1 const char filters_rcs[] = "$Id: filters.c,v 1.68 2006/12/05 14:45:48 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 1.68 2006/12/05 14:45:48 fabiankeil
44 * Make sure get_last_url() behaves like advertised
45 * and fast-redirects{} can be combined with redirect{}.
47 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
48 * Implemented +redirect{s@foo@bar@} to generate
49 * a redirect based on a rewritten version of the
52 * Revision 1.66 2006/09/23 13:26:38 roro
53 * Replace TABs by spaces in source code.
55 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
56 * Fix +redirect{}. Didn't work with -fast-redirects.
58 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
59 * Block requests for untrusted URLs with status
60 * code 403 instead of 200.
62 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
63 * Don't free p which is still in use and will be later
64 * freed by free_map(). Don't claim the referrer is unknown
65 * when the client didn't set one.
67 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
68 * Feature request 595948: Re-Filter logging in single line
70 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
71 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
73 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
74 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
75 * with what was really the latest development (the v_3_0_branch branch)
77 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
78 * Multiple filter file support
80 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
81 * A few memory leaks plugged (mostly on error paths)
83 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
84 * Add the ability to check jpeg images for invalid
85 * lengths of comment blocks. Defensive strategy
86 * against the exploit:
87 * Microsoft Security Bulletin MS04-028
88 * Buffer Overrun in JPEG Processing (GDI+) Could
89 * Allow Code Execution (833987)
90 * Enabled with +inspect-jpegs in actions files.
92 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
93 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
95 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
96 * Fixed bug #839859: "See why" link URL now gets url-encoded.
98 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
101 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
102 * Added basic support for OPTIONS and TRACE HTTP methods:
103 * New function direct_response which handles OPTIONS and
104 * TRACE requests whose Max-Forwards header field is zero.
106 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
107 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
109 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
110 * - Bugfix: Executing a filters without jobs no longer results in
111 * turing off *all* filters.
112 * - Security fix: Malicious web servers can't cause a seg fault
113 * through bogus chunk sizes anymore
115 * Revision 1.58 2002/04/24 02:11:17 oes
116 * Jon's multiple AF patch: url_actions now evaluates rules
119 * Revision 1.57 2002/04/08 20:38:34 swa
122 * Revision 1.56 2002/04/05 15:51:24 oes
123 * - bugfix: error-pages now get correct request protocol
124 * - fix for invalid HTML in trust info
126 * Revision 1.55 2002/04/02 16:13:51 oes
127 * Fix: No "Go there anyway" for SSL
129 * Revision 1.54 2002/04/02 14:55:56 oes
130 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
132 * Revision 1.53 2002/03/26 22:29:54 swa
133 * we have a new homepage!
135 * Revision 1.52 2002/03/24 16:35:57 jongfoster
138 * Revision 1.51 2002/03/24 15:23:33 jongfoster
141 * Revision 1.50 2002/03/24 13:25:43 swa
142 * name change related issues
144 * Revision 1.49 2002/03/16 20:29:14 oes
147 * Revision 1.48 2002/03/13 20:25:34 oes
148 * Better logging for content filters
150 * Revision 1.47 2002/03/13 00:30:52 jongfoster
152 * Added option of always sending redirect for imageblock,
153 * currently disabled with #if 0.
155 * Revision 1.46 2002/03/12 01:42:49 oes
156 * Introduced modular filters
158 * Revision 1.45 2002/03/08 16:47:50 oes
159 * Added choice beween GIF and PNG built-in images
161 * Revision 1.44 2002/03/07 03:49:31 oes
162 * - Fixed compiler warnings etc
163 * - Changed built-in images from GIF to PNG
164 * (with regard to Unisys patent issue)
165 * - Added a 4x4 pattern PNG which is less intrusive
166 * than the logo but also clearly marks the deleted banners
168 * Revision 1.43 2002/01/22 23:51:59 jongfoster
169 * Replacing strsav() with the safer string_append().
171 * Adding missing html_encode() to error message generators. Where encoded
172 * and unencoded versions of a string were provided, removing the unencoded
175 * Revision 1.42 2002/01/17 21:00:32 jongfoster
176 * Moving all our URL and URL pattern parsing code to urlmatch.c.
178 * Using a single, simple url_match(pattern,url) function - rather than
179 * the 3-line match routine which was repeated all over the place.
181 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
183 * Using parse_http_url() to parse URLs without faking a HTTP
184 * request line for parse_http_request().
186 * Revision 1.41 2001/11/13 00:14:07 jongfoster
187 * Fixing stupid bug now I've figured out what || means.
188 * (It always returns 0 or 1, not one of it's paramaters.)
190 * Revision 1.40 2001/10/26 17:37:55 oes
191 * - Re-enabled Netscape 200/404 bug workaround in block_url():
192 * - Removed OS/2 special case
193 * - Made block_url() independant from sed() having been run
194 * - Made trust_url independant from sed() having been run
195 * - Made is_imageurl independant from sed() having been run.
196 * It now checks User-Agent: and Accept: by itself.
199 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
200 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
201 * threads to call select() simultaneously. So, it's time to do a real, live,
202 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
203 * (native). Both versions will work, but using __OS2__ offers multi-threading.
205 * Revision 1.38 2001/10/23 21:32:33 jongfoster
206 * Adding error-checking to selected functions
208 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
209 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
210 * filters.c. Added a FIXME in front of the offending code. I'll gladly
211 * put in a better/more robust fix for all parties if one is presented...
