1 const char filters_rcs[] = "$Id: filters.c,v 1.52 2002/03/24 16:35:57 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://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 1.52 2002/03/24 16:35:57 jongfoster
44 * Revision 1.51 2002/03/24 15:23:33 jongfoster
47 * Revision 1.50 2002/03/24 13:25:43 swa
48 * name change related issues
50 * Revision 1.49 2002/03/16 20:29:14 oes
53 * Revision 1.48 2002/03/13 20:25:34 oes
54 * Better logging for content filters
56 * Revision 1.47 2002/03/13 00:30:52 jongfoster
58 * Added option of always sending redirect for imageblock,
59 * currently disabled with #if 0.
61 * Revision 1.46 2002/03/12 01:42:49 oes
62 * Introduced modular filters
64 * Revision 1.45 2002/03/08 16:47:50 oes
65 * Added choice beween GIF and PNG built-in images
67 * Revision 1.44 2002/03/07 03:49:31 oes
68 * - Fixed compiler warnings etc
69 * - Changed built-in images from GIF to PNG
70 * (with regard to Unisys patent issue)
71 * - Added a 4x4 pattern PNG which is less intrusive
72 * than the logo but also clearly marks the deleted banners
74 * Revision 1.43 2002/01/22 23:51:59 jongfoster
75 * Replacing strsav() with the safer string_append().
77 * Adding missing html_encode() to error message generators. Where encoded
78 * and unencoded versions of a string were provided, removing the unencoded
81 * Revision 1.42 2002/01/17 21:00:32 jongfoster
82 * Moving all our URL and URL pattern parsing code to urlmatch.c.
84 * Using a single, simple url_match(pattern,url) function - rather than
85 * the 3-line match routine which was repeated all over the place.
87 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
89 * Using parse_http_url() to parse URLs without faking a HTTP
90 * request line for parse_http_request().
92 * Revision 1.41 2001/11/13 00:14:07 jongfoster
93 * Fixing stupid bug now I've figured out what || means.
94 * (It always returns 0 or 1, not one of it's paramaters.)
96 * Revision 1.40 2001/10/26 17:37:55 oes
97 * - Re-enabled Netscape 200/404 bug workaround in block_url():
98 * - Removed OS/2 special case
99 * - Made block_url() independant from sed() having been run
100 * - Made trust_url independant from sed() having been run
101 * - Made is_imageurl independant from sed() having been run.
102 * It now checks User-Agent: and Accept: by itself.
105 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
106 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
107 * threads to call select() simultaneously. So, it's time to do a real, live,
108 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
109 * (native). Both versions will work, but using __OS2__ offers multi-threading.
111 * Revision 1.38 2001/10/23 21:32:33 jongfoster
112 * Adding error-checking to selected functions
114 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
115 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
116 * filters.c. Added a FIXME in front of the offending code. I'll gladly
117 * put in a better/more robust fix for all parties if one is presented...
118 * It seems that just returning 200 instead of 404 would pretty much fix
119 * it for everyone, but I don't know all the history of the problem.
121 * Revision 1.36 2001/10/10 16:44:16 oes
122 * Added match_portlist function
124 * Revision 1.35 2001/10/07 15:41:23 oes
125 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
127 * New function remove_chunked_transfer_coding that strips chunked
128 * transfer coding to plain and is called by pcrs_filter_response
129 * and gif_deanimate_response if neccessary
131 * Improved handling of zero-change re_filter runs
133 * pcrs_filter_response and gif_deanimate_response now remove
134 * chunked transfer codeing before processing the body.
136 * Revision 1.34 2001/09/20 15:49:36 steudten
138 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
139 * See cgi.c fill_template().
141 * Revision 1.33 2001/09/16 17:05:14 jongfoster
142 * Removing unused #include showarg.h
144 * Revision 1.32 2001/09/16 13:21:27 jongfoster
145 * Changes to use new list functions.
147 * Revision 1.31 2001/09/16 11:38:02 jongfoster
148 * Splitting fill_template() into 2 functions:
149 * template_load() loads the file
150 * template_fill() performs the PCRS regexps.
