1 const char filters_rcs[] = "$Id: filters.c,v 1.79 2007/01/31 16:21:38 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, 2004-2007 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 1.79 2007/01/31 16:21:38 fabiankeil
44 * Search for Max-Forwards headers case-insensitive,
45 * don't generate the "501 unsupported" message for invalid
46 * Max-Forwards values and don't increase negative ones.
48 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
49 * - Add HEAD support to finish_http_response.
50 * - Add error favicon to internal HTML error messages.
52 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
53 * Mark *csp as immutable for is_untrusted_url()
54 * and is_imageurl(). Closes FR 1237736.
56 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
57 * Integrate a modified version of Wil Mahan's
58 * zlib patch (PR #895531).
60 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
61 * Fixed gcc43 conversion warnings,
62 * changed sprintf calls to snprintf.
64 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
65 * Adjust comment in pcrs_filter_response()
66 * to recent pcrs changes. Hohoho.
68 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
69 * Don't crash if pcre returns an error code
70 * that pcrs didn't expect. Fixes BR 1621173.
72 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
73 * Modified is_untrusted_url to complain in case of
74 * write errors and to give a reason when adding new
75 * entries to the trustfile. Closes FR 1097611.
77 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
78 * Skip empty filter files in pcrs_filter_response,
79 * but don't ignore the ones that come afterwards.
80 * Fixes parts of BR 1619208.
82 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
83 * Added some sanity checks for get_last_url().
84 * Fixed possible segfault caused by my last commit.
86 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
87 * Let get_last_url() catch https URLs as well.
89 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
90 * Make sure get_last_url() behaves like advertised
91 * and fast-redirects{} can be combined with redirect{}.
93 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
94 * Implemented +redirect{s@foo@bar@} to generate
95 * a redirect based on a rewritten version of the
98 * Revision 1.66 2006/09/23 13:26:38 roro
99 * Replace TABs by spaces in source code.
101 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
102 * Fix +redirect{}. Didn't work with -fast-redirects.
104 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
105 * Block requests for untrusted URLs with status
106 * code 403 instead of 200.
108 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
109 * Don't free p which is still in use and will be later
110 * freed by free_map(). Don't claim the referrer is unknown
111 * when the client didn't set one.
113 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
114 * Feature request 595948: Re-Filter logging in single line
116 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
117 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
119 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
120 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
121 * with what was really the latest development (the v_3_0_branch branch)
123 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
124 * Multiple filter file support
126 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
127 * A few memory leaks plugged (mostly on error paths)
129 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
130 * Add the ability to check jpeg images for invalid
131 * lengths of comment blocks. Defensive strategy
132 * against the exploit:
133 * Microsoft Security Bulletin MS04-028
134 * Buffer Overrun in JPEG Processing (GDI+) Could
135 * Allow Code Execution (833987)
136 * Enabled with +inspect-jpegs in actions files.
138 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
139 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
141 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
142 * Fixed bug #839859: "See why" link URL now gets url-encoded.
144 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
147 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
148 * Added basic support for OPTIONS and TRACE HTTP methods:
149 * New function direct_response which handles OPTIONS and
150 * TRACE requests whose Max-Forwards header field is zero.
152 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
153 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
155 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
156 * - Bugfix: Executing a filters without jobs no longer results in
157 * turing off *all* filters.
158 * - Security fix: Malicious web servers can't cause a seg fault
159 * through bogus chunk sizes anymore
161 * Revision 1.58 2002/04/24 02:11:17 oes
162 * Jon's multiple AF patch: url_actions now evaluates rules
165 * Revision 1.57 2002/04/08 20:38:34 swa
168 * Revision 1.56 2002/04/05 15:51:24 oes
169 * - bugfix: error-pages now get correct request protocol
170 * - fix for invalid HTML in trust info
172 * Revision 1.55 2002/04/02 16:13:51 oes
173 * Fix: No "Go there anyway" for SSL
175 * Revision 1.54 2002/04/02 14:55:56 oes
176 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
178 * Revision 1.53 2002/03/26 22:29:54 swa
179 * we have a new homepage!
181 * Revision 1.52 2002/03/24 16:35:57 jongfoster
184 * Revision 1.51 2002/03/24 15:23:33 jongfoster
187 * Revision 1.50 2002/03/24 13:25:43 swa
188 * name change related issues
190 * Revision 1.49 2002/03/16 20:29:14 oes
193 * Revision 1.48 2002/03/13 20:25:34 oes
194 * Better logging for content filters
196 * Revision 1.47 2002/03/13 00:30:52 jongfoster
198 * Added option of always sending redirect for imageblock,
199 * currently disabled with #if 0.
201 * Revision 1.46 2002/03/12 01:42:49 oes
202 * Introduced modular filters
204 * Revision 1.45 2002/03/08 16:47:50 oes
205 * Added choice beween GIF and PNG built-in images
207 * Revision 1.44 2002/03/07 03:49:31 oes
208 * - Fixed compiler warnings etc
209 * - Changed built-in images from GIF to PNG
210 * (with regard to Unisys patent issue)
211 * - Added a 4x4 pattern PNG which is less intrusive
212 * than the logo but also clearly marks the deleted banners
214 * Revision 1.43 2002/01/22 23:51:59 jongfoster
215 * Replacing strsav() with the safer string_append().
217 * Adding missing html_encode() to error message generators. Where encoded
218 * and unencoded versions of a string were provided, removing the unencoded
221 * Revision 1.42 2002/01/17 21:00:32 jongfoster
222 * Moving all our URL and URL pattern parsing code to urlmatch.c.
224 * Using a single, simple url_match(pattern,url) function - rather than
225 * the 3-line match routine which was repeated all over the place.
227 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
229 * Using parse_http_url() to parse URLs without faking a HTTP
230 * request line for parse_http_request().
232 * Revision 1.41 2001/11/13 00:14:07 jongfoster
233 * Fixing stupid bug now I've figured out what || means.
234 * (It always returns 0 or 1, not one of it's paramaters.)
