provided, removing the unencoded
75 * Revision 1.42 2002/01/17 21:00:32 jongfoster
76 * Moving all our URL and URL pattern parsing code to urlmatch.c.
78 * Using a single, simple url_match(pattern,url) function - rather than
79 * the 3-line match routine which was repeated all over the place.
81 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
83 * Using parse_http_url() to parse URLs without faking a HTTP
84 * request line for parse_http_request().
86 * Revision 1.41 2001/11/13 00:14:07 jongfoster
87 * Fixing stupid bug now I've figured out what || means.
88 * (It always returns 0 or 1, not one of it's paramaters.)
90 * Revision 1.40 2001/10/26 17:37:55 oes
91 * - Re-enabled Netscape 200/404 bug workaround in block_url():
92 * - Removed OS/2 special case
93 * - Made block_url() independant from sed() having been run
94 * - Made trust_url independant from sed() having been run
95 * - Made is_imageurl independant from sed() having been run.
96 * It now checks User-Agent: and Accept: by itself.
99 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
100 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
101 * threads to call select() simultaneously. So, it's time to do a real, live,
102 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
103 * (native). Both versions will work, but using __OS2__ offers multi-threading.
105 * Revision 1.38 2001/10/23 21:32:33 jongfoster
106 * Adding error-checking to selected functions
108 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
109 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
110 * filters.c. Added a FIXME in front of the offending code. I'll gladly
111 * put in a better/more robust fix for all parties if one is presented...
112 * It seems that just returning 200 instead of 404 would pretty much fix
113 * it for everyone, but I don't know all the history of the problem.
115 * Revision 1.36 2001/10/10 16:44:16 oes
116 * Added match_portlist function
118 * Revision 1.35 2001/10/07 15:41:23 oes
119 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
121 * New function remove_chunked_transfer_coding that strips chunked
122 * transfer coding to plain and is called by pcrs_filter_response
123 * and gif_deanimate_response if neccessary
125 * Improved handling of zero-change re_filter runs
127 * pcrs_filter_response and gif_deanimate_response now remove
128 * chunked transfer codeing before processing the body.
130 * Revision 1.34 2001/09/20 15:49:36 steudten
132 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
133 * See cgi.c fill_template().
135 * Revision 1.33 2001/09/16 17:05:14 jongfoster
136 * Removing unused #include showarg.h
138 * Revision 1.32 2001/09/16 13:21:27 jongfoster
139 * Changes to use new list functions.
141 * Revision 1.31 2001/09/16 11:38:02 jongfoster
142 * Splitting fill_template() into 2 functions:
143 * template_load() loads the file
144 * template_fill() performs the PCRS regexps.
145 * This is because the CGI edit interface has a "table row"
146 * template which is used many times in the page - this
147 * change means it's only loaded from disk once.
149 * Revision 1.30 2001/09/16 11:00:10 jongfoster
150 * New function alloc_http_response, for symmetry with free_http_response
152 * Revision 1.29 2001/09/13 23:32:40 jongfoster
153 * Moving image data to cgi.c rather than cgi.h
154 * Fixing a GPF under Win32 (and any other OS that protects global
155 * constants from being written to).
157 * Revision 1.28 2001/09/10 10:18:51 oes
158 * Silenced compiler warnings
160 * Revision 1.27 2001/08/05 16:06:20 jongfoster
161 * Modifiying "struct map" so that there are now separate header and
162 * "map_entry" structures. This means that functions which modify a
163 * map no longer need to return a pointer to the modified map.
164 * Also, it no longer reverses the order of the entries (which may be
165 * important with some advanced template substitutions).