151 * This is because the CGI edit interface has a "table row"
152 * template which is used many times in the page - this
153 * change means it's only loaded from disk once.
155 * Revision 1.30 2001/09/16 11:00:10 jongfoster
156 * New function alloc_http_response, for symmetry with free_http_response
158 * Revision 1.29 2001/09/13 23:32:40 jongfoster
159 * Moving image data to cgi.c rather than cgi.h
160 * Fixing a GPF under Win32 (and any other OS that protects global
161 * constants from being written to).
163 * Revision 1.28 2001/09/10 10:18:51 oes
164 * Silenced compiler warnings
166 * Revision 1.27 2001/08/05 16:06:20 jongfoster
167 * Modifiying "struct map" so that there are now separate header and
168 * "map_entry" structures. This means that functions which modify a
169 * map no longer need to return a pointer to the modified map.
170 * Also, it no longer reverses the order of the entries (which may be
171 * important with some advanced template substitutions).
173 * Revision 1.26 2001/07/30 22:08:36 jongfoster
174 * Tidying up #defines:
175 * - All feature #defines are now of the form FEATURE_xxx
176 * - Permanently turned off WIN_GUI_EDIT
177 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
179 * Revision 1.25 2001/07/26 10:09:46 oes
180 * Made browser detection a little less naive
182 * Revision 1.24 2001/07/25 17:22:51 oes
183 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
185 * Revision 1.23 2001/07/23 13:40:12 oes
186 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
188 * Revision 1.22 2001/07/18 12:29:34 oes
189 * - Made gif_deanimate_response respect
190 * csp->action->string[ACTION_STRING_DEANIMATE]
191 * - Logging cosmetics
193 * Revision 1.21 2001/07/13 13:59:53 oes
194 * - Introduced gif_deanimate_response which shares the
195 * generic content modification interface of pcrs_filter_response
196 * and acts as a wrapper to deanimate.c:gif_deanimate()
197 * - Renamed re_process_buffer to pcrs_filter_response
198 * - pcrs_filter_response now returns NULL on failiure
199 * - Removed all #ifdef PCRS
201 * Revision 1.20 2001/07/01 17:01:04 oes
202 * Added comments and missing return statement in is_untrusted_url()
204 * Revision 1.19 2001/06/29 21:45:41 oes
205 * Indentation, CRLF->LF, Tab-> Space
207 * Revision 1.18 2001/06/29 13:27:38 oes
208 * - Cleaned up, renamed and reorderd functions
209 * and improved comments
212 * - Ported to CGI platform. Now delivers
213 * http_response or NULL
214 * - Unified HTML and GIF generation (moved image detection
215 * and GIF generation here from jcc.c:chat())
216 * - Fixed HTTP status to:
217 * - 403 (Forbidden) for the "blocked" HTML message
218 * - 200 (OK) for GIF answers
219 * - 302 (Redirect) for redirect to GIF
222 * - Ported to CGI platform. Now delivers
223 * http_response or NULL
224 * - Separated detection of untrusted URL into
225 * (bool)is_untrusted_url
226 * - Added enforcement of untrusted requests
228 * - Moved redirect_url() from cgi.c to here
229 * and ported it to the CGI platform
231 * - Removed logentry from cancelled commit
233 * Revision 1.17 2001/06/09 10:55:28 jongfoster
234 * Changing BUFSIZ ==> BUFFER_SIZE
236 * Revision 1.16 2001/06/07 23:10:26 jongfoster
237 * Allowing unanchored domain patterns to back off and retry
238 * if they partially match. Optimized right-anchored patterns.
239 * Moving ACL and forward files into config file.
240 * Replacing struct gateway with struct forward_spec
242 * Revision 1.15 2001/06/03 19:12:00 oes
243 * extracted-CGI relevant stuff
245 * Revision 1.14 2001/06/01 10:30:55 oes
246 * Added optional left-anchoring to domaincmp
248 * Revision 1.13 2001/05/31 21:21:30 jongfoster
249 * Permissionsfile / actions file changes:
250 * - Changed "permission" to "action" throughout
251 * - changes to file format to allow string parameters
252 * - Moved helper functions to actions.c
254 * Revision 1.12 2001/05/31 17:35:20 oes
256 * - Enhanced domain part globbing with infix and prefix asterisk
257 * matching and optional unanchored operation
259 * Revision 1.11 2001/05/29 11:53:23 oes
260 * "See why" link added to "blocked" page
262 * Revision 1.10 2001/05/29 09:50:24 jongfoster
263 * Unified blocklist/imagelist/permissionslist.
