1 const char filters_rcs[] = "$Id: filters.c,v 1.46 2002/03/12 01:42:49 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 * IJBSWA 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 1.46 2002/03/12 01:42:49 oes
42 * Introduced modular filters
44 * Revision 1.45 2002/03/08 16:47:50 oes
45 * Added choice beween GIF and PNG built-in images
47 * Revision 1.44 2002/03/07 03:49:31 oes
48 * - Fixed compiler warnings etc
49 * - Changed built-in images from GIF to PNG
50 * (with regard to Unisys patent issue)
51 * - Added a 4x4 pattern PNG which is less intrusive
52 * than the logo but also clearly marks the deleted banners
54 * Revision 1.43 2002/01/22 23:51:59 jongfoster
55 * Replacing strsav() with the safer string_append().
57 * Adding missing html_encode() to error message generators. Where encoded
58 * and unencoded versions of a string were provided, removing the unencoded
61 * Revision 1.42 2002/01/17 21:00:32 jongfoster
62 * Moving all our URL and URL pattern parsing code to urlmatch.c.
64 * Using a single, simple url_match(pattern,url) function - rather than
65 * the 3-line match routine which was repeated all over the place.
67 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
69 * Using parse_http_url() to parse URLs without faking a HTTP
70 * request line for parse_http_request().
72 * Revision 1.41 2001/11/13 00:14:07 jongfoster
73 * Fixing stupid bug now I've figured out what || means.
74 * (It always returns 0 or 1, not one of it's paramaters.)
76 * Revision 1.40 2001/10/26 17:37:55 oes
77 * - Re-enabled Netscape 200/404 bug workaround in block_url():
78 * - Removed OS/2 special case
79 * - Made block_url() independant from sed() having been run
80 * - Made trust_url independant from sed() having been run
81 * - Made is_imageurl independant from sed() having been run.
82 * It now checks User-Agent: and Accept: by itself.
85 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
86 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
87 * threads to call select() simultaneously. So, it's time to do a real, live,
88 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
89 * (native). Both versions will work, but using __OS2__ offers multi-threading.
91 * Revision 1.38 2001/10/23 21:32:33 jongfoster
92 * Adding error-checking to selected functions
94 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
95 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
96 * filters.c. Added a FIXME in front of the offending code. I'll gladly
97 * put in a better/more robust fix for all parties if one is presented...
98 * It seems that just returning 200 instead of 404 would pretty much fix
99 * it for everyone, but I don't know all the history of the problem.
101 * Revision 1.36 2001/10/10 16:44:16 oes
102 * Added match_portlist function
104 * Revision 1.35 2001/10/07 15:41:23 oes
105 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
107 * New function remove_chunked_transfer_coding that strips chunked
108 * transfer coding to plain and is called by pcrs_filter_response
109 * and gif_deanimate_response if neccessary
111 * Improved handling of zero-change re_filter runs
113 * pcrs_filter_response and gif_deanimate_response now remove
114 * chunked transfer codeing before processing the body.
116 * Revision 1.34 2001/09/20 15:49:36 steudten
118 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
119 * See cgi.c fill_template().
121 * Revision 1.33 2001/09/16 17:05:14 jongfoster
122 * Removing unused #include showarg.h
124 * Revision 1.32 2001/09/16 13:21:27 jongfoster
125 * Changes to use new list functions.
127 * Revision 1.31 2001/09/16 11:38:02 jongfoster
128 * Splitting fill_template() into 2 functions:
129 * template_load() loads the file
130 * template_fill() performs the PCRS regexps.
131 * This is because the CGI edit interface has a "table row"
132 * template which is used many times in the page - this
133 * change means it's only loaded from disk once.
135 * Revision 1.30 2001/09/16 11:00:10 jongfoster
136 * New function alloc_http_response, for symmetry with free_http_response
138 * Revision 1.29 2001/09/13 23:32:40 jongfoster
139 * Moving image data to cgi.c rather than cgi.h
140 * Fixing a GPF under Win32 (and any other OS that protects global
141 * constants from being written to).
