1 const char parsers_rcs[] = "$Id: parsers.c,v 1.134 2008/05/21 19:27:25 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `add_to_iob', `client_cookie_adder', `client_from',
9 * `client_referrer', `client_send_cookie', `client_ua',
10 * `client_uagent', `client_x_forwarded',
11 * `client_x_forwarded_adder', `client_xtra_adder',
12 * `content_type', `crumble', `destroy_list', `enlist',
13 * `flush_socket', ``get_header', `sed', `filter_header'
14 * `server_content_encoding', `server_content_disposition',
15 * `server_last_modified', `client_accept_language',
16 * `crunch_client_header', `client_if_modified_since',
17 * `client_if_none_match', `get_destination_from_headers',
18 * `parse_header_time', `decompress_iob' and `server_set_cookie'.
20 * Copyright : Written by and Copyright (C) 2001-2008 the SourceForge
21 * Privoxy team. http://www.privoxy.org/
23 * Based on the Internet Junkbuster originally written
24 * by and Copyright (C) 1997 Anonymous Coders and
25 * Junkbusters Corporation. http://www.junkbusters.com
27 * This program is free software; you can redistribute it
28 * and/or modify it under the terms of the GNU General
29 * Public License as published by the Free Software
30 * Foundation; either version 2 of the License, or (at
31 * your option) any later version.
33 * This program is distributed in the hope that it will
34 * be useful, but WITHOUT ANY WARRANTY; without even the
35 * implied warranty of MERCHANTABILITY or FITNESS FOR A
36 * PARTICULAR PURPOSE. See the GNU General Public
37 * License for more details.
39 * The GNU General Public License should be included with
40 * this file. If not, you can view it at
41 * http://www.gnu.org/copyleft/gpl.html
42 * or write to the Free Software Foundation, Inc., 59
43 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
47 * Revision 1.134 2008/05/21 19:27:25 fabiankeil
48 * As the wafer actions are gone, we can stop including encode.h.
50 * Revision 1.133 2008/05/21 15:50:47 fabiankeil
51 * Ditch cast from (char **) to (char **).
53 * Revision 1.132 2008/05/21 15:47:14 fabiankeil
54 * Streamline sed()'s prototype and declare
55 * the header parse and add structures static.
57 * Revision 1.131 2008/05/20 20:13:30 fabiankeil
58 * Factor update_server_headers() out of sed(), ditch the
59 * first_run hack and make server_patterns_light static.
61 * Revision 1.130 2008/05/19 17:18:04 fabiankeil
62 * Wrap memmove() calls in string_move()
63 * to document the purpose in one place.
65 * Revision 1.129 2008/05/17 14:02:07 fabiankeil
66 * Normalize linear header white space.
68 * Revision 1.128 2008/05/16 16:39:03 fabiankeil
69 * If a header is split across multiple lines,
70 * merge them to a single line before parsing them.
72 * Revision 1.127 2008/05/10 13:23:38 fabiankeil
73 * Don't provide get_header() with the whole client state
74 * structure when it only needs access to csp->iob.
76 * Revision 1.126 2008/05/03 16:40:45 fabiankeil
77 * Change content_filters_enabled()'s parameter from
78 * csp->action to action so it can be also used in the
79 * CGI code. Don't bother checking if there are filters
80 * loaded, as that's somewhat besides the point.
82 * Revision 1.125 2008/04/17 14:40:49 fabiankeil
83 * Provide get_http_time() with the buffer size so it doesn't
84 * have to blindly assume that the buffer is big enough.
86 * Revision 1.124 2008/04/16 16:38:21 fabiankeil
87 * Don't pass the whole csp structure to flush_socket()
88 * when it only needs a file descriptor and a buffer.
90 * Revision 1.123 2008/03/29 12:13:46 fabiankeil
91 * Remove send-wafer and send-vanilla-wafer actions.
93 * Revision 1.122 2008/03/28 15:13:39 fabiankeil
94 * Remove inspect-jpegs action.
96 * Revision 1.121 2008/01/05 21:37:03 fabiankeil
97 * Let client_range() also handle Request-Range headers
98 * which apparently are still supported by many servers.
100 * Revision 1.120 2008/01/04 17:43:45 fabiankeil
101 * Improve the warning messages that get logged if the action files
102 * "enable" filters but no filters of that type have been loaded.
104 * Revision 1.119 2007/12/28 18:32:51 fabiankeil
105 * In server_content_type():
106 * - Don't require leading white space when detecting image content types.
107 * - Change '... not replaced ...' message to sound less crazy if the text
108 * type actually is 'text/plain'.
109 * - Mark the 'text/plain == binary data' assumption for removal.
110 * - Remove a bunch of trailing white space.
112 * Revision 1.118 2007/12/28 16:56:35 fabiankeil
113 * Minor server_content_disposition() changes:
114 * - Don't regenerate the header name all lower-case.
115 * - Some white space fixes.
116 * - Remove useless log message in case of ENOMEM.
118 * Revision 1.117 2007/12/06 18:11:50 fabiankeil
119 * Garbage-collect the code to add a X-Forwarded-For
120 * header as it seems to be mostly used by accident.
122 * Revision 1.116 2007/12/01 13:04:22 fabiankeil
123 * Fix a crash on mingw32 with some Last Modified times in the future.
125 * Revision 1.115 2007/11/02 16:52:50 fabiankeil
126 * Remove a "can't happen" error block which, over
127 * time, mutated into a "guaranteed to happen" block.
129 * Revision 1.114 2007/10/19 16:56:26 fabiankeil
130 * - Downgrade "Buffer limit reached" message to LOG_LEVEL_INFO.
131 * - Use shiny new content_filters_enabled() in client_range().
133 * Revision 1.113 2007/10/10 17:29:57 fabiankeil
134 * I forgot about Poland.
136 * Revision 1.112 2007/10/09 16:38:40 fabiankeil
137 * Remove Range and If-Range headers if content filtering is enabled.
139 * Revision 1.111 2007/10/04 18:07:00 fabiankeil
140 * Move ACTION_VANILLA_WAFER handling from jcc's chat() into
141 * client_cookie_adder() to make sure send-vanilla-wafer can be
142 * controlled through tags (and thus regression-tested).
144 * Revision 1.110 2007/09/29 10:42:37 fabiankeil
145 * - Remove "scanning headers for" log message again.
146 * - Some more whitespace fixes.
148 * Revision 1.109 2007/09/08 14:25:48 fabiankeil
149 * Refactor client_referrer() and add conditional-forge parameter.
151 * Revision 1.108 2007/08/28 18:21:03 fabiankeil
152 * A bunch of whitespace fixes, pointy hat to me.
154 * Revision 1.107 2007/08/28 18:16:32 fabiankeil
155 * Fix possible memory corruption in server_http, make sure it's not
156 * executed for ordinary server headers and mark some problems for later.
158 * Revision 1.106 2007/08/18 14:30:32 fabiankeil
159 * Let content-type-overwrite{} honour force-text-mode again.
161 * Revision 1.105 2007/08/11 14:49:49 fabiankeil
162 * - Add prototpyes for the header parsers and make them static.
163 * - Comment out client_accept_encoding_adder() which isn't used right now.
165 * Revision 1.104 2007/07/14 07:38:19 fabiankeil
166 * Move the ACTION_FORCE_TEXT_MODE check out of
167 * server_content_type(). Signal other functions
168 * whether or not a content type has been declared.
169 * Part of the fix for BR#1750917.
171 * Revision 1.103 2007/06/01 16:31:54 fabiankeil
172 * Change sed() to return a jb_err in preparation for forward-override{}.
174 * Revision 1.102 2007/05/27 12:39:32 fabiankeil
175 * Adjust "X-Filter: No" to disable dedicated header filters.
177 * Revision 1.101 2007/05/14 10:16:41 fabiankeil
178 * Streamline client_cookie_adder().
180 * Revision 1.100 2007/04/30 15:53:11 fabiankeil
181 * Make sure filters with dynamic jobs actually use them.
183 * Revision 1.99 2007/04/30 15:06:26 fabiankeil
184 * - Introduce dynamic pcrs jobs that can resolve variables.
185 * - Remove unnecessary update_action_bits_for_all_tags() call.
187 * Revision 1.98 2007/04/17 18:32:10 fabiankeil
188 * - Make tagging based on tags set by earlier taggers
189 * of the same kind possible.
190 * - Log whether or not new tags cause action bits updates
191 * (in which case a matching tag-pattern section exists).
192 * - Log if the user tries to set a tag that is already set.
194 * Revision 1.97 2007/04/15 16:39:21 fabiankeil
195 * Introduce tags as alternative way to specify which
196 * actions apply to a request. At the moment tags can be
197 * created based on client and server headers.
199 * Revision 1.96 2007/04/12 12:53:58 fabiankeil
200 * Log a warning if the content is compressed, filtering is
201 * enabled and Privoxy was compiled without zlib support.
204 * Revision 1.95 2007/03/25 14:26:40 fabiankeil
205 * - Fix warnings when compiled with glibc.
206 * - Don't use crumble() for cookie crunching.
207 * - Move cookie time parsing into parse_header_time().
208 * - Let parse_header_time() return a jb_err code
209 * instead of a pointer that can only be used to
210 * check for NULL anyway.
212 * Revision 1.94 2007/03/21 12:23:53 fabiankeil
213 * - Add better protection against malicious gzip headers.
214 * - Stop logging the first hundred bytes of decompressed content.
215 * It looks like it's working and there is always debug 16.
216 * - Log the content size after decompression in decompress_iob()
217 * instead of pcrs_filter_response().
219 * Revision 1.93 2007/03/20 15:21:44 fabiankeil
220 * - Use dedicated header filter actions instead of abusing "filter".
221 * Replace "filter-client-headers" and "filter-client-headers"
222 * with "server-header-filter" and "client-header-filter".
223 * - Remove filter_client_header() and filter_client_header(),
224 * filter_header() now checks the shiny new
225 * CSP_FLAG_CLIENT_HEADER_PARSING_DONE flag instead.
227 * Revision 1.92 2007/03/05 13:25:32 fabiankeil
228 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
229 * - Handle "Cookie:" and "Connection:" headers a bit smarter
230 * (don't crunch them just to recreate them later on).
231 * - Add another non-standard time format for the cookie
232 * expiration date detection.
233 * - Fix a valgrind warning.
235 * Revision 1.91 2007/02/24 12:27:32 fabiankeil
236 * Improve cookie expiration date detection.
238 * Revision 1.90 2007/02/08 19:12:35 fabiankeil
239 * Don't run server_content_length() the first time
240 * sed() parses server headers; only adjust the
241 * Content-Length header if the page was modified.
243 * Revision 1.89 2007/02/07 16:52:11 fabiankeil
244 * Fix log messages regarding the cookie time format
245 * (cookie and request URL were mixed up).
247 * Revision 1.88 2007/02/07 11:27:12 fabiankeil
248 * - Let decompress_iob()
249 * - not corrupt the content if decompression fails
250 * early. (the first byte(s) were lost).
251 * - use pointer arithmetics with defined outcome for
253 * - Use a different kludge to remember a failed decompression.
255 * Revision 1.87 2007/01/31 16:21:38 fabiankeil
256 * Search for Max-Forwards headers case-insensitive,
257 * don't generate the "501 unsupported" message for invalid
258 * Max-Forwards values and don't increase negative ones.
260 * Revision 1.86 2007/01/30 13:05:26 fabiankeil
261 * - Let server_set_cookie() check the expiration date
262 * of cookies and don't touch the ones that are already
263 * expired. Fixes problems with low quality web applications
264 * as described in BR 932612.
266 * - Adjust comment in client_max_forwards to reality;
267 * remove invalid Max-Forwards headers.
269 * Revision 1.85 2007/01/26 15:33:46 fabiankeil
270 * Stop filter_header() from unintentionally removing
271 * empty header lines that were enlisted by the continue
274 * Revision 1.84 2007/01/24 12:56:52 fabiankeil
275 * - Repeat the request URL before logging any headers.
276 * Makes reading the log easier in case of simultaneous requests.
277 * - If there are more than one Content-Type headers in one request,
278 * use the first one and remove the others.
279 * - Remove "newval" variable in server_content_type().
280 * It's only used once.
282 * Revision 1.83 2007/01/12 15:03:02 fabiankeil
283 * Correct a cast, check inflateEnd() exit code
284 * to see if we have to, replace sprintf calls
287 * Revision 1.82 2007/01/01 19:36:37 fabiankeil
288 * Integrate a modified version of Wil Mahan's
289 * zlib patch (PR #895531).
291 * Revision 1.81 2006/12/31 22:21:33 fabiankeil
292 * Skip empty filter files in filter_header()
293 * but don't ignore the ones that come afterwards.
294 * Fixes BR 1619208, this time for real.
296 * Revision 1.80 2006/12/29 19:08:22 fabiankeil
297 * Reverted parts of my last commit
298 * to keep error handling working.
300 * Revision 1.79 2006/12/29 18:04:40 fabiankeil
301 * Fixed gcc43 conversion warnings.
303 * Revision 1.78 2006/12/26 17:19:20 fabiankeil
304 * Bringing back the "useless" localtime() call
305 * I removed in revision 1.67. On some platforms
306 * it's necessary to prevent time zone offsets.
308 * Revision 1.77 2006/12/07 18:44:26 fabiankeil
309 * Rebuild request URL in get_destination_from_headers()
310 * to make sure redirect{pcrs command} works as expected
311 * for intercepted requests.
313 * Revision 1.76 2006/12/06 19:52:25 fabiankeil
314 * Added get_destination_from_headers().
316 * Revision 1.75 2006/11/13 19:05:51 fabiankeil
317 * Make pthread mutex locking more generic. Instead of
318 * checking for OSX and OpenBSD, check for FEATURE_PTHREAD
319 * and use mutex locking unless there is an _r function
320 * available. Better safe than sorry.
322 * Fixes "./configure --disable-pthread" and should result
323 * in less threading-related problems on pthread-using platforms,
324 * but it still doesn't fix BR#1122404.