212 * It seems that just returning 200 instead of 404 would pretty much fix
213 * it for everyone, but I don't know all the history of the problem.
215 * Revision 1.36 2001/10/10 16:44:16 oes
216 * Added match_portlist function
218 * Revision 1.35 2001/10/07 15:41:23 oes
219 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
221 * New function remove_chunked_transfer_coding that strips chunked
222 * transfer coding to plain and is called by pcrs_filter_response
223 * and gif_deanimate_response if neccessary
225 * Improved handling of zero-change re_filter runs
227 * pcrs_filter_response and gif_deanimate_response now remove
228 * chunked transfer codeing before processing the body.
230 * Revision 1.34 2001/09/20 15:49:36 steudten
232 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
233 * See cgi.c fill_template().
235 * Revision 1.33 2001/09/16 17:05:14 jongfoster
236 * Removing unused #include showarg.h
238 * Revision 1.32 2001/09/16 13:21:27 jongfoster
239 * Changes to use new list functions.
241 * Revision 1.31 2001/09/16 11:38:02 jongfoster
242 * Splitting fill_template() into 2 functions:
243 * template_load() loads the file
244 * template_fill() performs the PCRS regexps.
245 * This is because the CGI edit interface has a "table row"
246 * template which is used many times in the page - this
247 * change means it's only loaded from disk once.
249 * Revision 1.30 2001/09/16 11:00:10 jongfoster
250 * New function alloc_http_response, for symmetry with free_http_response
252 * Revision 1.29 2001/09/13 23:32:40 jongfoster
253 * Moving image data to cgi.c rather than cgi.h
254 * Fixing a GPF under Win32 (and any other OS that protects global
255 * constants from being written to).
257 * Revision 1.28 2001/09/10 10:18:51 oes
258 * Silenced compiler warnings
260 * Revision 1.27 2001/08/05 16:06:20 jongfoster
261 * Modifiying "struct map" so that there are now separate header and
262 * "map_entry" structures. This means that functions which modify a
263 * map no longer need to return a pointer to the modified map.
264 * Also, it no longer reverses the order of the entries (which may be
265 * important with some advanced template substitutions).
267 * Revision 1.26 2001/07/30 22:08:36 jongfoster
268 * Tidying up #defines:
269 * - All feature #defines are now of the form FEATURE_xxx
270 * - Permanently turned off WIN_GUI_EDIT
271 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
273 * Revision 1.25 2001/07/26 10:09:46 oes
274 * Made browser detection a little less naive
276 * Revision 1.24 2001/07/25 17:22:51 oes
277 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
279 * Revision 1.23 2001/07/23 13:40:12 oes
280 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
282 * Revision 1.22 2001/07/18 12:29:34 oes
283 * - Made gif_deanimate_response respect
284 * csp->action->string[ACTION_STRING_DEANIMATE]
285 * - Logging cosmetics
287 * Revision 1.21 2001/07/13 13:59:53 oes
288 * - Introduced gif_deanimate_response which shares the
289 * generic content modification interface of pcrs_filter_response
290 * and acts as a wrapper to deanimate.c:gif_deanimate()
291 * - Renamed re_process_buffer to pcrs_filter_response
292 * - pcrs_filter_response now returns NULL on failiure
293 * - Removed all #ifdef PCRS
295 * Revision 1.20 2001/07/01 17:01:04 oes
296 * Added comments and missing return statement in is_untrusted_url()
298 * Revision 1.19 2001/06/29 21:45:41 oes
299 * Indentation, CRLF->LF, Tab-> Space
301 * Revision 1.18 2001/06/29 13:27:38 oes
302 * - Cleaned up, renamed and reorderd functions
303 * and improved comments
306 * - Ported to CGI platform. Now delivers
307 * http_response or NULL
308 * - Unified HTML and GIF generation (moved image detection
309 * and GIF generation here from jcc.c:chat())
310 * - Fixed HTTP status to:
311 * - 403 (Forbidden) for the "blocked" HTML message
312 * - 200 (OK) for GIF answers
313 * - 302 (Redirect) for redirect to GIF
316 * - Ported to CGI platform. Now delivers
317 * http_response or NULL
318 * - Separated detection of untrusted URL into
319 * (bool)is_untrusted_url
320 * - Added enforcement of untrusted requests
322 * - Moved redirect_url() from cgi.c to here
323 * and ported it to the CGI platform
325 * - Removed logentry from cancelled commit
327 * Revision 1.17 2001/06/09 10:55:28 jongfoster
328 * Changing BUFSIZ ==> BUFFER_SIZE
330 * Revision 1.16 2001/06/07 23:10:26 jongfoster
331 * Allowing unanchored domain patterns to back off and retry
332 * if they partially match. Optimized right-anchored patterns.
333 * Moving ACL and forward files into config file.
334 * Replacing struct gateway with struct forward_spec
336 * Revision 1.15 2001/06/03 19:12:00 oes
337 * extracted-CGI relevant stuff
339 * Revision 1.14 2001/06/01 10:30:55 oes
340 * Added optional left-anchoring to domaincmp
342 * Revision 1.13 2001/05/31 21:21:30 jongfoster
343 * Permissionsfile / actions file changes:
344 * - Changed "permission" to "action" throughout
345 * - changes to file format to allow string parameters
346 * - Moved helper functions to actions.c
348 * Revision 1.12 2001/05/31 17:35:20 oes
350 * - Enhanced domain part globbing with infix and prefix asterisk
351 * matching and optional unanchored operation
353 * Revision 1.11 2001/05/29 11:53:23 oes
354 * "See why" link added to "blocked" page
356 * Revision 1.10 2001/05/29 09:50:24 jongfoster
357 * Unified blocklist/imagelist/permissionslist.
