1 const char filters_rcs[] = "$Id: filters.c,v 1.70 2006/12/09 13:33:15 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 1.70 2006/12/09 13:33:15 fabiankeil
44 * Added some sanity checks for get_last_url().
45 * Fixed possible segfault caused by my last commit.
47 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
48 * Let get_last_url() catch https URLs as well.
50 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
51 * Make sure get_last_url() behaves like advertised
52 * and fast-redirects{} can be combined with redirect{}.
54 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
55 * Implemented +redirect{s@foo@bar@} to generate
56 * a redirect based on a rewritten version of the
59 * Revision 1.66 2006/09/23 13:26:38 roro
60 * Replace TABs by spaces in source code.
62 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
63 * Fix +redirect{}. Didn't work with -fast-redirects.
65 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
66 * Block requests for untrusted URLs with status
67 * code 403 instead of 200.
69 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
70 * Don't free p which is still in use and will be later
71 * freed by free_map(). Don't claim the referrer is unknown
72 * when the client didn't set one.
74 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
75 * Feature request 595948: Re-Filter logging in single line
77 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
78 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
80 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
81 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
82 * with what was really the latest development (the v_3_0_branch branch)
84 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
85 * Multiple filter file support
87 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
88 * A few memory leaks plugged (mostly on error paths)
90 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
91 * Add the ability to check jpeg images for invalid
92 * lengths of comment blocks. Defensive strategy
93 * against the exploit:
94 * Microsoft Security Bulletin MS04-028
95 * Buffer Overrun in JPEG Processing (GDI+) Could
96 * Allow Code Execution (833987)
97 * Enabled with +inspect-jpegs in actions files.
99 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
100 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
102 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
103 * Fixed bug #839859: "See why" link URL now gets url-encoded.
105 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
108 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
109 * Added basic support for OPTIONS and TRACE HTTP methods:
110 * New function direct_response which handles OPTIONS and
111 * TRACE requests whose Max-Forwards header field is zero.
113 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
114 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
116 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
117 * - Bugfix: Executing a filters without jobs no longer results in
118 * turing off *all* filters.
119 * - Security fix: Malicious web servers can't cause a seg fault
120 * through bogus chunk sizes anymore
122 * Revision 1.58 2002/04/24 02:11:17 oes
123 * Jon's multiple AF patch: url_actions now evaluates rules
126 * Revision 1.57 2002/04/08 20:38:34 swa
129 * Revision 1.56 2002/04/05 15:51:24 oes
130 * - bugfix: error-pages now get correct request protocol
131 * - fix for invalid HTML in trust info
133 * Revision 1.55 2002/04/02 16:13:51 oes
134 * Fix: No "Go there anyway" for SSL
136 * Revision 1.54 2002/04/02 14:55:56 oes
137 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
139 * Revision 1.53 2002/03/26 22:29:54 swa
140 * we have a new homepage!
142 * Revision 1.52 2002/03/24 16:35:57 jongfoster
145 * Revision 1.51 2002/03/24 15:23:33 jongfoster
148 * Revision 1.50 2002/03/24 13:25:43 swa
149 * name change related issues
151 * Revision 1.49 2002/03/16 20:29:14 oes
154 * Revision 1.48 2002/03/13 20:25:34 oes
155 * Better logging for content filters
157 * Revision 1.47 2002/03/13 00:30:52 jongfoster
159 * Added option of always sending redirect for imageblock,
160 * currently disabled with #if 0.
162 * Revision 1.46 2002/03/12 01:42:49 oes
163 * Introduced modular filters
165 * Revision 1.45 2002/03/08 16:47:50 oes
166 * Added choice beween GIF and PNG built-in images
168 * Revision 1.44 2002/03/07 03:49:31 oes
169 * - Fixed compiler warnings etc
170 * - Changed built-in images from GIF to PNG
171 * (with regard to Unisys patent issue)
172 * - Added a 4x4 pattern PNG which is less intrusive
173 * than the logo but also clearly marks the deleted banners
175 * Revision 1.43 2002/01/22 23:51:59 jongfoster
176 * Replacing strsav() with the safer string_append().
178 * Adding missing html_encode() to error message generators. Where encoded
179 * and unencoded versions of a string were provided, removing the unencoded
182 * Revision 1.42 2002/01/17 21:00:32 jongfoster
183 * Moving all our URL and URL pattern parsing code to urlmatch.c.
185 * Using a single, simple url_match(pattern,url) function - rather than
186 * the 3-line match routine which was repeated all over the place.
188 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
190 * Using parse_http_url() to parse URLs without faking a HTTP
191 * request line for parse_http_request().
193 * Revision 1.41 2001/11/13 00:14:07 jongfoster
194 * Fixing stupid bug now I've figured out what || means.
195 * (It always returns 0 or 1, not one of it's paramaters.)
197 * Revision 1.40 2001/10/26 17:37:55 oes
198 * - Re-enabled Netscape 200/404 bug workaround in block_url():
199 * - Removed OS/2 special case
200 * - Made block_url() independant from sed() having been run
201 * - Made trust_url independant from sed() having been run
202 * - Made is_imageurl independant from sed() having been run.
203 * It now checks User-Agent: and Accept: by itself.
206 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
207 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
208 * threads to call select() simultaneously. So, it's time to do a real, live,
209 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
210 * (native). Both versions will work, but using __OS2__ offers multi-threading.
212 * Revision 1.38 2001/10/23 21:32:33 jongfoster
213 * Adding error-checking to selected functions
215 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
216 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
217 * filters.c. Added a FIXME in front of the offending code. I'll gladly
218 * put in a better/more robust fix for all parties if one is presented...
