1 const char filters_rcs[] = "$Id: filters.c,v 1.73 2006/12/23 16:01:02 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 1.73 2006/12/23 16:01:02 fabiankeil
44 * Don't crash if pcre returns an error code
45 * that pcrs didn't expect. Fixes BR 1621173.
47 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
48 * Modified is_untrusted_url to complain in case of
49 * write errors and to give a reason when adding new
50 * entries to the trustfile. Closes FR 1097611.
52 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
53 * Skip empty filter files in pcrs_filter_response,
54 * but don't ignore the ones that come afterwards.
55 * Fixes parts of BR 1619208.
57 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
58 * Added some sanity checks for get_last_url().
59 * Fixed possible segfault caused by my last commit.
61 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
62 * Let get_last_url() catch https URLs as well.
64 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
65 * Make sure get_last_url() behaves like advertised
66 * and fast-redirects{} can be combined with redirect{}.
68 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
69 * Implemented +redirect{s@foo@bar@} to generate
70 * a redirect based on a rewritten version of the
73 * Revision 1.66 2006/09/23 13:26:38 roro
74 * Replace TABs by spaces in source code.
76 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
77 * Fix +redirect{}. Didn't work with -fast-redirects.
79 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
80 * Block requests for untrusted URLs with status
81 * code 403 instead of 200.
83 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
84 * Don't free p which is still in use and will be later
85 * freed by free_map(). Don't claim the referrer is unknown
86 * when the client didn't set one.
88 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
89 * Feature request 595948: Re-Filter logging in single line
91 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
92 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
94 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
95 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
96 * with what was really the latest development (the v_3_0_branch branch)
98 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
99 * Multiple filter file support
101 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
102 * A few memory leaks plugged (mostly on error paths)
104 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
105 * Add the ability to check jpeg images for invalid
106 * lengths of comment blocks. Defensive strategy
107 * against the exploit:
108 * Microsoft Security Bulletin MS04-028
109 * Buffer Overrun in JPEG Processing (GDI+) Could
110 * Allow Code Execution (833987)
111 * Enabled with +inspect-jpegs in actions files.
113 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
114 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
116 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
117 * Fixed bug #839859: "See why" link URL now gets url-encoded.
119 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
122 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
123 * Added basic support for OPTIONS and TRACE HTTP methods:
124 * New function direct_response which handles OPTIONS and
125 * TRACE requests whose Max-Forwards header field is zero.
127 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
128 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
130 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
131 * - Bugfix: Executing a filters without jobs no longer results in
132 * turing off *all* filters.
133 * - Security fix: Malicious web servers can't cause a seg fault
134 * through bogus chunk sizes anymore
136 * Revision 1.58 2002/04/24 02:11:17 oes
137 * Jon's multiple AF patch: url_actions now evaluates rules
140 * Revision 1.57 2002/04/08 20:38:34 swa
143 * Revision 1.56 2002/04/05 15:51:24 oes
144 * - bugfix: error-pages now get correct request protocol
145 * - fix for invalid HTML in trust info
147 * Revision 1.55 2002/04/02 16:13:51 oes
148 * Fix: No "Go there anyway" for SSL
150 * Revision 1.54 2002/04/02 14:55:56 oes
151 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
153 * Revision 1.53 2002/03/26 22:29:54 swa
154 * we have a new homepage!
156 * Revision 1.52 2002/03/24 16:35:57 jongfoster
159 * Revision 1.51 2002/03/24 15:23:33 jongfoster
162 * Revision 1.50 2002/03/24 13:25:43 swa
163 * name change related issues
165 * Revision 1.49 2002/03/16 20:29:14 oes
168 * Revision 1.48 2002/03/13 20:25:34 oes
169 * Better logging for content filters
171 * Revision 1.47 2002/03/13 00:30:52 jongfoster
173 * Added option of always sending redirect for imageblock,
174 * currently disabled with #if 0.
176 * Revision 1.46 2002/03/12 01:42:49 oes
177 * Introduced modular filters
179 * Revision 1.45 2002/03/08 16:47:50 oes
180 * Added choice beween GIF and PNG built-in images
182 * Revision 1.44 2002/03/07 03:49:31 oes
183 * - Fixed compiler warnings etc
184 * - Changed built-in images from GIF to PNG
185 * (with regard to Unisys patent issue)
186 * - Added a 4x4 pattern PNG which is less intrusive
187 * than the logo but also clearly marks the deleted banners
189 * Revision 1.43 2002/01/22 23:51:59 jongfoster
190 * Replacing strsav() with the safer string_append().
192 * Adding missing html_encode() to error message generators. Where encoded
193 * and unencoded versions of a string were provided, removing the unencoded
196 * Revision 1.42 2002/01/17 21:00:32 jongfoster
197 * Moving all our URL and URL pattern parsing code to urlmatch.c.
199 * Using a single, simple url_match(pattern,url) function - rather than
200 * the 3-line match routine which was repeated all over the place.
202 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
204 * Using parse_http_url() to parse URLs without faking a HTTP
205 * request line for parse_http_request().
207 * Revision 1.41 2001/11/13 00:14:07 jongfoster
208 * Fixing stupid bug now I've figured out what || means.
209 * (It always returns 0 or 1, not one of it's paramaters.)
211 * Revision 1.40 2001/10/26 17:37:55 oes
212 * - Re-enabled Netscape 200/404 bug workaround in block_url():
213 * - Removed OS/2 special case
214 * - Made block_url() independant from sed() having been run
215 * - Made trust_url independant from sed() having been run
216 * - Made is_imageurl independant from sed() having been run.
217 * It now checks User-Agent: and Accept: by itself.
220 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
221 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
222 * threads to call select() simultaneously. So, it's time to do a real, live,
223 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
224 * (native). Both versions will work, but using __OS2__ offers multi-threading.