326 * Revision 1.74 2006/10/02 16:59:12 fabiankeil
327 * The special header "X-Filter: No" now disables
328 * header filtering as well.
330 * Revision 1.73 2006/09/23 13:26:38 roro
331 * Replace TABs by spaces in source code.
333 * Revision 1.72 2006/09/23 12:37:21 fabiankeil
334 * Don't print a log message every time filter_headers is
335 * entered or left. It only creates noise without any real
338 * Revision 1.71 2006/09/21 19:55:17 fabiankeil
339 * Fix +hide-if-modified-since{-n}.
341 * Revision 1.70 2006/09/08 12:06:34 fabiankeil
342 * Have hide-if-modified-since interpret the random
343 * range value as minutes instead of hours. Allows
344 * more fine-grained configuration.
346 * Revision 1.69 2006/09/06 16:25:51 fabiankeil
347 * Always have parse_header_time return a pointer
348 * that actual makes sense, even though we currently
349 * only need it to detect problems.
351 * Revision 1.68 2006/09/06 10:43:32 fabiankeil
352 * Added config option enable-remote-http-toggle
353 * to specify if Privoxy should recognize special
354 * headers (currently only X-Filter) to change its
355 * behaviour. Disabled by default.
357 * Revision 1.67 2006/09/04 11:01:26 fabiankeil
358 * After filtering de-chunked instances, remove
359 * "Transfer-Encoding" header entirely instead of changing
360 * it to "Transfer-Encoding: identity", which is invalid.
361 * Thanks Michael Shields <shields@msrl.com>. Fixes PR 1318658.
363 * Don't use localtime in parse_header_time. An empty time struct
364 * is good enough, it gets overwritten by strptime anyway.
366 * Revision 1.66 2006/09/03 19:38:28 fabiankeil
367 * Use gmtime_r if available, fallback to gmtime with mutex
368 * protection for MacOSX and use vanilla gmtime for the rest.
370 * Revision 1.65 2006/08/22 10:55:56 fabiankeil
371 * Changed client_referrer to use the right type (size_t) for
372 * hostlenght and to shorten the temporary referrer string with
373 * '\0' instead of adding a useless line break.
375 * Revision 1.64 2006/08/17 17:15:10 fabiankeil
376 * - Back to timegm() using GnuPG's replacement if necessary.
377 * Using mktime() and localtime() could add a on hour offset if
378 * the randomize factor was big enough to lead to a summer/wintertime
381 * - Removed now-useless Privoxy 3.0.3 compatibility glue.
383 * - Moved randomization code into pick_from_range().
385 * - Changed parse_header_time definition.
386 * time_t isn't guaranteed to be signed and
387 * if it isn't, -1 isn't available as error code.
388 * Changed some variable types in client_if_modified_since()
389 * because of the same reason.
391 * Revision 1.63 2006/08/14 13:18:08 david__schmidt
392 * OS/2 compilation compatibility fixups
394 * Revision 1.62 2006/08/14 08:58:42 fabiankeil
395 * Changed include from strptime.c to strptime.h
397 * Revision 1.61 2006/08/14 08:25:19 fabiankeil
398 * Split filter-headers{} into filter-client-headers{}
399 * and filter-server-headers{}.
400 * Added parse_header_time() to share some code.
401 * Replaced timegm() with mktime().
403 * Revision 1.60 2006/08/12 03:54:37 david__schmidt
404 * Windows service integration
406 * Revision 1.59 2006/08/03 02:46:41 david__schmidt
407 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
409 * Revision 1.58 2006/07/18 14:48:47 david__schmidt
410 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
411 * with what was really the latest development (the v_3_0_branch branch)
413 * Revision 1.56.2.10 2006/01/21 16:16:08 david__schmidt
414 * Thanks to Edward Carrel for his patch to modernize OSX's
\rpthreads support. See bug #1409623.
416 * Revision 1.56.2.9 2004/10/03 12:53:45 david__schmidt
417 * Add the ability to check jpeg images for invalid
418 * lengths of comment blocks. Defensive strategy
419 * against the exploit:
420 * Microsoft Security Bulletin MS04-028
421 * Buffer Overrun in JPEG Processing (GDI+) Could
422 * Allow Code Execution (833987)
423 * Enabled with +inspect-jpegs in actions files.
425 * Revision 1.56.2.8 2003/07/11 13:21:25 oes
426 * Excluded text/plain objects from filtering. This fixes a
427 * couple of client-crashing, download corruption and
428 * Privoxy performance issues, whose root cause lies in
429 * web servers labelling content of unknown type as text/plain.
431 * Revision 1.56.2.7 2003/05/06 12:07:26 oes
432 * Fixed bug #729900: Suspicious HOST: headers are now killed and regenerated if necessary
434 * Revision 1.56.2.6 2003/04/14 21:28:30 oes
435 * Completing the previous change
437 * Revision 1.56.2.5 2003/04/14 12:08:16 oes
438 * Added temporary workaround for bug in PHP < 4.2.3
440 * Revision 1.56.2.4 2003/03/07 03:41:05 david__schmidt
441 * Wrapping all *_r functions (the non-_r versions of them) with mutex semaphores for OSX. Hopefully this will take care of all of those pesky crash reports.
443 * Revision 1.56.2.3 2002/11/10 04:20:02 hal9
444 * Fix typo: supressed -> suppressed
446 * Revision 1.56.2.2 2002/09/25 14:59:53 oes
447 * Improved cookie logging
449 * Revision 1.56.2.1 2002/09/25 14:52:45 oes
450 * Added basic support for OPTIONS and TRACE HTTP methods:
451 * - New parser function client_max_forwards which decrements
452 * the Max-Forwards HTTP header field of OPTIONS and TRACE
453 * requests by one before forwarding
454 * - New parser function client_host which extracts the host
455 * and port information from the HTTP header field if the
456 * request URI was not absolute
457 * - Don't crumble and re-add the Host: header, but only generate
458 * and append if missing
460 * Revision 1.56 2002/05/12 15:34:22 jongfoster
461 * Fixing typo in a comment
463 * Revision 1.55 2002/05/08 16:01:07 oes
464 * Optimized add_to_iob:
465 * - Use realloc instead of malloc(), memcpy(), free()
466 * - Expand to powers of two if possible, to get
467 * O(log n) reallocs instead of O(n).
468 * - Moved check for buffer limit here from chat
469 * - Report failure via returncode
471 * Revision 1.54 2002/04/02 15:03:16 oes
472 * Tiny code cosmetics
474 * Revision 1.53 2002/03/26 22:29:55 swa
475 * we have a new homepage!
477 * Revision 1.52 2002/03/24 13:25:43 swa
478 * name change related issues
480 * Revision 1.51 2002/03/13 00:27:05 jongfoster
483 * Revision 1.50 2002/03/12 01:45:35 oes
484 * More verbose logging
486 * Revision 1.49 2002/03/09 20:03:52 jongfoster
487 * - Making various functions return int rather than size_t.
488 * (Undoing a recent change). Since size_t is unsigned on
489 * Windows, functions like read_socket that return -1 on
490 * error cannot return a size_t.
492 * THIS WAS A MAJOR BUG - it caused frequent, unpredictable
493 * crashes, and also frequently caused JB to jump to 100%
494 * CPU and stay there. (Because it thought it had just
495 * read ((unsigned)-1) == 4Gb of data...)
497 * - The signature of write_socket has changed, it now simply
498 * returns success=0/failure=nonzero.
500 * - Trying to get rid of a few warnings --with-debug on
501 * Windows, I've introduced a new type "jb_socket". This is
502 * used for the socket file descriptors. On Windows, this
503 * is SOCKET (a typedef for unsigned). Everywhere else, it's
504 * an int. The error value can't be -1 any more, so it's
505 * now JB_INVALID_SOCKET (which is -1 on UNIX, and in
506 * Windows it maps to the #define INVALID_SOCKET.)
508 * - The signature of bind_port has changed.
510 * Revision 1.48 2002/03/07 03:46:53 oes
511 * Fixed compiler warnings etc
513 * Revision 1.47 2002/02/20 23:15:13 jongfoster
514 * Parsing functions now handle out-of-memory gracefully by returning
517 * Revision 1.46 2002/01/17 21:03:47 jongfoster
518 * Moving all our URL and URL pattern parsing code to urlmatch.c.
520 * Revision 1.45 2002/01/09 14:33:03 oes
521 * Added support for localtime_r.
523 * Revision 1.44 2001/12/14 01:22:54 steudten
524 * Remove 'user:pass@' from 'proto://user:pass@host' for the
525 * new added header 'Host: ..'. (See Req ID 491818)
527 * Revision 1.43 2001/11/23 00:26:38 jongfoster
528 * Fixing two really stupid errors in my previous commit
530 * Revision 1.42 2001/11/22 21:59:30 jongfoster
531 * Adding code to handle +no-cookies-keep
533 * Revision 1.41 2001/11/05 23:43:05 steudten
534 * Add time+date to log files.
536 * Revision 1.40 2001/10/26 20:13:09 jongfoster
537 * ctype.h is needed in Windows, too.
539 * Revision 1.39 2001/10/26 17:40:04 oes
540 * Introduced get_header_value()
541 * Removed http->user_agent, csp->referrer and csp->accept_types
542 * Removed client_accept()
544 * Revision 1.38 2001/10/25 03:40:48 david__schmidt
545 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
546 * threads to call select() simultaneously. So, it's time to do a real, live,
547 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
548 * (native). Both versions will work, but using __OS2__ offers multi-threading.
550 * Revision 1.37 2001/10/23 21:36:02 jongfoster
551 * Documenting sed()'s error behaviou (doc change only)
553 * Revision 1.36 2001/10/13 12:51:51 joergs
554 * Removed client_host, (was only required for the old 2.0.2-11 http://noijb.
555 * force-load), instead crumble Host: and add it (again) in client_host_adder
556 * (in case we get a HTTP/1.0 request without Host: header and forward it to
557 * a HTTP/1.1 server/proxy).
559 * Revision 1.35 2001/10/09 22:39:21 jongfoster
560 * assert.h is also required under Win32, so moving out of #ifndef _WIN32
563 * Revision 1.34 2001/10/07 18:50:55 oes
564 * Added server_content_encoding, renamed server_transfer_encoding
566 * Revision 1.33 2001/10/07 18:04:49 oes
567 * Changed server_http11 to server_http and its pattern to "HTTP".
568 * Additional functionality: it now saves the HTTP status into
569 * csp->http->status and sets CT_TABOO for Status 206 (partial range)
571 * Revision 1.32 2001/10/07 15:43:28 oes
572 * Removed FEATURE_DENY_GZIP and replaced it with client_accept_encoding,
573 * client_te and client_accept_encoding_adder, triggered by the new
574 * +no-compression action. For HTTP/1.1 the Accept-Encoding header is
575 * changed to allow only identity and chunked, and the TE header is
576 * crunched. For HTTP/1.0, Accept-Encoding is crunched.
578 * parse_http_request no longer does anything than parsing. The rewriting
579 * of http->cmd and version mangling are gone. It now also recognizes
580 * the put and delete methods and saves the url in http->url. Removed
583 * renamed content_type and content_length to have the server_ prefix
585 * server_content_type now only works if csp->content_type != CT_TABOO
587 * added server_transfer_encoding, which
588 * - Sets CT_TABOO to prohibit filtering if encoding compresses
589 * - Raises the CSP_FLAG_CHUNKED flag if Encoding is "chunked"
590 * - Change from "chunked" to "identity" if body was chunked
591 * but has been de-chunked for filtering.
593 * added server_content_md5 which crunches any Content-MD5 headers
594 * if the body was modified.
596 * made server_http11 conditional on +downgrade action
598 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
600 * Revision 1.31 2001/10/05 14:25:02 oes
601 * Crumble Keep-Alive from Server
603 * Revision 1.30 2001/09/29 12:56:03 joergs
604 * IJB now changes HTTP/1.1 to HTTP/1.0 in requests and answers.
606 * Revision 1.29 2001/09/24 21:09:24 jongfoster
607 * Fixing 2 memory leaks that Guy spotted, where the paramater to
608 * enlist() was not being free()d.
610 * Revision 1.28 2001/09/22 16:32:28 jongfoster
611 * Removing unused #includes.
613 * Revision 1.27 2001/09/20 15:45:25 steudten
615 * add casting from size_t to int for printf()
616 * remove local variable shadow s2
618 * Revision 1.26 2001/09/16 17:05:14 jongfoster
619 * Removing unused #include showarg.h
621 * Revision 1.25 2001/09/16 13:21:27 jongfoster
622 * Changes to use new list functions.
624 * Revision 1.24 2001/09/13 23:05:50 jongfoster
625 * Changing the string paramater to the header parsers a "const".
627 * Revision 1.23 2001/09/12 18:08:19 steudten
629 * In parse_http_request() header rewriting miss the host value, so
630 * from http://www.mydomain.com the result was just " / " not
631 * http://www.mydomain.com/ in case we forward.
633 * Revision 1.22 2001/09/10 10:58:53 oes
634 * Silenced compiler warnings
636 * Revision 1.21 2001/07/31 14:46:00 oes
637 * - Persistant connections now suppressed
638 * - sed() no longer appends empty header to csp->headers
640 * Revision 1.20 2001/07/30 22:08:36 jongfoster
641 * Tidying up #defines:
642 * - All feature #defines are now of the form FEATURE_xxx
643 * - Permanently turned off WIN_GUI_EDIT
644 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
646 * Revision 1.19 2001/07/25 17:21:54 oes
647 * client_uagent now saves copy of User-Agent: header value
649 * Revision 1.18 2001/07/13 14:02:46 oes
650 * - Included fix to repair broken HTTP requests that
651 * don't contain a path, not even '/'.
652 * - Removed all #ifdef PCRS
653 * - content_type now always inspected and classified as
654 * text, gif or other.