219 * It seems that just returning 200 instead of 404 would pretty much fix
220 * it for everyone, but I don't know all the history of the problem.
222 * Revision 1.36 2001/10/10 16:44:16 oes
223 * Added match_portlist function
225 * Revision 1.35 2001/10/07 15:41:23 oes
226 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
228 * New function remove_chunked_transfer_coding that strips chunked
229 * transfer coding to plain and is called by pcrs_filter_response
230 * and gif_deanimate_response if neccessary
232 * Improved handling of zero-change re_filter runs
234 * pcrs_filter_response and gif_deanimate_response now remove
235 * chunked transfer codeing before processing the body.
237 * Revision 1.34 2001/09/20 15:49:36 steudten
239 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
240 * See cgi.c fill_template().
242 * Revision 1.33 2001/09/16 17:05:14 jongfoster
243 * Removing unused #include showarg.h
245 * Revision 1.32 2001/09/16 13:21:27 jongfoster
246 * Changes to use new list functions.
248 * Revision 1.31 2001/09/16 11:38:02 jongfoster
249 * Splitting fill_template() into 2 functions:
250 * template_load() loads the file
251 * template_fill() performs the PCRS regexps.
252 * This is because the CGI edit interface has a "table row"
253 * template which is used many times in the page - this
254 * change means it's only loaded from disk once.
256 * Revision 1.30 2001/09/16 11:00:10 jongfoster
257 * New function alloc_http_response, for symmetry with free_http_response
259 * Revision 1.29 2001/09/13 23:32:40 jongfoster
260 * Moving image data to cgi.c rather than cgi.h
261 * Fixing a GPF under Win32 (and any other OS that protects global
262 * constants from being written to).
264 * Revision 1.28 2001/09/10 10:18:51 oes
265 * Silenced compiler warnings
267 * Revision 1.27 2001/08/05 16:06:20 jongfoster
268 * Modifiying "struct map" so that there are now separate header and
269 * "map_entry" structures. This means that functions which modify a
270 * map no longer need to return a pointer to the modified map.
271 * Also, it no longer reverses the order of the entries (which may be
272 * important with some advanced template substitutions).
274 * Revision 1.26 2001/07/30 22:08:36 jongfoster
275 * Tidying up #defines:
276 * - All feature #defines are now of the form FEATURE_xxx
277 * - Permanently turned off WIN_GUI_EDIT
278 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
280 * Revision 1.25 2001/07/26 10:09:46 oes
281 * Made browser detection a little less naive
283 * Revision 1.24 2001/07/25 17:22:51 oes
284 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
286 * Revision 1.23 2001/07/23 13:40:12 oes
287 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
289 * Revision 1.22 2001/07/18 12:29:34 oes
290 * - Made gif_deanimate_response respect
291 * csp->action->string[ACTION_STRING_DEANIMATE]
292 * - Logging cosmetics
294 * Revision 1.21 2001/07/13 13:59:53 oes
295 * - Introduced gif_deanimate_response which shares the
296 * generic content modification interface of pcrs_filter_response
297 * and acts as a wrapper to deanimate.c:gif_deanimate()
298 * - Renamed re_process_buffer to pcrs_filter_response
299 * - pcrs_filter_response now returns NULL on failiure
300 * - Removed all #ifdef PCRS
302 * Revision 1.20 2001/07/01 17:01:04 oes
303 * Added comments and missing return statement in is_untrusted_url()
305 * Revision 1.19 2001/06/29 21:45:41 oes
306 * Indentation, CRLF->LF, Tab-> Space
308 * Revision 1.18 2001/06/29 13:27:38 oes
309 * - Cleaned up, renamed and reorderd functions
310 * and improved comments
313 * - Ported to CGI platform. Now delivers
314 * http_response or NULL
315 * - Unified HTML and GIF generation (moved image detection
316 * and GIF generation here from jcc.c:chat())
317 * - Fixed HTTP status to:
318 * - 403 (Forbidden) for the "blocked" HTML message
319 * - 200 (OK) for GIF answers
320 * - 302 (Redirect) for redirect to GIF
323 * - Ported to CGI platform. Now delivers
324 * http_response or NULL
325 * - Separated detection of untrusted URL into
326 * (bool)is_untrusted_url
327 * - Added enforcement of untrusted requests
329 * - Moved redirect_url() from cgi.c to here
330 * and ported it to the CGI platform
332 * - Removed logentry from cancelled commit
334 * Revision 1.17 2001/06/09 10:55:28 jongfoster
335 * Changing BUFSIZ ==> BUFFER_SIZE
337 * Revision 1.16 2001/06/07 23:10:26 jongfoster
338 * Allowing unanchored domain patterns to back off and retry
339 * if they partially match. Optimized right-anchored patterns.
340 * Moving ACL and forward files into config file.
341 * Replacing struct gateway with struct forward_spec
343 * Revision 1.15 2001/06/03 19:12:00 oes
344 * extracted-CGI relevant stuff
346 * Revision 1.14 2001/06/01 10:30:55 oes
347 * Added optional left-anchoring to domaincmp
349 * Revision 1.13 2001/05/31 21:21:30 jongfoster
350 * Permissionsfile / actions file changes:
351 * - Changed "permission" to "action" throughout
352 * - changes to file format to allow string parameters
353 * - Moved helper functions to actions.c
355 * Revision 1.12 2001/05/31 17:35:20 oes
357 * - Enhanced domain part globbing with infix and prefix asterisk
358 * matching and optional unanchored operation
360 * Revision 1.11 2001/05/29 11:53:23 oes
361 * "See why" link added to "blocked" page
363 * Revision 1.10 2001/05/29 09:50:24 jongfoster
364 * Unified blocklist/imagelist/permissionslist.
