1 const char filters_rcs[] = "$Id: filters.c,v 1.55 2002/04/02 16:13:51 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 1.55 2002/04/02 16:13:51 oes
42 * Fix: No "Go there anyway" for SSL
44 * Revision 1.54 2002/04/02 14:55:56 oes
45 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
47 * Revision 1.53 2002/03/26 22:29:54 swa
48 * we have a new homepage!
50 * Revision 1.52 2002/03/24 16:35:57 jongfoster
53 * Revision 1.51 2002/03/24 15:23:33 jongfoster
56 * Revision 1.50 2002/03/24 13:25:43 swa
57 * name change related issues
59 * Revision 1.49 2002/03/16 20:29:14 oes
62 * Revision 1.48 2002/03/13 20:25:34 oes
63 * Better logging for content filters
65 * Revision 1.47 2002/03/13 00:30:52 jongfoster
67 * Added option of always sending redirect for imageblock,
68 * currently disabled with #if 0.
70 * Revision 1.46 2002/03/12 01:42:49 oes
71 * Introduced modular filters
73 * Revision 1.45 2002/03/08 16:47:50 oes
74 * Added choice beween GIF and PNG built-in images
76 * Revision 1.44 2002/03/07 03:49:31 oes
77 * - Fixed compiler warnings etc
78 * - Changed built-in images from GIF to PNG
79 * (with regard to Unisys patent issue)
80 * - Added a 4x4 pattern PNG which is less intrusive
81 * than the logo but also clearly marks the deleted banners
83 * Revision 1.43 2002/01/22 23:51:59 jongfoster
84 * Replacing strsav() with the safer string_append().
86 * Adding missing html_encode() to error message generators. Where encoded
87 * and unencoded versions of a string were provided, removing the unencoded
90 * Revision 1.42 2002/01/17 21:00:32 jongfoster
91 * Moving all our URL and URL pattern parsing code to urlmatch.c.
93 * Using a single, simple url_match(pattern,url) function - rather than
94 * the 3-line match routine which was repeated all over the place.
96 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
98 * Using parse_http_url() to parse URLs without faking a HTTP
99 * request line for parse_http_request().
101 * Revision 1.41 2001/11/13 00:14:07 jongfoster
102 * Fixing stupid bug now I've figured out what || means.
103 * (It always returns 0 or 1, not one of it's paramaters.)
105 * Revision 1.40 2001/10/26 17:37:55 oes
106 * - Re-enabled Netscape 200/404 bug workaround in block_url():
107 * - Removed OS/2 special case
108 * - Made block_url() independant from sed() having been run
109 * - Made trust_url independant from sed() having been run
110 * - Made is_imageurl independant from sed() having been run.
111 * It now checks User-Agent: and Accept: by itself.
114 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
115 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
116 * threads to call select() simultaneously. So, it's time to do a real, live,
117 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
118 * (native). Both versions will work, but using __OS2__ offers multi-threading.
120 * Revision 1.38 2001/10/23 21:32:33 jongfoster
121 * Adding error-checking to selected functions
123 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
124 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
125 * filters.c. Added a FIXME in front of the offending code. I'll gladly
126 * put in a better/more robust fix for all parties if one is presented...
127 * It seems that just returning 200 instead of 404 would pretty much fix
128 * it for everyone, but I don't know all the history of the problem.
130 * Revision 1.36 2001/10/10 16:44:16 oes
131 * Added match_portlist function
133 * Revision 1.35 2001/10/07 15:41:23 oes
134 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
136 * New function remove_chunked_transfer_coding that strips chunked
137 * transfer coding to plain and is called by pcrs_filter_response
138 * and gif_deanimate_response if neccessary
140 * Improved handling of zero-change re_filter runs
142 * pcrs_filter_response and gif_deanimate_response now remove
143 * chunked transfer codeing before processing the body.
145 * Revision 1.34 2001/09/20 15:49:36 steudten
147 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
148 * See cgi.c fill_template().
150 * Revision 1.33 2001/09/16 17:05:14 jongfoster
151 * Removing unused #include showarg.h
153 * Revision 1.32 2001/09/16 13:21:27 jongfoster
154 * Changes to use new list functions.
