1 const char filters_rcs[] = "$Id: filters.c,v 1.69 2006/12/08 12:39:13 fabiankeil 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', `trust_url', `gif_deanimate_response',
13 * `jpeg_inspect_response', `execute_single_pcrs_command',
14 * `rewrite_url', `get_last_url'
16 * Copyright : Written by and Copyright (C) 2001, 2004-2006 the SourceForge
17 * Privoxy team. http://www.privoxy.org/
19 * Based on the Internet Junkbuster originally written
20 * by and Copyright (C) 1997 Anonymous Coders and
21 * Junkbusters Corporation. http://www.junkbusters.com
23 * This program is free software; you can redistribute it
24 * and/or modify it under the terms of the GNU General
25 * Public License as published by the Free Software
26 * Foundation; either version 2 of the License, or (at
27 * your option) any later version.
29 * This program is distributed in the hope that it will
30 * be useful, but WITHOUT ANY WARRANTY; without even the
31 * implied warranty of MERCHANTABILITY or FITNESS FOR A
32 * PARTICULAR PURPOSE. See the GNU General Public
33 * License for more details.
35 * The GNU General Public License should be included with
36 * this file. If not, you can view it at
37 * http://www.gnu.org/copyleft/gpl.html
38 * or write to the Free Software Foundation, Inc., 59
39 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
43 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
44 * Let get_last_url() catch https URLs as well.
46 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
47 * Make sure get_last_url() behaves like advertised
48 * and fast-redirects{} can be combined with redirect{}.
50 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
51 * Implemented +redirect{s@foo@bar@} to generate
52 * a redirect based on a rewritten version of the
55 * Revision 1.66 2006/09/23 13:26:38 roro
56 * Replace TABs by spaces in source code.
58 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
59 * Fix +redirect{}. Didn't work with -fast-redirects.
61 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
62 * Block requests for untrusted URLs with status
63 * code 403 instead of 200.
65 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
66 * Don't free p which is still in use and will be later
67 * freed by free_map(). Don't claim the referrer is unknown
68 * when the client didn't set one.
70 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
71 * Feature request 595948: Re-Filter logging in single line
73 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
74 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
76 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
77 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
78 * with what was really the latest development (the v_3_0_branch branch)
80 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
81 * Multiple filter file support
83 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
84 * A few memory leaks plugged (mostly on error paths)
86 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
87 * Add the ability to check jpeg images for invalid
88 * lengths of comment blocks. Defensive strategy
89 * against the exploit:
90 * Microsoft Security Bulletin MS04-028
91 * Buffer Overrun in JPEG Processing (GDI+) Could
92 * Allow Code Execution (833987)
93 * Enabled with +inspect-jpegs in actions files.
95 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
96 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
98 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
99 * Fixed bug #839859: "See why" link URL now gets url-encoded.
101 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
104 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
105 * Added basic support for OPTIONS and TRACE HTTP methods:
106 * New function direct_response which handles OPTIONS and
107 * TRACE requests whose Max-Forwards header field is zero.
109 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
110 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
112 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
113 * - Bugfix: Executing a filters without jobs no longer results in
114 * turing off *all* filters.
115 * - Security fix: Malicious web servers can't cause a seg fault
116 * through bogus chunk sizes anymore
118 * Revision 1.58 2002/04/24 02:11:17 oes
119 * Jon's multiple AF patch: url_actions now evaluates rules
122 * Revision 1.57 2002/04/08 20:38:34 swa
125 * Revision 1.56 2002/04/05 15:51:24 oes
126 * - bugfix: error-pages now get correct request protocol
127 * - fix for invalid HTML in trust info
129 * Revision 1.55 2002/04/02 16:13:51 oes
130 * Fix: No "Go there anyway" for SSL
132 * Revision 1.54 2002/04/02 14:55:56 oes
133 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
135 * Revision 1.53 2002/03/26 22:29:54 swa
136 * we have a new homepage!
138 * Revision 1.52 2002/03/24 16:35:57 jongfoster
141 * Revision 1.51 2002/03/24 15:23:33 jongfoster
144 * Revision 1.50 2002/03/24 13:25:43 swa
145 * name change related issues
147 * Revision 1.49 2002/03/16 20:29:14 oes
150 * Revision 1.48 2002/03/13 20:25:34 oes
151 * Better logging for content filters
153 * Revision 1.47 2002/03/13 00:30:52 jongfoster
155 * Added option of always sending redirect for imageblock,
156 * currently disabled with #if 0.
158 * Revision 1.46 2002/03/12 01:42:49 oes
159 * Introduced modular filters
161 * Revision 1.45 2002/03/08 16:47:50 oes
162 * Added choice beween GIF and PNG built-in images
164 * Revision 1.44 2002/03/07 03:49:31 oes
165 * - Fixed compiler warnings etc
166 * - Changed built-in images from GIF to PNG
167 * (with regard to Unisys patent issue)
168 * - Added a 4x4 pattern PNG which is less intrusive
169 * than the logo but also clearly marks the deleted banners
171 * Revision 1.43 2002/01/22 23:51:59 jongfoster
172 * Replacing strsav() with the safer string_append().
174 * Adding missing html_encode() to error message generators. Where encoded
175 * and unencoded versions of a string were provided, removing the unencoded
178 * Revision 1.42 2002/01/17 21:00:32 jongfoster
179 * Moving all our URL and URL pattern parsing code to urlmatch.c.
181 * Using a single, simple url_match(pattern,url) function - rather than
182 * the 3-line match routine which was repeated all over the place.
184 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
186 * Using parse_http_url() to parse URLs without faking a HTTP
187 * request line for parse_http_request().