365 * File format is still under discussion, but the internal changes
368 * Also modified interceptor behaviour:
369 * - We now intercept all URLs beginning with one of the following
370 * prefixes (and *only* these prefixes):
372 * * http://ijbswa.sf.net/config/
373 * * http://ijbswa.sourceforge.net/config/
374 * - New interceptors "home page" - go to http://i.j.b/ to see it.
375 * - Internal changes so that intercepted and fast redirect pages
376 * are not replaced with an image.
377 * - Interceptors now have the option to send a binary page direct
378 * to the client. (i.e. ijb-send-banner uses this)
379 * - Implemented show-url-info interceptor. (Which is why I needed
380 * the above interceptors changes - a typical URL is
381 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
382 * The previous mechanism would not have intercepted that, and
383 * if it had been intercepted then it then it would have replaced
386 * Revision 1.9 2001/05/27 22:17:04 oes
388 * - re_process_buffer no longer writes the modified buffer
389 * to the client, which was very ugly. It now returns the
390 * buffer, which it is then written by chat.
392 * - content_length now adjusts the Content-Length: header
393 * for modified documents rather than crunch()ing it.
394 * (Length info in csp->content_length, which is 0 for
395 * unmodified documents)
397 * - For this to work, sed() is called twice when filtering.
399 * Revision 1.8 2001/05/26 17:13:28 jongfoster
400 * Filled in a function comment.
402 * Revision 1.7 2001/05/26 15:26:15 jongfoster
403 * ACL feature now provides more security by immediately dropping
404 * connections from untrusted hosts.
406 * Revision 1.6 2001/05/26 00:28:36 jongfoster
407 * Automatic reloading of config file.
408 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
409 * Most of the global variables have been moved to a new
410 * struct configuration_spec, accessed through csp->config->globalname
411 * Most of the globals remaining are used by the Win32 GUI.
413 * Revision 1.5 2001/05/25 22:34:30 jongfoster
416 * Revision 1.4 2001/05/22 18:46:04 oes
418 * - Enabled filtering banners by size rather than URL
419 * by adding patterns that replace all standard banner
420 * sizes with the "Junkbuster" gif to the re_filterfile
422 * - Enabled filtering WebBugs by providing a pattern
423 * which kills all 1x1 images
425 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
426 * which is selected by the (nonstandard and therefore
427 * capital) letter 'U' in the option string.
428 * It causes the quantifiers to be ungreedy by default.
429 * Appending a ? turns back to greedy (!).
431 * - Added a new interceptor ijb-send-banner, which
432 * sends back the "Junkbuster" gif. Without imagelist or
433 * MSIE detection support, or if tinygif = 1, or the
434 * URL isn't recognized as an imageurl, a lame HTML
435 * explanation is sent instead.
437 * - Added new feature, which permits blocking remote
438 * script redirects and firing back a local redirect
440 * The feature is conditionally compiled, i.e. it
441 * can be disabled with --disable-fast-redirects,
442 * plus it must be activated by a "fast-redirects"
443 * line in the config file, has its own log level
444 * and of course wants to be displayed by show-proxy-args
445 * Note: Boy, all the #ifdefs in 1001 locations and
446 * all the fumbling with configure.in and acconfig.h
447 * were *way* more work than the feature itself :-(
449 * - Because a generic redirect template was needed for
450 * this, tinygif = 3 now uses the same.
452 * - Moved GIFs, and other static HTTP response templates
457 * - Removed some >400 CRs again (Jon, you really worked
460 * Revision 1.3 2001/05/20 16:44:47 jongfoster
461 * Removing last hardcoded Junkbusters.com URLs.
463 * Revision 1.2 2001/05/20 01:21:20 jongfoster
464 * Version 2.9.4 checkin.
465 * - Merged popupfile and cookiefile, and added control over PCRS
466 * filtering, in new "permissionsfile".
467 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
468 * file error you now get a message box (in the Win32 GUI) rather
469 * than the program exiting with no explanation.
470 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
472 * - Removed tabs from "config"
473 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
474 * - Bumped up version number.
476 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
477 * Initial import of version 2.9.3 source tree
480 *********************************************************************/
486 #include <sys/types.h>
495 #endif /* ndef __OS2__ */
496 #include <netinet/in.h>
498 #include <winsock2.h>
499 #endif /* ndef _WIN32 */
503 #endif /* def __OS2__ */
511 #include "jbsockets.h"
512 #include "miscutil.h"
516 #include "deanimate.h"
517 #include "urlmatch.h"
523 const char filters_h_rcs[] = FILTERS_H_VERSION;
525 /* Fix a problem with Solaris. There should be no effect on other
527 * Solaris's isspace() is a macro which uses it's argument directly
528 * as an array index. Therefore we need to make sure that high-bit
529 * characters generate +ve values, and ideally we also want to make
530 * the argument match the declared parameter type of "int".
532 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
536 /*********************************************************************
538 * Function : block_acl
540 * Description : Block this request?
541 * Decide yes or no based on ACL file.
544 * 1 : dst = The proxy or gateway address this is going to.
545 * Or NULL to check all possible targets.
546 * 2 : csp = Current client state (buffers, headers, etc...)
547 * Also includes the client IP address.