226 * Revision 1.38 2001/10/23 21:32:33 jongfoster
227 * Adding error-checking to selected functions
229 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
230 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
231 * filters.c. Added a FIXME in front of the offending code. I'll gladly
232 * put in a better/more robust fix for all parties if one is presented...
233 * It seems that just returning 200 instead of 404 would pretty much fix
234 * it for everyone, but I don't know all the history of the problem.
236 * Revision 1.36 2001/10/10 16:44:16 oes
237 * Added match_portlist function
239 * Revision 1.35 2001/10/07 15:41:23 oes
240 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
242 * New function remove_chunked_transfer_coding that strips chunked
243 * transfer coding to plain and is called by pcrs_filter_response
244 * and gif_deanimate_response if neccessary
246 * Improved handling of zero-change re_filter runs
248 * pcrs_filter_response and gif_deanimate_response now remove
249 * chunked transfer codeing before processing the body.
251 * Revision 1.34 2001/09/20 15:49:36 steudten
253 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
254 * See cgi.c fill_template().
256 * Revision 1.33 2001/09/16 17:05:14 jongfoster
257 * Removing unused #include showarg.h
259 * Revision 1.32 2001/09/16 13:21:27 jongfoster
260 * Changes to use new list functions.
262 * Revision 1.31 2001/09/16 11:38:02 jongfoster
263 * Splitting fill_template() into 2 functions:
264 * template_load() loads the file
265 * template_fill() performs the PCRS regexps.
266 * This is because the CGI edit interface has a "table row"
267 * template which is used many times in the page - this
268 * change means it's only loaded from disk once.
270 * Revision 1.30 2001/09/16 11:00:10 jongfoster
271 * New function alloc_http_response, for symmetry with free_http_response
273 * Revision 1.29 2001/09/13 23:32:40 jongfoster
274 * Moving image data to cgi.c rather than cgi.h
275 * Fixing a GPF under Win32 (and any other OS that protects global
276 * constants from being written to).
278 * Revision 1.28 2001/09/10 10:18:51 oes
279 * Silenced compiler warnings
281 * Revision 1.27 2001/08/05 16:06:20 jongfoster
282 * Modifiying "struct map" so that there are now separate header and
283 * "map_entry" structures. This means that functions which modify a
284 * map no longer need to return a pointer to the modified map.
285 * Also, it no longer reverses the order of the entries (which may be
286 * important with some advanced template substitutions).
288 * Revision 1.26 2001/07/30 22:08:36 jongfoster
289 * Tidying up #defines:
290 * - All feature #defines are now of the form FEATURE_xxx
291 * - Permanently turned off WIN_GUI_EDIT
292 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
294 * Revision 1.25 2001/07/26 10:09:46 oes
295 * Made browser detection a little less naive
297 * Revision 1.24 2001/07/25 17:22:51 oes
298 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
300 * Revision 1.23 2001/07/23 13:40:12 oes
301 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
303 * Revision 1.22 2001/07/18 12:29:34 oes
304 * - Made gif_deanimate_response respect
305 * csp->action->string[ACTION_STRING_DEANIMATE]
306 * - Logging cosmetics
308 * Revision 1.21 2001/07/13 13:59:53 oes
309 * - Introduced gif_deanimate_response which shares the
310 * generic content modification interface of pcrs_filter_response
311 * and acts as a wrapper to deanimate.c:gif_deanimate()
312 * - Renamed re_process_buffer to pcrs_filter_response
313 * - pcrs_filter_response now returns NULL on failiure
314 * - Removed all #ifdef PCRS
316 * Revision 1.20 2001/07/01 17:01:04 oes
317 * Added comments and missing return statement in is_untrusted_url()
319 * Revision 1.19 2001/06/29 21:45:41 oes
320 * Indentation, CRLF->LF, Tab-> Space
322 * Revision 1.18 2001/06/29 13:27:38 oes
323 * - Cleaned up, renamed and reorderd functions
324 * and improved comments
327 * - Ported to CGI platform. Now delivers
328 * http_response or NULL
329 * - Unified HTML and GIF generation (moved image detection
330 * and GIF generation here from jcc.c:chat())
331 * - Fixed HTTP status to:
332 * - 403 (Forbidden) for the "blocked" HTML message
333 * - 200 (OK) for GIF answers
334 * - 302 (Redirect) for redirect to GIF
337 * - Ported to CGI platform. Now delivers
338 * http_response or NULL
339 * - Separated detection of untrusted URL into
340 * (bool)is_untrusted_url
341 * - Added enforcement of untrusted requests
343 * - Moved redirect_url() from cgi.c to here
344 * and ported it to the CGI platform
346 * - Removed logentry from cancelled commit
348 * Revision 1.17 2001/06/09 10:55:28 jongfoster
349 * Changing BUFSIZ ==> BUFFER_SIZE
351 * Revision 1.16 2001/06/07 23:10:26 jongfoster
352 * Allowing unanchored domain patterns to back off and retry
353 * if they partially match. Optimized right-anchored patterns.
354 * Moving ACL and forward files into config file.
355 * Replacing struct gateway with struct forward_spec
357 * Revision 1.15 2001/06/03 19:12:00 oes
358 * extracted-CGI relevant stuff
360 * Revision 1.14 2001/06/01 10:30:55 oes
361 * Added optional left-anchoring to domaincmp
363 * Revision 1.13 2001/05/31 21:21:30 jongfoster
364 * Permissionsfile / actions file changes:
365 * - Changed "permission" to "action" throughout
366 * - changes to file format to allow string parameters
367 * - Moved helper functions to actions.c
369 * Revision 1.12 2001/05/31 17:35:20 oes
371 * - Enhanced domain part globbing with infix and prefix asterisk
372 * matching and optional unanchored operation
374 * Revision 1.11 2001/05/29 11:53:23 oes
375 * "See why" link added to "blocked" page
377 * Revision 1.10 2001/05/29 09:50:24 jongfoster
378 * Unified blocklist/imagelist/permissionslist.