156 * Revision 1.31 2001/09/16 11:38:02 jongfoster
157 * Splitting fill_template() into 2 functions:
158 * template_load() loads the file
159 * template_fill() performs the PCRS regexps.
160 * This is because the CGI edit interface has a "table row"
161 * template which is used many times in the page - this
162 * change means it's only loaded from disk once.
164 * Revision 1.30 2001/09/16 11:00:10 jongfoster
165 * New function alloc_http_response, for symmetry with free_http_response
167 * Revision 1.29 2001/09/13 23:32:40 jongfoster
168 * Moving image data to cgi.c rather than cgi.h
169 * Fixing a GPF under Win32 (and any other OS that protects global
170 * constants from being written to).
172 * Revision 1.28 2001/09/10 10:18:51 oes
173 * Silenced compiler warnings
175 * Revision 1.27 2001/08/05 16:06:20 jongfoster
176 * Modifiying "struct map" so that there are now separate header and
177 * "map_entry" structures. This means that functions which modify a
178 * map no longer need to return a pointer to the modified map.
179 * Also, it no longer reverses the order of the entries (which may be
180 * important with some advanced template substitutions).
182 * Revision 1.26 2001/07/30 22:08:36 jongfoster
183 * Tidying up #defines:
184 * - All feature #defines are now of the form FEATURE_xxx
185 * - Permanently turned off WIN_GUI_EDIT
186 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
188 * Revision 1.25 2001/07/26 10:09:46 oes
189 * Made browser detection a little less naive
191 * Revision 1.24 2001/07/25 17:22:51 oes
192 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
194 * Revision 1.23 2001/07/23 13:40:12 oes
195 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
197 * Revision 1.22 2001/07/18 12:29:34 oes
198 * - Made gif_deanimate_response respect
199 * csp->action->string[ACTION_STRING_DEANIMATE]
200 * - Logging cosmetics
202 * Revision 1.21 2001/07/13 13:59:53 oes
203 * - Introduced gif_deanimate_response which shares the
204 * generic content modification interface of pcrs_filter_response
205 * and acts as a wrapper to deanimate.c:gif_deanimate()
206 * - Renamed re_process_buffer to pcrs_filter_response
207 * - pcrs_filter_response now returns NULL on failiure
208 * - Removed all #ifdef PCRS
210 * Revision 1.20 2001/07/01 17:01:04 oes
211 * Added comments and missing return statement in is_untrusted_url()
213 * Revision 1.19 2001/06/29 21:45:41 oes
214 * Indentation, CRLF->LF, Tab-> Space
216 * Revision 1.18 2001/06/29 13:27:38 oes
217 * - Cleaned up, renamed and reorderd functions
218 * and improved comments
221 * - Ported to CGI platform. Now delivers
222 * http_response or NULL
223 * - Unified HTML and GIF generation (moved image detection
224 * and GIF generation here from jcc.c:chat())
225 * - Fixed HTTP status to:
226 * - 403 (Forbidden) for the "blocked" HTML message
227 * - 200 (OK) for GIF answers
228 * - 302 (Redirect) for redirect to GIF
231 * - Ported to CGI platform. Now delivers
232 * http_response or NULL
233 * - Separated detection of untrusted URL into
234 * (bool)is_untrusted_url
235 * - Added enforcement of untrusted requests
237 * - Moved redirect_url() from cgi.c to here
238 * and ported it to the CGI platform
240 * - Removed logentry from cancelled commit
242 * Revision 1.17 2001/06/09 10:55:28 jongfoster
243 * Changing BUFSIZ ==> BUFFER_SIZE
245 * Revision 1.16 2001/06/07 23:10:26 jongfoster
246 * Allowing unanchored domain patterns to back off and retry
247 * if they partially match. Optimized right-anchored patterns.
248 * Moving ACL and forward files into config file.