549 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
551 *********************************************************************/
552 int block_acl(struct access_control_addr *dst, struct client_state *csp)
554 struct access_control_list *acl = csp->config->acl;
556 /* if not using an access control list, then permit the connection */
562 /* search the list */
565 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
569 /* Just want to check if they have any access */
570 if (acl->action == ACL_PERMIT)
575 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
576 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
578 if (acl->action == ACL_PERMIT)
596 /*********************************************************************
598 * Function : acl_addr
600 * Description : Called from `load_config' to parse an ACL address.
603 * 1 : aspec = String specifying ACL address.
604 * 2 : aca = struct access_control_addr to fill in.
606 * Returns : 0 => Ok, everything else is an error.
608 *********************************************************************/
609 int acl_addr(char *aspec, struct access_control_addr *aca)
611 int i, masklength, port;
617 if ((p = strchr(aspec, '/')) != NULL)
621 if (ijb_isdigit(*p) == 0)
625 masklength = atoi(p);
628 if ((masklength < 0) || (masklength > 32))
633 if ((p = strchr(aspec, ':')) != NULL)
637 if (ijb_isdigit(*p) == 0)
646 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
648 if (aca->addr == INADDR_NONE)
653 /* build the netmask */
655 for (i=1; i <= masklength ; i++)
657 aca->mask |= (1 << (32 - i));
660 /* now mask off the host portion of the ip address
661 * (i.e. save on the network portion of the address).
663 aca->addr = aca->addr & aca->mask;
668 #endif /* def FEATURE_ACL */
671 /*********************************************************************
673 * Function : match_portlist
675 * Description : Check if a given number is covered by a comma
676 * separated list of numbers and ranges (a,b-c,d,..)
679 * 1 : portlist = String with list
680 * 2 : port = port to check
682 * Returns : 0 => no match
685 *********************************************************************/
686 int match_portlist(const char *portlist, int port)
688 char *min, *max, *next, *portlist_copy;
690 min = next = portlist_copy = strdup(portlist);
693 * Zero-terminate first item and remember offset for next
695 if (NULL != (next = strchr(portlist_copy, (int) ',')))
701 * Loop through all items, checking for match
705 if (NULL == (max = strchr(min, (int) '-')))
708 * No dash, check for equality
710 if (port == atoi(min))
719 * This is a range, so check if between min and max,
720 * or, if max was omitted, between min and 65K
723 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
737 * Zero-terminate next item and remember offset for n+1
739 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
751 /*********************************************************************
753 * Function : block_url
755 * Description : Called from `chat'. Check to see if we need to block this.
758 * 1 : csp = Current client state (buffers, headers, etc...)
760 * Returns : NULL => unblocked, else HTTP block response
762 *********************************************************************/
763 struct http_response *block_url(struct client_state *csp)
765 struct http_response *rsp;
766 const char *new_content_type = NULL;
769 * If it's not blocked, don't block it ;-)
771 if ((csp->action->flags & ACTION_BLOCK) == 0)
775 if (csp->action->flags & ACTION_REDIRECT)
777 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
780 * Else, prepare a response
782 if (NULL == (rsp = alloc_http_response()))
784 return cgi_error_memory();
788 * If it's an image-url, send back an image or redirect
789 * as specified by the relevant +image action
791 #ifdef FEATURE_IMAGE_BLOCKING
792 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
796 /* determine HOW images should be blocked */
797 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
799 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
801 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
803 #if 1 /* Two alternative strategies, use this one for now: */
805 /* and handle accordingly: */
806 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
808 rsp->status = strdup("403 Request blocked by Privoxy");
809 if (rsp->status == NULL)
811 free_http_response(rsp);
812 return cgi_error_memory();
814 rsp->body = bindup(image_pattern_data, image_pattern_length);
815 if (rsp->body == NULL)
817 free_http_response(rsp);
818 return cgi_error_memory();
820 rsp->content_length = image_pattern_length;
822 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
824 free_http_response(rsp);
825 return cgi_error_memory();
829 else if (0 == strcmpic(p, "blank"))
831 rsp->status = strdup("403 Request blocked by Privoxy");
832 if (rsp->status == NULL)
834 free_http_response(rsp);
835 return cgi_error_memory();
837 rsp->body = bindup(image_blank_data, image_blank_length);
838 if (rsp->body == NULL)
840 free_http_response(rsp);
841 return cgi_error_memory();
843 rsp->content_length = image_blank_length;
845 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
847 free_http_response(rsp);
848 return cgi_error_memory();
854 rsp->status = strdup("302 Local Redirect from Privoxy");
855 if (rsp->status == NULL)
857 free_http_response(rsp);
858 return cgi_error_memory();
861 if (enlist_unique_header(rsp->headers, "Location", p))
863 free_http_response(rsp);
864 return cgi_error_memory();
868 #else /* Following code is disabled for now */
870 /* and handle accordingly: */
871 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
873 p = CGI_PREFIX "send-banner?type=pattern";
875 else if (0 == strcmpic(p, "blank"))
877 p = CGI_PREFIX "send-banner?type=blank";
879 rsp->status = strdup("302 Local Redirect from Privoxy");
880 if (rsp->status == NULL)
882 free_http_response(rsp);
883 return cgi_error_memory();
886 if (enlist_unique_header(rsp->headers, "Location", p))
888 free_http_response(rsp);
889 return cgi_error_memory();
891 #endif /* Preceeding code is disabled for now */
893 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
896 * Send empty document.
