1 const char parsers_rcs[] = "$Id: parsers.c,v 1.144 2008/09/21 13:59:33 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 1.144 2008/09/21 13:59:33 fabiankeil
48 * Treat unknown change-x-forwarded-for parameters as fatal errors.
50 * Revision 1.143 2008/09/21 13:36:52 fabiankeil
51 * If change-x-forwarded-for{add} is used and the client
52 * sends multiple X-Forwarded-For headers, append the client's
53 * IP address to each one of them. "Traditionally" we would
54 * lose all but the last one.
56 * Revision 1.142 2008/09/20 10:04:33 fabiankeil
57 * Remove hide-forwarded-for-headers action which has
58 * been obsoleted by change-x-forwarded-for{block}.
60 * Revision 1.141 2008/09/19 15:26:28 fabiankeil
61 * Add change-x-forwarded-for{} action to block or add
62 * X-Forwarded-For headers. Mostly based on code removed
65 * Revision 1.140 2008/09/12 17:51:43 fabiankeil
66 * - A few style fixes.
67 * - Remove a pointless cast.
69 * Revision 1.139 2008/09/04 08:13:58 fabiankeil
70 * Prepare for critical sections on Windows by adding a
71 * layer of indirection before the pthread mutex functions.
73 * Revision 1.138 2008/08/30 12:03:07 fabiankeil
74 * Remove FEATURE_COOKIE_JAR.
76 * Revision 1.137 2008/05/30 15:50:08 fabiankeil
77 * Remove questionable micro-optimizations
78 * whose usefulness has never been measured.
80 * Revision 1.136 2008/05/26 16:02:24 fabiankeil
81 * s@Insufficent@Insufficient@
83 * Revision 1.135 2008/05/21 20:12:10 fabiankeil
84 * The whole point of strclean() is to modify the
85 * first parameter, so don't mark it immutable,
86 * even though the compiler lets us get away with it.
88 * Revision 1.134 2008/05/21 19:27:25 fabiankeil
89 * As the wafer actions are gone, we can stop including encode.h.
91 * Revision 1.133 2008/05/21 15:50:47 fabiankeil
92 * Ditch cast from (char **) to (char **).
94 * Revision 1.132 2008/05/21 15:47:14 fabiankeil
95 * Streamline sed()'s prototype and declare
96 * the header parse and add structures static.
98 * Revision 1.131 2008/05/20 20:13:30 fabiankeil
99 * Factor update_server_headers() out of sed(), ditch the
100 * first_run hack and make server_patterns_light static.
102 * Revision 1.130 2008/05/19 17:18:04 fabiankeil
103 * Wrap memmove() calls in string_move()
104 * to document the purpose in one place.
106 * Revision 1.129 2008/05/17 14:02:07 fabiankeil
107 * Normalize linear header white space.
109 * Revision 1.128 2008/05/16 16:39:03 fabiankeil
110 * If a header is split across multiple lines,
111 * merge them to a single line before parsing them.
113 * Revision 1.127 2008/05/10 13:23:38 fabiankeil
114 * Don't provide get_header() with the whole client state
115 * structure when it only needs access to csp->iob.
117 * Revision 1.126 2008/05/03 16:40:45 fabiankeil
118 * Change content_filters_enabled()'s parameter from
119 * csp->action to action so it can be also used in the
120 * CGI code. Don't bother checking if there are filters
121 * loaded, as that's somewhat besides the point.
123 * Revision 1.125 2008/04/17 14:40:49 fabiankeil
124 * Provide get_http_time() with the buffer size so it doesn't
125 * have to blindly assume that the buffer is big enough.
127 * Revision 1.124 2008/04/16 16:38:21 fabiankeil
128 * Don't pass the whole csp structure to flush_socket()
129 * when it only needs a file descriptor and a buffer.
131 * Revision 1.123 2008/03/29 12:13:46 fabiankeil
132 * Remove send-wafer and send-vanilla-wafer actions.
134 * Revision 1.122 2008/03/28 15:13:39 fabiankeil
135 * Remove inspect-jpegs action.
137 * Revision 1.121 2008/01/05 21:37:03 fabiankeil
138 * Let client_range() also handle Request-Range headers
139 * which apparently are still supported by many servers.
141 * Revision 1.120 2008/01/04 17:43:45 fabiankeil
142 * Improve the warning messages that get logged if the action files
143 * "enable" filters but no filters of that type have been loaded.
145 * Revision 1.119 2007/12/28 18:32:51 fabiankeil
146 * In server_content_type():
147 * - Don't require leading white space when detecting image content types.
148 * - Change '... not replaced ...' message to sound less crazy if the text
149 * type actually is 'text/plain'.
150 * - Mark the 'text/plain == binary data' assumption for removal.
151 * - Remove a bunch of trailing white space.
153 * Revision 1.118 2007/12/28 16:56:35 fabiankeil
154 * Minor server_content_disposition() changes:
155 * - Don't regenerate the header name all lower-case.
156 * - Some white space fixes.
157 * - Remove useless log message in case of ENOMEM.
159 * Revision 1.117 2007/12/06 18:11:50 fabiankeil
160 * Garbage-collect the code to add a X-Forwarded-For
161 * header as it seems to be mostly used by accident.
163 * Revision 1.116 2007/12/01 13:04:22 fabiankeil
164 * Fix a crash on mingw32 with some Last Modified times in the future.
166 * Revision 1.115 2007/11/02 16:52:50 fabiankeil
167 * Remove a "can't happen" error block which, over
168 * time, mutated into a "guaranteed to happen" block.
170 * Revision 1.114 2007/10/19 16:56:26 fabiankeil
171 * - Downgrade "Buffer limit reached" message to LOG_LEVEL_INFO.
172 * - Use shiny new content_filters_enabled() in client_range().
174 * Revision 1.113 2007/10/10 17:29:57 fabiankeil
175 * I forgot about Poland.
177 * Revision 1.112 2007/10/09 16:38:40 fabiankeil
178 * Remove Range and If-Range headers if content filtering is enabled.
180 * Revision 1.111 2007/10/04 18:07:00 fabiankeil
181 * Move ACTION_VANILLA_WAFER handling from jcc's chat() into
182 * client_cookie_adder() to make sure send-vanilla-wafer can be
183 * controlled through tags (and thus regression-tested).
185 * Revision 1.110 2007/09/29 10:42:37 fabiankeil
186 * - Remove "scanning headers for" log message again.
187 * - Some more whitespace fixes.
189 * Revision 1.109 2007/09/08 14:25:48 fabiankeil
190 * Refactor client_referrer() and add conditional-forge parameter.
192 * Revision 1.108 2007/08/28 18:21:03 fabiankeil
193 * A bunch of whitespace fixes, pointy hat to me.
195 * Revision 1.107 2007/08/28 18:16:32 fabiankeil
196 * Fix possible memory corruption in server_http, make sure it's not
197 * executed for ordinary server headers and mark some problems for later.
199 * Revision 1.106 2007/08/18 14:30:32 fabiankeil
200 * Let content-type-overwrite{} honour force-text-mode again.
202 * Revision 1.105 2007/08/11 14:49:49 fabiankeil
203 * - Add prototpyes for the header parsers and make them static.
204 * - Comment out client_accept_encoding_adder() which isn't used right now.
206 * Revision 1.104 2007/07/14 07:38:19 fabiankeil
207 * Move the ACTION_FORCE_TEXT_MODE check out of
208 * server_content_type(). Signal other functions
209 * whether or not a content type has been declared.
210 * Part of the fix for BR#1750917.
212 * Revision 1.103 2007/06/01 16:31:54 fabiankeil
213 * Change sed() to return a jb_err in preparation for forward-override{}.
215 * Revision 1.102 2007/05/27 12:39:32 fabiankeil
216 * Adjust "X-Filter: No" to disable dedicated header filters.
218 * Revision 1.101 2007/05/14 10:16:41 fabiankeil
219 * Streamline client_cookie_adder().
221 * Revision 1.100 2007/04/30 15:53:11 fabiankeil
222 * Make sure filters with dynamic jobs actually use them.
224 * Revision 1.99 2007/04/30 15:06:26 fabiankeil
225 * - Introduce dynamic pcrs jobs that can resolve variables.
226 * - Remove unnecessary update_action_bits_for_all_tags() call.
228 * Revision 1.98 2007/04/17 18:32:10 fabiankeil
229 * - Make tagging based on tags set by earlier taggers
230 * of the same kind possible.
231 * - Log whether or not new tags cause action bits updates
232 * (in which case a matching tag-pattern section exists).
233 * - Log if the user tries to set a tag that is already set.
235 * Revision 1.97 2007/04/15 16:39:21 fabiankeil
236 * Introduce tags as alternative way to specify which
237 * actions apply to a request. At the moment tags can be
238 * created based on client and server headers.
240 * Revision 1.96 2007/04/12 12:53:58 fabiankeil
241 * Log a warning if the content is compressed, filtering is
242 * enabled and Privoxy was compiled without zlib support.
245 * Revision 1.95 2007/03/25 14:26:40 fabiankeil
246 * - Fix warnings when compiled with glibc.
247 * - Don't use crumble() for cookie crunching.
248 * - Move cookie time parsing into parse_header_time().
249 * - Let parse_header_time() return a jb_err code
250 * instead of a pointer that can only be used to
251 * check for NULL anyway.
253 * Revision 1.94 2007/03/21 12:23:53 fabiankeil
254 * - Add better protection against malicious gzip headers.
255 * - Stop logging the first hundred bytes of decompressed content.
256 * It looks like it's working and there is always debug 16.
257 * - Log the content size after decompression in decompress_iob()
258 * instead of pcrs_filter_response().
260 * Revision 1.93 2007/03/20 15:21:44 fabiankeil
261 * - Use dedicated header filter actions instead of abusing "filter".
262 * Replace "filter-client-headers" and "filter-client-headers"
263 * with "server-header-filter" and "client-header-filter".
264 * - Remove filter_client_header() and filter_client_header(),
265 * filter_header() now checks the shiny new
266 * CSP_FLAG_CLIENT_HEADER_PARSING_DONE flag instead.
268 * Revision 1.92 2007/03/05 13:25:32 fabiankeil
269 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
270 * - Handle "Cookie:" and "Connection:" headers a bit smarter
271 * (don't crunch them just to recreate them later on).
272 * - Add another non-standard time format for the cookie
273 * expiration date detection.
274 * - Fix a valgrind warning.
276 * Revision 1.91 2007/02/24 12:27:32 fabiankeil
277 * Improve cookie expiration date detection.
279 * Revision 1.90 2007/02/08 19:12:35 fabiankeil
280 * Don't run server_content_length() the first time
281 * sed() parses server headers; only adjust the
282 * Content-Length header if the page was modified.
284 * Revision 1.89 2007/02/07 16:52:11 fabiankeil
285 * Fix log messages regarding the cookie time format
286 * (cookie and request URL were mixed up).
288 * Revision 1.88 2007/02/07 11:27:12 fabiankeil
289 * - Let decompress_iob()
290 * - not corrupt the content if decompression fails
291 * early. (the first byte(s) were lost).
292 * - use pointer arithmetics with defined outcome for
294 * - Use a different kludge to remember a failed decompression.
296 * Revision 1.87 2007/01/31 16:21:38 fabiankeil
297 * Search for Max-Forwards headers case-insensitive,
298 * don't generate the "501 unsupported" message for invalid
299 * Max-Forwards values and don't increase negative ones.
301 * Revision 1.86 2007/01/30 13:05:26 fabiankeil
302 * - Let server_set_cookie() check the expiration date
303 * of cookies and don't touch the ones that are already
304 * expired. Fixes problems with low quality web applications
305 * as described in BR 932612.
307 * - Adjust comment in client_max_forwards to reality;
308 * remove invalid Max-Forwards headers.
310 * Revision 1.85 2007/01/26 15:33:46 fabiankeil
311 * Stop filter_header() from unintentionally removing
312 * empty header lines that were enlisted by the continue
315 * Revision 1.84 2007/01/24 12:56:52 fabiankeil
316 * - Repeat the request URL before logging any headers.
317 * Makes reading the log easier in case of simultaneous requests.
318 * - If there are more than one Content-Type headers in one request,
319 * use the first one and remove the others.
320 * - Remove "newval" variable in server_content_type().
321 * It's only used once.
323 * Revision 1.83 2007/01/12 15:03:02 fabiankeil
324 * Correct a cast, check inflateEnd() exit code
325 * to see if we have to, replace sprintf calls
328 * Revision 1.82 2007/01/01 19:36:37 fabiankeil
329 * Integrate a modified version of Wil Mahan's
330 * zlib patch (PR #895531).
332 * Revision 1.81 2006/12/31 22:21:33 fabiankeil
333 * Skip empty filter files in filter_header()
334 * but don't ignore the ones that come afterwards.
335 * Fixes BR 1619208, this time for real.
337 * Revision 1.80 2006/12/29 19:08:22 fabiankeil
338 * Reverted parts of my last commit
339 * to keep error handling working.
341 * Revision 1.79 2006/12/29 18:04:40 fabiankeil
342 * Fixed gcc43 conversion warnings.
344 * Revision 1.78 2006/12/26 17:19:20 fabiankeil
345 * Bringing back the "useless" localtime() call
346 * I removed in revision 1.67. On some platforms
347 * it's necessary to prevent time zone offsets.
349 * Revision 1.77 2006/12/07 18:44:26 fabiankeil
350 * Rebuild request URL in get_destination_from_headers()
351 * to make sure redirect{pcrs command} works as expected
352 * for intercepted requests.
354 * Revision 1.76 2006/12/06 19:52:25 fabiankeil
355 * Added get_destination_from_headers().
357 * Revision 1.75 2006/11/13 19:05:51 fabiankeil
358 * Make pthread mutex locking more generic. Instead of
359 * checking for OSX and OpenBSD, check for FEATURE_PTHREAD
360 * and use mutex locking unless there is an _r function
361 * available. Better safe than sorry.
363 * Fixes "./configure --disable-pthread" and should result
364 * in less threading-related problems on pthread-using platforms,
365 * but it still doesn't fix BR#1122404.
367 * Revision 1.74 2006/10/02 16:59:12 fabiankeil
368 * The special header "X-Filter: No" now disables
369 * header filtering as well.
371 * Revision 1.73 2006/09/23 13:26:38 roro
372 * Replace TABs by spaces in source code.