264 * File format is still under discussion, but the internal changes
267 * Also modified interceptor behaviour:
268 * - We now intercept all URLs beginning with one of the following
269 * prefixes (and *only* these prefixes):
271 * * http://ijbswa.sf.net/config/
272 * * http://ijbswa.sourceforge.net/config/
273 * - New interceptors "home page" - go to http://i.j.b/ to see it.
274 * - Internal changes so that intercepted and fast redirect pages
275 * are not replaced with an image.
276 * - Interceptors now have the option to send a binary page direct
277 * to the client. (i.e. ijb-send-banner uses this)
278 * - Implemented show-url-info interceptor. (Which is why I needed
279 * the above interceptors changes - a typical URL is
280 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
281 * The previous mechanism would not have intercepted that, and
282 * if it had been intercepted then it then it would have replaced
285 * Revision 1.9 2001/05/27 22:17:04 oes
287 * - re_process_buffer no longer writes the modified buffer
288 * to the client, which was very ugly. It now returns the
289 * buffer, which it is then written by chat.
291 * - content_length now adjusts the Content-Length: header
292 * for modified documents rather than crunch()ing it.
293 * (Length info in csp->content_length, which is 0 for
294 * unmodified documents)
296 * - For this to work, sed() is called twice when filtering.
298 * Revision 1.8 2001/05/26 17:13:28 jongfoster
299 * Filled in a function comment.
301 * Revision 1.7 2001/05/26 15:26:15 jongfoster
302 * ACL feature now provides more security by immediately dropping
303 * connections from untrusted hosts.
305 * Revision 1.6 2001/05/26 00:28:36 jongfoster
306 * Automatic reloading of config file.
307 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
308 * Most of the global variables have been moved to a new
309 * struct configuration_spec, accessed through csp->config->globalname
310 * Most of the globals remaining are used by the Win32 GUI.
312 * Revision 1.5 2001/05/25 22:34:30 jongfoster
315 * Revision 1.4 2001/05/22 18:46:04 oes
317 * - Enabled filtering banners by size rather than URL
318 * by adding patterns that replace all standard banner
319 * sizes with the "Junkbuster" gif to the re_filterfile
321 * - Enabled filtering WebBugs by providing a pattern
322 * which kills all 1x1 images
324 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
325 * which is selected by the (nonstandard and therefore
326 * capital) letter 'U' in the option string.
327 * It causes the quantifiers to be ungreedy by default.
328 * Appending a ? turns back to greedy (!).
330 * - Added a new interceptor ijb-send-banner, which
331 * sends back the "Junkbuster" gif. Without imagelist or
332 * MSIE detection support, or if tinygif = 1, or the
333 * URL isn't recognized as an imageurl, a lame HTML
334 * explanation is sent instead.
336 * - Added new feature, which permits blocking remote
337 * script redirects and firing back a local redirect
339 * The feature is conditionally compiled, i.e. it
340 * can be disabled with --disable-fast-redirects,
341 * plus it must be activated by a "fast-redirects"
342 * line in the config file, has its own log level
343 * and of course wants to be displayed by show-proxy-args
344 * Note: Boy, all the #ifdefs in 1001 locations and
345 * all the fumbling with configure.in and acconfig.h
346 * were *way* more work than the feature itself :-(
348 * - Because a generic redirect template was needed for
349 * this, tinygif = 3 now uses the same.
351 * - Moved GIFs, and other static HTTP response templates
356 * - Removed some >400 CRs again (Jon, you really worked
359 * Revision 1.3 2001/05/20 16:44:47 jongfoster
360 * Removing last hardcoded JunkBusters.com URLs.
362 * Revision 1.2 2001/05/20 01:21:20 jongfoster
363 * Version 2.9.4 checkin.