143 * Revision 1.28 2001/09/10 10:18:51 oes
144 * Silenced compiler warnings
146 * Revision 1.27 2001/08/05 16:06:20 jongfoster
147 * Modifiying "struct map" so that there are now separate header and
148 * "map_entry" structures. This means that functions which modify a
149 * map no longer need to return a pointer to the modified map.
150 * Also, it no longer reverses the order of the entries (which may be
151 * important with some advanced template substitutions).
153 * Revision 1.26 2001/07/30 22:08:36 jongfoster
154 * Tidying up #defines:
155 * - All feature #defines are now of the form FEATURE_xxx
156 * - Permanently turned off WIN_GUI_EDIT
157 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
159 * Revision 1.25 2001/07/26 10:09:46 oes
160 * Made browser detection a little less naive
162 * Revision 1.24 2001/07/25 17:22:51 oes
163 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
165 * Revision 1.23 2001/07/23 13:40:12 oes
166 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
168 * Revision 1.22 2001/07/18 12:29:34 oes
169 * - Made gif_deanimate_response respect
170 * csp->action->string[ACTION_STRING_DEANIMATE]
171 * - Logging cosmetics
173 * Revision 1.21 2001/07/13 13:59:53 oes
174 * - Introduced gif_deanimate_response which shares the
175 * generic content modification interface of pcrs_filter_response
176 * and acts as a wrapper to deanimate.c:gif_deanimate()
177 * - Renamed re_process_buffer to pcrs_filter_response
178 * - pcrs_filter_response now returns NULL on failiure
179 * - Removed all #ifdef PCRS
181 * Revision 1.20 2001/07/01 17:01:04 oes
182 * Added comments and missing return statement in is_untrusted_url()
184 * Revision 1.19 2001/06/29 21:45:41 oes
185 * Indentation, CRLF->LF, Tab-> Space
187 * Revision 1.18 2001/06/29 13:27:38 oes
188 * - Cleaned up, renamed and reorderd functions
189 * and improved comments
192 * - Ported to CGI platform. Now delivers
193 * http_response or NULL
194 * - Unified HTML and GIF generation (moved image detection
195 * and GIF generation here from jcc.c:chat())
196 * - Fixed HTTP status to:
197 * - 403 (Forbidden) for the "blocked" HTML message
198 * - 200 (OK) for GIF answers
199 * - 302 (Redirect) for redirect to GIF
202 * - Ported to CGI platform. Now delivers
203 * http_response or NULL
204 * - Separated detection of untrusted URL into
205 * (bool)is_untrusted_url
206 * - Added enforcement of untrusted requests
208 * - Moved redirect_url() from cgi.c to here
209 * and ported it to the CGI platform
211 * - Removed logentry from cancelled commit
213 * Revision 1.17 2001/06/09 10:55:28 jongfoster
214 * Changing BUFSIZ ==> BUFFER_SIZE
216 * Revision 1.16 2001/06/07 23:10:26 jongfoster
217 * Allowing unanchored domain patterns to back off and retry
218 * if they partially match. Optimized right-anchored patterns.
219 * Moving ACL and forward files into config file.
220 * Replacing struct gateway with struct forward_spec
222 * Revision 1.15 2001/06/03 19:12:00 oes
223 * extracted-CGI relevant stuff
225 * Revision 1.14 2001/06/01 10:30:55 oes
226 * Added optional left-anchoring to domaincmp
228 * Revision 1.13 2001/05/31 21:21:30 jongfoster
229 * Permissionsfile / actions file changes:
230 * - Changed "permission" to "action" throughout
231 * - changes to file format to allow string parameters
232 * - Moved helper functions to actions.c
234 * Revision 1.12 2001/05/31 17:35:20 oes
236 * - Enhanced domain part globbing with infix and prefix asterisk
237 * matching and optional unanchored operation
239 * Revision 1.11 2001/05/29 11:53:23 oes
240 * "See why" link added to "blocked" page
242 * Revision 1.10 2001/05/29 09:50:24 jongfoster
243 * Unified blocklist/imagelist/permissionslist.
244 * File format is still under discussion, but the internal changes
247 * Also modified interceptor behaviour:
248 * - We now intercept all URLs beginning with one of the following
249 * prefixes (and *only* these prefixes):
251 * * http://ijbswa.sf.net/config/
252 * * http://ijbswa.sourceforge.net/config/
253 * - New interceptors "home page" - go to http://i.j.b/ to see it.
