1 const char parsers_rcs[] = "$Id: parsers.c,v 1.142 2008/09/20 10:04:33 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 1.142 2008/09/20 10:04:33 fabiankeil
48 * Remove hide-forwarded-for-headers action which has
49 * been obsoleted by change-x-forwarded-for{block}.
51 * Revision 1.141 2008/09/19 15:26:28 fabiankeil
52 * Add change-x-forwarded-for{} action to block or add
53 * X-Forwarded-For headers. Mostly based on code removed
56 * Revision 1.140 2008/09/12 17:51:43 fabiankeil
57 * - A few style fixes.
58 * - Remove a pointless cast.
60 * Revision 1.139 2008/09/04 08:13:58 fabiankeil
61 * Prepare for critical sections on Windows by adding a
62 * layer of indirection before the pthread mutex functions.
64 * Revision 1.138 2008/08/30 12:03:07 fabiankeil
65 * Remove FEATURE_COOKIE_JAR.
67 * Revision 1.137 2008/05/30 15:50:08 fabiankeil
68 * Remove questionable micro-optimizations
69 * whose usefulness has never been measured.
71 * Revision 1.136 2008/05/26 16:02:24 fabiankeil
72 * s@Insufficent@Insufficient@
74 * Revision 1.135 2008/05/21 20:12:10 fabiankeil
75 * The whole point of strclean() is to modify the
76 * first parameter, so don't mark it immutable,
77 * even though the compiler lets us get away with it.
79 * Revision 1.134 2008/05/21 19:27:25 fabiankeil
80 * As the wafer actions are gone, we can stop including encode.h.
82 * Revision 1.133 2008/05/21 15:50:47 fabiankeil
83 * Ditch cast from (char **) to (char **).
85 * Revision 1.132 2008/05/21 15:47:14 fabiankeil
86 * Streamline sed()'s prototype and declare
87 * the header parse and add structures static.
89 * Revision 1.131 2008/05/20 20:13:30 fabiankeil
90 * Factor update_server_headers() out of sed(), ditch the
91 * first_run hack and make server_patterns_light static.
93 * Revision 1.130 2008/05/19 17:18:04 fabiankeil
94 * Wrap memmove() calls in string_move()
95 * to document the purpose in one place.
97 * Revision 1.129 2008/05/17 14:02:07 fabiankeil
98 * Normalize linear header white space.
100 * Revision 1.128 2008/05/16 16:39:03 fabiankeil
101 * If a header is split across multiple lines,
102 * merge them to a single line before parsing them.
104 * Revision 1.127 2008/05/10 13:23:38 fabiankeil
105 * Don't provide get_header() with the whole client state
106 * structure when it only needs access to csp->iob.
108 * Revision 1.126 2008/05/03 16:40:45 fabiankeil
109 * Change content_filters_enabled()'s parameter from
110 * csp->action to action so it can be also used in the
111 * CGI code. Don't bother checking if there are filters
112 * loaded, as that's somewhat besides the point.
114 * Revision 1.125 2008/04/17 14:40:49 fabiankeil
115 * Provide get_http_time() with the buffer size so it doesn't
116 * have to blindly assume that the buffer is big enough.
118 * Revision 1.124 2008/04/16 16:38:21 fabiankeil
119 * Don't pass the whole csp structure to flush_socket()
120 * when it only needs a file descriptor and a buffer.
122 * Revision 1.123 2008/03/29 12:13:46 fabiankeil
123 * Remove send-wafer and send-vanilla-wafer actions.
125 * Revision 1.122 2008/03/28 15:13:39 fabiankeil
126 * Remove inspect-jpegs action.
128 * Revision 1.121 2008/01/05 21:37:03 fabiankeil
129 * Let client_range() also handle Request-Range headers
130 * which apparently are still supported by many servers.
132 * Revision 1.120 2008/01/04 17:43:45 fabiankeil
133 * Improve the warning messages that get logged if the action files
134 * "enable" filters but no filters of that type have been loaded.
136 * Revision 1.119 2007/12/28 18:32:51 fabiankeil
137 * In server_content_type():
138 * - Don't require leading white space when detecting image content types.
139 * - Change '... not replaced ...' message to sound less crazy if the text
140 * type actually is 'text/plain'.
141 * - Mark the 'text/plain == binary data' assumption for removal.
142 * - Remove a bunch of trailing white space.
144 * Revision 1.118 2007/12/28 16:56:35 fabiankeil
145 * Minor server_content_disposition() changes:
146 * - Don't regenerate the header name all lower-case.
147 * - Some white space fixes.
148 * - Remove useless log message in case of ENOMEM.
150 * Revision 1.117 2007/12/06 18:11:50 fabiankeil
151 * Garbage-collect the code to add a X-Forwarded-For
152 * header as it seems to be mostly used by accident.
154 * Revision 1.116 2007/12/01 13:04:22 fabiankeil
155 * Fix a crash on mingw32 with some Last Modified times in the future.
157 * Revision 1.115 2007/11/02 16:52:50 fabiankeil
158 * Remove a "can't happen" error block which, over
159 * time, mutated into a "guaranteed to happen" block.
161 * Revision 1.114 2007/10/19 16:56:26 fabiankeil
162 * - Downgrade "Buffer limit reached" message to LOG_LEVEL_INFO.
163 * - Use shiny new content_filters_enabled() in client_range().
165 * Revision 1.113 2007/10/10 17:29:57 fabiankeil
166 * I forgot about Poland.
168 * Revision 1.112 2007/10/09 16:38:40 fabiankeil
169 * Remove Range and If-Range headers if content filtering is enabled.
171 * Revision 1.111 2007/10/04 18:07:00 fabiankeil
172 * Move ACTION_VANILLA_WAFER handling from jcc's chat() into
173 * client_cookie_adder() to make sure send-vanilla-wafer can be
174 * controlled through tags (and thus regression-tested).
176 * Revision 1.110 2007/09/29 10:42:37 fabiankeil
177 * - Remove "scanning headers for" log message again.
178 * - Some more whitespace fixes.
180 * Revision 1.109 2007/09/08 14:25:48 fabiankeil
181 * Refactor client_referrer() and add conditional-forge parameter.
183 * Revision 1.108 2007/08/28 18:21:03 fabiankeil
184 * A bunch of whitespace fixes, pointy hat to me.
186 * Revision 1.107 2007/08/28 18:16:32 fabiankeil
187 * Fix possible memory corruption in server_http, make sure it's not
188 * executed for ordinary server headers and mark some problems for later.
190 * Revision 1.106 2007/08/18 14:30:32 fabiankeil
191 * Let content-type-overwrite{} honour force-text-mode again.
193 * Revision 1.105 2007/08/11 14:49:49 fabiankeil
194 * - Add prototpyes for the header parsers and make them static.
195 * - Comment out client_accept_encoding_adder() which isn't used right now.
197 * Revision 1.104 2007/07/14 07:38:19 fabiankeil
198 * Move the ACTION_FORCE_TEXT_MODE check out of
199 * server_content_type(). Signal other functions
200 * whether or not a content type has been declared.
201 * Part of the fix for BR#1750917.
203 * Revision 1.103 2007/06/01 16:31:54 fabiankeil
204 * Change sed() to return a jb_err in preparation for forward-override{}.
206 * Revision 1.102 2007/05/27 12:39:32 fabiankeil
207 * Adjust "X-Filter: No" to disable dedicated header filters.
209 * Revision 1.101 2007/05/14 10:16:41 fabiankeil
210 * Streamline client_cookie_adder().
212 * Revision 1.100 2007/04/30 15:53:11 fabiankeil
213 * Make sure filters with dynamic jobs actually use them.
215 * Revision 1.99 2007/04/30 15:06:26 fabiankeil
216 * - Introduce dynamic pcrs jobs that can resolve variables.
217 * - Remove unnecessary update_action_bits_for_all_tags() call.
219 * Revision 1.98 2007/04/17 18:32:10 fabiankeil
220 * - Make tagging based on tags set by earlier taggers
221 * of the same kind possible.
222 * - Log whether or not new tags cause action bits updates
223 * (in which case a matching tag-pattern section exists).
224 * - Log if the user tries to set a tag that is already set.
226 * Revision 1.97 2007/04/15 16:39:21 fabiankeil
227 * Introduce tags as alternative way to specify which
228 * actions apply to a request. At the moment tags can be
229 * created based on client and server headers.
231 * Revision 1.96 2007/04/12 12:53:58 fabiankeil
232 * Log a warning if the content is compressed, filtering is
233 * enabled and Privoxy was compiled without zlib support.
236 * Revision 1.95 2007/03/25 14:26:40 fabiankeil
237 * - Fix warnings when compiled with glibc.
238 * - Don't use crumble() for cookie crunching.
239 * - Move cookie time parsing into parse_header_time().
240 * - Let parse_header_time() return a jb_err code
241 * instead of a pointer that can only be used to
242 * check for NULL anyway.
244 * Revision 1.94 2007/03/21 12:23:53 fabiankeil
245 * - Add better protection against malicious gzip headers.
246 * - Stop logging the first hundred bytes of decompressed content.
247 * It looks like it's working and there is always debug 16.
248 * - Log the content size after decompression in decompress_iob()
249 * instead of pcrs_filter_response().
251 * Revision 1.93 2007/03/20 15:21:44 fabiankeil
252 * - Use dedicated header filter actions instead of abusing "filter".
253 * Replace "filter-client-headers" and "filter-client-headers"
254 * with "server-header-filter" and "client-header-filter".
255 * - Remove filter_client_header() and filter_client_header(),
256 * filter_header() now checks the shiny new
257 * CSP_FLAG_CLIENT_HEADER_PARSING_DONE flag instead.
259 * Revision 1.92 2007/03/05 13:25:32 fabiankeil
260 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
261 * - Handle "Cookie:" and "Connection:" headers a bit smarter
262 * (don't crunch them just to recreate them later on).
263 * - Add another non-standard time format for the cookie
264 * expiration date detection.
265 * - Fix a valgrind warning.
267 * Revision 1.91 2007/02/24 12:27:32 fabiankeil
268 * Improve cookie expiration date detection.
270 * Revision 1.90 2007/02/08 19:12:35 fabiankeil
271 * Don't run server_content_length() the first time
272 * sed() parses server headers; only adjust the
273 * Content-Length header if the page was modified.
275 * Revision 1.89 2007/02/07 16:52:11 fabiankeil
276 * Fix log messages regarding the cookie time format
277 * (cookie and request URL were mixed up).
279 * Revision 1.88 2007/02/07 11:27:12 fabiankeil
280 * - Let decompress_iob()
281 * - not corrupt the content if decompression fails
282 * early. (the first byte(s) were lost).
283 * - use pointer arithmetics with defined outcome for
285 * - Use a different kludge to remember a failed decompression.
287 * Revision 1.87 2007/01/31 16:21:38 fabiankeil
288 * Search for Max-Forwards headers case-insensitive,
289 * don't generate the "501 unsupported" message for invalid
290 * Max-Forwards values and don't increase negative ones.
292 * Revision 1.86 2007/01/30 13:05:26 fabiankeil
293 * - Let server_set_cookie() check the expiration date
294 * of cookies and don't touch the ones that are already
295 * expired. Fixes problems with low quality web applications
296 * as described in BR 932612.
298 * - Adjust comment in client_max_forwards to reality;
299 * remove invalid Max-Forwards headers.
301 * Revision 1.85 2007/01/26 15:33:46 fabiankeil
302 * Stop filter_header() from unintentionally removing
303 * empty header lines that were enlisted by the continue
306 * Revision 1.84 2007/01/24 12:56:52 fabiankeil
307 * - Repeat the request URL before logging any headers.
308 * Makes reading the log easier in case of simultaneous requests.
309 * - If there are more than one Content-Type headers in one request,
310 * use the first one and remove the others.
311 * - Remove "newval" variable in server_content_type().
312 * It's only used once.
314 * Revision 1.83 2007/01/12 15:03:02 fabiankeil
315 * Correct a cast, check inflateEnd() exit code
316 * to see if we have to, replace sprintf calls
319 * Revision 1.82 2007/01/01 19:36:37 fabiankeil
320 * Integrate a modified version of Wil Mahan's
321 * zlib patch (PR #895531).
323 * Revision 1.81 2006/12/31 22:21:33 fabiankeil
324 * Skip empty filter files in filter_header()
325 * but don't ignore the ones that come afterwards.
326 * Fixes BR 1619208, this time for real.
328 * Revision 1.80 2006/12/29 19:08:22 fabiankeil
329 * Reverted parts of my last commit
330 * to keep error handling working.
332 * Revision 1.79 2006/12/29 18:04:40 fabiankeil
333 * Fixed gcc43 conversion warnings.
335 * Revision 1.78 2006/12/26 17:19:20 fabiankeil
336 * Bringing back the "useless" localtime() call
337 * I removed in revision 1.67. On some platforms
338 * it's necessary to prevent time zone offsets.
340 * Revision 1.77 2006/12/07 18:44:26 fabiankeil
341 * Rebuild request URL in get_destination_from_headers()
342 * to make sure redirect{pcrs command} works as expected
343 * for intercepted requests.
345 * Revision 1.76 2006/12/06 19:52:25 fabiankeil
346 * Added get_destination_from_headers().
348 * Revision 1.75 2006/11/13 19:05:51 fabiankeil
349 * Make pthread mutex locking more generic. Instead of
350 * checking for OSX and OpenBSD, check for FEATURE_PTHREAD
351 * and use mutex locking unless there is an _r function
352 * available. Better safe than sorry.
354 * Fixes "./configure --disable-pthread" and should result
355 * in less threading-related problems on pthread-using platforms,
356 * but it still doesn't fix BR#1122404.
358 * Revision 1.74 2006/10/02 16:59:12 fabiankeil
359 * The special header "X-Filter: No" now disables
360 * header filtering as well.
362 * Revision 1.73 2006/09/23 13:26:38 roro
363 * Replace TABs by spaces in source code.