364 * - Merged popupfile and cookiefile, and added control over PCRS
365 * filtering, in new "permissionsfile".
366 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
367 * file error you now get a message box (in the Win32 GUI) rather
368 * than the program exiting with no explanation.
369 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
371 * - Removed tabs from "config"
372 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
373 * - Bumped up version number.
375 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
376 * Initial import of version 2.9.3 source tree
379 *********************************************************************/
385 #include <sys/types.h>
394 #endif /* ndef __OS2__ */
395 #include <netinet/in.h>
397 #include <winsock2.h>
398 #endif /* ndef _WIN32 */
402 #endif /* def __OS2__ */
410 #include "jbsockets.h"
411 #include "miscutil.h"
415 #include "deanimate.h"
416 #include "urlmatch.h"
422 const char filters_h_rcs[] = FILTERS_H_VERSION;
424 /* Fix a problem with Solaris. There should be no effect on other
426 * Solaris's isspace() is a macro which uses it's argument directly
427 * as an array index. Therefore we need to make sure that high-bit
428 * characters generate +ve values, and ideally we also want to make
429 * the argument match the declared parameter type of "int".
431 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
435 /*********************************************************************
437 * Function : block_acl
439 * Description : Block this request?
440 * Decide yes or no based on ACL file.
443 * 1 : dst = The proxy or gateway address this is going to.
444 * Or NULL to check all possible targets.
445 * 2 : csp = Current client state (buffers, headers, etc...)
446 * Also includes the client IP address.
448 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
450 *********************************************************************/
451 int block_acl(struct access_control_addr *dst, struct client_state *csp)
453 struct access_control_list *acl = csp->config->acl;
455 /* if not using an access control list, then permit the connection */
461 /* search the list */
464 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
468 /* Just want to check if they have any access */
469 if (acl->action == ACL_PERMIT)
474 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
475 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
477 if (acl->action == ACL_PERMIT)
495 /*********************************************************************
497 * Function : acl_addr
499 * Description : Called from `load_config' to parse an ACL address.
502 * 1 : aspec = String specifying ACL address.
503 * 2 : aca = struct access_control_addr to fill in.
505 * Returns : 0 => Ok, everything else is an error.
507 *********************************************************************/
508 int acl_addr(char *aspec, struct access_control_addr *aca)
510 int i, masklength, port;
516 if ((p = strchr(aspec, '/')) != NULL)
520 if (ijb_isdigit(*p) == 0)
524 masklength = atoi(p);
527 if ((masklength < 0) || (masklength > 32))
532 if ((p = strchr(aspec, ':')) != NULL)
536 if (ijb_isdigit(*p) == 0)
545 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
547 if (aca->addr == INADDR_NONE)
552 /* build the netmask */
554 for (i=1; i <= masklength ; i++)
556 aca->mask |= (1 << (32 - i));
559 /* now mask off the host portion of the ip address
560 * (i.e. save on the network portion of the address).
562 aca->addr = aca->addr & aca->mask;
567 #endif /* def FEATURE_ACL */
570 /*********************************************************************
572 * Function : match_portlist
574 * Description : Check if a given number is covered by a comma
575 * separated list of numbers and ranges (a,b-c,d,..)
578 * 1 : portlist = String with list
579 * 2 : port = port to check
581 * Returns : 0 => no match
584 *********************************************************************/
585 int match_portlist(const char *portlist, int port)
587 char *min, *max, *next, *portlist_copy;
589 min = next = portlist_copy = strdup(portlist);
592 * Zero-terminate first item and remember offset for next
594 if (NULL != (next = strchr(portlist_copy, (int) ',')))
600 * Loop through all items, checking for match
604 if (NULL == (max = strchr(min, (int) '-')))
607 * No dash, check for equality
609 if (port == atoi(min))
618 * This is a range, so check if between min and max,
619 * or, if max was omitted, between min and 65K
622 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
636 * Zero-terminate next item and remember offset for n+1
638 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
650 /*********************************************************************
652 * Function : block_url
654 * Description : Called from `chat'. Check to see if we need to block this.
657 * 1 : csp = Current client state (buffers, headers, etc...)