374 * Revision 1.72 2006/09/23 12:37:21 fabiankeil
375 * Don't print a log message every time filter_headers is
376 * entered or left. It only creates noise without any real
379 * Revision 1.71 2006/09/21 19:55:17 fabiankeil
380 * Fix +hide-if-modified-since{-n}.
382 * Revision 1.70 2006/09/08 12:06:34 fabiankeil
383 * Have hide-if-modified-since interpret the random
384 * range value as minutes instead of hours. Allows
385 * more fine-grained configuration.
387 * Revision 1.69 2006/09/06 16:25:51 fabiankeil
388 * Always have parse_header_time return a pointer
389 * that actual makes sense, even though we currently
390 * only need it to detect problems.
392 * Revision 1.68 2006/09/06 10:43:32 fabiankeil
393 * Added config option enable-remote-http-toggle
394 * to specify if Privoxy should recognize special
395 * headers (currently only X-Filter) to change its
396 * behaviour. Disabled by default.
398 * Revision 1.67 2006/09/04 11:01:26 fabiankeil
399 * After filtering de-chunked instances, remove
400 * "Transfer-Encoding" header entirely instead of changing
401 * it to "Transfer-Encoding: identity", which is invalid.
402 * Thanks Michael Shields <shields@msrl.com>. Fixes PR 1318658.
404 * Don't use localtime in parse_header_time. An empty time struct
405 * is good enough, it gets overwritten by strptime anyway.
407 * Revision 1.66 2006/09/03 19:38:28 fabiankeil
408 * Use gmtime_r if available, fallback to gmtime with mutex
409 * protection for MacOSX and use vanilla gmtime for the rest.
411 * Revision 1.65 2006/08/22 10:55:56 fabiankeil
412 * Changed client_referrer to use the right type (size_t) for
413 * hostlenght and to shorten the temporary referrer string with
414 * '\0' instead of adding a useless line break.
416 * Revision 1.64 2006/08/17 17:15:10 fabiankeil
417 * - Back to timegm() using GnuPG's replacement if necessary.
418 * Using mktime() and localtime() could add a on hour offset if
419 * the randomize factor was big enough to lead to a summer/wintertime
422 * - Removed now-useless Privoxy 3.0.3 compatibility glue.
424 * - Moved randomization code into pick_from_range().
426 * - Changed parse_header_time definition.
427 * time_t isn't guaranteed to be signed and
428 * if it isn't, -1 isn't available as error code.
429 * Changed some variable types in client_if_modified_since()
430 * because of the same reason.
432 * Revision 1.63 2006/08/14 13:18:08 david__schmidt
433 * OS/2 compilation compatibility fixups
435 * Revision 1.62 2006/08/14 08:58:42 fabiankeil
436 * Changed include from strptime.c to strptime.h
438 * Revision 1.61 2006/08/14 08:25:19 fabiankeil
439 * Split filter-headers{} into filter-client-headers{}
440 * and filter-server-headers{}.
441 * Added parse_header_time() to share some code.
442 * Replaced timegm() with mktime().
444 * Revision 1.60 2006/08/12 03:54:37 david__schmidt
445 * Windows service integration
447 * Revision 1.59 2006/08/03 02:46:41 david__schmidt
448 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
450 * Revision 1.58 2006/07/18 14:48:47 david__schmidt
451 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
452 * with what was really the latest development (the v_3_0_branch branch)
454 * Revision 1.56.2.10 2006/01/21 16:16:08 david__schmidt
455 * Thanks to Edward Carrel for his patch to modernize OSX's
\rpthreads support. See bug #1409623.
457 * Revision 1.56.2.9 2004/10/03 12:53:45 david__schmidt
458 * Add the ability to check jpeg images for invalid
459 * lengths of comment blocks. Defensive strategy
460 * against the exploit:
461 * Microsoft Security Bulletin MS04-028
462 * Buffer Overrun in JPEG Processing (GDI+) Could
463 * Allow Code Execution (833987)
464 * Enabled with +inspect-jpegs in actions files.
466 * Revision 1.56.2.8 2003/07/11 13:21:25 oes
467 * Excluded text/plain objects from filtering. This fixes a
468 * couple of client-crashing, download corruption and
469 * Privoxy performance issues, whose root cause lies in
470 * web servers labelling content of unknown type as text/plain.
472 * Revision 1.56.2.7 2003/05/06 12:07:26 oes
473 * Fixed bug #729900: Suspicious HOST: headers are now killed and regenerated if necessary
475 * Revision 1.56.2.6 2003/04/14 21:28:30 oes
476 * Completing the previous change
478 * Revision 1.56.2.5 2003/04/14 12:08:16 oes
479 * Added temporary workaround for bug in PHP < 4.2.3
481 * Revision 1.56.2.4 2003/03/07 03:41:05 david__schmidt
482 * 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.
484 * Revision 1.56.2.3 2002/11/10 04:20:02 hal9
485 * Fix typo: supressed -> suppressed
487 * Revision 1.56.2.2 2002/09/25 14:59:53 oes
488 * Improved cookie logging
490 * Revision 1.56.2.1 2002/09/25 14:52:45 oes
491 * Added basic support for OPTIONS and TRACE HTTP methods:
492 * - New parser function client_max_forwards which decrements
493 * the Max-Forwards HTTP header field of OPTIONS and TRACE
494 * requests by one before forwarding
495 * - New parser function client_host which extracts the host
496 * and port information from the HTTP header field if the
497 * request URI was not absolute
498 * - Don't crumble and re-add the Host: header, but only generate
499 * and append if missing
501 * Revision 1.56 2002/05/12 15:34:22 jongfoster
502 * Fixing typo in a comment
504 * Revision 1.55 2002/05/08 16:01:07 oes
505 * Optimized add_to_iob:
506 * - Use realloc instead of malloc(), memcpy(), free()
507 * - Expand to powers of two if possible, to get
508 * O(log n) reallocs instead of O(n).
509 * - Moved check for buffer limit here from chat
510 * - Report failure via returncode
512 * Revision 1.54 2002/04/02 15:03:16 oes
513 * Tiny code cosmetics
515 * Revision 1.53 2002/03/26 22:29:55 swa
516 * we have a new homepage!
518 * Revision 1.52 2002/03/24 13:25:43 swa
519 * name change related issues
521 * Revision 1.51 2002/03/13 00:27:05 jongfoster
524 * Revision 1.50 2002/03/12 01:45:35 oes
525 * More verbose logging
527 * Revision 1.49 2002/03/09 20:03:52 jongfoster
528 * - Making various functions return int rather than size_t.
529 * (Undoing a recent change). Since size_t is unsigned on
530 * Windows, functions like read_socket that return -1 on
531 * error cannot return a size_t.
533 * THIS WAS A MAJOR BUG - it caused frequent, unpredictable
534 * crashes, and also frequently caused JB to jump to 100%
535 * CPU and stay there. (Because it thought it had just
536 * read ((unsigned)-1) == 4Gb of data...)
538 * - The signature of write_socket has changed, it now simply
539 * returns success=0/failure=nonzero.
541 * - Trying to get rid of a few warnings --with-debug on
542 * Windows, I've introduced a new type "jb_socket". This is
543 * used for the socket file descriptors. On Windows, this
544 * is SOCKET (a typedef for unsigned). Everywhere else, it's
545 * an int. The error value can't be -1 any more, so it's
546 * now JB_INVALID_SOCKET (which is -1 on UNIX, and in
547 * Windows it maps to the #define INVALID_SOCKET.)
549 * - The signature of bind_port has changed.
551 * Revision 1.48 2002/03/07 03:46:53 oes
552 * Fixed compiler warnings etc
554 * Revision 1.47 2002/02/20 23:15:13 jongfoster
555 * Parsing functions now handle out-of-memory gracefully by returning
558 * Revision 1.46 2002/01/17 21:03:47 jongfoster
559 * Moving all our URL and URL pattern parsing code to urlmatch.c.
561 * Revision 1.45 2002/01/09 14:33:03 oes
562 * Added support for localtime_r.
564 * Revision 1.44 2001/12/14 01:22:54 steudten
565 * Remove 'user:pass@' from 'proto://user:pass@host' for the
566 * new added header 'Host: ..'. (See Req ID 491818)
568 * Revision 1.43 2001/11/23 00:26:38 jongfoster
569 * Fixing two really stupid errors in my previous commit
571 * Revision 1.42 2001/11/22 21:59:30 jongfoster
572 * Adding code to handle +no-cookies-keep
574 * Revision 1.41 2001/11/05 23:43:05 steudten
575 * Add time+date to log files.
577 * Revision 1.40 2001/10/26 20:13:09 jongfoster
578 * ctype.h is needed in Windows, too.
580 * Revision 1.39 2001/10/26 17:40:04 oes
581 * Introduced get_header_value()
582 * Removed http->user_agent, csp->referrer and csp->accept_types
583 * Removed client_accept()
585 * Revision 1.38 2001/10/25 03:40:48 david__schmidt
586 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
587 * threads to call select() simultaneously. So, it's time to do a real, live,
588 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
589 * (native). Both versions will work, but using __OS2__ offers multi-threading.
591 * Revision 1.37 2001/10/23 21:36:02 jongfoster
592 * Documenting sed()'s error behaviou (doc change only)
594 * Revision 1.36 2001/10/13 12:51:51 joergs
595 * Removed client_host, (was only required for the old 2.0.2-11 http://noijb.
596 * force-load), instead crumble Host: and add it (again) in client_host_adder
597 * (in case we get a HTTP/1.0 request without Host: header and forward it to
598 * a HTTP/1.1 server/proxy).
600 * Revision 1.35 2001/10/09 22:39:21 jongfoster
601 * assert.h is also required under Win32, so moving out of #ifndef _WIN32
604 * Revision 1.34 2001/10/07 18:50:55 oes
605 * Added server_content_encoding, renamed server_transfer_encoding
607 * Revision 1.33 2001/10/07 18:04:49 oes
608 * Changed server_http11 to server_http and its pattern to "HTTP".
609 * Additional functionality: it now saves the HTTP status into
610 * csp->http->status and sets CT_TABOO for Status 206 (partial range)
612 * Revision 1.32 2001/10/07 15:43:28 oes
613 * Removed FEATURE_DENY_GZIP and replaced it with client_accept_encoding,
614 * client_te and client_accept_encoding_adder, triggered by the new
615 * +no-compression action. For HTTP/1.1 the Accept-Encoding header is
616 * changed to allow only identity and chunked, and the TE header is
617 * crunched. For HTTP/1.0, Accept-Encoding is crunched.
619 * parse_http_request no longer does anything than parsing. The rewriting
620 * of http->cmd and version mangling are gone. It now also recognizes
621 * the put and delete methods and saves the url in http->url. Removed
624 * renamed content_type and content_length to have the server_ prefix
626 * server_content_type now only works if csp->content_type != CT_TABOO
628 * added server_transfer_encoding, which
629 * - Sets CT_TABOO to prohibit filtering if encoding compresses
630 * - Raises the CSP_FLAG_CHUNKED flag if Encoding is "chunked"
631 * - Change from "chunked" to "identity" if body was chunked
632 * but has been de-chunked for filtering.
634 * added server_content_md5 which crunches any Content-MD5 headers
635 * if the body was modified.
637 * made server_http11 conditional on +downgrade action
639 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
641 * Revision 1.31 2001/10/05 14:25:02 oes
642 * Crumble Keep-Alive from Server
644 * Revision 1.30 2001/09/29 12:56:03 joergs
645 * IJB now changes HTTP/1.1 to HTTP/1.0 in requests and answers.
647 * Revision 1.29 2001/09/24 21:09:24 jongfoster
648 * Fixing 2 memory leaks that Guy spotted, where the paramater to
649 * enlist() was not being free()d.
651 * Revision 1.28 2001/09/22 16:32:28 jongfoster
652 * Removing unused #includes.
654 * Revision 1.27 2001/09/20 15:45:25 steudten
656 * add casting from size_t to int for printf()
657 * remove local variable shadow s2
659 * Revision 1.26 2001/09/16 17:05:14 jongfoster
660 * Removing unused #include showarg.h
662 * Revision 1.25 2001/09/16 13:21:27 jongfoster
663 * Changes to use new list functions.
665 * Revision 1.24 2001/09/13 23:05:50 jongfoster
666 * Changing the string paramater to the header parsers a "const".
668 * Revision 1.23 2001/09/12 18:08:19 steudten
670 * In parse_http_request() header rewriting miss the host value, so
671 * from http://www.mydomain.com the result was just " / " not
672 * http://www.mydomain.com/ in case we forward.
674 * Revision 1.22 2001/09/10 10:58:53 oes
675 * Silenced compiler warnings
677 * Revision 1.21 2001/07/31 14:46:00 oes
678 * - Persistant connections now suppressed
679 * - sed() no longer appends empty header to csp->headers
681 * Revision 1.20 2001/07/30 22:08:36 jongfoster
682 * Tidying up #defines:
683 * - All feature #defines are now of the form FEATURE_xxx
684 * - Permanently turned off WIN_GUI_EDIT
685 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
687 * Revision 1.19 2001/07/25 17:21:54 oes
688 * client_uagent now saves copy of User-Agent: header value
690 * Revision 1.18 2001/07/13 14:02:46 oes
691 * - Included fix to repair broken HTTP requests that
692 * don't contain a path, not even '/'.
693 * - Removed all #ifdef PCRS
694 * - content_type now always inspected and classified as
695 * text, gif or other.
696 * - formatting / comments
698 * Revision 1.17 2001/06/29 21:45:41 oes
699 * Indentation, CRLF->LF, Tab-> Space
701 * Revision 1.16 2001/06/29 13:32:42 oes
703 * - Adapted free_http_request
704 * - Removed logentry from cancelled commit
706 * Revision 1.15 2001/06/03 19:12:38 oes
707 * deleted const struct interceptors
709 * Revision 1.14 2001/06/01 18:49:17 jongfoster
710 * Replaced "list_share" with "list" - the tiny memory gain was not
711 * worth the extra complexity.
713 * Revision 1.13 2001/05/31 21:30:33 jongfoster
714 * Removed list code - it's now in list.[ch]
715 * Renamed "permission" to "action", and changed many features
716 * to use the actions file rather than the global config.
718 * Revision 1.12 2001/05/31 17:33:13 oes
722 * Revision 1.11 2001/05/29 20:11:19 joergs
723 * '/ * inside comment' warning removed.
725 * Revision 1.10 2001/05/29 09:50:24 jongfoster
726 * Unified blocklist/imagelist/permissionslist.