365 * Revision 1.72 2006/09/23 12:37:21 fabiankeil
366 * Don't print a log message every time filter_headers is
367 * entered or left. It only creates noise without any real
370 * Revision 1.71 2006/09/21 19:55:17 fabiankeil
371 * Fix +hide-if-modified-since{-n}.
373 * Revision 1.70 2006/09/08 12:06:34 fabiankeil
374 * Have hide-if-modified-since interpret the random
375 * range value as minutes instead of hours. Allows
376 * more fine-grained configuration.
378 * Revision 1.69 2006/09/06 16:25:51 fabiankeil
379 * Always have parse_header_time return a pointer
380 * that actual makes sense, even though we currently
381 * only need it to detect problems.
383 * Revision 1.68 2006/09/06 10:43:32 fabiankeil
384 * Added config option enable-remote-http-toggle
385 * to specify if Privoxy should recognize special
386 * headers (currently only X-Filter) to change its
387 * behaviour. Disabled by default.
389 * Revision 1.67 2006/09/04 11:01:26 fabiankeil
390 * After filtering de-chunked instances, remove
391 * "Transfer-Encoding" header entirely instead of changing
392 * it to "Transfer-Encoding: identity", which is invalid.
393 * Thanks Michael Shields <shields@msrl.com>. Fixes PR 1318658.
395 * Don't use localtime in parse_header_time. An empty time struct
396 * is good enough, it gets overwritten by strptime anyway.
398 * Revision 1.66 2006/09/03 19:38:28 fabiankeil
399 * Use gmtime_r if available, fallback to gmtime with mutex
400 * protection for MacOSX and use vanilla gmtime for the rest.
402 * Revision 1.65 2006/08/22 10:55:56 fabiankeil
403 * Changed client_referrer to use the right type (size_t) for
404 * hostlenght and to shorten the temporary referrer string with
405 * '\0' instead of adding a useless line break.
407 * Revision 1.64 2006/08/17 17:15:10 fabiankeil
408 * - Back to timegm() using GnuPG's replacement if necessary.
409 * Using mktime() and localtime() could add a on hour offset if
410 * the randomize factor was big enough to lead to a summer/wintertime
413 * - Removed now-useless Privoxy 3.0.3 compatibility glue.
415 * - Moved randomization code into pick_from_range().
417 * - Changed parse_header_time definition.
418 * time_t isn't guaranteed to be signed and
419 * if it isn't, -1 isn't available as error code.
420 * Changed some variable types in client_if_modified_since()
421 * because of the same reason.
423 * Revision 1.63 2006/08/14 13:18:08 david__schmidt
424 * OS/2 compilation compatibility fixups
426 * Revision 1.62 2006/08/14 08:58:42 fabiankeil
427 * Changed include from strptime.c to strptime.h
429 * Revision 1.61 2006/08/14 08:25:19 fabiankeil
430 * Split filter-headers{} into filter-client-headers{}
431 * and filter-server-headers{}.
432 * Added parse_header_time() to share some code.
433 * Replaced timegm() with mktime().
435 * Revision 1.60 2006/08/12 03:54:37 david__schmidt
436 * Windows service integration
438 * Revision 1.59 2006/08/03 02:46:41 david__schmidt
439 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
441 * Revision 1.58 2006/07/18 14:48:47 david__schmidt
442 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
443 * with what was really the latest development (the v_3_0_branch branch)
445 * Revision 1.56.2.10 2006/01/21 16:16:08 david__schmidt
446 * Thanks to Edward Carrel for his patch to modernize OSX's
\rpthreads support. See bug #1409623.
448 * Revision 1.56.2.9 2004/10/03 12:53:45 david__schmidt
449 * Add the ability to check jpeg images for invalid
450 * lengths of comment blocks. Defensive strategy
451 * against the exploit:
452 * Microsoft Security Bulletin MS04-028
453 * Buffer Overrun in JPEG Processing (GDI+) Could
454 * Allow Code Execution (833987)
455 * Enabled with +inspect-jpegs in actions files.
457 * Revision 1.56.2.8 2003/07/11 13:21:25 oes
458 * Excluded text/plain objects from filtering. This fixes a
459 * couple of client-crashing, download corruption and
460 * Privoxy performance issues, whose root cause lies in
461 * web servers labelling content of unknown type as text/plain.
463 * Revision 1.56.2.7 2003/05/06 12:07:26 oes
464 * Fixed bug #729900: Suspicious HOST: headers are now killed and regenerated if necessary
466 * Revision 1.56.2.6 2003/04/14 21:28:30 oes
467 * Completing the previous change
469 * Revision 1.56.2.5 2003/04/14 12:08:16 oes
470 * Added temporary workaround for bug in PHP < 4.2.3
472 * Revision 1.56.2.4 2003/03/07 03:41:05 david__schmidt
473 * 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.
475 * Revision 1.56.2.3 2002/11/10 04:20:02 hal9
476 * Fix typo: supressed -> suppressed
478 * Revision 1.56.2.2 2002/09/25 14:59:53 oes
479 * Improved cookie logging
481 * Revision 1.56.2.1 2002/09/25 14:52:45 oes
482 * Added basic support for OPTIONS and TRACE HTTP methods:
483 * - New parser function client_max_forwards which decrements
484 * the Max-Forwards HTTP header field of OPTIONS and TRACE
485 * requests by one before forwarding
486 * - New parser function client_host which extracts the host
487 * and port information from the HTTP header field if the
488 * request URI was not absolute
489 * - Don't crumble and re-add the Host: header, but only generate
490 * and append if missing
492 * Revision 1.56 2002/05/12 15:34:22 jongfoster
493 * Fixing typo in a comment
495 * Revision 1.55 2002/05/08 16:01:07 oes
496 * Optimized add_to_iob:
497 * - Use realloc instead of malloc(), memcpy(), free()
498 * - Expand to powers of two if possible, to get
499 * O(log n) reallocs instead of O(n).
500 * - Moved check for buffer limit here from chat
501 * - Report failure via returncode
503 * Revision 1.54 2002/04/02 15:03:16 oes
504 * Tiny code cosmetics
506 * Revision 1.53 2002/03/26 22:29:55 swa
507 * we have a new homepage!
509 * Revision 1.52 2002/03/24 13:25:43 swa
510 * name change related issues
512 * Revision 1.51 2002/03/13 00:27:05 jongfoster
515 * Revision 1.50 2002/03/12 01:45:35 oes
516 * More verbose logging
518 * Revision 1.49 2002/03/09 20:03:52 jongfoster
519 * - Making various functions return int rather than size_t.
520 * (Undoing a recent change). Since size_t is unsigned on
521 * Windows, functions like read_socket that return -1 on
522 * error cannot return a size_t.
524 * THIS WAS A MAJOR BUG - it caused frequent, unpredictable
525 * crashes, and also frequently caused JB to jump to 100%
526 * CPU and stay there. (Because it thought it had just
527 * read ((unsigned)-1) == 4Gb of data...)
529 * - The signature of write_socket has changed, it now simply
530 * returns success=0/failure=nonzero.
532 * - Trying to get rid of a few warnings --with-debug on
533 * Windows, I've introduced a new type "jb_socket". This is
534 * used for the socket file descriptors. On Windows, this
535 * is SOCKET (a typedef for unsigned). Everywhere else, it's
536 * an int. The error value can't be -1 any more, so it's
537 * now JB_INVALID_SOCKET (which is -1 on UNIX, and in
538 * Windows it maps to the #define INVALID_SOCKET.)
540 * - The signature of bind_port has changed.
542 * Revision 1.48 2002/03/07 03:46:53 oes
543 * Fixed compiler warnings etc
545 * Revision 1.47 2002/02/20 23:15:13 jongfoster
546 * Parsing functions now handle out-of-memory gracefully by returning
549 * Revision 1.46 2002/01/17 21:03:47 jongfoster
550 * Moving all our URL and URL pattern parsing code to urlmatch.c.
552 * Revision 1.45 2002/01/09 14:33:03 oes
553 * Added support for localtime_r.
555 * Revision 1.44 2001/12/14 01:22:54 steudten
556 * Remove 'user:pass@' from 'proto://user:pass@host' for the
557 * new added header 'Host: ..'. (See Req ID 491818)
559 * Revision 1.43 2001/11/23 00:26:38 jongfoster
560 * Fixing two really stupid errors in my previous commit
562 * Revision 1.42 2001/11/22 21:59:30 jongfoster
563 * Adding code to handle +no-cookies-keep
565 * Revision 1.41 2001/11/05 23:43:05 steudten
566 * Add time+date to log files.
568 * Revision 1.40 2001/10/26 20:13:09 jongfoster
569 * ctype.h is needed in Windows, too.
571 * Revision 1.39 2001/10/26 17:40:04 oes
572 * Introduced get_header_value()
573 * Removed http->user_agent, csp->referrer and csp->accept_types
574 * Removed client_accept()
576 * Revision 1.38 2001/10/25 03:40:48 david__schmidt
577 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
578 * threads to call select() simultaneously. So, it's time to do a real, live,
579 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
580 * (native). Both versions will work, but using __OS2__ offers multi-threading.
582 * Revision 1.37 2001/10/23 21:36:02 jongfoster
583 * Documenting sed()'s error behaviou (doc change only)
585 * Revision 1.36 2001/10/13 12:51:51 joergs
586 * Removed client_host, (was only required for the old 2.0.2-11 http://noijb.
587 * force-load), instead crumble Host: and add it (again) in client_host_adder
588 * (in case we get a HTTP/1.0 request without Host: header and forward it to
589 * a HTTP/1.1 server/proxy).
591 * Revision 1.35 2001/10/09 22:39:21 jongfoster
592 * assert.h is also required under Win32, so moving out of #ifndef _WIN32
595 * Revision 1.34 2001/10/07 18:50:55 oes
596 * Added server_content_encoding, renamed server_transfer_encoding
598 * Revision 1.33 2001/10/07 18:04:49 oes
599 * Changed server_http11 to server_http and its pattern to "HTTP".
600 * Additional functionality: it now saves the HTTP status into
601 * csp->http->status and sets CT_TABOO for Status 206 (partial range)
603 * Revision 1.32 2001/10/07 15:43:28 oes
604 * Removed FEATURE_DENY_GZIP and replaced it with client_accept_encoding,
605 * client_te and client_accept_encoding_adder, triggered by the new
606 * +no-compression action. For HTTP/1.1 the Accept-Encoding header is
607 * changed to allow only identity and chunked, and the TE header is
608 * crunched. For HTTP/1.0, Accept-Encoding is crunched.
610 * parse_http_request no longer does anything than parsing. The rewriting
611 * of http->cmd and version mangling are gone. It now also recognizes
612 * the put and delete methods and saves the url in http->url. Removed
615 * renamed content_type and content_length to have the server_ prefix
617 * server_content_type now only works if csp->content_type != CT_TABOO
619 * added server_transfer_encoding, which
620 * - Sets CT_TABOO to prohibit filtering if encoding compresses
621 * - Raises the CSP_FLAG_CHUNKED flag if Encoding is "chunked"
622 * - Change from "chunked" to "identity" if body was chunked
623 * but has been de-chunked for filtering.
625 * added server_content_md5 which crunches any Content-MD5 headers
626 * if the body was modified.
628 * made server_http11 conditional on +downgrade action
630 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
632 * Revision 1.31 2001/10/05 14:25:02 oes
633 * Crumble Keep-Alive from Server
635 * Revision 1.30 2001/09/29 12:56:03 joergs
636 * IJB now changes HTTP/1.1 to HTTP/1.0 in requests and answers.
638 * Revision 1.29 2001/09/24 21:09:24 jongfoster
639 * Fixing 2 memory leaks that Guy spotted, where the paramater to
640 * enlist() was not being free()d.
642 * Revision 1.28 2001/09/22 16:32:28 jongfoster
643 * Removing unused #includes.
645 * Revision 1.27 2001/09/20 15:45:25 steudten
647 * add casting from size_t to int for printf()
648 * remove local variable shadow s2
650 * Revision 1.26 2001/09/16 17:05:14 jongfoster
651 * Removing unused #include showarg.h
653 * Revision 1.25 2001/09/16 13:21:27 jongfoster
654 * Changes to use new list functions.
656 * Revision 1.24 2001/09/13 23:05:50 jongfoster
657 * Changing the string paramater to the header parsers a "const".
659 * Revision 1.23 2001/09/12 18:08:19 steudten
661 * In parse_http_request() header rewriting miss the host value, so
662 * from http://www.mydomain.com the result was just " / " not
663 * http://www.mydomain.com/ in case we forward.
665 * Revision 1.22 2001/09/10 10:58:53 oes
666 * Silenced compiler warnings
668 * Revision 1.21 2001/07/31 14:46:00 oes
669 * - Persistant connections now suppressed
670 * - sed() no longer appends empty header to csp->headers
672 * Revision 1.20 2001/07/30 22:08:36 jongfoster
673 * Tidying up #defines:
674 * - All feature #defines are now of the form FEATURE_xxx
675 * - Permanently turned off WIN_GUI_EDIT
676 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
678 * Revision 1.19 2001/07/25 17:21:54 oes
679 * client_uagent now saves copy of User-Agent: header value
681 * Revision 1.18 2001/07/13 14:02:46 oes
682 * - Included fix to repair broken HTTP requests that
683 * don't contain a path, not even '/'.
684 * - Removed all #ifdef PCRS
685 * - content_type now always inspected and classified as
686 * text, gif or other.
687 * - formatting / comments
689 * Revision 1.17 2001/06/29 21:45:41 oes
690 * Indentation, CRLF->LF, Tab-> Space
692 * Revision 1.16 2001/06/29 13:32:42 oes
694 * - Adapted free_http_request
695 * - Removed logentry from cancelled commit
697 * Revision 1.15 2001/06/03 19:12:38 oes
698 * deleted const struct interceptors
700 * Revision 1.14 2001/06/01 18:49:17 jongfoster
701 * Replaced "list_share" with "list" - the tiny memory gain was not
702 * worth the extra complexity.