655 * - formatting / comments
657 * Revision 1.17 2001/06/29 21:45:41 oes
658 * Indentation, CRLF->LF, Tab-> Space
660 * Revision 1.16 2001/06/29 13:32:42 oes
662 * - Adapted free_http_request
663 * - Removed logentry from cancelled commit
665 * Revision 1.15 2001/06/03 19:12:38 oes
666 * deleted const struct interceptors
668 * Revision 1.14 2001/06/01 18:49:17 jongfoster
669 * Replaced "list_share" with "list" - the tiny memory gain was not
670 * worth the extra complexity.
672 * Revision 1.13 2001/05/31 21:30:33 jongfoster
673 * Removed list code - it's now in list.[ch]
674 * Renamed "permission" to "action", and changed many features
675 * to use the actions file rather than the global config.
677 * Revision 1.12 2001/05/31 17:33:13 oes
681 * Revision 1.11 2001/05/29 20:11:19 joergs
682 * '/ * inside comment' warning removed.
684 * Revision 1.10 2001/05/29 09:50:24 jongfoster
685 * Unified blocklist/imagelist/permissionslist.
686 * File format is still under discussion, but the internal changes
689 * Also modified interceptor behaviour:
690 * - We now intercept all URLs beginning with one of the following
691 * prefixes (and *only* these prefixes):
693 * * http://ijbswa.sf.net/config/
694 * * http://ijbswa.sourceforge.net/config/
695 * - New interceptors "home page" - go to http://i.j.b/ to see it.
696 * - Internal changes so that intercepted and fast redirect pages
697 * are not replaced with an image.
698 * - Interceptors now have the option to send a binary page direct
699 * to the client. (i.e. ijb-send-banner uses this)
700 * - Implemented show-url-info interceptor. (Which is why I needed
701 * the above interceptors changes - a typical URL is
702 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
703 * The previous mechanism would not have intercepted that, and
704 * if it had been intercepted then it then it would have replaced
707 * Revision 1.9 2001/05/28 17:26:33 jongfoster
708 * Fixing segfault if last header was crunched.
709 * Fixing Windows build (snprintf() is _snprintf() under Win32, but we
710 * can use the cross-platform sprintf() instead.)
712 * Revision 1.8 2001/05/27 22:17:04 oes
714 * - re_process_buffer no longer writes the modified buffer
715 * to the client, which was very ugly. It now returns the
716 * buffer, which it is then written by chat.
718 * - content_length now adjusts the Content-Length: header
719 * for modified documents rather than crunch()ing it.
720 * (Length info in csp->content_length, which is 0 for
721 * unmodified documents)
723 * - For this to work, sed() is called twice when filtering.
725 * Revision 1.7 2001/05/27 13:19:06 oes
726 * Patched Joergs solution for the content-length in.
728 * Revision 1.6 2001/05/26 13:39:32 jongfoster
729 * Only crunches Content-Length header if applying RE filtering.
730 * Without this fix, Microsoft Windows Update wouldn't work.
732 * Revision 1.5 2001/05/26 00:28:36 jongfoster
733 * Automatic reloading of config file.
734 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
735 * Most of the global variables have been moved to a new
736 * struct configuration_spec, accessed through csp->config->globalname
737 * Most of the globals remaining are used by the Win32 GUI.
739 * Revision 1.4 2001/05/22 18:46:04 oes
741 * - Enabled filtering banners by size rather than URL
742 * by adding patterns that replace all standard banner
743 * sizes with the "Junkbuster" gif to the re_filterfile
745 * - Enabled filtering WebBugs by providing a pattern
746 * which kills all 1x1 images
748 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
749 * which is selected by the (nonstandard and therefore
750 * capital) letter 'U' in the option string.
751 * It causes the quantifiers to be ungreedy by default.
752 * Appending a ? turns back to greedy (!).
754 * - Added a new interceptor ijb-send-banner, which
755 * sends back the "Junkbuster" gif. Without imagelist or
756 * MSIE detection support, or if tinygif = 1, or the
757 * URL isn't recognized as an imageurl, a lame HTML
758 * explanation is sent instead.
760 * - Added new feature, which permits blocking remote
761 * script redirects and firing back a local redirect
763 * The feature is conditionally compiled, i.e. it
764 * can be disabled with --disable-fast-redirects,
765 * plus it must be activated by a "fast-redirects"
766 * line in the config file, has its own log level
767 * and of course wants to be displayed by show-proxy-args
768 * Note: Boy, all the #ifdefs in 1001 locations and
769 * all the fumbling with configure.in and acconfig.h
770 * were *way* more work than the feature itself :-(
772 * - Because a generic redirect template was needed for
773 * this, tinygif = 3 now uses the same.
775 * - Moved GIFs, and other static HTTP response templates
780 * - Removed some >400 CRs again (Jon, you really worked
783 * Revision 1.3 2001/05/20 01:21:20 jongfoster
784 * Version 2.9.4 checkin.
785 * - Merged popupfile and cookiefile, and added control over PCRS
786 * filtering, in new "permissionsfile".
787 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
788 * file error you now get a message box (in the Win32 GUI) rather
789 * than the program exiting with no explanation.
790 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
792 * - Removed tabs from "config"
793 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
794 * - Bumped up version number.
796 * Revision 1.2 2001/05/17 23:02:36 oes
797 * - Made referrer option accept 'L' as a substitute for '§'
799 * Revision 1.1.1.1 2001/05/15 13:59:01 oes
800 * Initial import of version 2.9.3 source tree
803 *********************************************************************/
810 #include <sys/types.h>
820 * Convince GNU's libc to provide a strptime prototype.
823 #endif /*__GLIBC__ */
830 #if !defined(_WIN32) && !defined(__OS2__)
836 #ifdef FEATURE_PTHREAD
838 /* jcc.h is for mutex semapores only */
839 #endif /* def FEATURE_PTHREAD */
844 #include "jbsockets.h"
845 #include "miscutil.h"
850 #ifndef HAVE_STRPTIME
851 #include "strptime.h"
854 const char parsers_h_rcs[] = PARSERS_H_VERSION;
856 /* Fix a problem with Solaris. There should be no effect on other
858 * Solaris's isspace() is a macro which uses its argument directly
859 * as an array index. Therefore we need to make sure that high-bit
860 * characters generate +ve values, and ideally we also want to make
861 * the argument match the declared parameter type of "int".
863 * Why did they write a character function that can't take a simple
864 * "char" argument? Doh!
866 #define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
867 #define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
869 static char *get_header_line(struct iob *iob);
870 static jb_err scan_headers(struct client_state *csp);
871 static jb_err header_tagger(struct client_state *csp, char *header);
872 static jb_err parse_header_time(const char *header_time, time_t *result);
874 static jb_err crumble (struct client_state *csp, char **header);
875 static jb_err connection (struct client_state *csp, char **header);
876 static jb_err filter_header (struct client_state *csp, char **header);
877 static jb_err client_referrer (struct client_state *csp, char **header);
878 static jb_err client_uagent (struct client_state *csp, char **header);
879 static jb_err client_ua (struct client_state *csp, char **header);
880 static jb_err client_from (struct client_state *csp, char **header);
881 static jb_err client_send_cookie (struct client_state *csp, char **header);
882 static jb_err client_x_forwarded (struct client_state *csp, char **header);
883 static jb_err client_accept_encoding (struct client_state *csp, char **header);
884 static jb_err client_te (struct client_state *csp, char **header);
885 static jb_err client_max_forwards (struct client_state *csp, char **header);
886 static jb_err client_host (struct client_state *csp, char **header);
887 static jb_err client_if_modified_since (struct client_state *csp, char **header);
888 static jb_err client_accept_language (struct client_state *csp, char **header);
889 static jb_err client_if_none_match (struct client_state *csp, char **header);
890 static jb_err crunch_client_header (struct client_state *csp, char **header);
891 static jb_err client_x_filter (struct client_state *csp, char **header);
892 static jb_err client_range (struct client_state *csp, char **header);
893 static jb_err server_set_cookie (struct client_state *csp, char **header);
894 static jb_err server_content_type (struct client_state *csp, char **header);
895 static jb_err server_content_length (struct client_state *csp, char **header);
896 static jb_err server_content_md5 (struct client_state *csp, char **header);
897 static jb_err server_content_encoding (struct client_state *csp, char **header);
898 static jb_err server_transfer_coding (struct client_state *csp, char **header);
899 static jb_err server_http (struct client_state *csp, char **header);
900 static jb_err crunch_server_header (struct client_state *csp, char **header);
901 static jb_err server_last_modified (struct client_state *csp, char **header);
902 static jb_err server_content_disposition(struct client_state *csp, char **header);
904 static jb_err client_host_adder (struct client_state *csp);
905 static jb_err client_xtra_adder (struct client_state *csp);
906 static jb_err connection_close_adder (struct client_state *csp);
908 static jb_err create_forged_referrer(char **header, const char *hostport);
909 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
910 static jb_err handle_conditional_hide_referrer_parameter(char **header,
911 const char *host, const int parameter_conditional_block);
914 * List of functions to run on a list of headers.
918 /** The header prefix to match */
921 /** The length of the prefix to match */
924 /** The function to apply to this line */
925 const parser_func_ptr parser;
928 static const struct parsers client_patterns[] = {
929 { "referer:", 8, client_referrer },
930 { "user-agent:", 11, client_uagent },
931 { "ua-", 3, client_ua },
932 { "from:", 5, client_from },
933 { "cookie:", 7, client_send_cookie },
934 { "x-forwarded-for:", 16, client_x_forwarded },
935 { "Accept-Encoding:", 16, client_accept_encoding },
936 { "TE:", 3, client_te },
937 { "Host:", 5, client_host },
938 { "if-modified-since:", 18, client_if_modified_since },
939 { "Keep-Alive:", 11, crumble },
940 { "connection:", 11, connection },
941 { "proxy-connection:", 17, crumble },
942 { "max-forwards:", 13, client_max_forwards },
943 { "Accept-Language:", 16, client_accept_language },
944 { "if-none-match:", 14, client_if_none_match },
945 { "Range:", 6, client_range },
946 { "Request-Range:", 14, client_range },
947 { "If-Range:", 9, client_range },
948 { "X-Filter:", 9, client_x_filter },
949 { "*", 0, crunch_client_header },
950 { "*", 0, filter_header },
954 static const struct parsers server_patterns[] = {
955 { "HTTP/", 5, server_http },
956 { "set-cookie:", 11, server_set_cookie },
957 { "connection:", 11, connection },
958 { "Content-Type:", 13, server_content_type },
959 { "Content-MD5:", 12, server_content_md5 },
960 { "Content-Encoding:", 17, server_content_encoding },
961 { "Transfer-Encoding:", 18, server_transfer_coding },
962 { "Keep-Alive:", 11, crumble },
963 { "content-disposition:", 20, server_content_disposition },
964 { "Last-Modified:", 14, server_last_modified },
965 { "*", 0, crunch_server_header },
966 { "*", 0, filter_header },
970 static const add_header_func_ptr add_client_headers[] = {
973 /* Temporarily disabled: client_accept_encoding_adder, */
974 connection_close_adder,
978 static const add_header_func_ptr add_server_headers[] = {
979 connection_close_adder,
983 /*********************************************************************
985 * Function : flush_socket
987 * Description : Write any pending "buffered" content.
990 * 1 : fd = file descriptor of the socket to read
991 * 2 : iob = The I/O buffer to flush, usually csp->iob.
993 * Returns : On success, the number of bytes written are returned (zero
994 * indicates nothing was written). On error, -1 is returned,
995 * and errno is set appropriately. If count is zero and the
996 * file descriptor refers to a regular file, 0 will be
997 * returned without causing any other effect. For a special
998 * file, the results are not portable.
1000 *********************************************************************/
1001 int flush_socket(jb_socket fd, struct iob *iob)
1003 int len = iob->eod - iob->cur;
1010 if (write_socket(fd, iob->cur, (size_t)len))
1014 iob->eod = iob->cur = iob->buf;
1020 /*********************************************************************
1022 * Function : add_to_iob
1024 * Description : Add content to the buffered page, expanding the
1025 * buffer if necessary.
1028 * 1 : csp = Current client state (buffers, headers, etc...)
1029 * 2 : buf = holds the content to be added to the page
1030 * 3 : n = number of bytes to be added
1032 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
1033 * or buffer limit reached.
1035 *********************************************************************/
1036 jb_err add_to_iob(struct client_state *csp, char *buf, int n)
1038 struct iob *iob = csp->iob;
1039 size_t used, offset, need, want;
1042 if (n <= 0) return JB_ERR_OK;
1044 used = (size_t)(iob->eod - iob->buf);
1045 offset = (size_t)(iob->cur - iob->buf);
1046 need = used + (size_t)n + 1;
1049 * If the buffer can't hold the new data, extend it first.
1050 * Use the next power of two if possible, else use the actual need.
1052 if (need > csp->config->buffer_limit)
1054 log_error(LOG_LEVEL_INFO, "Buffer limit reached while extending the buffer (iob)");
1055 return JB_ERR_MEMORY;
1058 if (need > iob->size)
1060 for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2;
1062 if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
1066 else if (NULL != (p = (char *)realloc(iob->buf, need)))
1072 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
1073 return JB_ERR_MEMORY;
1076 /* Update the iob pointers */
1077 iob->cur = p + offset;
1078 iob->eod = p + used;
1082 /* copy the new data into the iob buffer */
1083 memcpy(iob->eod, buf, (size_t)n);
1085 /* point to the end of the data */
1088 /* null terminate == cheap insurance */
1097 /*********************************************************************
1099 * Function : decompress_iob
1101 * Description : Decompress buffered page, expanding the
1102 * buffer as necessary. csp->iob->cur
1103 * should point to the the beginning of the
1104 * compressed data block.
1107 * 1 : csp = Current client state (buffers, headers, etc...)
1109 * Returns : JB_ERR_OK on success,
1110 * JB_ERR_MEMORY if out-of-memory limit reached, and
1111 * JB_ERR_COMPRESS if error decompressing buffer.