727 * File format is still under discussion, but the internal changes
730 * Also modified interceptor behaviour:
731 * - We now intercept all URLs beginning with one of the following
732 * prefixes (and *only* these prefixes):
734 * * http://ijbswa.sf.net/config/
735 * * http://ijbswa.sourceforge.net/config/
736 * - New interceptors "home page" - go to http://i.j.b/ to see it.
737 * - Internal changes so that intercepted and fast redirect pages
738 * are not replaced with an image.
739 * - Interceptors now have the option to send a binary page direct
740 * to the client. (i.e. ijb-send-banner uses this)
741 * - Implemented show-url-info interceptor. (Which is why I needed
742 * the above interceptors changes - a typical URL is
743 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
744 * The previous mechanism would not have intercepted that, and
745 * if it had been intercepted then it then it would have replaced
748 * Revision 1.9 2001/05/28 17:26:33 jongfoster
749 * Fixing segfault if last header was crunched.
750 * Fixing Windows build (snprintf() is _snprintf() under Win32, but we
751 * can use the cross-platform sprintf() instead.)
753 * Revision 1.8 2001/05/27 22:17:04 oes
755 * - re_process_buffer no longer writes the modified buffer
756 * to the client, which was very ugly. It now returns the
757 * buffer, which it is then written by chat.
759 * - content_length now adjusts the Content-Length: header
760 * for modified documents rather than crunch()ing it.
761 * (Length info in csp->content_length, which is 0 for
762 * unmodified documents)
764 * - For this to work, sed() is called twice when filtering.
766 * Revision 1.7 2001/05/27 13:19:06 oes
767 * Patched Joergs solution for the content-length in.
769 * Revision 1.6 2001/05/26 13:39:32 jongfoster
770 * Only crunches Content-Length header if applying RE filtering.
771 * Without this fix, Microsoft Windows Update wouldn't work.
773 * Revision 1.5 2001/05/26 00:28:36 jongfoster
774 * Automatic reloading of config file.
775 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
776 * Most of the global variables have been moved to a new
777 * struct configuration_spec, accessed through csp->config->globalname
778 * Most of the globals remaining are used by the Win32 GUI.
780 * Revision 1.4 2001/05/22 18:46:04 oes
782 * - Enabled filtering banners by size rather than URL
783 * by adding patterns that replace all standard banner
784 * sizes with the "Junkbuster" gif to the re_filterfile
786 * - Enabled filtering WebBugs by providing a pattern
787 * which kills all 1x1 images
789 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
790 * which is selected by the (nonstandard and therefore
791 * capital) letter 'U' in the option string.
792 * It causes the quantifiers to be ungreedy by default.
793 * Appending a ? turns back to greedy (!).
795 * - Added a new interceptor ijb-send-banner, which
796 * sends back the "Junkbuster" gif. Without imagelist or
797 * MSIE detection support, or if tinygif = 1, or the
798 * URL isn't recognized as an imageurl, a lame HTML
799 * explanation is sent instead.
801 * - Added new feature, which permits blocking remote
802 * script redirects and firing back a local redirect
804 * The feature is conditionally compiled, i.e. it
805 * can be disabled with --disable-fast-redirects,
806 * plus it must be activated by a "fast-redirects"
807 * line in the config file, has its own log level
808 * and of course wants to be displayed by show-proxy-args
809 * Note: Boy, all the #ifdefs in 1001 locations and
810 * all the fumbling with configure.in and acconfig.h
811 * were *way* more work than the feature itself :-(
813 * - Because a generic redirect template was needed for
814 * this, tinygif = 3 now uses the same.
816 * - Moved GIFs, and other static HTTP response templates
821 * - Removed some >400 CRs again (Jon, you really worked
824 * Revision 1.3 2001/05/20 01:21:20 jongfoster
825 * Version 2.9.4 checkin.
826 * - Merged popupfile and cookiefile, and added control over PCRS
827 * filtering, in new "permissionsfile".
828 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
829 * file error you now get a message box (in the Win32 GUI) rather
830 * than the program exiting with no explanation.
831 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
833 * - Removed tabs from "config"
834 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
835 * - Bumped up version number.
837 * Revision 1.2 2001/05/17 23:02:36 oes
838 * - Made referrer option accept 'L' as a substitute for '§'
840 * Revision 1.1.1.1 2001/05/15 13:59:01 oes
841 * Initial import of version 2.9.3 source tree
844 *********************************************************************/
851 #include <sys/types.h>
861 * Convince GNU's libc to provide a strptime prototype.
864 #endif /*__GLIBC__ */
871 #if !defined(_WIN32) && !defined(__OS2__)
877 #ifdef FEATURE_PTHREAD
879 /* jcc.h is for mutex semapores only */
880 #endif /* def FEATURE_PTHREAD */
885 #include "jbsockets.h"
886 #include "miscutil.h"
891 #ifndef HAVE_STRPTIME
892 #include "strptime.h"
895 const char parsers_h_rcs[] = PARSERS_H_VERSION;
897 /* Fix a problem with Solaris. There should be no effect on other
899 * Solaris's isspace() is a macro which uses its argument directly
900 * as an array index. Therefore we need to make sure that high-bit
901 * characters generate +ve values, and ideally we also want to make
902 * the argument match the declared parameter type of "int".
904 * Why did they write a character function that can't take a simple
905 * "char" argument? Doh!
907 #define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
908 #define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
910 static char *get_header_line(struct iob *iob);
911 static jb_err scan_headers(struct client_state *csp);
912 static jb_err header_tagger(struct client_state *csp, char *header);
913 static jb_err parse_header_time(const char *header_time, time_t *result);
915 static jb_err crumble (struct client_state *csp, char **header);
916 static jb_err filter_header (struct client_state *csp, char **header);
917 static jb_err client_connection (struct client_state *csp, char **header);
918 static jb_err client_referrer (struct client_state *csp, char **header);
919 static jb_err client_uagent (struct client_state *csp, char **header);
920 static jb_err client_ua (struct client_state *csp, char **header);
921 static jb_err client_from (struct client_state *csp, char **header);
922 static jb_err client_send_cookie (struct client_state *csp, char **header);
923 static jb_err client_x_forwarded (struct client_state *csp, char **header);
924 static jb_err client_accept_encoding (struct client_state *csp, char **header);
925 static jb_err client_te (struct client_state *csp, char **header);
926 static jb_err client_max_forwards (struct client_state *csp, char **header);
927 static jb_err client_host (struct client_state *csp, char **header);
928 static jb_err client_if_modified_since (struct client_state *csp, char **header);
929 static jb_err client_accept_language (struct client_state *csp, char **header);
930 static jb_err client_if_none_match (struct client_state *csp, char **header);
931 static jb_err crunch_client_header (struct client_state *csp, char **header);
932 static jb_err client_x_filter (struct client_state *csp, char **header);
933 static jb_err client_range (struct client_state *csp, char **header);
934 static jb_err server_set_cookie (struct client_state *csp, char **header);
935 static jb_err server_connection (struct client_state *csp, char **header);
936 static jb_err server_content_type (struct client_state *csp, char **header);
937 static jb_err server_adjust_content_length(struct client_state *csp, char **header);
938 static jb_err server_content_md5 (struct client_state *csp, char **header);
939 static jb_err server_content_encoding (struct client_state *csp, char **header);
940 static jb_err server_transfer_coding (struct client_state *csp, char **header);
941 static jb_err server_http (struct client_state *csp, char **header);
942 static jb_err crunch_server_header (struct client_state *csp, char **header);
943 static jb_err server_last_modified (struct client_state *csp, char **header);
944 static jb_err server_content_disposition(struct client_state *csp, char **header);
946 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
947 static jb_err server_save_content_length(struct client_state *csp, char **header);
948 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
950 static jb_err client_host_adder (struct client_state *csp);
951 static jb_err client_xtra_adder (struct client_state *csp);
952 static jb_err client_x_forwarded_for_adder(struct client_state *csp);
953 static jb_err client_connection_header_adder(struct client_state *csp);
954 static jb_err server_connection_close_adder(struct client_state *csp);
956 static jb_err create_forged_referrer(char **header, const char *hostport);
957 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
958 static jb_err handle_conditional_hide_referrer_parameter(char **header,
959 const char *host, const int parameter_conditional_block);
962 * List of functions to run on a list of headers.
966 /** The header prefix to match */
969 /** The length of the prefix to match */
972 /** The function to apply to this line */
973 const parser_func_ptr parser;
976 static const struct parsers client_patterns[] = {
977 { "referer:", 8, client_referrer },
978 { "user-agent:", 11, client_uagent },
979 { "ua-", 3, client_ua },
980 { "from:", 5, client_from },
981 { "cookie:", 7, client_send_cookie },
982 { "x-forwarded-for:", 16, client_x_forwarded },
983 { "Accept-Encoding:", 16, client_accept_encoding },
984 { "TE:", 3, client_te },
985 { "Host:", 5, client_host },
986 { "if-modified-since:", 18, client_if_modified_since },
987 { "Keep-Alive:", 11, crumble },
988 { "connection:", 11, client_connection },
989 { "proxy-connection:", 17, crumble },
990 { "max-forwards:", 13, client_max_forwards },
991 { "Accept-Language:", 16, client_accept_language },
992 { "if-none-match:", 14, client_if_none_match },
993 { "Range:", 6, client_range },
994 { "Request-Range:", 14, client_range },
995 { "If-Range:", 9, client_range },
996 { "X-Filter:", 9, client_x_filter },
997 { "*", 0, crunch_client_header },
998 { "*", 0, filter_header },
1002 static const struct parsers server_patterns[] = {
1003 { "HTTP/", 5, server_http },
1004 { "set-cookie:", 11, server_set_cookie },
1005 { "connection:", 11, server_connection },
1006 { "Content-Type:", 13, server_content_type },
1007 { "Content-MD5:", 12, server_content_md5 },
1008 { "Content-Encoding:", 17, server_content_encoding },
1009 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1010 { "Content-Length:", 15, server_save_content_length },
1011 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1012 { "Transfer-Encoding:", 18, server_transfer_coding },
1013 { "Keep-Alive:", 11, crumble },
1014 { "content-disposition:", 20, server_content_disposition },
1015 { "Last-Modified:", 14, server_last_modified },
1016 { "*", 0, crunch_server_header },
1017 { "*", 0, filter_header },
1021 static const add_header_func_ptr add_client_headers[] = {
1023 client_x_forwarded_for_adder,
1025 /* Temporarily disabled: client_accept_encoding_adder, */
1026 client_connection_header_adder,
1030 static const add_header_func_ptr add_server_headers[] = {
1031 server_connection_close_adder,
1035 /*********************************************************************
1037 * Function : flush_socket
1039 * Description : Write any pending "buffered" content.
1042 * 1 : fd = file descriptor of the socket to read
1043 * 2 : iob = The I/O buffer to flush, usually csp->iob.
1045 * Returns : On success, the number of bytes written are returned (zero
1046 * indicates nothing was written). On error, -1 is returned,
1047 * and errno is set appropriately. If count is zero and the
1048 * file descriptor refers to a regular file, 0 will be
1049 * returned without causing any other effect. For a special
1050 * file, the results are not portable.
1052 *********************************************************************/
1053 int flush_socket(jb_socket fd, struct iob *iob)
1055 int len = iob->eod - iob->cur;
1062 if (write_socket(fd, iob->cur, (size_t)len))
1066 iob->eod = iob->cur = iob->buf;
1072 /*********************************************************************
1074 * Function : add_to_iob
1076 * Description : Add content to the buffered page, expanding the
1077 * buffer if necessary.
1080 * 1 : csp = Current client state (buffers, headers, etc...)
1081 * 2 : buf = holds the content to be added to the page
1082 * 3 : n = number of bytes to be added
1084 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
1085 * or buffer limit reached.
1087 *********************************************************************/
1088 jb_err add_to_iob(struct client_state *csp, char *buf, int n)
1090 struct iob *iob = csp->iob;
1091 size_t used, offset, need, want;
1094 if (n <= 0) return JB_ERR_OK;
1096 used = (size_t)(iob->eod - iob->buf);
1097 offset = (size_t)(iob->cur - iob->buf);
1098 need = used + (size_t)n + 1;
1101 * If the buffer can't hold the new data, extend it first.
1102 * Use the next power of two if possible, else use the actual need.
1104 if (need > csp->config->buffer_limit)
1106 log_error(LOG_LEVEL_INFO,
1107 "Buffer limit reached while extending the buffer (iob). Needed: %d. Limit: %d",
1108 need, csp->config->buffer_limit);
1109 return JB_ERR_MEMORY;
1112 if (need > iob->size)
1114 for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2;
1116 if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
1120 else if (NULL != (p = (char *)realloc(iob->buf, need)))
1126 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
1127 return JB_ERR_MEMORY;
1130 /* Update the iob pointers */
1131 iob->cur = p + offset;
1132 iob->eod = p + used;
1136 /* copy the new data into the iob buffer */
1137 memcpy(iob->eod, buf, (size_t)n);
1139 /* point to the end of the data */
1142 /* null terminate == cheap insurance */
1151 /*********************************************************************
1153 * Function : decompress_iob
1155 * Description : Decompress buffered page, expanding the
1156 * buffer as necessary. csp->iob->cur
1157 * should point to the the beginning of the
1158 * compressed data block.
1161 * 1 : csp = Current client state (buffers, headers, etc...)
1163 * Returns : JB_ERR_OK on success,
1164 * JB_ERR_MEMORY if out-of-memory limit reached, and
1165 * JB_ERR_COMPRESS if error decompressing buffer.
1167 *********************************************************************/
1168 jb_err decompress_iob(struct client_state *csp)
1170 char *buf; /* new, uncompressed buffer */
1171 char *cur; /* Current iob position (to keep the original
1172 * iob->cur unmodified if we return early) */
1173 size_t bufsize; /* allocated size of the new buffer */
1174 size_t old_size; /* Content size before decompression */
1175 size_t skip_size; /* Number of bytes at the beginning of the iob
1176 that we should NOT decompress. */
1177 int status; /* return status of the inflate() call */
1178 z_stream zstr; /* used by calls to zlib */
1180 assert(csp->iob->cur - csp->iob->buf > 0);
1181 assert(csp->iob->eod - csp->iob->cur > 0);
1183 bufsize = csp->iob->size;
1184 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
1185 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
1187 cur = csp->iob->cur;
1192 * This is to protect the parsing of gzipped data,
1193 * but it should(?) be valid for deflated data also.