249 * Replacing struct gateway with struct forward_spec
251 * Revision 1.15 2001/06/03 19:12:00 oes
252 * extracted-CGI relevant stuff
254 * Revision 1.14 2001/06/01 10:30:55 oes
255 * Added optional left-anchoring to domaincmp
257 * Revision 1.13 2001/05/31 21:21:30 jongfoster
258 * Permissionsfile / actions file changes:
259 * - Changed "permission" to "action" throughout
260 * - changes to file format to allow string parameters
261 * - Moved helper functions to actions.c
263 * Revision 1.12 2001/05/31 17:35:20 oes
265 * - Enhanced domain part globbing with infix and prefix asterisk
266 * matching and optional unanchored operation
268 * Revision 1.11 2001/05/29 11:53:23 oes
269 * "See why" link added to "blocked" page
271 * Revision 1.10 2001/05/29 09:50:24 jongfoster
272 * Unified blocklist/imagelist/permissionslist.
273 * File format is still under discussion, but the internal changes
276 * Also modified interceptor behaviour:
277 * - We now intercept all URLs beginning with one of the following
278 * prefixes (and *only* these prefixes):
280 * * http://ijbswa.sf.net/config/
281 * * http://ijbswa.sourceforge.net/config/
282 * - New interceptors "home page" - go to http://i.j.b/ to see it.
283 * - Internal changes so that intercepted and fast redirect pages
284 * are not replaced with an image.
285 * - Interceptors now have the option to send a binary page direct
286 * to the client. (i.e. ijb-send-banner uses this)
287 * - Implemented show-url-info interceptor. (Which is why I needed
288 * the above interceptors changes - a typical URL is
289 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
290 * The previous mechanism would not have intercepted that, and
291 * if it had been intercepted then it then it would have replaced
294 * Revision 1.9 2001/05/27 22:17:04 oes
296 * - re_process_buffer no longer writes the modified buffer
297 * to the client, which was very ugly. It now returns the
298 * buffer, which it is then written by chat.
300 * - content_length now adjusts the Content-Length: header
301 * for modified documents rather than crunch()ing it.
302 * (Length info in csp->content_length, which is 0 for
303 * unmodified documents)
305 * - For this to work, sed() is called twice when filtering.
307 * Revision 1.8 2001/05/26 17:13:28 jongfoster
308 * Filled in a function comment.
310 * Revision 1.7 2001/05/26 15:26:15 jongfoster
311 * ACL feature now provides more security by immediately dropping
312 * connections from untrusted hosts.
314 * Revision 1.6 2001/05/26 00:28:36 jongfoster
315 * Automatic reloading of config file.
316 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
317 * Most of the global variables have been moved to a new
318 * struct configuration_spec, accessed through csp->config->globalname
319 * Most of the globals remaining are used by the Win32 GUI.
321 * Revision 1.5 2001/05/25 22:34:30 jongfoster
324 * Revision 1.4 2001/05/22 18:46:04 oes
326 * - Enabled filtering banners by size rather than URL
327 * by adding patterns that replace all standard banner
328 * sizes with the "Junkbuster" gif to the re_filterfile
330 * - Enabled filtering WebBugs by providing a pattern
331 * which kills all 1x1 images
333 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
334 * which is selected by the (nonstandard and therefore
335 * capital) letter 'U' in the option string.
336 * It causes the quantifiers to be ungreedy by default.
337 * Appending a ? turns back to greedy (!).
339 * - Added a new interceptor ijb-send-banner, which
340 * sends back the "Junkbuster" gif. Without imagelist or
341 * MSIE detection support, or if tinygif = 1, or the
342 * URL isn't recognized as an imageurl, a lame HTML
343 * explanation is sent instead.
345 * - Added new feature, which permits blocking remote
346 * script redirects and firing back a local redirect
348 * The feature is conditionally compiled, i.e. it
349 * can be disabled with --disable-fast-redirects,
350 * plus it must be activated by a "fast-redirects"
351 * line in the config file, has its own log level
352 * and of course wants to be displayed by show-proxy-args
353 * Note: Boy, all the #ifdefs in 1001 locations and
354 * all the fumbling with configure.in and acconfig.h
355 * were *way* more work than the feature itself :-(
357 * - Because a generic redirect template was needed for
358 * this, tinygif = 3 now uses the same.
360 * - Moved GIFs, and other static HTTP response templates
365 * - Removed some >400 CRs again (Jon, you really worked
368 * Revision 1.3 2001/05/20 16:44:47 jongfoster
369 * Removing last hardcoded JunkBusters.com URLs.