1113 *********************************************************************/
1114 jb_err decompress_iob(struct client_state *csp)
1116 char *buf; /* new, uncompressed buffer */
1117 char *cur; /* Current iob position (to keep the original
1118 * iob->cur unmodified if we return early) */
1119 size_t bufsize; /* allocated size of the new buffer */
1120 size_t old_size; /* Content size before decompression */
1121 size_t skip_size; /* Number of bytes at the beginning of the iob
1122 that we should NOT decompress. */
1123 int status; /* return status of the inflate() call */
1124 z_stream zstr; /* used by calls to zlib */
1126 assert(csp->iob->cur - csp->iob->buf > 0);
1127 assert(csp->iob->eod - csp->iob->cur > 0);
1129 bufsize = csp->iob->size;
1130 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
1131 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
1133 cur = csp->iob->cur;
1138 * This is to protect the parsing of gzipped data,
1139 * but it should(?) be valid for deflated data also.
1141 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
1142 return JB_ERR_COMPRESS;
1145 if (csp->content_type & CT_GZIP)
1148 * Our task is slightly complicated by the facts that data
1149 * compressed by gzip does not include a zlib header, and
1150 * that there is no easily accessible interface in zlib to
1151 * handle a gzip header. We strip off the gzip header by
1152 * hand, and later inform zlib not to expect a header.
1156 * Strip off the gzip header. Please see RFC 1952 for more
1157 * explanation of the appropriate fields.
1159 if ((*cur++ != (char)0x1f)
1160 || (*cur++ != (char)0x8b)
1161 || (*cur++ != Z_DEFLATED))
1163 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
1164 return JB_ERR_COMPRESS;
1170 * XXX: These magic numbers should be replaced
1171 * with macros to give a better idea what they do.
1175 /* The gzip header has reserved bits set; bail out. */
1176 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
1177 return JB_ERR_COMPRESS;
1181 /* Skip extra fields if necessary. */
1185 * Skip a given number of bytes, specified
1186 * as a 16-bit little-endian value.
1189 * XXX: This code used to be:
1191 * csp->iob->cur += *csp->iob->cur++ + (*csp->iob->cur++ << 8);
1193 * which I had to change into:
1195 * cur += *cur++ + (*cur++ << 8);
1197 * at which point gcc43 finally noticed that the value
1198 * of cur is undefined (it depends on which of the
1199 * summands is evaluated first).
1201 * I haven't come across a site where this
1202 * code is actually executed yet, but I hope
1206 skip_bytes = *cur++;
1207 skip_bytes = *cur++ << 8;
1209 assert(skip_bytes == *csp->iob->cur - 2 + ((*csp->iob->cur - 1) << 8));
1212 * The number of bytes to skip should be positive
1213 * and we'd like to stay in the buffer.
1215 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
1217 log_error(LOG_LEVEL_ERROR,
1218 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
1220 return JB_ERR_COMPRESS;
1222 log_error(LOG_LEVEL_INFO,
1223 "Skipping %d bytes for gzip compression. Does this sound right?",
1228 /* Skip the filename if necessary. */
1231 /* A null-terminated string is supposed to follow. */
1232 while (*cur++ && (cur < csp->iob->eod));
1236 /* Skip the comment if necessary. */
1239 /* A null-terminated string is supposed to follow. */
1240 while (*cur++ && (cur < csp->iob->eod));
1243 /* Skip the CRC if necessary. */
1249 if (cur >= csp->iob->eod)
1252 * If the current position pointer reached or passed
1253 * the buffer end, we were obviously tricked to skip
1256 log_error(LOG_LEVEL_ERROR,
1257 "Malformed gzip header detected. Aborting decompression.");
1258 return JB_ERR_COMPRESS;
1262 else if (csp->content_type & CT_DEFLATE)
1265 * XXX: The debug level should be lowered
1266 * before the next stable release.
1268 log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
1270 * In theory (that is, according to RFC 1950), deflate-compressed
1271 * data should begin with a two-byte zlib header and have an
1272 * adler32 checksum at the end. It seems that in practice only
1273 * the raw compressed data is sent. Note that this means that
1274 * we are not RFC 1950-compliant here, but the advantage is that
1275 * this actually works. :)
1277 * We add a dummy null byte to tell zlib where the data ends,
1278 * and later inform it not to expect a header.
1280 * Fortunately, add_to_iob() has thoughtfully null-terminated
1281 * the buffer; we can just increment the end pointer to include
1288 log_error(LOG_LEVEL_ERROR,
1289 "Unable to determine compression format for decompression");
1290 return JB_ERR_COMPRESS;
1293 /* Set up the fields required by zlib. */
1294 zstr.next_in = (Bytef *)cur;
1295 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
1296 zstr.zalloc = Z_NULL;
1297 zstr.zfree = Z_NULL;
1298 zstr.opaque = Z_NULL;
1301 * Passing -MAX_WBITS to inflateInit2 tells the library
1302 * that there is no zlib header.
1304 if (inflateInit2 (&zstr, -MAX_WBITS) != Z_OK)
1306 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
1307 return JB_ERR_COMPRESS;
1311 * Next, we allocate new storage for the inflated data.
1312 * We don't modify the existing iob yet, so in case there
1313 * is error in decompression we can recover gracefully.
1315 buf = zalloc(bufsize);
1318 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1319 return JB_ERR_MEMORY;
1322 assert(bufsize >= skip_size);
1323 memcpy(buf, csp->iob->buf, skip_size);
1324 zstr.avail_out = bufsize - skip_size;
1325 zstr.next_out = (Bytef *)buf + skip_size;
1327 /* Try to decompress the whole stream in one shot. */
1328 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
1330 /* We need to allocate more memory for the output buffer. */
1332 char *tmpbuf; /* used for realloc'ing the buffer */
1333 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
1336 * If zlib wants more data then there's a problem, because
1337 * the complete compressed file should have been buffered.
1339 if (0 == zstr.avail_in)
1341 log_error(LOG_LEVEL_ERROR, "Unexpected end of compressed iob");
1342 return JB_ERR_COMPRESS;
1346 * If we tried the limit and still didn't have enough
1347 * memory, just give up.
1349 if (bufsize == csp->config->buffer_limit)
1351 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
1352 return JB_ERR_MEMORY;
1355 /* Try doubling the buffer size each time. */
1358 /* Don't exceed the buffer limit. */
1359 if (bufsize > csp->config->buffer_limit)
1361 bufsize = csp->config->buffer_limit;
1364 /* Try to allocate the new buffer. */
1365 tmpbuf = realloc(buf, bufsize);
1368 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1370 return JB_ERR_MEMORY;
1374 char *oldnext_out = (char *)zstr.next_out;
1377 * Update the fields for inflate() to use the new
1378 * buffer, which may be in a location different from
1381 zstr.avail_out += bufsize - oldbufsize;
1382 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
1385 * Compare with an uglier method of calculating these values
1386 * that doesn't require the extra oldbufsize variable.
1388 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
1389 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
1390 assert(zstr.avail_out > 0);
1396 if (Z_STREAM_ERROR == inflateEnd(&zstr))
1398 log_error(LOG_LEVEL_ERROR,
1399 "Inconsistent stream state after decompression: %s", zstr.msg);
1401 * XXX: Intentionally no return.
1403 * According to zlib.h, Z_STREAM_ERROR is returned
1404 * "if the stream state was inconsistent".
1406 * I assume in this case inflate()'s status
1407 * would also be something different than Z_STREAM_END
1408 * so this check should be redundant, but lets see.
1412 if (status != Z_STREAM_END)
1414 /* We failed to decompress the stream. */
1415 log_error(LOG_LEVEL_ERROR,
1416 "Error in decompressing to the buffer (iob): %s", zstr.msg);
1417 return JB_ERR_COMPRESS;
1421 * Finally, we can actually update the iob, since the
1422 * decompression was successful. First, free the old
1425 freez(csp->iob->buf);
1427 /* Now, update the iob to use the new buffer. */
1428 csp->iob->buf = buf;
1429 csp->iob->cur = csp->iob->buf + skip_size;
1430 csp->iob->eod = (char *)zstr.next_out;
1431 csp->iob->size = bufsize;
1434 * Make sure the new uncompressed iob obeys some minimal
1435 * consistency conditions.
1437 if ((csp->iob->buf < csp->iob->cur)
1438 && (csp->iob->cur <= csp->iob->eod)
1439 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
1441 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
1444 log_error(LOG_LEVEL_RE_FILTER,
1445 "Decompression successful. Old size: %d, new size: %d.",
1446 old_size, new_size);
1450 /* zlib thinks this is OK, so lets do the same. */
1451 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
1456 /* It seems that zlib did something weird. */
1457 log_error(LOG_LEVEL_ERROR,
1458 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
1459 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
1460 csp->iob->eod, csp->iob->buf + csp->iob->size);
1461 return JB_ERR_COMPRESS;
1467 #endif /* defined(FEATURE_ZLIB) */
1470 /*********************************************************************
1472 * Function : string_move
1474 * Description : memmove wrapper to move the last part of a string
1475 * towards the beginning, overwriting the part in
1476 * the middle. strlcpy() can't be used here as the
1480 * 1 : dst = Destination to overwrite
1481 * 2 : src = Source to move.
1485 *********************************************************************/
1486 static void string_move(char *dst, char *src)
1490 /* +1 to copy the terminating nul as well. */
1491 memmove(dst, src, strlen(src)+1);
1495 /*********************************************************************
1497 * Function : normalize_lws
1499 * Description : Reduces unquoted linear white space in headers
1500 * to a single space in accordance with RFC 2616 2.2.
1501 * This simplifies parsing and filtering later on.
1503 * XXX: Remove log messages before
1504 * the next stable release?
1507 * 1 : header = A header with linear white space to reduce.
1511 *********************************************************************/
1512 static void normalize_lws(char *header)
1518 if (ijb_isspace(*p) && ijb_isspace(*(p+1)))
1522 while (ijb_isspace(*q))
1526 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
1527 string_move(p+1, q);
1532 log_error(LOG_LEVEL_HEADER,
1533 "Converting tab to space in '%s'", header);
1538 char *end_of_token = strstr(p+1, "\"");
1540 if (NULL != end_of_token)
1542 /* Don't mess with quoted text. */
1547 log_error(LOG_LEVEL_HEADER,
1548 "Ignoring single quote in '%s'", header);
1554 p = strchr(header, ':');
1555 if ((p != NULL) && (p != header) && ijb_isspace(*(p-1)))
1558 * There's still space before the colon.
1561 string_move(p-1, p);
1566 /*********************************************************************
1568 * Function : get_header
1570 * Description : This (odd) routine will parse the csp->iob
1571 * to get the next complete header.
1574 * 1 : iob = The I/O buffer to parse, usually csp->iob.
1576 * Returns : Any one of the following:
1578 * 1) a pointer to a dynamically allocated string that contains a header line
1579 * 2) NULL indicating that the end of the header was reached
1580 * 3) "" indicating that the end of the iob was reached before finding
1581 * a complete header line.
1583 *********************************************************************/
1584 char *get_header(struct iob *iob)
1588 header = get_header_line(iob);
1590 if ((header == NULL) || (*header == '\0'))
1593 * No complete header read yet, tell the client.
1598 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
1601 * Header spans multiple lines, append the next one.
1603 char *continued_header;
1605 continued_header = get_header_line(iob);
1606 if ((continued_header == NULL) || (*continued_header == '\0'))
1609 * No complete header read yet, return what we got.
1610 * XXX: Should "unread" header instead.
1612 log_error(LOG_LEVEL_INFO,
1613 "Failed to read a multi-line header properly: '%s'",
1618 if (JB_ERR_OK != string_join(&header, continued_header))
1620 log_error(LOG_LEVEL_FATAL,
1621 "Out of memory while appending multiple headers.");
1625 /* XXX: remove before next stable release. */
1626 log_error(LOG_LEVEL_HEADER,
1627 "Merged multiple header lines to: '%s'",
1632 normalize_lws(header);
1639 /*********************************************************************
1641 * Function : get_header_line
1643 * Description : This (odd) routine will parse the csp->iob
1644 * to get the next header line.
1647 * 1 : iob = The I/O buffer to parse, usually csp->iob.
1649 * Returns : Any one of the following:
1651 * 1) a pointer to a dynamically allocated string that contains a header line
1652 * 2) NULL indicating that the end of the header was reached
1653 * 3) "" indicating that the end of the iob was reached before finding
1654 * a complete header line.
1656 *********************************************************************/
1657 static char *get_header_line(struct iob *iob)
1661 if ((iob->cur == NULL)
1662 || ((p = strchr(iob->cur, '\n')) == NULL))
1664 return(""); /* couldn't find a complete header */
1669 ret = strdup(iob->cur);
1672 /* FIXME No way to handle error properly */
1673 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
1678 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
1680 /* is this a blank line (i.e. the end of the header) ? */
1692 /*********************************************************************
1694 * Function : get_header_value
1696 * Description : Get the value of a given header from a chained list
1697 * of header lines or return NULL if no such header is
1698 * present in the list.
1701 * 1 : header_list = pointer to list
1702 * 2 : header_name = string with name of header to look for.
1703 * Trailing colon required, capitalization
1706 * Returns : NULL if not found, else value of header
1708 *********************************************************************/
1709 char *get_header_value(const struct list *header_list, const char *header_name)
1711 struct list_entry *cur_entry;
1715 assert(header_list);
1716 assert(header_name);
1717 length = strlen(header_name);
1719 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
1723 if (!strncmpic(cur_entry->str, header_name, length))
1726 * Found: return pointer to start of value
1728 ret = (char *) (cur_entry->str + length);
1729 while (*ret && ijb_isspace(*ret)) ret++;
1743 /*********************************************************************
1745 * Function : scan_headers
1747 * Description : Scans headers, applies tags and updates action bits.
1750 * 1 : csp = Current client state (buffers, headers, etc...)
1752 * Returns : JB_ERR_OK
1754 *********************************************************************/
1755 static jb_err scan_headers(struct client_state *csp)
1757 struct list_entry *h; /* Header */
1758 jb_err err = JB_ERR_OK;
1760 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1762 /* Header crunch()ed in previous run? -> ignore */
1763 if (h->str == NULL) continue;
1764 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1765 err = header_tagger(csp, h->str);
1772 /*********************************************************************
1776 * Description : add, delete or modify lines in the HTTP header streams.