379 * File format is still under discussion, but the internal changes
382 * Also modified interceptor behaviour:
383 * - We now intercept all URLs beginning with one of the following
384 * prefixes (and *only* these prefixes):
386 * * http://ijbswa.sf.net/config/
387 * * http://ijbswa.sourceforge.net/config/
388 * - New interceptors "home page" - go to http://i.j.b/ to see it.
389 * - Internal changes so that intercepted and fast redirect pages
390 * are not replaced with an image.
391 * - Interceptors now have the option to send a binary page direct
392 * to the client. (i.e. ijb-send-banner uses this)
393 * - Implemented show-url-info interceptor. (Which is why I needed
394 * the above interceptors changes - a typical URL is
395 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
396 * The previous mechanism would not have intercepted that, and
397 * if it had been intercepted then it then it would have replaced
400 * Revision 1.9 2001/05/27 22:17:04 oes
402 * - re_process_buffer no longer writes the modified buffer
403 * to the client, which was very ugly. It now returns the
404 * buffer, which it is then written by chat.
406 * - content_length now adjusts the Content-Length: header
407 * for modified documents rather than crunch()ing it.
408 * (Length info in csp->content_length, which is 0 for
409 * unmodified documents)
411 * - For this to work, sed() is called twice when filtering.
413 * Revision 1.8 2001/05/26 17:13:28 jongfoster
414 * Filled in a function comment.
416 * Revision 1.7 2001/05/26 15:26:15 jongfoster
417 * ACL feature now provides more security by immediately dropping
418 * connections from untrusted hosts.
420 * Revision 1.6 2001/05/26 00:28:36 jongfoster
421 * Automatic reloading of config file.
422 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
423 * Most of the global variables have been moved to a new
424 * struct configuration_spec, accessed through csp->config->globalname
425 * Most of the globals remaining are used by the Win32 GUI.
427 * Revision 1.5 2001/05/25 22:34:30 jongfoster
430 * Revision 1.4 2001/05/22 18:46:04 oes
432 * - Enabled filtering banners by size rather than URL
433 * by adding patterns that replace all standard banner
434 * sizes with the "Junkbuster" gif to the re_filterfile
436 * - Enabled filtering WebBugs by providing a pattern
437 * which kills all 1x1 images
439 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
440 * which is selected by the (nonstandard and therefore
441 * capital) letter 'U' in the option string.
442 * It causes the quantifiers to be ungreedy by default.
443 * Appending a ? turns back to greedy (!).
445 * - Added a new interceptor ijb-send-banner, which
446 * sends back the "Junkbuster" gif. Without imagelist or
447 * MSIE detection support, or if tinygif = 1, or the
448 * URL isn't recognized as an imageurl, a lame HTML
449 * explanation is sent instead.
451 * - Added new feature, which permits blocking remote
452 * script redirects and firing back a local redirect
454 * The feature is conditionally compiled, i.e. it
455 * can be disabled with --disable-fast-redirects,
456 * plus it must be activated by a "fast-redirects"
457 * line in the config file, has its own log level
458 * and of course wants to be displayed by show-proxy-args
459 * Note: Boy, all the #ifdefs in 1001 locations and
460 * all the fumbling with configure.in and acconfig.h
461 * were *way* more work than the feature itself :-(
463 * - Because a generic redirect template was needed for
464 * this, tinygif = 3 now uses the same.
466 * - Moved GIFs, and other static HTTP response templates
471 * - Removed some >400 CRs again (Jon, you really worked
474 * Revision 1.3 2001/05/20 16:44:47 jongfoster
475 * Removing last hardcoded Junkbusters.com URLs.
477 * Revision 1.2 2001/05/20 01:21:20 jongfoster
478 * Version 2.9.4 checkin.
479 * - Merged popupfile and cookiefile, and added control over PCRS
480 * filtering, in new "permissionsfile".
481 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
482 * file error you now get a message box (in the Win32 GUI) rather
483 * than the program exiting with no explanation.
484 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
486 * - Removed tabs from "config"
487 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
488 * - Bumped up version number.
490 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
491 * Initial import of version 2.9.3 source tree
494 *********************************************************************/
500 #include <sys/types.h>
509 #endif /* ndef __OS2__ */
510 #include <netinet/in.h>
512 #include <winsock2.h>
513 #endif /* ndef _WIN32 */
517 #endif /* def __OS2__ */
525 #include "jbsockets.h"
526 #include "miscutil.h"
530 #include "deanimate.h"
531 #include "urlmatch.h"
537 const char filters_h_rcs[] = FILTERS_H_VERSION;
539 /* Fix a problem with Solaris. There should be no effect on other
541 * Solaris's isspace() is a macro which uses it's argument directly
542 * as an array index. Therefore we need to make sure that high-bit
543 * characters generate +ve values, and ideally we also want to make
544 * the argument match the declared parameter type of "int".
546 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
550 /*********************************************************************
552 * Function : block_acl
554 * Description : Block this request?
555 * Decide yes or no based on ACL file.
558 * 1 : dst = The proxy or gateway address this is going to.
559 * Or NULL to check all possible targets.
560 * 2 : csp = Current client state (buffers, headers, etc...)
561 * Also includes the client IP address.