898 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
901 rsp->body = strdup(" ");
902 rsp->content_length = 1;
904 rsp->status = strdup("403 Request blocked by Privoxy");
905 if (rsp->status == NULL)
907 free_http_response(rsp);
908 return cgi_error_memory();
910 if (new_content_type != 0)
912 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
913 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
915 free_http_response(rsp);
916 return cgi_error_memory();
922 #endif /* def FEATURE_IMAGE_BLOCKING */
925 * Else, generate an HTML "blocked" message:
929 struct map * exports;
933 * Workaround for stupid Netscape bug which prevents
934 * pages from being displayed if loading a referenced
935 * JavaScript or style sheet fails. So make it appear
936 * as if it succeeded.
938 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
939 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
940 && !strstr(p, "Gecko") /* save Mozilla, */
941 && !strstr(p, "compatible") /* MSIE */
942 && !strstr(p, "Opera")) /* and Opera. */
944 rsp->status = strdup("200 Request for blocked URL");
948 rsp->status = strdup("404 Request for blocked URL");
951 if (rsp->status == NULL)
953 free_http_response(rsp);
954 return cgi_error_memory();
957 exports = default_exports(csp, NULL);
960 free_http_response(rsp);
961 return cgi_error_memory();
964 #ifdef FEATURE_FORCE_LOAD
965 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
966 if (csp->http->ssl != 0)
967 #endif /* ndef FEATURE_FORCE_LOAD */
969 err = map_block_killer(exports, "force-support");
972 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
973 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
974 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
975 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
980 free_http_response(rsp);
981 return cgi_error_memory();
984 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
987 free_http_response(rsp);
988 return cgi_error_memory();
992 return finish_http_response(rsp);
998 /*********************************************************************
1000 * Function : trust_url FIXME: I should be called distrust_url
1002 * Description : Calls is_untrusted_url to determine if the URL is trusted
1003 * and if not, returns a HTTP 403 response with a reject message.
1006 * 1 : csp = Current client state (buffers, headers, etc...)
1008 * Returns : NULL => trusted, else http_response.
1010 *********************************************************************/
1011 struct http_response *trust_url(struct client_state *csp)
1013 struct http_response *rsp;
1014 struct map * exports;
1015 char buf[BUFFER_SIZE];
1017 struct url_spec **tl;
1022 * Don't bother to work on trusted URLs
1024 if (!is_untrusted_url(csp))
1030 * Else, prepare a response:
1032 if (NULL == (rsp = alloc_http_response()))
1034 return cgi_error_memory();
1037 rsp->status = strdup("403 Request blocked by Privoxy");
1038 exports = default_exports(csp, NULL);
1039 if (exports == NULL || rsp->status == NULL)
1041 free_http_response(rsp);
1042 return cgi_error_memory();
1046 * Export the protocol, host, port, and referrer information
1048 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1049 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1050 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1052 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1054 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1058 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1064 free_http_response(rsp);
1065 return cgi_error_memory();
1069 * Export the trust list
1072 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1074 sprintf(buf, "<li>%s</li>\n", t->spec);
1075 string_append(&p, buf);
1077 err = map(exports, "trusted-referrers", 1, p, 0);
1082 free_http_response(rsp);
1083 return cgi_error_memory();
1087 * Export the trust info, if available
1089 if (csp->config->trust_info->first)
1091 struct list_entry *l;
1094 for (l = csp->config->trust_info->first; l ; l = l->next)
1096 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1097 string_append(&p, buf);
1099 err = map(exports, "trust-info", 1, p, 0);
1103 err = map_block_killer(exports, "have-trust-info");
1109 free_http_response(rsp);
1110 return cgi_error_memory();
1114 * Export the force prefix or the force conditional block killer
1116 #ifdef FEATURE_FORCE_LOAD
1117 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1118 #else /* ifndef FEATURE_FORCE_LOAD */
1119 err = map_block_killer(exports, "force-support");
1120 #endif /* ndef FEATURE_FORCE_LOAD */
1125 free_http_response(rsp);
1126 return cgi_error_memory();
1130 * Build the response
1132 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1135 free_http_response(rsp);
1136 return cgi_error_memory();
1139 return finish_http_response(rsp);
1141 #endif /* def FEATURE_TRUST */
1143 /*********************************************************************
1145 * Function : execute_single_pcrs_command
1147 * Description : Apply single pcrs command to the subject.
1148 * The subject itself is left untouched, memory for the result
1149 * is malloc()ed and it is the caller's responsibility to free
1150 * the result when it's no longer needed.
1153 * 1 : subject = the subject (== original) string
1154 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1155 * 3 : hits = int* for returning the number of modifications
1157 * Returns : NULL in case of errors, otherwise the
1158 * result of the pcrs command.
1160 *********************************************************************/
1161 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1165 char *result = NULL;
1169 assert(pcrs_command);
1172 size = strlen(subject);
1174 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1176 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1177 pcrs_command, error);
1179 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1181 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1188 job = pcrs_free_job(job);
1195 /*********************************************************************
1197 * Function : rewrite_url
1199 * Description : Rewrites a URL with a single pcrs command
1200 * and returns the result if it differs from the
1201 * original and isn't obviously invalid.
1204 * 1 : old_url = URL to rewrite.
1205 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1208 * Returns : NULL if the pcrs_command didn't change the url, or
1209 * the result of the modification.