1777 * On entry, it receives a linked list of headers space
1778 * that was allocated dynamically (both the list nodes
1779 * and the header contents).
1781 * As a side effect it frees the space used by the original
1785 * 1 : csp = Current client state (buffers, headers, etc...)
1786 * 2 : filter_server_headers = Boolean to switch between
1787 * server and header filtering.
1789 * Returns : JB_ERR_OK in case off success, or
1790 * JB_ERR_MEMORY on out-of-memory error.
1792 *********************************************************************/
1793 jb_err sed(struct client_state *csp, int filter_server_headers)
1795 /* XXX: use more descriptive names. */
1796 struct list_entry *p;
1797 const struct parsers *v;
1798 const add_header_func_ptr *f;
1799 jb_err err = JB_ERR_OK;
1801 if (filter_server_headers)
1803 v = server_patterns;
1804 f = add_server_headers;
1808 v = client_patterns;
1809 f = add_client_headers;
1814 while ((err == JB_ERR_OK) && (v->str != NULL))
1816 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1818 /* Header crunch()ed in previous run? -> ignore */
1819 if (p->str == NULL) continue;
1821 /* Does the current parser handle this header? */
1822 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1823 (v->len == CHECK_EVERY_HEADER_REMAINING))
1825 err = v->parser(csp, &(p->str));
1831 /* place additional headers on the csp->headers list */
1832 while ((err == JB_ERR_OK) && (*f))
1842 /*********************************************************************
1844 * Function : update_server_headers
1846 * Description : Updates server headers after the body has been modified.
1849 * 1 : csp = Current client state (buffers, headers, etc...)
1851 * Returns : JB_ERR_OK in case off success, or
1852 * JB_ERR_MEMORY on out-of-memory error.
1854 *********************************************************************/
1855 jb_err update_server_headers(struct client_state *csp)
1857 jb_err err = JB_ERR_OK;
1859 static const struct parsers server_patterns_light[] = {
1860 { "Content-Length:", 15, server_content_length },
1861 { "Transfer-Encoding:", 18, server_transfer_coding },
1863 { "Content-Encoding:", 17, server_content_encoding },
1864 #endif /* def FEATURE_ZLIB */
1868 if (strncmpic(csp->http->cmd, "HEAD", 4))
1870 const struct parsers *v;
1871 struct list_entry *p;
1873 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1875 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1877 /* Header crunch()ed in previous run? -> ignore */
1878 if (p->str == NULL) continue;
1880 /* Does the current parser handle this header? */
1881 if (strncmpic(p->str, v->str, v->len) == 0)
1883 err = v->parser(csp, (char **)&(p->str));
1893 /*********************************************************************
1895 * Function : header_tagger
1897 * Description : Executes all text substitutions from applying
1898 * tag actions and saves the result as tag.
1900 * XXX: Shares enough code with filter_header() and
1901 * pcrs_filter_response() to warrant some helper functions.
1904 * 1 : csp = Current client state (buffers, headers, etc...)
1905 * 2 : header = Header that is used as tagger input
1907 * Returns : JB_ERR_OK on success and always succeeds
1909 *********************************************************************/
1910 static jb_err header_tagger(struct client_state *csp, char *header)
1912 int wanted_filter_type;
1913 int multi_action_index;
1917 struct file_list *fl;
1918 struct re_filterfile_spec *b;
1919 struct list_entry *tag_name;
1921 int found_filters = 0;
1922 const size_t header_length = strlen(header);
1924 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1926 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1927 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1931 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1932 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1935 /* Check if there are any filters */
1936 for (i = 0; i < MAX_AF_FILES; i++)
1949 if (0 == found_filters)
1951 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1952 "tagging enabled, but no taggers available.");
1956 for (i = 0; i < MAX_AF_FILES; i++)
1959 if ((NULL == fl) || (NULL == fl->f))
1962 * Either there are no filter files
1963 * left, or this filter file just
1964 * contains no valid filters.
1966 * Continue to be sure we don't miss
1967 * valid filter files that are chained
1968 * after empty or invalid ones.
1973 /* For all filters, */
1974 for (b = fl->f; b; b = b->next)
1976 if (b->type != wanted_filter_type)
1978 /* skip the ones we don't care about, */
1981 /* leaving only taggers that could apply, of which we use the ones, */
1982 for (tag_name = csp->action->multi[multi_action_index]->first;
1983 NULL != tag_name; tag_name = tag_name->next)
1985 /* that do apply, and */
1986 if (strcmp(b->name, tag_name->str) == 0)
1988 char *modified_tag = NULL;
1990 size_t size = header_length;
1991 pcrs_job *joblist = b->joblist;
1993 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1995 if (NULL == joblist)
1997 log_error(LOG_LEVEL_RE_FILTER,
1998 "Tagger %s has empty joblist. Nothing to do.", b->name);
2002 /* execute their pcrs_joblist on the header. */
2003 for (job = joblist; NULL != job; job = job->next)
2005 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
2009 /* Success, continue with the modified version. */
2018 /* Tagger doesn't match */
2021 /* Regex failure, log it but continue anyway. */
2022 log_error(LOG_LEVEL_ERROR,
2023 "Problems with tagger \'%s\' and header \'%s\': %s",
2024 b->name, *header, pcrs_strerror(hits));
2026 freez(modified_tag);
2030 if (b->dynamic) pcrs_free_joblist(joblist);
2032 /* If this tagger matched */
2038 * There is to technical limitation which makes
2039 * it impossible to use empty tags, but I assume
2040 * no one would do it intentionally.
2043 log_error(LOG_LEVEL_INFO,
2044 "Tagger \'%s\' created an empty tag. Ignored.",
2049 if (!list_contains_item(csp->tags, tag))
2051 if (JB_ERR_OK != enlist(csp->tags, tag))
2053 log_error(LOG_LEVEL_ERROR,
2054 "Insufficient memory to add tag \'%s\', "
2055 "based on tagger \'%s\' and header \'%s\'",
2056 tag, b->name, *header);
2060 char *action_message;
2062 * update the action bits right away, to make
2063 * tagging based on tags set by earlier taggers
2064 * of the same kind possible.
2066 if (update_action_bits_for_tag(csp, tag))
2068 action_message = "Action bits updated accordingly.";
2072 action_message = "No action bits update necessary.";
2075 log_error(LOG_LEVEL_HEADER,
2076 "Tagger \'%s\' added tag \'%s\'. %s",
2077 b->name, tag, action_message);
2082 /* XXX: Is this log-worthy? */
2083 log_error(LOG_LEVEL_HEADER,
2084 "Tagger \'%s\' didn't add tag \'%s\'. "
2085 "Tag already present", b->name, tag);
2088 } /* if the tagger matched */
2089 } /* if the tagger applies */
2090 } /* for every tagger that could apply */
2091 } /* for all filters */
2092 } /* for all filter files */
2097 /* here begins the family of parser functions that reformat header lines */
2099 /*********************************************************************
2101 * Function : filter_header
2103 * Description : Executes all text substitutions from all applying
2104 * +(server|client)-header-filter actions on the header.
2105 * Most of the code was copied from pcrs_filter_response,
2106 * including the rather short variable names
2109 * 1 : csp = Current client state (buffers, headers, etc...)
2110 * 2 : header = On input, pointer to header to modify.
2111 * On output, pointer to the modified header, or NULL
2112 * to remove the header. This function frees the
2113 * original string if necessary.
2115 * Returns : JB_ERR_OK on success and always succeeds
2117 *********************************************************************/
2118 static jb_err filter_header(struct client_state *csp, char **header)
2122 size_t size = strlen(*header);
2124 char *newheader = NULL;
2127 struct file_list *fl;
2128 struct re_filterfile_spec *b;
2129 struct list_entry *filtername;
2131 int i, found_filters = 0;
2132 int wanted_filter_type;
2133 int multi_action_index;
2135 if (csp->flags & CSP_FLAG_NO_FILTERING)
2140 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
2142 wanted_filter_type = FT_SERVER_HEADER_FILTER;
2143 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
2147 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
2148 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
2152 * Need to check the set of re_filterfiles...
2154 for (i = 0; i < MAX_AF_FILES; i++)
2167 if (0 == found_filters)
2169 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2170 "header filtering enabled, but no matching filters available.");
2174 for (i = 0; i < MAX_AF_FILES; i++)
2177 if ((NULL == fl) || (NULL == fl->f))
2180 * Either there are no filter files
2181 * left, or this filter file just
2182 * contains no valid filters.
2184 * Continue to be sure we don't miss
2185 * valid filter files that are chained
2186 * after empty or invalid ones.
2191 * For all applying +filter actions, look if a filter by that
2192 * name exists and if yes, execute its pcrs_joblist on the
2195 for (b = fl->f; b; b = b->next)
2197 if (b->type != wanted_filter_type)
2199 /* Skip other filter types */
2203 for (filtername = csp->action->multi[multi_action_index]->first;
2204 filtername ; filtername = filtername->next)
2206 if (strcmp(b->name, filtername->str) == 0)
2208 int current_hits = 0;
2209 pcrs_job *joblist = b->joblist;
2211 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2213 if (NULL == joblist)
2215 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2219 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
2220 *header, size, b->name);
2222 /* Apply all jobs from the joblist */
2223 for (job = joblist; NULL != job; job = job->next)
2225 matches = pcrs_execute(job, *header, size, &newheader, &size);
2228 current_hits += matches;
2229 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
2231 *header = newheader;
2233 else if ( 0 == matches )
2235 /* Filter doesn't change header */
2241 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
2242 *header, b->name, pcrs_strerror(matches));
2243 if (newheader != NULL)
2245 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
2251 if (b->dynamic) pcrs_free_joblist(joblist);
2253 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
2254 hits += current_hits;
2261 * Additionally checking for hits is important because if
2262 * the continue hack is triggered, server headers can
2263 * arrive empty to separate multiple heads from each other.
2265 if ((0 == size) && hits)
2267 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
2275 /*********************************************************************
2277 * Function : connection
2279 * Description : Makes sure that the value of the Connection: header
2280 * is "close" and signals connection_close_adder
2284 * 1 : csp = Current client state (buffers, headers, etc...)
2285 * 2 : header = On input, pointer to header to modify.
2286 * On output, pointer to the modified header, or NULL
2287 * to remove the header. This function frees the
2288 * original string if necessary.
2290 * Returns : JB_ERR_OK on success, or
2291 * JB_ERR_MEMORY on out-of-memory error.
2293 *********************************************************************/
2294 static jb_err connection(struct client_state *csp, char **header)
2296 char *old_header = *header;
2298 /* Do we have a 'Connection: close' header? */
2299 if (strcmpic(*header, "Connection: close"))
2301 /* No, create one */
2302 *header = strdup("Connection: close");
2305 return JB_ERR_MEMORY;
2307 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
2311 /* Signal connection_close_adder() to return early. */
2312 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
2314 csp->flags |= CSP_FLAG_SERVER_CONNECTION_CLOSE_SET;
2318 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET;
2325 /*********************************************************************
2327 * Function : crumble
2329 * Description : This is called if a header matches a pattern to "crunch"
2332 * 1 : csp = Current client state (buffers, headers, etc...)
2333 * 2 : header = On input, pointer to header to modify.
2334 * On output, pointer to the modified header, or NULL
2335 * to remove the header. This function frees the
2336 * original string if necessary.
2338 * Returns : JB_ERR_OK on success, or
2339 * JB_ERR_MEMORY on out-of-memory error.
2341 *********************************************************************/
2342 static jb_err crumble(struct client_state *csp, char **header)
2344 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2350 /*********************************************************************
2352 * Function : crunch_server_header
2354 * Description : Crunch server header if it matches a string supplied by the
2355 * user. Called from `sed'.
2358 * 1 : csp = Current client state (buffers, headers, etc...)
2359 * 2 : header = On input, pointer to header to modify.
2360 * On output, pointer to the modified header, or NULL
2361 * to remove the header. This function frees the
2362 * original string if necessary.
2364 * Returns : JB_ERR_OK on success and always succeeds
2366 *********************************************************************/
2367 static jb_err crunch_server_header(struct client_state *csp, char **header)
2369 const char *crunch_pattern;
2371 /* Do we feel like crunching? */
2372 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2374 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2376 /* Is the current header the lucky one? */
2377 if (strstr(*header, crunch_pattern))
2379 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2388 /*********************************************************************
2390 * Function : server_content_type
2392 * Description : Set the content-type for filterable types (text/.*,
2393 * .*xml.*, javascript and image/gif) unless filtering has been
2394 * forbidden (CT_TABOO) while parsing earlier headers.
2395 * NOTE: Since text/plain is commonly used by web servers
2396 * for files whose correct type is unknown, we don't
2397 * set CT_TEXT for it.
2400 * 1 : csp = Current client state (buffers, headers, etc...)
2401 * 2 : header = On input, pointer to header to modify.
2402 * On output, pointer to the modified header, or NULL
2403 * to remove the header. This function frees the
2404 * original string if necessary.
2406 * Returns : JB_ERR_OK on success, or
2407 * JB_ERR_MEMORY on out-of-memory error.
2409 *********************************************************************/
2410 static jb_err server_content_type(struct client_state *csp, char **header)
2412 /* Remove header if it isn't the first Content-Type header */
2413 if ((csp->content_type & CT_DECLARED))
2416 * Another, slightly slower, way to see if
2417 * we already parsed another Content-Type header.
2419 assert(NULL != get_header_value(csp->headers, "Content-Type:"));
2421 log_error(LOG_LEVEL_ERROR,
2422 "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
2430 * Signal that the Content-Type has been set.
2432 csp->content_type |= CT_DECLARED;
2434 if (!(csp->content_type & CT_TABOO))
2437 * XXX: The assumption that text/plain is a sign of
2438 * binary data seems to be somewhat unreasonable nowadays
2439 * and should be dropped after 3.0.8 is out.