704 * Revision 1.13 2001/05/31 21:30:33 jongfoster
705 * Removed list code - it's now in list.[ch]
706 * Renamed "permission" to "action", and changed many features
707 * to use the actions file rather than the global config.
709 * Revision 1.12 2001/05/31 17:33:13 oes
713 * Revision 1.11 2001/05/29 20:11:19 joergs
714 * '/ * inside comment' warning removed.
716 * Revision 1.10 2001/05/29 09:50:24 jongfoster
717 * Unified blocklist/imagelist/permissionslist.
718 * File format is still under discussion, but the internal changes
721 * Also modified interceptor behaviour:
722 * - We now intercept all URLs beginning with one of the following
723 * prefixes (and *only* these prefixes):
725 * * http://ijbswa.sf.net/config/
726 * * http://ijbswa.sourceforge.net/config/
727 * - New interceptors "home page" - go to http://i.j.b/ to see it.
728 * - Internal changes so that intercepted and fast redirect pages
729 * are not replaced with an image.
730 * - Interceptors now have the option to send a binary page direct
731 * to the client. (i.e. ijb-send-banner uses this)
732 * - Implemented show-url-info interceptor. (Which is why I needed
733 * the above interceptors changes - a typical URL is
734 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
735 * The previous mechanism would not have intercepted that, and
736 * if it had been intercepted then it then it would have replaced
739 * Revision 1.9 2001/05/28 17:26:33 jongfoster
740 * Fixing segfault if last header was crunched.
741 * Fixing Windows build (snprintf() is _snprintf() under Win32, but we
742 * can use the cross-platform sprintf() instead.)
744 * Revision 1.8 2001/05/27 22:17:04 oes
746 * - re_process_buffer no longer writes the modified buffer
747 * to the client, which was very ugly. It now returns the
748 * buffer, which it is then written by chat.
750 * - content_length now adjusts the Content-Length: header
751 * for modified documents rather than crunch()ing it.
752 * (Length info in csp->content_length, which is 0 for
753 * unmodified documents)
755 * - For this to work, sed() is called twice when filtering.
757 * Revision 1.7 2001/05/27 13:19:06 oes
758 * Patched Joergs solution for the content-length in.
760 * Revision 1.6 2001/05/26 13:39:32 jongfoster
761 * Only crunches Content-Length header if applying RE filtering.
762 * Without this fix, Microsoft Windows Update wouldn't work.
764 * Revision 1.5 2001/05/26 00:28:36 jongfoster
765 * Automatic reloading of config file.
766 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
767 * Most of the global variables have been moved to a new
768 * struct configuration_spec, accessed through csp->config->globalname
769 * Most of the globals remaining are used by the Win32 GUI.
771 * Revision 1.4 2001/05/22 18:46:04 oes
773 * - Enabled filtering banners by size rather than URL
774 * by adding patterns that replace all standard banner
775 * sizes with the "Junkbuster" gif to the re_filterfile
777 * - Enabled filtering WebBugs by providing a pattern
778 * which kills all 1x1 images
780 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
781 * which is selected by the (nonstandard and therefore
782 * capital) letter 'U' in the option string.
783 * It causes the quantifiers to be ungreedy by default.
784 * Appending a ? turns back to greedy (!).
786 * - Added a new interceptor ijb-send-banner, which
787 * sends back the "Junkbuster" gif. Without imagelist or
788 * MSIE detection support, or if tinygif = 1, or the
789 * URL isn't recognized as an imageurl, a lame HTML
790 * explanation is sent instead.
792 * - Added new feature, which permits blocking remote
793 * script redirects and firing back a local redirect
795 * The feature is conditionally compiled, i.e. it
796 * can be disabled with --disable-fast-redirects,
797 * plus it must be activated by a "fast-redirects"
798 * line in the config file, has its own log level
799 * and of course wants to be displayed by show-proxy-args
800 * Note: Boy, all the #ifdefs in 1001 locations and
801 * all the fumbling with configure.in and acconfig.h
802 * were *way* more work than the feature itself :-(
804 * - Because a generic redirect template was needed for
805 * this, tinygif = 3 now uses the same.
807 * - Moved GIFs, and other static HTTP response templates
812 * - Removed some >400 CRs again (Jon, you really worked
815 * Revision 1.3 2001/05/20 01:21:20 jongfoster
816 * Version 2.9.4 checkin.
817 * - Merged popupfile and cookiefile, and added control over PCRS
818 * filtering, in new "permissionsfile".
819 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
820 * file error you now get a message box (in the Win32 GUI) rather
821 * than the program exiting with no explanation.
822 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
824 * - Removed tabs from "config"
825 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
826 * - Bumped up version number.
828 * Revision 1.2 2001/05/17 23:02:36 oes
829 * - Made referrer option accept 'L' as a substitute for '§'
831 * Revision 1.1.1.1 2001/05/15 13:59:01 oes
832 * Initial import of version 2.9.3 source tree
835 *********************************************************************/
842 #include <sys/types.h>
852 * Convince GNU's libc to provide a strptime prototype.
855 #endif /*__GLIBC__ */
862 #if !defined(_WIN32) && !defined(__OS2__)
868 #ifdef FEATURE_PTHREAD
870 /* jcc.h is for mutex semapores only */
871 #endif /* def FEATURE_PTHREAD */
876 #include "jbsockets.h"
877 #include "miscutil.h"
882 #ifndef HAVE_STRPTIME
883 #include "strptime.h"
886 const char parsers_h_rcs[] = PARSERS_H_VERSION;
888 /* Fix a problem with Solaris. There should be no effect on other
890 * Solaris's isspace() is a macro which uses its argument directly
891 * as an array index. Therefore we need to make sure that high-bit
892 * characters generate +ve values, and ideally we also want to make
893 * the argument match the declared parameter type of "int".
895 * Why did they write a character function that can't take a simple
896 * "char" argument? Doh!
898 #define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
899 #define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
901 static char *get_header_line(struct iob *iob);
902 static jb_err scan_headers(struct client_state *csp);
903 static jb_err header_tagger(struct client_state *csp, char *header);
904 static jb_err parse_header_time(const char *header_time, time_t *result);
906 static jb_err crumble (struct client_state *csp, char **header);
907 static jb_err connection (struct client_state *csp, char **header);
908 static jb_err filter_header (struct client_state *csp, char **header);
909 static jb_err client_referrer (struct client_state *csp, char **header);
910 static jb_err client_uagent (struct client_state *csp, char **header);
911 static jb_err client_ua (struct client_state *csp, char **header);
912 static jb_err client_from (struct client_state *csp, char **header);
913 static jb_err client_send_cookie (struct client_state *csp, char **header);
914 static jb_err client_x_forwarded (struct client_state *csp, char **header);
915 static jb_err client_accept_encoding (struct client_state *csp, char **header);
916 static jb_err client_te (struct client_state *csp, char **header);
917 static jb_err client_max_forwards (struct client_state *csp, char **header);
918 static jb_err client_host (struct client_state *csp, char **header);
919 static jb_err client_if_modified_since (struct client_state *csp, char **header);
920 static jb_err client_accept_language (struct client_state *csp, char **header);
921 static jb_err client_if_none_match (struct client_state *csp, char **header);
922 static jb_err crunch_client_header (struct client_state *csp, char **header);
923 static jb_err client_x_filter (struct client_state *csp, char **header);
924 static jb_err client_range (struct client_state *csp, char **header);
925 static jb_err server_set_cookie (struct client_state *csp, char **header);
926 static jb_err server_content_type (struct client_state *csp, char **header);
927 static jb_err server_content_length (struct client_state *csp, char **header);
928 static jb_err server_content_md5 (struct client_state *csp, char **header);
929 static jb_err server_content_encoding (struct client_state *csp, char **header);
930 static jb_err server_transfer_coding (struct client_state *csp, char **header);
931 static jb_err server_http (struct client_state *csp, char **header);
932 static jb_err crunch_server_header (struct client_state *csp, char **header);
933 static jb_err server_last_modified (struct client_state *csp, char **header);
934 static jb_err server_content_disposition(struct client_state *csp, char **header);
936 static jb_err client_host_adder (struct client_state *csp);
937 static jb_err client_xtra_adder (struct client_state *csp);
938 static jb_err client_x_forwarded_for_adder(struct client_state *csp);
939 static jb_err connection_close_adder (struct client_state *csp);
941 static jb_err create_forged_referrer(char **header, const char *hostport);
942 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
943 static jb_err handle_conditional_hide_referrer_parameter(char **header,
944 const char *host, const int parameter_conditional_block);
947 * List of functions to run on a list of headers.
951 /** The header prefix to match */
954 /** The length of the prefix to match */
957 /** The function to apply to this line */
958 const parser_func_ptr parser;
961 static const struct parsers client_patterns[] = {
962 { "referer:", 8, client_referrer },
963 { "user-agent:", 11, client_uagent },
964 { "ua-", 3, client_ua },
965 { "from:", 5, client_from },
966 { "cookie:", 7, client_send_cookie },
967 { "x-forwarded-for:", 16, client_x_forwarded },
968 { "Accept-Encoding:", 16, client_accept_encoding },
969 { "TE:", 3, client_te },
970 { "Host:", 5, client_host },
971 { "if-modified-since:", 18, client_if_modified_since },
972 { "Keep-Alive:", 11, crumble },
973 { "connection:", 11, connection },
974 { "proxy-connection:", 17, crumble },
975 { "max-forwards:", 13, client_max_forwards },
976 { "Accept-Language:", 16, client_accept_language },
977 { "if-none-match:", 14, client_if_none_match },
978 { "Range:", 6, client_range },
979 { "Request-Range:", 14, client_range },
980 { "If-Range:", 9, client_range },
981 { "X-Filter:", 9, client_x_filter },
982 { "*", 0, crunch_client_header },
983 { "*", 0, filter_header },
987 static const struct parsers server_patterns[] = {
988 { "HTTP/", 5, server_http },
989 { "set-cookie:", 11, server_set_cookie },
990 { "connection:", 11, connection },
991 { "Content-Type:", 13, server_content_type },
992 { "Content-MD5:", 12, server_content_md5 },
993 { "Content-Encoding:", 17, server_content_encoding },
994 { "Transfer-Encoding:", 18, server_transfer_coding },
995 { "Keep-Alive:", 11, crumble },
996 { "content-disposition:", 20, server_content_disposition },
997 { "Last-Modified:", 14, server_last_modified },
998 { "*", 0, crunch_server_header },
999 { "*", 0, filter_header },
1003 static const add_header_func_ptr add_client_headers[] = {
1005 client_x_forwarded_for_adder,
1007 /* Temporarily disabled: client_accept_encoding_adder, */
1008 connection_close_adder,
1012 static const add_header_func_ptr add_server_headers[] = {
1013 connection_close_adder,
1017 /*********************************************************************
1019 * Function : flush_socket
1021 * Description : Write any pending "buffered" content.
1024 * 1 : fd = file descriptor of the socket to read
1025 * 2 : iob = The I/O buffer to flush, usually csp->iob.
1027 * Returns : On success, the number of bytes written are returned (zero
1028 * indicates nothing was written). On error, -1 is returned,
1029 * and errno is set appropriately. If count is zero and the
1030 * file descriptor refers to a regular file, 0 will be
1031 * returned without causing any other effect. For a special
1032 * file, the results are not portable.
1034 *********************************************************************/
1035 int flush_socket(jb_socket fd, struct iob *iob)
1037 int len = iob->eod - iob->cur;
1044 if (write_socket(fd, iob->cur, (size_t)len))
1048 iob->eod = iob->cur = iob->buf;
1054 /*********************************************************************
1056 * Function : add_to_iob
1058 * Description : Add content to the buffered page, expanding the
1059 * buffer if necessary.
1062 * 1 : csp = Current client state (buffers, headers, etc...)
1063 * 2 : buf = holds the content to be added to the page
1064 * 3 : n = number of bytes to be added
1066 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
1067 * or buffer limit reached.
1069 *********************************************************************/
1070 jb_err add_to_iob(struct client_state *csp, char *buf, int n)
1072 struct iob *iob = csp->iob;
1073 size_t used, offset, need, want;
1076 if (n <= 0) return JB_ERR_OK;
1078 used = (size_t)(iob->eod - iob->buf);
1079 offset = (size_t)(iob->cur - iob->buf);
1080 need = used + (size_t)n + 1;
1083 * If the buffer can't hold the new data, extend it first.
1084 * Use the next power of two if possible, else use the actual need.
1086 if (need > csp->config->buffer_limit)
1088 log_error(LOG_LEVEL_INFO, "Buffer limit reached while extending the buffer (iob)");
1089 return JB_ERR_MEMORY;
1092 if (need > iob->size)
1094 for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2;
1096 if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
1100 else if (NULL != (p = (char *)realloc(iob->buf, need)))
1106 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
1107 return JB_ERR_MEMORY;
1110 /* Update the iob pointers */
1111 iob->cur = p + offset;
1112 iob->eod = p + used;
1116 /* copy the new data into the iob buffer */
1117 memcpy(iob->eod, buf, (size_t)n);
1119 /* point to the end of the data */
1122 /* null terminate == cheap insurance */
1131 /*********************************************************************
1133 * Function : decompress_iob
1135 * Description : Decompress buffered page, expanding the
1136 * buffer as necessary. csp->iob->cur
1137 * should point to the the beginning of the
1138 * compressed data block.
1141 * 1 : csp = Current client state (buffers, headers, etc...)
1143 * Returns : JB_ERR_OK on success,
1144 * JB_ERR_MEMORY if out-of-memory limit reached, and
1145 * JB_ERR_COMPRESS if error decompressing buffer.