1211 *********************************************************************/
1212 char *rewrite_url(char *old_url, const char *pcrs_command)
1214 char *new_url = NULL;
1218 assert(pcrs_command);
1220 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1224 log_error(LOG_LEVEL_REDIRECTS,
1225 "pcrs command \"%s\" didn't change \"%s\".",
1226 pcrs_command, old_url, new_url);
1229 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1231 log_error(LOG_LEVEL_ERROR,
1232 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1233 "but the result doesn't look like a valid URL and will be ignored.",
1234 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1239 log_error(LOG_LEVEL_REDIRECTS,
1240 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1241 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1249 #ifdef FEATURE_FAST_REDIRECTS
1250 /*********************************************************************
1252 * Function : get_last_url
1254 * Description : Search for the last URL inside a string.
1255 * If the string already is a URL, it will
1256 * be the first URL found.
1259 * 1 : subject = the string to check
1260 * 2 : redirect_mode = +fast-redirect{} mode
1262 * Returns : NULL if no URL was found, or
1263 * the last URL found.
1265 *********************************************************************/
1266 char *get_last_url(char *subject, const char *redirect_mode)
1268 char *new_url = NULL;
1272 assert(redirect_mode);
1274 subject = strdup(subject);
1275 if (subject == NULL)
1277 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1281 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1283 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1284 new_url = url_decode(subject);
1285 if (new_url != NULL)
1292 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1296 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1299 * Find the last URL encoded in the request
1302 while ((tmp = strstr(tmp, "http://")) != NULL)
1306 tmp = (new_url != NULL) ? new_url : subject;
1307 while ((tmp = strstr(tmp, "https://")) != NULL)
1312 if ((new_url != NULL)
1313 && ( (new_url != subject)
1314 || (0 == strncmpic(subject, "http://", 7))
1315 || (0 == strncmpic(subject, "https://", 8))
1319 * Return new URL if we found a redirect
1320 * or if the subject already was a URL.
1322 * The second case makes sure that we can
1323 * chain get_last_url after another redirection check
1324 * (like rewrite_url) without losing earlier redirects.
1326 new_url = strdup(new_url);
1335 #endif /* def FEATURE_FAST_REDIRECTS */
1338 /*********************************************************************
1340 * Function : redirect_url
1342 * Description : Checks if Privoxy should answer the request with
1343 * a HTTP redirect and generates the redirect if
1347 * 1 : csp = Current client state (buffers, headers, etc...)
1349 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1351 *********************************************************************/
1352 struct http_response *redirect_url(struct client_state *csp)
1354 struct http_response *rsp;
1355 #ifdef FEATURE_FAST_REDIRECTS
1357 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1358 * as compile-time option? The user can easily disable
1359 * it in his action file.
1361 char * redirect_mode;
1362 #endif /* def FEATURE_FAST_REDIRECTS */
1363 char *old_url = NULL;
1364 char *new_url = NULL;
1365 char *redirection_string;
1367 if ((csp->action->flags & ACTION_REDIRECT))
1369 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1372 * If the redirection string begins with 's',
1373 * assume it's a pcrs command, otherwise treat it as
1374 * properly formatted URL and use it for the redirection
1377 * According to RFC 2616 section 14.30 the URL
1378 * has to be absolute and if the user tries:
1379 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1380 * she would get undefined results anyway.
1384 if (*redirection_string == 's')
1386 old_url = csp->http->url;
1387 new_url = rewrite_url(old_url, redirection_string);
1391 log_error(LOG_LEVEL_REDIRECTS,
1392 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1393 redirection_string);
1394 new_url = strdup(redirection_string);
1398 #ifdef FEATURE_FAST_REDIRECTS
1399 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1401 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1404 * If it exists, use the previously rewritten URL as input
1405 * otherwise just use the old path.
1407 old_url = new_url ? new_url : strdup(csp->http->path);
1408 new_url = get_last_url(old_url, redirect_mode);
1411 #endif /* def FEATURE_FAST_REDIRECTS */
1413 /* Did any redirect action trigger? */
1416 if (0 == strcmpic(new_url, csp->http->url))
1418 log_error(LOG_LEVEL_ERROR,
1419 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1420 csp->http->url, new_url);
1425 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1427 if (NULL == (rsp = alloc_http_response()))
1430 return cgi_error_memory();
1433 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1434 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1437 free_http_response(rsp);
1438 return cgi_error_memory();
1441 return finish_http_response(rsp);
1445 /* Only reached if no redirect is required */
1451 #ifdef FEATURE_IMAGE_BLOCKING
1452 /*********************************************************************
1454 * Function : is_imageurl
1456 * Description : Given a URL, decide whether it is an image or not,
1457 * using either the info from a previous +image action
1458 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1459 * is MSIE and not on a Mac, tell from the browser's accept
1463 * 1 : csp = Current client state (buffers, headers, etc...)
1465 * Returns : True (nonzero) if URL is an image, false (0)
1468 *********************************************************************/
1469 int is_imageurl(struct client_state *csp)
1471 #ifdef FEATURE_IMAGE_DETECT_MSIE
1474 tmp = get_header_value(csp->headers, "User-Agent:");
1475 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1477 tmp = get_header_value(csp->headers, "Accept:");
1478 if (tmp && strstr(tmp, "image/gif"))
1480 /* Client will accept HTML. If this seems counterintuitive,
1490 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1492 return ((csp->action->flags & ACTION_IMAGE) != 0);
1495 #endif /* def FEATURE_IMAGE_BLOCKING */
1498 #ifdef FEATURE_TRUST
1499 /*********************************************************************
1501 * Function : is_untrusted_url
1503 * Description : Should we "distrust" this URL (and block it)?