2441 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2442 || strstr(*header, "xml")
2443 || strstr(*header, "application/x-javascript"))
2445 csp->content_type |= CT_TEXT;
2447 else if (strstr(*header, "image/gif"))
2449 csp->content_type |= CT_GIF;
2454 * Are we messing with the content type?
2456 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2459 * Make sure the user doesn't accidently
2460 * change the content type of binary documents.
2462 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2465 *header = strdup("Content-Type: ");
2466 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2470 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2471 return JB_ERR_MEMORY;
2473 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2477 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2478 "It doesn't look like a content type that should be filtered. "
2479 "Enable force-text-mode if you know what you're doing.", *header);
2487 /*********************************************************************
2489 * Function : server_transfer_coding
2491 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2492 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2493 * - Remove header if body was chunked but has been
2494 * de-chunked for filtering.
2497 * 1 : csp = Current client state (buffers, headers, etc...)
2498 * 2 : header = On input, pointer to header to modify.
2499 * On output, pointer to the modified header, or NULL
2500 * to remove the header. This function frees the
2501 * original string if necessary.
2503 * Returns : JB_ERR_OK on success, or
2504 * JB_ERR_MEMORY on out-of-memory error.
2506 *********************************************************************/
2507 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2510 * Turn off pcrs and gif filtering if body compressed
2512 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2516 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2518 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2519 csp->http->cmd, *header);
2520 #endif /* def FEATURE_ZLIB */
2521 csp->content_type = CT_TABOO;
2525 * Raise flag if body chunked
2527 if (strstr(*header, "chunked"))
2529 csp->flags |= CSP_FLAG_CHUNKED;
2532 * If the body was modified, it has been de-chunked first
2533 * and the header must be removed.
2535 * FIXME: If there is more than one transfer encoding,
2536 * only the "chunked" part should be removed here.
2538 if (csp->flags & CSP_FLAG_MODIFIED)
2540 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2549 /*********************************************************************
2551 * Function : server_content_encoding
2553 * Description : This function is run twice for each request,
2554 * unless FEATURE_ZLIB and filtering are disabled.
2556 * The first run is used to check if the content
2557 * is compressed, if FEATURE_ZLIB is disabled
2558 * filtering is then disabled as well, if FEATURE_ZLIB
2559 * is enabled the content is marked for decompression.
2561 * The second run is used to remove the Content-Encoding
2562 * header if the decompression was successful.
2565 * 1 : csp = Current client state (buffers, headers, etc...)
2566 * 2 : header = On input, pointer to header to modify.
2567 * On output, pointer to the modified header, or NULL
2568 * to remove the header. This function frees the
2569 * original string if necessary.
2571 * Returns : JB_ERR_OK on success, or
2572 * JB_ERR_MEMORY on out-of-memory error.
2574 *********************************************************************/
2575 static jb_err server_content_encoding(struct client_state *csp, char **header)
2578 if ((csp->flags & CSP_FLAG_MODIFIED)
2579 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2582 * We successfully decompressed the content,
2583 * and have to clean the header now, so the
2584 * client no longer expects compressed data..
2586 * XXX: There is a difference between cleaning
2587 * and removing it completely.
2589 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2592 else if (strstr(*header, "gzip"))
2594 /* Mark for gzip decompression */
2595 csp->content_type |= CT_GZIP;
2597 else if (strstr(*header, "deflate"))
2599 /* Mark for zlib decompression */
2600 csp->content_type |= CT_DEFLATE;
2602 else if (strstr(*header, "compress"))
2605 * We can't decompress this; therefore we can't filter
2608 csp->content_type |= CT_TABOO;
2610 #else /* !defined(FEATURE_ZLIB) */
2611 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2614 * Body is compressed, turn off pcrs and gif filtering.
2616 csp->content_type |= CT_TABOO;
2619 * Log a warning if the user expects the content to be filtered.
2621 if ((csp->rlist != NULL) &&
2622 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2624 log_error(LOG_LEVEL_INFO,
2625 "Compressed content detected, content filtering disabled. "
2626 "Consider recompiling Privoxy with zlib support or "
2627 "enable the prevent-compression action.");
2630 #endif /* defined(FEATURE_ZLIB) */
2637 /*********************************************************************
2639 * Function : server_content_length
2641 * Description : Adjust Content-Length header if we modified
2645 * 1 : csp = Current client state (buffers, headers, etc...)
2646 * 2 : header = On input, pointer to header to modify.
2647 * On output, pointer to the modified header, or NULL
2648 * to remove the header. This function frees the
2649 * original string if necessary.
2651 * Returns : JB_ERR_OK on success, or
2652 * JB_ERR_MEMORY on out-of-memory error.
2654 *********************************************************************/
2655 static jb_err server_content_length(struct client_state *csp, char **header)
2657 const size_t max_header_length = 80;
2659 /* Regenerate header if the content was modified. */
2660 if (csp->flags & CSP_FLAG_MODIFIED)
2663 *header = (char *) zalloc(max_header_length);
2664 if (*header == NULL)
2666 return JB_ERR_MEMORY;
2669 snprintf(*header, max_header_length, "Content-Length: %d",
2670 (int)csp->content_length);
2671 log_error(LOG_LEVEL_HEADER, "Adjusted Content-Length to %d",
2672 (int)csp->content_length);
2679 /*********************************************************************
2681 * Function : server_content_md5
2683 * Description : Crumble any Content-MD5 headers if the document was
2684 * modified. FIXME: Should we re-compute instead?
2687 * 1 : csp = Current client state (buffers, headers, etc...)
2688 * 2 : header = On input, pointer to header to modify.
2689 * On output, pointer to the modified header, or NULL
2690 * to remove the header. This function frees the
2691 * original string if necessary.
2693 * Returns : JB_ERR_OK on success, or
2694 * JB_ERR_MEMORY on out-of-memory error.
2696 *********************************************************************/
2697 static jb_err server_content_md5(struct client_state *csp, char **header)
2699 if (csp->flags & CSP_FLAG_MODIFIED)
2701 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2709 /*********************************************************************
2711 * Function : server_content_disposition
2713 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2714 * Called from `sed'.
2717 * 1 : csp = Current client state (buffers, headers, etc...)
2718 * 2 : header = On input, pointer to header to modify.
2719 * On output, pointer to the modified header, or NULL
2720 * to remove the header. This function frees the
2721 * original string if necessary.
2723 * Returns : JB_ERR_OK on success, or
2724 * JB_ERR_MEMORY on out-of-memory error.
2726 *********************************************************************/
2727 static jb_err server_content_disposition(struct client_state *csp, char **header)
2732 * Are we messing with the Content-Disposition header?
2734 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2740 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2742 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2745 * Blocking content-disposition header
2747 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2754 * Replacing Content-Disposition header
2757 *header = strdup("Content-Disposition: ");
2758 string_append(header, newval);
2760 if (*header != NULL)
2762 log_error(LOG_LEVEL_HEADER,
2763 "Content-Disposition header crunched and replaced with: %s", *header);
2766 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2770 /*********************************************************************
2772 * Function : server_last_modified
2774 * Description : Changes Last-Modified header to the actual date
2775 * to help hide-if-modified-since.
2776 * Called from `sed'.
2779 * 1 : csp = Current client state (buffers, headers, etc...)
2780 * 2 : header = On input, pointer to header to modify.
2781 * On output, pointer to the modified header, or NULL
2782 * to remove the header. This function frees the
2783 * original string if necessary.
2785 * Returns : JB_ERR_OK on success, or
2786 * JB_ERR_MEMORY on out-of-memory error.
2788 *********************************************************************/
2789 static jb_err server_last_modified(struct client_state *csp, char **header)
2792 char buf[BUFFER_SIZE];
2795 #ifdef HAVE_GMTIME_R
2798 struct tm *timeptr = NULL;
2799 time_t now, last_modified;
2801 long int days, hours, minutes, seconds;
2804 * Are we messing with the Last-Modified header?
2806 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2812 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2814 if (0 == strcmpic(newval, "block") )
2817 * Blocking Last-Modified header. Useless but why not.
2819 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2823 else if (0 == strcmpic(newval, "reset-to-request-time"))
2826 * Setting Last-Modified Header to now.
2828 get_http_time(0, buf, sizeof(buf));
2830 *header = strdup("Last-Modified: ");
2831 string_append(header, buf);
2833 if (*header == NULL)
2835 log_error(LOG_LEVEL_HEADER, "Insufficent memory. Last-Modified header got lost, boohoo.");
2839 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2842 else if (0 == strcmpic(newval, "randomize"))
2844 const char *header_time = *header + sizeof("Last-Modified:");
2846 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2848 #ifdef HAVE_GMTIME_R
2849 timeptr = gmtime_r(&now, &gmt);
2850 #elif FEATURE_PTHREAD
2851 pthread_mutex_lock(&gmtime_mutex);
2852 timeptr = gmtime(&now);
2853 pthread_mutex_unlock(&gmtime_mutex);
2855 timeptr = gmtime(&now);
2857 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2859 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2864 rtime = (long int)difftime(now, last_modified);
2873 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2875 rtime = pick_from_range(rtime);
2876 if (negative) rtime *= -1;
2877 last_modified += rtime;
2878 #ifdef HAVE_GMTIME_R
2879 timeptr = gmtime_r(&last_modified, &gmt);
2880 #elif FEATURE_PTHREAD
2881 pthread_mutex_lock(&gmtime_mutex);
2882 timeptr = gmtime(&last_modified);
2883 pthread_mutex_unlock(&gmtime_mutex);
2885 timeptr = gmtime(&last_modified);
2887 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
2889 *header = strdup("Last-Modified: ");
2890 string_append(header, newheader);
2892 if (*header == NULL)
2894 log_error(LOG_LEVEL_ERROR, "Insufficent memory, header crunched without replacement.");
2895 return JB_ERR_MEMORY;
2898 if (LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */
2900 days = rtime / (3600 * 24);
2901 hours = rtime / 3600 % 24;
2902 minutes = rtime / 60 % 60;
2903 seconds = rtime % 60;
2905 log_error(LOG_LEVEL_HEADER, "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2906 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2907 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2912 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2921 /*********************************************************************
2923 * Function : client_accept_encoding
2925 * Description : Rewrite the client's Accept-Encoding header so that
2926 * if doesn't allow compression, if the action applies.
2927 * Note: For HTTP/1.0 the absence of the header is enough.
2930 * 1 : csp = Current client state (buffers, headers, etc...)
2931 * 2 : header = On input, pointer to header to modify.
2932 * On output, pointer to the modified header, or NULL
2933 * to remove the header. This function frees the
2934 * original string if necessary.
2936 * Returns : JB_ERR_OK on success, or
2937 * JB_ERR_MEMORY on out-of-memory error.
2939 *********************************************************************/
2940 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2942 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2944 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2948 /* Temporarily disable the correct behaviour to
2949 * work around a PHP bug.
2951 * if (!strcmpic(csp->http->ver, "HTTP/1.1"))
2953 * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0");
2954 * if (*header == NULL)
2956 * return JB_ERR_MEMORY;
2967 /*********************************************************************
2969 * Function : client_te
2971 * Description : Rewrite the client's TE header so that
2972 * if doesn't allow compression, if the action applies.
2975 * 1 : csp = Current client state (buffers, headers, etc...)
2976 * 2 : header = On input, pointer to header to modify.
2977 * On output, pointer to the modified header, or NULL
2978 * to remove the header. This function frees the
2979 * original string if necessary.
2981 * Returns : JB_ERR_OK on success, or
2982 * JB_ERR_MEMORY on out-of-memory error.
2984 *********************************************************************/
2985 static jb_err client_te(struct client_state *csp, char **header)
2987 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2990 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2997 /*********************************************************************
2999 * Function : client_referrer
3001 * Description : Handle the "referer" config setting properly.
3002 * Called from `sed'.
3005 * 1 : csp = Current client state (buffers, headers, etc...)
3006 * 2 : header = On input, pointer to header to modify.
3007 * On output, pointer to the modified header, or NULL
3008 * to remove the header. This function frees the
3009 * original string if necessary.
3011 * Returns : JB_ERR_OK on success, or
3012 * JB_ERR_MEMORY on out-of-memory error.
3014 *********************************************************************/
3015 static jb_err client_referrer(struct client_state *csp, char **header)
3017 const char *parameter;
3018 /* booleans for parameters we have to check multiple times */
3019 int parameter_conditional_block;
3020 int parameter_conditional_forge;
3022 #ifdef FEATURE_FORCE_LOAD
3024 * Since the referrer can include the prefix even
3025 * if the request itself is non-forced, we must
3026 * clean it unconditionally.
3028 * XXX: strclean is too broad
3030 strclean(*header, FORCE_PREFIX);
3031 #endif /* def FEATURE_FORCE_LOAD */
3033 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
3035 /* Nothing left to do */
3039 parameter = csp->action->string[ACTION_STRING_REFERER];
3040 assert(parameter != NULL);
3041 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
3042 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
3044 if (!parameter_conditional_block && !parameter_conditional_forge)
3047 * As conditional-block and conditional-forge are the only
3048 * parameters that rely on the original referrer, we can
3049 * remove it now for all the others.
3054 if (0 == strcmpic(parameter, "block"))
3056 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
3059 else if (parameter_conditional_block || parameter_conditional_forge)
3061 return handle_conditional_hide_referrer_parameter(header,
3062 csp->http->hostport, parameter_conditional_block);
3064 else if (0 == strcmpic(parameter, "forge"))
3066 return create_forged_referrer(header, csp->http->hostport);
3070 /* interpret parameter as user-supplied referer to fake */
3071 return create_fake_referrer(header, parameter);
3076 /*********************************************************************
3078 * Function : client_accept_language
3080 * Description : Handle the "Accept-Language" config setting properly.
3081 * Called from `sed'.
3084 * 1 : csp = Current client state (buffers, headers, etc...)
3085 * 2 : header = On input, pointer to header to modify.
3086 * On output, pointer to the modified header, or NULL
3087 * to remove the header. This function frees the
3088 * original string if necessary.