371 * Revision 1.2 2001/05/20 01:21:20 jongfoster
372 * Version 2.9.4 checkin.
373 * - Merged popupfile and cookiefile, and added control over PCRS
374 * filtering, in new "permissionsfile".
375 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
376 * file error you now get a message box (in the Win32 GUI) rather
377 * than the program exiting with no explanation.
378 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
380 * - Removed tabs from "config"
381 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
382 * - Bumped up version number.
384 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
385 * Initial import of version 2.9.3 source tree
388 *********************************************************************/
394 #include <sys/types.h>
403 #endif /* ndef __OS2__ */
404 #include <netinet/in.h>
406 #include <winsock2.h>
407 #endif /* ndef _WIN32 */
411 #endif /* def __OS2__ */
419 #include "jbsockets.h"
420 #include "miscutil.h"
424 #include "deanimate.h"
425 #include "urlmatch.h"
431 const char filters_h_rcs[] = FILTERS_H_VERSION;
433 /* Fix a problem with Solaris. There should be no effect on other
435 * Solaris's isspace() is a macro which uses it's argument directly
436 * as an array index. Therefore we need to make sure that high-bit
437 * characters generate +ve values, and ideally we also want to make
438 * the argument match the declared parameter type of "int".
440 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
444 /*********************************************************************
446 * Function : block_acl
448 * Description : Block this request?
449 * Decide yes or no based on ACL file.
452 * 1 : dst = The proxy or gateway address this is going to.
453 * Or NULL to check all possible targets.
454 * 2 : csp = Current client state (buffers, headers, etc...)
455 * Also includes the client IP address.
457 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
459 *********************************************************************/
460 int block_acl(struct access_control_addr *dst, struct client_state *csp)
462 struct access_control_list *acl = csp->config->acl;
464 /* if not using an access control list, then permit the connection */
470 /* search the list */
473 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
477 /* Just want to check if they have any access */
478 if (acl->action == ACL_PERMIT)
483 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
484 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
486 if (acl->action == ACL_PERMIT)
504 /*********************************************************************
506 * Function : acl_addr
508 * Description : Called from `load_config' to parse an ACL address.
511 * 1 : aspec = String specifying ACL address.
512 * 2 : aca = struct access_control_addr to fill in.
514 * Returns : 0 => Ok, everything else is an error.
516 *********************************************************************/
517 int acl_addr(char *aspec, struct access_control_addr *aca)
519 int i, masklength, port;
525 if ((p = strchr(aspec, '/')) != NULL)
529 if (ijb_isdigit(*p) == 0)
533 masklength = atoi(p);
536 if ((masklength < 0) || (masklength > 32))
541 if ((p = strchr(aspec, ':')) != NULL)
545 if (ijb_isdigit(*p) == 0)
554 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
556 if (aca->addr == INADDR_NONE)
561 /* build the netmask */
563 for (i=1; i <= masklength ; i++)
565 aca->mask |= (1 << (32 - i));
568 /* now mask off the host portion of the ip address
569 * (i.e. save on the network portion of the address).
571 aca->addr = aca->addr & aca->mask;
576 #endif /* def FEATURE_ACL */
579 /*********************************************************************
581 * Function : match_portlist
583 * Description : Check if a given number is covered by a comma
584 * separated list of numbers and ranges (a,b-c,d,..)
587 * 1 : portlist = String with list
588 * 2 : port = port to check
590 * Returns : 0 => no match
593 *********************************************************************/
594 int match_portlist(const char *portlist, int port)
596 char *min, *max, *next, *portlist_copy;
598 min = next = portlist_copy = strdup(portlist);
601 * Zero-terminate first item and remember offset for next
603 if (NULL != (next = strchr(portlist_copy, (int) ',')))
609 * Loop through all items, checking for match
613 if (NULL == (max = strchr(min, (int) '-')))
616 * No dash, check for equality
618 if (port == atoi(min))
627 * This is a range, so check if between min and max,
628 * or, if max was omitted, between min and 65K
631 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
645 * Zero-terminate next item and remember offset for n+1
647 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
659 /*********************************************************************
661 * Function : block_url
663 * Description : Called from `chat'. Check to see if we need to block this.