189 * Revision 1.41 2001/11/13 00:14:07 jongfoster
190 * Fixing stupid bug now I've figured out what || means.
191 * (It always returns 0 or 1, not one of it's paramaters.)
193 * Revision 1.40 2001/10/26 17:37:55 oes
194 * - Re-enabled Netscape 200/404 bug workaround in block_url():
195 * - Removed OS/2 special case
196 * - Made block_url() independant from sed() having been run
197 * - Made trust_url independant from sed() having been run
198 * - Made is_imageurl independant from sed() having been run.
199 * It now checks User-Agent: and Accept: by itself.
202 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
203 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
204 * threads to call select() simultaneously. So, it's time to do a real, live,
205 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
206 * (native). Both versions will work, but using __OS2__ offers multi-threading.
208 * Revision 1.38 2001/10/23 21:32:33 jongfoster
209 * Adding error-checking to selected functions
211 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
212 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
213 * filters.c. Added a FIXME in front of the offending code. I'll gladly
214 * put in a better/more robust fix for all parties if one is presented...
215 * It seems that just returning 200 instead of 404 would pretty much fix
216 * it for everyone, but I don't know all the history of the problem.
218 * Revision 1.36 2001/10/10 16:44:16 oes
219 * Added match_portlist function
221 * Revision 1.35 2001/10/07 15:41:23 oes
222 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
224 * New function remove_chunked_transfer_coding that strips chunked
225 * transfer coding to plain and is called by pcrs_filter_response
226 * and gif_deanimate_response if neccessary
228 * Improved handling of zero-change re_filter runs
230 * pcrs_filter_response and gif_deanimate_response now remove
231 * chunked transfer codeing before processing the body.
233 * Revision 1.34 2001/09/20 15:49:36 steudten
235 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
236 * See cgi.c fill_template().
238 * Revision 1.33 2001/09/16 17:05:14 jongfoster
239 * Removing unused #include showarg.h
241 * Revision 1.32 2001/09/16 13:21:27 jongfoster
242 * Changes to use new list functions.
244 * Revision 1.31 2001/09/16 11:38:02 jongfoster
245 * Splitting fill_template() into 2 functions:
246 * template_load() loads the file
247 * template_fill() performs the PCRS regexps.
248 * This is because the CGI edit interface has a "table row"
249 * template which is used many times in the page - this
250 * change means it's only loaded from disk once.
252 * Revision 1.30 2001/09/16 11:00:10 jongfoster
253 * New function alloc_http_response, for symmetry with free_http_response
255 * Revision 1.29 2001/09/13 23:32:40 jongfoster
256 * Moving image data to cgi.c rather than cgi.h
257 * Fixing a GPF under Win32 (and any other OS that protects global
258 * constants from being written to).
260 * Revision 1.28 2001/09/10 10:18:51 oes
261 * Silenced compiler warnings
263 * Revision 1.27 2001/08/05 16:06:20 jongfoster
264 * Modifiying "struct map" so that there are now separate header and
265 * "map_entry" structures. This means that functions which modify a
266 * map no longer need to return a pointer to the modified map.
267 * Also, it no longer reverses the order of the entries (which may be
268 * important with some advanced template substitutions).
270 * Revision 1.26 2001/07/30 22:08:36 jongfoster
271 * Tidying up #defines:
272 * - All feature #defines are now of the form FEATURE_xxx
273 * - Permanently turned off WIN_GUI_EDIT
274 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
276 * Revision 1.25 2001/07/26 10:09:46 oes
277 * Made browser detection a little less naive
279 * Revision 1.24 2001/07/25 17:22:51 oes
280 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
282 * Revision 1.23 2001/07/23 13:40:12 oes
283 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
285 * Revision 1.22 2001/07/18 12:29:34 oes
286 * - Made gif_deanimate_response respect
287 * csp->action->string[ACTION_STRING_DEANIMATE]
288 * - Logging cosmetics
290 * Revision 1.21 2001/07/13 13:59:53 oes
291 * - Introduced gif_deanimate_response which shares the
292 * generic content modification interface of pcrs_filter_response
293 * and acts as a wrapper to deanimate.c:gif_deanimate()
294 * - Renamed re_process_buffer to pcrs_filter_response
295 * - pcrs_filter_response now returns NULL on failiure
296 * - Removed all #ifdef PCRS
298 * Revision 1.20 2001/07/01 17:01:04 oes
299 * Added comments and missing return statement in is_untrusted_url()
301 * Revision 1.19 2001/06/29 21:45:41 oes
302 * Indentation, CRLF->LF, Tab-> Space
304 * Revision 1.18 2001/06/29 13:27:38 oes
305 * - Cleaned up, renamed and reorderd functions
306 * and improved comments
309 * - Ported to CGI platform. Now delivers
310 * http_response or NULL
311 * - Unified HTML and GIF generation (moved image detection
312 * and GIF generation here from jcc.c:chat())
313 * - Fixed HTTP status to:
314 * - 403 (Forbidden) for the "blocked" HTML message
315 * - 200 (OK) for GIF answers
316 * - 302 (Redirect) for redirect to GIF
319 * - Ported to CGI platform. Now delivers
320 * http_response or NULL
321 * - Separated detection of untrusted URL into
322 * (bool)is_untrusted_url
323 * - Added enforcement of untrusted requests
325 * - Moved redirect_url() from cgi.c to here
326 * and ported it to the CGI platform
328 * - Removed logentry from cancelled commit
330 * Revision 1.17 2001/06/09 10:55:28 jongfoster
331 * Changing BUFSIZ ==> BUFFER_SIZE
333 * Revision 1.16 2001/06/07 23:10:26 jongfoster
334 * Allowing unanchored domain patterns to back off and retry
335 * if they partially match. Optimized right-anchored patterns.