3090 * Returns : JB_ERR_OK on success, or
3091 * JB_ERR_MEMORY on out-of-memory error.
3093 *********************************************************************/
3094 static jb_err client_accept_language(struct client_state *csp, char **header)
3099 * Are we messing with the Accept-Language?
3101 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
3103 /*I don't think so*/
3107 newval = csp->action->string[ACTION_STRING_LANGUAGE];
3109 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3112 * Blocking Accept-Language header
3114 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
3121 * Replacing Accept-Language header
3124 *header = strdup("Accept-Language: ");
3125 string_append(header, newval);
3127 if (*header == NULL)
3129 log_error(LOG_LEVEL_ERROR,
3130 "Insufficent memory. Accept-Language header crunched without replacement.");
3134 log_error(LOG_LEVEL_HEADER,
3135 "Accept-Language header crunched and replaced with: %s", *header);
3138 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3142 /*********************************************************************
3144 * Function : crunch_client_header
3146 * Description : Crunch client header if it matches a string supplied by the
3147 * user. Called from `sed'.
3150 * 1 : csp = Current client state (buffers, headers, etc...)
3151 * 2 : header = On input, pointer to header to modify.
3152 * On output, pointer to the modified header, or NULL
3153 * to remove the header. This function frees the
3154 * original string if necessary.
3156 * Returns : JB_ERR_OK on success and always succeeds
3158 *********************************************************************/
3159 static jb_err crunch_client_header(struct client_state *csp, char **header)
3161 const char *crunch_pattern;
3163 /* Do we feel like crunching? */
3164 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
3166 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
3168 /* Is the current header the lucky one? */
3169 if (strstr(*header, crunch_pattern))
3171 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
3179 /*********************************************************************
3181 * Function : client_uagent
3183 * Description : Handle the "user-agent" config setting properly
3184 * and remember its original value to enable browser
3185 * bug workarounds. Called from `sed'.
3188 * 1 : csp = Current client state (buffers, headers, etc...)
3189 * 2 : header = On input, pointer to header to modify.
3190 * On output, pointer to the modified header, or NULL
3191 * to remove the header. This function frees the
3192 * original string if necessary.
3194 * Returns : JB_ERR_OK on success, or
3195 * JB_ERR_MEMORY on out-of-memory error.
3197 *********************************************************************/
3198 static jb_err client_uagent(struct client_state *csp, char **header)
3202 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
3207 newval = csp->action->string[ACTION_STRING_USER_AGENT];
3214 *header = strdup("User-Agent: ");
3215 string_append(header, newval);
3217 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
3219 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3223 /*********************************************************************
3225 * Function : client_ua
3227 * Description : Handle "ua-" headers properly. Called from `sed'.
3230 * 1 : csp = Current client state (buffers, headers, etc...)
3231 * 2 : header = On input, pointer to header to modify.
3232 * On output, pointer to the modified header, or NULL
3233 * to remove the header. This function frees the
3234 * original string if necessary.
3236 * Returns : JB_ERR_OK on success, or
3237 * JB_ERR_MEMORY on out-of-memory error.
3239 *********************************************************************/
3240 static jb_err client_ua(struct client_state *csp, char **header)
3242 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3244 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3252 /*********************************************************************
3254 * Function : client_from
3256 * Description : Handle the "from" config setting properly.
3257 * Called from `sed'.
3260 * 1 : csp = Current client state (buffers, headers, etc...)
3261 * 2 : header = On input, pointer to header to modify.
3262 * On output, pointer to the modified header, or NULL
3263 * to remove the header. This function frees the
3264 * original string if necessary.
3266 * Returns : JB_ERR_OK on success, or
3267 * JB_ERR_MEMORY on out-of-memory error.
3269 *********************************************************************/
3270 static jb_err client_from(struct client_state *csp, char **header)
3274 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3281 newval = csp->action->string[ACTION_STRING_FROM];
3284 * Are we blocking the e-mail address?
3286 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3288 log_error(LOG_LEVEL_HEADER, "crunched From!");
3292 log_error(LOG_LEVEL_HEADER, " modified");
3294 *header = strdup("From: ");
3295 string_append(header, newval);
3297 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3301 /*********************************************************************
3303 * Function : client_send_cookie
3305 * Description : Crunches the "cookie" header if necessary.
3306 * Called from `sed'.
3308 * XXX: Stupid name, doesn't send squat.
3311 * 1 : csp = Current client state (buffers, headers, etc...)
3312 * 2 : header = On input, pointer to header to modify.
3313 * On output, pointer to the modified header, or NULL
3314 * to remove the header. This function frees the
3315 * original string if necessary.
3317 * Returns : JB_ERR_OK on success, or
3318 * JB_ERR_MEMORY on out-of-memory error.
3320 *********************************************************************/
3321 static jb_err client_send_cookie(struct client_state *csp, char **header)
3323 if (csp->action->flags & ACTION_NO_COOKIE_READ)
3325 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3333 /*********************************************************************
3335 * Function : client_x_forwarded
3337 * Description : Handle the "x-forwarded-for" config setting properly,
3338 * also used in the add_client_headers list. Called from `sed'.
3341 * 1 : csp = Current client state (buffers, headers, etc...)
3342 * 2 : header = On input, pointer to header to modify.
3343 * On output, pointer to the modified header, or NULL
3344 * to remove the header. This function frees the
3345 * original string if necessary.
3347 * Returns : JB_ERR_OK on success, or
3348 * JB_ERR_MEMORY on out-of-memory error.
3350 *********************************************************************/
3351 jb_err client_x_forwarded(struct client_state *csp, char **header)
3353 if ((csp->action->flags & ACTION_HIDE_FORWARDED) != 0)
3356 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3363 /*********************************************************************
3365 * Function : client_max_forwards
3367 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3368 * from the value of the Max-Forwards header field.
3371 * 1 : csp = Current client state (buffers, headers, etc...)
3372 * 2 : header = On input, pointer to header to modify.
3373 * On output, pointer to the modified header, or NULL
3374 * to remove the header. This function frees the
3375 * original string if necessary.
3377 * Returns : JB_ERR_OK on success, or
3378 * JB_ERR_MEMORY on out-of-memory error.
3380 *********************************************************************/
3381 static jb_err client_max_forwards(struct client_state *csp, char **header)
3385 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3386 (0 == strcmpic(csp->http->gpc, "options")))
3388 assert(*(*header+12) == ':');
3389 if (1 == sscanf(*header+12, ": %u", &max_forwards))
3391 if (max_forwards > 0)
3393 snprintf(*header, strlen(*header)+1, "Max-Forwards: %u", --max_forwards);
3394 log_error(LOG_LEVEL_HEADER, "Max-Forwards value for %s request reduced to %u.",
3395 csp->http->gpc, max_forwards);
3397 else if (max_forwards < 0)
3399 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3405 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3414 /*********************************************************************
3416 * Function : client_host
3418 * Description : If the request URI did not contain host and
3419 * port information, parse and evaluate the Host
3422 * Also, kill ill-formed HOST: headers as sent by
3423 * Apple's iTunes software when used with a proxy.
3426 * 1 : csp = Current client state (buffers, headers, etc...)
3427 * 2 : header = On input, pointer to header to modify.
3428 * On output, pointer to the modified header, or NULL
3429 * to remove the header. This function frees the
3430 * original string if necessary.
3432 * Returns : JB_ERR_OK on success, or
3433 * JB_ERR_MEMORY on out-of-memory error.
3435 *********************************************************************/
3436 static jb_err client_host(struct client_state *csp, char **header)
3441 * If the header field name is all upper-case, chances are that it's
3442 * an ill-formed one from iTunes. BTW, killing innocent headers here is
3443 * not a problem -- they are regenerated later.
3445 if ((*header)[1] == 'O')
3447 log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
3452 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3453 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3456 if (NULL == (p = strdup((*header)+6)))
3458 return JB_ERR_MEMORY;
3461 if (NULL == (q = strdup(p)))
3464 return JB_ERR_MEMORY;
3467 freez(csp->http->hostport);
3468 csp->http->hostport = p;
3469 freez(csp->http->host);
3470 csp->http->host = q;
3471 q = strchr(csp->http->host, ':');
3474 /* Terminate hostname and evaluate port string */
3476 csp->http->port = atoi(q);
3480 csp->http->port = csp->http->ssl ? 443 : 80;
3483 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3484 csp->http->hostport, csp->http->host, csp->http->port);
3487 /* Signal client_host_adder() to return right away */
3488 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3494 /*********************************************************************
3496 * Function : client_if_modified_since
3498 * Description : Remove or modify the If-Modified-Since header.
3501 * 1 : csp = Current client state (buffers, headers, etc...)
3502 * 2 : header = On input, pointer to header to modify.
3503 * On output, pointer to the modified header, or NULL
3504 * to remove the header. This function frees the
3505 * original string if necessary.
3507 * Returns : JB_ERR_OK on success, or
3508 * JB_ERR_MEMORY on out-of-memory error.
3510 *********************************************************************/
3511 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3514 #ifdef HAVE_GMTIME_R
3517 struct tm *timeptr = NULL;
3521 long int hours, minutes, seconds;
3525 if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3528 * The client got an error message because of a temporary problem,
3529 * the problem is gone and the client now tries to revalidate our
3530 * error message on the real server. The revalidation would always
3531 * end with the transmission of the whole document and there is
3532 * no need to expose the bogus If-Modified-Since header.
3534 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3537 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3539 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3541 if ((0 == strcmpic(newval, "block")))
3543 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3546 else /* add random value */
3548 const char *header_time = *header + sizeof("If-Modified-Since:");
3550 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3552 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3557 rtime = strtol(newval, &endptr, 0);
3560 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3561 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3568 rtime = pick_from_range(rtime);
3572 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3575 tm += rtime * (negative ? -1 : 1);
3576 #ifdef HAVE_GMTIME_R
3577 timeptr = gmtime_r(&tm, &gmt);
3578 #elif FEATURE_PTHREAD
3579 pthread_mutex_lock(&gmtime_mutex);
3580 timeptr = gmtime(&tm);
3581 pthread_mutex_unlock(&gmtime_mutex);
3583 timeptr = gmtime(&tm);
3585 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
3588 *header = strdup("If-Modified-Since: ");
3589 string_append(header, newheader);
3591 if (*header == NULL)
3593 log_error(LOG_LEVEL_HEADER, "Insufficent memory, header crunched without replacement.");
3594 return JB_ERR_MEMORY;
3597 if (LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */
3599 hours = rtime / 3600;
3600 minutes = rtime / 60 % 60;
3601 seconds = rtime % 60;
3603 log_error(LOG_LEVEL_HEADER, "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3604 *header, (negative) ? "subtracted" : "added", hours, (hours == 1) ? "r" : "rs",
3605 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
3615 /*********************************************************************
3617 * Function : client_if_none_match
3619 * Description : Remove the If-None-Match header.
3622 * 1 : csp = Current client state (buffers, headers, etc...)
3623 * 2 : header = On input, pointer to header to modify.
3624 * On output, pointer to the modified header, or NULL
3625 * to remove the header. This function frees the
3626 * original string if necessary.
3628 * Returns : JB_ERR_OK on success, or
3629 * JB_ERR_MEMORY on out-of-memory error.
3631 *********************************************************************/
3632 static jb_err client_if_none_match(struct client_state *csp, char **header)
3634 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3636 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3644 /*********************************************************************
3646 * Function : client_x_filter
3648 * Description : Disables filtering if the client set "X-Filter: No".
3649 * Called from `sed'.
3652 * 1 : csp = Current client state (buffers, headers, etc...)
3653 * 2 : header = On input, pointer to header to modify.
3654 * On output, pointer to the modified header, or NULL
3655 * to remove the header. This function frees the
3656 * original string if necessary.
3658 * Returns : JB_ERR_OK on success
3660 *********************************************************************/
3661 jb_err client_x_filter(struct client_state *csp, char **header)
3663 if ( 0 == strcmpic(*header, "X-Filter: No"))
3665 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3667 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3671 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3673 log_error(LOG_LEVEL_HEADER,
3674 "force-text-mode overruled the client's request to fetch without filtering!");
3678 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3679 csp->flags |= CSP_FLAG_NO_FILTERING;
3680 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3682 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3690 /*********************************************************************
3692 * Function : client_range
3694 * Description : Removes Range, Request-Range and If-Range headers if
3695 * content filtering is enabled. If the client's version
3696 * of the document has been altered by Privoxy, the server
3697 * could interpret the range differently than the client
3698 * intended in which case the user could end up with
3699 * corrupted content.
3702 * 1 : csp = Current client state (buffers, headers, etc...)
3703 * 2 : header = On input, pointer to header to modify.
3704 * On output, pointer to the modified header, or NULL
3705 * to remove the header. This function frees the
3706 * original string if necessary.
3708 * Returns : JB_ERR_OK
3710 *********************************************************************/
3711 static jb_err client_range(struct client_state *csp, char **header)
3713 if (content_filters_enabled(csp->action))
3715 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3716 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3723 /* the following functions add headers directly to the header list */
3725 /*********************************************************************
3727 * Function : client_host_adder
3729 * Description : Adds the Host: header field if it is missing.
3730 * Called from `sed'.
3733 * 1 : csp = Current client state (buffers, headers, etc...)
3735 * Returns : JB_ERR_OK on success, or
3736 * JB_ERR_MEMORY on out-of-memory error.
3738 *********************************************************************/
3739 static jb_err client_host_adder(struct client_state *csp)
3744 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3746 /* Header already set by the client, nothing to do. */
3750 if ( !csp->http->hostport || !*(csp->http->hostport))
3752 /* XXX: When does this happen and why is it OK? */
3753 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3758 * remove 'user:pass@' from 'proto://user:pass@host'
3760 if ( (p = strchr( csp->http->hostport, '@')) != NULL )
3766 p = csp->http->hostport;
3769 /* XXX: Just add it, we already made sure that it will be unique */
3770 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3771 err = enlist_unique_header(csp->headers, "Host", p);
3778 /*********************************************************************
3780 * Function : client_accept_encoding_adder
3782 * Description : Add an Accept-Encoding header to the client's request
3783 * that disables compression if the action applies, and
3784 * the header is not already there. Called from `sed'.