1147 *********************************************************************/
1148 jb_err decompress_iob(struct client_state *csp)
1150 char *buf; /* new, uncompressed buffer */
1151 char *cur; /* Current iob position (to keep the original
1152 * iob->cur unmodified if we return early) */
1153 size_t bufsize; /* allocated size of the new buffer */
1154 size_t old_size; /* Content size before decompression */
1155 size_t skip_size; /* Number of bytes at the beginning of the iob
1156 that we should NOT decompress. */
1157 int status; /* return status of the inflate() call */
1158 z_stream zstr; /* used by calls to zlib */
1160 assert(csp->iob->cur - csp->iob->buf > 0);
1161 assert(csp->iob->eod - csp->iob->cur > 0);
1163 bufsize = csp->iob->size;
1164 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
1165 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
1167 cur = csp->iob->cur;
1172 * This is to protect the parsing of gzipped data,
1173 * but it should(?) be valid for deflated data also.
1175 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
1176 return JB_ERR_COMPRESS;
1179 if (csp->content_type & CT_GZIP)
1182 * Our task is slightly complicated by the facts that data
1183 * compressed by gzip does not include a zlib header, and
1184 * that there is no easily accessible interface in zlib to
1185 * handle a gzip header. We strip off the gzip header by
1186 * hand, and later inform zlib not to expect a header.
1190 * Strip off the gzip header. Please see RFC 1952 for more
1191 * explanation of the appropriate fields.
1193 if ((*cur++ != (char)0x1f)
1194 || (*cur++ != (char)0x8b)
1195 || (*cur++ != Z_DEFLATED))
1197 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
1198 return JB_ERR_COMPRESS;
1204 * XXX: These magic numbers should be replaced
1205 * with macros to give a better idea what they do.
1209 /* The gzip header has reserved bits set; bail out. */
1210 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
1211 return JB_ERR_COMPRESS;
1215 /* Skip extra fields if necessary. */
1219 * Skip a given number of bytes, specified
1220 * as a 16-bit little-endian value.
1223 * XXX: This code used to be:
1225 * csp->iob->cur += *csp->iob->cur++ + (*csp->iob->cur++ << 8);
1227 * which I had to change into:
1229 * cur += *cur++ + (*cur++ << 8);
1231 * at which point gcc43 finally noticed that the value
1232 * of cur is undefined (it depends on which of the
1233 * summands is evaluated first).
1235 * I haven't come across a site where this
1236 * code is actually executed yet, but I hope
1240 skip_bytes = *cur++;
1241 skip_bytes = *cur++ << 8;
1243 assert(skip_bytes == *csp->iob->cur - 2 + ((*csp->iob->cur - 1) << 8));
1246 * The number of bytes to skip should be positive
1247 * and we'd like to stay in the buffer.
1249 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
1251 log_error(LOG_LEVEL_ERROR,
1252 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
1254 return JB_ERR_COMPRESS;
1256 log_error(LOG_LEVEL_INFO,
1257 "Skipping %d bytes for gzip compression. Does this sound right?",
1262 /* Skip the filename if necessary. */
1265 /* A null-terminated string is supposed to follow. */
1266 while (*cur++ && (cur < csp->iob->eod));
1270 /* Skip the comment if necessary. */
1273 /* A null-terminated string is supposed to follow. */
1274 while (*cur++ && (cur < csp->iob->eod));
1277 /* Skip the CRC if necessary. */
1283 if (cur >= csp->iob->eod)
1286 * If the current position pointer reached or passed
1287 * the buffer end, we were obviously tricked to skip
1290 log_error(LOG_LEVEL_ERROR,
1291 "Malformed gzip header detected. Aborting decompression.");
1292 return JB_ERR_COMPRESS;
1296 else if (csp->content_type & CT_DEFLATE)
1299 * XXX: The debug level should be lowered
1300 * before the next stable release.
1302 log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
1304 * In theory (that is, according to RFC 1950), deflate-compressed
1305 * data should begin with a two-byte zlib header and have an
1306 * adler32 checksum at the end. It seems that in practice only
1307 * the raw compressed data is sent. Note that this means that
1308 * we are not RFC 1950-compliant here, but the advantage is that
1309 * this actually works. :)
1311 * We add a dummy null byte to tell zlib where the data ends,
1312 * and later inform it not to expect a header.
1314 * Fortunately, add_to_iob() has thoughtfully null-terminated
1315 * the buffer; we can just increment the end pointer to include
1322 log_error(LOG_LEVEL_ERROR,
1323 "Unable to determine compression format for decompression");
1324 return JB_ERR_COMPRESS;
1327 /* Set up the fields required by zlib. */
1328 zstr.next_in = (Bytef *)cur;
1329 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
1330 zstr.zalloc = Z_NULL;
1331 zstr.zfree = Z_NULL;
1332 zstr.opaque = Z_NULL;
1335 * Passing -MAX_WBITS to inflateInit2 tells the library
1336 * that there is no zlib header.
1338 if (inflateInit2 (&zstr, -MAX_WBITS) != Z_OK)
1340 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
1341 return JB_ERR_COMPRESS;
1345 * Next, we allocate new storage for the inflated data.
1346 * We don't modify the existing iob yet, so in case there
1347 * is error in decompression we can recover gracefully.
1349 buf = zalloc(bufsize);
1352 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1353 return JB_ERR_MEMORY;
1356 assert(bufsize >= skip_size);
1357 memcpy(buf, csp->iob->buf, skip_size);
1358 zstr.avail_out = bufsize - skip_size;
1359 zstr.next_out = (Bytef *)buf + skip_size;
1361 /* Try to decompress the whole stream in one shot. */
1362 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
1364 /* We need to allocate more memory for the output buffer. */
1366 char *tmpbuf; /* used for realloc'ing the buffer */
1367 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
1370 * If zlib wants more data then there's a problem, because
1371 * the complete compressed file should have been buffered.
1373 if (0 == zstr.avail_in)
1375 log_error(LOG_LEVEL_ERROR, "Unexpected end of compressed iob");
1376 return JB_ERR_COMPRESS;
1380 * If we tried the limit and still didn't have enough
1381 * memory, just give up.
1383 if (bufsize == csp->config->buffer_limit)
1385 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
1386 return JB_ERR_MEMORY;
1389 /* Try doubling the buffer size each time. */
1392 /* Don't exceed the buffer limit. */
1393 if (bufsize > csp->config->buffer_limit)
1395 bufsize = csp->config->buffer_limit;
1398 /* Try to allocate the new buffer. */
1399 tmpbuf = realloc(buf, bufsize);
1402 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1404 return JB_ERR_MEMORY;
1408 char *oldnext_out = (char *)zstr.next_out;
1411 * Update the fields for inflate() to use the new
1412 * buffer, which may be in a location different from
1415 zstr.avail_out += bufsize - oldbufsize;
1416 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
1419 * Compare with an uglier method of calculating these values
1420 * that doesn't require the extra oldbufsize variable.
1422 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
1423 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
1424 assert(zstr.avail_out > 0);
1430 if (Z_STREAM_ERROR == inflateEnd(&zstr))
1432 log_error(LOG_LEVEL_ERROR,
1433 "Inconsistent stream state after decompression: %s", zstr.msg);
1435 * XXX: Intentionally no return.
1437 * According to zlib.h, Z_STREAM_ERROR is returned
1438 * "if the stream state was inconsistent".
1440 * I assume in this case inflate()'s status
1441 * would also be something different than Z_STREAM_END
1442 * so this check should be redundant, but lets see.
1446 if (status != Z_STREAM_END)
1448 /* We failed to decompress the stream. */
1449 log_error(LOG_LEVEL_ERROR,
1450 "Error in decompressing to the buffer (iob): %s", zstr.msg);
1451 return JB_ERR_COMPRESS;
1455 * Finally, we can actually update the iob, since the
1456 * decompression was successful. First, free the old
1459 freez(csp->iob->buf);
1461 /* Now, update the iob to use the new buffer. */
1462 csp->iob->buf = buf;
1463 csp->iob->cur = csp->iob->buf + skip_size;
1464 csp->iob->eod = (char *)zstr.next_out;
1465 csp->iob->size = bufsize;
1468 * Make sure the new uncompressed iob obeys some minimal
1469 * consistency conditions.
1471 if ((csp->iob->buf < csp->iob->cur)
1472 && (csp->iob->cur <= csp->iob->eod)
1473 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
1475 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
1478 log_error(LOG_LEVEL_RE_FILTER,
1479 "Decompression successful. Old size: %d, new size: %d.",
1480 old_size, new_size);
1484 /* zlib thinks this is OK, so lets do the same. */
1485 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
1490 /* It seems that zlib did something weird. */
1491 log_error(LOG_LEVEL_ERROR,
1492 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
1493 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
1494 csp->iob->eod, csp->iob->buf + csp->iob->size);
1495 return JB_ERR_COMPRESS;
1501 #endif /* defined(FEATURE_ZLIB) */
1504 /*********************************************************************
1506 * Function : string_move
1508 * Description : memmove wrapper to move the last part of a string
1509 * towards the beginning, overwriting the part in
1510 * the middle. strlcpy() can't be used here as the
1514 * 1 : dst = Destination to overwrite
1515 * 2 : src = Source to move.
1519 *********************************************************************/
1520 static void string_move(char *dst, char *src)
1524 /* +1 to copy the terminating nul as well. */
1525 memmove(dst, src, strlen(src)+1);
1529 /*********************************************************************
1531 * Function : normalize_lws
1533 * Description : Reduces unquoted linear white space in headers
1534 * to a single space in accordance with RFC 2616 2.2.
1535 * This simplifies parsing and filtering later on.
1537 * XXX: Remove log messages before
1538 * the next stable release?
1541 * 1 : header = A header with linear white space to reduce.
1545 *********************************************************************/
1546 static void normalize_lws(char *header)
1552 if (ijb_isspace(*p) && ijb_isspace(*(p+1)))
1556 while (ijb_isspace(*q))
1560 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
1561 string_move(p+1, q);
1566 log_error(LOG_LEVEL_HEADER,
1567 "Converting tab to space in '%s'", header);
1572 char *end_of_token = strstr(p+1, "\"");
1574 if (NULL != end_of_token)
1576 /* Don't mess with quoted text. */
1581 log_error(LOG_LEVEL_HEADER,
1582 "Ignoring single quote in '%s'", header);
1588 p = strchr(header, ':');
1589 if ((p != NULL) && (p != header) && ijb_isspace(*(p-1)))
1592 * There's still space before the colon.
1595 string_move(p-1, p);
1600 /*********************************************************************
1602 * Function : get_header
1604 * Description : This (odd) routine will parse the csp->iob
1605 * to get the next complete header.
1608 * 1 : iob = The I/O buffer to parse, usually csp->iob.
1610 * Returns : Any one of the following:
1612 * 1) a pointer to a dynamically allocated string that contains a header line
1613 * 2) NULL indicating that the end of the header was reached
1614 * 3) "" indicating that the end of the iob was reached before finding
1615 * a complete header line.
1617 *********************************************************************/
1618 char *get_header(struct iob *iob)
1622 header = get_header_line(iob);
1624 if ((header == NULL) || (*header == '\0'))
1627 * No complete header read yet, tell the client.
1632 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
1635 * Header spans multiple lines, append the next one.
1637 char *continued_header;
1639 continued_header = get_header_line(iob);
1640 if ((continued_header == NULL) || (*continued_header == '\0'))
1643 * No complete header read yet, return what we got.
1644 * XXX: Should "unread" header instead.
1646 log_error(LOG_LEVEL_INFO,
1647 "Failed to read a multi-line header properly: '%s'",
1652 if (JB_ERR_OK != string_join(&header, continued_header))
1654 log_error(LOG_LEVEL_FATAL,
1655 "Out of memory while appending multiple headers.");
1659 /* XXX: remove before next stable release. */
1660 log_error(LOG_LEVEL_HEADER,
1661 "Merged multiple header lines to: '%s'",
1666 normalize_lws(header);
1673 /*********************************************************************
1675 * Function : get_header_line
1677 * Description : This (odd) routine will parse the csp->iob
1678 * to get the next header line.
1681 * 1 : iob = The I/O buffer to parse, usually csp->iob.
1683 * Returns : Any one of the following:
1685 * 1) a pointer to a dynamically allocated string that contains a header line
1686 * 2) NULL indicating that the end of the header was reached
1687 * 3) "" indicating that the end of the iob was reached before finding
1688 * a complete header line.
1690 *********************************************************************/
1691 static char *get_header_line(struct iob *iob)
1695 if ((iob->cur == NULL)
1696 || ((p = strchr(iob->cur, '\n')) == NULL))
1698 return(""); /* couldn't find a complete header */
1703 ret = strdup(iob->cur);
1706 /* FIXME No way to handle error properly */
1707 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
1712 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
1714 /* is this a blank line (i.e. the end of the header) ? */
1726 /*********************************************************************
1728 * Function : get_header_value
1730 * Description : Get the value of a given header from a chained list
1731 * of header lines or return NULL if no such header is
1732 * present in the list.
1735 * 1 : header_list = pointer to list
1736 * 2 : header_name = string with name of header to look for.
1737 * Trailing colon required, capitalization
1740 * Returns : NULL if not found, else value of header
1742 *********************************************************************/
1743 char *get_header_value(const struct list *header_list, const char *header_name)
1745 struct list_entry *cur_entry;
1749 assert(header_list);
1750 assert(header_name);
1751 length = strlen(header_name);
1753 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
1757 if (!strncmpic(cur_entry->str, header_name, length))
1760 * Found: return pointer to start of value
1762 ret = cur_entry->str + length;
1763 while (*ret && ijb_isspace(*ret)) ret++;
1777 /*********************************************************************
1779 * Function : scan_headers
1781 * Description : Scans headers, applies tags and updates action bits.
1784 * 1 : csp = Current client state (buffers, headers, etc...)
1786 * Returns : JB_ERR_OK
1788 *********************************************************************/
1789 static jb_err scan_headers(struct client_state *csp)
1791 struct list_entry *h; /* Header */
1792 jb_err err = JB_ERR_OK;
1794 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1796 /* Header crunch()ed in previous run? -> ignore */
1797 if (h->str == NULL) continue;
1798 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1799 err = header_tagger(csp, h->str);
1806 /*********************************************************************
1810 * Description : add, delete or modify lines in the HTTP header streams.