666 * 1 : csp = Current client state (buffers, headers, etc...)
668 * Returns : NULL => unblocked, else HTTP block response
670 *********************************************************************/
671 struct http_response *block_url(struct client_state *csp)
673 #ifdef FEATURE_IMAGE_BLOCKING
675 #endif /* def FEATURE_IMAGE_BLOCKING */
676 struct http_response *rsp;
679 * If it's not blocked, don't block it ;-)
681 if ((csp->action->flags & ACTION_BLOCK) == 0)
687 * Else, prepare a response
689 if (NULL == (rsp = alloc_http_response()))
691 return cgi_error_memory();
695 * If it's an image-url, send back an image or redirect
696 * as specified by the relevant +image action
698 #ifdef FEATURE_IMAGE_BLOCKING
699 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
702 /* determine HOW images should be blocked */
703 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
705 #if 1 /* Two alternative strategies, use this one for now: */
707 /* and handle accordingly: */
708 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
710 rsp->body = bindup(image_pattern_data, image_pattern_length);
711 if (rsp->body == NULL)
713 free_http_response(rsp);
714 return cgi_error_memory();
716 rsp->content_length = image_pattern_length;
718 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
720 free_http_response(rsp);
721 return cgi_error_memory();
725 else if (0 == strcmpic(p, "blank"))
727 rsp->body = bindup(image_blank_data, image_blank_length);
728 if (rsp->body == NULL)
730 free_http_response(rsp);
731 return cgi_error_memory();
733 rsp->content_length = image_blank_length;
735 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
737 free_http_response(rsp);
738 return cgi_error_memory();
744 rsp->status = strdup("302 Local Redirect from Privoxy");
745 if (rsp->status == NULL)
747 free_http_response(rsp);
748 return cgi_error_memory();
751 if (enlist_unique_header(rsp->headers, "Location", p))
753 free_http_response(rsp);
754 return cgi_error_memory();
758 #else /* Following code is disabled for now */
760 /* and handle accordingly: */
761 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
763 p = CGI_PREFIX "send-banner?type=pattern";
765 else if (0 == strcmpic(p, "blank"))
767 p = CGI_PREFIX "send-banner?type=blank";
769 rsp->status = strdup("302 Local Redirect from Privoxy");
770 if (rsp->status == NULL)
772 free_http_response(rsp);
773 return cgi_error_memory();
776 if (enlist_unique_header(rsp->headers, "Location", p))
778 free_http_response(rsp);
779 return cgi_error_memory();
781 #endif /* Preceeding code is disabled for now */
784 #endif /* def FEATURE_IMAGE_BLOCKING */
787 * Else, generate an HTML "blocked" message:
791 struct map * exports;
794 * Workaround for stupid Netscape bug which prevents
795 * pages from being displayed if loading a referenced
796 * JavaScript or style sheet fails. So make it appear
797 * as if it succeeded.
799 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
800 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
801 && !strstr(p, "Gecko") /* save Mozilla, */
802 && !strstr(p, "compatible") /* MSIE */
803 && !strstr(p, "Opera")) /* and Opera. */
805 rsp->status = strdup("200 Request for blocked URL");
809 rsp->status = strdup("404 Request for blocked URL");
812 if (rsp->status == NULL)
814 free_http_response(rsp);
815 return cgi_error_memory();
818 exports = default_exports(csp, NULL);
821 free_http_response(rsp);
822 return cgi_error_memory();
825 #ifdef FEATURE_FORCE_LOAD
826 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
827 if (csp->http->ssl != 0)
828 #endif /* ndef FEATURE_FORCE_LOAD */
830 err = map_block_killer(exports, "force-support");
833 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
834 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
835 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
840 free_http_response(rsp);
841 return cgi_error_memory();
844 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
847 free_http_response(rsp);
848 return cgi_error_memory();
852 return finish_http_response(rsp);
858 /*********************************************************************
860 * Function : trust_url FIXME: I should be called distrust_url
862 * Description : Calls is_untrusted_url to determine if the URL is trusted
863 * and if not, returns a HTTP 304 response with a reject message.
866 * 1 : csp = Current client state (buffers, headers, etc...)