254 * - Internal changes so that intercepted and fast redirect pages
255 * are not replaced with an image.
256 * - Interceptors now have the option to send a binary page direct
257 * to the client. (i.e. ijb-send-banner uses this)
258 * - Implemented show-url-info interceptor. (Which is why I needed
259 * the above interceptors changes - a typical URL is
260 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
261 * The previous mechanism would not have intercepted that, and
262 * if it had been intercepted then it then it would have replaced
265 * Revision 1.9 2001/05/27 22:17:04 oes
267 * - re_process_buffer no longer writes the modified buffer
268 * to the client, which was very ugly. It now returns the
269 * buffer, which it is then written by chat.
271 * - content_length now adjusts the Content-Length: header
272 * for modified documents rather than crunch()ing it.
273 * (Length info in csp->content_length, which is 0 for
274 * unmodified documents)
276 * - For this to work, sed() is called twice when filtering.
278 * Revision 1.8 2001/05/26 17:13:28 jongfoster
279 * Filled in a function comment.
281 * Revision 1.7 2001/05/26 15:26:15 jongfoster
282 * ACL feature now provides more security by immediately dropping
283 * connections from untrusted hosts.
285 * Revision 1.6 2001/05/26 00:28:36 jongfoster
286 * Automatic reloading of config file.
287 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
288 * Most of the global variables have been moved to a new
289 * struct configuration_spec, accessed through csp->config->globalname
290 * Most of the globals remaining are used by the Win32 GUI.
292 * Revision 1.5 2001/05/25 22:34:30 jongfoster
295 * Revision 1.4 2001/05/22 18:46:04 oes
297 * - Enabled filtering banners by size rather than URL
298 * by adding patterns that replace all standard banner
299 * sizes with the "Junkbuster" gif to the re_filterfile
301 * - Enabled filtering WebBugs by providing a pattern
302 * which kills all 1x1 images
304 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
305 * which is selected by the (nonstandard and therefore
306 * capital) letter 'U' in the option string.
307 * It causes the quantifiers to be ungreedy by default.
308 * Appending a ? turns back to greedy (!).
310 * - Added a new interceptor ijb-send-banner, which
311 * sends back the "Junkbuster" gif. Without imagelist or
312 * MSIE detection support, or if tinygif = 1, or the
313 * URL isn't recognized as an imageurl, a lame HTML
314 * explanation is sent instead.
316 * - Added new feature, which permits blocking remote
317 * script redirects and firing back a local redirect
319 * The feature is conditionally compiled, i.e. it
320 * can be disabled with --disable-fast-redirects,
321 * plus it must be activated by a "fast-redirects"
322 * line in the config file, has its own log level
323 * and of course wants to be displayed by show-proxy-args
324 * Note: Boy, all the #ifdefs in 1001 locations and
325 * all the fumbling with configure.in and acconfig.h
326 * were *way* more work than the feature itself :-(
328 * - Because a generic redirect template was needed for
329 * this, tinygif = 3 now uses the same.
331 * - Moved GIFs, and other static HTTP response templates
336 * - Removed some >400 CRs again (Jon, you really worked
339 * Revision 1.3 2001/05/20 16:44:47 jongfoster
340 * Removing last hardcoded JunkBusters.com URLs.
342 * Revision 1.2 2001/05/20 01:21:20 jongfoster
343 * Version 2.9.4 checkin.
344 * - Merged popupfile and cookiefile, and added control over PCRS
345 * filtering, in new "permissionsfile".
346 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
347 * file error you now get a message box (in the Win32 GUI) rather
348 * than the program exiting with no explanation.
349 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
351 * - Removed tabs from "config"
352 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
353 * - Bumped up version number.