167 * Revision 1.26 2001/07/30 22:08:36 jongfoster
168 * Tidying up #defines:
169 * - All feature #defines are now of the form FEATURE_xxx
170 * - Permanently turned off WIN_GUI_EDIT
171 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
173 * Revision 1.25 2001/07/26 10:09:46 oes
174 * Made browser detection a little less naive
176 * Revision 1.24 2001/07/25 17:22:51 oes
177 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
179 * Revision 1.23 2001/07/23 13:40:12 oes
180 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
182 * Revision 1.22 2001/07/18 12:29:34 oes
183 * - Made gif_deanimate_response respect
184 * csp->action->string[ACTION_STRING_DEANIMATE]
185 * - Logging cosmetics
187 * Revision 1.21 2001/07/13 13:59:53 oes
188 * - Introduced gif_deanimate_response which shares the
189 * generic content modification interface of pcrs_filter_response
190 * and acts as a wrapper to deanimate.c:gif_deanimate()
191 * - Renamed re_process_buffer to pcrs_filter_response
192 * - pcrs_filter_response now returns NULL on failiure
193 * - Removed all #ifdef PCRS
195 * Revision 1.20 2001/07/01 17:01:04 oes
196 * Added comments and missing return statement in is_untrusted_url()
198 * Revision 1.19 2001/06/29 21:45:41 oes
199 * Indentation, CRLF->LF, Tab-> Space
201 * Revision 1.18 2001/06/29 13:27:38 oes
202 * - Cleaned up, renamed and reorderd functions
203 * and improved comments
206 * - Ported to CGI platform. Now delivers
207 * http_response or NULL
208 * - Unified HTML and GIF generation (moved image detection
209 * and GIF generation here from jcc.c:chat())
210 * - Fixed HTTP status to:
211 * - 403 (Forbidden) for the "blocked" HTML message
212 * - 200 (OK) for GIF answers
213 * - 302 (Redirect) for redirect to GIF
216 * - Ported to CGI platform. Now delivers
217 * http_response or NULL
218 * - Separated detection of untrusted URL into
219 * (bool)is_untrusted_url
220 * - Added enforcement of untrusted requests
222 * - Moved redirect_url() from cgi.c to here
223 * and ported it to the CGI platform
225 * - Removed logentry from cancelled commit
227 * Revision 1.17 2001/06/09 10:55:28 jongfoster
228 * Changing BUFSIZ ==> BUFFER_SIZE
230 * Revision 1.16 2001/06/07 23:10:26 jongfoster
231 * Allowing unanchored domain patterns to back off and retry
232 * if they partially match. Optimized right-anchored patterns.
233 * Moving ACL and forward files into config file.
234 * Replacing struct gateway with struct forward_spec
236 * Revision 1.15 2001/06/03 19:12:00 oes
237 * extracted-CGI relevant stuff
239 * Revision 1.14 2001/06/01 10:30:55 oes
240 * Added optional left-anchoring to domaincmp
242 * Revision 1.13 2001/05/31 21:21:30 jongfoster
243 * Permissionsfile / actions file changes:
244 * - Changed "permission" to "action" throughout
245 * - changes to file format to allow string parameters
246 * - Moved helper functions to actions.c
248 * Revision 1.12 2001/05/31 17:35:20 oes
250 * - Enhanced domain part globbing with infix and prefix asterisk
251 * matching and optional unanchored operation
253 * Revision 1.11 2001/05/29 11:53:23 oes
254 * "See why" link added to "blocked" page
256 * Revision 1.10 2001/05/29 09:50:24 jongfoster
257 * Unified blocklist/imagelist/permissionslist.
258 * File format is still under discussion, but the internal changes
261 * Also modified interceptor behaviour:
262 * - We now intercept all URLs beginning with one of the following
263 * prefixes (and *only* these prefixes):
265 * * http://ijbswa.sf.net/config/
266 * * http://ijbswa.sourceforge.net/config/
267 * - New interceptors "home page" - go to http://i.j.b/ to see it.
268 * - Internal changes so that intercepted and fast redirect pages
269 * are not replaced with an image.
270 * - Interceptors now have the option to send a binary page direct
271 * to the client. (i.e. ijb-send-banner uses this)
272 * - Implemented show-url-info interceptor. (Which is why I needed
273 * the above interceptors changes - a typical URL is
274 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
275 * The previous mechanism would not have intercepted that, and
276 * if it had been intercepted then it then it would have replaced
279 * Revision 1.9 2001/05/27 22:17:04 oes
281 * - re_process_buffer no longer writes the modified buffer
282 * to the client, which was very ugly. It now returns the
283 * buffer, which it is then written by chat.
285 * - content_length now adjusts the Content-Length: header
286 * for modified documents rather than crunch()ing it.
287 * (Length info in csp->content_length, which is 0 for
288 * unmodified documents)
290 * - For this to work, sed() is called twice when filtering.
292 * Revision 1.8 2001/05/26 17:13:28 jongfoster
293 * Filled in a function comment.
295 * Revision 1.7 2001/05/26 15:26:15 jongfoster
296 * ACL feature now provides more security by immediately dropping
297 * connections from untrusted hosts.