358 * File format is still under discussion, but the internal changes
361 * Also modified interceptor behaviour:
362 * - We now intercept all URLs beginning with one of the following
363 * prefixes (and *only* these prefixes):
365 * * http://ijbswa.sf.net/config/
366 * * http://ijbswa.sourceforge.net/config/
367 * - New interceptors "home page" - go to http://i.j.b/ to see it.
368 * - Internal changes so that intercepted and fast redirect pages
369 * are not replaced with an image.
370 * - Interceptors now have the option to send a binary page direct
371 * to the client. (i.e. ijb-send-banner uses this)
372 * - Implemented show-url-info interceptor. (Which is why I needed
373 * the above interceptors changes - a typical URL is
374 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
375 * The previous mechanism would not have intercepted that, and
376 * if it had been intercepted then it then it would have replaced
379 * Revision 1.9 2001/05/27 22:17:04 oes
381 * - re_process_buffer no longer writes the modified buffer
382 * to the client, which was very ugly. It now returns the
383 * buffer, which it is then written by chat.
385 * - content_length now adjusts the Content-Length: header
386 * for modified documents rather than crunch()ing it.
387 * (Length info in csp->content_length, which is 0 for
388 * unmodified documents)
390 * - For this to work, sed() is called twice when filtering.
392 * Revision 1.8 2001/05/26 17:13:28 jongfoster
393 * Filled in a function comment.
395 * Revision 1.7 2001/05/26 15:26:15 jongfoster
396 * ACL feature now provides more security by immediately dropping
397 * connections from untrusted hosts.
399 * Revision 1.6 2001/05/26 00:28:36 jongfoster
400 * Automatic reloading of config file.
401 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
402 * Most of the global variables have been moved to a new
403 * struct configuration_spec, accessed through csp->config->globalname
404 * Most of the globals remaining are used by the Win32 GUI.
406 * Revision 1.5 2001/05/25 22:34:30 jongfoster
409 * Revision 1.4 2001/05/22 18:46:04 oes
411 * - Enabled filtering banners by size rather than URL
412 * by adding patterns that replace all standard banner
413 * sizes with the "Junkbuster" gif to the re_filterfile
415 * - Enabled filtering WebBugs by providing a pattern
416 * which kills all 1x1 images
418 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
419 * which is selected by the (nonstandard and therefore
420 * capital) letter 'U' in the option string.
421 * It causes the quantifiers to be ungreedy by default.
422 * Appending a ? turns back to greedy (!).
424 * - Added a new interceptor ijb-send-banner, which
425 * sends back the "Junkbuster" gif. Without imagelist or
426 * MSIE detection support, or if tinygif = 1, or the
427 * URL isn't recognized as an imageurl, a lame HTML
428 * explanation is sent instead.
430 * - Added new feature, which permits blocking remote
431 * script redirects and firing back a local redirect
433 * The feature is conditionally compiled, i.e. it
434 * can be disabled with --disable-fast-redirects,
435 * plus it must be activated by a "fast-redirects"
436 * line in the config file, has its own log level
437 * and of course wants to be displayed by show-proxy-args
438 * Note: Boy, all the #ifdefs in 1001 locations and
439 * all the fumbling with configure.in and acconfig.h
440 * were *way* more work than the feature itself :-(
442 * - Because a generic redirect template was needed for
443 * this, tinygif = 3 now uses the same.
445 * - Moved GIFs, and other static HTTP response templates
450 * - Removed some >400 CRs again (Jon, you really worked
453 * Revision 1.3 2001/05/20 16:44:47 jongfoster
454 * Removing last hardcoded Junkbusters.com URLs.
456 * Revision 1.2 2001/05/20 01:21:20 jongfoster
457 * Version 2.9.4 checkin.
458 * - Merged popupfile and cookiefile, and added control over PCRS
459 * filtering, in new "permissionsfile".
460 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
461 * file error you now get a message box (in the Win32 GUI) rather
462 * than the program exiting with no explanation.
463 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
465 * - Removed tabs from "config"
466 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
467 * - Bumped up version number.
469 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
470 * Initial import of version 2.9.3 source tree
473 *********************************************************************/
479 #include <sys/types.h>
488 #endif /* ndef __OS2__ */
489 #include <netinet/in.h>
491 #include <winsock2.h>
492 #endif /* ndef _WIN32 */
496 #endif /* def __OS2__ */
504 #include "jbsockets.h"
505 #include "miscutil.h"
509 #include "deanimate.h"
510 #include "urlmatch.h"
516 const char filters_h_rcs[] = FILTERS_H_VERSION;
518 /* Fix a problem with Solaris. There should be no effect on other
520 * Solaris's isspace() is a macro which uses it's argument directly
521 * as an array index. Therefore we need to make sure that high-bit
522 * characters generate +ve values, and ideally we also want to make
523 * the argument match the declared parameter type of "int".
525 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
529 /*********************************************************************
531 * Function : block_acl
533 * Description : Block this request?
534 * Decide yes or no based on ACL file.
537 * 1 : dst = The proxy or gateway address this is going to.
538 * Or NULL to check all possible targets.
539 * 2 : csp = Current client state (buffers, headers, etc...)