659 * Returns : NULL => unblocked, else HTTP block response
661 *********************************************************************/
662 struct http_response *block_url(struct client_state *csp)
664 #ifdef FEATURE_IMAGE_BLOCKING
666 #endif /* def FEATURE_IMAGE_BLOCKING */
667 struct http_response *rsp;
670 * If it's not blocked, don't block it ;-)
672 if ((csp->action->flags & ACTION_BLOCK) == 0)
678 * Else, prepare a response
680 if (NULL == (rsp = alloc_http_response()))
682 return cgi_error_memory();
686 * If it's an image-url, send back an image or redirect
687 * as specified by the relevant +image action
689 #ifdef FEATURE_IMAGE_BLOCKING
690 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
693 /* determine HOW images should be blocked */
694 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
696 #if 1 /* Two alternative strategies, use this one for now: */
698 /* and handle accordingly: */
699 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
701 rsp->body = bindup(image_pattern_data, image_pattern_length);
702 if (rsp->body == NULL)
704 free_http_response(rsp);
705 return cgi_error_memory();
707 rsp->content_length = image_pattern_length;
709 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
711 free_http_response(rsp);
712 return cgi_error_memory();
716 else if (0 == strcmpic(p, "blank"))
718 rsp->body = bindup(image_blank_data, image_blank_length);
719 if (rsp->body == NULL)
721 free_http_response(rsp);
722 return cgi_error_memory();
724 rsp->content_length = image_blank_length;
726 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
728 free_http_response(rsp);
729 return cgi_error_memory();
735 rsp->status = strdup("302 Local Redirect from Privoxy");
736 if (rsp->status == NULL)
738 free_http_response(rsp);
739 return cgi_error_memory();
742 if (enlist_unique_header(rsp->headers, "Location", p))
744 free_http_response(rsp);
745 return cgi_error_memory();
749 #else /* Following code is disabled for now */
751 /* and handle accordingly: */
752 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
754 p = CGI_PREFIX "send-banner?type=pattern";
756 else if (0 == strcmpic(p, "blank"))
758 p = CGI_PREFIX "send-banner?type=blank";
760 rsp->status = strdup("302 Local Redirect from Privoxy");
761 if (rsp->status == NULL)
763 free_http_response(rsp);
764 return cgi_error_memory();
767 if (enlist_unique_header(rsp->headers, "Location", p))
769 free_http_response(rsp);
770 return cgi_error_memory();
772 #endif /* Preceeding code is disabled for now */
775 #endif /* def FEATURE_IMAGE_BLOCKING */
778 * Else, generate an HTML "blocked" message:
782 struct map * exports;
785 * Workaround for stupid Netscape bug which prevents
786 * pages from being displayed if loading a referenced
787 * JavaScript or style sheet fails. So make it appear
788 * as if it succeeded.
790 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
791 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
792 && !strstr(p, "Gecko") /* save Mozilla, */
793 && !strstr(p, "compatible") /* MSIE */
794 && !strstr(p, "Opera")) /* and Opera. */
796 rsp->status = strdup("200 Request for blocked URL");
800 rsp->status = strdup("404 Request for blocked URL");
803 if (rsp->status == NULL)
805 free_http_response(rsp);
806 return cgi_error_memory();
809 exports = default_exports(csp, NULL);
812 free_http_response(rsp);
813 return cgi_error_memory();
816 #ifdef FEATURE_FORCE_LOAD
817 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
818 #else /* ifndef FEATURE_FORCE_LOAD */
819 err = map_block_killer(exports, "force-support");
820 #endif /* ndef FEATURE_FORCE_LOAD */
822 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
823 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
828 free_http_response(rsp);
829 return cgi_error_memory();
832 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
835 free_http_response(rsp);
836 return cgi_error_memory();
840 return finish_http_response(rsp);
846 /*********************************************************************
848 * Function : trust_url FIXME: I should be called distrust_url
850 * Description : Calls is_untrusted_url to determine if the URL is trusted
851 * and if not, returns a HTTP 304 response with a reject message.
854 * 1 : csp = Current client state (buffers, headers, etc...)
856 * Returns : NULL => trusted, else http_response.