299 * Revision 1.6 2001/05/26 00:28:36 jongfoster
300 * Automatic reloading of config file.
301 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
302 * Most of the global variables have been moved to a new
303 * struct configuration_spec, accessed through csp->config->globalname
304 * Most of the globals remaining are used by the Win32 GUI.
306 * Revision 1.5 2001/05/25 22:34:30 jongfoster
309 * Revision 1.4 2001/05/22 18:46:04 oes
311 * - Enabled filtering banners by size rather than URL
312 * by adding patterns that replace all standard banner
313 * sizes with the "Junkbuster" gif to the re_filterfile
315 * - Enabled filtering WebBugs by providing a pattern
316 * which kills all 1x1 images
318 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
319 * which is selected by the (nonstandard and therefore
320 * capital) letter 'U' in the option string.
321 * It causes the quantifiers to be ungreedy by default.
322 * Appending a ? turns back to greedy (!).
324 * - Added a new interceptor ijb-send-banner, which
325 * sends back the "Junkbuster" gif. Without imagelist or
326 * MSIE detection support, or if tinygif = 1, or the
327 * URL isn't recognized as an imageurl, a lame HTML
328 * explanation is sent instead.
330 * - Added new feature, which permits blocking remote
331 * script redirects and firing back a local redirect
333 * The feature is conditionally compiled, i.e. it
334 * can be disabled with --disable-fast-redirects,
335 * plus it must be activated by a "fast-redirects"
336 * line in the config file, has its own log level
337 * and of course wants to be displayed by show-proxy-args
338 * Note: Boy, all the #ifdefs in 1001 locations and
339 * all the fumbling with configure.in and acconfig.h
340 * were *way* more work than the feature itself :-(
342 * - Because a generic redirect template was needed for
343 * this, tinygif = 3 now uses the same.
345 * - Moved GIFs, and other static HTTP response templates
350 * - Removed some >400 CRs again (Jon, you really worked
353 * Revision 1.3 2001/05/20 16:44:47 jongfoster
354 * Removing last hardcoded JunkBusters.com URLs.
356 * Revision 1.2 2001/05/20 01:21:20 jongfoster
357 * Version 2.9.4 checkin.
358 * - Merged popupfile and cookiefile, and added control over PCRS
359 * filtering, in new "permissionsfile".
360 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
361 * file error you now get a message box (in the Win32 GUI) rather
362 * than the program exiting with no explanation.
363 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
365 * - Removed tabs from "config"
366 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
367 * - Bumped up version number.
369 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
370 * Initial import of version 2.9.3 source tree
373 *********************************************************************/
379 #include <sys/types.h>
388 #endif /* ndef __OS2__ */
389 #include <netinet/in.h>
391 #include <winsock2.h>
392 #endif /* ndef _WIN32 */
396 #endif /* def __OS2__ */
404 #include "jbsockets.h"
405 #include "miscutil.h"
409 #include "deanimate.h"
410 #include "urlmatch.h"
416 const char filters_h_rcs[] = FILTERS_H_VERSION;
418 /* Fix a problem with Solaris. There should be no effect on other
420 * Solaris's isspace() is a macro which uses it's argument directly
421 * as an array index. Therefore we need to make sure that high-bit
422 * characters generate +ve values, and ideally we also want to make
423 * the argument match the declared parameter type of "int".
425 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
429 /*********************************************************************
431 * Function : block_acl
433 * Description : Block this request?
434 * Decide yes or no based on ACL file.
437 * 1 : dst = The proxy or gateway address this is going to.
438 * Or NULL to check all possible targets.
439 * 2 : csp = Current client state (buffers, headers, etc...)
440 * Also includes the client IP address.
442 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
444 *********************************************************************/
445 int block_acl(struct access_control_addr *dst, struct client_state *csp)
447 struct access_control_list *acl = csp->config->acl;
449 /* if not using an access control list, then permit the connection */
455 /* search the list */
458 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
462 /* Just want to check if they have any access */
463 if (acl->action == ACL_PERMIT)
468 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
469 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
471 if (acl->action == ACL_PERMIT)
489 /*********************************************************************
491 * Function : acl_addr
493 * Description : Called from `load_config' to parse an ACL address.
496 * 1 : aspec = String specifying ACL address.
497 * 2 : aca = struct access_control_addr to fill in.
499 * Returns : 0 => Ok, everything else is an error.