355 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
356 * Initial import of version 2.9.3 source tree
359 *********************************************************************/
365 #include <sys/types.h>
374 #endif /* ndef __OS2__ */
375 #include <netinet/in.h>
377 #include <winsock2.h>
378 #endif /* ndef _WIN32 */
382 #endif /* def __OS2__ */
390 #include "jbsockets.h"
391 #include "miscutil.h"
395 #include "deanimate.h"
396 #include "urlmatch.h"
402 const char filters_h_rcs[] = FILTERS_H_VERSION;
404 /* Fix a problem with Solaris. There should be no effect on other
406 * Solaris's isspace() is a macro which uses it's argument directly
407 * as an array index. Therefore we need to make sure that high-bit
408 * characters generate +ve values, and ideally we also want to make
409 * the argument match the declared parameter type of "int".
411 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
415 /*********************************************************************
417 * Function : block_acl
419 * Description : Block this request?
420 * Decide yes or no based on ACL file.
423 * 1 : dst = The proxy or gateway address this is going to.
424 * Or NULL to check all possible targets.
425 * 2 : csp = Current client state (buffers, headers, etc...)
426 * Also includes the client IP address.
428 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
430 *********************************************************************/
431 int block_acl(struct access_control_addr *dst, struct client_state *csp)
433 struct access_control_list *acl = csp->config->acl;
435 /* if not using an access control list, then permit the connection */
441 /* search the list */
444 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
448 /* Just want to check if they have any access */
449 if (acl->action == ACL_PERMIT)
454 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
455 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
457 if (acl->action == ACL_PERMIT)
475 /*********************************************************************
477 * Function : acl_addr
479 * Description : Called from `load_config' to parse an ACL address.
482 * 1 : aspec = String specifying ACL address.
483 * 2 : aca = struct access_control_addr to fill in.
485 * Returns : 0 => Ok, everything else is an error.
487 *********************************************************************/
488 int acl_addr(char *aspec, struct access_control_addr *aca)
490 int i, masklength, port;
496 if ((p = strchr(aspec, '/')) != NULL)
500 if (ijb_isdigit(*p) == 0)
504 masklength = atoi(p);
507 if ((masklength < 0) || (masklength > 32))
512 if ((p = strchr(aspec, ':')) != NULL)
516 if (ijb_isdigit(*p) == 0)
525 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
527 if (aca->addr == INADDR_NONE)
532 /* build the netmask */
534 for (i=1; i <= masklength ; i++)
536 aca->mask |= (1 << (32 - i));
539 /* now mask off the host portion of the ip address
540 * (i.e. save on the network portion of the address).
542 aca->addr = aca->addr & aca->mask;
547 #endif /* def FEATURE_ACL */
550 /*********************************************************************
552 * Function : match_portlist
554 * Description : Check if a given number is covered by a comma
555 * separated list of numbers and ranges (a,b-c,d,..)
558 * 1 : portlist = String with list
559 * 2 : port = port to check
561 * Returns : 0 => no match
564 *********************************************************************/
565 int match_portlist(const char *portlist, int port)
567 char *min, *max, *next, *portlist_copy;
569 min = next = portlist_copy = strdup(portlist);
572 * Zero-terminate first item and remember offset for next
574 if (NULL != (next = strchr(portlist_copy, (int) ',')))
580 * Loop through all items, checking for match
584 if (NULL == (max = strchr(min, (int) '-')))
587 * No dash, check for equality
589 if (port == atoi(min))
598 * This is a range, so check if between min and max,
599 * or, if max was omitted, between min and 65K
602 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
616 * Zero-terminate next item and remember offset for n+1
618 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
630 /*********************************************************************
632 * Function : block_url
634 * Description : Called from `chat'. Check to see if we need to block this.
637 * 1 : csp = Current client state (buffers, headers, etc...)