540 * Also includes the client IP address.
542 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
544 *********************************************************************/
545 int block_acl(struct access_control_addr *dst, struct client_state *csp)
547 struct access_control_list *acl = csp->config->acl;
549 /* if not using an access control list, then permit the connection */
555 /* search the list */
558 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
562 /* Just want to check if they have any access */
563 if (acl->action == ACL_PERMIT)
568 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
569 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
571 if (acl->action == ACL_PERMIT)
589 /*********************************************************************
591 * Function : acl_addr
593 * Description : Called from `load_config' to parse an ACL address.
596 * 1 : aspec = String specifying ACL address.
597 * 2 : aca = struct access_control_addr to fill in.
599 * Returns : 0 => Ok, everything else is an error.
601 *********************************************************************/
602 int acl_addr(char *aspec, struct access_control_addr *aca)
604 int i, masklength, port;
610 if ((p = strchr(aspec, '/')) != NULL)
614 if (ijb_isdigit(*p) == 0)
618 masklength = atoi(p);
621 if ((masklength < 0) || (masklength > 32))
626 if ((p = strchr(aspec, ':')) != NULL)
630 if (ijb_isdigit(*p) == 0)
639 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
641 if (aca->addr == INADDR_NONE)
646 /* build the netmask */
648 for (i=1; i <= masklength ; i++)
650 aca->mask |= (1 << (32 - i));
653 /* now mask off the host portion of the ip address
654 * (i.e. save on the network portion of the address).
656 aca->addr = aca->addr & aca->mask;
661 #endif /* def FEATURE_ACL */
664 /*********************************************************************
666 * Function : match_portlist
668 * Description : Check if a given number is covered by a comma
669 * separated list of numbers and ranges (a,b-c,d,..)
672 * 1 : portlist = String with list
673 * 2 : port = port to check
675 * Returns : 0 => no match
678 *********************************************************************/
679 int match_portlist(const char *portlist, int port)
681 char *min, *max, *next, *portlist_copy;
683 min = next = portlist_copy = strdup(portlist);
686 * Zero-terminate first item and remember offset for next
688 if (NULL != (next = strchr(portlist_copy, (int) ',')))
694 * Loop through all items, checking for match
698 if (NULL == (max = strchr(min, (int) '-')))
701 * No dash, check for equality
703 if (port == atoi(min))
712 * This is a range, so check if between min and max,
713 * or, if max was omitted, between min and 65K
716 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
730 * Zero-terminate next item and remember offset for n+1
732 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
744 /*********************************************************************
746 * Function : block_url
748 * Description : Called from `chat'. Check to see if we need to block this.
751 * 1 : csp = Current client state (buffers, headers, etc...)
753 * Returns : NULL => unblocked, else HTTP block response
755 *********************************************************************/
756 struct http_response *block_url(struct client_state *csp)
758 struct http_response *rsp;
759 const char *new_content_type = NULL;
762 * If it's not blocked, don't block it ;-)
764 if ((csp->action->flags & ACTION_BLOCK) == 0)
768 if (csp->action->flags & ACTION_REDIRECT)
770 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
773 * Else, prepare a response
775 if (NULL == (rsp = alloc_http_response()))
777 return cgi_error_memory();
781 * If it's an image-url, send back an image or redirect
782 * as specified by the relevant +image action
784 #ifdef FEATURE_IMAGE_BLOCKING
785 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
789 /* determine HOW images should be blocked */
790 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
792 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
794 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
796 #if 1 /* Two alternative strategies, use this one for now: */
798 /* and handle accordingly: */
799 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
801 rsp->status = strdup("403 Request blocked by Privoxy");
802 if (rsp->status == NULL)
804 free_http_response(rsp);
805 return cgi_error_memory();
807 rsp->body = bindup(image_pattern_data, image_pattern_length);
808 if (rsp->body == NULL)
810 free_http_response(rsp);
811 return cgi_error_memory();
813 rsp->content_length = image_pattern_length;
815 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
817 free_http_response(rsp);
818 return cgi_error_memory();
822 else if (0 == strcmpic(p, "blank"))
824 rsp->status = strdup("403 Request blocked by Privoxy");
825 if (rsp->status == NULL)
827 free_http_response(rsp);
828 return cgi_error_memory();
830 rsp->body = bindup(image_blank_data, image_blank_length);
831 if (rsp->body == NULL)
833 free_http_response(rsp);
834 return cgi_error_memory();
836 rsp->content_length = image_blank_length;
838 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
840 free_http_response(rsp);
841 return cgi_error_memory();
847 rsp->status = strdup("302 Local Redirect from Privoxy");
848 if (rsp->status == NULL)
850 free_http_response(rsp);
851 return cgi_error_memory();
854 if (enlist_unique_header(rsp->headers, "Location", p))
856 free_http_response(rsp);
857 return cgi_error_memory();
861 #else /* Following code is disabled for now */
863 /* and handle accordingly: */
864 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
866 p = CGI_PREFIX "send-banner?type=pattern";
868 else if (0 == strcmpic(p, "blank"))
870 p = CGI_PREFIX "send-banner?type=blank";
872 rsp->status = strdup("302 Local Redirect from Privoxy");
873 if (rsp->status == NULL)
875 free_http_response(rsp);
876 return cgi_error_memory();
879 if (enlist_unique_header(rsp->headers, "Location", p))
881 free_http_response(rsp);
882 return cgi_error_memory();
884 #endif /* Preceeding code is disabled for now */
886 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
889 * Send empty document.