1195 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
1196 return JB_ERR_COMPRESS;
1199 if (csp->content_type & CT_GZIP)
1202 * Our task is slightly complicated by the facts that data
1203 * compressed by gzip does not include a zlib header, and
1204 * that there is no easily accessible interface in zlib to
1205 * handle a gzip header. We strip off the gzip header by
1206 * hand, and later inform zlib not to expect a header.
1210 * Strip off the gzip header. Please see RFC 1952 for more
1211 * explanation of the appropriate fields.
1213 if ((*cur++ != (char)0x1f)
1214 || (*cur++ != (char)0x8b)
1215 || (*cur++ != Z_DEFLATED))
1217 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
1218 return JB_ERR_COMPRESS;
1224 * XXX: These magic numbers should be replaced
1225 * with macros to give a better idea what they do.
1229 /* The gzip header has reserved bits set; bail out. */
1230 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
1231 return JB_ERR_COMPRESS;
1235 /* Skip extra fields if necessary. */
1239 * Skip a given number of bytes, specified
1240 * as a 16-bit little-endian value.
1243 * XXX: This code used to be:
1245 * csp->iob->cur += *csp->iob->cur++ + (*csp->iob->cur++ << 8);
1247 * which I had to change into:
1249 * cur += *cur++ + (*cur++ << 8);
1251 * at which point gcc43 finally noticed that the value
1252 * of cur is undefined (it depends on which of the
1253 * summands is evaluated first).
1255 * I haven't come across a site where this
1256 * code is actually executed yet, but I hope
1260 skip_bytes = *cur++;
1261 skip_bytes = *cur++ << 8;
1263 assert(skip_bytes == *csp->iob->cur - 2 + ((*csp->iob->cur - 1) << 8));
1266 * The number of bytes to skip should be positive
1267 * and we'd like to stay in the buffer.
1269 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
1271 log_error(LOG_LEVEL_ERROR,
1272 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
1274 return JB_ERR_COMPRESS;
1276 log_error(LOG_LEVEL_INFO,
1277 "Skipping %d bytes for gzip compression. Does this sound right?",
1282 /* Skip the filename if necessary. */
1285 /* A null-terminated string is supposed to follow. */
1286 while (*cur++ && (cur < csp->iob->eod));
1290 /* Skip the comment if necessary. */
1293 /* A null-terminated string is supposed to follow. */
1294 while (*cur++ && (cur < csp->iob->eod));
1297 /* Skip the CRC if necessary. */
1303 if (cur >= csp->iob->eod)
1306 * If the current position pointer reached or passed
1307 * the buffer end, we were obviously tricked to skip
1310 log_error(LOG_LEVEL_ERROR,
1311 "Malformed gzip header detected. Aborting decompression.");
1312 return JB_ERR_COMPRESS;
1316 else if (csp->content_type & CT_DEFLATE)
1319 * XXX: The debug level should be lowered
1320 * before the next stable release.
1322 log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
1324 * In theory (that is, according to RFC 1950), deflate-compressed
1325 * data should begin with a two-byte zlib header and have an
1326 * adler32 checksum at the end. It seems that in practice only
1327 * the raw compressed data is sent. Note that this means that
1328 * we are not RFC 1950-compliant here, but the advantage is that
1329 * this actually works. :)
1331 * We add a dummy null byte to tell zlib where the data ends,
1332 * and later inform it not to expect a header.
1334 * Fortunately, add_to_iob() has thoughtfully null-terminated
1335 * the buffer; we can just increment the end pointer to include
1342 log_error(LOG_LEVEL_ERROR,
1343 "Unable to determine compression format for decompression");
1344 return JB_ERR_COMPRESS;
1347 /* Set up the fields required by zlib. */
1348 zstr.next_in = (Bytef *)cur;
1349 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
1350 zstr.zalloc = Z_NULL;
1351 zstr.zfree = Z_NULL;
1352 zstr.opaque = Z_NULL;
1355 * Passing -MAX_WBITS to inflateInit2 tells the library
1356 * that there is no zlib header.
1358 if (inflateInit2 (&zstr, -MAX_WBITS) != Z_OK)
1360 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
1361 return JB_ERR_COMPRESS;
1365 * Next, we allocate new storage for the inflated data.
1366 * We don't modify the existing iob yet, so in case there
1367 * is error in decompression we can recover gracefully.
1369 buf = zalloc(bufsize);
1372 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1373 return JB_ERR_MEMORY;
1376 assert(bufsize >= skip_size);
1377 memcpy(buf, csp->iob->buf, skip_size);
1378 zstr.avail_out = bufsize - skip_size;
1379 zstr.next_out = (Bytef *)buf + skip_size;
1381 /* Try to decompress the whole stream in one shot. */
1382 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
1384 /* We need to allocate more memory for the output buffer. */
1386 char *tmpbuf; /* used for realloc'ing the buffer */
1387 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
1390 * If zlib wants more data then there's a problem, because
1391 * the complete compressed file should have been buffered.
1393 if (0 == zstr.avail_in)
1395 log_error(LOG_LEVEL_ERROR, "Unexpected end of compressed iob");
1396 return JB_ERR_COMPRESS;
1400 * If we tried the limit and still didn't have enough
1401 * memory, just give up.
1403 if (bufsize == csp->config->buffer_limit)
1405 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
1406 return JB_ERR_MEMORY;
1409 /* Try doubling the buffer size each time. */
1412 /* Don't exceed the buffer limit. */
1413 if (bufsize > csp->config->buffer_limit)
1415 bufsize = csp->config->buffer_limit;
1418 /* Try to allocate the new buffer. */
1419 tmpbuf = realloc(buf, bufsize);
1422 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1424 return JB_ERR_MEMORY;
1428 char *oldnext_out = (char *)zstr.next_out;
1431 * Update the fields for inflate() to use the new
1432 * buffer, which may be in a location different from
1435 zstr.avail_out += bufsize - oldbufsize;
1436 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
1439 * Compare with an uglier method of calculating these values
1440 * that doesn't require the extra oldbufsize variable.
1442 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
1443 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
1444 assert(zstr.avail_out > 0);
1450 if (Z_STREAM_ERROR == inflateEnd(&zstr))
1452 log_error(LOG_LEVEL_ERROR,
1453 "Inconsistent stream state after decompression: %s", zstr.msg);
1455 * XXX: Intentionally no return.
1457 * According to zlib.h, Z_STREAM_ERROR is returned
1458 * "if the stream state was inconsistent".
1460 * I assume in this case inflate()'s status
1461 * would also be something different than Z_STREAM_END
1462 * so this check should be redundant, but lets see.
1466 if (status != Z_STREAM_END)
1468 /* We failed to decompress the stream. */
1469 log_error(LOG_LEVEL_ERROR,
1470 "Error in decompressing to the buffer (iob): %s", zstr.msg);
1471 return JB_ERR_COMPRESS;
1475 * Finally, we can actually update the iob, since the
1476 * decompression was successful. First, free the old
1479 freez(csp->iob->buf);
1481 /* Now, update the iob to use the new buffer. */
1482 csp->iob->buf = buf;
1483 csp->iob->cur = csp->iob->buf + skip_size;
1484 csp->iob->eod = (char *)zstr.next_out;
1485 csp->iob->size = bufsize;
1488 * Make sure the new uncompressed iob obeys some minimal
1489 * consistency conditions.
1491 if ((csp->iob->buf < csp->iob->cur)
1492 && (csp->iob->cur <= csp->iob->eod)
1493 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
1495 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
1498 log_error(LOG_LEVEL_RE_FILTER,
1499 "Decompression successful. Old size: %d, new size: %d.",
1500 old_size, new_size);
1504 /* zlib thinks this is OK, so lets do the same. */
1505 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
1510 /* It seems that zlib did something weird. */
1511 log_error(LOG_LEVEL_ERROR,
1512 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
1513 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
1514 csp->iob->eod, csp->iob->buf + csp->iob->size);
1515 return JB_ERR_COMPRESS;
1521 #endif /* defined(FEATURE_ZLIB) */
1524 /*********************************************************************
1526 * Function : string_move
1528 * Description : memmove wrapper to move the last part of a string
1529 * towards the beginning, overwriting the part in
1530 * the middle. strlcpy() can't be used here as the
1534 * 1 : dst = Destination to overwrite
1535 * 2 : src = Source to move.
1539 *********************************************************************/
1540 static void string_move(char *dst, char *src)
1544 /* +1 to copy the terminating nul as well. */
1545 memmove(dst, src, strlen(src)+1);
1549 /*********************************************************************
1551 * Function : normalize_lws
1553 * Description : Reduces unquoted linear white space in headers
1554 * to a single space in accordance with RFC 2616 2.2.
1555 * This simplifies parsing and filtering later on.
1557 * XXX: Remove log messages before
1558 * the next stable release?
1561 * 1 : header = A header with linear white space to reduce.
1565 *********************************************************************/
1566 static void normalize_lws(char *header)
1572 if (ijb_isspace(*p) && ijb_isspace(*(p+1)))
1576 while (ijb_isspace(*q))
1580 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
1581 string_move(p+1, q);
1586 log_error(LOG_LEVEL_HEADER,
1587 "Converting tab to space in '%s'", header);
1592 char *end_of_token = strstr(p+1, "\"");
1594 if (NULL != end_of_token)
1596 /* Don't mess with quoted text. */
1601 log_error(LOG_LEVEL_HEADER,
1602 "Ignoring single quote in '%s'", header);
1608 p = strchr(header, ':');
1609 if ((p != NULL) && (p != header) && ijb_isspace(*(p-1)))
1612 * There's still space before the colon.
1615 string_move(p-1, p);
1620 /*********************************************************************
1622 * Function : get_header
1624 * Description : This (odd) routine will parse the csp->iob
1625 * to get the next complete header.
1628 * 1 : iob = The I/O buffer to parse, usually csp->iob.
1630 * Returns : Any one of the following:
1632 * 1) a pointer to a dynamically allocated string that contains a header line
1633 * 2) NULL indicating that the end of the header was reached
1634 * 3) "" indicating that the end of the iob was reached before finding
1635 * a complete header line.
1637 *********************************************************************/
1638 char *get_header(struct iob *iob)
1642 header = get_header_line(iob);
1644 if ((header == NULL) || (*header == '\0'))
1647 * No complete header read yet, tell the client.
1652 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
1655 * Header spans multiple lines, append the next one.
1657 char *continued_header;
1659 continued_header = get_header_line(iob);
1660 if ((continued_header == NULL) || (*continued_header == '\0'))
1663 * No complete header read yet, return what we got.
1664 * XXX: Should "unread" header instead.
1666 log_error(LOG_LEVEL_INFO,
1667 "Failed to read a multi-line header properly: '%s'",
1672 if (JB_ERR_OK != string_join(&header, continued_header))
1674 log_error(LOG_LEVEL_FATAL,
1675 "Out of memory while appending multiple headers.");
1679 /* XXX: remove before next stable release. */
1680 log_error(LOG_LEVEL_HEADER,
1681 "Merged multiple header lines to: '%s'",
1686 normalize_lws(header);
1693 /*********************************************************************
1695 * Function : get_header_line
1697 * Description : This (odd) routine will parse the csp->iob
1698 * to get the next header line.
1701 * 1 : iob = The I/O buffer to parse, usually csp->iob.
1703 * Returns : Any one of the following:
1705 * 1) a pointer to a dynamically allocated string that contains a header line
1706 * 2) NULL indicating that the end of the header was reached
1707 * 3) "" indicating that the end of the iob was reached before finding
1708 * a complete header line.
1710 *********************************************************************/
1711 static char *get_header_line(struct iob *iob)
1715 if ((iob->cur == NULL)
1716 || ((p = strchr(iob->cur, '\n')) == NULL))
1718 return(""); /* couldn't find a complete header */
1723 ret = strdup(iob->cur);
1726 /* FIXME No way to handle error properly */
1727 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
1732 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
1734 /* is this a blank line (i.e. the end of the header) ? */
1746 /*********************************************************************
1748 * Function : get_header_value
1750 * Description : Get the value of a given header from a chained list
1751 * of header lines or return NULL if no such header is
1752 * present in the list.
1755 * 1 : header_list = pointer to list
1756 * 2 : header_name = string with name of header to look for.
1757 * Trailing colon required, capitalization
1760 * Returns : NULL if not found, else value of header
1762 *********************************************************************/
1763 char *get_header_value(const struct list *header_list, const char *header_name)
1765 struct list_entry *cur_entry;
1769 assert(header_list);
1770 assert(header_name);
1771 length = strlen(header_name);
1773 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
1777 if (!strncmpic(cur_entry->str, header_name, length))
1780 * Found: return pointer to start of value
1782 ret = cur_entry->str + length;
1783 while (*ret && ijb_isspace(*ret)) ret++;
1797 /*********************************************************************
1799 * Function : scan_headers
1801 * Description : Scans headers, applies tags and updates action bits.
1804 * 1 : csp = Current client state (buffers, headers, etc...)
1806 * Returns : JB_ERR_OK
1808 *********************************************************************/
1809 static jb_err scan_headers(struct client_state *csp)
1811 struct list_entry *h; /* Header */
1812 jb_err err = JB_ERR_OK;
1814 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1816 /* Header crunch()ed in previous run? -> ignore */
1817 if (h->str == NULL) continue;
1818 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1819 err = header_tagger(csp, h->str);
1826 /*********************************************************************
1830 * Description : add, delete or modify lines in the HTTP header streams.
1831 * On entry, it receives a linked list of headers space
1832 * that was allocated dynamically (both the list nodes
1833 * and the header contents).
1835 * As a side effect it frees the space used by the original
1839 * 1 : csp = Current client state (buffers, headers, etc...)
1840 * 2 : filter_server_headers = Boolean to switch between
1841 * server and header filtering.
1843 * Returns : JB_ERR_OK in case off success, or
1844 * JB_ERR_MEMORY on out-of-memory error.