1811 * On entry, it receives a linked list of headers space
1812 * that was allocated dynamically (both the list nodes
1813 * and the header contents).
1815 * As a side effect it frees the space used by the original
1819 * 1 : csp = Current client state (buffers, headers, etc...)
1820 * 2 : filter_server_headers = Boolean to switch between
1821 * server and header filtering.
1823 * Returns : JB_ERR_OK in case off success, or
1824 * JB_ERR_MEMORY on out-of-memory error.
1826 *********************************************************************/
1827 jb_err sed(struct client_state *csp, int filter_server_headers)
1829 /* XXX: use more descriptive names. */
1830 struct list_entry *p;
1831 const struct parsers *v;
1832 const add_header_func_ptr *f;
1833 jb_err err = JB_ERR_OK;
1835 if (filter_server_headers)
1837 v = server_patterns;
1838 f = add_server_headers;
1842 v = client_patterns;
1843 f = add_client_headers;
1848 while ((err == JB_ERR_OK) && (v->str != NULL))
1850 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1852 /* Header crunch()ed in previous run? -> ignore */
1853 if (p->str == NULL) continue;
1855 /* Does the current parser handle this header? */
1856 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1857 (v->len == CHECK_EVERY_HEADER_REMAINING))
1859 err = v->parser(csp, &(p->str));
1865 /* place additional headers on the csp->headers list */
1866 while ((err == JB_ERR_OK) && (*f))
1876 /*********************************************************************
1878 * Function : update_server_headers
1880 * Description : Updates server headers after the body has been modified.
1883 * 1 : csp = Current client state (buffers, headers, etc...)
1885 * Returns : JB_ERR_OK in case off success, or
1886 * JB_ERR_MEMORY on out-of-memory error.
1888 *********************************************************************/
1889 jb_err update_server_headers(struct client_state *csp)
1891 jb_err err = JB_ERR_OK;
1893 static const struct parsers server_patterns_light[] = {
1894 { "Content-Length:", 15, server_content_length },
1895 { "Transfer-Encoding:", 18, server_transfer_coding },
1897 { "Content-Encoding:", 17, server_content_encoding },
1898 #endif /* def FEATURE_ZLIB */
1902 if (strncmpic(csp->http->cmd, "HEAD", 4))
1904 const struct parsers *v;
1905 struct list_entry *p;
1907 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1909 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1911 /* Header crunch()ed in previous run? -> ignore */
1912 if (p->str == NULL) continue;
1914 /* Does the current parser handle this header? */
1915 if (strncmpic(p->str, v->str, v->len) == 0)
1917 err = v->parser(csp, (char **)&(p->str));
1927 /*********************************************************************
1929 * Function : header_tagger
1931 * Description : Executes all text substitutions from applying
1932 * tag actions and saves the result as tag.
1934 * XXX: Shares enough code with filter_header() and
1935 * pcrs_filter_response() to warrant some helper functions.
1938 * 1 : csp = Current client state (buffers, headers, etc...)
1939 * 2 : header = Header that is used as tagger input
1941 * Returns : JB_ERR_OK on success and always succeeds
1943 *********************************************************************/
1944 static jb_err header_tagger(struct client_state *csp, char *header)
1946 int wanted_filter_type;
1947 int multi_action_index;
1951 struct file_list *fl;
1952 struct re_filterfile_spec *b;
1953 struct list_entry *tag_name;
1955 int found_filters = 0;
1956 const size_t header_length = strlen(header);
1958 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1960 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1961 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1965 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1966 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1969 /* Check if there are any filters */
1970 for (i = 0; i < MAX_AF_FILES; i++)
1983 if (0 == found_filters)
1985 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1986 "tagging enabled, but no taggers available.");
1990 for (i = 0; i < MAX_AF_FILES; i++)
1993 if ((NULL == fl) || (NULL == fl->f))
1996 * Either there are no filter files
1997 * left, or this filter file just
1998 * contains no valid filters.
2000 * Continue to be sure we don't miss
2001 * valid filter files that are chained
2002 * after empty or invalid ones.
2007 /* For all filters, */
2008 for (b = fl->f; b; b = b->next)
2010 if (b->type != wanted_filter_type)
2012 /* skip the ones we don't care about, */
2015 /* leaving only taggers that could apply, of which we use the ones, */
2016 for (tag_name = csp->action->multi[multi_action_index]->first;
2017 NULL != tag_name; tag_name = tag_name->next)
2019 /* that do apply, and */
2020 if (strcmp(b->name, tag_name->str) == 0)
2022 char *modified_tag = NULL;
2024 size_t size = header_length;
2025 pcrs_job *joblist = b->joblist;
2027 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2029 if (NULL == joblist)
2031 log_error(LOG_LEVEL_RE_FILTER,
2032 "Tagger %s has empty joblist. Nothing to do.", b->name);
2036 /* execute their pcrs_joblist on the header. */
2037 for (job = joblist; NULL != job; job = job->next)
2039 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
2043 /* Success, continue with the modified version. */
2052 /* Tagger doesn't match */
2055 /* Regex failure, log it but continue anyway. */
2056 log_error(LOG_LEVEL_ERROR,
2057 "Problems with tagger \'%s\' and header \'%s\': %s",
2058 b->name, *header, pcrs_strerror(hits));
2060 freez(modified_tag);
2064 if (b->dynamic) pcrs_free_joblist(joblist);
2066 /* If this tagger matched */
2072 * There is to technical limitation which makes
2073 * it impossible to use empty tags, but I assume
2074 * no one would do it intentionally.
2077 log_error(LOG_LEVEL_INFO,
2078 "Tagger \'%s\' created an empty tag. Ignored.",
2083 if (!list_contains_item(csp->tags, tag))
2085 if (JB_ERR_OK != enlist(csp->tags, tag))
2087 log_error(LOG_LEVEL_ERROR,
2088 "Insufficient memory to add tag \'%s\', "
2089 "based on tagger \'%s\' and header \'%s\'",
2090 tag, b->name, *header);
2094 char *action_message;
2096 * update the action bits right away, to make
2097 * tagging based on tags set by earlier taggers
2098 * of the same kind possible.
2100 if (update_action_bits_for_tag(csp, tag))
2102 action_message = "Action bits updated accordingly.";
2106 action_message = "No action bits update necessary.";
2109 log_error(LOG_LEVEL_HEADER,
2110 "Tagger \'%s\' added tag \'%s\'. %s",
2111 b->name, tag, action_message);
2116 /* XXX: Is this log-worthy? */
2117 log_error(LOG_LEVEL_HEADER,
2118 "Tagger \'%s\' didn't add tag \'%s\'. "
2119 "Tag already present", b->name, tag);
2122 } /* if the tagger matched */
2123 } /* if the tagger applies */
2124 } /* for every tagger that could apply */
2125 } /* for all filters */
2126 } /* for all filter files */
2131 /* here begins the family of parser functions that reformat header lines */
2133 /*********************************************************************
2135 * Function : filter_header
2137 * Description : Executes all text substitutions from all applying
2138 * +(server|client)-header-filter actions on the header.
2139 * Most of the code was copied from pcrs_filter_response,
2140 * including the rather short variable names
2143 * 1 : csp = Current client state (buffers, headers, etc...)
2144 * 2 : header = On input, pointer to header to modify.
2145 * On output, pointer to the modified header, or NULL
2146 * to remove the header. This function frees the
2147 * original string if necessary.
2149 * Returns : JB_ERR_OK on success and always succeeds
2151 *********************************************************************/
2152 static jb_err filter_header(struct client_state *csp, char **header)
2156 size_t size = strlen(*header);
2158 char *newheader = NULL;
2161 struct file_list *fl;
2162 struct re_filterfile_spec *b;
2163 struct list_entry *filtername;
2165 int i, found_filters = 0;
2166 int wanted_filter_type;
2167 int multi_action_index;
2169 if (csp->flags & CSP_FLAG_NO_FILTERING)
2174 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
2176 wanted_filter_type = FT_SERVER_HEADER_FILTER;
2177 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
2181 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
2182 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
2186 * Need to check the set of re_filterfiles...
2188 for (i = 0; i < MAX_AF_FILES; i++)
2201 if (0 == found_filters)
2203 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2204 "header filtering enabled, but no matching filters available.");
2208 for (i = 0; i < MAX_AF_FILES; i++)
2211 if ((NULL == fl) || (NULL == fl->f))
2214 * Either there are no filter files
2215 * left, or this filter file just
2216 * contains no valid filters.
2218 * Continue to be sure we don't miss
2219 * valid filter files that are chained
2220 * after empty or invalid ones.
2225 * For all applying +filter actions, look if a filter by that
2226 * name exists and if yes, execute its pcrs_joblist on the
2229 for (b = fl->f; b; b = b->next)
2231 if (b->type != wanted_filter_type)
2233 /* Skip other filter types */
2237 for (filtername = csp->action->multi[multi_action_index]->first;
2238 filtername ; filtername = filtername->next)
2240 if (strcmp(b->name, filtername->str) == 0)
2242 int current_hits = 0;
2243 pcrs_job *joblist = b->joblist;
2245 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2247 if (NULL == joblist)
2249 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2253 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
2254 *header, size, b->name);
2256 /* Apply all jobs from the joblist */
2257 for (job = joblist; NULL != job; job = job->next)
2259 matches = pcrs_execute(job, *header, size, &newheader, &size);
2262 current_hits += matches;
2263 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
2265 *header = newheader;
2267 else if ( 0 == matches )
2269 /* Filter doesn't change header */
2275 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
2276 *header, b->name, pcrs_strerror(matches));
2277 if (newheader != NULL)
2279 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
2285 if (b->dynamic) pcrs_free_joblist(joblist);
2287 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
2288 hits += current_hits;
2295 * Additionally checking for hits is important because if
2296 * the continue hack is triggered, server headers can
2297 * arrive empty to separate multiple heads from each other.
2299 if ((0 == size) && hits)
2301 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
2309 /*********************************************************************
2311 * Function : connection
2313 * Description : Makes sure that the value of the Connection: header
2314 * is "close" and signals connection_close_adder
2318 * 1 : csp = Current client state (buffers, headers, etc...)
2319 * 2 : header = On input, pointer to header to modify.
2320 * On output, pointer to the modified header, or NULL
2321 * to remove the header. This function frees the
2322 * original string if necessary.
2324 * Returns : JB_ERR_OK on success, or
2325 * JB_ERR_MEMORY on out-of-memory error.
2327 *********************************************************************/
2328 static jb_err connection(struct client_state *csp, char **header)
2330 char *old_header = *header;
2332 /* Do we have a 'Connection: close' header? */
2333 if (strcmpic(*header, "Connection: close"))
2335 /* No, create one */
2336 *header = strdup("Connection: close");
2339 return JB_ERR_MEMORY;
2341 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
2345 /* Signal connection_close_adder() to return early. */
2346 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
2348 csp->flags |= CSP_FLAG_SERVER_CONNECTION_CLOSE_SET;
2352 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET;
2359 /*********************************************************************
2361 * Function : crumble
2363 * Description : This is called if a header matches a pattern to "crunch"
2366 * 1 : csp = Current client state (buffers, headers, etc...)
2367 * 2 : header = On input, pointer to header to modify.
2368 * On output, pointer to the modified header, or NULL
2369 * to remove the header. This function frees the
2370 * original string if necessary.
2372 * Returns : JB_ERR_OK on success, or
2373 * JB_ERR_MEMORY on out-of-memory error.
2375 *********************************************************************/
2376 static jb_err crumble(struct client_state *csp, char **header)
2378 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2384 /*********************************************************************
2386 * Function : crunch_server_header
2388 * Description : Crunch server header if it matches a string supplied by the
2389 * user. Called from `sed'.
2392 * 1 : csp = Current client state (buffers, headers, etc...)
2393 * 2 : header = On input, pointer to header to modify.
2394 * On output, pointer to the modified header, or NULL
2395 * to remove the header. This function frees the
2396 * original string if necessary.
2398 * Returns : JB_ERR_OK on success and always succeeds
2400 *********************************************************************/
2401 static jb_err crunch_server_header(struct client_state *csp, char **header)
2403 const char *crunch_pattern;
2405 /* Do we feel like crunching? */
2406 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2408 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2410 /* Is the current header the lucky one? */
2411 if (strstr(*header, crunch_pattern))
2413 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2422 /*********************************************************************
2424 * Function : server_content_type
2426 * Description : Set the content-type for filterable types (text/.*,
2427 * .*xml.*, javascript and image/gif) unless filtering has been
2428 * forbidden (CT_TABOO) while parsing earlier headers.
2429 * NOTE: Since text/plain is commonly used by web servers
2430 * for files whose correct type is unknown, we don't
2431 * set CT_TEXT for it.
2434 * 1 : csp = Current client state (buffers, headers, etc...)
2435 * 2 : header = On input, pointer to header to modify.
2436 * On output, pointer to the modified header, or NULL
2437 * to remove the header. This function frees the
2438 * original string if necessary.
2440 * Returns : JB_ERR_OK on success, or
2441 * JB_ERR_MEMORY on out-of-memory error.
2443 *********************************************************************/
2444 static jb_err server_content_type(struct client_state *csp, char **header)
2446 /* Remove header if it isn't the first Content-Type header */
2447 if ((csp->content_type & CT_DECLARED))
2450 * Another, slightly slower, way to see if
2451 * we already parsed another Content-Type header.
2453 assert(NULL != get_header_value(csp->headers, "Content-Type:"));
2455 log_error(LOG_LEVEL_ERROR,
2456 "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
2464 * Signal that the Content-Type has been set.
2466 csp->content_type |= CT_DECLARED;
2468 if (!(csp->content_type & CT_TABOO))
2471 * XXX: The assumption that text/plain is a sign of
2472 * binary data seems to be somewhat unreasonable nowadays
2473 * and should be dropped after 3.0.8 is out.