563 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
565 *********************************************************************/
566 int block_acl(struct access_control_addr *dst, struct client_state *csp)
568 struct access_control_list *acl = csp->config->acl;
570 /* if not using an access control list, then permit the connection */
576 /* search the list */
579 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
583 /* Just want to check if they have any access */
584 if (acl->action == ACL_PERMIT)
589 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
590 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
592 if (acl->action == ACL_PERMIT)
610 /*********************************************************************
612 * Function : acl_addr
614 * Description : Called from `load_config' to parse an ACL address.
617 * 1 : aspec = String specifying ACL address.
618 * 2 : aca = struct access_control_addr to fill in.
620 * Returns : 0 => Ok, everything else is an error.
622 *********************************************************************/
623 int acl_addr(char *aspec, struct access_control_addr *aca)
625 int i, masklength, port;
631 if ((p = strchr(aspec, '/')) != NULL)
635 if (ijb_isdigit(*p) == 0)
639 masklength = atoi(p);
642 if ((masklength < 0) || (masklength > 32))
647 if ((p = strchr(aspec, ':')) != NULL)
651 if (ijb_isdigit(*p) == 0)
660 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
662 if (aca->addr == INADDR_NONE)
667 /* build the netmask */
669 for (i=1; i <= masklength ; i++)
671 aca->mask |= (1 << (32 - i));
674 /* now mask off the host portion of the ip address
675 * (i.e. save on the network portion of the address).
677 aca->addr = aca->addr & aca->mask;
682 #endif /* def FEATURE_ACL */
685 /*********************************************************************
687 * Function : match_portlist
689 * Description : Check if a given number is covered by a comma
690 * separated list of numbers and ranges (a,b-c,d,..)
693 * 1 : portlist = String with list
694 * 2 : port = port to check
696 * Returns : 0 => no match
699 *********************************************************************/
700 int match_portlist(const char *portlist, int port)
702 char *min, *max, *next, *portlist_copy;
704 min = next = portlist_copy = strdup(portlist);
707 * Zero-terminate first item and remember offset for next
709 if (NULL != (next = strchr(portlist_copy, (int) ',')))
715 * Loop through all items, checking for match
719 if (NULL == (max = strchr(min, (int) '-')))
722 * No dash, check for equality
724 if (port == atoi(min))
733 * This is a range, so check if between min and max,
734 * or, if max was omitted, between min and 65K
737 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
751 * Zero-terminate next item and remember offset for n+1
753 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
765 /*********************************************************************
767 * Function : block_url
769 * Description : Called from `chat'. Check to see if we need to block this.
772 * 1 : csp = Current client state (buffers, headers, etc...)
774 * Returns : NULL => unblocked, else HTTP block response
776 *********************************************************************/
777 struct http_response *block_url(struct client_state *csp)
779 struct http_response *rsp;
780 const char *new_content_type = NULL;
783 * If it's not blocked, don't block it ;-)
785 if ((csp->action->flags & ACTION_BLOCK) == 0)
789 if (csp->action->flags & ACTION_REDIRECT)
791 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
794 * Else, prepare a response
796 if (NULL == (rsp = alloc_http_response()))
798 return cgi_error_memory();
802 * If it's an image-url, send back an image or redirect
803 * as specified by the relevant +image action
805 #ifdef FEATURE_IMAGE_BLOCKING
806 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
810 /* determine HOW images should be blocked */
811 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
813 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
815 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
817 #if 1 /* Two alternative strategies, use this one for now: */
819 /* and handle accordingly: */
820 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
822 rsp->status = strdup("403 Request blocked by Privoxy");
823 if (rsp->status == NULL)
825 free_http_response(rsp);
826 return cgi_error_memory();
828 rsp->body = bindup(image_pattern_data, image_pattern_length);
829 if (rsp->body == NULL)
831 free_http_response(rsp);
832 return cgi_error_memory();
834 rsp->content_length = image_pattern_length;
836 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
838 free_http_response(rsp);
839 return cgi_error_memory();
843 else if (0 == strcmpic(p, "blank"))
845 rsp->status = strdup("403 Request blocked by Privoxy");
846 if (rsp->status == NULL)
848 free_http_response(rsp);
849 return cgi_error_memory();
851 rsp->body = bindup(image_blank_data, image_blank_length);
852 if (rsp->body == NULL)
854 free_http_response(rsp);
855 return cgi_error_memory();
857 rsp->content_length = image_blank_length;
859 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
861 free_http_response(rsp);
862 return cgi_error_memory();
868 rsp->status = strdup("302 Local Redirect from Privoxy");
869 if (rsp->status == NULL)
871 free_http_response(rsp);
872 return cgi_error_memory();
875 if (enlist_unique_header(rsp->headers, "Location", p))
877 free_http_response(rsp);
878 return cgi_error_memory();
882 #else /* Following code is disabled for now */
884 /* and handle accordingly: */
885 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
887 p = CGI_PREFIX "send-banner?type=pattern";
889 else if (0 == strcmpic(p, "blank"))
891 p = CGI_PREFIX "send-banner?type=blank";
893 rsp->status = strdup("302 Local Redirect from Privoxy");
894 if (rsp->status == NULL)
896 free_http_response(rsp);
897 return cgi_error_memory();
900 if (enlist_unique_header(rsp->headers, "Location", p))
902 free_http_response(rsp);
903 return cgi_error_memory();
905 #endif /* Preceeding code is disabled for now */
907 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
910 * Send empty document.
912 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
915 rsp->body = strdup(" ");
916 rsp->content_length = 1;
918 rsp->status = strdup("403 Request blocked by Privoxy");
919 if (rsp->status == NULL)
921 free_http_response(rsp);
922 return cgi_error_memory();
924 if (new_content_type != 0)
926 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
927 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
929 free_http_response(rsp);
930 return cgi_error_memory();
936 #endif /* def FEATURE_IMAGE_BLOCKING */
939 * Else, generate an HTML "blocked" message:
943 struct map * exports;
947 * Workaround for stupid Netscape bug which prevents
948 * pages from being displayed if loading a referenced
949 * JavaScript or style sheet fails. So make it appear
950 * as if it succeeded.