336 * Moving ACL and forward files into config file.
337 * Replacing struct gateway with struct forward_spec
339 * Revision 1.15 2001/06/03 19:12:00 oes
340 * extracted-CGI relevant stuff
342 * Revision 1.14 2001/06/01 10:30:55 oes
343 * Added optional left-anchoring to domaincmp
345 * Revision 1.13 2001/05/31 21:21:30 jongfoster
346 * Permissionsfile / actions file changes:
347 * - Changed "permission" to "action" throughout
348 * - changes to file format to allow string parameters
349 * - Moved helper functions to actions.c
351 * Revision 1.12 2001/05/31 17:35:20 oes
353 * - Enhanced domain part globbing with infix and prefix asterisk
354 * matching and optional unanchored operation
356 * Revision 1.11 2001/05/29 11:53:23 oes
357 * "See why" link added to "blocked" page
359 * Revision 1.10 2001/05/29 09:50:24 jongfoster
360 * Unified blocklist/imagelist/permissionslist.
361 * File format is still under discussion, but the internal changes
364 * Also modified interceptor behaviour:
365 * - We now intercept all URLs beginning with one of the following
366 * prefixes (and *only* these prefixes):
368 * * http://ijbswa.sf.net/config/
369 * * http://ijbswa.sourceforge.net/config/
370 * - New interceptors "home page" - go to http://i.j.b/ to see it.
371 * - Internal changes so that intercepted and fast redirect pages
372 * are not replaced with an image.
373 * - Interceptors now have the option to send a binary page direct
374 * to the client. (i.e. ijb-send-banner uses this)
375 * - Implemented show-url-info interceptor. (Which is why I needed
376 * the above interceptors changes - a typical URL is
377 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
378 * The previous mechanism would not have intercepted that, and
379 * if it had been intercepted then it then it would have replaced
382 * Revision 1.9 2001/05/27 22:17:04 oes
384 * - re_process_buffer no longer writes the modified buffer
385 * to the client, which was very ugly. It now returns the
386 * buffer, which it is then written by chat.
388 * - content_length now adjusts the Content-Length: header
389 * for modified documents rather than crunch()ing it.
390 * (Length info in csp->content_length, which is 0 for
391 * unmodified documents)
393 * - For this to work, sed() is called twice when filtering.
395 * Revision 1.8 2001/05/26 17:13:28 jongfoster
396 * Filled in a function comment.
398 * Revision 1.7 2001/05/26 15:26:15 jongfoster
399 * ACL feature now provides more security by immediately dropping
400 * connections from untrusted hosts.
402 * Revision 1.6 2001/05/26 00:28:36 jongfoster
403 * Automatic reloading of config file.
404 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
405 * Most of the global variables have been moved to a new
406 * struct configuration_spec, accessed through csp->config->globalname
407 * Most of the globals remaining are used by the Win32 GUI.
409 * Revision 1.5 2001/05/25 22:34:30 jongfoster
412 * Revision 1.4 2001/05/22 18:46:04 oes
414 * - Enabled filtering banners by size rather than URL
415 * by adding patterns that replace all standard banner
416 * sizes with the "Junkbuster" gif to the re_filterfile
418 * - Enabled filtering WebBugs by providing a pattern
419 * which kills all 1x1 images
421 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
422 * which is selected by the (nonstandard and therefore
423 * capital) letter 'U' in the option string.
424 * It causes the quantifiers to be ungreedy by default.
425 * Appending a ? turns back to greedy (!).
427 * - Added a new interceptor ijb-send-banner, which
428 * sends back the "Junkbuster" gif. Without imagelist or
429 * MSIE detection support, or if tinygif = 1, or the
430 * URL isn't recognized as an imageurl, a lame HTML
431 * explanation is sent instead.
433 * - Added new feature, which permits blocking remote
434 * script redirects and firing back a local redirect
436 * The feature is conditionally compiled, i.e. it
437 * can be disabled with --disable-fast-redirects,
438 * plus it must be activated by a "fast-redirects"
439 * line in the config file, has its own log level
440 * and of course wants to be displayed by show-proxy-args
441 * Note: Boy, all the #ifdefs in 1001 locations and
442 * all the fumbling with configure.in and acconfig.h
443 * were *way* more work than the feature itself :-(
445 * - Because a generic redirect template was needed for
446 * this, tinygif = 3 now uses the same.
448 * - Moved GIFs, and other static HTTP response templates
453 * - Removed some >400 CRs again (Jon, you really worked
456 * Revision 1.3 2001/05/20 16:44:47 jongfoster
457 * Removing last hardcoded Junkbusters.com URLs.
459 * Revision 1.2 2001/05/20 01:21:20 jongfoster
460 * Version 2.9.4 checkin.
461 * - Merged popupfile and cookiefile, and added control over PCRS
462 * filtering, in new "permissionsfile".
463 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
464 * file error you now get a message box (in the Win32 GUI) rather
465 * than the program exiting with no explanation.
466 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
468 * - Removed tabs from "config"
469 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
470 * - Bumped up version number.
472 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
473 * Initial import of version 2.9.3 source tree
476 *********************************************************************/
482 #include <sys/types.h>
491 #endif /* ndef __OS2__ */
492 #include <netinet/in.h>
494 #include <winsock2.h>
495 #endif /* ndef _WIN32 */
499 #endif /* def __OS2__ */
507 #include "jbsockets.h"
508 #include "miscutil.h"
512 #include "deanimate.h"
513 #include "urlmatch.h"
519 const char filters_h_rcs[] = FILTERS_H_VERSION;
521 /* Fix a problem with Solaris. There should be no effect on other
523 * Solaris's isspace() is a macro which uses it's argument directly
524 * as an array index. Therefore we need to make sure that high-bit
525 * characters generate +ve values, and ideally we also want to make
526 * the argument match the declared parameter type of "int".