858 *********************************************************************/
859 struct http_response *trust_url(struct client_state *csp)
861 struct http_response *rsp;
862 struct map * exports;
863 char buf[BUFFER_SIZE];
865 struct url_spec **tl;
870 * Don't bother to work on trusted URLs
872 if (!is_untrusted_url(csp))
878 * Else, prepare a response:
880 if (NULL == (rsp = alloc_http_response()))
882 return cgi_error_memory();
885 exports = default_exports(csp, NULL);
888 free_http_response(rsp);
889 return cgi_error_memory();
893 * Export the host, port, and referrer information
895 err = map(exports, "hostport", 1, csp->http->hostport, 1);
896 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
898 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
900 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
904 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
910 free_http_response(rsp);
911 return cgi_error_memory();
915 * Export the trust list
918 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
920 sprintf(buf, "<li>%s</li>\n", t->spec);
921 string_append(&p, buf);
923 err = map(exports, "trusted-referrers", 1, p, 0);
928 free_http_response(rsp);
929 return cgi_error_memory();
933 * Export the trust info, if available
935 if (csp->config->trust_info->first)
937 struct list_entry *l;
940 for (l = csp->config->trust_info->first; l ; l = l->next)
942 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
943 string_append(&p, buf);
945 err = map(exports, "trust-info", 1, p, 0);
949 err = map_block_killer(exports, "have-trust-info");
955 free_http_response(rsp);
956 return cgi_error_memory();
960 * Export the force prefix or the force conditional block killer
962 #ifdef FEATURE_FORCE_LOAD
963 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
964 #else /* ifndef FEATURE_FORCE_LOAD */
965 err = map_block_killer(exports, "force-support");
966 #endif /* ndef FEATURE_FORCE_LOAD */
971 free_http_response(rsp);
972 return cgi_error_memory();
978 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
981 free_http_response(rsp);
982 return cgi_error_memory();
985 return finish_http_response(rsp);
987 #endif /* def FEATURE_TRUST */
990 #ifdef FEATURE_FAST_REDIRECTS
991 /*********************************************************************
993 * Function : redirect_url
995 * Description : Checks for redirection URLs and returns a HTTP redirect
996 * to the destination URL, if necessary
999 * 1 : csp = Current client state (buffers, headers, etc...)
1001 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1003 *********************************************************************/
1004 struct http_response *redirect_url(struct client_state *csp)
1007 struct http_response *rsp;
1009 p = q = csp->http->path;
1010 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1013 * find the last URL encoded in the request
1015 while ((p = strstr(p, "http://")) != NULL)
1021 * if there was any, generate and return a HTTP redirect
1023 if (q != csp->http->path)
1025 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1027 if (NULL == (rsp = alloc_http_response()))
1029 return cgi_error_memory();
1032 if ( enlist_unique_header(rsp->headers, "Location", q)
1033 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1035 free_http_response(rsp);
1036 return cgi_error_memory();
1039 return finish_http_response(rsp);
1047 #endif /* def FEATURE_FAST_REDIRECTS */
1050 #ifdef FEATURE_IMAGE_BLOCKING
1051 /*********************************************************************
1053 * Function : is_imageurl
1055 * Description : Given a URL, decide whether it is an image or not,
1056 * using either the info from a previous +image action
1057 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1058 * the browser's accept header.
1061 * 1 : csp = Current client state (buffers, headers, etc...)
1063 * Returns : True (nonzero) if URL is an image, false (0)
1066 *********************************************************************/
1067 int is_imageurl(struct client_state *csp)
1069 #ifdef FEATURE_IMAGE_DETECT_MSIE
1072 tmp = get_header_value(csp->headers, "User-Agent:");
1073 if (tmp && strstr(tmp, "MSIE"))
1075 tmp = get_header_value(csp->headers, "Accept:");
1076 if (tmp && strstr(tmp, "image/gif"))
1078 /* Client will accept HTML. If this seems counterintuitive,
1088 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1090 return ((csp->action->flags & ACTION_IMAGE) != 0);
1093 #endif /* def FEATURE_IMAGE_BLOCKING */
1096 #ifdef FEATURE_COOKIE_JAR
1097 /*********************************************************************
1099 * Function : is_untrusted_url
1101 * Description : Should we "distrust" this URL (and block it)?