952 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
953 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
954 && !strstr(p, "Gecko") /* save Mozilla, */
955 && !strstr(p, "compatible") /* MSIE */
956 && !strstr(p, "Opera")) /* and Opera. */
958 rsp->status = strdup("200 Request for blocked URL");
962 rsp->status = strdup("404 Request for blocked URL");
965 if (rsp->status == NULL)
967 free_http_response(rsp);
968 return cgi_error_memory();
971 exports = default_exports(csp, NULL);
974 free_http_response(rsp);
975 return cgi_error_memory();
978 #ifdef FEATURE_FORCE_LOAD
979 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
980 if (csp->http->ssl != 0)
981 #endif /* ndef FEATURE_FORCE_LOAD */
983 err = map_block_killer(exports, "force-support");
986 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
987 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
988 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
989 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
994 free_http_response(rsp);
995 return cgi_error_memory();
998 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1001 free_http_response(rsp);
1002 return cgi_error_memory();
1006 return finish_http_response(rsp);
1011 #ifdef FEATURE_TRUST
1012 /*********************************************************************
1014 * Function : trust_url FIXME: I should be called distrust_url
1016 * Description : Calls is_untrusted_url to determine if the URL is trusted
1017 * and if not, returns a HTTP 403 response with a reject message.
1020 * 1 : csp = Current client state (buffers, headers, etc...)
1022 * Returns : NULL => trusted, else http_response.
1024 *********************************************************************/
1025 struct http_response *trust_url(struct client_state *csp)
1027 struct http_response *rsp;
1028 struct map * exports;
1029 char buf[BUFFER_SIZE];
1031 struct url_spec **tl;
1036 * Don't bother to work on trusted URLs
1038 if (!is_untrusted_url(csp))
1044 * Else, prepare a response:
1046 if (NULL == (rsp = alloc_http_response()))
1048 return cgi_error_memory();
1051 rsp->status = strdup("403 Request blocked by Privoxy");
1052 exports = default_exports(csp, NULL);
1053 if (exports == NULL || rsp->status == NULL)
1055 free_http_response(rsp);
1056 return cgi_error_memory();
1060 * Export the protocol, host, port, and referrer information
1062 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1063 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1064 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1066 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1068 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1072 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1078 free_http_response(rsp);
1079 return cgi_error_memory();
1083 * Export the trust list
1086 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1088 sprintf(buf, "<li>%s</li>\n", t->spec);
1089 string_append(&p, buf);
1091 err = map(exports, "trusted-referrers", 1, p, 0);
1096 free_http_response(rsp);
1097 return cgi_error_memory();
1101 * Export the trust info, if available
1103 if (csp->config->trust_info->first)
1105 struct list_entry *l;
1108 for (l = csp->config->trust_info->first; l ; l = l->next)
1110 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1111 string_append(&p, buf);
1113 err = map(exports, "trust-info", 1, p, 0);
1117 err = map_block_killer(exports, "have-trust-info");
1123 free_http_response(rsp);
1124 return cgi_error_memory();
1128 * Export the force prefix or the force conditional block killer
1130 #ifdef FEATURE_FORCE_LOAD
1131 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1132 #else /* ifndef FEATURE_FORCE_LOAD */
1133 err = map_block_killer(exports, "force-support");
1134 #endif /* ndef FEATURE_FORCE_LOAD */
1139 free_http_response(rsp);
1140 return cgi_error_memory();
1144 * Build the response
1146 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1149 free_http_response(rsp);
1150 return cgi_error_memory();
1153 return finish_http_response(rsp);
1155 #endif /* def FEATURE_TRUST */
1157 /*********************************************************************
1159 * Function : execute_single_pcrs_command
1161 * Description : Apply single pcrs command to the subject.
1162 * The subject itself is left untouched, memory for the result
1163 * is malloc()ed and it is the caller's responsibility to free
1164 * the result when it's no longer needed.
1167 * 1 : subject = the subject (== original) string
1168 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1169 * 3 : hits = int* for returning the number of modifications
1171 * Returns : NULL in case of errors, otherwise the
1172 * result of the pcrs command.
1174 *********************************************************************/
1175 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1179 char *result = NULL;
1183 assert(pcrs_command);
1186 size = strlen(subject);
1188 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1190 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1191 pcrs_command, error);
1193 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1195 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1202 job = pcrs_free_job(job);
1209 /*********************************************************************
1211 * Function : rewrite_url
1213 * Description : Rewrites a URL with a single pcrs command
1214 * and returns the result if it differs from the
1215 * original and isn't obviously invalid.
1218 * 1 : old_url = URL to rewrite.
1219 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1222 * Returns : NULL if the pcrs_command didn't change the url, or
1223 * the result of the modification.
1225 *********************************************************************/
1226 char *rewrite_url(char *old_url, const char *pcrs_command)
1228 char *new_url = NULL;
1232 assert(pcrs_command);
1234 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1238 log_error(LOG_LEVEL_REDIRECTS,
1239 "pcrs command \"%s\" didn't change \"%s\".",
1240 pcrs_command, old_url, new_url);
1243 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1245 log_error(LOG_LEVEL_ERROR,
1246 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1247 "but the result doesn't look like a valid URL and will be ignored.",
1248 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1253 log_error(LOG_LEVEL_REDIRECTS,
1254 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1255 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1263 #ifdef FEATURE_FAST_REDIRECTS
1264 /*********************************************************************
1266 * Function : get_last_url
1268 * Description : Search for the last URL inside a string.
1269 * If the string already is a URL, it will
1270 * be the first URL found.