639 * Returns : NULL => unblocked, else HTTP block response
641 *********************************************************************/
642 struct http_response *block_url(struct client_state *csp)
644 #ifdef FEATURE_IMAGE_BLOCKING
646 #endif /* def FEATURE_IMAGE_BLOCKING */
647 struct http_response *rsp;
650 * If it's not blocked, don't block it ;-)
652 if ((csp->action->flags & ACTION_BLOCK) == 0)
658 * Else, prepare a response
660 if (NULL == (rsp = alloc_http_response()))
662 return cgi_error_memory();
666 * If it's an image-url, send back an image or redirect
667 * as specified by the relevant +image action
669 #ifdef FEATURE_IMAGE_BLOCKING
670 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
673 /* determine HOW images should be blocked */
674 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
676 #if 1 /* Two alternative strategies, use this one for now: */
678 /* and handle accordingly: */
679 if ((p == NULL) || (0 == strcmpic(p, "logo")))
681 rsp->body = bindup(image_logo_data, image_logo_length);
682 if (rsp->body == NULL)
684 free_http_response(rsp);
685 return cgi_error_memory();
687 rsp->content_length = image_logo_length;
689 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
691 free_http_response(rsp);
692 return cgi_error_memory();
696 else if (0 == strcmpic(p, "blank"))
698 rsp->body = bindup(image_blank_data, image_blank_length);
699 if (rsp->body == NULL)
701 free_http_response(rsp);
702 return cgi_error_memory();
704 rsp->content_length = image_blank_length;
706 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
708 free_http_response(rsp);
709 return cgi_error_memory();
713 else if (0 == strcmpic(p, "pattern"))
715 rsp->body = bindup(image_pattern_data, image_pattern_length);
716 if (rsp->body == NULL)
718 free_http_response(rsp);
719 return cgi_error_memory();
721 rsp->content_length = image_pattern_length;
723 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
725 free_http_response(rsp);
726 return cgi_error_memory();
732 rsp->status = strdup("302 Local Redirect from Junkbuster");
733 if (rsp->status == NULL)
735 free_http_response(rsp);
736 return cgi_error_memory();
739 if (enlist_unique_header(rsp->headers, "Location", p))
741 free_http_response(rsp);
742 return cgi_error_memory();
746 #else /* Following code is disabled for now */
748 /* and handle accordingly: */
749 if ((p == NULL) || (0 == strcmpic(p, "logo")))
751 p = CGI_PREFIX "send-banner?type=logo";
753 else if (0 == strcmpic(p, "blank"))
755 p = CGI_PREFIX "send-banner?type=blank";
757 else if (0 == strcmpic(p, "pattern"))
759 p = CGI_PREFIX "send-banner?type=pattern";
761 rsp->status = strdup("302 Local Redirect from Junkbuster");
762 if (rsp->status == NULL)
764 free_http_response(rsp);
765 return cgi_error_memory();
768 if (enlist_unique_header(rsp->headers, "Location", p))
770 free_http_response(rsp);
771 return cgi_error_memory();
773 #endif /* Preceeding code is disabled for now */
776 #endif /* def FEATURE_IMAGE_BLOCKING */
779 * Else, generate an HTML "blocked" message:
783 struct map * exports;
786 * Workaround for stupid Netscape bug which prevents
787 * pages from being displayed if loading a referenced
788 * JavaScript or style sheet fails. So make it appear
789 * as if it succeeded.
791 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
792 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
793 && !strstr(p, "Gecko") /* save Mozilla, */
794 && !strstr(p, "compatible") /* MSIE */
795 && !strstr(p, "Opera")) /* and Opera. */
797 rsp->status = strdup("200 Request for blocked URL");
801 rsp->status = strdup("404 Request for blocked URL");
804 if (rsp->status == NULL)
806 free_http_response(rsp);
807 return cgi_error_memory();
810 exports = default_exports(csp, NULL);
813 free_http_response(rsp);
814 return cgi_error_memory();
817 #ifdef FEATURE_FORCE_LOAD
818 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
819 #else /* ifndef FEATURE_FORCE_LOAD */
820 err = map_block_killer(exports, "force-support");
821 #endif /* ndef FEATURE_FORCE_LOAD */
823 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
824 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
829 free_http_response(rsp);
830 return cgi_error_memory();
833 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
836 free_http_response(rsp);
837 return cgi_error_memory();
841 return finish_http_response(rsp);
847 /*********************************************************************
849 * Function : trust_url FIXME: I should be called distrust_url
851 * Description : Calls is_untrusted_url to determine if the URL is trusted
852 * and if not, returns a HTTP 304 response with a reject message.
855 * 1 : csp = Current client state (buffers, headers, etc...)
857 * Returns : NULL => trusted, else http_response.