501 *********************************************************************/
502 int acl_addr(char *aspec, struct access_control_addr *aca)
504 int i, masklength, port;
510 if ((p = strchr(aspec, '/')) != NULL)
514 if (ijb_isdigit(*p) == 0)
518 masklength = atoi(p);
521 if ((masklength < 0) || (masklength > 32))
526 if ((p = strchr(aspec, ':')) != NULL)
530 if (ijb_isdigit(*p) == 0)
539 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
541 if (aca->addr == INADDR_NONE)
546 /* build the netmask */
548 for (i=1; i <= masklength ; i++)
550 aca->mask |= (1 << (32 - i));
553 /* now mask off the host portion of the ip address
554 * (i.e. save on the network portion of the address).
556 aca->addr = aca->addr & aca->mask;
561 #endif /* def FEATURE_ACL */
564 /*********************************************************************
566 * Function : match_portlist
568 * Description : Check if a given number is covered by a comma
569 * separated list of numbers and ranges (a,b-c,d,..)
572 * 1 : portlist = String with list
573 * 2 : port = port to check
575 * Returns : 0 => no match
578 *********************************************************************/
579 int match_portlist(const char *portlist, int port)
581 char *min, *max, *next, *portlist_copy;
583 min = next = portlist_copy = strdup(portlist);
586 * Zero-terminate first item and remember offset for next
588 if (NULL != (next = strchr(portlist_copy, (int) ',')))
594 * Loop through all items, checking for match
598 if (NULL == (max = strchr(min, (int) '-')))
601 * No dash, check for equality
603 if (port == atoi(min))
612 * This is a range, so check if between min and max,
613 * or, if max was omitted, between min and 65K
616 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
630 * Zero-terminate next item and remember offset for n+1
632 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
644 /*********************************************************************
646 * Function : block_url
648 * Description : Called from `chat'. Check to see if we need to block this.
651 * 1 : csp = Current client state (buffers, headers, etc...)
653 * Returns : NULL => unblocked, else HTTP block response
655 *********************************************************************/
656 struct http_response *block_url(struct client_state *csp)
658 #ifdef FEATURE_IMAGE_BLOCKING
660 #endif /* def FEATURE_IMAGE_BLOCKING */
661 struct http_response *rsp;
664 * If it's not blocked, don't block it ;-)
666 if ((csp->action->flags & ACTION_BLOCK) == 0)
672 * Else, prepare a response
674 if (NULL == (rsp = alloc_http_response()))
676 return cgi_error_memory();
680 * If it's an image-url, send back an image or redirect
681 * as specified by the relevant +image action
683 #ifdef FEATURE_IMAGE_BLOCKING
684 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
687 /* determine HOW images should be blocked */
688 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
690 #if 1 /* Two alternative strategies, use this one for now: */
692 /* and handle accordingly: */
693 if ((p == NULL) || (0 == strcmpic(p, "logo")))
695 rsp->body = bindup(image_logo_data, image_logo_length);
696 if (rsp->body == NULL)
698 free_http_response(rsp);
699 return cgi_error_memory();
701 rsp->content_length = image_logo_length;
703 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
705 free_http_response(rsp);
706 return cgi_error_memory();
710 else if (0 == strcmpic(p, "blank"))
712 rsp->body = bindup(image_blank_data, image_blank_length);
713 if (rsp->body == NULL)
715 free_http_response(rsp);
716 return cgi_error_memory();
718 rsp->content_length = image_blank_length;
720 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
722 free_http_response(rsp);
723 return cgi_error_memory();
727 else if (0 == strcmpic(p, "pattern"))
729 rsp->body = bindup(image_pattern_data, image_pattern_length);
730 if (rsp->body == NULL)
732 free_http_response(rsp);
733 return cgi_error_memory();
735 rsp->content_length = image_pattern_length;
737 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
739 free_http_response(rsp);
740 return cgi_error_memory();
746 rsp->status = strdup("302 Local Redirect from Privoxy");
747 if (rsp->status == NULL)
749 free_http_response(rsp);
750 return cgi_error_memory();
753 if (enlist_unique_header(rsp->headers, "Location", p))
755 free_http_response(rsp);
756 return cgi_error_memory();
760 #else /* Following code is disabled for now */
762 /* and handle accordingly: */
763 if ((p == NULL) || (0 == strcmpic(p, "logo")))
765 p = CGI_PREFIX "send-banner?type=logo";
767 else if (0 == strcmpic(p, "blank"))
769 p = CGI_PREFIX "send-banner?type=blank";
771 else if (0 == strcmpic(p, "pattern"))
773 p = CGI_PREFIX "send-banner?type=pattern";
775 rsp->status = strdup("302 Local Redirect from Privoxy");
776 if (rsp->status == NULL)
778 free_http_response(rsp);
779 return cgi_error_memory();
782 if (enlist_unique_header(rsp->headers, "Location", p))
784 free_http_response(rsp);
785 return cgi_error_memory();
787 #endif /* Preceeding code is disabled for now */
790 #endif /* def FEATURE_IMAGE_BLOCKING */
793 * Else, generate an HTML "blocked" message:
797 struct map * exports;
800 * Workaround for stupid Netscape bug which prevents
801 * pages from being displayed if loading a referenced
802 * JavaScript or style sheet fails. So make it appear
803 * as if it succeeded.