1273 * 1 : subject = the string to check
1274 * 2 : redirect_mode = +fast-redirect{} mode
1276 * Returns : NULL if no URL was found, or
1277 * the last URL found.
1279 *********************************************************************/
1280 char *get_last_url(char *subject, const char *redirect_mode)
1282 char *new_url = NULL;
1286 assert(redirect_mode);
1288 subject = strdup(subject);
1289 if (subject == NULL)
1291 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1295 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1297 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1298 new_url = url_decode(subject);
1299 if (new_url != NULL)
1306 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1310 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1313 * Find the last URL encoded in the request
1316 while ((tmp = strstr(tmp, "http://")) != NULL)
1320 tmp = (new_url != NULL) ? new_url : subject;
1321 while ((tmp = strstr(tmp, "https://")) != NULL)
1326 if ((new_url != NULL)
1327 && ( (new_url != subject)
1328 || (0 == strncmpic(subject, "http://", 7))
1329 || (0 == strncmpic(subject, "https://", 8))
1333 * Return new URL if we found a redirect
1334 * or if the subject already was a URL.
1336 * The second case makes sure that we can
1337 * chain get_last_url after another redirection check
1338 * (like rewrite_url) without losing earlier redirects.
1340 new_url = strdup(new_url);
1349 #endif /* def FEATURE_FAST_REDIRECTS */
1352 /*********************************************************************
1354 * Function : redirect_url
1356 * Description : Checks if Privoxy should answer the request with
1357 * a HTTP redirect and generates the redirect if
1361 * 1 : csp = Current client state (buffers, headers, etc...)
1363 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1365 *********************************************************************/
1366 struct http_response *redirect_url(struct client_state *csp)
1368 struct http_response *rsp;
1369 #ifdef FEATURE_FAST_REDIRECTS
1371 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1372 * as compile-time option? The user can easily disable
1373 * it in his action file.
1375 char * redirect_mode;
1376 #endif /* def FEATURE_FAST_REDIRECTS */
1377 char *old_url = NULL;
1378 char *new_url = NULL;
1379 char *redirection_string;
1381 if ((csp->action->flags & ACTION_REDIRECT))
1383 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1386 * If the redirection string begins with 's',
1387 * assume it's a pcrs command, otherwise treat it as
1388 * properly formatted URL and use it for the redirection
1391 * According to RFC 2616 section 14.30 the URL
1392 * has to be absolute and if the user tries:
1393 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1394 * she would get undefined results anyway.
1398 if (*redirection_string == 's')
1400 old_url = csp->http->url;
1401 new_url = rewrite_url(old_url, redirection_string);
1405 log_error(LOG_LEVEL_REDIRECTS,
1406 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1407 redirection_string);
1408 new_url = strdup(redirection_string);
1412 #ifdef FEATURE_FAST_REDIRECTS
1413 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1415 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1418 * If it exists, use the previously rewritten URL as input
1419 * otherwise just use the old path.
1421 old_url = new_url ? new_url : strdup(csp->http->path);
1422 new_url = get_last_url(old_url, redirect_mode);
1425 #endif /* def FEATURE_FAST_REDIRECTS */
1427 /* Did any redirect action trigger? */
1430 if (0 == strcmpic(new_url, csp->http->url))
1432 log_error(LOG_LEVEL_ERROR,
1433 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1434 csp->http->url, new_url);
1439 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1441 if (NULL == (rsp = alloc_http_response()))
1444 return cgi_error_memory();
1447 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1448 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1451 free_http_response(rsp);
1452 return cgi_error_memory();
1455 return finish_http_response(rsp);
1459 /* Only reached if no redirect is required */
1465 #ifdef FEATURE_IMAGE_BLOCKING
1466 /*********************************************************************
1468 * Function : is_imageurl
1470 * Description : Given a URL, decide whether it is an image or not,
1471 * using either the info from a previous +image action
1472 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1473 * is MSIE and not on a Mac, tell from the browser's accept
1477 * 1 : csp = Current client state (buffers, headers, etc...)
1479 * Returns : True (nonzero) if URL is an image, false (0)
1482 *********************************************************************/
1483 int is_imageurl(struct client_state *csp)
1485 #ifdef FEATURE_IMAGE_DETECT_MSIE
1488 tmp = get_header_value(csp->headers, "User-Agent:");
1489 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1491 tmp = get_header_value(csp->headers, "Accept:");
1492 if (tmp && strstr(tmp, "image/gif"))
1494 /* Client will accept HTML. If this seems counterintuitive,
1504 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1506 return ((csp->action->flags & ACTION_IMAGE) != 0);
1509 #endif /* def FEATURE_IMAGE_BLOCKING */
1512 #ifdef FEATURE_TRUST
1513 /*********************************************************************
1515 * Function : is_untrusted_url
1517 * Description : Should we "distrust" this URL (and block it)?
1519 * Yes if it matches a line in the trustfile, or if the
1520 * referrer matches a line starting with "+" in the
1525 * 1 : csp = Current client state (buffers, headers, etc...)
1527 * Returns : 0 => trusted, 1 => untrusted
1529 *********************************************************************/
1530 int is_untrusted_url(struct client_state *csp)
1532 struct file_list *fl;
1533 struct block_spec *b;
1534 struct url_spec **trusted_url;
1535 struct http_request rhttp[1];
1536 const char * referer;
1540 * If we don't have a trustlist, we trust everybody
1542 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1547 memset(rhttp, '\0', sizeof(*rhttp));
1550 * Do we trust the request URL itself?
1552 for (b = b->next; b ; b = b->next)
1554 if (url_match(b->url, csp->http))
1560 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1562 /* no referrer was supplied */
1568 * If not, do we maybe trust its referrer?