859 *********************************************************************/
860 struct http_response *trust_url(struct client_state *csp)
862 struct http_response *rsp;
863 struct map * exports;
864 char buf[BUFFER_SIZE];
866 struct url_spec **tl;
871 * Don't bother to work on trusted URLs
873 if (!is_untrusted_url(csp))
879 * Else, prepare a response:
881 if (NULL == (rsp = alloc_http_response()))
883 return cgi_error_memory();
886 exports = default_exports(csp, NULL);
889 free_http_response(rsp);
890 return cgi_error_memory();
894 * Export the host, port, and referrer information
896 err = map(exports, "hostport", 1, csp->http->hostport, 1);
897 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
899 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
901 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
905 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
911 free_http_response(rsp);
912 return cgi_error_memory();
916 * Export the trust list
919 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
921 sprintf(buf, "<li>%s</li>\n", t->spec);
922 string_append(&p, buf);
924 err = map(exports, "trusted-referrers", 1, p, 0);
929 free_http_response(rsp);
930 return cgi_error_memory();
934 * Export the trust info, if available
936 if (csp->config->trust_info->first)
938 struct list_entry *l;
941 for (l = csp->config->trust_info->first; l ; l = l->next)
943 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
944 string_append(&p, buf);
946 err = map(exports, "trust-info", 1, p, 0);
950 err = map_block_killer(exports, "have-trust-info");
956 free_http_response(rsp);
957 return cgi_error_memory();
961 * Export the force prefix or the force conditional block killer
963 #ifdef FEATURE_FORCE_LOAD
964 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
965 #else /* ifndef FEATURE_FORCE_LOAD */
966 err = map_block_killer(exports, "force-support");
967 #endif /* ndef FEATURE_FORCE_LOAD */
972 free_http_response(rsp);
973 return cgi_error_memory();
979 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
982 free_http_response(rsp);
983 return cgi_error_memory();
986 return finish_http_response(rsp);
988 #endif /* def FEATURE_TRUST */
991 #ifdef FEATURE_FAST_REDIRECTS
992 /*********************************************************************
994 * Function : redirect_url
996 * Description : Checks for redirection URLs and returns a HTTP redirect
997 * to the destination URL, if necessary
1000 * 1 : csp = Current client state (buffers, headers, etc...)
1002 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1004 *********************************************************************/
1005 struct http_response *redirect_url(struct client_state *csp)
1008 struct http_response *rsp;
1010 p = q = csp->http->path;
1011 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1014 * find the last URL encoded in the request
1016 while ((p = strstr(p, "http://")) != NULL)
1022 * if there was any, generate and return a HTTP redirect
1024 if (q != csp->http->path)
1026 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1028 if (NULL == (rsp = alloc_http_response()))
1030 return cgi_error_memory();
1033 if ( enlist_unique_header(rsp->headers, "Location", q)
1034 || (NULL == (rsp->status = strdup("302 Local Redirect from Junkbuster"))) )
1036 free_http_response(rsp);
1037 return cgi_error_memory();
1040 return finish_http_response(rsp);
1048 #endif /* def FEATURE_FAST_REDIRECTS */
1051 #ifdef FEATURE_IMAGE_BLOCKING
1052 /*********************************************************************
1054 * Function : is_imageurl
1056 * Description : Given a URL, decide whether it is an image or not,
1057 * using either the info from a previous +image action
1058 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1059 * the browser's accept header.
1062 * 1 : csp = Current client state (buffers, headers, etc...)
1064 * Returns : True (nonzero) if URL is an image, false (0)
1067 *********************************************************************/
1068 int is_imageurl(struct client_state *csp)
1070 #ifdef FEATURE_IMAGE_DETECT_MSIE
1073 tmp = get_header_value(csp->headers, "User-Agent:");
1074 if (tmp && strstr(tmp, "MSIE"))
1076 tmp = get_header_value(csp->headers, "Accept:");
1077 if (tmp && strstr(tmp, "image/gif"))
1079 /* Client will accept HTML. If this seems counterintuitive,
1089 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1091 return ((csp->action->flags & ACTION_IMAGE) != 0);
1094 #endif /* def FEATURE_IMAGE_BLOCKING */
1097 #ifdef FEATURE_COOKIE_JAR
1098 /*********************************************************************
1100 * Function : is_untrusted_url
1102 * Description : Should we "distrust" this URL (and block it)?