1846 *********************************************************************/
1847 jb_err sed(struct client_state *csp, int filter_server_headers)
1849 /* XXX: use more descriptive names. */
1850 struct list_entry *p;
1851 const struct parsers *v;
1852 const add_header_func_ptr *f;
1853 jb_err err = JB_ERR_OK;
1855 if (filter_server_headers)
1857 v = server_patterns;
1858 f = add_server_headers;
1862 v = client_patterns;
1863 f = add_client_headers;
1868 while ((err == JB_ERR_OK) && (v->str != NULL))
1870 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1872 /* Header crunch()ed in previous run? -> ignore */
1873 if (p->str == NULL) continue;
1875 /* Does the current parser handle this header? */
1876 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1877 (v->len == CHECK_EVERY_HEADER_REMAINING))
1879 err = v->parser(csp, &(p->str));
1885 /* place additional headers on the csp->headers list */
1886 while ((err == JB_ERR_OK) && (*f))
1896 /*********************************************************************
1898 * Function : update_server_headers
1900 * Description : Updates server headers after the body has been modified.
1903 * 1 : csp = Current client state (buffers, headers, etc...)
1905 * Returns : JB_ERR_OK in case off success, or
1906 * JB_ERR_MEMORY on out-of-memory error.
1908 *********************************************************************/
1909 jb_err update_server_headers(struct client_state *csp)
1911 jb_err err = JB_ERR_OK;
1913 static const struct parsers server_patterns_light[] = {
1914 { "Content-Length:", 15, server_adjust_content_length },
1915 { "Transfer-Encoding:", 18, server_transfer_coding },
1917 { "Content-Encoding:", 17, server_content_encoding },
1918 #endif /* def FEATURE_ZLIB */
1922 if (strncmpic(csp->http->cmd, "HEAD", 4))
1924 const struct parsers *v;
1925 struct list_entry *p;
1927 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1929 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1931 /* Header crunch()ed in previous run? -> ignore */
1932 if (p->str == NULL) continue;
1934 /* Does the current parser handle this header? */
1935 if (strncmpic(p->str, v->str, v->len) == 0)
1937 err = v->parser(csp, (char **)&(p->str));
1947 /*********************************************************************
1949 * Function : header_tagger
1951 * Description : Executes all text substitutions from applying
1952 * tag actions and saves the result as tag.
1954 * XXX: Shares enough code with filter_header() and
1955 * pcrs_filter_response() to warrant some helper functions.
1958 * 1 : csp = Current client state (buffers, headers, etc...)
1959 * 2 : header = Header that is used as tagger input
1961 * Returns : JB_ERR_OK on success and always succeeds
1963 *********************************************************************/
1964 static jb_err header_tagger(struct client_state *csp, char *header)
1966 int wanted_filter_type;
1967 int multi_action_index;
1971 struct file_list *fl;
1972 struct re_filterfile_spec *b;
1973 struct list_entry *tag_name;
1975 int found_filters = 0;
1976 const size_t header_length = strlen(header);
1978 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1980 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1981 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1985 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1986 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1989 /* Check if there are any filters */
1990 for (i = 0; i < MAX_AF_FILES; i++)
2003 if (0 == found_filters)
2005 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2006 "tagging enabled, but no taggers available.");
2010 for (i = 0; i < MAX_AF_FILES; i++)
2013 if ((NULL == fl) || (NULL == fl->f))
2016 * Either there are no filter files
2017 * left, or this filter file just
2018 * contains no valid filters.
2020 * Continue to be sure we don't miss
2021 * valid filter files that are chained
2022 * after empty or invalid ones.
2027 /* For all filters, */
2028 for (b = fl->f; b; b = b->next)
2030 if (b->type != wanted_filter_type)
2032 /* skip the ones we don't care about, */
2035 /* leaving only taggers that could apply, of which we use the ones, */
2036 for (tag_name = csp->action->multi[multi_action_index]->first;
2037 NULL != tag_name; tag_name = tag_name->next)
2039 /* that do apply, and */
2040 if (strcmp(b->name, tag_name->str) == 0)
2042 char *modified_tag = NULL;
2044 size_t size = header_length;
2045 pcrs_job *joblist = b->joblist;
2047 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2049 if (NULL == joblist)
2051 log_error(LOG_LEVEL_RE_FILTER,
2052 "Tagger %s has empty joblist. Nothing to do.", b->name);
2056 /* execute their pcrs_joblist on the header. */
2057 for (job = joblist; NULL != job; job = job->next)
2059 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
2063 /* Success, continue with the modified version. */
2072 /* Tagger doesn't match */
2075 /* Regex failure, log it but continue anyway. */
2076 log_error(LOG_LEVEL_ERROR,
2077 "Problems with tagger \'%s\' and header \'%s\': %s",
2078 b->name, *header, pcrs_strerror(hits));
2080 freez(modified_tag);
2084 if (b->dynamic) pcrs_free_joblist(joblist);
2086 /* If this tagger matched */
2092 * There is to technical limitation which makes
2093 * it impossible to use empty tags, but I assume
2094 * no one would do it intentionally.
2097 log_error(LOG_LEVEL_INFO,
2098 "Tagger \'%s\' created an empty tag. Ignored.",
2103 if (!list_contains_item(csp->tags, tag))
2105 if (JB_ERR_OK != enlist(csp->tags, tag))
2107 log_error(LOG_LEVEL_ERROR,
2108 "Insufficient memory to add tag \'%s\', "
2109 "based on tagger \'%s\' and header \'%s\'",
2110 tag, b->name, *header);
2114 char *action_message;
2116 * update the action bits right away, to make
2117 * tagging based on tags set by earlier taggers
2118 * of the same kind possible.
2120 if (update_action_bits_for_tag(csp, tag))
2122 action_message = "Action bits updated accordingly.";
2126 action_message = "No action bits update necessary.";
2129 log_error(LOG_LEVEL_HEADER,
2130 "Tagger \'%s\' added tag \'%s\'. %s",
2131 b->name, tag, action_message);
2136 /* XXX: Is this log-worthy? */
2137 log_error(LOG_LEVEL_HEADER,
2138 "Tagger \'%s\' didn't add tag \'%s\'. "
2139 "Tag already present", b->name, tag);
2142 } /* if the tagger matched */
2143 } /* if the tagger applies */
2144 } /* for every tagger that could apply */
2145 } /* for all filters */
2146 } /* for all filter files */
2151 /* here begins the family of parser functions that reformat header lines */
2153 /*********************************************************************
2155 * Function : filter_header
2157 * Description : Executes all text substitutions from all applying
2158 * +(server|client)-header-filter actions on the header.
2159 * Most of the code was copied from pcrs_filter_response,
2160 * including the rather short variable names
2163 * 1 : csp = Current client state (buffers, headers, etc...)
2164 * 2 : header = On input, pointer to header to modify.
2165 * On output, pointer to the modified header, or NULL
2166 * to remove the header. This function frees the
2167 * original string if necessary.
2169 * Returns : JB_ERR_OK on success and always succeeds
2171 *********************************************************************/
2172 static jb_err filter_header(struct client_state *csp, char **header)
2176 size_t size = strlen(*header);
2178 char *newheader = NULL;
2181 struct file_list *fl;
2182 struct re_filterfile_spec *b;
2183 struct list_entry *filtername;
2185 int i, found_filters = 0;
2186 int wanted_filter_type;
2187 int multi_action_index;
2189 if (csp->flags & CSP_FLAG_NO_FILTERING)
2194 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
2196 wanted_filter_type = FT_SERVER_HEADER_FILTER;
2197 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
2201 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
2202 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
2206 * Need to check the set of re_filterfiles...
2208 for (i = 0; i < MAX_AF_FILES; i++)
2221 if (0 == found_filters)
2223 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
2224 "header filtering enabled, but no matching filters available.");
2228 for (i = 0; i < MAX_AF_FILES; i++)
2231 if ((NULL == fl) || (NULL == fl->f))
2234 * Either there are no filter files
2235 * left, or this filter file just
2236 * contains no valid filters.
2238 * Continue to be sure we don't miss
2239 * valid filter files that are chained
2240 * after empty or invalid ones.
2245 * For all applying +filter actions, look if a filter by that
2246 * name exists and if yes, execute its pcrs_joblist on the
2249 for (b = fl->f; b; b = b->next)
2251 if (b->type != wanted_filter_type)
2253 /* Skip other filter types */
2257 for (filtername = csp->action->multi[multi_action_index]->first;
2258 filtername ; filtername = filtername->next)
2260 if (strcmp(b->name, filtername->str) == 0)
2262 int current_hits = 0;
2263 pcrs_job *joblist = b->joblist;
2265 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2267 if (NULL == joblist)
2269 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
2273 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
2274 *header, size, b->name);
2276 /* Apply all jobs from the joblist */
2277 for (job = joblist; NULL != job; job = job->next)
2279 matches = pcrs_execute(job, *header, size, &newheader, &size);
2282 current_hits += matches;
2283 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
2285 *header = newheader;
2287 else if ( 0 == matches )
2289 /* Filter doesn't change header */
2295 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
2296 *header, b->name, pcrs_strerror(matches));
2297 if (newheader != NULL)
2299 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
2305 if (b->dynamic) pcrs_free_joblist(joblist);
2307 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
2308 hits += current_hits;
2315 * Additionally checking for hits is important because if
2316 * the continue hack is triggered, server headers can
2317 * arrive empty to separate multiple heads from each other.
2319 if ((0 == size) && hits)
2321 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
2329 /*********************************************************************
2331 * Function : server_connection
2333 * Description : Makes sure that the value of the Connection: header
2334 * is "close" and signals server_connection_close_adder
2338 * 1 : csp = Current client state (buffers, headers, etc...)
2339 * 2 : header = On input, pointer to header to modify.
2340 * On output, pointer to the modified header, or NULL
2341 * to remove the header. This function frees the
2342 * original string if necessary.
2344 * Returns : JB_ERR_OK on success, or
2345 * JB_ERR_MEMORY on out-of-memory error.
2347 *********************************************************************/
2348 static jb_err server_connection(struct client_state *csp, char **header)
2350 char *old_header = *header;
2352 /* Do we have a 'Connection: close' header? */
2353 if (strcmpic(*header, "Connection: close"))
2355 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2356 if (!strcmpic(*header, "Connection: keep-alive"))
2358 /* Remember to keep the connection alive. */
2359 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
2361 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
2363 *header = strdup("Connection: close");
2366 return JB_ERR_MEMORY;
2368 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
2372 /* Signal server_connection_close_adder() to return early. */
2373 csp->flags |= CSP_FLAG_SERVER_CONNECTION_CLOSE_SET;
2378 /*********************************************************************
2380 * Function : client_connection
2382 * Description : Makes sure a proper "Connection:" header is
2383 * set and signals connection_header_adder
2387 * 1 : csp = Current client state (buffers, headers, etc...)
2388 * 2 : header = On input, pointer to header to modify.
2389 * On output, pointer to the modified header, or NULL
2390 * to remove the header. This function frees the
2391 * original string if necessary.
2393 * Returns : JB_ERR_OK on success, or
2394 * JB_ERR_MEMORY on out-of-memory error.
2396 *********************************************************************/
2397 static jb_err client_connection(struct client_state *csp, char **header)
2399 char *old_header = *header;
2400 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2401 static const char wanted_header[] = "Connection: keep-alive";
2403 static const char wanted_header[] = "Connection: close";
2404 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
2406 if (strcmpic(*header, wanted_header))
2408 *header = strdup(wanted_header);
2411 return JB_ERR_MEMORY;
2413 log_error(LOG_LEVEL_HEADER,
2414 "Replaced: \'%s\' with \'%s\'", old_header, *header);
2418 /* Signal client_connection_close_adder() to return early. */
2419 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
2425 /*********************************************************************
2427 * Function : crumble
2429 * Description : This is called if a header matches a pattern to "crunch"
2432 * 1 : csp = Current client state (buffers, headers, etc...)
2433 * 2 : header = On input, pointer to header to modify.
2434 * On output, pointer to the modified header, or NULL
2435 * to remove the header. This function frees the
2436 * original string if necessary.
2438 * Returns : JB_ERR_OK on success, or
2439 * JB_ERR_MEMORY on out-of-memory error.
2441 *********************************************************************/
2442 static jb_err crumble(struct client_state *csp, char **header)
2444 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2450 /*********************************************************************
2452 * Function : crunch_server_header
2454 * Description : Crunch server header if it matches a string supplied by the
2455 * user. Called from `sed'.
2458 * 1 : csp = Current client state (buffers, headers, etc...)
2459 * 2 : header = On input, pointer to header to modify.
2460 * On output, pointer to the modified header, or NULL
2461 * to remove the header. This function frees the
2462 * original string if necessary.
2464 * Returns : JB_ERR_OK on success and always succeeds
2466 *********************************************************************/
2467 static jb_err crunch_server_header(struct client_state *csp, char **header)
2469 const char *crunch_pattern;
2471 /* Do we feel like crunching? */
2472 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2474 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2476 /* Is the current header the lucky one? */
2477 if (strstr(*header, crunch_pattern))
2479 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2488 /*********************************************************************
2490 * Function : server_content_type
2492 * Description : Set the content-type for filterable types (text/.*,
2493 * .*xml.*, javascript and image/gif) unless filtering has been
2494 * forbidden (CT_TABOO) while parsing earlier headers.
2495 * NOTE: Since text/plain is commonly used by web servers
2496 * for files whose correct type is unknown, we don't
2497 * set CT_TEXT for it.
2500 * 1 : csp = Current client state (buffers, headers, etc...)
2501 * 2 : header = On input, pointer to header to modify.
2502 * On output, pointer to the modified header, or NULL
2503 * to remove the header. This function frees the
2504 * original string if necessary.
2506 * Returns : JB_ERR_OK on success, or
2507 * JB_ERR_MEMORY on out-of-memory error.
2509 *********************************************************************/
2510 static jb_err server_content_type(struct client_state *csp, char **header)
2512 /* Remove header if it isn't the first Content-Type header */
2513 if ((csp->content_type & CT_DECLARED))
2516 * Another, slightly slower, way to see if
2517 * we already parsed another Content-Type header.
2519 assert(NULL != get_header_value(csp->headers, "Content-Type:"));
2521 log_error(LOG_LEVEL_ERROR,
2522 "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
2530 * Signal that the Content-Type has been set.
2532 csp->content_type |= CT_DECLARED;
2534 if (!(csp->content_type & CT_TABOO))
2537 * XXX: The assumption that text/plain is a sign of
2538 * binary data seems to be somewhat unreasonable nowadays
2539 * and should be dropped after 3.0.8 is out.
2541 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2542 || strstr(*header, "xml")
2543 || strstr(*header, "application/x-javascript"))
2545 csp->content_type |= CT_TEXT;
2547 else if (strstr(*header, "image/gif"))
2549 csp->content_type |= CT_GIF;
2554 * Are we messing with the content type?