236 * Revision 1.40 2001/10/26 17:37:55 oes
237 * - Re-enabled Netscape 200/404 bug workaround in block_url():
238 * - Removed OS/2 special case
239 * - Made block_url() independant from sed() having been run
240 * - Made trust_url independant from sed() having been run
241 * - Made is_imageurl independant from sed() having been run.
242 * It now checks User-Agent: and Accept: by itself.
245 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
246 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
247 * threads to call select() simultaneously. So, it's time to do a real, live,
248 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
249 * (native). Both versions will work, but using __OS2__ offers multi-threading.
251 * Revision 1.38 2001/10/23 21:32:33 jongfoster
252 * Adding error-checking to selected functions
254 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
255 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
256 * filters.c. Added a FIXME in front of the offending code. I'll gladly
257 * put in a better/more robust fix for all parties if one is presented...
258 * It seems that just returning 200 instead of 404 would pretty much fix
259 * it for everyone, but I don't know all the history of the problem.
261 * Revision 1.36 2001/10/10 16:44:16 oes
262 * Added match_portlist function
264 * Revision 1.35 2001/10/07 15:41:23 oes
265 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
267 * New function remove_chunked_transfer_coding that strips chunked
268 * transfer coding to plain and is called by pcrs_filter_response
269 * and gif_deanimate_response if neccessary
271 * Improved handling of zero-change re_filter runs
273 * pcrs_filter_response and gif_deanimate_response now remove
274 * chunked transfer codeing before processing the body.
276 * Revision 1.34 2001/09/20 15:49:36 steudten
278 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
279 * See cgi.c fill_template().
281 * Revision 1.33 2001/09/16 17:05:14 jongfoster
282 * Removing unused #include showarg.h
284 * Revision 1.32 2001/09/16 13:21:27 jongfoster
285 * Changes to use new list functions.
287 * Revision 1.31 2001/09/16 11:38:02 jongfoster
288 * Splitting fill_template() into 2 functions:
289 * template_load() loads the file
290 * template_fill() performs the PCRS regexps.
291 * This is because the CGI edit interface has a "table row"
292 * template which is used many times in the page - this
293 * change means it's only loaded from disk once.
295 * Revision 1.30 2001/09/16 11:00:10 jongfoster
296 * New function alloc_http_response, for symmetry with free_http_response
298 * Revision 1.29 2001/09/13 23:32:40 jongfoster
299 * Moving image data to cgi.c rather than cgi.h
300 * Fixing a GPF under Win32 (and any other OS that protects global
301 * constants from being written to).
303 * Revision 1.28 2001/09/10 10:18:51 oes
304 * Silenced compiler warnings
306 * Revision 1.27 2001/08/05 16:06:20 jongfoster
307 * Modifiying "struct map" so that there are now separate header and
308 * "map_entry" structures. This means that functions which modify a
309 * map no longer need to return a pointer to the modified map.
310 * Also, it no longer reverses the order of the entries (which may be
311 * important with some advanced template substitutions).
313 * Revision 1.26 2001/07/30 22:08:36 jongfoster
314 * Tidying up #defines:
315 * - All feature #defines are now of the form FEATURE_xxx
316 * - Permanently turned off WIN_GUI_EDIT
317 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
319 * Revision 1.25 2001/07/26 10:09:46 oes
320 * Made browser detection a little less naive
322 * Revision 1.24 2001/07/25 17:22:51 oes
323 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
325 * Revision 1.23 2001/07/23 13:40:12 oes
326 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
328 * Revision 1.22 2001/07/18 12:29:34 oes
329 * - Made gif_deanimate_response respect
330 * csp->action->string[ACTION_STRING_DEANIMATE]
331 * - Logging cosmetics
333 * Revision 1.21 2001/07/13 13:59:53 oes
334 * - Introduced gif_deanimate_response which shares the
335 * generic content modification interface of pcrs_filter_response
336 * and acts as a wrapper to deanimate.c:gif_deanimate()
337 * - Renamed re_process_buffer to pcrs_filter_response
338 * - pcrs_filter_response now returns NULL on failiure
339 * - Removed all #ifdef PCRS
341 * Revision 1.20 2001/07/01 17:01:04 oes
342 * Added comments and missing return statement in is_untrusted_url()
344 * Revision 1.19 2001/06/29 21:45:41 oes
345 * Indentation, CRLF->LF, Tab-> Space
347 * Revision 1.18 2001/06/29 13:27:38 oes
348 * - Cleaned up, renamed and reorderd functions
349 * and improved comments
352 * - Ported to CGI platform. Now delivers
353 * http_response or NULL
354 * - Unified HTML and GIF generation (moved image detection
355 * and GIF generation here from jcc.c:chat())
356 * - Fixed HTTP status to:
357 * - 403 (Forbidden) for the "blocked" HTML message
358 * - 200 (OK) for GIF answers
359 * - 302 (Redirect) for redirect to GIF
362 * - Ported to CGI platform. Now delivers
363 * http_response or NULL
364 * - Separated detection of untrusted URL into
365 * (bool)is_untrusted_url
366 * - Added enforcement of untrusted requests
368 * - Moved redirect_url() from cgi.c to here
369 * and ported it to the CGI platform
371 * - Removed logentry from cancelled commit
373 * Revision 1.17 2001/06/09 10:55:28 jongfoster
374 * Changing BUFSIZ ==> BUFFER_SIZE
376 * Revision 1.16 2001/06/07 23:10:26 jongfoster
377 * Allowing unanchored domain patterns to back off and retry
378 * if they partially match. Optimized right-anchored patterns.
379 * Moving ACL and forward files into config file.