2475 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2476 || strstr(*header, "xml")
2477 || strstr(*header, "application/x-javascript"))
2479 csp->content_type |= CT_TEXT;
2481 else if (strstr(*header, "image/gif"))
2483 csp->content_type |= CT_GIF;
2488 * Are we messing with the content type?
2490 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2493 * Make sure the user doesn't accidently
2494 * change the content type of binary documents.
2496 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2499 *header = strdup("Content-Type: ");
2500 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2504 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2505 return JB_ERR_MEMORY;
2507 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2511 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2512 "It doesn't look like a content type that should be filtered. "
2513 "Enable force-text-mode if you know what you're doing.", *header);
2521 /*********************************************************************
2523 * Function : server_transfer_coding
2525 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2526 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2527 * - Remove header if body was chunked but has been
2528 * de-chunked for filtering.
2531 * 1 : csp = Current client state (buffers, headers, etc...)
2532 * 2 : header = On input, pointer to header to modify.
2533 * On output, pointer to the modified header, or NULL
2534 * to remove the header. This function frees the
2535 * original string if necessary.
2537 * Returns : JB_ERR_OK on success, or
2538 * JB_ERR_MEMORY on out-of-memory error.
2540 *********************************************************************/
2541 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2544 * Turn off pcrs and gif filtering if body compressed
2546 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2550 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2552 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2553 csp->http->cmd, *header);
2554 #endif /* def FEATURE_ZLIB */
2555 csp->content_type = CT_TABOO;
2559 * Raise flag if body chunked
2561 if (strstr(*header, "chunked"))
2563 csp->flags |= CSP_FLAG_CHUNKED;
2566 * If the body was modified, it has been de-chunked first
2567 * and the header must be removed.
2569 * FIXME: If there is more than one transfer encoding,
2570 * only the "chunked" part should be removed here.
2572 if (csp->flags & CSP_FLAG_MODIFIED)
2574 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2583 /*********************************************************************
2585 * Function : server_content_encoding
2587 * Description : This function is run twice for each request,
2588 * unless FEATURE_ZLIB and filtering are disabled.
2590 * The first run is used to check if the content
2591 * is compressed, if FEATURE_ZLIB is disabled
2592 * filtering is then disabled as well, if FEATURE_ZLIB
2593 * is enabled the content is marked for decompression.
2595 * The second run is used to remove the Content-Encoding
2596 * header if the decompression was successful.
2599 * 1 : csp = Current client state (buffers, headers, etc...)
2600 * 2 : header = On input, pointer to header to modify.
2601 * On output, pointer to the modified header, or NULL
2602 * to remove the header. This function frees the
2603 * original string if necessary.
2605 * Returns : JB_ERR_OK on success, or
2606 * JB_ERR_MEMORY on out-of-memory error.
2608 *********************************************************************/
2609 static jb_err server_content_encoding(struct client_state *csp, char **header)
2612 if ((csp->flags & CSP_FLAG_MODIFIED)
2613 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2616 * We successfully decompressed the content,
2617 * and have to clean the header now, so the
2618 * client no longer expects compressed data..
2620 * XXX: There is a difference between cleaning
2621 * and removing it completely.
2623 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2626 else if (strstr(*header, "gzip"))
2628 /* Mark for gzip decompression */
2629 csp->content_type |= CT_GZIP;
2631 else if (strstr(*header, "deflate"))
2633 /* Mark for zlib decompression */
2634 csp->content_type |= CT_DEFLATE;
2636 else if (strstr(*header, "compress"))
2639 * We can't decompress this; therefore we can't filter
2642 csp->content_type |= CT_TABOO;
2644 #else /* !defined(FEATURE_ZLIB) */
2645 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2648 * Body is compressed, turn off pcrs and gif filtering.
2650 csp->content_type |= CT_TABOO;
2653 * Log a warning if the user expects the content to be filtered.
2655 if ((csp->rlist != NULL) &&
2656 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2658 log_error(LOG_LEVEL_INFO,
2659 "Compressed content detected, content filtering disabled. "
2660 "Consider recompiling Privoxy with zlib support or "
2661 "enable the prevent-compression action.");
2664 #endif /* defined(FEATURE_ZLIB) */
2671 /*********************************************************************
2673 * Function : server_content_length
2675 * Description : Adjust Content-Length header if we modified
2679 * 1 : csp = Current client state (buffers, headers, etc...)
2680 * 2 : header = On input, pointer to header to modify.
2681 * On output, pointer to the modified header, or NULL
2682 * to remove the header. This function frees the
2683 * original string if necessary.
2685 * Returns : JB_ERR_OK on success, or
2686 * JB_ERR_MEMORY on out-of-memory error.
2688 *********************************************************************/
2689 static jb_err server_content_length(struct client_state *csp, char **header)
2691 const size_t max_header_length = 80;
2693 /* Regenerate header if the content was modified. */
2694 if (csp->flags & CSP_FLAG_MODIFIED)
2697 *header = (char *) zalloc(max_header_length);
2698 if (*header == NULL)
2700 return JB_ERR_MEMORY;
2703 snprintf(*header, max_header_length, "Content-Length: %d",
2704 (int)csp->content_length);
2705 log_error(LOG_LEVEL_HEADER, "Adjusted Content-Length to %d",
2706 (int)csp->content_length);
2713 /*********************************************************************
2715 * Function : server_content_md5
2717 * Description : Crumble any Content-MD5 headers if the document was
2718 * modified. FIXME: Should we re-compute instead?
2721 * 1 : csp = Current client state (buffers, headers, etc...)
2722 * 2 : header = On input, pointer to header to modify.
2723 * On output, pointer to the modified header, or NULL
2724 * to remove the header. This function frees the
2725 * original string if necessary.
2727 * Returns : JB_ERR_OK on success, or
2728 * JB_ERR_MEMORY on out-of-memory error.
2730 *********************************************************************/
2731 static jb_err server_content_md5(struct client_state *csp, char **header)
2733 if (csp->flags & CSP_FLAG_MODIFIED)
2735 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2743 /*********************************************************************
2745 * Function : server_content_disposition
2747 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2748 * Called from `sed'.
2751 * 1 : csp = Current client state (buffers, headers, etc...)
2752 * 2 : header = On input, pointer to header to modify.
2753 * On output, pointer to the modified header, or NULL
2754 * to remove the header. This function frees the
2755 * original string if necessary.
2757 * Returns : JB_ERR_OK on success, or
2758 * JB_ERR_MEMORY on out-of-memory error.
2760 *********************************************************************/
2761 static jb_err server_content_disposition(struct client_state *csp, char **header)
2766 * Are we messing with the Content-Disposition header?
2768 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2774 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2776 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2779 * Blocking content-disposition header
2781 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2788 * Replacing Content-Disposition header
2791 *header = strdup("Content-Disposition: ");
2792 string_append(header, newval);
2794 if (*header != NULL)
2796 log_error(LOG_LEVEL_HEADER,
2797 "Content-Disposition header crunched and replaced with: %s", *header);
2800 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2804 /*********************************************************************
2806 * Function : server_last_modified
2808 * Description : Changes Last-Modified header to the actual date
2809 * to help hide-if-modified-since.
2810 * Called from `sed'.
2813 * 1 : csp = Current client state (buffers, headers, etc...)
2814 * 2 : header = On input, pointer to header to modify.
2815 * On output, pointer to the modified header, or NULL
2816 * to remove the header. This function frees the
2817 * original string if necessary.
2819 * Returns : JB_ERR_OK on success, or
2820 * JB_ERR_MEMORY on out-of-memory error.
2822 *********************************************************************/
2823 static jb_err server_last_modified(struct client_state *csp, char **header)
2826 char buf[BUFFER_SIZE];
2829 #ifdef HAVE_GMTIME_R
2832 struct tm *timeptr = NULL;
2833 time_t now, last_modified;
2835 long int days, hours, minutes, seconds;
2838 * Are we messing with the Last-Modified header?
2840 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2846 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2848 if (0 == strcmpic(newval, "block") )
2851 * Blocking Last-Modified header. Useless but why not.
2853 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2857 else if (0 == strcmpic(newval, "reset-to-request-time"))
2860 * Setting Last-Modified Header to now.
2862 get_http_time(0, buf, sizeof(buf));
2864 *header = strdup("Last-Modified: ");
2865 string_append(header, buf);
2867 if (*header == NULL)
2869 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2873 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2876 else if (0 == strcmpic(newval, "randomize"))
2878 const char *header_time = *header + sizeof("Last-Modified:");
2880 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2882 #ifdef HAVE_GMTIME_R
2883 timeptr = gmtime_r(&now, &gmt);
2884 #elif FEATURE_PTHREAD
2885 privoxy_mutex_lock(&gmtime_mutex);
2886 timeptr = gmtime(&now);
2887 privoxy_mutex_unlock(&gmtime_mutex);
2889 timeptr = gmtime(&now);
2891 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2893 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2898 rtime = (long int)difftime(now, last_modified);
2907 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2909 rtime = pick_from_range(rtime);
2910 if (negative) rtime *= -1;
2911 last_modified += rtime;
2912 #ifdef HAVE_GMTIME_R
2913 timeptr = gmtime_r(&last_modified, &gmt);
2914 #elif FEATURE_PTHREAD
2915 privoxy_mutex_lock(&gmtime_mutex);
2916 timeptr = gmtime(&last_modified);
2917 privoxy_mutex_unlock(&gmtime_mutex);
2919 timeptr = gmtime(&last_modified);
2921 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
2923 *header = strdup("Last-Modified: ");
2924 string_append(header, newheader);
2926 if (*header == NULL)
2928 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2929 return JB_ERR_MEMORY;
2932 days = rtime / (3600 * 24);
2933 hours = rtime / 3600 % 24;
2934 minutes = rtime / 60 % 60;
2935 seconds = rtime % 60;
2937 log_error(LOG_LEVEL_HEADER,
2938 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2939 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2940 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2944 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2953 /*********************************************************************
2955 * Function : client_accept_encoding
2957 * Description : Rewrite the client's Accept-Encoding header so that
2958 * if doesn't allow compression, if the action applies.
2959 * Note: For HTTP/1.0 the absence of the header is enough.
2962 * 1 : csp = Current client state (buffers, headers, etc...)
2963 * 2 : header = On input, pointer to header to modify.
2964 * On output, pointer to the modified header, or NULL
2965 * to remove the header. This function frees the
2966 * original string if necessary.
2968 * Returns : JB_ERR_OK on success, or
2969 * JB_ERR_MEMORY on out-of-memory error.
2971 *********************************************************************/
2972 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2974 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2976 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2980 /* Temporarily disable the correct behaviour to
2981 * work around a PHP bug.
2983 * if (!strcmpic(csp->http->ver, "HTTP/1.1"))
2985 * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0");
2986 * if (*header == NULL)
2988 * return JB_ERR_MEMORY;
2999 /*********************************************************************
3001 * Function : client_te
3003 * Description : Rewrite the client's TE header so that
3004 * if doesn't allow compression, if the action applies.
3007 * 1 : csp = Current client state (buffers, headers, etc...)
3008 * 2 : header = On input, pointer to header to modify.
3009 * On output, pointer to the modified header, or NULL
3010 * to remove the header. This function frees the
3011 * original string if necessary.
3013 * Returns : JB_ERR_OK on success, or
3014 * JB_ERR_MEMORY on out-of-memory error.
3016 *********************************************************************/
3017 static jb_err client_te(struct client_state *csp, char **header)
3019 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3022 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
3029 /*********************************************************************
3031 * Function : client_referrer
3033 * Description : Handle the "referer" config setting properly.
3034 * Called from `sed'.
3037 * 1 : csp = Current client state (buffers, headers, etc...)
3038 * 2 : header = On input, pointer to header to modify.
3039 * On output, pointer to the modified header, or NULL
3040 * to remove the header. This function frees the
3041 * original string if necessary.
3043 * Returns : JB_ERR_OK on success, or
3044 * JB_ERR_MEMORY on out-of-memory error.
3046 *********************************************************************/
3047 static jb_err client_referrer(struct client_state *csp, char **header)
3049 const char *parameter;
3050 /* booleans for parameters we have to check multiple times */
3051 int parameter_conditional_block;
3052 int parameter_conditional_forge;
3054 #ifdef FEATURE_FORCE_LOAD
3056 * Since the referrer can include the prefix even
3057 * if the request itself is non-forced, we must
3058 * clean it unconditionally.
3060 * XXX: strclean is too broad
3062 strclean(*header, FORCE_PREFIX);
3063 #endif /* def FEATURE_FORCE_LOAD */
3065 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
3067 /* Nothing left to do */
3071 parameter = csp->action->string[ACTION_STRING_REFERER];
3072 assert(parameter != NULL);
3073 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
3074 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
3076 if (!parameter_conditional_block && !parameter_conditional_forge)
3079 * As conditional-block and conditional-forge are the only
3080 * parameters that rely on the original referrer, we can
3081 * remove it now for all the others.
3086 if (0 == strcmpic(parameter, "block"))
3088 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
3091 else if (parameter_conditional_block || parameter_conditional_forge)
3093 return handle_conditional_hide_referrer_parameter(header,
3094 csp->http->hostport, parameter_conditional_block);
3096 else if (0 == strcmpic(parameter, "forge"))
3098 return create_forged_referrer(header, csp->http->hostport);
3102 /* interpret parameter as user-supplied referer to fake */
3103 return create_fake_referrer(header, parameter);
3108 /*********************************************************************
3110 * Function : client_accept_language
3112 * Description : Handle the "Accept-Language" config setting properly.
3113 * Called from `sed'.
3116 * 1 : csp = Current client state (buffers, headers, etc...)
3117 * 2 : header = On input, pointer to header to modify.
3118 * On output, pointer to the modified header, or NULL
3119 * to remove the header. This function frees the
3120 * original string if necessary.
3122 * Returns : JB_ERR_OK on success, or
3123 * JB_ERR_MEMORY on out-of-memory error.