868 * Returns : NULL => trusted, else http_response.
870 *********************************************************************/
871 struct http_response *trust_url(struct client_state *csp)
873 struct http_response *rsp;
874 struct map * exports;
875 char buf[BUFFER_SIZE];
877 struct url_spec **tl;
882 * Don't bother to work on trusted URLs
884 if (!is_untrusted_url(csp))
890 * Else, prepare a response:
892 if (NULL == (rsp = alloc_http_response()))
894 return cgi_error_memory();
897 exports = default_exports(csp, NULL);
900 free_http_response(rsp);
901 return cgi_error_memory();
905 * Export the protocol, host, port, and referrer information
907 err = map(exports, "hostport", 1, csp->http->hostport, 1);
908 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
909 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
911 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
913 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
917 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
923 free_http_response(rsp);
924 return cgi_error_memory();
928 * Export the trust list
931 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
933 sprintf(buf, "<li>%s</li>\n", t->spec);
934 string_append(&p, buf);
936 err = map(exports, "trusted-referrers", 1, p, 0);
941 free_http_response(rsp);
942 return cgi_error_memory();
946 * Export the trust info, if available
948 if (csp->config->trust_info->first)
950 struct list_entry *l;
953 for (l = csp->config->trust_info->first; l ; l = l->next)
955 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
956 string_append(&p, buf);
958 err = map(exports, "trust-info", 1, p, 0);
962 err = map_block_killer(exports, "have-trust-info");
968 free_http_response(rsp);
969 return cgi_error_memory();
973 * Export the force prefix or the force conditional block killer
975 #ifdef FEATURE_FORCE_LOAD
976 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
977 #else /* ifndef FEATURE_FORCE_LOAD */
978 err = map_block_killer(exports, "force-support");
979 #endif /* ndef FEATURE_FORCE_LOAD */
984 free_http_response(rsp);
985 return cgi_error_memory();
991 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
994 free_http_response(rsp);
995 return cgi_error_memory();
998 return finish_http_response(rsp);
1000 #endif /* def FEATURE_TRUST */
1003 #ifdef FEATURE_FAST_REDIRECTS
1004 /*********************************************************************
1006 * Function : redirect_url
1008 * Description : Checks for redirection URLs and returns a HTTP redirect
1009 * to the destination URL, if necessary
1012 * 1 : csp = Current client state (buffers, headers, etc...)
1014 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1016 *********************************************************************/
1017 struct http_response *redirect_url(struct client_state *csp)
1020 struct http_response *rsp;
1022 p = q = csp->http->path;
1023 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1026 * find the last URL encoded in the request
1028 while ((p = strstr(p, "http://")) != NULL)
1034 * if there was any, generate and return a HTTP redirect
1036 if (q != csp->http->path)
1038 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1040 if (NULL == (rsp = alloc_http_response()))
1042 return cgi_error_memory();
1045 if ( enlist_unique_header(rsp->headers, "Location", q)
1046 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1048 free_http_response(rsp);
1049 return cgi_error_memory();
1052 return finish_http_response(rsp);
1060 #endif /* def FEATURE_FAST_REDIRECTS */
1063 #ifdef FEATURE_IMAGE_BLOCKING
1064 /*********************************************************************
1066 * Function : is_imageurl
1068 * Description : Given a URL, decide whether it is an image or not,
1069 * using either the info from a previous +image action
1070 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1071 * the browser's accept header.
1074 * 1 : csp = Current client state (buffers, headers, etc...)
1076 * Returns : True (nonzero) if URL is an image, false (0)
1079 *********************************************************************/
1080 int is_imageurl(struct client_state *csp)
1082 #ifdef FEATURE_IMAGE_DETECT_MSIE
1085 tmp = get_header_value(csp->headers, "User-Agent:");
1086 if (tmp && strstr(tmp, "MSIE"))
1088 tmp = get_header_value(csp->headers, "Accept:");
1089 if (tmp && strstr(tmp, "image/gif"))
1091 /* Client will accept HTML. If this seems counterintuitive,
1101 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1103 return ((csp->action->flags & ACTION_IMAGE) != 0);
1106 #endif /* def FEATURE_IMAGE_BLOCKING */
1109 #ifdef FEATURE_TRUST
1110 /*********************************************************************
1112 * Function : is_untrusted_url
1114 * Description : Should we "distrust" this URL (and block it)?