380 * Replacing struct gateway with struct forward_spec
382 * Revision 1.15 2001/06/03 19:12:00 oes
383 * extracted-CGI relevant stuff
385 * Revision 1.14 2001/06/01 10:30:55 oes
386 * Added optional left-anchoring to domaincmp
388 * Revision 1.13 2001/05/31 21:21:30 jongfoster
389 * Permissionsfile / actions file changes:
390 * - Changed "permission" to "action" throughout
391 * - changes to file format to allow string parameters
392 * - Moved helper functions to actions.c
394 * Revision 1.12 2001/05/31 17:35:20 oes
396 * - Enhanced domain part globbing with infix and prefix asterisk
397 * matching and optional unanchored operation
399 * Revision 1.11 2001/05/29 11:53:23 oes
400 * "See why" link added to "blocked" page
402 * Revision 1.10 2001/05/29 09:50:24 jongfoster
403 * Unified blocklist/imagelist/permissionslist.
404 * File format is still under discussion, but the internal changes
407 * Also modified interceptor behaviour:
408 * - We now intercept all URLs beginning with one of the following
409 * prefixes (and *only* these prefixes):
411 * * http://ijbswa.sf.net/config/
412 * * http://ijbswa.sourceforge.net/config/
413 * - New interceptors "home page" - go to http://i.j.b/ to see it.
414 * - Internal changes so that intercepted and fast redirect pages
415 * are not replaced with an image.
416 * - Interceptors now have the option to send a binary page direct
417 * to the client. (i.e. ijb-send-banner uses this)
418 * - Implemented show-url-info interceptor. (Which is why I needed
419 * the above interceptors changes - a typical URL is
420 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
421 * The previous mechanism would not have intercepted that, and
422 * if it had been intercepted then it then it would have replaced
425 * Revision 1.9 2001/05/27 22:17:04 oes
427 * - re_process_buffer no longer writes the modified buffer
428 * to the client, which was very ugly. It now returns the
429 * buffer, which it is then written by chat.
431 * - content_length now adjusts the Content-Length: header
432 * for modified documents rather than crunch()ing it.
433 * (Length info in csp->content_length, which is 0 for
434 * unmodified documents)
436 * - For this to work, sed() is called twice when filtering.
438 * Revision 1.8 2001/05/26 17:13:28 jongfoster
439 * Filled in a function comment.
441 * Revision 1.7 2001/05/26 15:26:15 jongfoster
442 * ACL feature now provides more security by immediately dropping
443 * connections from untrusted hosts.
445 * Revision 1.6 2001/05/26 00:28:36 jongfoster
446 * Automatic reloading of config file.
447 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
448 * Most of the global variables have been moved to a new
449 * struct configuration_spec, accessed through csp->config->globalname
450 * Most of the globals remaining are used by the Win32 GUI.
452 * Revision 1.5 2001/05/25 22:34:30 jongfoster
455 * Revision 1.4 2001/05/22 18:46:04 oes
457 * - Enabled filtering banners by size rather than URL
458 * by adding patterns that replace all standard banner
459 * sizes with the "Junkbuster" gif to the re_filterfile
461 * - Enabled filtering WebBugs by providing a pattern
462 * which kills all 1x1 images
464 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
465 * which is selected by the (nonstandard and therefore
466 * capital) letter 'U' in the option string.
467 * It causes the quantifiers to be ungreedy by default.
468 * Appending a ? turns back to greedy (!).
470 * - Added a new interceptor ijb-send-banner, which
471 * sends back the "Junkbuster" gif. Without imagelist or
472 * MSIE detection support, or if tinygif = 1, or the
473 * URL isn't recognized as an imageurl, a lame HTML
474 * explanation is sent instead.
476 * - Added new feature, which permits blocking remote
477 * script redirects and firing back a local redirect
479 * The feature is conditionally compiled, i.e. it
480 * can be disabled with --disable-fast-redirects,
481 * plus it must be activated by a "fast-redirects"
482 * line in the config file, has its own log level
483 * and of course wants to be displayed by show-proxy-args
484 * Note: Boy, all the #ifdefs in 1001 locations and
485 * all the fumbling with configure.in and acconfig.h
486 * were *way* more work than the feature itself :-(
488 * - Because a generic redirect template was needed for
489 * this, tinygif = 3 now uses the same.
491 * - Moved GIFs, and other static HTTP response templates
496 * - Removed some >400 CRs again (Jon, you really worked
499 * Revision 1.3 2001/05/20 16:44:47 jongfoster
500 * Removing last hardcoded Junkbusters.com URLs.
502 * Revision 1.2 2001/05/20 01:21:20 jongfoster
503 * Version 2.9.4 checkin.
504 * - Merged popupfile and cookiefile, and added control over PCRS
505 * filtering, in new "permissionsfile".
506 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
507 * file error you now get a message box (in the Win32 GUI) rather
508 * than the program exiting with no explanation.
509 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
511 * - Removed tabs from "config"
512 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
513 * - Bumped up version number.
515 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
516 * Initial import of version 2.9.3 source tree
519 *********************************************************************/
525 #include <sys/types.h>
534 #endif /* ndef __OS2__ */
535 #include <netinet/in.h>
537 #include <winsock2.h>
538 #endif /* ndef _WIN32 */
542 #endif /* def __OS2__ */
550 #include "jbsockets.h"
551 #include "miscutil.h"
555 #include "deanimate.h"
556 #include "urlmatch.h"
562 const char filters_h_rcs[] = FILTERS_H_VERSION;
564 /* Fix a problem with Solaris. There should be no effect on other
566 * Solaris's isspace() is a macro which uses it's argument directly
567 * as an array index. Therefore we need to make sure that high-bit
568 * characters generate +ve values, and ideally we also want to make
569 * the argument match the declared parameter type of "int".
571 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
575 /*********************************************************************
577 * Function : block_acl
579 * Description : Block this request?
580 * Decide yes or no based on ACL file.
583 * 1 : dst = The proxy or gateway address this is going to.
584 * Or NULL to check all possible targets.
585 * 2 : csp = Current client state (buffers, headers, etc...)
586 * Also includes the client IP address.
588 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
590 *********************************************************************/
591 int block_acl(struct access_control_addr *dst, struct client_state *csp)
593 struct access_control_list *acl = csp->config->acl;
595 /* if not using an access control list, then permit the connection */
601 /* search the list */
604 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
608 /* Just want to check if they have any access */
609 if (acl->action == ACL_PERMIT)
614 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
615 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
617 if (acl->action == ACL_PERMIT)
635 /*********************************************************************
637 * Function : acl_addr
639 * Description : Called from `load_config' to parse an ACL address.