3125 *********************************************************************/
3126 static jb_err client_accept_language(struct client_state *csp, char **header)
3131 * Are we messing with the Accept-Language?
3133 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
3135 /*I don't think so*/
3139 newval = csp->action->string[ACTION_STRING_LANGUAGE];
3141 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3144 * Blocking Accept-Language header
3146 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
3153 * Replacing Accept-Language header
3156 *header = strdup("Accept-Language: ");
3157 string_append(header, newval);
3159 if (*header == NULL)
3161 log_error(LOG_LEVEL_ERROR,
3162 "Insufficient memory. Accept-Language header crunched without replacement.");
3166 log_error(LOG_LEVEL_HEADER,
3167 "Accept-Language header crunched and replaced with: %s", *header);
3170 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3174 /*********************************************************************
3176 * Function : crunch_client_header
3178 * Description : Crunch client header if it matches a string supplied by the
3179 * user. Called from `sed'.
3182 * 1 : csp = Current client state (buffers, headers, etc...)
3183 * 2 : header = On input, pointer to header to modify.
3184 * On output, pointer to the modified header, or NULL
3185 * to remove the header. This function frees the
3186 * original string if necessary.
3188 * Returns : JB_ERR_OK on success and always succeeds
3190 *********************************************************************/
3191 static jb_err crunch_client_header(struct client_state *csp, char **header)
3193 const char *crunch_pattern;
3195 /* Do we feel like crunching? */
3196 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
3198 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
3200 /* Is the current header the lucky one? */
3201 if (strstr(*header, crunch_pattern))
3203 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
3211 /*********************************************************************
3213 * Function : client_uagent
3215 * Description : Handle the "user-agent" config setting properly
3216 * and remember its original value to enable browser
3217 * bug workarounds. Called from `sed'.
3220 * 1 : csp = Current client state (buffers, headers, etc...)
3221 * 2 : header = On input, pointer to header to modify.
3222 * On output, pointer to the modified header, or NULL
3223 * to remove the header. This function frees the
3224 * original string if necessary.
3226 * Returns : JB_ERR_OK on success, or
3227 * JB_ERR_MEMORY on out-of-memory error.
3229 *********************************************************************/
3230 static jb_err client_uagent(struct client_state *csp, char **header)
3234 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
3239 newval = csp->action->string[ACTION_STRING_USER_AGENT];
3246 *header = strdup("User-Agent: ");
3247 string_append(header, newval);
3249 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
3251 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3255 /*********************************************************************
3257 * Function : client_ua
3259 * Description : Handle "ua-" headers properly. Called from `sed'.
3262 * 1 : csp = Current client state (buffers, headers, etc...)
3263 * 2 : header = On input, pointer to header to modify.
3264 * On output, pointer to the modified header, or NULL
3265 * to remove the header. This function frees the
3266 * original string if necessary.
3268 * Returns : JB_ERR_OK on success, or
3269 * JB_ERR_MEMORY on out-of-memory error.
3271 *********************************************************************/
3272 static jb_err client_ua(struct client_state *csp, char **header)
3274 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3276 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3284 /*********************************************************************
3286 * Function : client_from
3288 * Description : Handle the "from" config setting properly.
3289 * Called from `sed'.
3292 * 1 : csp = Current client state (buffers, headers, etc...)
3293 * 2 : header = On input, pointer to header to modify.
3294 * On output, pointer to the modified header, or NULL
3295 * to remove the header. This function frees the
3296 * original string if necessary.
3298 * Returns : JB_ERR_OK on success, or
3299 * JB_ERR_MEMORY on out-of-memory error.
3301 *********************************************************************/
3302 static jb_err client_from(struct client_state *csp, char **header)
3306 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3313 newval = csp->action->string[ACTION_STRING_FROM];
3316 * Are we blocking the e-mail address?
3318 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3320 log_error(LOG_LEVEL_HEADER, "crunched From!");
3324 log_error(LOG_LEVEL_HEADER, " modified");
3326 *header = strdup("From: ");
3327 string_append(header, newval);
3329 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3333 /*********************************************************************
3335 * Function : client_send_cookie
3337 * Description : Crunches the "cookie" header if necessary.
3338 * Called from `sed'.
3340 * XXX: Stupid name, doesn't send squat.
3343 * 1 : csp = Current client state (buffers, headers, etc...)
3344 * 2 : header = On input, pointer to header to modify.
3345 * On output, pointer to the modified header, or NULL
3346 * to remove the header. This function frees the
3347 * original string if necessary.
3349 * Returns : JB_ERR_OK on success, or
3350 * JB_ERR_MEMORY on out-of-memory error.
3352 *********************************************************************/
3353 static jb_err client_send_cookie(struct client_state *csp, char **header)
3355 if (csp->action->flags & ACTION_NO_COOKIE_READ)
3357 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3365 /*********************************************************************
3367 * Function : client_x_forwarded
3369 * Description : Handle the "x-forwarded-for" config setting properly,
3370 * also used in the add_client_headers list. Called from `sed'.
3373 * 1 : csp = Current client state (buffers, headers, etc...)
3374 * 2 : header = On input, pointer to header to modify.
3375 * On output, pointer to the modified header, or NULL
3376 * to remove the header. This function frees the
3377 * original string if necessary.
3379 * Returns : JB_ERR_OK on success, or
3380 * JB_ERR_MEMORY on out-of-memory error.
3382 *********************************************************************/
3383 jb_err client_x_forwarded(struct client_state *csp, char **header)
3385 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
3387 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
3389 if (0 == strcmpic(parameter, "block"))
3392 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3394 else if (0 == strcmpic(parameter, "add"))
3396 string_append(header, ", ");
3397 string_append(header, csp->ip_addr_str);
3399 if (*header == NULL)
3401 return JB_ERR_MEMORY;
3403 log_error(LOG_LEVEL_HEADER,
3404 "Appended client IP address to %s", *header);
3405 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
3413 /*********************************************************************
3415 * Function : client_max_forwards
3417 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3418 * from the value of the Max-Forwards header field.
3421 * 1 : csp = Current client state (buffers, headers, etc...)
3422 * 2 : header = On input, pointer to header to modify.
3423 * On output, pointer to the modified header, or NULL
3424 * to remove the header. This function frees the
3425 * original string if necessary.
3427 * Returns : JB_ERR_OK on success, or
3428 * JB_ERR_MEMORY on out-of-memory error.
3430 *********************************************************************/
3431 static jb_err client_max_forwards(struct client_state *csp, char **header)
3435 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3436 (0 == strcmpic(csp->http->gpc, "options")))
3438 assert(*(*header+12) == ':');
3439 if (1 == sscanf(*header+12, ": %u", &max_forwards))
3441 if (max_forwards > 0)
3443 snprintf(*header, strlen(*header)+1, "Max-Forwards: %u", --max_forwards);
3444 log_error(LOG_LEVEL_HEADER, "Max-Forwards value for %s request reduced to %u.",
3445 csp->http->gpc, max_forwards);
3447 else if (max_forwards < 0)
3449 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3455 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3464 /*********************************************************************
3466 * Function : client_host
3468 * Description : If the request URI did not contain host and
3469 * port information, parse and evaluate the Host
3472 * Also, kill ill-formed HOST: headers as sent by
3473 * Apple's iTunes software when used with a proxy.
3476 * 1 : csp = Current client state (buffers, headers, etc...)
3477 * 2 : header = On input, pointer to header to modify.
3478 * On output, pointer to the modified header, or NULL
3479 * to remove the header. This function frees the
3480 * original string if necessary.
3482 * Returns : JB_ERR_OK on success, or
3483 * JB_ERR_MEMORY on out-of-memory error.
3485 *********************************************************************/
3486 static jb_err client_host(struct client_state *csp, char **header)
3491 * If the header field name is all upper-case, chances are that it's
3492 * an ill-formed one from iTunes. BTW, killing innocent headers here is
3493 * not a problem -- they are regenerated later.
3495 if ((*header)[1] == 'O')
3497 log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
3502 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3503 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3506 if (NULL == (p = strdup((*header)+6)))
3508 return JB_ERR_MEMORY;
3511 if (NULL == (q = strdup(p)))
3514 return JB_ERR_MEMORY;
3517 freez(csp->http->hostport);
3518 csp->http->hostport = p;
3519 freez(csp->http->host);
3520 csp->http->host = q;
3521 q = strchr(csp->http->host, ':');
3524 /* Terminate hostname and evaluate port string */
3526 csp->http->port = atoi(q);
3530 csp->http->port = csp->http->ssl ? 443 : 80;
3533 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3534 csp->http->hostport, csp->http->host, csp->http->port);
3537 /* Signal client_host_adder() to return right away */
3538 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3544 /*********************************************************************
3546 * Function : client_if_modified_since
3548 * Description : Remove or modify the If-Modified-Since header.
3551 * 1 : csp = Current client state (buffers, headers, etc...)
3552 * 2 : header = On input, pointer to header to modify.
3553 * On output, pointer to the modified header, or NULL
3554 * to remove the header. This function frees the
3555 * original string if necessary.
3557 * Returns : JB_ERR_OK on success, or
3558 * JB_ERR_MEMORY on out-of-memory error.
3560 *********************************************************************/
3561 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3564 #ifdef HAVE_GMTIME_R
3567 struct tm *timeptr = NULL;
3571 long int hours, minutes, seconds;
3575 if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3578 * The client got an error message because of a temporary problem,
3579 * the problem is gone and the client now tries to revalidate our
3580 * error message on the real server. The revalidation would always
3581 * end with the transmission of the whole document and there is
3582 * no need to expose the bogus If-Modified-Since header.
3584 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3587 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3589 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3591 if ((0 == strcmpic(newval, "block")))
3593 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3596 else /* add random value */
3598 const char *header_time = *header + sizeof("If-Modified-Since:");
3600 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3602 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3607 rtime = strtol(newval, &endptr, 0);
3610 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3611 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3618 rtime = pick_from_range(rtime);
3622 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3625 tm += rtime * (negative ? -1 : 1);
3626 #ifdef HAVE_GMTIME_R
3627 timeptr = gmtime_r(&tm, &gmt);
3628 #elif FEATURE_PTHREAD
3629 privoxy_mutex_lock(&gmtime_mutex);
3630 timeptr = gmtime(&tm);
3631 privoxy_mutex_unlock(&gmtime_mutex);
3633 timeptr = gmtime(&tm);
3635 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
3638 *header = strdup("If-Modified-Since: ");
3639 string_append(header, newheader);
3641 if (*header == NULL)
3643 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3644 return JB_ERR_MEMORY;
3647 hours = rtime / 3600;
3648 minutes = rtime / 60 % 60;
3649 seconds = rtime % 60;
3651 log_error(LOG_LEVEL_HEADER,
3652 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3653 *header, (negative) ? "subtracted" : "added", hours,
3654 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3655 seconds, (seconds == 1) ? ")" : "s)");
3664 /*********************************************************************
3666 * Function : client_if_none_match
3668 * Description : Remove the If-None-Match header.
3671 * 1 : csp = Current client state (buffers, headers, etc...)
3672 * 2 : header = On input, pointer to header to modify.
3673 * On output, pointer to the modified header, or NULL
3674 * to remove the header. This function frees the
3675 * original string if necessary.
3677 * Returns : JB_ERR_OK on success, or
3678 * JB_ERR_MEMORY on out-of-memory error.
3680 *********************************************************************/
3681 static jb_err client_if_none_match(struct client_state *csp, char **header)
3683 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3685 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3693 /*********************************************************************
3695 * Function : client_x_filter
3697 * Description : Disables filtering if the client set "X-Filter: No".
3698 * Called from `sed'.
3701 * 1 : csp = Current client state (buffers, headers, etc...)
3702 * 2 : header = On input, pointer to header to modify.
3703 * On output, pointer to the modified header, or NULL
3704 * to remove the header. This function frees the
3705 * original string if necessary.
3707 * Returns : JB_ERR_OK on success
3709 *********************************************************************/
3710 jb_err client_x_filter(struct client_state *csp, char **header)
3712 if ( 0 == strcmpic(*header, "X-Filter: No"))
3714 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3716 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3720 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3722 log_error(LOG_LEVEL_HEADER,
3723 "force-text-mode overruled the client's request to fetch without filtering!");
3727 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3728 csp->flags |= CSP_FLAG_NO_FILTERING;
3729 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3731 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3739 /*********************************************************************
3741 * Function : client_range
3743 * Description : Removes Range, Request-Range and If-Range headers if
3744 * content filtering is enabled. If the client's version
3745 * of the document has been altered by Privoxy, the server
3746 * could interpret the range differently than the client
3747 * intended in which case the user could end up with
3748 * corrupted content.
3751 * 1 : csp = Current client state (buffers, headers, etc...)
3752 * 2 : header = On input, pointer to header to modify.
3753 * On output, pointer to the modified header, or NULL
3754 * to remove the header. This function frees the
3755 * original string if necessary.
3757 * Returns : JB_ERR_OK
3759 *********************************************************************/
3760 static jb_err client_range(struct client_state *csp, char **header)
3762 if (content_filters_enabled(csp->action))
3764 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3765 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3772 /* the following functions add headers directly to the header list */
3774 /*********************************************************************
3776 * Function : client_host_adder
3778 * Description : Adds the Host: header field if it is missing.
3779 * Called from `sed'.
3782 * 1 : csp = Current client state (buffers, headers, etc...)
3784 * Returns : JB_ERR_OK on success, or
3785 * JB_ERR_MEMORY on out-of-memory error.
3787 *********************************************************************/
3788 static jb_err client_host_adder(struct client_state *csp)
3793 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3795 /* Header already set by the client, nothing to do. */
3799 if ( !csp->http->hostport || !*(csp->http->hostport))
3801 /* XXX: When does this happen and why is it OK? */
3802 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3807 * remove 'user:pass@' from 'proto://user:pass@host'
3809 if ( (p = strchr( csp->http->hostport, '@')) != NULL )
3815 p = csp->http->hostport;
3818 /* XXX: Just add it, we already made sure that it will be unique */
3819 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3820 err = enlist_unique_header(csp->headers, "Host", p);
3827 /*********************************************************************
3829 * Function : client_accept_encoding_adder
3831 * Description : Add an Accept-Encoding header to the client's request
3832 * that disables compression if the action applies, and
3833 * the header is not already there. Called from `sed'.