1104 * Yes if it matches a line in the trustfile, or if the
1105 * referrer matches a line starting with "+" in the
1110 * 1 : csp = Current client state (buffers, headers, etc...)
1112 * Returns : 0 => trusted, 1 => untrusted
1114 *********************************************************************/
1115 int is_untrusted_url(struct client_state *csp)
1117 struct file_list *fl;
1118 struct block_spec *b;
1119 struct url_spec **trusted_url;
1120 struct http_request rhttp[1];
1121 const char * referer;
1125 * If we don't have a trustlist, we trust everybody
1127 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1132 memset(rhttp, '\0', sizeof(*rhttp));
1135 * Do we trust the request URL itself?
1137 for (b = b->next; b ; b = b->next)
1139 if (url_match(b->url, csp->http))
1145 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1147 /* no referrer was supplied */
1152 * If not, do we maybe trust its referrer?
1157 * Parse the URL from the referrer
1160 err = parse_http_url(referer, rhttp, csp);
1166 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1168 if (url_match(*trusted_url, rhttp))
1170 /* if the URL's referrer is from a trusted referrer, then
1171 * add the target spec to the trustfile as an unblocked
1172 * domain and return NULL (which means it's OK).
1177 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1181 char * new_entry = strdup("~");
1183 string_append(&new_entry, csp->http->hostport);
1185 path = csp->http->path;
1186 if ( (path[0] == '/')
1188 && ((path_end = strchr(path + 2, '/')) != NULL))
1190 /* since this path points into a user's home space
1191 * be sure to include this spec in the trustfile.
1193 int path_len = path_end - path; /* save offset */
1194 path = strdup(path); /* Copy string */
1197 path_end = path + path_len; /* regenerate ptr to new buffer */
1198 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1200 string_join(&new_entry, path);
1203 if (new_entry != NULL)
1205 fprintf(fp, "%s\n", new_entry);
1210 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1211 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1221 #endif /* def FEATURE_COOKIE_JAR */
1224 /*********************************************************************
1226 * Function : pcrs_filter_response
1228 * Description : Ecexute all text substitutions from all applying
1229 * +filter actions on the text buffer that's been accumulated
1230 * in csp->iob->buf. If this changes the contents, set
1231 * csp->content_length to the modified size and raise the
1232 * CSP_FLAG_MODIFIED flag.
1235 * 1 : csp = Current client state (buffers, headers, etc...)
1237 * Returns : a pointer to the (newly allocated) modified buffer.
1238 * or NULL if there were no hits or something went wrong
1240 *********************************************************************/
1241 char *pcrs_filter_response(struct client_state *csp)
1246 char *old = csp->iob->cur, *new = NULL;
1249 struct file_list *fl;
1250 struct re_filterfile_spec *b;
1251 struct list_entry *filtername;
1256 if (csp->iob->cur >= csp->iob->eod)
1260 size = csp->iob->eod - csp->iob->cur;
1262 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1264 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1269 * If the body has a "chunked" transfer-encoding,
1270 * get rid of it first, adjusting size and iob->eod
1272 if (csp->flags & CSP_FLAG_CHUNKED)
1274 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1275 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1279 csp->iob->eod = csp->iob->cur + size;
1280 csp->flags |= CSP_FLAG_MODIFIED;
1284 * For all applying +filter actions, look if a filter by that
1285 * name exists and if yes, execute it's pcrs_joblist on the
1288 for (b = fl->f; b; b = b->next)
1290 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1291 filtername ; filtername = filtername->next)
1293 if (strcmp(b->filtername, filtername->str) == 0)
1295 if ( NULL == b->joblist )
1297 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->filtername);
1301 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1302 csp->http->hostport, csp->http->path, size, b->filtername);
1304 /* Apply all jobs from the joblist */
1305 for (job = b->joblist; NULL != job; job = job->next)
1307 hits += pcrs_execute(job, old, size, &new, &size);
1308 if (old != csp->iob->cur) free(old);
1312 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", hits, size);
1318 * If there were no hits, destroy our copy and let
1319 * chat() use the original in csp->iob
1327 csp->flags |= CSP_FLAG_MODIFIED;
1328 csp->content_length = size;
1336 /*********************************************************************
1338 * Function : gif_deanimate_response
1340 * Description : Deanimate the GIF image that has been accumulated in
1341 * csp->iob->buf, set csp->content_length to the modified
1342 * size and raise the CSP_FLAG_MODIFIED flag.
