X-Git-Url: http://www.privoxy.org/gitweb/?p=privoxy.git;a=blobdiff_plain;f=parsers.c;h=92d1567f9985dfa00663d1a8f48b79ca96efbce4;hp=2b2043c32e0ba64216b020f2628ffa5e3315bd8c;hb=3445ac73d9a18f4df13df1f40682a884ccebd572;hpb=b28e9205bb0d4ac98d591c6e506219438799d56c diff --git a/parsers.c b/parsers.c index 2b2043c3..92d1567f 100644 --- a/parsers.c +++ b/parsers.c @@ -1,4 +1,4 @@ -const char parsers_rcs[] = "$Id: parsers.c,v 1.16 2001/06/29 13:32:42 oes Exp $"; +const char parsers_rcs[] = "$Id: parsers.c,v 1.71 2006/09/21 19:55:17 fabiankeil Exp $"; /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $ @@ -10,18 +10,17 @@ const char parsers_rcs[] = "$Id: parsers.c,v 1.16 2001/06/29 13:32:42 oes Exp $" * `client_uagent', `client_x_forwarded', * `client_x_forwarded_adder', `client_xtra_adder', * `content_type', `crumble', `destroy_list', `enlist', - * `flush_socket', `free_http_request', `get_header', - * `list_to_text', `parse_http_request', `sed', + * `flush_socket', ``get_header', `sed', * and `server_set_cookie'. * * Copyright : Written by and Copyright (C) 2001 the SourceForge - * IJBSWA team. http://ijbswa.sourceforge.net + * Privoxy team. http://www.privoxy.org/ * * Based on the Internet Junkbuster originally written - * by and Copyright (C) 1997 Anonymous Coders and + * by and Copyright (C) 1997 Anonymous Coders and * Junkbusters Corporation. http://www.junkbusters.com * - * This program is free software; you can redistribute it + * This program is free software; you can redistribute it * and/or modify it under the terms of the GNU General * Public License as published by the Free Software * Foundation; either version 2 of the License, or (at @@ -41,6 +40,328 @@ const char parsers_rcs[] = "$Id: parsers.c,v 1.16 2001/06/29 13:32:42 oes Exp $" * * Revisions : * $Log: parsers.c,v $ + * Revision 1.71 2006/09/21 19:55:17 fabiankeil + * Fix +hide-if-modified-since{-n}. + * + * Revision 1.70 2006/09/08 12:06:34 fabiankeil + * Have hide-if-modified-since interpret the random + * range value as minutes instead of hours. Allows + * more fine-grained configuration. + * + * Revision 1.69 2006/09/06 16:25:51 fabiankeil + * Always have parse_header_time return a pointer + * that actual makes sense, even though we currently + * only need it to detect problems. + * + * Revision 1.68 2006/09/06 10:43:32 fabiankeil + * Added config option enable-remote-http-toggle + * to specify if Privoxy should recognize special + * headers (currently only X-Filter) to change its + * behaviour. Disabled by default. + * + * Revision 1.67 2006/09/04 11:01:26 fabiankeil + * After filtering de-chunked instances, remove + * "Transfer-Encoding" header entirely instead of changing + * it to "Transfer-Encoding: identity", which is invalid. + * Thanks Michael Shields . Fixes PR 1318658. + * + * Don't use localtime in parse_header_time. An empty time struct + * is good enough, it gets overwritten by strptime anyway. + * + * Revision 1.66 2006/09/03 19:38:28 fabiankeil + * Use gmtime_r if available, fallback to gmtime with mutex + * protection for MacOSX and use vanilla gmtime for the rest. + * + * Revision 1.65 2006/08/22 10:55:56 fabiankeil + * Changed client_referrer to use the right type (size_t) for + * hostlenght and to shorten the temporary referrer string with + * '\0' instead of adding a useless line break. + * + * Revision 1.64 2006/08/17 17:15:10 fabiankeil + * - Back to timegm() using GnuPG's replacement if necessary. + * Using mktime() and localtime() could add a on hour offset if + * the randomize factor was big enough to lead to a summer/wintertime + * switch. + * + * - Removed now-useless Privoxy 3.0.3 compatibility glue. + * + * - Moved randomization code into pick_from_range(). + * + * - Changed parse_header_time definition. + * time_t isn't guaranteed to be signed and + * if it isn't, -1 isn't available as error code. + * Changed some variable types in client_if_modified_since() + * because of the same reason. + * + * Revision 1.63 2006/08/14 13:18:08 david__schmidt + * OS/2 compilation compatibility fixups + * + * Revision 1.62 2006/08/14 08:58:42 fabiankeil + * Changed include from strptime.c to strptime.h + * + * Revision 1.61 2006/08/14 08:25:19 fabiankeil + * Split filter-headers{} into filter-client-headers{} + * and filter-server-headers{}. + * Added parse_header_time() to share some code. + * Replaced timegm() with mktime(). + * + * Revision 1.60 2006/08/12 03:54:37 david__schmidt + * Windows service integration + * + * Revision 1.59 2006/08/03 02:46:41 david__schmidt + * Incorporate Fabian Keil's patch work: http://www.fabiankeil.de/sourcecode/privoxy/ + * + * Revision 1.58 2006/07/18 14:48:47 david__schmidt + * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch) + * with what was really the latest development (the v_3_0_branch branch) + * + * Revision 1.56.2.10 2006/01/21 16:16:08 david__schmidt + * Thanks to Edward Carrel for his patch to modernize OSX's pthreads support. See bug #1409623. + * + * Revision 1.56.2.9 2004/10/03 12:53:45 david__schmidt + * Add the ability to check jpeg images for invalid + * lengths of comment blocks. Defensive strategy + * against the exploit: + * Microsoft Security Bulletin MS04-028 + * Buffer Overrun in JPEG Processing (GDI+) Could + * Allow Code Execution (833987) + * Enabled with +inspect-jpegs in actions files. + * + * Revision 1.56.2.8 2003/07/11 13:21:25 oes + * Excluded text/plain objects from filtering. This fixes a + * couple of client-crashing, download corruption and + * Privoxy performance issues, whose root cause lies in + * web servers labelling content of unknown type as text/plain. + * + * Revision 1.56.2.7 2003/05/06 12:07:26 oes + * Fixed bug #729900: Suspicious HOST: headers are now killed and regenerated if necessary + * + * Revision 1.56.2.6 2003/04/14 21:28:30 oes + * Completing the previous change + * + * Revision 1.56.2.5 2003/04/14 12:08:16 oes + * Added temporary workaround for bug in PHP < 4.2.3 + * + * Revision 1.56.2.4 2003/03/07 03:41:05 david__schmidt + * 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. + * + * Revision 1.56.2.3 2002/11/10 04:20:02 hal9 + * Fix typo: supressed -> suppressed + * + * Revision 1.56.2.2 2002/09/25 14:59:53 oes + * Improved cookie logging + * + * Revision 1.56.2.1 2002/09/25 14:52:45 oes + * Added basic support for OPTIONS and TRACE HTTP methods: + * - New parser function client_max_forwards which decrements + * the Max-Forwards HTTP header field of OPTIONS and TRACE + * requests by one before forwarding + * - New parser function client_host which extracts the host + * and port information from the HTTP header field if the + * request URI was not absolute + * - Don't crumble and re-add the Host: header, but only generate + * and append if missing + * + * Revision 1.56 2002/05/12 15:34:22 jongfoster + * Fixing typo in a comment + * + * Revision 1.55 2002/05/08 16:01:07 oes + * Optimized add_to_iob: + * - Use realloc instead of malloc(), memcpy(), free() + * - Expand to powers of two if possible, to get + * O(log n) reallocs instead of O(n). + * - Moved check for buffer limit here from chat + * - Report failure via returncode + * + * Revision 1.54 2002/04/02 15:03:16 oes + * Tiny code cosmetics + * + * Revision 1.53 2002/03/26 22:29:55 swa + * we have a new homepage! + * + * Revision 1.52 2002/03/24 13:25:43 swa + * name change related issues + * + * Revision 1.51 2002/03/13 00:27:05 jongfoster + * Killing warnings + * + * Revision 1.50 2002/03/12 01:45:35 oes + * More verbose logging + * + * Revision 1.49 2002/03/09 20:03:52 jongfoster + * - Making various functions return int rather than size_t. + * (Undoing a recent change). Since size_t is unsigned on + * Windows, functions like read_socket that return -1 on + * error cannot return a size_t. + * + * THIS WAS A MAJOR BUG - it caused frequent, unpredictable + * crashes, and also frequently caused JB to jump to 100% + * CPU and stay there. (Because it thought it had just + * read ((unsigned)-1) == 4Gb of data...) + * + * - The signature of write_socket has changed, it now simply + * returns success=0/failure=nonzero. + * + * - Trying to get rid of a few warnings --with-debug on + * Windows, I've introduced a new type "jb_socket". This is + * used for the socket file descriptors. On Windows, this + * is SOCKET (a typedef for unsigned). Everywhere else, it's + * an int. The error value can't be -1 any more, so it's + * now JB_INVALID_SOCKET (which is -1 on UNIX, and in + * Windows it maps to the #define INVALID_SOCKET.) + * + * - The signature of bind_port has changed. + * + * Revision 1.48 2002/03/07 03:46:53 oes + * Fixed compiler warnings etc + * + * Revision 1.47 2002/02/20 23:15:13 jongfoster + * Parsing functions now handle out-of-memory gracefully by returning + * an error code. + * + * Revision 1.46 2002/01/17 21:03:47 jongfoster + * Moving all our URL and URL pattern parsing code to urlmatch.c. + * + * Revision 1.45 2002/01/09 14:33:03 oes + * Added support for localtime_r. + * + * Revision 1.44 2001/12/14 01:22:54 steudten + * Remove 'user:pass@' from 'proto://user:pass@host' for the + * new added header 'Host: ..'. (See Req ID 491818) + * + * Revision 1.43 2001/11/23 00:26:38 jongfoster + * Fixing two really stupid errors in my previous commit + * + * Revision 1.42 2001/11/22 21:59:30 jongfoster + * Adding code to handle +no-cookies-keep + * + * Revision 1.41 2001/11/05 23:43:05 steudten + * Add time+date to log files. + * + * Revision 1.40 2001/10/26 20:13:09 jongfoster + * ctype.h is needed in Windows, too. + * + * Revision 1.39 2001/10/26 17:40:04 oes + * Introduced get_header_value() + * Removed http->user_agent, csp->referrer and csp->accept_types + * Removed client_accept() + * + * Revision 1.38 2001/10/25 03:40:48 david__schmidt + * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple + * threads to call select() simultaneously. So, it's time to do a real, live, + * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__ + * (native). Both versions will work, but using __OS2__ offers multi-threading. + * + * Revision 1.37 2001/10/23 21:36:02 jongfoster + * Documenting sed()'s error behaviou (doc change only) + * + * Revision 1.36 2001/10/13 12:51:51 joergs + * Removed client_host, (was only required for the old 2.0.2-11 http://noijb. + * force-load), instead crumble Host: and add it (again) in client_host_adder + * (in case we get a HTTP/1.0 request without Host: header and forward it to + * a HTTP/1.1 server/proxy). + * + * Revision 1.35 2001/10/09 22:39:21 jongfoster + * assert.h is also required under Win32, so moving out of #ifndef _WIN32 + * block. + * + * Revision 1.34 2001/10/07 18:50:55 oes + * Added server_content_encoding, renamed server_transfer_encoding + * + * Revision 1.33 2001/10/07 18:04:49 oes + * Changed server_http11 to server_http and its pattern to "HTTP". + * Additional functionality: it now saves the HTTP status into + * csp->http->status and sets CT_TABOO for Status 206 (partial range) + * + * Revision 1.32 2001/10/07 15:43:28 oes + * Removed FEATURE_DENY_GZIP and replaced it with client_accept_encoding, + * client_te and client_accept_encoding_adder, triggered by the new + * +no-compression action. For HTTP/1.1 the Accept-Encoding header is + * changed to allow only identity and chunked, and the TE header is + * crunched. For HTTP/1.0, Accept-Encoding is crunched. + * + * parse_http_request no longer does anything than parsing. The rewriting + * of http->cmd and version mangling are gone. It now also recognizes + * the put and delete methods and saves the url in http->url. Removed + * unused variable. + * + * renamed content_type and content_length to have the server_ prefix + * + * server_content_type now only works if csp->content_type != CT_TABOO + * + * added server_transfer_encoding, which + * - Sets CT_TABOO to prohibit filtering if encoding compresses + * - Raises the CSP_FLAG_CHUNKED flag if Encoding is "chunked" + * - Change from "chunked" to "identity" if body was chunked + * but has been de-chunked for filtering. + * + * added server_content_md5 which crunches any Content-MD5 headers + * if the body was modified. + * + * made server_http11 conditional on +downgrade action + * + * Replaced 6 boolean members of csp with one bitmap (csp->flags) + * + * Revision 1.31 2001/10/05 14:25:02 oes + * Crumble Keep-Alive from Server + * + * Revision 1.30 2001/09/29 12:56:03 joergs + * IJB now changes HTTP/1.1 to HTTP/1.0 in requests and answers. + * + * Revision 1.29 2001/09/24 21:09:24 jongfoster + * Fixing 2 memory leaks that Guy spotted, where the paramater to + * enlist() was not being free()d. + * + * Revision 1.28 2001/09/22 16:32:28 jongfoster + * Removing unused #includes. + * + * Revision 1.27 2001/09/20 15:45:25 steudten + * + * add casting from size_t to int for printf() + * remove local variable shadow s2 + * + * Revision 1.26 2001/09/16 17:05:14 jongfoster + * Removing unused #include showarg.h + * + * Revision 1.25 2001/09/16 13:21:27 jongfoster + * Changes to use new list functions. + * + * Revision 1.24 2001/09/13 23:05:50 jongfoster + * Changing the string paramater to the header parsers a "const". + * + * Revision 1.23 2001/09/12 18:08:19 steudten + * + * In parse_http_request() header rewriting miss the host value, so + * from http://www.mydomain.com the result was just " / " not + * http://www.mydomain.com/ in case we forward. + * + * Revision 1.22 2001/09/10 10:58:53 oes + * Silenced compiler warnings + * + * Revision 1.21 2001/07/31 14:46:00 oes + * - Persistant connections now suppressed + * - sed() no longer appends empty header to csp->headers + * + * Revision 1.20 2001/07/30 22:08:36 jongfoster + * Tidying up #defines: + * - All feature #defines are now of the form FEATURE_xxx + * - Permanently turned off WIN_GUI_EDIT + * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS + * + * Revision 1.19 2001/07/25 17:21:54 oes + * client_uagent now saves copy of User-Agent: header value + * + * Revision 1.18 2001/07/13 14:02:46 oes + * - Included fix to repair broken HTTP requests that + * don't contain a path, not even '/'. + * - Removed all #ifdef PCRS + * - content_type now always inspected and classified as + * text, gif or other. + * - formatting / comments + * + * Revision 1.17 2001/06/29 21:45:41 oes + * Indentation, CRLF->LF, Tab-> Space + * * Revision 1.16 2001/06/29 13:32:42 oes * - Fixed a comment * - Adapted free_http_request @@ -63,7 +384,7 @@ const char parsers_rcs[] = "$Id: parsers.c,v 1.16 2001/06/29 13:32:42 oes Exp $" * CRLF -> LF * * Revision 1.11 2001/05/29 20:11:19 joergs - * '/* inside comment' warning removed. + * '/ * inside comment' warning removed. * * Revision 1.10 2001/05/29 09:50:24 jongfoster * Unified blocklist/imagelist/permissionslist. @@ -189,29 +510,40 @@ const char parsers_rcs[] = "$Id: parsers.c,v 1.16 2001/06/29 13:32:42 oes Exp $" #include "config.h" +#ifndef _WIN32 #include #include +#endif + #include #include +#include #include +#include -#ifndef _WIN32 +#if !defined(_WIN32) && !defined(__OS2__) #include #endif #include "project.h" + +#ifdef OSX_DARWIN +#include +#include "jcc.h" +/* jcc.h is for mutex semapores only */ +#endif /* def OSX_DARWIN */ #include "list.h" #include "parsers.h" #include "encode.h" -#include "filters.h" -#include "loaders.h" -#include "showargs.h" -#include "jcc.h" #include "ssplit.h" #include "errlog.h" #include "jbsockets.h" #include "miscutil.h" -#include "cgi.h" +#include "list.h" + +#ifndef HAVE_STRPTIME +#include "strptime.h" +#endif const char parsers_h_rcs[] = PARSERS_H_VERSION; @@ -221,8 +553,8 @@ const char parsers_h_rcs[] = PARSERS_H_VERSION; * as an array index. Therefore we need to make sure that high-bit * characters generate +ve values, and ideally we also want to make * the argument match the declared parameter type of "int". - * - * Why did they write a character function that can't take a simple + * + * Why did they write a character function that can't take a simple * "char" argument? Doh! */ #define ijb_isupper(__X) isupper((int)(unsigned char)(__X)) @@ -230,53 +562,67 @@ const char parsers_h_rcs[] = PARSERS_H_VERSION; const struct parsers client_patterns[] = { - { "referer:", 8, client_referrer }, + { "referer:", 8, client_referrer }, { "user-agent:", 11, client_uagent }, - { "ua-", 3, client_ua }, - { "from:", 5, client_from }, - { "cookie:", 7, client_send_cookie }, + { "ua-", 3, client_ua }, + { "from:", 5, client_from }, + { "cookie:", 7, client_send_cookie }, { "x-forwarded-for:", 16, client_x_forwarded }, - { "proxy-connection:", 17, crumble }, -#ifdef DENY_GZIP - { "Accept-Encoding: gzip", 21, crumble }, -#endif /* def DENY_GZIP */ -#if defined(DETECT_MSIE_IMAGES) - { "Accept:", 7, client_accept }, -#endif /* defined(DETECT_MSIE_IMAGES) */ -#ifdef FORCE_LOAD + { "Accept-Encoding:", 16, client_accept_encoding }, + { "TE:", 3, client_te }, { "Host:", 5, client_host }, -#endif /* def FORCE_LOAD */ -/* { "if-modified-since:", 18, crumble }, */ - { NULL, 0, NULL } + { "if-modified-since:", 18, client_if_modified_since }, + { "Keep-Alive:", 11, crumble }, + { "connection:", 11, crumble }, + { "proxy-connection:", 17, crumble }, + { "max-forwards:", 13, client_max_forwards }, + { "Accept-Language:", 16, client_accept_language }, + { "if-none-match:", 14, client_if_none_match }, + { "X-Filter:", 9, client_x_filter }, + { "*", 0, crunch_client_header }, + { "*", 0, filter_client_header }, + { NULL, 0, NULL } }; - const struct parsers server_patterns[] = { - { "set-cookie:", 11, server_set_cookie }, - { "connection:", 11, crumble }, -#if defined(PCRS) || defined(KILLPOPUPS) - { "Content-Type:", 13, content_type }, -#endif /* defined(PCRS) || defined(KILLPOPUPS) */ -#ifdef PCRS - { "Content-Length:", 15, content_length }, -#endif /* def PCRS */ + { "HTTP", 4, server_http }, + { "set-cookie:", 11, server_set_cookie }, + { "connection:", 11, crumble }, + { "Content-Type:", 13, server_content_type }, + { "Content-Length:", 15, server_content_length }, + { "Content-MD5:", 12, server_content_md5 }, + { "Content-Encoding:", 17, server_content_encoding }, + { "Transfer-Encoding:", 18, server_transfer_coding }, + { "Keep-Alive:", 11, crumble }, + { "content-disposition:", 20, server_content_disposition }, + { "Last-Modified:", 14, server_last_modified }, + { "*", 0, crunch_server_header }, + { "*", 0, filter_server_header }, { NULL, 0, NULL } }; +const struct parsers server_patterns_light[] = { + { "Content-Length:", 15, server_content_length }, + { "Transfer-Encoding:", 18, server_transfer_coding }, + { NULL, 0, NULL } +}; -void (* const add_client_headers[])(struct client_state *) = { +const add_header_func_ptr add_client_headers[] = { + client_host_adder, client_cookie_adder, client_x_forwarded_adder, client_xtra_adder, + /* Temporarily disabled: client_accept_encoding_adder, */ + connection_close_adder, NULL }; -void (* const add_server_headers[])(struct client_state *) = { +const add_header_func_ptr add_server_headers[] = { + connection_close_adder, NULL }; - /********************************************************************* * * Function : flush_socket @@ -295,19 +641,22 @@ void (* const add_server_headers[])(struct client_state *) = { * file, the results are not portable. * *********************************************************************/ -int flush_socket(int fd, struct client_state *csp) +int flush_socket(jb_socket fd, struct client_state *csp) { struct iob *iob = csp->iob; - int n = iob->eod - iob->cur; + int len = iob->eod - iob->cur; - if (n <= 0) + if (len <= 0) { return(0); } - n = write_socket(fd, iob->cur, n); + if (write_socket(fd, iob->cur, (size_t)len)) + { + return(-1); + } iob->eod = iob->cur = iob->buf; - return(n); + return(len); } @@ -316,70 +665,74 @@ int flush_socket(int fd, struct client_state *csp) * * Function : add_to_iob * - * Description : Add content to the buffered page. + * Description : Add content to the buffered page, expanding the + * buffer if necessary. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * 2 : buf = holds the content to be added to the page * 3 : n = number of bytes to be added * - * Returns : Number of bytes in the content buffer. + * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory + * or buffer limit reached. * *********************************************************************/ -int add_to_iob(struct client_state *csp, char *buf, int n) +jb_err add_to_iob(struct client_state *csp, char *buf, int n) { struct iob *iob = csp->iob; - int have, need; + size_t used, offset, need, want; char *p; - have = iob->eod - iob->cur; - - if (n <= 0) - { - return(have); - } + if (n <= 0) return JB_ERR_OK; - need = have + n; + used = iob->eod - iob->buf; + offset = iob->cur - iob->buf; + need = used + n + 1; - if ((p = (char *)malloc(need + 1)) == NULL) + /* + * If the buffer can't hold the new data, extend it first. + * Use the next power of two if possible, else use the actual need. + */ + if (need > csp->config->buffer_limit) { - log_error(LOG_LEVEL_ERROR, "malloc() iob failed: %E"); - return(-1); + log_error(LOG_LEVEL_ERROR, "Buffer limit reached while extending the buffer (iob)"); + return JB_ERR_MEMORY; } - if (have) + if (need > iob->size) { - /* there is something in the buffer - save it */ - memcpy(p, iob->cur, have); - - /* replace the buffer with the new space */ - freez(iob->buf); - iob->buf = p; + for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2; + + if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want))) + { + iob->size = want; + } + else if (NULL != (p = (char *)realloc(iob->buf, need))) + { + iob->size = need; + } + else + { + log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E"); + return JB_ERR_MEMORY; + } - /* point to the end of the data */ - p += have; - } - else - { - /* the buffer is empty, free it and reinitialize */ - freez(iob->buf); + /* Update the iob pointers */ + iob->cur = p + offset; + iob->eod = p + used; iob->buf = p; } /* copy the new data into the iob buffer */ - memcpy(p, buf, n); + memcpy(iob->eod, buf, (size_t)n); /* point to the end of the data */ - p += n; + iob->eod += n; /* null terminate == cheap insurance */ - *p = '\0'; - - /* set the pointers to the new values */ - iob->cur = iob->buf; - iob->eod = p; + *iob->eod = '\0'; - return(need); + return JB_ERR_OK; } @@ -416,12 +769,17 @@ char *get_header(struct client_state *csp) *p = '\0'; ret = strdup(iob->cur); + if (ret == NULL) + { + /* FIXME No way to handle error properly */ + log_error(LOG_LEVEL_FATAL, "Out of memory in get_header()"); + } iob->cur = p+1; - if ((q = strchr(ret, '\r'))) *q = '\0'; + if ((q = strchr(ret, '\r')) != NULL) *q = '\0'; - /* is this a blank linke (i.e. the end of the header) ? */ + /* is this a blank line (i.e. the end of the header) ? */ if (*ret == '\0') { freez(ret); @@ -433,6 +791,56 @@ char *get_header(struct client_state *csp) } +/********************************************************************* + * + * Function : get_header_value + * + * Description : Get the value of a given header from a chained list + * of header lines or return NULL if no such header is + * present in the list. + * + * Parameters : + * 1 : header_list = pointer to list + * 2 : header_name = string with name of header to look for. + * Trailing colon required, capitalization + * doesn't matter. + * + * Returns : NULL if not found, else value of header + * + *********************************************************************/ +char *get_header_value(const struct list *header_list, const char *header_name) +{ + struct list_entry *cur_entry; + char *ret = NULL; + size_t length = 0; + + assert(header_list); + assert(header_name); + length = strlen(header_name); + + for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next) + { + if (cur_entry->str) + { + if (!strncmpic(cur_entry->str, header_name, length)) + { + /* + * Found: return pointer to start of value + */ + ret = (char *) (cur_entry->str + length); + while (*ret && ijb_isspace(*ret)) ret++; + return(ret); + } + } + } + + /* + * Not found + */ + return NULL; + +} + /********************************************************************* * * Function : sed @@ -451,230 +859,268 @@ char *get_header(struct client_state *csp) * headers (client or server) * 3 : csp = Current client state (buffers, headers, etc...) * - * Returns : Single pointer to a fully formed header. + * Returns : Single pointer to a fully formed header, or NULL + * on out-of-memory error. * *********************************************************************/ -char *sed(const struct parsers pats[], void (* const more_headers[])(struct client_state *), struct client_state *csp) +char *sed(const struct parsers pats[], + const add_header_func_ptr more_headers[], + struct client_state *csp) { - struct list *p; + struct list_entry *p; const struct parsers *v; - char *hdr; - void (* const *f)(); + const add_header_func_ptr *f; + jb_err err = JB_ERR_OK; + int first_run; - for (v = pats; v->str ; v++) + /* + * If filtering is enabled, sed is run twice, + * but most of the work needs to be done only once. + */ + first_run = (more_headers != NULL ) ? 1 : 0; + + if (first_run) /* Parse and print */ { - for (p = csp->headers->next; p ; p = p->next) + for (v = pats; (err == JB_ERR_OK) && (v->str != NULL) ; v++) { - /* Header crunch()ed in previous run? -> ignore */ - if (p->str == NULL) continue; + for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL) ; p = p->next) + { + /* Header crunch()ed in previous run? -> ignore */ + if (p->str == NULL) continue; - if (v == pats) log_error(LOG_LEVEL_HEADER, "scan: %s", p->str); + if (v == pats) log_error(LOG_LEVEL_HEADER, "scan: %s", p->str); - if (strncmpic(p->str, v->str, v->len) == 0) - { - hdr = v->parser(v, p->str, csp); - freez(p->str); - p->str = hdr; + /* Does the current parser handle this header? */ + if ((strncmpic(p->str, v->str, v->len) == 0) || (v->len == CHECK_EVERY_HEADER_REMAINING)) + { + err = v->parser(csp, (char **)&(p->str)); + } } } + /* place any additional headers on the csp->headers list */ + for (f = more_headers; (err == JB_ERR_OK) && (*f) ; f++) + { + err = (*f)(csp); + } } - - /* place any additional headers on the csp->headers list */ - for (f = more_headers; *f ; f++) + else /* Parse only */ { - (*f)(csp); + /* + * The second run is only needed if the body was modified + * and the content-lenght has changed. + */ + if (strncmpic(csp->http->cmd, "HEAD", 4)) + { + /*XXX: Code duplication*/ + for (v = pats; (err == JB_ERR_OK) && (v->str != NULL) ; v++) + { + for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL) ; p = p->next) + { + /* Header crunch()ed in previous run? -> ignore */ + if (p->str == NULL) continue; + + /* Does the current parser handle this header? */ + if (strncmpic(p->str, v->str, v->len) == 0) + { + err = v->parser(csp, (char **)&(p->str)); + } + } + } + } } - /* add the blank line at the end of the header, if necessary */ - if ( (csp->headers->last == NULL) - || (csp->headers->last->str == NULL) - || (*csp->headers->last->str != '\0') ) + if (err != JB_ERR_OK) { - enlist(csp->headers, ""); + return NULL; } - hdr = list_to_text(csp->headers); - - return(hdr); - + return list_to_text(csp->headers); } +/* here begins the family of parser functions that reformat header lines */ + /********************************************************************* * - * Function : free_http_request + * Function : filter_server_header * - * Description : Freez a http_request structure + * Description : Checks if server header filtering is enabled. + * If it is, filter_header is called to do the work. * * Parameters : - * 1 : http = points to a http_request structure to free + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : N/A + * Returns : JB_ERR_OK on success and always succeeds * *********************************************************************/ -void free_http_request(struct http_request *http) +jb_err filter_server_header(struct client_state *csp, char **header) { - freez(http->cmd); - freez(http->gpc); - freez(http->host); - freez(http->hostport); - freez(http->path); - freez(http->ver); - freez(http->host_ip_addr_str); - + if (csp->action->flags & ACTION_FILTER_SERVER_HEADERS) + { + filter_header(csp, header); + } + return(JB_ERR_OK); } - /********************************************************************* * - * Function : parse_http_request + * Function : filter_client_header * - * Description : Parse out the host and port from the URL. Find the - * hostname & path, port (if ':'), and/or password (if '@') + * Description : Checks if client header filtering is enabled. + * If it is, filter_header is called to do the work. * * Parameters : - * 1 : req = URL (or is it URI?) to break down - * 2 : http = pointer to the http structure to hold elements - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : N/A + * Returns : JB_ERR_OK on success and always succeeds * *********************************************************************/ -void parse_http_request(char *req, struct http_request *http, struct client_state *csp) +jb_err filter_client_header(struct client_state *csp, char **header) { - char *buf, *v[10], *url, *p; - int n; - - memset(http, '\0', sizeof(*http)); - - http->cmd = strdup(req); - - buf = strdup(req); - - n = ssplit(buf, " \r\n", v, SZ(v), 1, 1); - - if (n == 3) + if (csp->action->flags & ACTION_FILTER_CLIENT_HEADERS) { - /* this could be a CONNECT request */ - if (strcmpic(v[0], "connect") == 0) - { - http->ssl = 1; - http->gpc = strdup(v[0]); - http->hostport = strdup(v[1]); - http->ver = strdup(v[2]); - } - -#ifdef WEBDAV - -/* This next line is a little ugly, but it simplifies the if statement below. */ -/* Basically if using webDAV, we want the OR condition to use these too. */ - -/* - * by haroon - * These are the headers as defined in RFC2518 to add webDAV support - */ + filter_header(csp, header); + } + return(JB_ERR_OK); +} -#define OR_WEBDAV || \ - (0 == strcmpic(v[0], "propfind")) || \ - (0 == strcmpic(v[0], "proppatch")) || \ - (0 == strcmpic(v[0], "move")) || \ - (0 == strcmpic(v[0], "copy")) || \ - (0 == strcmpic(v[0], "mkcol")) || \ - (0 == strcmpic(v[0], "lock")) || \ - (0 == strcmpic(v[0], "unlock")) +/********************************************************************* + * + * Function : filter_header + * + * Description : Executes all text substitutions from all applying + * +filter actions on the header. + * Most of the code was copied from pcrs_filter_response, + * including the rather short variable names + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success and always succeeds + * + *********************************************************************/ +jb_err filter_header(struct client_state *csp, char **header) +{ + int hits=0; + int matches; + size_t size = strlen(*header); -#else /* No webDAV support is enabled. Provide an empty OR_WEBDAV macro. */ + char *newheader = NULL; + pcrs_job *job; -#define OR_WEBDAV + struct file_list *fl; + struct re_filterfile_spec *b; + struct list_entry *filtername; -#endif + int i, found_filters = 0; - /* or it could be a GET or a POST (possibly webDAV too) */ - if ((strcmpic(v[0], "get") == 0) || - (strcmpic(v[0], "head") == 0) OR_WEBDAV || - (strcmpic(v[0], "post") == 0)) + /* + * Need to check the set of re_filterfiles... + */ + for (i = 0; i < MAX_AF_FILES; i++) + { + fl = csp->rlist[i]; + if (NULL != fl) { - http->ssl = 0; - http->gpc = strdup(v[0]); - url = v[1]; - http->ver = strdup(v[2]); - - if (strncmpic(url, "http://", 7) == 0) - { - url += 7; - } - else if (strncmpic(url, "https://", 8) == 0) - { - url += 8; - } - else - { - url = NULL; - } - - if (url && (p = strchr(url, '/'))) + if (NULL != fl->f) { - http->path = strdup(p); - *p = '\0'; - http->hostport = strdup(url); + found_filters = 1; + break; } } } - freez(buf); - - - if (http->hostport == NULL) - { - free_http_request(http); - return; - } - - buf = strdup(http->hostport); - - - /* check if url contains password */ - n = ssplit(buf, "@", v, SZ(v), 1, 1); - if (n == 2) - { - char * newbuf = NULL; - newbuf = strdup(v[1]); - freez(buf); - buf = newbuf; - } - - n = ssplit(buf, ":", v, SZ(v), 1, 1); - - if (n == 1) - { - http->host = strdup(v[0]); - http->port = 80; - } - - if (n == 2) + if (0 == found_filters) { - http->host = strdup(v[0]); - http->port = atoi(v[1]); + log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering."); + return(JB_ERR_OK); } - freez(buf); - - if (http->host == NULL) + for (i = 0; i < MAX_AF_FILES; i++) { - free_http_request(http); + fl = csp->rlist[i]; + if ((NULL == fl) || (NULL == fl->f)) + break; + /* + * For all applying +filter actions, look if a filter by that + * name exists and if yes, execute its pcrs_joblist on the + * buffer. + */ + for (b = fl->f; b; b = b->next) + { + for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first; + filtername ; filtername = filtername->next) + { + if (strcmp(b->name, filtername->str) == 0) + { + int current_hits = 0; + + if ( NULL == b->joblist ) + { + log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name); + continue; + } + + log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s (size %d) with filter %s...", + *header, size, b->name); + + /* Apply all jobs from the joblist */ + for (job = b->joblist; NULL != job; job = job->next) + { + matches = pcrs_execute(job, *header, size, &newheader, &size); + if ( 0 < matches ) + { + current_hits += matches; + log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader); + freez(*header); + *header = newheader; + } + else if ( 0 == matches ) + { + /* Filter doesn't change header */ + freez(newheader); + } + else + { + /* RegEx failure */ + log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s", + *header, b->name, pcrs_strerror(matches)); + if( newheader != NULL) + { + log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader); + freez(newheader); + } + } + } + log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size); + hits += current_hits; + } + } + } } - if (http->path == NULL) + if ( 0 == size ) { - http->path = strdup(""); + log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header); + freez(*header); } + return(JB_ERR_OK); } -/* here begins the family of parser functions that reformat header lines */ - - /********************************************************************* * * Function : crumble @@ -682,183 +1128,806 @@ void parse_http_request(char *req, struct http_request *http, struct client_stat * Description : This is called if a header matches a pattern to "crunch" * * Parameters : - * 1 : v = Pointer to parsers structure, which basically holds - * headers (client or server) that we want to "crunch" - * 2 : s = header (from sed) to "crunch" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : Always NULL. + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *crumble(const struct parsers *v, char *s, struct client_state *csp) +jb_err crumble(struct client_state *csp, char **header) { - log_error(LOG_LEVEL_HEADER, "crunch!"); - return(NULL); - + log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header); + freez(*header); + return JB_ERR_OK; } - -#if defined(PCRS) || defined(KILLPOPUPS) - /********************************************************************* * - * Function : content_type + * Function : crunch_server_header * - * Description : Is this a text/.