805 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
806 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
807 && !strstr(p, "Gecko") /* save Mozilla, */
808 && !strstr(p, "compatible") /* MSIE */
809 && !strstr(p, "Opera")) /* and Opera. */
811 rsp->status = strdup("200 Request for blocked URL");
815 rsp->status = strdup("404 Request for blocked URL");
818 if (rsp->status == NULL)
820 free_http_response(rsp);
821 return cgi_error_memory();
824 exports = default_exports(csp, NULL);
827 free_http_response(rsp);
828 return cgi_error_memory();
831 #ifdef FEATURE_FORCE_LOAD
832 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
833 #else /* ifndef FEATURE_FORCE_LOAD */
834 err = map_block_killer(exports, "force-support");
835 #endif /* ndef FEATURE_FORCE_LOAD */
837 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
838 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
843 free_http_response(rsp);
844 return cgi_error_memory();
847 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
850 free_http_response(rsp);
851 return cgi_error_memory();
855 return finish_http_response(rsp);
861 /*********************************************************************
863 * Function : trust_url FIXME: I should be called distrust_url
865 * Description : Calls is_untrusted_url to determine if the URL is trusted
866 * and if not, returns a HTTP 304 response with a reject message.
869 * 1 : csp = Current client state (buffers, headers, etc...)
871 * Returns : NULL => trusted, else http_response.
873 *********************************************************************/
874 struct http_response *trust_url(struct client_state *csp)
876 struct http_response *rsp;
877 struct map * exports;
878 char buf[BUFFER_SIZE];
880 struct url_spec **tl;
885 * Don't bother to work on trusted URLs
887 if (!is_untrusted_url(csp))
893 * Else, prepare a response:
895 if (NULL == (rsp = alloc_http_response()))
897 return cgi_error_memory();
900 exports = default_exports(csp, NULL);
903 free_http_response(rsp);
904 return cgi_error_memory();
908 * Export the host, port, and referrer information
910 err = map(exports, "hostport", 1, csp->http->hostport, 1);
911 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
913 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
915 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
919 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
925 free_http_response(rsp);
926 return cgi_error_memory();
930 * Export the trust list
933 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
935 sprintf(buf, "<li>%s</li>\n", t->spec);
936 string_append(&p, buf);
938 err = map(exports, "trusted-referrers", 1, p, 0);
943 free_http_response(rsp);
944 return cgi_error_memory();
948 * Export the trust info, if available
950 if (csp->config->trust_info->first)
952 struct list_entry *l;
955 for (l = csp->config->trust_info->first; l ; l = l->next)
957 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
958 string_append(&p, buf);
960 err = map(exports, "trust-info", 1, p, 0);
964 err = map_block_killer(exports, "have-trust-info");
970 free_http_response(rsp);
971 return cgi_error_memory();
975 * Export the force prefix or the force conditional block killer
977 #ifdef FEATURE_FORCE_LOAD
978 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
979 #else /* ifndef FEATURE_FORCE_LOAD */
980 err = map_block_killer(exports, "force-support");
981 #endif /* ndef FEATURE_FORCE_LOAD */
986 free_http_response(rsp);
987 return cgi_error_memory();
993 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
996 free_http_response(rsp);
997 return cgi_error_memory();
1000 return finish_http_response(rsp);
1002 #endif /* def FEATURE_TRUST */
1005 #ifdef FEATURE_FAST_REDIRECTS
1006 /*********************************************************************
1008 * Function : redirect_url
1010 * Description : Checks for redirection URLs and returns a HTTP redirect
1011 * to the destination URL, if necessary
1014 * 1 : csp = Current client state (buffers, headers, etc...)