2556 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2559 * Make sure the user doesn't accidently
2560 * change the content type of binary documents.
2562 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2565 *header = strdup("Content-Type: ");
2566 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2570 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2571 return JB_ERR_MEMORY;
2573 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2577 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2578 "It doesn't look like a content type that should be filtered. "
2579 "Enable force-text-mode if you know what you're doing.", *header);
2587 /*********************************************************************
2589 * Function : server_transfer_coding
2591 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2592 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2593 * - Remove header if body was chunked but has been
2594 * de-chunked for filtering.
2597 * 1 : csp = Current client state (buffers, headers, etc...)
2598 * 2 : header = On input, pointer to header to modify.
2599 * On output, pointer to the modified header, or NULL
2600 * to remove the header. This function frees the
2601 * original string if necessary.
2603 * Returns : JB_ERR_OK on success, or
2604 * JB_ERR_MEMORY on out-of-memory error.
2606 *********************************************************************/
2607 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2610 * Turn off pcrs and gif filtering if body compressed
2612 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2616 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2618 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2619 csp->http->cmd, *header);
2620 #endif /* def FEATURE_ZLIB */
2621 csp->content_type = CT_TABOO;
2625 * Raise flag if body chunked
2627 if (strstr(*header, "chunked"))
2629 csp->flags |= CSP_FLAG_CHUNKED;
2632 * If the body was modified, it has been de-chunked first
2633 * and the header must be removed.
2635 * FIXME: If there is more than one transfer encoding,
2636 * only the "chunked" part should be removed here.
2638 if (csp->flags & CSP_FLAG_MODIFIED)
2640 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2644 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2645 log_error(LOG_LEVEL_ERROR,
2646 "Chunked transfer encoding detected with experimental "
2647 "keep-alive support enabled. Expect delayed delivery.");
2648 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
2656 /*********************************************************************
2658 * Function : server_content_encoding
2660 * Description : This function is run twice for each request,
2661 * unless FEATURE_ZLIB and filtering are disabled.
2663 * The first run is used to check if the content
2664 * is compressed, if FEATURE_ZLIB is disabled
2665 * filtering is then disabled as well, if FEATURE_ZLIB
2666 * is enabled the content is marked for decompression.
2668 * The second run is used to remove the Content-Encoding
2669 * header if the decompression was successful.
2672 * 1 : csp = Current client state (buffers, headers, etc...)
2673 * 2 : header = On input, pointer to header to modify.
2674 * On output, pointer to the modified header, or NULL
2675 * to remove the header. This function frees the
2676 * original string if necessary.
2678 * Returns : JB_ERR_OK on success, or
2679 * JB_ERR_MEMORY on out-of-memory error.
2681 *********************************************************************/
2682 static jb_err server_content_encoding(struct client_state *csp, char **header)
2685 if ((csp->flags & CSP_FLAG_MODIFIED)
2686 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2689 * We successfully decompressed the content,
2690 * and have to clean the header now, so the
2691 * client no longer expects compressed data..
2693 * XXX: There is a difference between cleaning
2694 * and removing it completely.
2696 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2699 else if (strstr(*header, "gzip"))
2701 /* Mark for gzip decompression */
2702 csp->content_type |= CT_GZIP;
2704 else if (strstr(*header, "deflate"))
2706 /* Mark for zlib decompression */
2707 csp->content_type |= CT_DEFLATE;
2709 else if (strstr(*header, "compress"))
2712 * We can't decompress this; therefore we can't filter
2715 csp->content_type |= CT_TABOO;
2717 #else /* !defined(FEATURE_ZLIB) */
2718 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2721 * Body is compressed, turn off pcrs and gif filtering.
2723 csp->content_type |= CT_TABOO;
2726 * Log a warning if the user expects the content to be filtered.
2728 if ((csp->rlist != NULL) &&
2729 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2731 log_error(LOG_LEVEL_INFO,
2732 "Compressed content detected, content filtering disabled. "
2733 "Consider recompiling Privoxy with zlib support or "
2734 "enable the prevent-compression action.");
2737 #endif /* defined(FEATURE_ZLIB) */
2744 /*********************************************************************
2746 * Function : server_adjust_content_length
2748 * Description : Adjust Content-Length header if we modified
2752 * 1 : csp = Current client state (buffers, headers, etc...)
2753 * 2 : header = On input, pointer to header to modify.
2754 * On output, pointer to the modified header, or NULL
2755 * to remove the header. This function frees the
2756 * original string if necessary.
2758 * Returns : JB_ERR_OK on success, or
2759 * JB_ERR_MEMORY on out-of-memory error.
2761 *********************************************************************/
2762 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2764 const size_t max_header_length = 80;
2766 /* Regenerate header if the content was modified. */
2767 if (csp->flags & CSP_FLAG_MODIFIED)
2770 *header = (char *) zalloc(max_header_length);
2771 if (*header == NULL)
2773 return JB_ERR_MEMORY;
2776 snprintf(*header, max_header_length, "Content-Length: %d",
2777 (int)csp->content_length);
2778 log_error(LOG_LEVEL_HEADER, "Adjusted Content-Length to %d",
2779 (int)csp->content_length);
2786 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2787 /*********************************************************************
2789 * Function : server_save_content_length
2791 * Description : Save the Content-Length sent by the server.
2794 * 1 : csp = Current client state (buffers, headers, etc...)
2795 * 2 : header = On input, pointer to header to modify.
2796 * On output, pointer to the modified header, or NULL
2797 * to remove the header. This function frees the
2798 * original string if necessary.
2800 * Returns : JB_ERR_OK on success, or
2801 * JB_ERR_MEMORY on out-of-memory error.
2803 *********************************************************************/
2804 static jb_err server_save_content_length(struct client_state *csp, char **header)
2806 unsigned int content_length = 0;
2808 assert(*(*header+14) == ':');
2810 if (1 != sscanf(*header+14, ": %u", &content_length))
2812 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2817 csp->expected_content_length = content_length;
2818 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2823 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2826 /*********************************************************************
2828 * Function : server_content_md5
2830 * Description : Crumble any Content-MD5 headers if the document was
2831 * modified. FIXME: Should we re-compute instead?
2834 * 1 : csp = Current client state (buffers, headers, etc...)
2835 * 2 : header = On input, pointer to header to modify.
2836 * On output, pointer to the modified header, or NULL
2837 * to remove the header. This function frees the
2838 * original string if necessary.
2840 * Returns : JB_ERR_OK on success, or
2841 * JB_ERR_MEMORY on out-of-memory error.
2843 *********************************************************************/
2844 static jb_err server_content_md5(struct client_state *csp, char **header)
2846 if (csp->flags & CSP_FLAG_MODIFIED)
2848 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2856 /*********************************************************************
2858 * Function : server_content_disposition
2860 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2861 * Called from `sed'.
2864 * 1 : csp = Current client state (buffers, headers, etc...)
2865 * 2 : header = On input, pointer to header to modify.
2866 * On output, pointer to the modified header, or NULL
2867 * to remove the header. This function frees the
2868 * original string if necessary.
2870 * Returns : JB_ERR_OK on success, or
2871 * JB_ERR_MEMORY on out-of-memory error.
2873 *********************************************************************/
2874 static jb_err server_content_disposition(struct client_state *csp, char **header)
2879 * Are we messing with the Content-Disposition header?
2881 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2887 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2889 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2892 * Blocking content-disposition header
2894 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2901 * Replacing Content-Disposition header
2904 *header = strdup("Content-Disposition: ");
2905 string_append(header, newval);
2907 if (*header != NULL)
2909 log_error(LOG_LEVEL_HEADER,
2910 "Content-Disposition header crunched and replaced with: %s", *header);
2913 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2917 /*********************************************************************
2919 * Function : server_last_modified
2921 * Description : Changes Last-Modified header to the actual date
2922 * to help hide-if-modified-since.
2923 * Called from `sed'.
2926 * 1 : csp = Current client state (buffers, headers, etc...)
2927 * 2 : header = On input, pointer to header to modify.
2928 * On output, pointer to the modified header, or NULL
2929 * to remove the header. This function frees the
2930 * original string if necessary.
2932 * Returns : JB_ERR_OK on success, or
2933 * JB_ERR_MEMORY on out-of-memory error.
2935 *********************************************************************/
2936 static jb_err server_last_modified(struct client_state *csp, char **header)
2939 char buf[BUFFER_SIZE];
2942 #ifdef HAVE_GMTIME_R
2945 struct tm *timeptr = NULL;
2946 time_t now, last_modified;
2948 long int days, hours, minutes, seconds;
2951 * Are we messing with the Last-Modified header?
2953 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2959 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2961 if (0 == strcmpic(newval, "block") )
2964 * Blocking Last-Modified header. Useless but why not.
2966 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2970 else if (0 == strcmpic(newval, "reset-to-request-time"))
2973 * Setting Last-Modified Header to now.
2975 get_http_time(0, buf, sizeof(buf));
2977 *header = strdup("Last-Modified: ");
2978 string_append(header, buf);
2980 if (*header == NULL)
2982 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2986 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2989 else if (0 == strcmpic(newval, "randomize"))
2991 const char *header_time = *header + sizeof("Last-Modified:");
2993 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2995 #ifdef HAVE_GMTIME_R
2996 timeptr = gmtime_r(&now, &gmt);
2997 #elif FEATURE_PTHREAD
2998 privoxy_mutex_lock(&gmtime_mutex);
2999 timeptr = gmtime(&now);
3000 privoxy_mutex_unlock(&gmtime_mutex);
3002 timeptr = gmtime(&now);
3004 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
3006 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3011 rtime = (long int)difftime(now, last_modified);
3020 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
3022 rtime = pick_from_range(rtime);
3023 if (negative) rtime *= -1;
3024 last_modified += rtime;
3025 #ifdef HAVE_GMTIME_R
3026 timeptr = gmtime_r(&last_modified, &gmt);
3027 #elif FEATURE_PTHREAD
3028 privoxy_mutex_lock(&gmtime_mutex);
3029 timeptr = gmtime(&last_modified);
3030 privoxy_mutex_unlock(&gmtime_mutex);
3032 timeptr = gmtime(&last_modified);
3034 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
3036 *header = strdup("Last-Modified: ");
3037 string_append(header, newheader);
3039 if (*header == NULL)
3041 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
3042 return JB_ERR_MEMORY;
3045 days = rtime / (3600 * 24);
3046 hours = rtime / 3600 % 24;
3047 minutes = rtime / 60 % 60;
3048 seconds = rtime % 60;
3050 log_error(LOG_LEVEL_HEADER,
3051 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
3052 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
3053 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
3057 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
3066 /*********************************************************************
3068 * Function : client_accept_encoding
3070 * Description : Rewrite the client's Accept-Encoding header so that
3071 * if doesn't allow compression, if the action applies.
3072 * Note: For HTTP/1.0 the absence of the header is enough.
3075 * 1 : csp = Current client state (buffers, headers, etc...)
3076 * 2 : header = On input, pointer to header to modify.
3077 * On output, pointer to the modified header, or NULL
3078 * to remove the header. This function frees the
3079 * original string if necessary.
3081 * Returns : JB_ERR_OK on success, or
3082 * JB_ERR_MEMORY on out-of-memory error.
3084 *********************************************************************/
3085 static jb_err client_accept_encoding(struct client_state *csp, char **header)
3087 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3089 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
3093 /* Temporarily disable the correct behaviour to
3094 * work around a PHP bug.
3096 * if (!strcmpic(csp->http->ver, "HTTP/1.1"))
3098 * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0");
3099 * if (*header == NULL)
3101 * return JB_ERR_MEMORY;
3112 /*********************************************************************
3114 * Function : client_te
3116 * Description : Rewrite the client's TE header so that
3117 * if doesn't allow compression, if the action applies.
3120 * 1 : csp = Current client state (buffers, headers, etc...)
3121 * 2 : header = On input, pointer to header to modify.
3122 * On output, pointer to the modified header, or NULL
3123 * to remove the header. This function frees the
3124 * original string if necessary.
3126 * Returns : JB_ERR_OK on success, or
3127 * JB_ERR_MEMORY on out-of-memory error.
3129 *********************************************************************/
3130 static jb_err client_te(struct client_state *csp, char **header)
3132 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3135 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
3142 /*********************************************************************
3144 * Function : client_referrer
3146 * Description : Handle the "referer" config setting properly.
3147 * Called from `sed'.
3150 * 1 : csp = Current client state (buffers, headers, etc...)
3151 * 2 : header = On input, pointer to header to modify.
3152 * On output, pointer to the modified header, or NULL
3153 * to remove the header. This function frees the
3154 * original string if necessary.
3156 * Returns : JB_ERR_OK on success, or
3157 * JB_ERR_MEMORY on out-of-memory error.
3159 *********************************************************************/
3160 static jb_err client_referrer(struct client_state *csp, char **header)
3162 const char *parameter;
3163 /* booleans for parameters we have to check multiple times */
3164 int parameter_conditional_block;
3165 int parameter_conditional_forge;
3167 #ifdef FEATURE_FORCE_LOAD
3169 * Since the referrer can include the prefix even
3170 * if the request itself is non-forced, we must
3171 * clean it unconditionally.
3173 * XXX: strclean is too broad
3175 strclean(*header, FORCE_PREFIX);
3176 #endif /* def FEATURE_FORCE_LOAD */
3178 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
3180 /* Nothing left to do */
3184 parameter = csp->action->string[ACTION_STRING_REFERER];
3185 assert(parameter != NULL);
3186 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
3187 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
3189 if (!parameter_conditional_block && !parameter_conditional_forge)
3192 * As conditional-block and conditional-forge are the only
3193 * parameters that rely on the original referrer, we can
3194 * remove it now for all the others.
3199 if (0 == strcmpic(parameter, "block"))
3201 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
3204 else if (parameter_conditional_block || parameter_conditional_forge)
3206 return handle_conditional_hide_referrer_parameter(header,
3207 csp->http->hostport, parameter_conditional_block);
3209 else if (0 == strcmpic(parameter, "forge"))
3211 return create_forged_referrer(header, csp->http->hostport);
3215 /* interpret parameter as user-supplied referer to fake */
3216 return create_fake_referrer(header, parameter);
3221 /*********************************************************************
3223 * Function : client_accept_language
3225 * Description : Handle the "Accept-Language" config setting properly.
3226 * Called from `sed'.
3229 * 1 : csp = Current client state (buffers, headers, etc...)
3230 * 2 : header = On input, pointer to header to modify.