891 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
894 rsp->body = strdup(" ");
895 rsp->content_length = 1;
897 rsp->status = strdup("403 Request blocked by Privoxy");
898 if (rsp->status == NULL)
900 free_http_response(rsp);
901 return cgi_error_memory();
903 if (new_content_type != 0)
905 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
906 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
908 free_http_response(rsp);
909 return cgi_error_memory();
915 #endif /* def FEATURE_IMAGE_BLOCKING */
918 * Else, generate an HTML "blocked" message:
922 struct map * exports;
926 * Workaround for stupid Netscape bug which prevents
927 * pages from being displayed if loading a referenced
928 * JavaScript or style sheet fails. So make it appear
929 * as if it succeeded.
931 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
932 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
933 && !strstr(p, "Gecko") /* save Mozilla, */
934 && !strstr(p, "compatible") /* MSIE */
935 && !strstr(p, "Opera")) /* and Opera. */
937 rsp->status = strdup("200 Request for blocked URL");
941 rsp->status = strdup("404 Request for blocked URL");
944 if (rsp->status == NULL)
946 free_http_response(rsp);
947 return cgi_error_memory();
950 exports = default_exports(csp, NULL);
953 free_http_response(rsp);
954 return cgi_error_memory();
957 #ifdef FEATURE_FORCE_LOAD
958 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
959 if (csp->http->ssl != 0)
960 #endif /* ndef FEATURE_FORCE_LOAD */
962 err = map_block_killer(exports, "force-support");
965 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
966 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
967 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
968 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
973 free_http_response(rsp);
974 return cgi_error_memory();
977 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
980 free_http_response(rsp);
981 return cgi_error_memory();
985 return finish_http_response(rsp);
991 /*********************************************************************
993 * Function : trust_url FIXME: I should be called distrust_url
995 * Description : Calls is_untrusted_url to determine if the URL is trusted
996 * and if not, returns a HTTP 403 response with a reject message.
999 * 1 : csp = Current client state (buffers, headers, etc...)
1001 * Returns : NULL => trusted, else http_response.
1003 *********************************************************************/
1004 struct http_response *trust_url(struct client_state *csp)
1006 struct http_response *rsp;
1007 struct map * exports;
1008 char buf[BUFFER_SIZE];
1010 struct url_spec **tl;
1015 * Don't bother to work on trusted URLs
1017 if (!is_untrusted_url(csp))
1023 * Else, prepare a response:
1025 if (NULL == (rsp = alloc_http_response()))
1027 return cgi_error_memory();
1030 rsp->status = strdup("403 Request blocked by Privoxy");
1031 exports = default_exports(csp, NULL);
1032 if (exports == NULL || rsp->status == NULL)
1034 free_http_response(rsp);
1035 return cgi_error_memory();
1039 * Export the protocol, host, port, and referrer information
1041 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1042 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1043 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1045 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1047 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1051 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1057 free_http_response(rsp);
1058 return cgi_error_memory();
1062 * Export the trust list
1065 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1067 sprintf(buf, "<li>%s</li>\n", t->spec);
1068 string_append(&p, buf);
1070 err = map(exports, "trusted-referrers", 1, p, 0);
1075 free_http_response(rsp);
1076 return cgi_error_memory();
1080 * Export the trust info, if available
1082 if (csp->config->trust_info->first)
1084 struct list_entry *l;
1087 for (l = csp->config->trust_info->first; l ; l = l->next)
1089 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1090 string_append(&p, buf);
1092 err = map(exports, "trust-info", 1, p, 0);
1096 err = map_block_killer(exports, "have-trust-info");
1102 free_http_response(rsp);
1103 return cgi_error_memory();
1107 * Export the force prefix or the force conditional block killer
1109 #ifdef FEATURE_FORCE_LOAD
1110 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1111 #else /* ifndef FEATURE_FORCE_LOAD */
1112 err = map_block_killer(exports, "force-support");
1113 #endif /* ndef FEATURE_FORCE_LOAD */
1118 free_http_response(rsp);
1119 return cgi_error_memory();
1123 * Build the response
1125 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1128 free_http_response(rsp);
1129 return cgi_error_memory();
1132 return finish_http_response(rsp);
1134 #endif /* def FEATURE_TRUST */
1136 /*********************************************************************
1138 * Function : execute_single_pcrs_command
1140 * Description : Apply single pcrs command to the subject.
1141 * The subject itself is left untouched, memory for the result
1142 * is malloc()ed and it is the caller's responsibility to free
1143 * the result when it's no longer needed.
1146 * 1 : subject = the subject (== original) string
1147 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1148 * 3 : hits = int* for returning the number of modifications
1150 * Returns : NULL in case of errors, otherwise the
1151 * result of the pcrs command.
1153 *********************************************************************/
1154 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1158 char *result = NULL;
1162 assert(pcrs_command);
1165 size = strlen(subject);
1167 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1169 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1170 pcrs_command, error);
1172 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1174 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1181 job = pcrs_free_job(job);
1188 /*********************************************************************
1190 * Function : rewrite_url
1192 * Description : Rewrites a URL with a single pcrs command
1193 * and returns the result if it differs from the
1194 * original and isn't obviously invalid.
1197 * 1 : old_url = URL to rewrite.
1198 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1201 * Returns : NULL if the pcrs_command didn't change the url, or
1202 * the result of the modification.