1016 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1018 *********************************************************************/
1019 struct http_response *redirect_url(struct client_state *csp)
1022 struct http_response *rsp;
1024 p = q = csp->http->path;
1025 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1028 * find the last URL encoded in the request
1030 while ((p = strstr(p, "http://")) != NULL)
1036 * if there was any, generate and return a HTTP redirect
1038 if (q != csp->http->path)
1040 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1042 if (NULL == (rsp = alloc_http_response()))
1044 return cgi_error_memory();
1047 if ( enlist_unique_header(rsp->headers, "Location", q)
1048 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1050 free_http_response(rsp);
1051 return cgi_error_memory();
1054 return finish_http_response(rsp);
1062 #endif /* def FEATURE_FAST_REDIRECTS */
1065 #ifdef FEATURE_IMAGE_BLOCKING
1066 /*********************************************************************
1068 * Function : is_imageurl
1070 * Description : Given a URL, decide whether it is an image or not,
1071 * using either the info from a previous +image action
1072 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1073 * the browser's accept header.
1076 * 1 : csp = Current client state (buffers, headers, etc...)
1078 * Returns : True (nonzero) if URL is an image, false (0)
1081 *********************************************************************/
1082 int is_imageurl(struct client_state *csp)
1084 #ifdef FEATURE_IMAGE_DETECT_MSIE
1087 tmp = get_header_value(csp->headers, "User-Agent:");
1088 if (tmp && strstr(tmp, "MSIE"))
1090 tmp = get_header_value(csp->headers, "Accept:");
1091 if (tmp && strstr(tmp, "image/gif"))
1093 /* Client will accept HTML. If this seems counterintuitive,
1103 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1105 return ((csp->action->flags & ACTION_IMAGE) != 0);
1108 #endif /* def FEATURE_IMAGE_BLOCKING */
1111 #ifdef FEATURE_COOKIE_JAR
1112 /*********************************************************************
1114 * Function : is_untrusted_url
1116 * Description : Should we "distrust" this URL (and block it)?
1118 * Yes if it matches a line in the trustfile, or if the
1119 * referrer matches a line starting with "+" in the
1124 * 1 : csp = Current client state (buffers, headers, etc...)
1126 * Returns : 0 => trusted, 1 => untrusted
1128 *********************************************************************/
1129 int is_untrusted_url(struct client_state *csp)
1131 struct file_list *fl;
1132 struct block_spec *b;
1133 struct url_spec **trusted_url;
1134 struct http_request rhttp[1];
1135 const char * referer;
1139 * If we don't have a trustlist, we trust everybody
1141 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1146 memset(rhttp, '\0', sizeof(*rhttp));
1149 * Do we trust the request URL itself?
1151 for (b = b->next; b ; b = b->next)
1153 if (url_match(b->url, csp->http))
1159 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1161 /* no referrer was supplied */
1166 * If not, do we maybe trust its referrer?
1171 * Parse the URL from the 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_COOKIE_JAR */
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;
1505 init_current_action(csp->action);
1507 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1512 apply_url_actions(csp->action, http, b);
1517 /*********************************************************************
1519 * Function : apply_url_actions
1521 * Description : Applies a list of URL actions.
1524 * 1 : action = Destination.
1525 * 2 : http = Current URL
1526 * 3 : b = list of URL actions to apply
1530 *********************************************************************/
1531 void apply_url_actions(struct current_action_spec *action,
1532 struct http_request *http,
1533 struct url_actions *b)
1537 /* Should never happen */
1541 for (b = b->next; NULL != b; b = b->next)
1543 if (url_match(b->url, http))
1545 merge_current_action(action, b->action);
1551 /*********************************************************************
1553 * Function : forward_url
1555 * Description : Should we forward this to another proxy?
1558 * 1 : http = http_request request for current URL
1559 * 2 : csp = Current client state (buffers, headers, etc...)
1561 * Returns : Pointer to forwarding information.
1563 *********************************************************************/
1564 const struct forward_spec * forward_url(struct http_request *http,
1565 struct client_state *csp)
1567 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1568 struct forward_spec *fwd = csp->config->forward;
1577 if (url_match(fwd->url, http))