642 * 1 : aspec = String specifying ACL address.
643 * 2 : aca = struct access_control_addr to fill in.
645 * Returns : 0 => Ok, everything else is an error.
647 *********************************************************************/
648 int acl_addr(char *aspec, struct access_control_addr *aca)
650 int i, masklength, port;
656 if ((p = strchr(aspec, '/')) != NULL)
660 if (ijb_isdigit(*p) == 0)
664 masklength = atoi(p);
667 if ((masklength < 0) || (masklength > 32))
672 if ((p = strchr(aspec, ':')) != NULL)
676 if (ijb_isdigit(*p) == 0)
685 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
687 if (aca->addr == INADDR_NONE)
692 /* build the netmask */
694 for (i=1; i <= masklength ; i++)
696 aca->mask |= (1 << (32 - i));
699 /* now mask off the host portion of the ip address
700 * (i.e. save on the network portion of the address).
702 aca->addr = aca->addr & aca->mask;
707 #endif /* def FEATURE_ACL */
710 /*********************************************************************
712 * Function : match_portlist
714 * Description : Check if a given number is covered by a comma
715 * separated list of numbers and ranges (a,b-c,d,..)
718 * 1 : portlist = String with list
719 * 2 : port = port to check
721 * Returns : 0 => no match
724 *********************************************************************/
725 int match_portlist(const char *portlist, int port)
727 char *min, *max, *next, *portlist_copy;
729 min = next = portlist_copy = strdup(portlist);
732 * Zero-terminate first item and remember offset for next
734 if (NULL != (next = strchr(portlist_copy, (int) ',')))
740 * Loop through all items, checking for match
744 if (NULL == (max = strchr(min, (int) '-')))
747 * No dash, check for equality
749 if (port == atoi(min))
758 * This is a range, so check if between min and max,
759 * or, if max was omitted, between min and 65K
762 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
776 * Zero-terminate next item and remember offset for n+1
778 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
790 /*********************************************************************
792 * Function : block_url
794 * Description : Called from `chat'. Check to see if we need to block this.
797 * 1 : csp = Current client state (buffers, headers, etc...)
799 * Returns : NULL => unblocked, else HTTP block response
801 *********************************************************************/
802 struct http_response *block_url(struct client_state *csp)
804 struct http_response *rsp;
805 const char *new_content_type = NULL;
808 * If it's not blocked, don't block it ;-)
810 if ((csp->action->flags & ACTION_BLOCK) == 0)
814 if (csp->action->flags & ACTION_REDIRECT)
816 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
819 * Else, prepare a response
821 if (NULL == (rsp = alloc_http_response()))
823 return cgi_error_memory();
827 * If it's an image-url, send back an image or redirect
828 * as specified by the relevant +image action
830 #ifdef FEATURE_IMAGE_BLOCKING
831 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
835 /* determine HOW images should be blocked */
836 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
838 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
840 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
842 #if 1 /* Two alternative strategies, use this one for now: */
844 /* and handle accordingly: */
845 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
847 rsp->status = strdup("403 Request blocked by Privoxy");
848 if (rsp->status == NULL)
850 free_http_response(rsp);
851 return cgi_error_memory();
853 rsp->body = bindup(image_pattern_data, image_pattern_length);
854 if (rsp->body == NULL)
856 free_http_response(rsp);
857 return cgi_error_memory();
859 rsp->content_length = image_pattern_length;
861 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
863 free_http_response(rsp);
864 return cgi_error_memory();
868 else if (0 == strcmpic(p, "blank"))
870 rsp->status = strdup("403 Request blocked by Privoxy");
871 if (rsp->status == NULL)
873 free_http_response(rsp);
874 return cgi_error_memory();
876 rsp->body = bindup(image_blank_data, image_blank_length);
877 if (rsp->body == NULL)
879 free_http_response(rsp);
880 return cgi_error_memory();
882 rsp->content_length = image_blank_length;
884 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
886 free_http_response(rsp);
887 return cgi_error_memory();
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();
907 #else /* Following code is disabled for now */
909 /* and handle accordingly: */
910 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
912 p = CGI_PREFIX "send-banner?type=pattern";
914 else if (0 == strcmpic(p, "blank"))
916 p = CGI_PREFIX "send-banner?type=blank";
918 rsp->status = strdup("302 Local Redirect from Privoxy");
919 if (rsp->status == NULL)
921 free_http_response(rsp);
922 return cgi_error_memory();
925 if (enlist_unique_header(rsp->headers, "Location", p))
927 free_http_response(rsp);
928 return cgi_error_memory();
930 #endif /* Preceeding code is disabled for now */
932 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
935 * Send empty document.
937 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
940 rsp->body = strdup(" ");
941 rsp->content_length = 1;
943 rsp->status = strdup("403 Request blocked by Privoxy");
944 if (rsp->status == NULL)
946 free_http_response(rsp);
947 return cgi_error_memory();
949 if (new_content_type != 0)
951 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
952 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
954 free_http_response(rsp);
955 return cgi_error_memory();
960 #endif /* def FEATURE_IMAGE_BLOCKING */
963 * Else, generate an HTML "blocked" message:
967 struct map * exports;
971 * Workaround for stupid Netscape bug which prevents
972 * pages from being displayed if loading a referenced
973 * JavaScript or style sheet fails. So make it appear
974 * as if it succeeded.