1204 *********************************************************************/
1205 char *rewrite_url(char *old_url, const char *pcrs_command)
1207 char *new_url = NULL;
1211 assert(pcrs_command);
1213 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1217 log_error(LOG_LEVEL_REDIRECTS,
1218 "pcrs command \"%s\" didn't change \"%s\".",
1219 pcrs_command, old_url, new_url);
1222 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1224 log_error(LOG_LEVEL_ERROR,
1225 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1226 "but the result doesn't look like a valid URL and will be ignored.",
1227 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1232 log_error(LOG_LEVEL_REDIRECTS,
1233 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1234 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1242 #ifdef FEATURE_FAST_REDIRECTS
1243 /*********************************************************************
1245 * Function : get_last_url
1247 * Description : Search for the last URL inside a string.
1248 * If the string already is a URL, it will
1249 * be the first URL found.
1252 * 1 : subject = the string to check
1253 * 2 : redirect_mode = +fast-redirect{} mode
1255 * Returns : NULL if no URL was found, or
1256 * the last URL found.
1258 *********************************************************************/
1259 char *get_last_url(char *subject, const char *redirect_mode)
1265 assert(redirect_mode);
1267 subject = strdup(subject);
1269 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1271 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1272 new_url = url_decode(subject);
1277 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1280 * Find the last URL encoded in the request
1283 while ((tmp = strstr(tmp, "http://")) != NULL)
1288 while ((tmp = strstr(tmp, "https://")) != NULL)
1293 if ((new_url != subject)
1294 || (0 == strncmpic(subject, "http://", 7))
1295 || (0 == strncmpic(subject, "https://", 8)))
1298 * Return new URL if we found a redirect
1299 * or if the subject already was a URL.
1301 * The second case makes sure that we can
1302 * chain get_last_url after another redirection check
1303 * (like rewrite_url) without losing earlier redirects.
1305 new_url = strdup(new_url);
1314 #endif /* def FEATURE_FAST_REDIRECTS */
1317 /*********************************************************************
1319 * Function : redirect_url
1321 * Description : Checks if Privoxy should answer the request with
1322 * a HTTP redirect and generates the redirect if
1326 * 1 : csp = Current client state (buffers, headers, etc...)
1328 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1330 *********************************************************************/
1331 struct http_response *redirect_url(struct client_state *csp)
1333 struct http_response *rsp;
1334 #ifdef FEATURE_FAST_REDIRECTS
1336 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1337 * as compile-time option? The user can easily disable
1338 * it in his action file.
1340 char * redirect_mode;
1341 #endif /* def FEATURE_FAST_REDIRECTS */
1342 char *old_url = NULL;
1343 char *new_url = NULL;
1344 char *redirection_string;
1346 if ((csp->action->flags & ACTION_REDIRECT))
1348 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1351 * If the redirection string begins with 's',
1352 * assume it's a pcrs command, otherwise treat it as
1353 * properly formatted URL and use it for the redirection
1356 * According to RFC 2616 section 14.30 the URL
1357 * has to be absolute and if the user tries:
1358 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1359 * she would get undefined results anyway.
1363 if (*redirection_string == 's')
1365 old_url = csp->http->url;
1366 new_url = rewrite_url(old_url, redirection_string);
1370 log_error(LOG_LEVEL_REDIRECTS,
1371 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1372 redirection_string);
1373 new_url = strdup(redirection_string);
1377 #ifdef FEATURE_FAST_REDIRECTS
1378 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1380 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1383 * If it exists, use the previously rewritten URL as input
1384 * otherwise just use the old path.
1386 old_url = new_url ? new_url : strdup(csp->http->path);
1387 new_url = get_last_url(old_url, redirect_mode);
1390 #endif /* def FEATURE_FAST_REDIRECTS */
1392 /* Did any redirect action trigger? */
1395 if (0 == strcmpic(new_url, csp->http->url))
1397 log_error(LOG_LEVEL_ERROR,
1398 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1399 csp->http->url, new_url);
1404 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1406 if (NULL == (rsp = alloc_http_response()))
1409 return cgi_error_memory();
1412 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1413 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1416 free_http_response(rsp);
1417 return cgi_error_memory();
1420 return finish_http_response(rsp);
1424 /* Only reached if no redirect is required */
1430 #ifdef FEATURE_IMAGE_BLOCKING
1431 /*********************************************************************
1433 * Function : is_imageurl
1435 * Description : Given a URL, decide whether it is an image or not,
1436 * using either the info from a previous +image action
1437 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1438 * is MSIE and not on a Mac, tell from the browser's accept
1442 * 1 : csp = Current client state (buffers, headers, etc...)
1444 * Returns : True (nonzero) if URL is an image, false (0)
1447 *********************************************************************/
1448 int is_imageurl(struct client_state *csp)
1450 #ifdef FEATURE_IMAGE_DETECT_MSIE
1453 tmp = get_header_value(csp->headers, "User-Agent:");
1454 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1456 tmp = get_header_value(csp->headers, "Accept:");
1457 if (tmp && strstr(tmp, "image/gif"))
1459 /* Client will accept HTML. If this seems counterintuitive,
1469 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1471 return ((csp->action->flags & ACTION_IMAGE) != 0);
1474 #endif /* def FEATURE_IMAGE_BLOCKING */
1477 #ifdef FEATURE_TRUST
1478 /*********************************************************************
1480 * Function : is_untrusted_url
1482 * Description : Should we "distrust" this URL (and block it)?