528 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
532 /*********************************************************************
534 * Function : block_acl
536 * Description : Block this request?
537 * Decide yes or no based on ACL file.
540 * 1 : dst = The proxy or gateway address this is going to.
541 * Or NULL to check all possible targets.
542 * 2 : csp = Current client state (buffers, headers, etc...)
543 * Also includes the client IP address.
545 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
547 *********************************************************************/
548 int block_acl(struct access_control_addr *dst, struct client_state *csp)
550 struct access_control_list *acl = csp->config->acl;
552 /* if not using an access control list, then permit the connection */
558 /* search the list */
561 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
565 /* Just want to check if they have any access */
566 if (acl->action == ACL_PERMIT)
571 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
572 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
574 if (acl->action == ACL_PERMIT)
592 /*********************************************************************
594 * Function : acl_addr
596 * Description : Called from `load_config' to parse an ACL address.
599 * 1 : aspec = String specifying ACL address.
600 * 2 : aca = struct access_control_addr to fill in.
602 * Returns : 0 => Ok, everything else is an error.
604 *********************************************************************/
605 int acl_addr(char *aspec, struct access_control_addr *aca)
607 int i, masklength, port;
613 if ((p = strchr(aspec, '/')) != NULL)
617 if (ijb_isdigit(*p) == 0)
621 masklength = atoi(p);
624 if ((masklength < 0) || (masklength > 32))
629 if ((p = strchr(aspec, ':')) != NULL)
633 if (ijb_isdigit(*p) == 0)
642 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
644 if (aca->addr == INADDR_NONE)
649 /* build the netmask */
651 for (i=1; i <= masklength ; i++)
653 aca->mask |= (1 << (32 - i));
656 /* now mask off the host portion of the ip address
657 * (i.e. save on the network portion of the address).
659 aca->addr = aca->addr & aca->mask;
664 #endif /* def FEATURE_ACL */
667 /*********************************************************************
669 * Function : match_portlist
671 * Description : Check if a given number is covered by a comma
672 * separated list of numbers and ranges (a,b-c,d,..)
675 * 1 : portlist = String with list
676 * 2 : port = port to check
678 * Returns : 0 => no match
681 *********************************************************************/
682 int match_portlist(const char *portlist, int port)
684 char *min, *max, *next, *portlist_copy;
686 min = next = portlist_copy = strdup(portlist);
689 * Zero-terminate first item and remember offset for next
691 if (NULL != (next = strchr(portlist_copy, (int) ',')))
697 * Loop through all items, checking for match
701 if (NULL == (max = strchr(min, (int) '-')))
704 * No dash, check for equality
706 if (port == atoi(min))
715 * This is a range, so check if between min and max,
716 * or, if max was omitted, between min and 65K
719 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
733 * Zero-terminate next item and remember offset for n+1
735 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
747 /*********************************************************************
749 * Function : block_url
751 * Description : Called from `chat'. Check to see if we need to block this.
754 * 1 : csp = Current client state (buffers, headers, etc...)
756 * Returns : NULL => unblocked, else HTTP block response
758 *********************************************************************/
759 struct http_response *block_url(struct client_state *csp)
761 struct http_response *rsp;
762 const char *new_content_type = NULL;
765 * If it's not blocked, don't block it ;-)
767 if ((csp->action->flags & ACTION_BLOCK) == 0)
771 if (csp->action->flags & ACTION_REDIRECT)
773 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
776 * Else, prepare a response
778 if (NULL == (rsp = alloc_http_response()))
780 return cgi_error_memory();
784 * If it's an image-url, send back an image or redirect
785 * as specified by the relevant +image action
787 #ifdef FEATURE_IMAGE_BLOCKING
788 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
792 /* determine HOW images should be blocked */
793 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
795 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
797 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
799 #if 1 /* Two alternative strategies, use this one for now: */
801 /* and handle accordingly: */
802 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
804 rsp->status = strdup("403 Request blocked by Privoxy");
805 if (rsp->status == NULL)
807 free_http_response(rsp);
808 return cgi_error_memory();
810 rsp->body = bindup(image_pattern_data, image_pattern_length);
811 if (rsp->body == NULL)
813 free_http_response(rsp);
814 return cgi_error_memory();
816 rsp->content_length = image_pattern_length;
818 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
820 free_http_response(rsp);
821 return cgi_error_memory();
825 else if (0 == strcmpic(p, "blank"))
827 rsp->status = strdup("403 Request blocked by Privoxy");
828 if (rsp->status == NULL)
830 free_http_response(rsp);
831 return cgi_error_memory();
833 rsp->body = bindup(image_blank_data, image_blank_length);
834 if (rsp->body == NULL)
836 free_http_response(rsp);
837 return cgi_error_memory();
839 rsp->content_length = image_blank_length;
841 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
843 free_http_response(rsp);
844 return cgi_error_memory();
850 rsp->status = strdup("302 Local Redirect from Privoxy");
851 if (rsp->status == NULL)
853 free_http_response(rsp);
854 return cgi_error_memory();
857 if (enlist_unique_header(rsp->headers, "Location", p))
859 free_http_response(rsp);
860 return cgi_error_memory();
864 #else /* Following code is disabled for now */
866 /* and handle accordingly: */
867 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
869 p = CGI_PREFIX "send-banner?type=pattern";
871 else if (0 == strcmpic(p, "blank"))
873 p = CGI_PREFIX "send-banner?type=blank";
875 rsp->status = strdup("302 Local Redirect from Privoxy");
876 if (rsp->status == NULL)
878 free_http_response(rsp);
879 return cgi_error_memory();
882 if (enlist_unique_header(rsp->headers, "Location", p))
884 free_http_response(rsp);
885 return cgi_error_memory();
887 #endif /* Preceeding code is disabled for now */
889 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
892 * Send empty document.