1103 * Yes if it matches a line in the trustfile, or if the
1104 * referrer matches a line starting with "+" in the
1109 * 1 : csp = Current client state (buffers, headers, etc...)
1111 * Returns : 0 => trusted, 1 => untrusted
1113 *********************************************************************/
1114 int is_untrusted_url(struct client_state *csp)
1116 struct file_list *fl;
1117 struct block_spec *b;
1118 struct url_spec **trusted_url;
1119 struct http_request rhttp[1];
1120 const char * referer;
1124 * If we don't have a trustlist, we trust everybody
1126 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1131 memset(rhttp, '\0', sizeof(*rhttp));
1134 * Do we trust the request URL itself?
1136 for (b = b->next; b ; b = b->next)
1138 if (url_match(b->url, csp->http))
1144 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1146 /* no referrer was supplied */
1151 * If not, do we maybe trust its referrer?
1156 * Parse the URL from the referrer
1159 err = parse_http_url(referer, rhttp, csp);
1165 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1167 if (url_match(*trusted_url, rhttp))
1169 /* if the URL's referrer is from a trusted referrer, then
1170 * add the target spec to the trustfile as an unblocked
1171 * domain and return NULL (which means it's OK).
1176 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1180 char * new_entry = strdup("~");
1182 string_append(&new_entry, csp->http->hostport);
1184 path = csp->http->path;
1185 if ( (path[0] == '/')
1187 && ((path_end = strchr(path + 2, '/')) != NULL))
1189 /* since this path points into a user's home space
1190 * be sure to include this spec in the trustfile.
1192 int path_len = path_end - path; /* save offset */
1193 path = strdup(path); /* Copy string */
1196 path_end = path + path_len; /* regenerate ptr to new buffer */
1197 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1199 string_join(&new_entry, path);
1202 if (new_entry != NULL)
1204 fprintf(fp, "%s\n", new_entry);
1209 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1210 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1220 #endif /* def FEATURE_COOKIE_JAR */
1223 /*********************************************************************
1225 * Function : pcrs_filter_response
1227 * Description : Ecexute all text substitutions from all applying
1228 * +filter actions on the text buffer that's been accumulated
1229 * in csp->iob->buf. If this changes the contents, set
1230 * csp->content_length to the modified size and raise the
1231 * CSP_FLAG_MODIFIED flag.
1234 * 1 : csp = Current client state (buffers, headers, etc...)
1236 * Returns : a pointer to the (newly allocated) modified buffer.
1237 * or NULL if there were no hits or something went wrong
1239 *********************************************************************/
1240 char *pcrs_filter_response(struct client_state *csp)
1245 char *old = csp->iob->cur, *new = NULL;
1248 struct file_list *fl;
1249 struct re_filterfile_spec *b;
1250 struct list_entry *filtername;
1255 if (csp->iob->cur >= csp->iob->eod)
1259 size = csp->iob->eod - csp->iob->cur;
1261 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1263 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1268 * If the body has a "chunked" transfer-encoding,
1269 * get rid of it first, adjusting size and iob->eod
1271 if (csp->flags & CSP_FLAG_CHUNKED)
1273 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1274 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1278 csp->iob->eod = csp->iob->cur + size;
1279 csp->flags |= CSP_FLAG_MODIFIED;
1283 * For all applying +filter actions, look if a filter by that
1284 * name exists and if yes, execute it's pcrs_joblist on the
1287 for (b = fl->f; b; b = b->next)
1289 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1290 filtername ; filtername = filtername->next)
1292 if (strcmp(b->name, filtername->str) == 0)
1294 int current_hits = 0;
1296 if ( NULL == b->joblist )
1298 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1302 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1303 csp->http->hostport, csp->http->path, size, b->name);
1305 /* Apply all jobs from the joblist */
1306 for (job = b->joblist; NULL != job; job = job->next)
1308 current_hits += pcrs_execute(job, old, size, &new, &size);
1309 if (old != csp->iob->cur) free(old);
1313 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1314 hits += current_hits;
1320 * If there were no hits, destroy our copy and let
1321 * chat() use the original in csp->iob
1329 csp->flags |= CSP_FLAG_MODIFIED;
1330 csp->content_length = size;
1338 /*********************************************************************
1340 * Function : gif_deanimate_response
1342 * Description : Deanimate the GIF image that has been accumulated in
1343 * csp->iob->buf, set csp->content_length to the modified
1344 * size and raise the CSP_FLAG_MODIFIED flag.