1570 err = parse_http_url(referer, rhttp, csp);
1576 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1578 if (url_match(*trusted_url, rhttp))
1580 /* if the URL's referrer is from a trusted referrer, then
1581 * add the target spec to the trustfile as an unblocked
1582 * domain and return 0 (which means it's OK).
1587 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1591 char * new_entry = strdup("~");
1593 string_append(&new_entry, csp->http->hostport);
1595 path = csp->http->path;
1596 if ( (path[0] == '/')
1598 && ((path_end = strchr(path + 2, '/')) != NULL))
1600 /* since this path points into a user's home space
1601 * be sure to include this spec in the trustfile.
1603 int path_len = path_end - path; /* save offset */
1604 path = strdup(path); /* Copy string */
1607 path_end = path + path_len; /* regenerate ptr to new buffer */
1608 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1610 string_join(&new_entry, path);
1614 * Give a reason for generating this entry.
1616 string_append(&new_entry, " # Trusted referrer was: ");
1617 string_append(&new_entry, referer);
1619 if (new_entry != NULL)
1621 if (-1 == fprintf(fp, "%s\n", new_entry))
1623 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1624 new_entry, csp->config->trustfile);
1630 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1631 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1638 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1639 csp->http->hostport, csp->config->trustfile);
1646 #endif /* def FEATURE_TRUST */
1649 /*********************************************************************
1651 * Function : pcrs_filter_response
1653 * Description : Execute all text substitutions from all applying
1654 * +filter actions on the text buffer that's been accumulated
1655 * in csp->iob->buf. If this changes the contents, set
1656 * csp->content_length to the modified size and raise the
1657 * CSP_FLAG_MODIFIED flag.
1660 * 1 : csp = Current client state (buffers, headers, etc...)
1662 * Returns : a pointer to the (newly allocated) modified buffer.
1663 * or NULL if there were no hits or something went wrong
1665 *********************************************************************/
1666 char *pcrs_filter_response(struct client_state *csp)
1669 size_t size, prev_size;
1671 char *old = csp->iob->cur, *new = NULL;
1674 struct file_list *fl;
1675 struct re_filterfile_spec *b;
1676 struct list_entry *filtername;
1678 int i, found_filters = 0;
1683 if (csp->iob->cur >= csp->iob->eod)
1687 size = csp->iob->eod - csp->iob->cur;
1690 * Need to check the set of re_filterfiles...
1692 for (i = 0; i < MAX_AF_FILES; i++)
1705 if (0 == found_filters)
1707 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1712 * If the body has a "chunked" transfer-encoding,
1713 * get rid of it first, adjusting size and iob->eod
1715 if (csp->flags & CSP_FLAG_CHUNKED)
1717 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1718 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1722 csp->iob->eod = csp->iob->cur + size;
1723 csp->flags |= CSP_FLAG_MODIFIED;
1726 for (i = 0; i < MAX_AF_FILES; i++)
1729 if ((NULL == fl) || (NULL == fl->f))
1732 * Either there are no filter files
1733 * left, or this filter file just
1734 * contains no valid filters.
1736 * Continue to be sure we don't miss
1737 * valid filter files that are chained
1738 * after empty or invalid ones.
1743 * For all applying +filter actions, look if a filter by that
1744 * name exists and if yes, execute it's pcrs_joblist on the
1747 for (b = fl->f; b; b = b->next)
1749 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1750 filtername ; filtername = filtername->next)
1752 if (strcmp(b->name, filtername->str) == 0)
1754 int current_hits = 0; /* Number of hits caused by this filter */
1755 int job_number = 0; /* Which job we're currently executing */
1756 int job_hits = 0; /* How many hits the current job caused */
1758 if ( NULL == b->joblist )
1760 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1765 /* Apply all jobs from the joblist */
1766 for (job = b->joblist; NULL != job; job = job->next)
1769 job_hits = pcrs_execute(job, old, size, &new, &size);
1774 * That went well. Continue filtering
1775 * and use the result of this job as
1776 * input for the next one.
1778 current_hits += job_hits;
1779 if (old != csp->iob->cur)
1788 * This job caused an unexpected error. Inform the user
1789 * and skip the rest of the jobs in this filter. We could
1790 * continue with the next job, but usually the jobs
1791 * depend on each other or are similar enough to
1792 * fail for the same reason.
1794 * At the moment our pcrs expects the error codes of pcre 3.4,
1795 * but newer pcre versions can return additional error codes.
1796 * As a result pcrs_strerror()'s error message might be
1797 * "Unknown error ...", therefore we print the numerical value
1800 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1801 * should we use LOG_LEVEL_RE_FILTER instead?
1803 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1804 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1809 log_error(LOG_LEVEL_RE_FILTER,
1810 "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1811 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1813 hits += current_hits;
1820 * If there were no hits, destroy our copy and let
1821 * chat() use the original in csp->iob
1829 csp->flags |= CSP_FLAG_MODIFIED;
1830 csp->content_length = size;
1838 /*********************************************************************
1840 * Function : gif_deanimate_response
1842 * Description : Deanimate the GIF image that has been accumulated in
1843 * csp->iob->buf, set csp->content_length to the modified
1844 * size and raise the CSP_FLAG_MODIFIED flag.
1847 * 1 : csp = Current client state (buffers, headers, etc...)
1849 * Returns : a pointer to the (newly allocated) modified buffer.
1850 * or NULL in case something went wrong.