894 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
897 rsp->body = strdup(" ");
898 rsp->content_length = 1;
900 rsp->status = strdup("403 Request blocked by Privoxy");
901 if (rsp->status == NULL)
903 free_http_response(rsp);
904 return cgi_error_memory();
906 if (new_content_type != 0)
908 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
909 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
911 free_http_response(rsp);
912 return cgi_error_memory();
918 #endif /* def FEATURE_IMAGE_BLOCKING */
921 * Else, generate an HTML "blocked" message:
925 struct map * exports;
929 * Workaround for stupid Netscape bug which prevents
930 * pages from being displayed if loading a referenced
931 * JavaScript or style sheet fails. So make it appear
932 * as if it succeeded.
934 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
935 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
936 && !strstr(p, "Gecko") /* save Mozilla, */
937 && !strstr(p, "compatible") /* MSIE */
938 && !strstr(p, "Opera")) /* and Opera. */
940 rsp->status = strdup("200 Request for blocked URL");
944 rsp->status = strdup("404 Request for blocked URL");
947 if (rsp->status == NULL)
949 free_http_response(rsp);
950 return cgi_error_memory();
953 exports = default_exports(csp, NULL);
956 free_http_response(rsp);
957 return cgi_error_memory();
960 #ifdef FEATURE_FORCE_LOAD
961 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
962 if (csp->http->ssl != 0)
963 #endif /* ndef FEATURE_FORCE_LOAD */
965 err = map_block_killer(exports, "force-support");
968 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
969 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
970 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
971 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
976 free_http_response(rsp);
977 return cgi_error_memory();
980 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
983 free_http_response(rsp);
984 return cgi_error_memory();
988 return finish_http_response(rsp);
994 /*********************************************************************
996 * Function : trust_url FIXME: I should be called distrust_url
998 * Description : Calls is_untrusted_url to determine if the URL is trusted
999 * and if not, returns a HTTP 403 response with a reject message.
1002 * 1 : csp = Current client state (buffers, headers, etc...)
1004 * Returns : NULL => trusted, else http_response.
1006 *********************************************************************/
1007 struct http_response *trust_url(struct client_state *csp)
1009 struct http_response *rsp;
1010 struct map * exports;
1011 char buf[BUFFER_SIZE];
1013 struct url_spec **tl;
1018 * Don't bother to work on trusted URLs
1020 if (!is_untrusted_url(csp))
1026 * Else, prepare a response:
1028 if (NULL == (rsp = alloc_http_response()))
1030 return cgi_error_memory();
1033 rsp->status = strdup("403 Request blocked by Privoxy");
1034 exports = default_exports(csp, NULL);
1035 if (exports == NULL || rsp->status == NULL)
1037 free_http_response(rsp);
1038 return cgi_error_memory();
1042 * Export the protocol, host, port, and referrer information
1044 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1045 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1046 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1048 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1050 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1054 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1060 free_http_response(rsp);
1061 return cgi_error_memory();
1065 * Export the trust list
1068 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1070 sprintf(buf, "<li>%s</li>\n", t->spec);
1071 string_append(&p, buf);
1073 err = map(exports, "trusted-referrers", 1, p, 0);
1078 free_http_response(rsp);
1079 return cgi_error_memory();
1083 * Export the trust info, if available
1085 if (csp->config->trust_info->first)
1087 struct list_entry *l;
1090 for (l = csp->config->trust_info->first; l ; l = l->next)
1092 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1093 string_append(&p, buf);
1095 err = map(exports, "trust-info", 1, p, 0);
1099 err = map_block_killer(exports, "have-trust-info");
1105 free_http_response(rsp);
1106 return cgi_error_memory();
1110 * Export the force prefix or the force conditional block killer
1112 #ifdef FEATURE_FORCE_LOAD
1113 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1114 #else /* ifndef FEATURE_FORCE_LOAD */
1115 err = map_block_killer(exports, "force-support");
1116 #endif /* ndef FEATURE_FORCE_LOAD */
1121 free_http_response(rsp);
1122 return cgi_error_memory();
1126 * Build the response
1128 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1131 free_http_response(rsp);
1132 return cgi_error_memory();
1135 return finish_http_response(rsp);
1137 #endif /* def FEATURE_TRUST */
1139 /*********************************************************************
1141 * Function : execute_single_pcrs_command
1143 * Description : Apply single pcrs command to the subject.
1144 * The subject itself is left untouched, memory for the result
1145 * is malloc()ed and it is the caller's responsibility to free
1146 * the result when it's no longer needed.
1149 * 1 : subject = the subject (== original) string
1150 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1151 * 3 : hits = int* for returning the number of modifications
1153 * Returns : NULL in case of errors, otherwise the
1154 * result of the pcrs command.
1156 *********************************************************************/
1157 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1161 char *result = NULL;
1165 assert(pcrs_command);
1168 size = strlen(subject);
1170 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1172 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1173 pcrs_command, error);
1175 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1177 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1184 job = pcrs_free_job(job);
1191 /*********************************************************************
1193 * Function : rewrite_url
1195 * Description : Rewrites a URL with a single pcrs command
1196 * and returns the result if it differs from the
1197 * original and isn't obviously invalid.