3231 * On output, pointer to the modified header, or NULL
3232 * to remove the header. This function frees the
3233 * original string if necessary.
3235 * Returns : JB_ERR_OK on success, or
3236 * JB_ERR_MEMORY on out-of-memory error.
3238 *********************************************************************/
3239 static jb_err client_accept_language(struct client_state *csp, char **header)
3244 * Are we messing with the Accept-Language?
3246 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
3248 /*I don't think so*/
3252 newval = csp->action->string[ACTION_STRING_LANGUAGE];
3254 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3257 * Blocking Accept-Language header
3259 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
3266 * Replacing Accept-Language header
3269 *header = strdup("Accept-Language: ");
3270 string_append(header, newval);
3272 if (*header == NULL)
3274 log_error(LOG_LEVEL_ERROR,
3275 "Insufficient memory. Accept-Language header crunched without replacement.");
3279 log_error(LOG_LEVEL_HEADER,
3280 "Accept-Language header crunched and replaced with: %s", *header);
3283 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3287 /*********************************************************************
3289 * Function : crunch_client_header
3291 * Description : Crunch client header if it matches a string supplied by the
3292 * user. Called from `sed'.
3295 * 1 : csp = Current client state (buffers, headers, etc...)
3296 * 2 : header = On input, pointer to header to modify.
3297 * On output, pointer to the modified header, or NULL
3298 * to remove the header. This function frees the
3299 * original string if necessary.
3301 * Returns : JB_ERR_OK on success and always succeeds
3303 *********************************************************************/
3304 static jb_err crunch_client_header(struct client_state *csp, char **header)
3306 const char *crunch_pattern;
3308 /* Do we feel like crunching? */
3309 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
3311 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
3313 /* Is the current header the lucky one? */
3314 if (strstr(*header, crunch_pattern))
3316 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
3324 /*********************************************************************
3326 * Function : client_uagent
3328 * Description : Handle the "user-agent" config setting properly
3329 * and remember its original value to enable browser
3330 * bug workarounds. Called from `sed'.
3333 * 1 : csp = Current client state (buffers, headers, etc...)
3334 * 2 : header = On input, pointer to header to modify.
3335 * On output, pointer to the modified header, or NULL
3336 * to remove the header. This function frees the
3337 * original string if necessary.
3339 * Returns : JB_ERR_OK on success, or
3340 * JB_ERR_MEMORY on out-of-memory error.
3342 *********************************************************************/
3343 static jb_err client_uagent(struct client_state *csp, char **header)
3347 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
3352 newval = csp->action->string[ACTION_STRING_USER_AGENT];
3359 *header = strdup("User-Agent: ");
3360 string_append(header, newval);
3362 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
3364 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3368 /*********************************************************************
3370 * Function : client_ua
3372 * Description : Handle "ua-" headers properly. Called from `sed'.
3375 * 1 : csp = Current client state (buffers, headers, etc...)
3376 * 2 : header = On input, pointer to header to modify.
3377 * On output, pointer to the modified header, or NULL
3378 * to remove the header. This function frees the
3379 * original string if necessary.
3381 * Returns : JB_ERR_OK on success, or
3382 * JB_ERR_MEMORY on out-of-memory error.
3384 *********************************************************************/
3385 static jb_err client_ua(struct client_state *csp, char **header)
3387 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3389 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3397 /*********************************************************************
3399 * Function : client_from
3401 * Description : Handle the "from" config setting properly.
3402 * Called from `sed'.
3405 * 1 : csp = Current client state (buffers, headers, etc...)
3406 * 2 : header = On input, pointer to header to modify.
3407 * On output, pointer to the modified header, or NULL
3408 * to remove the header. This function frees the
3409 * original string if necessary.
3411 * Returns : JB_ERR_OK on success, or
3412 * JB_ERR_MEMORY on out-of-memory error.
3414 *********************************************************************/
3415 static jb_err client_from(struct client_state *csp, char **header)
3419 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3426 newval = csp->action->string[ACTION_STRING_FROM];
3429 * Are we blocking the e-mail address?
3431 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3433 log_error(LOG_LEVEL_HEADER, "crunched From!");
3437 log_error(LOG_LEVEL_HEADER, " modified");
3439 *header = strdup("From: ");
3440 string_append(header, newval);
3442 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3446 /*********************************************************************
3448 * Function : client_send_cookie
3450 * Description : Crunches the "cookie" header if necessary.
3451 * Called from `sed'.
3453 * XXX: Stupid name, doesn't send squat.
3456 * 1 : csp = Current client state (buffers, headers, etc...)
3457 * 2 : header = On input, pointer to header to modify.
3458 * On output, pointer to the modified header, or NULL
3459 * to remove the header. This function frees the
3460 * original string if necessary.
3462 * Returns : JB_ERR_OK on success, or
3463 * JB_ERR_MEMORY on out-of-memory error.
3465 *********************************************************************/
3466 static jb_err client_send_cookie(struct client_state *csp, char **header)
3468 if (csp->action->flags & ACTION_NO_COOKIE_READ)
3470 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3478 /*********************************************************************
3480 * Function : client_x_forwarded
3482 * Description : Handle the "x-forwarded-for" config setting properly,
3483 * also used in the add_client_headers list. Called from `sed'.
3486 * 1 : csp = Current client state (buffers, headers, etc...)
3487 * 2 : header = On input, pointer to header to modify.
3488 * On output, pointer to the modified header, or NULL
3489 * to remove the header. This function frees the
3490 * original string if necessary.
3492 * Returns : JB_ERR_OK on success, or
3493 * JB_ERR_MEMORY on out-of-memory error.
3495 *********************************************************************/
3496 jb_err client_x_forwarded(struct client_state *csp, char **header)
3498 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
3500 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
3502 if (0 == strcmpic(parameter, "block"))
3505 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3507 else if (0 == strcmpic(parameter, "add"))
3509 string_append(header, ", ");
3510 string_append(header, csp->ip_addr_str);
3512 if (*header == NULL)
3514 return JB_ERR_MEMORY;
3516 log_error(LOG_LEVEL_HEADER,
3517 "Appended client IP address to %s", *header);
3518 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
3522 log_error(LOG_LEVEL_FATAL,
3523 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
3531 /*********************************************************************
3533 * Function : client_max_forwards
3535 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3536 * from the value of the Max-Forwards header field.
3539 * 1 : csp = Current client state (buffers, headers, etc...)
3540 * 2 : header = On input, pointer to header to modify.
3541 * On output, pointer to the modified header, or NULL
3542 * to remove the header. This function frees the
3543 * original string if necessary.
3545 * Returns : JB_ERR_OK on success, or
3546 * JB_ERR_MEMORY on out-of-memory error.
3548 *********************************************************************/
3549 static jb_err client_max_forwards(struct client_state *csp, char **header)
3553 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3554 (0 == strcmpic(csp->http->gpc, "options")))
3556 assert(*(*header+12) == ':');
3557 if (1 == sscanf(*header+12, ": %u", &max_forwards))
3559 if (max_forwards > 0)
3561 snprintf(*header, strlen(*header)+1, "Max-Forwards: %u", --max_forwards);
3562 log_error(LOG_LEVEL_HEADER, "Max-Forwards value for %s request reduced to %u.",
3563 csp->http->gpc, max_forwards);
3565 else if (max_forwards < 0)
3567 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3573 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3582 /*********************************************************************
3584 * Function : client_host
3586 * Description : If the request URI did not contain host and
3587 * port information, parse and evaluate the Host
3590 * Also, kill ill-formed HOST: headers as sent by
3591 * Apple's iTunes software when used with a proxy.
3594 * 1 : csp = Current client state (buffers, headers, etc...)
3595 * 2 : header = On input, pointer to header to modify.
3596 * On output, pointer to the modified header, or NULL
3597 * to remove the header. This function frees the
3598 * original string if necessary.
3600 * Returns : JB_ERR_OK on success, or
3601 * JB_ERR_MEMORY on out-of-memory error.
3603 *********************************************************************/
3604 static jb_err client_host(struct client_state *csp, char **header)
3609 * If the header field name is all upper-case, chances are that it's
3610 * an ill-formed one from iTunes. BTW, killing innocent headers here is
3611 * not a problem -- they are regenerated later.
3613 if ((*header)[1] == 'O')
3615 log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
3620 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3621 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3624 if (NULL == (p = strdup((*header)+6)))
3626 return JB_ERR_MEMORY;
3629 if (NULL == (q = strdup(p)))
3632 return JB_ERR_MEMORY;
3635 freez(csp->http->hostport);
3636 csp->http->hostport = p;
3637 freez(csp->http->host);
3638 csp->http->host = q;
3639 q = strchr(csp->http->host, ':');
3642 /* Terminate hostname and evaluate port string */
3644 csp->http->port = atoi(q);
3648 csp->http->port = csp->http->ssl ? 443 : 80;
3651 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3652 csp->http->hostport, csp->http->host, csp->http->port);
3655 /* Signal client_host_adder() to return right away */
3656 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3662 /*********************************************************************
3664 * Function : client_if_modified_since
3666 * Description : Remove or modify the If-Modified-Since header.
3669 * 1 : csp = Current client state (buffers, headers, etc...)
3670 * 2 : header = On input, pointer to header to modify.
3671 * On output, pointer to the modified header, or NULL
3672 * to remove the header. This function frees the
3673 * original string if necessary.
3675 * Returns : JB_ERR_OK on success, or
3676 * JB_ERR_MEMORY on out-of-memory error.
3678 *********************************************************************/
3679 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3682 #ifdef HAVE_GMTIME_R
3685 struct tm *timeptr = NULL;
3689 long int hours, minutes, seconds;
3693 if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3696 * The client got an error message because of a temporary problem,
3697 * the problem is gone and the client now tries to revalidate our
3698 * error message on the real server. The revalidation would always
3699 * end with the transmission of the whole document and there is
3700 * no need to expose the bogus If-Modified-Since header.
3702 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3705 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3707 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3709 if ((0 == strcmpic(newval, "block")))
3711 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3714 else /* add random value */
3716 const char *header_time = *header + sizeof("If-Modified-Since:");
3718 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3720 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3725 rtime = strtol(newval, &endptr, 0);
3728 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3729 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3736 rtime = pick_from_range(rtime);
3740 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3743 tm += rtime * (negative ? -1 : 1);
3744 #ifdef HAVE_GMTIME_R
3745 timeptr = gmtime_r(&tm, &gmt);
3746 #elif FEATURE_PTHREAD
3747 privoxy_mutex_lock(&gmtime_mutex);
3748 timeptr = gmtime(&tm);
3749 privoxy_mutex_unlock(&gmtime_mutex);
3751 timeptr = gmtime(&tm);
3753 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
3756 *header = strdup("If-Modified-Since: ");
3757 string_append(header, newheader);
3759 if (*header == NULL)
3761 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3762 return JB_ERR_MEMORY;
3765 hours = rtime / 3600;
3766 minutes = rtime / 60 % 60;
3767 seconds = rtime % 60;
3769 log_error(LOG_LEVEL_HEADER,
3770 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3771 *header, (negative) ? "subtracted" : "added", hours,
3772 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3773 seconds, (seconds == 1) ? ")" : "s)");
3782 /*********************************************************************
3784 * Function : client_if_none_match
3786 * Description : Remove the If-None-Match header.
3789 * 1 : csp = Current client state (buffers, headers, etc...)
3790 * 2 : header = On input, pointer to header to modify.
3791 * On output, pointer to the modified header, or NULL
3792 * to remove the header. This function frees the
3793 * original string if necessary.
3795 * Returns : JB_ERR_OK on success, or
3796 * JB_ERR_MEMORY on out-of-memory error.
3798 *********************************************************************/
3799 static jb_err client_if_none_match(struct client_state *csp, char **header)
3801 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3803 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3811 /*********************************************************************
3813 * Function : client_x_filter
3815 * Description : Disables filtering if the client set "X-Filter: No".
3816 * Called from `sed'.
3819 * 1 : csp = Current client state (buffers, headers, etc...)
3820 * 2 : header = On input, pointer to header to modify.
3821 * On output, pointer to the modified header, or NULL
3822 * to remove the header. This function frees the
3823 * original string if necessary.
3825 * Returns : JB_ERR_OK on success
3827 *********************************************************************/
3828 jb_err client_x_filter(struct client_state *csp, char **header)
3830 if ( 0 == strcmpic(*header, "X-Filter: No"))
3832 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3834 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3838 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3840 log_error(LOG_LEVEL_HEADER,
3841 "force-text-mode overruled the client's request to fetch without filtering!");
3845 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3846 csp->flags |= CSP_FLAG_NO_FILTERING;
3847 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3849 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3857 /*********************************************************************
3859 * Function : client_range
3861 * Description : Removes Range, Request-Range and If-Range headers if
3862 * content filtering is enabled. If the client's version
3863 * of the document has been altered by Privoxy, the server
3864 * could interpret the range differently than the client
3865 * intended in which case the user could end up with
3866 * corrupted content.
3869 * 1 : csp = Current client state (buffers, headers, etc...)
3870 * 2 : header = On input, pointer to header to modify.
3871 * On output, pointer to the modified header, or NULL
3872 * to remove the header. This function frees the
3873 * original string if necessary.
3875 * Returns : JB_ERR_OK
3877 *********************************************************************/
3878 static jb_err client_range(struct client_state *csp, char **header)
3880 if (content_filters_enabled(csp->action))
3882 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3883 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3890 /* the following functions add headers directly to the header list */
3892 /*********************************************************************
3894 * Function : client_host_adder
3896 * Description : Adds the Host: header field if it is missing.
3897 * Called from `sed'.
3900 * 1 : csp = Current client state (buffers, headers, etc...)
3902 * Returns : JB_ERR_OK on success, or
3903 * JB_ERR_MEMORY on out-of-memory error.
3905 *********************************************************************/
3906 static jb_err client_host_adder(struct client_state *csp)
3911 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3913 /* Header already set by the client, nothing to do. */
3917 if ( !csp->http->hostport || !*(csp->http->hostport))
3919 /* XXX: When does this happen and why is it OK? */
3920 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3925 * remove 'user:pass@' from 'proto://user:pass@host'
3927 if ( (p = strchr( csp->http->hostport, '@')) != NULL )
3933 p = csp->http->hostport;
3936 /* XXX: Just add it, we already made sure that it will be unique */
3937 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3938 err = enlist_unique_header(csp->headers, "Host", p);
3945 /*********************************************************************
3947 * Function : client_accept_encoding_adder
3949 * Description : Add an Accept-Encoding header to the client's request
3950 * that disables compression if the action applies, and
3951 * the header is not already there. Called from `sed'.