1347 * 1 : csp = Current client state (buffers, headers, etc...)
1349 * Returns : a pointer to the (newly allocated) modified buffer.
1350 * or NULL in case something went wrong.
1352 *********************************************************************/
1353 char *gif_deanimate_response(struct client_state *csp)
1355 struct binbuffer *in, *out;
1357 size_t size = csp->iob->eod - csp->iob->cur;
1360 * If the body has a "chunked" transfer-encoding,
1361 * get rid of it first, adjusting size and iob->eod
1363 if (csp->flags & CSP_FLAG_CHUNKED)
1365 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1366 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1370 csp->iob->eod = csp->iob->cur + size;
1371 csp->flags |= CSP_FLAG_MODIFIED;
1374 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1375 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1377 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1381 in->buffer = csp->iob->cur;
1384 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1386 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1393 if ((int)size == out->offset)
1395 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1399 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1401 csp->content_length = out->offset;
1402 csp->flags |= CSP_FLAG_MODIFIED;
1412 /*********************************************************************
1414 * Function : remove_chunked_transfer_coding
1416 * Description : In-situ remove the "chunked" transfer coding as defined
1417 * in rfc2616 from a buffer.
1420 * 1 : buffer = Pointer to the text buffer
1421 * 2 : size = Number of bytes to be processed
1423 * Returns : The new size, i.e. the number of bytes from buffer which
1424 * are occupied by the stripped body, or 0 in case something
1427 *********************************************************************/
1428 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1431 unsigned int chunksize = 0;
1432 char *from_p, *to_p;
1435 from_p = to_p = buffer;
1437 if (sscanf(buffer, "%x", &chunksize) != 1)
1439 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1443 while (chunksize > 0)
1445 if (NULL == (from_p = strstr(from_p, "\r\n")))
1447 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1450 newsize += chunksize;
1453 memmove(to_p, from_p, (size_t) chunksize);
1454 to_p = buffer + newsize;
1455 from_p += chunksize + 2;
1457 if (sscanf(from_p, "%x", &chunksize) != 1)
1459 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1464 /* FIXME: Should this get its own loglevel? */
1465 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1471 /*********************************************************************
1473 * Function : url_actions
1475 * Description : Gets the actions for this URL.
1478 * 1 : http = http_request request for blocked URLs
1479 * 2 : csp = Current client state (buffers, headers, etc...)
1483 *********************************************************************/
1484 void url_actions(struct http_request *http,
1485 struct client_state *csp)
1487 struct file_list *fl;
1488 struct url_actions *b;
1490 init_current_action(csp->action);
1492 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1497 apply_url_actions(csp->action, http, b);
1502 /*********************************************************************
1504 * Function : apply_url_actions
1506 * Description : Applies a list of URL actions.
1509 * 1 : action = Destination.
1510 * 2 : http = Current URL
1511 * 3 : b = list of URL actions to apply
1515 *********************************************************************/
1516 void apply_url_actions(struct current_action_spec *action,
1517 struct http_request *http,
1518 struct url_actions *b)
1522 /* Should never happen */
1526 for (b = b->next; NULL != b; b = b->next)
1528 if (url_match(b->url, http))
1530 merge_current_action(action, b->action);
1536 /*********************************************************************
1538 * Function : forward_url
1540 * Description : Should we forward this to another proxy?
1543 * 1 : http = http_request request for current URL
1544 * 2 : csp = Current client state (buffers, headers, etc...)
1546 * Returns : Pointer to forwarding information.
1548 *********************************************************************/
1549 const struct forward_spec * forward_url(struct http_request *http,
1550 struct client_state *csp)
1552 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1553 struct forward_spec *fwd = csp->config->forward;
1562 if (url_match(fwd->url, http))