1200 * 1 : old_url = URL to rewrite.
1201 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1204 * Returns : NULL if the pcrs_command didn't change the url, or
1205 * the result of the modification.
1207 *********************************************************************/
1208 char *rewrite_url(char *old_url, const char *pcrs_command)
1210 char *new_url = NULL;
1214 assert(pcrs_command);
1216 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1220 log_error(LOG_LEVEL_REDIRECTS,
1221 "pcrs command \"%s\" didn't change \"%s\".",
1222 pcrs_command, old_url, new_url);
1225 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1227 log_error(LOG_LEVEL_ERROR,
1228 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1229 "but the result doesn't look like a valid URL and will be ignored.",
1230 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1235 log_error(LOG_LEVEL_REDIRECTS,
1236 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1237 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1245 #ifdef FEATURE_FAST_REDIRECTS
1246 /*********************************************************************
1248 * Function : get_last_url
1250 * Description : Search for the last URL inside a string.
1251 * If the string already is a URL, it will
1252 * be the first URL found.
1255 * 1 : subject = the string to check
1256 * 2 : redirect_mode = +fast-redirect{} mode
1258 * Returns : NULL if no URL was found, or
1259 * the last URL found.
1261 *********************************************************************/
1262 char *get_last_url(char *subject, const char *redirect_mode)
1264 char *new_url = NULL;
1268 assert(redirect_mode);
1270 subject = strdup(subject);
1271 if (subject == NULL)
1273 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1277 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1279 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1280 new_url = url_decode(subject);
1281 if (new_url != NULL)
1288 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1292 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1295 * Find the last URL encoded in the request
1298 while ((tmp = strstr(tmp, "http://")) != NULL)
1302 tmp = (new_url != NULL) ? new_url : subject;
1303 while ((tmp = strstr(tmp, "https://")) != NULL)
1308 if ((new_url != NULL)
1309 && ( (new_url != subject)
1310 || (0 == strncmpic(subject, "http://", 7))
1311 || (0 == strncmpic(subject, "https://", 8))
1315 * Return new URL if we found a redirect
1316 * or if the subject already was a URL.
1318 * The second case makes sure that we can
1319 * chain get_last_url after another redirection check
1320 * (like rewrite_url) without losing earlier redirects.
1322 new_url = strdup(new_url);
1331 #endif /* def FEATURE_FAST_REDIRECTS */
1334 /*********************************************************************
1336 * Function : redirect_url
1338 * Description : Checks if Privoxy should answer the request with
1339 * a HTTP redirect and generates the redirect if
1343 * 1 : csp = Current client state (buffers, headers, etc...)
1345 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1347 *********************************************************************/
1348 struct http_response *redirect_url(struct client_state *csp)
1350 struct http_response *rsp;
1351 #ifdef FEATURE_FAST_REDIRECTS
1353 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1354 * as compile-time option? The user can easily disable
1355 * it in his action file.
1357 char * redirect_mode;
1358 #endif /* def FEATURE_FAST_REDIRECTS */
1359 char *old_url = NULL;
1360 char *new_url = NULL;
1361 char *redirection_string;
1363 if ((csp->action->flags & ACTION_REDIRECT))
1365 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1368 * If the redirection string begins with 's',
1369 * assume it's a pcrs command, otherwise treat it as
1370 * properly formatted URL and use it for the redirection
1373 * According to RFC 2616 section 14.30 the URL
1374 * has to be absolute and if the user tries:
1375 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1376 * she would get undefined results anyway.
1380 if (*redirection_string == 's')
1382 old_url = csp->http->url;
1383 new_url = rewrite_url(old_url, redirection_string);
1387 log_error(LOG_LEVEL_REDIRECTS,
1388 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1389 redirection_string);
1390 new_url = strdup(redirection_string);
1394 #ifdef FEATURE_FAST_REDIRECTS
1395 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1397 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1400 * If it exists, use the previously rewritten URL as input
1401 * otherwise just use the old path.
1403 old_url = new_url ? new_url : strdup(csp->http->path);
1404 new_url = get_last_url(old_url, redirect_mode);
1407 #endif /* def FEATURE_FAST_REDIRECTS */
1409 /* Did any redirect action trigger? */
1412 if (0 == strcmpic(new_url, csp->http->url))
1414 log_error(LOG_LEVEL_ERROR,
1415 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1416 csp->http->url, new_url);
1421 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1423 if (NULL == (rsp = alloc_http_response()))
1426 return cgi_error_memory();
1429 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1430 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1433 free_http_response(rsp);
1434 return cgi_error_memory();
1437 return finish_http_response(rsp);
1441 /* Only reached if no redirect is required */
1447 #ifdef FEATURE_IMAGE_BLOCKING
1448 /*********************************************************************
1450 * Function : is_imageurl
1452 * Description : Given a URL, decide whether it is an image or not,
1453 * using either the info from a previous +image action
1454 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1455 * is MSIE and not on a Mac, tell from the browser's accept
1459 * 1 : csp = Current client state (buffers, headers, etc...)
1461 * Returns : True (nonzero) if URL is an image, false (0)
1464 *********************************************************************/
1465 int is_imageurl(struct client_state *csp)
1467 #ifdef FEATURE_IMAGE_DETECT_MSIE
1470 tmp = get_header_value(csp->headers, "User-Agent:");
1471 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1473 tmp = get_header_value(csp->headers, "Accept:");
1474 if (tmp && strstr(tmp, "image/gif"))
1476 /* Client will accept HTML. If this seems counterintuitive,
1486 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1488 return ((csp->action->flags & ACTION_IMAGE) != 0);
1491 #endif /* def FEATURE_IMAGE_BLOCKING */
1494 #ifdef FEATURE_TRUST
1495 /*********************************************************************
1497 * Function : is_untrusted_url
1499 * Description : Should we "distrust" this URL (and block it)?