1484 * Yes if it matches a line in the trustfile, or if the
1485 * referrer matches a line starting with "+" in the
1490 * 1 : csp = Current client state (buffers, headers, etc...)
1492 * Returns : 0 => trusted, 1 => untrusted
1494 *********************************************************************/
1495 int is_untrusted_url(struct client_state *csp)
1497 struct file_list *fl;
1498 struct block_spec *b;
1499 struct url_spec **trusted_url;
1500 struct http_request rhttp[1];
1501 const char * referer;
1505 * If we don't have a trustlist, we trust everybody
1507 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1512 memset(rhttp, '\0', sizeof(*rhttp));
1515 * Do we trust the request URL itself?
1517 for (b = b->next; b ; b = b->next)
1519 if (url_match(b->url, csp->http))
1525 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1527 /* no referrer was supplied */
1533 * If not, do we maybe trust its referrer?
1535 err = parse_http_url(referer, rhttp, csp);
1541 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1543 if (url_match(*trusted_url, rhttp))
1545 /* if the URL's referrer is from a trusted referrer, then
1546 * add the target spec to the trustfile as an unblocked
1547 * domain and return NULL (which means it's OK).
1552 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1556 char * new_entry = strdup("~");
1558 string_append(&new_entry, csp->http->hostport);
1560 path = csp->http->path;
1561 if ( (path[0] == '/')
1563 && ((path_end = strchr(path + 2, '/')) != NULL))
1565 /* since this path points into a user's home space
1566 * be sure to include this spec in the trustfile.
1568 int path_len = path_end - path; /* save offset */
1569 path = strdup(path); /* Copy string */
1572 path_end = path + path_len; /* regenerate ptr to new buffer */
1573 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1575 string_join(&new_entry, path);
1578 if (new_entry != NULL)
1580 fprintf(fp, "%s\n", new_entry);
1585 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1586 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1596 #endif /* def FEATURE_TRUST */
1599 /*********************************************************************
1601 * Function : pcrs_filter_response
1603 * Description : Execute all text substitutions from all applying
1604 * +filter actions on the text buffer that's been accumulated
1605 * in csp->iob->buf. If this changes the contents, set
1606 * csp->content_length to the modified size and raise the
1607 * CSP_FLAG_MODIFIED flag.
1610 * 1 : csp = Current client state (buffers, headers, etc...)
1612 * Returns : a pointer to the (newly allocated) modified buffer.
1613 * or NULL if there were no hits or something went wrong
1615 *********************************************************************/
1616 char *pcrs_filter_response(struct client_state *csp)
1619 size_t size, prev_size;
1621 char *old = csp->iob->cur, *new = NULL;
1624 struct file_list *fl;
1625 struct re_filterfile_spec *b;
1626 struct list_entry *filtername;
1628 int i, found_filters = 0;
1633 if (csp->iob->cur >= csp->iob->eod)
1637 size = csp->iob->eod - csp->iob->cur;
1640 * Need to check the set of re_filterfiles...
1642 for (i = 0; i < MAX_AF_FILES; i++)
1655 if (0 == found_filters)
1657 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1662 * If the body has a "chunked" transfer-encoding,
1663 * get rid of it first, adjusting size and iob->eod
1665 if (csp->flags & CSP_FLAG_CHUNKED)
1667 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1668 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1672 csp->iob->eod = csp->iob->cur + size;
1673 csp->flags |= CSP_FLAG_MODIFIED;
1676 for (i = 0; i < MAX_AF_FILES; i++)
1679 if ((NULL == fl) || (NULL == fl->f))
1682 * For all applying +filter actions, look if a filter by that
1683 * name exists and if yes, execute it's pcrs_joblist on the
1686 for (b = fl->f; b; b = b->next)
1688 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1689 filtername ; filtername = filtername->next)
1691 if (strcmp(b->name, filtername->str) == 0)
1693 int current_hits = 0;
1695 if ( NULL == b->joblist )
1697 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1702 /* Apply all jobs from the joblist */
1703 for (job = b->joblist; NULL != job; job = job->next)
1705 current_hits += pcrs_execute(job, old, size, &new, &size);
1706 if (old != csp->iob->cur) free(old);
1710 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1711 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1713 hits += current_hits;
1720 * If there were no hits, destroy our copy and let
1721 * chat() use the original in csp->iob
1729 csp->flags |= CSP_FLAG_MODIFIED;
1730 csp->content_length = size;
1738 /*********************************************************************
1740 * Function : gif_deanimate_response
1742 * Description : Deanimate the GIF image that has been accumulated in
1743 * csp->iob->buf, set csp->content_length to the modified
1744 * size and raise the CSP_FLAG_MODIFIED flag.
1747 * 1 : csp = Current client state (buffers, headers, etc...)
1749 * Returns : a pointer to the (newly allocated) modified buffer.
1750 * or NULL in case something went wrong.
1752 *********************************************************************/
1753 char *gif_deanimate_response(struct client_state *csp)
1755 struct binbuffer *in, *out;
1757 size_t size = csp->iob->eod - csp->iob->cur;
1760 * If the body has a "chunked" transfer-encoding,
1761 * get rid of it first, adjusting size and iob->eod
1763 if (csp->flags & CSP_FLAG_CHUNKED)
1765 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1766 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1770 csp->iob->eod = csp->iob->cur + size;
1771 csp->flags |= CSP_FLAG_MODIFIED;
1774 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1775 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1777 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1781 in->buffer = csp->iob->cur;
1784 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1786 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1793 if ((int)size == out->offset)
1795 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1799 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1801 csp->content_length = out->offset;
1802 csp->flags |= CSP_FLAG_MODIFIED;
1812 /*********************************************************************
1814 * Function : jpeg_inspect_response
1819 * 1 : csp = Current client state (buffers, headers, etc...)