1852 *********************************************************************/
1853 char *gif_deanimate_response(struct client_state *csp)
1855 struct binbuffer *in, *out;
1857 size_t size = csp->iob->eod - csp->iob->cur;
1860 * If the body has a "chunked" transfer-encoding,
1861 * get rid of it first, adjusting size and iob->eod
1863 if (csp->flags & CSP_FLAG_CHUNKED)
1865 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1866 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1870 csp->iob->eod = csp->iob->cur + size;
1871 csp->flags |= CSP_FLAG_MODIFIED;
1874 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1875 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1877 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1881 in->buffer = csp->iob->cur;
1884 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1886 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1893 if ((int)size == out->offset)
1895 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1899 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1901 csp->content_length = out->offset;
1902 csp->flags |= CSP_FLAG_MODIFIED;
1912 /*********************************************************************
1914 * Function : jpeg_inspect_response
1919 * 1 : csp = Current client state (buffers, headers, etc...)
1921 * Returns : a pointer to the (newly allocated) modified buffer
1922 * or NULL in case something went wrong.
1924 *********************************************************************/
1925 char *jpeg_inspect_response(struct client_state *csp)
1927 struct binbuffer *in = NULL, *out = NULL;
1929 size_t size = csp->iob->eod - csp->iob->cur;
1932 * If the body has a "chunked" transfer-encoding,
1933 * get rid of it first, adjusting size and iob->eod
1935 if (csp->flags & CSP_FLAG_CHUNKED)
1937 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1938 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1942 csp->iob->eod = csp->iob->cur + size;
1943 csp->flags |= CSP_FLAG_MODIFIED;
1946 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1948 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1952 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1954 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1958 in->buffer = csp->iob->cur;
1962 * Calling jpeg_inspect has the side-effect of creating and
1963 * modifying the image buffer of "out" directly.
1965 if (jpeg_inspect(in, out))
1967 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1975 csp->content_length = out->offset;
1976 csp->flags |= CSP_FLAG_MODIFIED;
1986 /*********************************************************************
1988 * Function : remove_chunked_transfer_coding
1990 * Description : In-situ remove the "chunked" transfer coding as defined
1991 * in rfc2616 from a buffer.
1994 * 1 : buffer = Pointer to the text buffer
1995 * 2 : size = Number of bytes to be processed
1997 * Returns : The new size, i.e. the number of bytes from buffer which
1998 * are occupied by the stripped body, or 0 in case something
2001 *********************************************************************/
2002 int remove_chunked_transfer_coding(char *buffer, const size_t size)
2005 unsigned int chunksize = 0;
2006 char *from_p, *to_p;
2009 from_p = to_p = buffer;
2011 if (sscanf(buffer, "%x", &chunksize) != 1)
2013 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2017 while (chunksize > 0)
2019 if (NULL == (from_p = strstr(from_p, "\r\n")))
2021 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2025 if ((newsize += chunksize) >= size)
2027 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2032 memmove(to_p, from_p, (size_t) chunksize);
2033 to_p = buffer + newsize;
2034 from_p += chunksize + 2;
2036 if (sscanf(from_p, "%x", &chunksize) != 1)
2038 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2043 /* FIXME: Should this get its own loglevel? */
2044 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2050 /*********************************************************************
2052 * Function : url_actions
2054 * Description : Gets the actions for this URL.
2057 * 1 : http = http_request request for blocked URLs
2058 * 2 : csp = Current client state (buffers, headers, etc...)
2062 *********************************************************************/
2063 void url_actions(struct http_request *http,
2064 struct client_state *csp)
2066 struct file_list *fl;
2067 struct url_actions *b;
2070 init_current_action(csp->action);
2072 for (i = 0; i < MAX_AF_FILES; i++)
2074 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2079 apply_url_actions(csp->action, http, b);
2086 /*********************************************************************
2088 * Function : apply_url_actions
2090 * Description : Applies a list of URL actions.
2093 * 1 : action = Destination.
2094 * 2 : http = Current URL
2095 * 3 : b = list of URL actions to apply
2099 *********************************************************************/
2100 void apply_url_actions(struct current_action_spec *action,
2101 struct http_request *http,
2102 struct url_actions *b)
2106 /* Should never happen */
2110 for (b = b->next; NULL != b; b = b->next)
2112 if (url_match(b->url, http))
2114 merge_current_action(action, b->action);
2120 /*********************************************************************
2122 * Function : forward_url
2124 * Description : Should we forward this to another proxy?
2127 * 1 : http = http_request request for current URL
2128 * 2 : csp = Current client state (buffers, headers, etc...)
2130 * Returns : Pointer to forwarding information.
2132 *********************************************************************/
2133 const struct forward_spec * forward_url(struct http_request *http,
2134 struct client_state *csp)
2136 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2137 struct forward_spec *fwd = csp->config->forward;
2146 if (url_match(fwd->url, http))
2157 /*********************************************************************
2159 * Function : direct_response
2161 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2162 * request and if so, return a HTTP 501 to the client.
2164 * FIXME: I have a stupid name and I should handle the
2165 * requests properly. Still, what we do here is rfc-
2166 * compliant, whereas ignoring or forwarding are not.
2169 * 1 : csp = Current client state (buffers, headers, etc...)
2171 * Returns : http_response if , NULL if nonmatch or handler fail
2173 *********************************************************************/
2174 struct http_response *direct_response(struct client_state *csp)
2176 struct http_response *rsp;
2177 struct list_entry *p;
2179 if ((0 == strcmpic(csp->http->gpc, "trace"))
2180 || (0 == strcmpic(csp->http->gpc, "options")))
2182 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2184 if (!strncmp("Max-Forwards:", p->str, 13)
2185 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
2187 /* FIXME: We could handle at least TRACE here,
2188 but that would require a verbatim copy of
2189 the request which we don't have anymore */
2191 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
2193 /* Get mem for response or fail*/
2194 if (NULL == (rsp = alloc_http_response()))
2196 return cgi_error_memory();
2199 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2201 free_http_response(rsp);
2202 return cgi_error_memory();
2206 return(finish_http_response(rsp));