1116 * Yes if it matches a line in the trustfile, or if the
1117 * referrer matches a line starting with "+" in the
1122 * 1 : csp = Current client state (buffers, headers, etc...)
1124 * Returns : 0 => trusted, 1 => untrusted
1126 *********************************************************************/
1127 int is_untrusted_url(struct client_state *csp)
1129 struct file_list *fl;
1130 struct block_spec *b;
1131 struct url_spec **trusted_url;
1132 struct http_request rhttp[1];
1133 const char * referer;
1137 * If we don't have a trustlist, we trust everybody
1139 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1144 memset(rhttp, '\0', sizeof(*rhttp));
1147 * Do we trust the request URL itself?
1149 for (b = b->next; b ; b = b->next)
1151 if (url_match(b->url, csp->http))
1157 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1159 /* no referrer was supplied */
1165 * If not, do we maybe trust its referrer?
1167 err = parse_http_url(referer, rhttp, csp);
1173 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1175 if (url_match(*trusted_url, rhttp))
1177 /* if the URL's referrer is from a trusted referrer, then
1178 * add the target spec to the trustfile as an unblocked
1179 * domain and return NULL (which means it's OK).
1184 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1188 char * new_entry = strdup("~");
1190 string_append(&new_entry, csp->http->hostport);
1192 path = csp->http->path;
1193 if ( (path[0] == '/')
1195 && ((path_end = strchr(path + 2, '/')) != NULL))
1197 /* since this path points into a user's home space
1198 * be sure to include this spec in the trustfile.
1200 int path_len = path_end - path; /* save offset */
1201 path = strdup(path); /* Copy string */
1204 path_end = path + path_len; /* regenerate ptr to new buffer */
1205 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1207 string_join(&new_entry, path);
1210 if (new_entry != NULL)
1212 fprintf(fp, "%s\n", new_entry);
1217 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1218 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1228 #endif /* def FEATURE_TRUST */
1231 /*********************************************************************
1233 * Function : pcrs_filter_response
1235 * Description : Ecexute all text substitutions from all applying
1236 * +filter actions on the text buffer that's been accumulated
1237 * in csp->iob->buf. If this changes the contents, set
1238 * csp->content_length to the modified size and raise the
1239 * CSP_FLAG_MODIFIED flag.
1242 * 1 : csp = Current client state (buffers, headers, etc...)
1244 * Returns : a pointer to the (newly allocated) modified buffer.
1245 * or NULL if there were no hits or something went wrong
1247 *********************************************************************/
1248 char *pcrs_filter_response(struct client_state *csp)
1253 char *old = csp->iob->cur, *new = NULL;
1256 struct file_list *fl;
1257 struct re_filterfile_spec *b;
1258 struct list_entry *filtername;
1263 if (csp->iob->cur >= csp->iob->eod)
1267 size = csp->iob->eod - csp->iob->cur;
1269 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1271 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1276 * If the body has a "chunked" transfer-encoding,
1277 * get rid of it first, adjusting size and iob->eod
1279 if (csp->flags & CSP_FLAG_CHUNKED)
1281 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1282 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1286 csp->iob->eod = csp->iob->cur + size;
1287 csp->flags |= CSP_FLAG_MODIFIED;
1291 * For all applying +filter actions, look if a filter by that
1292 * name exists and if yes, execute it's pcrs_joblist on the
1295 for (b = fl->f; b; b = b->next)
1297 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1298 filtername ; filtername = filtername->next)
1300 if (strcmp(b->name, filtername->str) == 0)
1302 int current_hits = 0;
1304 if ( NULL == b->joblist )
1306 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1310 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1311 csp->http->hostport, csp->http->path, size, b->name);
1313 /* Apply all jobs from the joblist */
1314 for (job = b->joblist; NULL != job; job = job->next)
1316 current_hits += pcrs_execute(job, old, size, &new, &size);
1317 if (old != csp->iob->cur) free(old);
1321 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1322 hits += current_hits;
1328 * If there were no hits, destroy our copy and let
1329 * chat() use the original in csp->iob
1337 csp->flags |= CSP_FLAG_MODIFIED;
1338 csp->content_length = size;
1346 /*********************************************************************
1348 * Function : gif_deanimate_response
1350 * Description : Deanimate the GIF image that has been accumulated in
1351 * csp->iob->buf, set csp->content_length to the modified
1352 * size and raise the CSP_FLAG_MODIFIED flag.