1821 * Returns : a pointer to the (newly allocated) modified buffer
1822 * or NULL in case something went wrong.
1824 *********************************************************************/
1825 char *jpeg_inspect_response(struct client_state *csp)
1827 struct binbuffer *in = NULL, *out = NULL;
1829 size_t size = csp->iob->eod - csp->iob->cur;
1832 * If the body has a "chunked" transfer-encoding,
1833 * get rid of it first, adjusting size and iob->eod
1835 if (csp->flags & CSP_FLAG_CHUNKED)
1837 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1838 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1842 csp->iob->eod = csp->iob->cur + size;
1843 csp->flags |= CSP_FLAG_MODIFIED;
1846 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1848 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1852 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1854 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1858 in->buffer = csp->iob->cur;
1862 * Calling jpeg_inspect has the side-effect of creating and
1863 * modifying the image buffer of "out" directly.
1865 if (jpeg_inspect(in, out))
1867 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1875 csp->content_length = out->offset;
1876 csp->flags |= CSP_FLAG_MODIFIED;
1886 /*********************************************************************
1888 * Function : remove_chunked_transfer_coding
1890 * Description : In-situ remove the "chunked" transfer coding as defined
1891 * in rfc2616 from a buffer.
1894 * 1 : buffer = Pointer to the text buffer
1895 * 2 : size = Number of bytes to be processed
1897 * Returns : The new size, i.e. the number of bytes from buffer which
1898 * are occupied by the stripped body, or 0 in case something
1901 *********************************************************************/
1902 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1905 unsigned int chunksize = 0;
1906 char *from_p, *to_p;
1909 from_p = to_p = buffer;
1911 if (sscanf(buffer, "%x", &chunksize) != 1)
1913 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1917 while (chunksize > 0)
1919 if (NULL == (from_p = strstr(from_p, "\r\n")))
1921 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1925 if ((newsize += chunksize) >= size)
1927 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1932 memmove(to_p, from_p, (size_t) chunksize);
1933 to_p = buffer + newsize;
1934 from_p += chunksize + 2;
1936 if (sscanf(from_p, "%x", &chunksize) != 1)
1938 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1943 /* FIXME: Should this get its own loglevel? */
1944 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1950 /*********************************************************************
1952 * Function : url_actions
1954 * Description : Gets the actions for this URL.
1957 * 1 : http = http_request request for blocked URLs
1958 * 2 : csp = Current client state (buffers, headers, etc...)
1962 *********************************************************************/
1963 void url_actions(struct http_request *http,
1964 struct client_state *csp)
1966 struct file_list *fl;
1967 struct url_actions *b;
1970 init_current_action(csp->action);
1972 for (i = 0; i < MAX_AF_FILES; i++)
1974 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1979 apply_url_actions(csp->action, http, b);
1986 /*********************************************************************
1988 * Function : apply_url_actions
1990 * Description : Applies a list of URL actions.
1993 * 1 : action = Destination.
1994 * 2 : http = Current URL
1995 * 3 : b = list of URL actions to apply
1999 *********************************************************************/
2000 void apply_url_actions(struct current_action_spec *action,
2001 struct http_request *http,
2002 struct url_actions *b)
2006 /* Should never happen */
2010 for (b = b->next; NULL != b; b = b->next)
2012 if (url_match(b->url, http))
2014 merge_current_action(action, b->action);
2020 /*********************************************************************
2022 * Function : forward_url
2024 * Description : Should we forward this to another proxy?
2027 * 1 : http = http_request request for current URL
2028 * 2 : csp = Current client state (buffers, headers, etc...)
2030 * Returns : Pointer to forwarding information.
2032 *********************************************************************/
2033 const struct forward_spec * forward_url(struct http_request *http,
2034 struct client_state *csp)
2036 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2037 struct forward_spec *fwd = csp->config->forward;
2046 if (url_match(fwd->url, http))
2057 /*********************************************************************
2059 * Function : direct_response
2061 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2062 * request and if so, return a HTTP 501 to the client.
2064 * FIXME: I have a stupid name and I should handle the
2065 * requests properly. Still, what we do here is rfc-
2066 * compliant, whereas ignoring or forwarding are not.
2069 * 1 : csp = Current client state (buffers, headers, etc...)
2071 * Returns : http_response if , NULL if nonmatch or handler fail
2073 *********************************************************************/
2074 struct http_response *direct_response(struct client_state *csp)
2076 struct http_response *rsp;
2077 struct list_entry *p;
2079 if ((0 == strcmpic(csp->http->gpc, "trace"))
2080 || (0 == strcmpic(csp->http->gpc, "options")))
2082 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2084 if (!strncmp("Max-Forwards:", p->str, 13)
2085 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
2087 /* FIXME: We could handle at least TRACE here,
2088 but that would require a verbatim copy of
2089 the request which we don't have anymore */
2091 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
2093 /* Get mem for response or fail*/
2094 if (NULL == (rsp = alloc_http_response()))
2096 return cgi_error_memory();
2099 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2101 free_http_response(rsp);
2102 return cgi_error_memory();
2106 return(finish_http_response(rsp));