3785 * Note: For HTTP/1.0, the absence of the header is enough.
3788 * 1 : csp = Current client state (buffers, headers, etc...)
3790 * Returns : JB_ERR_OK on success, or
3791 * JB_ERR_MEMORY on out-of-memory error.
3793 *********************************************************************/
3794 static jb_err client_accept_encoding_adder(struct client_state *csp)
3796 if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3797 && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
3799 return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
3807 /*********************************************************************
3809 * Function : client_xtra_adder
3811 * Description : Used in the add_client_headers list. Called from `sed'.
3814 * 1 : csp = Current client state (buffers, headers, etc...)
3816 * Returns : JB_ERR_OK on success, or
3817 * JB_ERR_MEMORY on out-of-memory error.
3819 *********************************************************************/
3820 static jb_err client_xtra_adder(struct client_state *csp)
3822 struct list_entry *lst;
3825 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3826 lst ; lst = lst->next)
3828 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3829 err = enlist(csp->headers, lst->str);
3841 /*********************************************************************
3843 * Function : connection_close_adder
3845 * Description : "Temporary" fix for the needed but missing HTTP/1.1
3846 * support. Adds a "Connection: close" header to csp->headers
3847 * unless the header was already present. Called from `sed'.
3849 * FIXME: This whole function shouldn't be neccessary!
3852 * 1 : csp = Current client state (buffers, headers, etc...)
3854 * Returns : JB_ERR_OK on success, or
3855 * JB_ERR_MEMORY on out-of-memory error.
3857 *********************************************************************/
3858 static jb_err connection_close_adder(struct client_state *csp)
3860 const unsigned int flags = csp->flags;
3863 * Return right away if
3865 * - we're parsing server headers and the server header
3866 * "Connection: close" is already set, or if
3868 * - we're parsing client headers and the client header
3869 * "Connection: close" is already set.
3871 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE
3872 && flags & CSP_FLAG_SERVER_CONNECTION_CLOSE_SET)
3873 ||(!(flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3874 && flags & CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET))
3879 log_error(LOG_LEVEL_HEADER, "Adding: Connection: close");
3881 return enlist(csp->headers, "Connection: close");
3885 /*********************************************************************
3887 * Function : server_http
3889 * Description : - Save the HTTP Status into csp->http->status
3890 * - Set CT_TABOO to prevent filtering if the answer
3891 * is a partial range (HTTP status 206)
3892 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3896 * 1 : csp = Current client state (buffers, headers, etc...)
3897 * 2 : header = On input, pointer to header to modify.
3898 * On output, pointer to the modified header, or NULL
3899 * to remove the header. This function frees the
3900 * original string if necessary.
3902 * Returns : JB_ERR_OK on success, or
3903 * JB_ERR_MEMORY on out-of-memory error.
3905 *********************************************************************/
3906 static jb_err server_http(struct client_state *csp, char **header)
3908 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3909 if (csp->http->status == 206)
3911 csp->content_type = CT_TABOO;
3914 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3916 /* XXX: Should we do a real validity check here? */
3917 if (strlen(*header) > 8)
3920 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3925 * XXX: Should we block the request or
3926 * enlist a valid status code line here?
3928 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3929 "Downgrading to HTTP/1.0 impossible.");
3937 /*********************************************************************
3939 * Function : server_set_cookie
3941 * Description : Handle the server "cookie" header properly.
3942 * Log cookie to the jar file. Then "crunch",
3943 * accept or rewrite it to a session cookie.
3944 * Called from `sed'.
3946 * TODO: Allow the user to specify a new expiration
3947 * time to cause the cookie to expire even before the
3948 * browser is closed.
3951 * 1 : csp = Current client state (buffers, headers, etc...)
3952 * 2 : header = On input, pointer to header to modify.
3953 * On output, pointer to the modified header, or NULL
3954 * to remove the header. This function frees the
3955 * original string if necessary.
3957 * Returns : JB_ERR_OK on success, or
3958 * JB_ERR_MEMORY on out-of-memory error.
3960 *********************************************************************/
3961 static jb_err server_set_cookie(struct client_state *csp, char **header)
3968 #ifdef FEATURE_COOKIE_JAR
3969 if (csp->config->jar)
3972 * Write timestamp into outbuf.
3974 * Complex because not all OSs have tm_gmtoff or
3975 * the %z field in strftime()
3977 char tempbuf[ BUFFER_SIZE ];
3979 #ifdef HAVE_LOCALTIME_R
3980 tm_now = *localtime_r(&now, &tm_now);
3981 #elif FEATURE_PTHREAD
3982 pthread_mutex_lock(&localtime_mutex);
3983 tm_now = *localtime (&now);
3984 pthread_mutex_unlock(&localtime_mutex);
3986 tm_now = *localtime (&now);
3988 strftime(tempbuf, BUFFER_SIZE-6, "%b %d %H:%M:%S ", &tm_now);
3990 /* strlen("set-cookie: ") = 12 */
3991 fprintf(csp->config->jar, "%s %s\t%s\n", tempbuf, csp->http->host, *header + 12);
3993 #endif /* def FEATURE_COOKIE_JAR */
3995 if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
3997 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
4000 else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
4002 /* Flag whether or not to log a message */
4005 /* A variable to store the tag we're working on */
4008 /* Skip "Set-Cookie:" (11 characters) in header */
4009 cur_tag = *header + 11;
4011 /* skip whitespace between "Set-Cookie:" and value */
4012 while (*cur_tag && ijb_isspace(*cur_tag))
4017 /* Loop through each tag in the cookie */
4021 char *next_tag = strchr(cur_tag, ';');
4022 if (next_tag != NULL)
4024 /* Skip the ';' character itself */
4027 /* skip whitespace ";" and start of tag */
4028 while (*next_tag && ijb_isspace(*next_tag))
4035 /* "Next tag" is the end of the string */
4036 next_tag = cur_tag + strlen(cur_tag);
4040 * Check the expiration date to see
4041 * if the cookie is still valid, if yes,
4042 * rewrite it to a session cookie.
4044 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
4046 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
4048 /* Did we detect the date properly? */
4049 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
4052 * Nope, treat it as if it was still valid.
4054 * XXX: Should we remove the whole cookie instead?
4056 log_error(LOG_LEVEL_ERROR,
4057 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
4058 string_move(cur_tag, next_tag);
4064 * Yes. Check if the cookie is still valid.
4066 * If the cookie is already expired it's probably
4067 * a delete cookie and even if it isn't, the browser
4068 * will discard it anyway.
4072 * XXX: timegm() isn't available on some AmigaOS
4073 * versions and our replacement doesn't work.
4075 * Our options are to either:
4077 * - disable session-cookies-only completely if timegm
4080 * - to simply remove all expired tags, like it has
4081 * been done until Privoxy 3.0.6 and to live with
4082 * the consequence that it can cause login/logout
4083 * problems on servers that don't validate their
4084 * input properly, or
4086 * - to replace it with mktime in which
4087 * case there is a slight chance of valid cookies
4088 * passing as already expired.
4090 * This is the way it's currently done and it's not
4091 * as bad as it sounds. If the missing GMT offset is
4092 * enough to change the result of the expiration check
4093 * the cookie will be only valid for a few hours
4094 * anyway, which in many cases will be shorter
4095 * than a browser session.
4097 if (cookie_time - now < 0)
4099 log_error(LOG_LEVEL_HEADER,
4100 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4101 /* Just in case some clown sets more then one expiration date */
4107 * Still valid, delete expiration date by copying
4108 * the rest of the string over it.
4110 string_move(cur_tag, next_tag);
4112 /* That changed the header, need to issue a log message */
4116 * Note that the next tag has now been moved to *cur_tag,
4117 * so we do not need to update the cur_tag pointer.
4125 /* Move on to next cookie tag */
4132 assert(NULL != *header);
4133 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
4142 #ifdef FEATURE_FORCE_LOAD
4143 /*********************************************************************
4145 * Function : strclean
4147 * Description : In-Situ-Eliminate all occurances of substring in
4151 * 1 : string = string to clean
4152 * 2 : substring = substring to eliminate
4154 * Returns : Number of eliminations
4156 *********************************************************************/
4157 int strclean(char *string, const char *substring)
4163 len = strlen(substring);
4165 while((pos = strstr(string, substring)) != NULL)
4172 while (*p++ != '\0');
4179 #endif /* def FEATURE_FORCE_LOAD */
4182 /*********************************************************************
4184 * Function : parse_header_time
4186 * Description : Parses time formats used in HTTP header strings
4187 * to get the numerical respresentation.
4190 * 1 : header_time = HTTP header time as string.
4191 * 2 : result = storage for header_time in seconds
4193 * Returns : JB_ERR_OK if the time format was recognized, or
4194 * JB_ERR_PARSE otherwise.
4196 *********************************************************************/
4197 static jb_err parse_header_time(const char *header_time, time_t *result)
4202 * Zero out gmt to prevent time zone offsets.
4204 * While this is only necessary on some platforms
4205 * (mingw32 for example), I don't know how to
4206 * detect these automatically and doing it everywhere
4209 memset(&gmt, 0, sizeof(gmt));
4211 /* Tue, 02 Jun 2037 20:00:00 */
4212 if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
4213 /* Tue, 02-Jun-2037 20:00:00 */
4214 && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
4215 /* Tue, 02-Jun-37 20:00:00 */
4216 && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
4217 /* Tuesday, 02-Jun-2037 20:00:00 */
4218 && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
4219 /* Tuesday Jun 02 20:00:00 2037 */
4220 && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
4222 return JB_ERR_PARSE;
4225 *result = timegm(&gmt);
4232 /*********************************************************************
4234 * Function : get_destination_from_headers
4236 * Description : Parse the "Host:" header to get the request's destination.
4237 * Only needed if the client's request was forcefully
4238 * redirected into Privoxy.
4240 * Code mainly copied from client_host() which is currently
4241 * run too late for this purpose.
4244 * 1 : headers = List of headers (one of them hopefully being
4245 * the "Host:" header)
4246 * 2 : http = storage for the result (host, port and hostport).
4248 * Returns : JB_ERR_MEMORY in case of memory problems,
4249 * JB_ERR_PARSE if the host header couldn't be found,
4250 * JB_ERR_OK otherwise.
4252 *********************************************************************/
4253 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4259 host = get_header_value(headers, "Host:");
4263 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4264 return JB_ERR_PARSE;
4267 if (NULL == (p = strdup((host))))
4269 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4270 return JB_ERR_MEMORY;
4273 if (NULL == (q = strdup(p)))
4276 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4277 return JB_ERR_MEMORY;
4280 freez(http->hostport);
4284 q = strchr(http->host, ':');
4287 /* Terminate hostname and evaluate port string */
4289 http->port = atoi(q);
4293 http->port = http->ssl ? 443 : 80;
4296 /* Rebuild request URL */
4298 http->url = strdup(http->ssl ? "https://" : "http://");
4299 string_append(&http->url, http->hostport);
4300 string_append(&http->url, http->path);
4301 if (http->url == NULL)
4303 return JB_ERR_MEMORY;
4306 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4314 /*********************************************************************
4316 * Function : create_forged_referrer
4318 * Description : Helper for client_referrer to forge a referer as
4319 * 'http://[hostname:port/' to fool stupid
4320 * checks for in-site links
4323 * 1 : header = Pointer to header pointer
4324 * 2 : hostport = Host and optionally port as string
4326 * Returns : JB_ERR_OK in case of success, or
4327 * JB_ERR_MEMORY in case of memory problems.
4329 *********************************************************************/
4330 static jb_err create_forged_referrer(char **header, const char *hostport)
4332 assert(NULL == *header);
4334 *header = strdup("Referer: http://");
4335 string_append(header, hostport);
4336 string_append(header, "/");
4338 if (NULL == *header)
4340 return JB_ERR_MEMORY;
4343 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4350 /*********************************************************************
4352 * Function : create_fake_referrer
4354 * Description : Helper for client_referrer to create a fake referrer
4355 * based on a string supplied by the user.
4358 * 1 : header = Pointer to header pointer
4359 * 2 : hosthost = Referrer to fake
4361 * Returns : JB_ERR_OK in case of success, or
4362 * JB_ERR_MEMORY in case of memory problems.
4364 *********************************************************************/
4365 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4367 assert(NULL == *header);
4369 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4371 log_error(LOG_LEVEL_HEADER,
4372 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4374 *header = strdup("Referer: ");
4375 string_append(header, fake_referrer);
4377 if (NULL == *header)
4379 return JB_ERR_MEMORY;
4382 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4389 /*********************************************************************
4391 * Function : handle_conditional_hide_referrer_parameter
4393 * Description : Helper for client_referrer to crunch or forge
4394 * the referrer header if the host has changed.
4397 * 1 : header = Pointer to header pointer
4398 * 2 : host = The target host (may include the port)
4399 * 3 : parameter_conditional_block = Boolean to signal
4400 * if we're in conditional-block mode. If not set,
4401 * we're in conditional-forge mode.
4403 * Returns : JB_ERR_OK in case of success, or
4404 * JB_ERR_MEMORY in case of memory problems.
4406 *********************************************************************/
4407 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4408 const char *host, const int parameter_conditional_block)
4410 char *referer = strdup(*header);
4411 const size_t hostlenght = strlen(host);
4413 if (NULL == referer)
4416 return JB_ERR_MEMORY;
4419 /* referer begins with 'Referer: http[s]://' */
4420 if (hostlenght < (strlen(referer)-17))
4423 * Shorten referer to make sure the referer is blocked
4424 * if www.example.org/www.example.com-shall-see-the-referer/
4425 * links to www.example.com/
4427 referer[hostlenght+17] = '\0';
4429 if (NULL == strstr(referer, host))
4431 /* Host has changed */
4432 if (parameter_conditional_block)
4434 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4441 return create_forged_referrer(header, host);