976 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
977 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
978 && !strstr(p, "Gecko") /* save Mozilla, */
979 && !strstr(p, "compatible") /* MSIE */
980 && !strstr(p, "Opera")) /* and Opera. */
982 rsp->status = strdup("200 Request for blocked URL");
986 rsp->status = strdup("403 Request for blocked URL");
989 if (rsp->status == NULL)
991 free_http_response(rsp);
992 return cgi_error_memory();
995 exports = default_exports(csp, NULL);
998 free_http_response(rsp);
999 return cgi_error_memory();
1002 #ifdef FEATURE_FORCE_LOAD
1003 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1004 if (csp->http->ssl != 0)
1005 #endif /* ndef FEATURE_FORCE_LOAD */
1007 err = map_block_killer(exports, "force-support");
1010 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1011 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1012 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1013 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1018 free_http_response(rsp);
1019 return cgi_error_memory();
1022 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1025 free_http_response(rsp);
1026 return cgi_error_memory();
1029 rsp->reason = RSP_REASON_BLOCKED;
1031 return finish_http_response(csp, rsp);
1036 #ifdef FEATURE_TRUST
1037 /*********************************************************************
1039 * Function : trust_url FIXME: I should be called distrust_url
1041 * Description : Calls is_untrusted_url to determine if the URL is trusted
1042 * and if not, returns a HTTP 403 response with a reject message.
1045 * 1 : csp = Current client state (buffers, headers, etc...)
1047 * Returns : NULL => trusted, else http_response.
1049 *********************************************************************/
1050 struct http_response *trust_url(struct client_state *csp)
1052 struct http_response *rsp;
1053 struct map * exports;
1054 char buf[BUFFER_SIZE];
1056 struct url_spec **tl;
1061 * Don't bother to work on trusted URLs
1063 if (!is_untrusted_url(csp))
1069 * Else, prepare a response:
1071 if (NULL == (rsp = alloc_http_response()))
1073 return cgi_error_memory();
1076 rsp->status = strdup("403 Request blocked by Privoxy");
1077 exports = default_exports(csp, NULL);
1078 if (exports == NULL || rsp->status == NULL)
1080 free_http_response(rsp);
1081 return cgi_error_memory();
1085 * Export the protocol, host, port, and referrer information
1087 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1088 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1089 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1091 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1093 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1097 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1103 free_http_response(rsp);
1104 return cgi_error_memory();
1108 * Export the trust list
1111 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1113 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1114 string_append(&p, buf);
1116 err = map(exports, "trusted-referrers", 1, p, 0);
1121 free_http_response(rsp);
1122 return cgi_error_memory();
1126 * Export the trust info, if available
1128 if (csp->config->trust_info->first)
1130 struct list_entry *l;
1133 for (l = csp->config->trust_info->first; l ; l = l->next)
1135 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1136 string_append(&p, buf);
1138 err = map(exports, "trust-info", 1, p, 0);
1142 err = map_block_killer(exports, "have-trust-info");
1148 free_http_response(rsp);
1149 return cgi_error_memory();
1153 * Export the force prefix or the force conditional block killer
1155 #ifdef FEATURE_FORCE_LOAD
1156 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1157 #else /* ifndef FEATURE_FORCE_LOAD */
1158 err = map_block_killer(exports, "force-support");
1159 #endif /* ndef FEATURE_FORCE_LOAD */
1164 free_http_response(rsp);
1165 return cgi_error_memory();
1169 * Build the response
1171 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1174 free_http_response(rsp);
1175 return cgi_error_memory();
1177 rsp->reason = RSP_REASON_UNTRUSTED;
1179 return finish_http_response(csp, rsp);
1181 #endif /* def FEATURE_TRUST */
1183 /*********************************************************************
1185 * Function : execute_single_pcrs_command
1187 * Description : Apply single pcrs command to the subject.
1188 * The subject itself is left untouched, memory for the result
1189 * is malloc()ed and it is the caller's responsibility to free
1190 * the result when it's no longer needed.
1193 * 1 : subject = the subject (== original) string
1194 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1195 * 3 : hits = int* for returning the number of modifications
1197 * Returns : NULL in case of errors, otherwise the
1198 * result of the pcrs command.
1200 *********************************************************************/
1201 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1205 char *result = NULL;
1209 assert(pcrs_command);
1212 size = strlen(subject);
1214 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1216 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1217 pcrs_command, error);
1219 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1221 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1228 job = pcrs_free_job(job);
1235 /*********************************************************************
1237 * Function : rewrite_url
1239 * Description : Rewrites a URL with a single pcrs command
1240 * and returns the result if it differs from the
1241 * original and isn't obviously invalid.
1244 * 1 : old_url = URL to rewrite.
1245 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1248 * Returns : NULL if the pcrs_command didn't change the url, or
1249 * the result of the modification.
1251 *********************************************************************/
1252 char *rewrite_url(char *old_url, const char *pcrs_command)
1254 char *new_url = NULL;
1258 assert(pcrs_command);
1260 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1264 log_error(LOG_LEVEL_REDIRECTS,
1265 "pcrs command \"%s\" didn't change \"%s\".",
1266 pcrs_command, old_url, new_url);
1269 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1271 log_error(LOG_LEVEL_ERROR,
1272 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1273 "but the result doesn't look like a valid URL and will be ignored.",
1274 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1279 log_error(LOG_LEVEL_REDIRECTS,
1280 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1281 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1289 #ifdef FEATURE_FAST_REDIRECTS
1290 /*********************************************************************
1292 * Function : get_last_url
1294 * Description : Search for the last URL inside a string.
1295 * If the string already is a URL, it will
1296 * be the first URL found.
1299 * 1 : subject = the string to check
1300 * 2 : redirect_mode = +fast-redirect{} mode
1302 * Returns : NULL if no URL was found, or
1303 * the last URL found.
1305 *********************************************************************/
1306 char *get_last_url(char *subject, const char *redirect_mode)
1308 char *new_url = NULL;
1312 assert(redirect_mode);
1314 subject = strdup(subject);
1315 if (subject == NULL)
1317 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1321 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1323 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1324 new_url = url_decode(subject);
1325 if (new_url != NULL)
1332 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1336 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1339 * Find the last URL encoded in the request
1342 while ((tmp = strstr(tmp, "http://")) != NULL)
1346 tmp = (new_url != NULL) ? new_url : subject;
1347 while ((tmp = strstr(tmp, "https://")) != NULL)
1352 if ((new_url != NULL)
1353 && ( (new_url != subject)
1354 || (0 == strncmpic(subject, "http://", 7))
1355 || (0 == strncmpic(subject, "https://", 8))
1359 * Return new URL if we found a redirect
1360 * or if the subject already was a URL.