1501 * Yes if it matches a line in the trustfile, or if the
1502 * referrer matches a line starting with "+" in the
1507 * 1 : csp = Current client state (buffers, headers, etc...)
1509 * Returns : 0 => trusted, 1 => untrusted
1511 *********************************************************************/
1512 int is_untrusted_url(struct client_state *csp)
1514 struct file_list *fl;
1515 struct block_spec *b;
1516 struct url_spec **trusted_url;
1517 struct http_request rhttp[1];
1518 const char * referer;
1522 * If we don't have a trustlist, we trust everybody
1524 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1529 memset(rhttp, '\0', sizeof(*rhttp));
1532 * Do we trust the request URL itself?
1534 for (b = b->next; b ; b = b->next)
1536 if (url_match(b->url, csp->http))
1542 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1544 /* no referrer was supplied */
1550 * If not, do we maybe trust its referrer?
1552 err = parse_http_url(referer, rhttp, csp);
1558 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1560 if (url_match(*trusted_url, rhttp))
1562 /* if the URL's referrer is from a trusted referrer, then
1563 * add the target spec to the trustfile as an unblocked
1564 * domain and return NULL (which means it's OK).
1569 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1573 char * new_entry = strdup("~");
1575 string_append(&new_entry, csp->http->hostport);
1577 path = csp->http->path;
1578 if ( (path[0] == '/')
1580 && ((path_end = strchr(path + 2, '/')) != NULL))
1582 /* since this path points into a user's home space
1583 * be sure to include this spec in the trustfile.
1585 int path_len = path_end - path; /* save offset */
1586 path = strdup(path); /* Copy string */
1589 path_end = path + path_len; /* regenerate ptr to new buffer */
1590 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1592 string_join(&new_entry, path);
1595 if (new_entry != NULL)
1597 fprintf(fp, "%s\n", new_entry);
1602 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1603 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1613 #endif /* def FEATURE_TRUST */
1616 /*********************************************************************
1618 * Function : pcrs_filter_response
1620 * Description : Execute all text substitutions from all applying
1621 * +filter actions on the text buffer that's been accumulated
1622 * in csp->iob->buf. If this changes the contents, set
1623 * csp->content_length to the modified size and raise the
1624 * CSP_FLAG_MODIFIED flag.
1627 * 1 : csp = Current client state (buffers, headers, etc...)
1629 * Returns : a pointer to the (newly allocated) modified buffer.
1630 * or NULL if there were no hits or something went wrong
1632 *********************************************************************/
1633 char *pcrs_filter_response(struct client_state *csp)
1636 size_t size, prev_size;
1638 char *old = csp->iob->cur, *new = NULL;
1641 struct file_list *fl;
1642 struct re_filterfile_spec *b;
1643 struct list_entry *filtername;
1645 int i, found_filters = 0;
1650 if (csp->iob->cur >= csp->iob->eod)
1654 size = csp->iob->eod - csp->iob->cur;
1657 * Need to check the set of re_filterfiles...
1659 for (i = 0; i < MAX_AF_FILES; i++)
1672 if (0 == found_filters)
1674 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1679 * If the body has a "chunked" transfer-encoding,
1680 * get rid of it first, adjusting size and iob->eod
1682 if (csp->flags & CSP_FLAG_CHUNKED)
1684 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1685 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1689 csp->iob->eod = csp->iob->cur + size;
1690 csp->flags |= CSP_FLAG_MODIFIED;
1693 for (i = 0; i < MAX_AF_FILES; i++)
1696 if ((NULL == fl) || (NULL == fl->f))
1699 * For all applying +filter actions, look if a filter by that
1700 * name exists and if yes, execute it's pcrs_joblist on the
1703 for (b = fl->f; b; b = b->next)
1705 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1706 filtername ; filtername = filtername->next)
1708 if (strcmp(b->name, filtername->str) == 0)
1710 int current_hits = 0;
1712 if ( NULL == b->joblist )
1714 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1719 /* Apply all jobs from the joblist */
1720 for (job = b->joblist; NULL != job; job = job->next)
1722 current_hits += pcrs_execute(job, old, size, &new, &size);
1723 if (old != csp->iob->cur) free(old);
1727 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1728 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1730 hits += current_hits;
1737 * If there were no hits, destroy our copy and let
1738 * chat() use the original in csp->iob
1746 csp->flags |= CSP_FLAG_MODIFIED;
1747 csp->content_length = size;
1755 /*********************************************************************
1757 * Function : gif_deanimate_response
1759 * Description : Deanimate the GIF image that has been accumulated in
1760 * csp->iob->buf, set csp->content_length to the modified
1761 * size and raise the CSP_FLAG_MODIFIED flag.
1764 * 1 : csp = Current client state (buffers, headers, etc...)
1766 * Returns : a pointer to the (newly allocated) modified buffer.
1767 * or NULL in case something went wrong.