1355 * 1 : csp = Current client state (buffers, headers, etc...)
1357 * Returns : a pointer to the (newly allocated) modified buffer.
1358 * or NULL in case something went wrong.
1360 *********************************************************************/
1361 char *gif_deanimate_response(struct client_state *csp)
1363 struct binbuffer *in, *out;
1365 size_t size = csp->iob->eod - csp->iob->cur;
1368 * If the body has a "chunked" transfer-encoding,
1369 * get rid of it first, adjusting size and iob->eod
1371 if (csp->flags & CSP_FLAG_CHUNKED)
1373 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1374 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1378 csp->iob->eod = csp->iob->cur + size;
1379 csp->flags |= CSP_FLAG_MODIFIED;
1382 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1383 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1385 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1389 in->buffer = csp->iob->cur;
1392 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1394 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1401 if ((int)size == out->offset)
1403 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1407 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1409 csp->content_length = out->offset;
1410 csp->flags |= CSP_FLAG_MODIFIED;
1420 /*********************************************************************
1422 * Function : remove_chunked_transfer_coding
1424 * Description : In-situ remove the "chunked" transfer coding as defined
1425 * in rfc2616 from a buffer.
1428 * 1 : buffer = Pointer to the text buffer
1429 * 2 : size = Number of bytes to be processed
1431 * Returns : The new size, i.e. the number of bytes from buffer which
1432 * are occupied by the stripped body, or 0 in case something
1435 *********************************************************************/
1436 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1439 unsigned int chunksize = 0;
1440 char *from_p, *to_p;
1443 from_p = to_p = buffer;
1445 if (sscanf(buffer, "%x", &chunksize) != 1)
1447 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1451 while (chunksize > 0)
1453 if (NULL == (from_p = strstr(from_p, "\r\n")))
1455 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1458 newsize += chunksize;
1461 memmove(to_p, from_p, (size_t) chunksize);
1462 to_p = buffer + newsize;
1463 from_p += chunksize + 2;
1465 if (sscanf(from_p, "%x", &chunksize) != 1)
1467 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1472 /* FIXME: Should this get its own loglevel? */
1473 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1479 /*********************************************************************
1481 * Function : url_actions
1483 * Description : Gets the actions for this URL.
1486 * 1 : http = http_request request for blocked URLs
1487 * 2 : csp = Current client state (buffers, headers, etc...)
1491 *********************************************************************/
1492 void url_actions(struct http_request *http,
1493 struct client_state *csp)
1495 struct file_list *fl;
1496 struct url_actions *b;
1498 init_current_action(csp->action);
1500 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1505 apply_url_actions(csp->action, http, b);
1510 /*********************************************************************
1512 * Function : apply_url_actions
1514 * Description : Applies a list of URL actions.
1517 * 1 : action = Destination.
1518 * 2 : http = Current URL
1519 * 3 : b = list of URL actions to apply
1523 *********************************************************************/
1524 void apply_url_actions(struct current_action_spec *action,
1525 struct http_request *http,
1526 struct url_actions *b)
1530 /* Should never happen */
1534 for (b = b->next; NULL != b; b = b->next)
1536 if (url_match(b->url, http))
1538 merge_current_action(action, b->action);
1544 /*********************************************************************
1546 * Function : forward_url
1548 * Description : Should we forward this to another proxy?
1551 * 1 : http = http_request request for current URL
1552 * 2 : csp = Current client state (buffers, headers, etc...)
1554 * Returns : Pointer to forwarding information.
1556 *********************************************************************/
1557 const struct forward_spec * forward_url(struct http_request *http,
1558 struct client_state *csp)
1560 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1561 struct forward_spec *fwd = csp->config->forward;
1570 if (url_match(fwd->url, http))