1362 * The second case makes sure that we can
1363 * chain get_last_url after another redirection check
1364 * (like rewrite_url) without losing earlier redirects.
1366 new_url = strdup(new_url);
1375 #endif /* def FEATURE_FAST_REDIRECTS */
1378 /*********************************************************************
1380 * Function : redirect_url
1382 * Description : Checks if Privoxy should answer the request with
1383 * a HTTP redirect and generates the redirect if
1387 * 1 : csp = Current client state (buffers, headers, etc...)
1389 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1391 *********************************************************************/
1392 struct http_response *redirect_url(struct client_state *csp)
1394 struct http_response *rsp;
1395 #ifdef FEATURE_FAST_REDIRECTS
1397 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1398 * as compile-time option? The user can easily disable
1399 * it in his action file.
1401 char * redirect_mode;
1402 #endif /* def FEATURE_FAST_REDIRECTS */
1403 char *old_url = NULL;
1404 char *new_url = NULL;
1405 char *redirection_string;
1407 if ((csp->action->flags & ACTION_REDIRECT))
1409 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1412 * If the redirection string begins with 's',
1413 * assume it's a pcrs command, otherwise treat it as
1414 * properly formatted URL and use it for the redirection
1417 * According to RFC 2616 section 14.30 the URL
1418 * has to be absolute and if the user tries:
1419 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1420 * she would get undefined results anyway.
1424 if (*redirection_string == 's')
1426 old_url = csp->http->url;
1427 new_url = rewrite_url(old_url, redirection_string);
1431 log_error(LOG_LEVEL_REDIRECTS,
1432 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1433 redirection_string);
1434 new_url = strdup(redirection_string);
1438 #ifdef FEATURE_FAST_REDIRECTS
1439 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1441 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1444 * If it exists, use the previously rewritten URL as input
1445 * otherwise just use the old path.
1447 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1448 new_url = get_last_url(old_url, redirect_mode);
1451 #endif /* def FEATURE_FAST_REDIRECTS */
1453 /* Did any redirect action trigger? */
1456 if (0 == strcmpic(new_url, csp->http->url))
1458 log_error(LOG_LEVEL_ERROR,
1459 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1460 csp->http->url, new_url);
1465 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1467 if (NULL == (rsp = alloc_http_response()))
1470 return cgi_error_memory();
1473 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1474 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1477 free_http_response(rsp);
1478 return cgi_error_memory();
1480 rsp->reason = RSP_REASON_REDIRECTED;
1483 return finish_http_response(csp, rsp);
1487 /* Only reached if no redirect is required */
1493 #ifdef FEATURE_IMAGE_BLOCKING
1494 /*********************************************************************
1496 * Function : is_imageurl
1498 * Description : Given a URL, decide whether it is an image or not,
1499 * using either the info from a previous +image action
1500 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1501 * is MSIE and not on a Mac, tell from the browser's accept
1505 * 1 : csp = Current client state (buffers, headers, etc...)
1507 * Returns : True (nonzero) if URL is an image, false (0)
1510 *********************************************************************/
1511 int is_imageurl(const struct client_state *csp)
1513 #ifdef FEATURE_IMAGE_DETECT_MSIE
1516 tmp = get_header_value(csp->headers, "User-Agent:");
1517 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1519 tmp = get_header_value(csp->headers, "Accept:");
1520 if (tmp && strstr(tmp, "image/gif"))
1522 /* Client will accept HTML. If this seems counterintuitive,
1532 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1534 return ((csp->action->flags & ACTION_IMAGE) != 0);
1537 #endif /* def FEATURE_IMAGE_BLOCKING */
1540 #ifdef FEATURE_TRUST
1541 /*********************************************************************
1543 * Function : is_untrusted_url
1545 * Description : Should we "distrust" this URL (and block it)?
1547 * Yes if it matches a line in the trustfile, or if the
1548 * referrer matches a line starting with "+" in the
1553 * 1 : csp = Current client state (buffers, headers, etc...)
1555 * Returns : 0 => trusted, 1 => untrusted
1557 *********************************************************************/
1558 int is_untrusted_url(const struct client_state *csp)
1560 struct file_list *fl;
1561 struct block_spec *b;
1562 struct url_spec **trusted_url;
1563 struct http_request rhttp[1];
1564 const char * referer;
1568 * If we don't have a trustlist, we trust everybody
1570 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1575 memset(rhttp, '\0', sizeof(*rhttp));
1578 * Do we trust the request URL itself?
1580 for (b = b->next; b ; b = b->next)
1582 if (url_match(b->url, csp->http))
1588 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1590 /* no referrer was supplied */
1596 * If not, do we maybe trust its referrer?
1598 err = parse_http_url(referer, rhttp, csp);
1604 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1606 if (url_match(*trusted_url, rhttp))
1608 /* if the URL's referrer is from a trusted referrer, then
1609 * add the target spec to the trustfile as an unblocked
1610 * domain and return 0 (which means it's OK).
1615 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1619 char * new_entry = strdup("~");
1621 string_append(&new_entry, csp->http->hostport);
1623 path = csp->http->path;
1624 if ( (path[0] == '/')
1626 && ((path_end = strchr(path + 2, '/')) != NULL))
1628 /* since this path points into a user's home space
1629 * be sure to include this spec in the trustfile.
1631 int path_len = path_end - path; /* save offset */
1632 path = strdup(path); /* Copy string */
1635 path_end = path + path_len; /* regenerate ptr to new buffer */
1636 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1638 string_join(&new_entry, path);
1642 * Give a reason for generating this entry.