3952 * Note: For HTTP/1.0, the absence of the header is enough.
3955 * 1 : csp = Current client state (buffers, headers, etc...)
3957 * Returns : JB_ERR_OK on success, or
3958 * JB_ERR_MEMORY on out-of-memory error.
3960 *********************************************************************/
3961 static jb_err client_accept_encoding_adder(struct client_state *csp)
3963 if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3964 && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
3966 return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
3974 /*********************************************************************
3976 * Function : client_xtra_adder
3978 * Description : Used in the add_client_headers list. Called from `sed'.
3981 * 1 : csp = Current client state (buffers, headers, etc...)
3983 * Returns : JB_ERR_OK on success, or
3984 * JB_ERR_MEMORY on out-of-memory error.
3986 *********************************************************************/
3987 static jb_err client_xtra_adder(struct client_state *csp)
3989 struct list_entry *lst;
3992 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3993 lst ; lst = lst->next)
3995 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3996 err = enlist(csp->headers, lst->str);
4008 /*********************************************************************
4010 * Function : client_x_forwarded_for_adder
4012 * Description : Used in the add_client_headers list. Called from `sed'.
4015 * 1 : csp = Current client state (buffers, headers, etc...)
4017 * Returns : JB_ERR_OK on success, or
4018 * JB_ERR_MEMORY on out-of-memory error.
4020 *********************************************************************/
4021 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
4023 char *header = NULL;
4026 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
4027 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
4028 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
4031 * If we aren't adding X-Forwarded-For headers,
4032 * or we already appended an existing X-Forwarded-For
4033 * header, there's nothing left to do here.
4038 header = strdup("X-Forwarded-For: ");
4039 string_append(&header, csp->ip_addr_str);
4043 return JB_ERR_MEMORY;
4046 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
4047 err = enlist(csp->headers, header);
4054 /*********************************************************************
4056 * Function : server_connection_close_adder
4058 * Description : "Temporary" fix for the needed but missing HTTP/1.1
4059 * support. Adds a "Connection: close" header to csp->headers
4060 * unless the header was already present. Called from `sed'.
4062 * FIXME: This whole function shouldn't be neccessary!
4065 * 1 : csp = Current client state (buffers, headers, etc...)
4067 * Returns : JB_ERR_OK on success, or
4068 * JB_ERR_MEMORY on out-of-memory error.
4070 *********************************************************************/
4071 static jb_err server_connection_close_adder(struct client_state *csp)
4073 const unsigned int flags = csp->flags;
4075 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
4076 && (flags & CSP_FLAG_SERVER_CONNECTION_CLOSE_SET))
4081 log_error(LOG_LEVEL_HEADER, "Adding: Connection: close");
4083 return enlist(csp->headers, "Connection: close");
4087 /*********************************************************************
4089 * Function : client_connection_header_adder
4091 * Description : Adds a proper "Connection:" header to csp->headers
4092 * unless the header was already present. Called from `sed'.
4095 * 1 : csp = Current client state (buffers, headers, etc...)
4097 * Returns : JB_ERR_OK on success, or
4098 * JB_ERR_MEMORY on out-of-memory error.
4100 *********************************************************************/
4101 static jb_err client_connection_header_adder(struct client_state *csp)
4103 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
4104 static const char wanted_header[] = "Connection: keep-alive";
4106 static const char wanted_header[] = "Connection: close";
4107 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
4108 const unsigned int flags = csp->flags;
4110 if (!(flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
4111 && (flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
4116 log_error(LOG_LEVEL_HEADER, "Adding: %s", wanted_header);
4118 return enlist(csp->headers, wanted_header);
4122 /*********************************************************************
4124 * Function : server_http
4126 * Description : - Save the HTTP Status into csp->http->status
4127 * - Set CT_TABOO to prevent filtering if the answer
4128 * is a partial range (HTTP status 206)
4129 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
4133 * 1 : csp = Current client state (buffers, headers, etc...)
4134 * 2 : header = On input, pointer to header to modify.
4135 * On output, pointer to the modified header, or NULL
4136 * to remove the header. This function frees the
4137 * original string if necessary.
4139 * Returns : JB_ERR_OK on success, or
4140 * JB_ERR_MEMORY on out-of-memory error.
4142 *********************************************************************/
4143 static jb_err server_http(struct client_state *csp, char **header)
4145 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
4146 if (csp->http->status == 206)
4148 csp->content_type = CT_TABOO;
4151 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
4153 /* XXX: Should we do a real validity check here? */
4154 if (strlen(*header) > 8)
4157 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
4162 * XXX: Should we block the request or
4163 * enlist a valid status code line here?
4165 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
4166 "Downgrading to HTTP/1.0 impossible.");
4174 /*********************************************************************
4176 * Function : server_set_cookie
4178 * Description : Handle the server "cookie" header properly.
4179 * Log cookie to the jar file. Then "crunch",
4180 * accept or rewrite it to a session cookie.
4181 * Called from `sed'.
4183 * TODO: Allow the user to specify a new expiration
4184 * time to cause the cookie to expire even before the
4185 * browser is closed.
4188 * 1 : csp = Current client state (buffers, headers, etc...)
4189 * 2 : header = On input, pointer to header to modify.
4190 * On output, pointer to the modified header, or NULL
4191 * to remove the header. This function frees the
4192 * original string if necessary.
4194 * Returns : JB_ERR_OK on success, or
4195 * JB_ERR_MEMORY on out-of-memory error.
4197 *********************************************************************/
4198 static jb_err server_set_cookie(struct client_state *csp, char **header)
4205 if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
4207 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
4210 else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
4212 /* Flag whether or not to log a message */
4215 /* A variable to store the tag we're working on */
4218 /* Skip "Set-Cookie:" (11 characters) in header */
4219 cur_tag = *header + 11;
4221 /* skip whitespace between "Set-Cookie:" and value */
4222 while (*cur_tag && ijb_isspace(*cur_tag))
4227 /* Loop through each tag in the cookie */
4231 char *next_tag = strchr(cur_tag, ';');
4232 if (next_tag != NULL)
4234 /* Skip the ';' character itself */
4237 /* skip whitespace ";" and start of tag */
4238 while (*next_tag && ijb_isspace(*next_tag))
4245 /* "Next tag" is the end of the string */
4246 next_tag = cur_tag + strlen(cur_tag);
4250 * Check the expiration date to see
4251 * if the cookie is still valid, if yes,
4252 * rewrite it to a session cookie.
4254 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
4256 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
4258 /* Did we detect the date properly? */
4259 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
4262 * Nope, treat it as if it was still valid.
4264 * XXX: Should we remove the whole cookie instead?
4266 log_error(LOG_LEVEL_ERROR,
4267 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
4268 string_move(cur_tag, next_tag);
4274 * Yes. Check if the cookie is still valid.
4276 * If the cookie is already expired it's probably
4277 * a delete cookie and even if it isn't, the browser
4278 * will discard it anyway.
4282 * XXX: timegm() isn't available on some AmigaOS
4283 * versions and our replacement doesn't work.
4285 * Our options are to either:
4287 * - disable session-cookies-only completely if timegm
4290 * - to simply remove all expired tags, like it has
4291 * been done until Privoxy 3.0.6 and to live with
4292 * the consequence that it can cause login/logout
4293 * problems on servers that don't validate their
4294 * input properly, or
4296 * - to replace it with mktime in which
4297 * case there is a slight chance of valid cookies
4298 * passing as already expired.
4300 * This is the way it's currently done and it's not
4301 * as bad as it sounds. If the missing GMT offset is
4302 * enough to change the result of the expiration check
4303 * the cookie will be only valid for a few hours
4304 * anyway, which in many cases will be shorter
4305 * than a browser session.
4307 if (cookie_time - now < 0)
4309 log_error(LOG_LEVEL_HEADER,
4310 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4311 /* Just in case some clown sets more then one expiration date */
4317 * Still valid, delete expiration date by copying
4318 * the rest of the string over it.
4320 string_move(cur_tag, next_tag);
4322 /* That changed the header, need to issue a log message */
4326 * Note that the next tag has now been moved to *cur_tag,
4327 * so we do not need to update the cur_tag pointer.
4335 /* Move on to next cookie tag */
4342 assert(NULL != *header);
4343 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
4352 #ifdef FEATURE_FORCE_LOAD
4353 /*********************************************************************
4355 * Function : strclean
4357 * Description : In-Situ-Eliminate all occurances of substring in
4361 * 1 : string = string to clean
4362 * 2 : substring = substring to eliminate
4364 * Returns : Number of eliminations
4366 *********************************************************************/
4367 int strclean(char *string, const char *substring)
4373 len = strlen(substring);
4375 while((pos = strstr(string, substring)) != NULL)
4382 while (*p++ != '\0');
4389 #endif /* def FEATURE_FORCE_LOAD */
4392 /*********************************************************************
4394 * Function : parse_header_time
4396 * Description : Parses time formats used in HTTP header strings
4397 * to get the numerical respresentation.
4400 * 1 : header_time = HTTP header time as string.
4401 * 2 : result = storage for header_time in seconds
4403 * Returns : JB_ERR_OK if the time format was recognized, or
4404 * JB_ERR_PARSE otherwise.
4406 *********************************************************************/
4407 static jb_err parse_header_time(const char *header_time, time_t *result)
4412 * Zero out gmt to prevent time zone offsets.
4414 * While this is only necessary on some platforms
4415 * (mingw32 for example), I don't know how to
4416 * detect these automatically and doing it everywhere
4419 memset(&gmt, 0, sizeof(gmt));
4421 /* Tue, 02 Jun 2037 20:00:00 */
4422 if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
4423 /* Tue, 02-Jun-2037 20:00:00 */
4424 && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
4425 /* Tue, 02-Jun-37 20:00:00 */
4426 && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
4427 /* Tuesday, 02-Jun-2037 20:00:00 */
4428 && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
4429 /* Tuesday Jun 02 20:00:00 2037 */
4430 && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
4432 return JB_ERR_PARSE;
4435 *result = timegm(&gmt);
4442 /*********************************************************************
4444 * Function : get_destination_from_headers
4446 * Description : Parse the "Host:" header to get the request's destination.
4447 * Only needed if the client's request was forcefully
4448 * redirected into Privoxy.
4450 * Code mainly copied from client_host() which is currently
4451 * run too late for this purpose.
4454 * 1 : headers = List of headers (one of them hopefully being
4455 * the "Host:" header)
4456 * 2 : http = storage for the result (host, port and hostport).
4458 * Returns : JB_ERR_MEMORY in case of memory problems,
4459 * JB_ERR_PARSE if the host header couldn't be found,
4460 * JB_ERR_OK otherwise.
4462 *********************************************************************/
4463 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4469 host = get_header_value(headers, "Host:");
4473 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4474 return JB_ERR_PARSE;
4477 if (NULL == (p = strdup((host))))
4479 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4480 return JB_ERR_MEMORY;
4483 if (NULL == (q = strdup(p)))
4486 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4487 return JB_ERR_MEMORY;
4490 freez(http->hostport);
4494 q = strchr(http->host, ':');
4497 /* Terminate hostname and evaluate port string */
4499 http->port = atoi(q);
4503 http->port = http->ssl ? 443 : 80;
4506 /* Rebuild request URL */
4508 http->url = strdup(http->ssl ? "https://" : "http://");
4509 string_append(&http->url, http->hostport);
4510 string_append(&http->url, http->path);
4511 if (http->url == NULL)
4513 return JB_ERR_MEMORY;
4516 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4524 /*********************************************************************
4526 * Function : create_forged_referrer
4528 * Description : Helper for client_referrer to forge a referer as
4529 * 'http://[hostname:port/' to fool stupid
4530 * checks for in-site links
4533 * 1 : header = Pointer to header pointer
4534 * 2 : hostport = Host and optionally port as string
4536 * Returns : JB_ERR_OK in case of success, or
4537 * JB_ERR_MEMORY in case of memory problems.
4539 *********************************************************************/
4540 static jb_err create_forged_referrer(char **header, const char *hostport)
4542 assert(NULL == *header);
4544 *header = strdup("Referer: http://");
4545 string_append(header, hostport);
4546 string_append(header, "/");
4548 if (NULL == *header)
4550 return JB_ERR_MEMORY;
4553 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4560 /*********************************************************************
4562 * Function : create_fake_referrer
4564 * Description : Helper for client_referrer to create a fake referrer
4565 * based on a string supplied by the user.
4568 * 1 : header = Pointer to header pointer
4569 * 2 : hosthost = Referrer to fake
4571 * Returns : JB_ERR_OK in case of success, or
4572 * JB_ERR_MEMORY in case of memory problems.
4574 *********************************************************************/
4575 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4577 assert(NULL == *header);
4579 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4581 log_error(LOG_LEVEL_HEADER,
4582 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4584 *header = strdup("Referer: ");
4585 string_append(header, fake_referrer);
4587 if (NULL == *header)
4589 return JB_ERR_MEMORY;
4592 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4599 /*********************************************************************
4601 * Function : handle_conditional_hide_referrer_parameter
4603 * Description : Helper for client_referrer to crunch or forge
4604 * the referrer header if the host has changed.
4607 * 1 : header = Pointer to header pointer
4608 * 2 : host = The target host (may include the port)
4609 * 3 : parameter_conditional_block = Boolean to signal
4610 * if we're in conditional-block mode. If not set,
4611 * we're in conditional-forge mode.
4613 * Returns : JB_ERR_OK in case of success, or
4614 * JB_ERR_MEMORY in case of memory problems.
4616 *********************************************************************/
4617 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4618 const char *host, const int parameter_conditional_block)
4620 char *referer = strdup(*header);
4621 const size_t hostlenght = strlen(host);
4623 if (NULL == referer)
4626 return JB_ERR_MEMORY;
4629 /* referer begins with 'Referer: http[s]://' */
4630 if (hostlenght < (strlen(referer)-17))
4633 * Shorten referer to make sure the referer is blocked
4634 * if www.example.org/www.example.com-shall-see-the-referer/
4635 * links to www.example.com/
4637 referer[hostlenght+17] = '\0';
4639 if (NULL == strstr(referer, host))
4641 /* Host has changed */
4642 if (parameter_conditional_block)
4644 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4651 return create_forged_referrer(header, host);