1505 * Yes if it matches a line in the trustfile, or if the
1506 * referrer matches a line starting with "+" in the
1511 * 1 : csp = Current client state (buffers, headers, etc...)
1513 * Returns : 0 => trusted, 1 => untrusted
1515 *********************************************************************/
1516 int is_untrusted_url(struct client_state *csp)
1518 struct file_list *fl;
1519 struct block_spec *b;
1520 struct url_spec **trusted_url;
1521 struct http_request rhttp[1];
1522 const char * referer;
1526 * If we don't have a trustlist, we trust everybody
1528 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1533 memset(rhttp, '\0', sizeof(*rhttp));
1536 * Do we trust the request URL itself?
1538 for (b = b->next; b ; b = b->next)
1540 if (url_match(b->url, csp->http))
1546 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1548 /* no referrer was supplied */
1554 * If not, do we maybe trust its referrer?
1556 err = parse_http_url(referer, rhttp, csp);
1562 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1564 if (url_match(*trusted_url, rhttp))
1566 /* if the URL's referrer is from a trusted referrer, then
1567 * add the target spec to the trustfile as an unblocked
1568 * domain and return NULL (which means it's OK).
1573 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1577 char * new_entry = strdup("~");
1579 string_append(&new_entry, csp->http->hostport);
1581 path = csp->http->path;
1582 if ( (path[0] == '/')
1584 && ((path_end = strchr(path + 2, '/')) != NULL))
1586 /* since this path points into a user's home space
1587 * be sure to include this spec in the trustfile.
1589 int path_len = path_end - path; /* save offset */
1590 path = strdup(path); /* Copy string */
1593 path_end = path + path_len; /* regenerate ptr to new buffer */
1594 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1596 string_join(&new_entry, path);
1599 if (new_entry != NULL)
1601 fprintf(fp, "%s\n", new_entry);
1606 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1607 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1617 #endif /* def FEATURE_TRUST */
1620 /*********************************************************************
1622 * Function : pcrs_filter_response
1624 * Description : Execute all text substitutions from all applying
1625 * +filter actions on the text buffer that's been accumulated
1626 * in csp->iob->buf. If this changes the contents, set
1627 * csp->content_length to the modified size and raise the
1628 * CSP_FLAG_MODIFIED flag.
1631 * 1 : csp = Current client state (buffers, headers, etc...)
1633 * Returns : a pointer to the (newly allocated) modified buffer.
1634 * or NULL if there were no hits or something went wrong
1636 *********************************************************************/
1637 char *pcrs_filter_response(struct client_state *csp)
1640 size_t size, prev_size;
1642 char *old = csp->iob->cur, *new = NULL;
1645 struct file_list *fl;
1646 struct re_filterfile_spec *b;
1647 struct list_entry *filtername;
1649 int i, found_filters = 0;
1654 if (csp->iob->cur >= csp->iob->eod)
1658 size = csp->iob->eod - csp->iob->cur;
1661 * Need to check the set of re_filterfiles...
1663 for (i = 0; i < MAX_AF_FILES; i++)
1676 if (0 == found_filters)
1678 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1683 * If the body has a "chunked" transfer-encoding,
1684 * get rid of it first, adjusting size and iob->eod
1686 if (csp->flags & CSP_FLAG_CHUNKED)
1688 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1689 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1693 csp->iob->eod = csp->iob->cur + size;
1694 csp->flags |= CSP_FLAG_MODIFIED;
1697 for (i = 0; i < MAX_AF_FILES; i++)
1700 if ((NULL == fl) || (NULL == fl->f))
1703 * Either there are no filter files
1704 * left, or this filter file just
1705 * contains no valid filters.
1707 * Continue to be sure we don't miss
1708 * valid filter files that are chained
1709 * after empty or invalid ones.
1714 * For all applying +filter actions, look if a filter by that
1715 * name exists and if yes, execute it's pcrs_joblist on the
1718 for (b = fl->f; b; b = b->next)
1720 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1721 filtername ; filtername = filtername->next)
1723 if (strcmp(b->name, filtername->str) == 0)
1725 int current_hits = 0;
1727 if ( NULL == b->joblist )
1729 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1734 /* Apply all jobs from the joblist */
1735 for (job = b->joblist; NULL != job; job = job->next)
1737 current_hits += pcrs_execute(job, old, size, &new, &size);
1738 if (old != csp->iob->cur) free(old);
1742 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1743 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1745 hits += current_hits;
1752 * If there were no hits, destroy our copy and let
1753 * chat() use the original in csp->iob
1761 csp->flags |= CSP_FLAG_MODIFIED;
1762 csp->content_length = size;
1770 /*********************************************************************
1772 * Function : gif_deanimate_response
1774 * Description : Deanimate the GIF image that has been accumulated in
1775 * csp->iob->buf, set csp->content_length to the modified
1776 * size and raise the CSP_FLAG_MODIFIED flag.
1779 * 1 : csp = Current client state (buffers, headers, etc...)
1781 * Returns : a pointer to the (newly allocated) modified buffer.
1782 * or NULL in case something went wrong.
1784 *********************************************************************/
1785 char *gif_deanimate_response(struct client_state *csp)
1787 struct binbuffer *in, *out;
1789 size_t size = csp->iob->eod - csp->iob->cur;
1792 * If the body has a "chunked" transfer-encoding,
1793 * get rid of it first, adjusting size and iob->eod
1795 if (csp->flags & CSP_FLAG_CHUNKED)
1797 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1798 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1802 csp->iob->eod = csp->iob->cur + size;
1803 csp->flags |= CSP_FLAG_MODIFIED;
1806 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1807 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1809 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1813 in->buffer = csp->iob->cur;
1816 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1818 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1825 if ((int)size == out->offset)
1827 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1831 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1833 csp->content_length = out->offset;
1834 csp->flags |= CSP_FLAG_MODIFIED;
1844 /*********************************************************************
1846 * Function : jpeg_inspect_response
1851 * 1 : csp = Current client state (buffers, headers, etc...)