1644 string_append(&new_entry, " # Trusted referrer was: ");
1645 string_append(&new_entry, referer);
1647 if (new_entry != NULL)
1649 if (-1 == fprintf(fp, "%s\n", new_entry))
1651 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1652 new_entry, csp->config->trustfile);
1658 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1659 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1666 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1667 csp->http->hostport, csp->config->trustfile);
1675 #endif /* def FEATURE_TRUST */
1678 /*********************************************************************
1680 * Function : pcrs_filter_response
1682 * Description : Execute all text substitutions from all applying
1683 * +filter actions on the text buffer that's been accumulated
1684 * in csp->iob->buf. If this changes the contents, set
1685 * csp->content_length to the modified size and raise the
1686 * CSP_FLAG_MODIFIED flag.
1688 * XXX: Currently pcrs_filter_response is also responsible
1689 * for dechunking and decompressing. Both should be
1690 * done in separate functions so other content modifiers
1691 * profit as well, even if pcrs filtering is disabled.
1694 * 1 : csp = Current client state (buffers, headers, etc...)
1696 * Returns : a pointer to the (newly allocated) modified buffer.
1697 * or NULL if there were no hits or something went wrong
1699 *********************************************************************/
1700 char *pcrs_filter_response(struct client_state *csp)
1703 size_t size, prev_size;
1705 char *old = csp->iob->cur, *new = NULL;
1708 struct file_list *fl;
1709 struct re_filterfile_spec *b;
1710 struct list_entry *filtername;
1712 int i, found_filters = 0;
1717 if (csp->iob->cur >= csp->iob->eod)
1721 size = (size_t)(csp->iob->eod - csp->iob->cur);
1724 * Need to check the set of re_filterfiles...
1726 for (i = 0; i < MAX_AF_FILES; i++)
1739 if (0 == found_filters)
1741 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1746 * If the body has a "chunked" transfer-encoding,
1747 * get rid of it first, adjusting size and iob->eod
1749 if (csp->flags & CSP_FLAG_CHUNKED)
1751 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1752 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1756 csp->iob->eod = csp->iob->cur + size;
1757 csp->flags |= CSP_FLAG_MODIFIED;
1762 * If the body has a compressed transfer-encoding,
1763 * uncompress it first, adjusting size and iob->eod.
1764 * Note that decompression occurs after de-chunking.
1766 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1768 /* Notice that we at least tried to decompress. */
1769 if (JB_ERR_OK != decompress_iob(csp))
1772 * We failed to decompress the data; there's no point
1773 * in continuing since we can't filter.
1775 * XXX: Actually the Accept-Encoding header may
1776 * just be incorrect in which case we could continue
1779 * Unset CT_GZIP and CT_DEFLATE to remember not
1780 * to modify the Content-Encoding header later.
1782 csp->content_type &= ~CT_GZIP;
1783 csp->content_type &= ~CT_DEFLATE;
1786 log_error(LOG_LEVEL_RE_FILTER, "Decompressing successful");
1789 * Decompression gives us a completely new iob,
1790 * so we need to update.
1792 size = (size_t)(csp->iob->eod - csp->iob->cur);
1793 old = csp->iob->cur;
1795 csp->flags |= CSP_FLAG_MODIFIED;
1796 csp->content_type &= ~CT_TABOO;
1800 for (i = 0; i < MAX_AF_FILES; i++)
1803 if ((NULL == fl) || (NULL == fl->f))
1806 * Either there are no filter files
1807 * left, or this filter file just
1808 * contains no valid filters.
1810 * Continue to be sure we don't miss
1811 * valid filter files that are chained
1812 * after empty or invalid ones.
1817 * For all applying +filter actions, look if a filter by that
1818 * name exists and if yes, execute it's pcrs_joblist on the
1821 for (b = fl->f; b; b = b->next)
1823 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1824 filtername ; filtername = filtername->next)
1826 if (strcmp(b->name, filtername->str) == 0)
1828 int current_hits = 0; /* Number of hits caused by this filter */
1829 int job_number = 0; /* Which job we're currently executing */
1830 int job_hits = 0; /* How many hits the current job caused */
1832 if ( NULL == b->joblist )
1834 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1839 /* Apply all jobs from the joblist */
1840 for (job = b->joblist; NULL != job; job = job->next)
1843 job_hits = pcrs_execute(job, old, size, &new, &size);
1848 * That went well. Continue filtering
1849 * and use the result of this job as
1850 * input for the next one.
1852 current_hits += job_hits;
1853 if (old != csp->iob->cur)
1862 * This job caused an unexpected error. Inform the user
1863 * and skip the rest of the jobs in this filter. We could
1864 * continue with the next job, but usually the jobs
1865 * depend on each other or are similar enough to
1866 * fail for the same reason.
1868 * At the moment our pcrs expects the error codes of pcre 3.4,
1869 * but newer pcre versions can return additional error codes.
1870 * As a result pcrs_strerror()'s error message might be
1871 * "Unknown error ...", therefore we print the numerical value
1874 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1875 * should we use LOG_LEVEL_RE_FILTER instead?
1877 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1878 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1883 log_error(LOG_LEVEL_RE_FILTER,
1884 "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1885 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1887 hits += current_hits;
1894 * If there were no hits, destroy our copy and let
1895 * chat() use the original in csp->iob
1903 csp->flags |= CSP_FLAG_MODIFIED;
1904 csp->content_length = size;
1912 /*********************************************************************
1914 * Function : gif_deanimate_response
1916 * Description : Deanimate the GIF image that has been accumulated in
1917 * csp->iob->buf, set csp->content_length to the modified
1918 * size and raise the CSP_FLAG_MODIFIED flag.
1921 * 1 : csp = Current client state (buffers, headers, etc...)
1923 * Returns : a pointer to the (newly allocated) modified buffer.
1924 * or NULL in case something went wrong.