1345 * 1 : csp = Current client state (buffers, headers, etc...)
1347 * Returns : a pointer to the (newly allocated) modified buffer.
1348 * or NULL in case something went wrong.
1350 *********************************************************************/
1351 char *gif_deanimate_response(struct client_state *csp)
1353 struct binbuffer *in, *out;
1355 size_t size = csp->iob->eod - csp->iob->cur;
1358 * If the body has a "chunked" transfer-encoding,
1359 * get rid of it first, adjusting size and iob->eod
1361 if (csp->flags & CSP_FLAG_CHUNKED)
1363 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1364 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1368 csp->iob->eod = csp->iob->cur + size;
1369 csp->flags |= CSP_FLAG_MODIFIED;
1372 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1373 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1375 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1379 in->buffer = csp->iob->cur;
1382 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1384 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1391 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1392 csp->content_length = out->offset;
1393 csp->flags |= CSP_FLAG_MODIFIED;
1403 /*********************************************************************
1405 * Function : remove_chunked_transfer_coding
1407 * Description : In-situ remove the "chunked" transfer coding as defined
1408 * in rfc2616 from a buffer.
1411 * 1 : buffer = Pointer to the text buffer
1412 * 2 : size = Number of bytes to be processed
1414 * Returns : The new size, i.e. the number of bytes from buffer which
1415 * are occupied by the stripped body, or 0 in case something
1418 *********************************************************************/
1419 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1422 unsigned int chunksize = 0;
1423 char *from_p, *to_p;
1426 from_p = to_p = buffer;
1428 if (sscanf(buffer, "%x", &chunksize) != 1)
1430 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1434 while (chunksize > 0)
1436 if (NULL == (from_p = strstr(from_p, "\r\n")))
1438 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1441 newsize += chunksize;
1444 memmove(to_p, from_p, (size_t) chunksize);
1445 to_p = buffer + newsize;
1446 from_p += chunksize + 2;
1448 if (sscanf(from_p, "%x", &chunksize) != 1)
1450 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1455 /* FIXME: Should this get its own loglevel? */
1456 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1462 /*********************************************************************
1464 * Function : url_actions
1466 * Description : Gets the actions for this URL.
1469 * 1 : http = http_request request for blocked URLs
1470 * 2 : csp = Current client state (buffers, headers, etc...)
1474 *********************************************************************/
1475 void url_actions(struct http_request *http,
1476 struct client_state *csp)
1478 struct file_list *fl;
1479 struct url_actions *b;
1481 init_current_action(csp->action);
1483 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1488 apply_url_actions(csp->action, http, b);
1493 /*********************************************************************
1495 * Function : apply_url_actions
1497 * Description : Applies a list of URL actions.
1500 * 1 : action = Destination.
1501 * 2 : http = Current URL
1502 * 3 : b = list of URL actions to apply
1506 *********************************************************************/
1507 void apply_url_actions(struct current_action_spec *action,
1508 struct http_request *http,
1509 struct url_actions *b)
1513 /* Should never happen */
1517 for (b = b->next; NULL != b; b = b->next)
1519 if (url_match(b->url, http))
1521 merge_current_action(action, b->action);
1527 /*********************************************************************
1529 * Function : forward_url
1531 * Description : Should we forward this to another proxy?
1534 * 1 : http = http_request request for current URL
1535 * 2 : csp = Current client state (buffers, headers, etc...)
1537 * Returns : Pointer to forwarding information.
1539 *********************************************************************/
1540 const struct forward_spec * forward_url(struct http_request *http,
1541 struct client_state *csp)
1543 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1544 struct forward_spec *fwd = csp->config->forward;
1553 if (url_match(fwd->url, http))