1853 * Returns : a pointer to the (newly allocated) modified buffer
1854 * or NULL in case something went wrong.
1856 *********************************************************************/
1857 char *jpeg_inspect_response(struct client_state *csp)
1859 struct binbuffer *in = NULL, *out = NULL;
1861 size_t size = csp->iob->eod - csp->iob->cur;
1864 * If the body has a "chunked" transfer-encoding,
1865 * get rid of it first, adjusting size and iob->eod
1867 if (csp->flags & CSP_FLAG_CHUNKED)
1869 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1870 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1874 csp->iob->eod = csp->iob->cur + size;
1875 csp->flags |= CSP_FLAG_MODIFIED;
1878 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1880 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1884 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1886 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1890 in->buffer = csp->iob->cur;
1894 * Calling jpeg_inspect has the side-effect of creating and
1895 * modifying the image buffer of "out" directly.
1897 if (jpeg_inspect(in, out))
1899 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1907 csp->content_length = out->offset;
1908 csp->flags |= CSP_FLAG_MODIFIED;
1918 /*********************************************************************
1920 * Function : remove_chunked_transfer_coding
1922 * Description : In-situ remove the "chunked" transfer coding as defined
1923 * in rfc2616 from a buffer.
1926 * 1 : buffer = Pointer to the text buffer
1927 * 2 : size = Number of bytes to be processed
1929 * Returns : The new size, i.e. the number of bytes from buffer which
1930 * are occupied by the stripped body, or 0 in case something
1933 *********************************************************************/
1934 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1937 unsigned int chunksize = 0;
1938 char *from_p, *to_p;
1941 from_p = to_p = buffer;
1943 if (sscanf(buffer, "%x", &chunksize) != 1)
1945 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1949 while (chunksize > 0)
1951 if (NULL == (from_p = strstr(from_p, "\r\n")))
1953 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1957 if ((newsize += chunksize) >= size)
1959 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1964 memmove(to_p, from_p, (size_t) chunksize);
1965 to_p = buffer + newsize;
1966 from_p += chunksize + 2;
1968 if (sscanf(from_p, "%x", &chunksize) != 1)
1970 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1975 /* FIXME: Should this get its own loglevel? */
1976 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1982 /*********************************************************************
1984 * Function : url_actions
1986 * Description : Gets the actions for this URL.
1989 * 1 : http = http_request request for blocked URLs
1990 * 2 : csp = Current client state (buffers, headers, etc...)
1994 *********************************************************************/
1995 void url_actions(struct http_request *http,
1996 struct client_state *csp)
1998 struct file_list *fl;
1999 struct url_actions *b;
2002 init_current_action(csp->action);
2004 for (i = 0; i < MAX_AF_FILES; i++)
2006 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2011 apply_url_actions(csp->action, http, b);
2018 /*********************************************************************
2020 * Function : apply_url_actions
2022 * Description : Applies a list of URL actions.
2025 * 1 : action = Destination.
2026 * 2 : http = Current URL
2027 * 3 : b = list of URL actions to apply
2031 *********************************************************************/
2032 void apply_url_actions(struct current_action_spec *action,
2033 struct http_request *http,
2034 struct url_actions *b)
2038 /* Should never happen */
2042 for (b = b->next; NULL != b; b = b->next)
2044 if (url_match(b->url, http))
2046 merge_current_action(action, b->action);
2052 /*********************************************************************
2054 * Function : forward_url
2056 * Description : Should we forward this to another proxy?
2059 * 1 : http = http_request request for current URL
2060 * 2 : csp = Current client state (buffers, headers, etc...)
2062 * Returns : Pointer to forwarding information.
2064 *********************************************************************/
2065 const struct forward_spec * forward_url(struct http_request *http,
2066 struct client_state *csp)
2068 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2069 struct forward_spec *fwd = csp->config->forward;
2078 if (url_match(fwd->url, http))
2089 /*********************************************************************
2091 * Function : direct_response
2093 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2094 * request and if so, return a HTTP 501 to the client.
2096 * FIXME: I have a stupid name and I should handle the
2097 * requests properly. Still, what we do here is rfc-
2098 * compliant, whereas ignoring or forwarding are not.
2101 * 1 : csp = Current client state (buffers, headers, etc...)
2103 * Returns : http_response if , NULL if nonmatch or handler fail
2105 *********************************************************************/
2106 struct http_response *direct_response(struct client_state *csp)
2108 struct http_response *rsp;
2109 struct list_entry *p;
2111 if ((0 == strcmpic(csp->http->gpc, "trace"))
2112 || (0 == strcmpic(csp->http->gpc, "options")))
2114 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2116 if (!strncmp("Max-Forwards:", p->str, 13)
2117 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
2119 /* FIXME: We could handle at least TRACE here,
2120 but that would require a verbatim copy of
2121 the request which we don't have anymore */
2123 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
2125 /* Get mem for response or fail*/
2126 if (NULL == (rsp = alloc_http_response()))
2128 return cgi_error_memory();
2131 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2133 free_http_response(rsp);
2134 return cgi_error_memory();
2138 return(finish_http_response(rsp));