1769 *********************************************************************/
1770 char *gif_deanimate_response(struct client_state *csp)
1772 struct binbuffer *in, *out;
1774 size_t size = csp->iob->eod - csp->iob->cur;
1777 * If the body has a "chunked" transfer-encoding,
1778 * get rid of it first, adjusting size and iob->eod
1780 if (csp->flags & CSP_FLAG_CHUNKED)
1782 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1783 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1787 csp->iob->eod = csp->iob->cur + size;
1788 csp->flags |= CSP_FLAG_MODIFIED;
1791 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1792 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1794 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1798 in->buffer = csp->iob->cur;
1801 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1803 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1810 if ((int)size == out->offset)
1812 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1816 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1818 csp->content_length = out->offset;
1819 csp->flags |= CSP_FLAG_MODIFIED;
1829 /*********************************************************************
1831 * Function : jpeg_inspect_response
1836 * 1 : csp = Current client state (buffers, headers, etc...)
1838 * Returns : a pointer to the (newly allocated) modified buffer
1839 * or NULL in case something went wrong.
1841 *********************************************************************/
1842 char *jpeg_inspect_response(struct client_state *csp)
1844 struct binbuffer *in = NULL, *out = NULL;
1846 size_t size = csp->iob->eod - csp->iob->cur;
1849 * If the body has a "chunked" transfer-encoding,
1850 * get rid of it first, adjusting size and iob->eod
1852 if (csp->flags & CSP_FLAG_CHUNKED)
1854 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1855 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1859 csp->iob->eod = csp->iob->cur + size;
1860 csp->flags |= CSP_FLAG_MODIFIED;
1863 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1865 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1869 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1871 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1875 in->buffer = csp->iob->cur;
1879 * Calling jpeg_inspect has the side-effect of creating and
1880 * modifying the image buffer of "out" directly.
1882 if (jpeg_inspect(in, out))
1884 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1892 csp->content_length = out->offset;
1893 csp->flags |= CSP_FLAG_MODIFIED;
1903 /*********************************************************************
1905 * Function : remove_chunked_transfer_coding
1907 * Description : In-situ remove the "chunked" transfer coding as defined
1908 * in rfc2616 from a buffer.
1911 * 1 : buffer = Pointer to the text buffer
1912 * 2 : size = Number of bytes to be processed
1914 * Returns : The new size, i.e. the number of bytes from buffer which
1915 * are occupied by the stripped body, or 0 in case something
1918 *********************************************************************/
1919 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1922 unsigned int chunksize = 0;
1923 char *from_p, *to_p;
1926 from_p = to_p = buffer;
1928 if (sscanf(buffer, "%x", &chunksize) != 1)
1930 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1934 while (chunksize > 0)
1936 if (NULL == (from_p = strstr(from_p, "\r\n")))
1938 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1942 if ((newsize += chunksize) >= size)
1944 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1949 memmove(to_p, from_p, (size_t) chunksize);
1950 to_p = buffer + newsize;
1951 from_p += chunksize + 2;
1953 if (sscanf(from_p, "%x", &chunksize) != 1)
1955 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1960 /* FIXME: Should this get its own loglevel? */
1961 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1967 /*********************************************************************
1969 * Function : url_actions
1971 * Description : Gets the actions for this URL.
1974 * 1 : http = http_request request for blocked URLs
1975 * 2 : csp = Current client state (buffers, headers, etc...)
1979 *********************************************************************/
1980 void url_actions(struct http_request *http,
1981 struct client_state *csp)
1983 struct file_list *fl;
1984 struct url_actions *b;
1987 init_current_action(csp->action);
1989 for (i = 0; i < MAX_AF_FILES; i++)
1991 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1996 apply_url_actions(csp->action, http, b);
2003 /*********************************************************************
2005 * Function : apply_url_actions
2007 * Description : Applies a list of URL actions.
2010 * 1 : action = Destination.
2011 * 2 : http = Current URL
2012 * 3 : b = list of URL actions to apply
2016 *********************************************************************/
2017 void apply_url_actions(struct current_action_spec *action,
2018 struct http_request *http,
2019 struct url_actions *b)
2023 /* Should never happen */
2027 for (b = b->next; NULL != b; b = b->next)
2029 if (url_match(b->url, http))
2031 merge_current_action(action, b->action);
2037 /*********************************************************************
2039 * Function : forward_url
2041 * Description : Should we forward this to another proxy?
2044 * 1 : http = http_request request for current URL
2045 * 2 : csp = Current client state (buffers, headers, etc...)
2047 * Returns : Pointer to forwarding information.
2049 *********************************************************************/
2050 const struct forward_spec * forward_url(struct http_request *http,
2051 struct client_state *csp)
2053 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2054 struct forward_spec *fwd = csp->config->forward;
2063 if (url_match(fwd->url, http))
2074 /*********************************************************************
2076 * Function : direct_response
2078 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2079 * request and if so, return a HTTP 501 to the client.
2081 * FIXME: I have a stupid name and I should handle the
2082 * requests properly. Still, what we do here is rfc-
2083 * compliant, whereas ignoring or forwarding are not.
2086 * 1 : csp = Current client state (buffers, headers, etc...)
2088 * Returns : http_response if , NULL if nonmatch or handler fail
2090 *********************************************************************/
2091 struct http_response *direct_response(struct client_state *csp)
2093 struct http_response *rsp;
2094 struct list_entry *p;
2096 if ((0 == strcmpic(csp->http->gpc, "trace"))
2097 || (0 == strcmpic(csp->http->gpc, "options")))
2099 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2101 if (!strncmp("Max-Forwards:", p->str, 13)
2102 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
2104 /* FIXME: We could handle at least TRACE here,
2105 but that would require a verbatim copy of
2106 the request which we don't have anymore */
2108 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
2110 /* Get mem for response or fail*/
2111 if (NULL == (rsp = alloc_http_response()))
2113 return cgi_error_memory();
2116 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2118 free_http_response(rsp);
2119 return cgi_error_memory();
2123 return(finish_http_response(rsp));