1926 *********************************************************************/
1927 char *gif_deanimate_response(struct client_state *csp)
1929 struct binbuffer *in, *out;
1933 size = (size_t)(csp->iob->eod - csp->iob->cur);
1936 * If the body has a "chunked" transfer-encoding,
1937 * get rid of it first, adjusting size and iob->eod
1939 if (csp->flags & CSP_FLAG_CHUNKED)
1941 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1942 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1946 csp->iob->eod = csp->iob->cur + size;
1947 csp->flags |= CSP_FLAG_MODIFIED;
1950 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1951 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1953 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1957 in->buffer = csp->iob->cur;
1960 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1962 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1969 if ((int)size == out->offset)
1971 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1975 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1977 csp->content_length = out->offset;
1978 csp->flags |= CSP_FLAG_MODIFIED;
1988 /*********************************************************************
1990 * Function : jpeg_inspect_response
1995 * 1 : csp = Current client state (buffers, headers, etc...)
1997 * Returns : a pointer to the (newly allocated) modified buffer
1998 * or NULL in case something went wrong.
2000 *********************************************************************/
2001 char *jpeg_inspect_response(struct client_state *csp)
2003 struct binbuffer *in = NULL;
2004 struct binbuffer *out = NULL;
2008 size = (size_t)(csp->iob->eod - csp->iob->cur);
2011 * If the body has a "chunked" transfer-encoding,
2012 * get rid of it first, adjusting size and iob->eod
2014 if (csp->flags & CSP_FLAG_CHUNKED)
2016 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2017 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2021 csp->iob->eod = csp->iob->cur + size;
2022 csp->flags |= CSP_FLAG_MODIFIED;
2025 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2027 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2031 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2033 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2037 in->buffer = csp->iob->cur;
2041 * Calling jpeg_inspect has the side-effect of creating and
2042 * modifying the image buffer of "out" directly.
2044 if (jpeg_inspect(in, out))
2046 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2054 csp->content_length = out->offset;
2055 csp->flags |= CSP_FLAG_MODIFIED;
2065 /*********************************************************************
2067 * Function : remove_chunked_transfer_coding
2069 * Description : In-situ remove the "chunked" transfer coding as defined
2070 * in rfc2616 from a buffer.
2073 * 1 : buffer = Pointer to the text buffer
2074 * 2 : size = Number of bytes to be processed
2076 * Returns : The new size, i.e. the number of bytes from buffer which
2077 * are occupied by the stripped body, or 0 in case something
2080 *********************************************************************/
2081 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2084 unsigned int chunksize = 0;
2085 char *from_p, *to_p;
2088 from_p = to_p = buffer;
2090 if (sscanf(buffer, "%x", &chunksize) != 1)
2092 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2096 while (chunksize > 0)
2098 if (NULL == (from_p = strstr(from_p, "\r\n")))
2100 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2104 if ((newsize += chunksize) >= size)
2106 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2111 memmove(to_p, from_p, (size_t) chunksize);
2112 to_p = buffer + newsize;
2113 from_p += chunksize + 2;
2115 if (sscanf(from_p, "%x", &chunksize) != 1)
2117 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2122 /* FIXME: Should this get its own loglevel? */
2123 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2129 /*********************************************************************
2131 * Function : url_actions
2133 * Description : Gets the actions for this URL.
2136 * 1 : http = http_request request for blocked URLs
2137 * 2 : csp = Current client state (buffers, headers, etc...)
2141 *********************************************************************/
2142 void url_actions(struct http_request *http,
2143 struct client_state *csp)
2145 struct file_list *fl;
2146 struct url_actions *b;
2149 init_current_action(csp->action);
2151 for (i = 0; i < MAX_AF_FILES; i++)
2153 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2158 apply_url_actions(csp->action, http, b);
2165 /*********************************************************************
2167 * Function : apply_url_actions
2169 * Description : Applies a list of URL actions.
2172 * 1 : action = Destination.
2173 * 2 : http = Current URL
2174 * 3 : b = list of URL actions to apply
2178 *********************************************************************/
2179 void apply_url_actions(struct current_action_spec *action,
2180 struct http_request *http,
2181 struct url_actions *b)
2185 /* Should never happen */
2189 for (b = b->next; NULL != b; b = b->next)
2191 if (url_match(b->url, http))
2193 merge_current_action(action, b->action);
2199 /*********************************************************************
2201 * Function : forward_url
2203 * Description : Should we forward this to another proxy?
2206 * 1 : http = http_request request for current URL
2207 * 2 : csp = Current client state (buffers, headers, etc...)
2209 * Returns : Pointer to forwarding information.
2211 *********************************************************************/
2212 const struct forward_spec * forward_url(struct http_request *http,
2213 struct client_state *csp)
2215 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2216 struct forward_spec *fwd = csp->config->forward;
2225 if (url_match(fwd->url, http))
2236 /*********************************************************************
2238 * Function : direct_response
2240 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2241 * request and if so, return a HTTP 501 to the client.
2243 * FIXME: I have a stupid name and I should handle the
2244 * requests properly. Still, what we do here is rfc-
2245 * compliant, whereas ignoring or forwarding are not.
2248 * 1 : csp = Current client state (buffers, headers, etc...)
2250 * Returns : http_response if , NULL if nonmatch or handler fail
2252 *********************************************************************/
2253 struct http_response *direct_response(struct client_state *csp)
2255 struct http_response *rsp;
2256 struct list_entry *p;
2258 if ((0 == strcmpic(csp->http->gpc, "trace"))
2259 || (0 == strcmpic(csp->http->gpc, "options")))
2261 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2263 if (!strncmpic("Max-Forwards:", p->str, 13))
2265 unsigned int max_forwards;
2268 * If it's a Max-Forwards value of zero,
2269 * we have to intercept the request.
2271 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2274 * FIXME: We could handle at least TRACE here,
2275 * but that would require a verbatim copy of
2276 * the request which we don't have anymore
2278 log_error(LOG_LEVEL_HEADER,
2279 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2282 /* Get mem for response or fail*/
2283 if (NULL == (rsp = alloc_http_response()))
2285 return cgi_error_memory();
2288 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2290 free_http_response(rsp);
2291 return cgi_error_memory();
2295 rsp->reason = RSP_REASON_UNSUPPORTED;
2297 return(finish_http_response(csp, rsp));