3834 * Note: For HTTP/1.0, the absence of the header is enough.
3837 * 1 : csp = Current client state (buffers, headers, etc...)
3839 * Returns : JB_ERR_OK on success, or
3840 * JB_ERR_MEMORY on out-of-memory error.
3842 *********************************************************************/
3843 static jb_err client_accept_encoding_adder(struct client_state *csp)
3845 if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3846 && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
3848 return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
3856 /*********************************************************************
3858 * Function : client_xtra_adder
3860 * Description : Used in the add_client_headers list. Called from `sed'.
3863 * 1 : csp = Current client state (buffers, headers, etc...)
3865 * Returns : JB_ERR_OK on success, or
3866 * JB_ERR_MEMORY on out-of-memory error.
3868 *********************************************************************/
3869 static jb_err client_xtra_adder(struct client_state *csp)
3871 struct list_entry *lst;
3874 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3875 lst ; lst = lst->next)
3877 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3878 err = enlist(csp->headers, lst->str);
3890 /*********************************************************************
3892 * Function : client_x_forwarded_for_adder
3894 * Description : Used in the add_client_headers list. Called from `sed'.
3897 * 1 : csp = Current client state (buffers, headers, etc...)
3899 * Returns : JB_ERR_OK on success, or
3900 * JB_ERR_MEMORY on out-of-memory error.
3902 *********************************************************************/
3903 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3905 char *header = NULL;
3908 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3909 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3910 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3913 * If we aren't adding X-Forwarded-For headers,
3914 * or we already appended an existing X-Forwarded-For
3915 * header, there's nothing left to do here.
3920 header = strdup("X-Forwarded-For: ");
3921 string_append(&header, csp->ip_addr_str);
3925 return JB_ERR_MEMORY;
3928 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3929 err = enlist(csp->headers, header);
3936 /*********************************************************************
3938 * Function : connection_close_adder
3940 * Description : "Temporary" fix for the needed but missing HTTP/1.1
3941 * support. Adds a "Connection: close" header to csp->headers
3942 * unless the header was already present. Called from `sed'.
3944 * FIXME: This whole function shouldn't be neccessary!
3947 * 1 : csp = Current client state (buffers, headers, etc...)
3949 * Returns : JB_ERR_OK on success, or
3950 * JB_ERR_MEMORY on out-of-memory error.
3952 *********************************************************************/
3953 static jb_err connection_close_adder(struct client_state *csp)
3955 const unsigned int flags = csp->flags;
3958 * Return right away if
3960 * - we're parsing server headers and the server header
3961 * "Connection: close" is already set, or if
3963 * - we're parsing client headers and the client header
3964 * "Connection: close" is already set.
3966 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE
3967 && flags & CSP_FLAG_SERVER_CONNECTION_CLOSE_SET)
3968 ||(!(flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3969 && flags & CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET))
3974 log_error(LOG_LEVEL_HEADER, "Adding: Connection: close");
3976 return enlist(csp->headers, "Connection: close");
3980 /*********************************************************************
3982 * Function : server_http
3984 * Description : - Save the HTTP Status into csp->http->status
3985 * - Set CT_TABOO to prevent filtering if the answer
3986 * is a partial range (HTTP status 206)
3987 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3991 * 1 : csp = Current client state (buffers, headers, etc...)
3992 * 2 : header = On input, pointer to header to modify.
3993 * On output, pointer to the modified header, or NULL
3994 * to remove the header. This function frees the
3995 * original string if necessary.
3997 * Returns : JB_ERR_OK on success, or
3998 * JB_ERR_MEMORY on out-of-memory error.
4000 *********************************************************************/
4001 static jb_err server_http(struct client_state *csp, char **header)
4003 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
4004 if (csp->http->status == 206)
4006 csp->content_type = CT_TABOO;
4009 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
4011 /* XXX: Should we do a real validity check here? */
4012 if (strlen(*header) > 8)
4015 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
4020 * XXX: Should we block the request or
4021 * enlist a valid status code line here?
4023 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
4024 "Downgrading to HTTP/1.0 impossible.");
4032 /*********************************************************************
4034 * Function : server_set_cookie
4036 * Description : Handle the server "cookie" header properly.
4037 * Log cookie to the jar file. Then "crunch",
4038 * accept or rewrite it to a session cookie.
4039 * Called from `sed'.
4041 * TODO: Allow the user to specify a new expiration
4042 * time to cause the cookie to expire even before the
4043 * browser is closed.
4046 * 1 : csp = Current client state (buffers, headers, etc...)
4047 * 2 : header = On input, pointer to header to modify.
4048 * On output, pointer to the modified header, or NULL
4049 * to remove the header. This function frees the
4050 * original string if necessary.
4052 * Returns : JB_ERR_OK on success, or
4053 * JB_ERR_MEMORY on out-of-memory error.
4055 *********************************************************************/
4056 static jb_err server_set_cookie(struct client_state *csp, char **header)
4063 if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
4065 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
4068 else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
4070 /* Flag whether or not to log a message */
4073 /* A variable to store the tag we're working on */
4076 /* Skip "Set-Cookie:" (11 characters) in header */
4077 cur_tag = *header + 11;
4079 /* skip whitespace between "Set-Cookie:" and value */
4080 while (*cur_tag && ijb_isspace(*cur_tag))
4085 /* Loop through each tag in the cookie */
4089 char *next_tag = strchr(cur_tag, ';');
4090 if (next_tag != NULL)
4092 /* Skip the ';' character itself */
4095 /* skip whitespace ";" and start of tag */
4096 while (*next_tag && ijb_isspace(*next_tag))
4103 /* "Next tag" is the end of the string */
4104 next_tag = cur_tag + strlen(cur_tag);
4108 * Check the expiration date to see
4109 * if the cookie is still valid, if yes,
4110 * rewrite it to a session cookie.
4112 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
4114 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
4116 /* Did we detect the date properly? */
4117 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
4120 * Nope, treat it as if it was still valid.
4122 * XXX: Should we remove the whole cookie instead?
4124 log_error(LOG_LEVEL_ERROR,
4125 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
4126 string_move(cur_tag, next_tag);
4132 * Yes. Check if the cookie is still valid.
4134 * If the cookie is already expired it's probably
4135 * a delete cookie and even if it isn't, the browser
4136 * will discard it anyway.
4140 * XXX: timegm() isn't available on some AmigaOS
4141 * versions and our replacement doesn't work.
4143 * Our options are to either:
4145 * - disable session-cookies-only completely if timegm
4148 * - to simply remove all expired tags, like it has
4149 * been done until Privoxy 3.0.6 and to live with
4150 * the consequence that it can cause login/logout
4151 * problems on servers that don't validate their
4152 * input properly, or
4154 * - to replace it with mktime in which
4155 * case there is a slight chance of valid cookies
4156 * passing as already expired.
4158 * This is the way it's currently done and it's not
4159 * as bad as it sounds. If the missing GMT offset is
4160 * enough to change the result of the expiration check
4161 * the cookie will be only valid for a few hours
4162 * anyway, which in many cases will be shorter
4163 * than a browser session.
4165 if (cookie_time - now < 0)
4167 log_error(LOG_LEVEL_HEADER,
4168 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4169 /* Just in case some clown sets more then one expiration date */
4175 * Still valid, delete expiration date by copying
4176 * the rest of the string over it.
4178 string_move(cur_tag, next_tag);
4180 /* That changed the header, need to issue a log message */
4184 * Note that the next tag has now been moved to *cur_tag,
4185 * so we do not need to update the cur_tag pointer.
4193 /* Move on to next cookie tag */
4200 assert(NULL != *header);
4201 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
4210 #ifdef FEATURE_FORCE_LOAD
4211 /*********************************************************************
4213 * Function : strclean
4215 * Description : In-Situ-Eliminate all occurances of substring in
4219 * 1 : string = string to clean
4220 * 2 : substring = substring to eliminate
4222 * Returns : Number of eliminations
4224 *********************************************************************/
4225 int strclean(char *string, const char *substring)
4231 len = strlen(substring);
4233 while((pos = strstr(string, substring)) != NULL)
4240 while (*p++ != '\0');
4247 #endif /* def FEATURE_FORCE_LOAD */
4250 /*********************************************************************
4252 * Function : parse_header_time
4254 * Description : Parses time formats used in HTTP header strings
4255 * to get the numerical respresentation.
4258 * 1 : header_time = HTTP header time as string.
4259 * 2 : result = storage for header_time in seconds
4261 * Returns : JB_ERR_OK if the time format was recognized, or
4262 * JB_ERR_PARSE otherwise.
4264 *********************************************************************/
4265 static jb_err parse_header_time(const char *header_time, time_t *result)
4270 * Zero out gmt to prevent time zone offsets.
4272 * While this is only necessary on some platforms
4273 * (mingw32 for example), I don't know how to
4274 * detect these automatically and doing it everywhere
4277 memset(&gmt, 0, sizeof(gmt));
4279 /* Tue, 02 Jun 2037 20:00:00 */
4280 if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
4281 /* Tue, 02-Jun-2037 20:00:00 */
4282 && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
4283 /* Tue, 02-Jun-37 20:00:00 */
4284 && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
4285 /* Tuesday, 02-Jun-2037 20:00:00 */
4286 && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
4287 /* Tuesday Jun 02 20:00:00 2037 */
4288 && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
4290 return JB_ERR_PARSE;
4293 *result = timegm(&gmt);
4300 /*********************************************************************
4302 * Function : get_destination_from_headers
4304 * Description : Parse the "Host:" header to get the request's destination.
4305 * Only needed if the client's request was forcefully
4306 * redirected into Privoxy.
4308 * Code mainly copied from client_host() which is currently
4309 * run too late for this purpose.
4312 * 1 : headers = List of headers (one of them hopefully being
4313 * the "Host:" header)
4314 * 2 : http = storage for the result (host, port and hostport).
4316 * Returns : JB_ERR_MEMORY in case of memory problems,
4317 * JB_ERR_PARSE if the host header couldn't be found,
4318 * JB_ERR_OK otherwise.
4320 *********************************************************************/
4321 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4327 host = get_header_value(headers, "Host:");
4331 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4332 return JB_ERR_PARSE;
4335 if (NULL == (p = strdup((host))))
4337 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4338 return JB_ERR_MEMORY;
4341 if (NULL == (q = strdup(p)))
4344 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4345 return JB_ERR_MEMORY;
4348 freez(http->hostport);
4352 q = strchr(http->host, ':');
4355 /* Terminate hostname and evaluate port string */
4357 http->port = atoi(q);
4361 http->port = http->ssl ? 443 : 80;
4364 /* Rebuild request URL */
4366 http->url = strdup(http->ssl ? "https://" : "http://");
4367 string_append(&http->url, http->hostport);
4368 string_append(&http->url, http->path);
4369 if (http->url == NULL)
4371 return JB_ERR_MEMORY;
4374 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4382 /*********************************************************************
4384 * Function : create_forged_referrer
4386 * Description : Helper for client_referrer to forge a referer as
4387 * 'http://[hostname:port/' to fool stupid
4388 * checks for in-site links
4391 * 1 : header = Pointer to header pointer
4392 * 2 : hostport = Host and optionally port as string
4394 * Returns : JB_ERR_OK in case of success, or
4395 * JB_ERR_MEMORY in case of memory problems.
4397 *********************************************************************/
4398 static jb_err create_forged_referrer(char **header, const char *hostport)
4400 assert(NULL == *header);
4402 *header = strdup("Referer: http://");
4403 string_append(header, hostport);
4404 string_append(header, "/");
4406 if (NULL == *header)
4408 return JB_ERR_MEMORY;
4411 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4418 /*********************************************************************
4420 * Function : create_fake_referrer
4422 * Description : Helper for client_referrer to create a fake referrer
4423 * based on a string supplied by the user.
4426 * 1 : header = Pointer to header pointer
4427 * 2 : hosthost = Referrer to fake
4429 * Returns : JB_ERR_OK in case of success, or
4430 * JB_ERR_MEMORY in case of memory problems.
4432 *********************************************************************/
4433 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4435 assert(NULL == *header);
4437 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4439 log_error(LOG_LEVEL_HEADER,
4440 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4442 *header = strdup("Referer: ");
4443 string_append(header, fake_referrer);
4445 if (NULL == *header)
4447 return JB_ERR_MEMORY;
4450 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4457 /*********************************************************************
4459 * Function : handle_conditional_hide_referrer_parameter
4461 * Description : Helper for client_referrer to crunch or forge
4462 * the referrer header if the host has changed.
4465 * 1 : header = Pointer to header pointer
4466 * 2 : host = The target host (may include the port)
4467 * 3 : parameter_conditional_block = Boolean to signal
4468 * if we're in conditional-block mode. If not set,
4469 * we're in conditional-forge mode.
4471 * Returns : JB_ERR_OK in case of success, or
4472 * JB_ERR_MEMORY in case of memory problems.
4474 *********************************************************************/
4475 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4476 const char *host, const int parameter_conditional_block)
4478 char *referer = strdup(*header);
4479 const size_t hostlenght = strlen(host);
4481 if (NULL == referer)
4484 return JB_ERR_MEMORY;
4487 /* referer begins with 'Referer: http[s]://' */
4488 if (hostlenght < (strlen(referer)-17))
4491 * Shorten referer to make sure the referer is blocked
4492 * if www.example.org/www.example.com-shall-see-the-referer/
4493 * links to www.example.com/
4495 referer[hostlenght+17] = '\0';
4497 if (NULL == strstr(referer, host))
4499 /* Host has changed */
4500 if (parameter_conditional_block)
4502 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4509 return create_forged_referrer(header, host);