* or javascript MIME Type? + * Description : Crunch server header if it matches a string supplied by the + * user. Called from `sed'. * * Parameters : - * 1 : v = ignored - * 2 : s = header string we are "considering" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : A duplicate string pointer to this header (ie. pass thru) + * Returns : JB_ERR_OK on success and always succeeds * *********************************************************************/ -char *content_type(const struct parsers *v, char *s, struct client_state *csp) +jb_err crunch_server_header(struct client_state *csp, char **header) { - if (strstr (s, " text/") || strstr (s, "application/x-javascript")) - csp->is_text = 1; - else - csp->is_text = 0; + const char *crunch_pattern; + /*Is there a header to crunch*/ - return(strdup(s)); - -} -#endif /* defined(PCRS) || defined(KILLPOPUPS) */ + if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER)) + { + crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER]; + /*Is the current header the lucky one?*/ + if (strstr(*header, crunch_pattern)) + { + log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern); + freez(*header); + } + } -#ifdef PCRS + return JB_ERR_OK; +} /********************************************************************* * - * Function : content_length + * Function : server_content_type * - * Description : Adjust Content-Length header if we have - * filtered this page through PCRS. + * Description : Set the content-type for filterable types (text/.*, + * javascript and image/gif) unless filtering has been + * forbidden (CT_TABOO) while parsing earlier headers. + * NOTE: Since text/plain is commonly used by web servers + * for files whose correct type is unknown, we don't + * set CT_TEXT for it. * * Parameters : - * 1 : v = ignored - * 2 : s = header string we are "considering" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : A duplicate string pointer to this header (ie. pass thru) + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *content_length(const struct parsers *v, char *s, struct client_state *csp) +jb_err server_content_type(struct client_state *csp, char **header) { - if (csp->content_length != 0) /* Content has been modified */ + const char *newval; + + newval = csp->action->string[ACTION_STRING_CONTENT_TYPE]; + + if (csp->content_type != CT_TABOO) { - s = (char *) zalloc(100); - sprintf(s, "Content-Length: %d", csp->content_length); - log_error(LOG_LEVEL_HEADER, "Adjust Content-Length to %d", csp->content_length); - return(s); + if ((strstr(*header, " text/") && !strstr(*header, "plain")) + || strstr(*header, "xml") + || strstr(*header, "application/x-javascript")) + csp->content_type = CT_TEXT; + else if (strstr(*header, " image/gif")) + csp->content_type = CT_GIF; + else if (strstr(*header, " image/jpeg")) + csp->content_type = CT_JPEG; + else + csp->content_type = 0; } - else + /* + * Are we enabling text mode by force? + */ + if (csp->action->flags & ACTION_FORCE_TEXT_MODE) { - return(strdup(s)); + /* + * Do we really have to? + */ + if (csp->content_type == CT_TEXT) + { + log_error(LOG_LEVEL_HEADER, "Text mode is already enabled."); + } + else + { + csp->content_type = CT_TEXT; + log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!"); + } } + /* + * Are we messing with the content type? + */ + if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE) + { + /* + * Make sure the user doesn't accidently + * change the content type of binary documents. + */ + if (csp->content_type == CT_TEXT) + { + freez(*header); + *header = strdup("Content-Type: "); + string_append(header, newval); + + if (header == NULL) + { + log_error(LOG_LEVEL_HEADER, "Insufficient memory. Conten-Type crunched without replacement!"); + return JB_ERR_MEMORY; + } + log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header); + } + else + { + log_error(LOG_LEVEL_HEADER, "%s not replaced. It doesn't look like text. " + "Enable force-text-mode if you know what you're doing.", *header); + } + } + return JB_ERR_OK; } -#endif /* def PCRS */ - /********************************************************************* * - * Function : client_referrer + * Function : server_transfer_coding * - * Description : Handle the "referer" config setting properly. - * Called from `sed'. + * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses + * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked" + * - Remove header if body was chunked but has been + * de-chunked for filtering. * * Parameters : - * 1 : v = ignored - * 2 : s = header (from sed) to "crunch" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : NULL if crunched, or a malloc'ed string with the original - * or modified header + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_referrer(const struct parsers *v, char *s, struct client_state *csp) +jb_err server_transfer_coding(struct client_state *csp, char **header) { - const char * newval; -#ifdef FORCE_LOAD - /* Since the referrer can include the prefix even - * even if the request itself is non-forced, we must - * clean it unconditionally - */ - strclean(s, FORCE_PREFIX); -#endif /* def FORCE_LOAD */ - -#ifdef TRUST_FILES - csp->referrer = strdup(s); -#endif /* def TRUST_FILES */ - /* - * Are we sending referer? + * Turn off pcrs and gif filtering if body compressed */ - if ((csp->action->flags & ACTION_HIDE_REFERER) == 0) + if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate")) { - return(strdup(s)); + csp->content_type = CT_TABOO; } - newval = csp->action->string[ACTION_STRING_REFERER]; - /* - * Are we blocking referer? + * Raise flag if body chunked */ - if ((newval == NULL) || (0 == strcmpic(newval, "block")) ) + if (strstr(*header, "chunked")) { - log_error(LOG_LEVEL_HEADER, "crunch!"); - return(NULL); - } + csp->flags |= CSP_FLAG_CHUNKED; - /* - * Are we forging referer? - */ - if (0 == strcmpic(newval, "forge")) - { /* - * Forge a referer as http://[hostname:port of REQUEST]/ - * to fool stupid checks for in-site links + * If the body was modified, it has been de-chunked first + * and the header must be removed. */ - log_error(LOG_LEVEL_HEADER, "crunch+forge!"); - s = strsav(NULL, "Referer: "); - s = strsav(s, "http://"); - s = strsav(s, csp->http->hostport); - s = strsav(s, "/"); - return(s); + if (csp->flags & CSP_FLAG_MODIFIED) + { + log_error(LOG_LEVEL_HEADER, "Removing: %s", *header); + freez(*header); + } } - /* - * Have we got a fixed referer? - */ - if (0 == strncmpic(newval, "http://", 7)) - { - /* - * We have a specific (fixed) referer we want to send. - */ - - log_error(LOG_LEVEL_HEADER, "modified"); - - s = strsav( NULL, "Referer: " ); - s = strsav( s, newval ); - return(s); - } + return JB_ERR_OK; +} - /* Should never get here! */ - log_error(LOG_LEVEL_ERROR, "Bad parameter: +referer{%s}", newval); - /* - * Forge is probably the best default. - * - * Forge a referer as http://[hostname:port of REQUEST]/ - * to fool stupid checks for in-site links - */ - log_error(LOG_LEVEL_HEADER, "crunch+forge!"); - s = strsav(NULL, "Referer: "); - s = strsav(s, "http://"); - s = strsav(s, csp->http->hostport); - s = strsav(s, "/"); - return(s); +/********************************************************************* + * + * Function : server_content_encoding + * + * Description : Prohibit filtering (CT_TABOO) if content encoding compresses + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err server_content_encoding(struct client_state *csp, char **header) +{ + /* + * Turn off pcrs and gif filtering if body compressed + */ + if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate")) + { + csp->content_type = CT_TABOO; + } + + return JB_ERR_OK; + +} + + +/********************************************************************* + * + * Function : server_content_length + * + * Description : Adjust Content-Length header if we modified + * the body. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err server_content_length(struct client_state *csp, char **header) +{ + if (csp->content_length != 0) /* Content length could have been modified */ + { + /* + * XXX: Shouldn't we check if csp->content_length + * is different than the original value? + */ + freez(*header); + *header = (char *) zalloc(100); + if (*header == NULL) + { + return JB_ERR_MEMORY; + } + + sprintf(*header, "Content-Length: %d", (int) csp->content_length); + + log_error(LOG_LEVEL_HEADER, "Adjust Content-Length to %d", (int) csp->content_length); + } + + return JB_ERR_OK; +} + + +/********************************************************************* + * + * Function : server_content_md5 + * + * Description : Crumble any Content-MD5 headers if the document was + * modified. FIXME: Should we re-compute instead? + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err server_content_md5(struct client_state *csp, char **header) +{ + if (csp->flags & CSP_FLAG_MODIFIED) + { + log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5"); + freez(*header); + } + + return JB_ERR_OK; +} + +/********************************************************************* + * + * Function : server_content_disposition + * + * Description : If enabled, blocks or modifies the "content-disposition" header. + * Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err server_content_disposition(struct client_state *csp, char **header) +{ + const char *newval; + + /* + * Are we messing with the content-disposition header? + */ + if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0) + { + /*Me tinks not*/ + return JB_ERR_OK; + } + + newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION]; + + if ((newval == NULL) || (0 == strcmpic(newval, "block")) ) + { + /* + * Blocking content-disposition header + */ + log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header); + freez(*header); + return JB_ERR_OK; + } + else + { + /* + * Replacing content-disposition header + */ + freez(*header); + *header = strdup("content-disposition: "); + string_append(header, newval); + + if (*header == NULL) + { + log_error(LOG_LEVEL_HEADER, "Insufficent memory. content-disposition header not fully replaced."); + } + else + { + log_error(LOG_LEVEL_HEADER, "content-disposition header crunched and replaced with: %s", *header); + } + } + return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK; +} + +/********************************************************************* + * + * Function : server_last_modified + * + * Description : Changes Last-Modified header to the actual date + * to help hide-if-modified-since. + * Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err server_last_modified(struct client_state *csp, char **header) +{ + const char *newval; + char buf[BUFFER_SIZE]; + + char newheader[50]; +#ifdef HAVE_GMTIME_R + struct tm gmt; +#endif + struct tm *timeptr = NULL; + time_t now, last_modified; + long int rtime; + long int days, hours, minutes, seconds; + + /* + * Are we messing with the Last-Modified header? + */ + if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0) + { + /*Nope*/ + return JB_ERR_OK; + } + + newval = csp->action->string[ACTION_STRING_LAST_MODIFIED]; + + if (0 == strcmpic(newval, "block") ) + { + /* + * Blocking Last-Modified header. Useless but why not. + */ + log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header); + freez(*header); + return JB_ERR_OK; + } + else if (0 == strcmpic(newval, "reset-to-request-time")) + { + /* + * Setting Last-Modified Header to now. + */ + get_http_time(0, buf); + freez(*header); + *header = strdup("Last-Modified: "); + string_append(header, buf); + + if (*header == NULL) + { + log_error(LOG_LEVEL_HEADER, "Insufficent memory. Last-Modified header got lost, boohoo."); + } + else + { + log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header); + } + } + else if (0 == strcmpic(newval, "randomize")) + { + log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header); + now = time(NULL); +#ifdef HAVE_GMTIME_R + timeptr = gmtime_r(&now, &gmt); +#elif OSX_DARWIN + pthread_mutex_lock(&gmtime_mutex); + timeptr = gmtime(&now); + pthread_mutex_unlock(&gmtime_mutex); +#else + timeptr = gmtime(&now); +#endif + if ((timeptr = parse_header_time(*header, &last_modified)) == NULL) + { + log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s (crunching!)", *header); + freez(*header); + } + else + { + rtime = difftime(now, last_modified); + if (rtime) + { + rtime = pick_from_range(rtime); + last_modified += rtime; +#ifdef HAVE_GMTIME_R + timeptr = gmtime_r(&last_modified, &gmt); +#elif OSX_DARWIN + pthread_mutex_lock(&gmtime_mutex); + timeptr = gmtime(&last_modified); + pthread_mutex_unlock(&gmtime_mutex); +#else + timeptr = gmtime(&last_modified); +#endif + strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr); + freez(*header); + *header = strdup("Last-Modified: "); + string_append(header, newheader); + + if (*header == NULL) + { + log_error(LOG_LEVEL_ERROR, "Insufficent memory, header crunched without replacement."); + return JB_ERR_MEMORY; + } + + if(LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */ + { + days = rtime / (3600 * 24); + hours = rtime / 3600 % 24; + minutes = rtime / 60 % 60; + seconds = rtime % 60; + + log_error(LOG_LEVEL_HEADER, "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s", + *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs", + minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)"); + } + } + else + { + log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with."); + } + } + } + + return JB_ERR_OK; +} + +/********************************************************************* + * + * Function : client_accept_encoding + * + * Description : Rewrite the client's Accept-Encoding header so that + * if doesn't allow compression, if the action applies. + * Note: For HTTP/1.0 the absence of the header is enough. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_accept_encoding(struct client_state *csp, char **header) +{ + if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0) + { + log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content"); + + freez(*header); + + /* Temporarily disable the correct behaviour to + * work around a PHP bug. + * + * if (!strcmpic(csp->http->ver, "HTTP/1.1")) + * { + * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0"); + * if (*header == NULL) + * { + * return JB_ERR_MEMORY; + * } + * } + * + */ + } + + return JB_ERR_OK; +} + + +/********************************************************************* + * + * Function : client_te + * + * Description : Rewrite the client's TE header so that + * if doesn't allow compression, if the action applies. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_te(struct client_state *csp, char **header) +{ + if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0) + { + freez(*header); + log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer"); + } + + return JB_ERR_OK; +} + +/********************************************************************* + * + * Function : client_referrer + * + * Description : Handle the "referer" config setting properly. + * Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_referrer(struct client_state *csp, char **header) +{ + const char *newval; + const char *host; + char *referer; + size_t hostlenght; + +#ifdef FEATURE_FORCE_LOAD + /* Since the referrer can include the prefix even + * if the request itself is non-forced, we must + * clean it unconditionally + */ + strclean(*header, FORCE_PREFIX); +#endif /* def FEATURE_FORCE_LOAD */ + + /* + * Are we sending referer? + */ + if ((csp->action->flags & ACTION_HIDE_REFERER) == 0) + { + return JB_ERR_OK; + } + + newval = csp->action->string[ACTION_STRING_REFERER]; + + if ((0 != strcmpic(newval, "conditional-block"))) + { + freez(*header); + } + if ((newval == NULL) || (0 == strcmpic(newval, "block")) ) + { + /* + * Blocking referer + */ + log_error(LOG_LEVEL_HEADER, "Referer crunched!"); + return JB_ERR_OK; + } + else if (0 == strcmpic(newval, "conditional-block")) + { + /* + * Block referer if host has changed. + */ + + if (NULL == (host = strdup(csp->http->hostport))) + { + freez(*header); + log_error(LOG_LEVEL_HEADER, "Referer crunched! Couldn't allocate memory for temporary host copy."); + return JB_ERR_MEMORY; + } + if (NULL == (referer = strdup(*header))) + { + freez(*header); + freez(host); + log_error(LOG_LEVEL_HEADER, "Referer crunched! Couldn't allocate memory for temporary referer copy."); + return JB_ERR_MEMORY; + } + hostlenght = strlen(host); + if ( hostlenght < (strlen(referer)-17) ) /*referer begins with 'Referer: http[s]://'*/ + { + /*Shorten referer to make sure the referer is blocked + *if www.example.org/www.example.com-shall-see-the-referer/ + *links to www.example.com/ + */ + referer[hostlenght+17] = '\0'; + } + if ( 0 == strstr(referer, host)) /*Host has changed*/ + { + log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header); + freez(*header); + } + else + { + log_error(LOG_LEVEL_HEADER, "%s (not modified, still on %s)", *header, host); + } + freez(referer); + freez(host); + return JB_ERR_OK; + } + else if (0 != strcmpic(newval, "forge")) + { + /* + * We have a specific (fixed) referer we want to send. + */ + if ((0 != strncmpic(newval, "http://", 7)) && (0 != strncmpic(newval, "https://", 8))) + { + log_error(LOG_LEVEL_HEADER, "Parameter: +referrer{%s} is a bad idea, but I don't care.", newval); + } + *header = strdup("Referer: "); + string_append(header, newval); + log_error(LOG_LEVEL_HEADER, "Referer overwritten with: %s", *header); + + return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK; + } + else + { + /* + * Forge a referer as http://[hostname:port of REQUEST]/ + * to fool stupid checks for in-site links + */ + + *header = strdup("Referer: http://"); + string_append(header, csp->http->hostport); + string_append(header, "/"); + log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header); + + return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK; + } +} + +/********************************************************************* + * + * Function : client_accept_language + * + * Description : Handle the "Accept-Language" config setting properly. + * Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_accept_language(struct client_state *csp, char **header) +{ + const char *newval; + + /* + * Are we messing with the Accept-Language? + */ + if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0) + { + /*I don't think so*/ + return JB_ERR_OK; + } + + newval = csp->action->string[ACTION_STRING_LANGUAGE]; + + if ((newval == NULL) || (0 == strcmpic(newval, "block")) ) + { + /* + * Blocking Accept-Language header + */ + log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!"); + freez(*header); + return JB_ERR_OK; + } + else + { + /* + * Replacing Accept-Language header + */ + freez(*header); + *header = strdup("Accept-Language: "); + string_append(header, newval); + + if (*header == NULL) + { + log_error(LOG_LEVEL_ERROR, " Insufficent memory. Accept-Language header crunched without replacement."); + } + else + { + log_error(LOG_LEVEL_HEADER, "Accept-Language header crunched and replaced with: %s", *header); + } + } + return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK; +} + +/********************************************************************* + * + * Function : crunch_client_header + * + * Description : Crunch client header if it matches a string supplied by the + * user. Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success and always succeeds + * + *********************************************************************/ +jb_err crunch_client_header(struct client_state *csp, char **header) +{ + const char *crunch_pattern; + /*Is there a header to crunch*/ + + if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER)) + { + crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER]; + + /*Is the current header the lucky one?*/ + if (strstr(*header, crunch_pattern)) + { + log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern); + freez(*header); + } + } + return JB_ERR_OK; } @@ -866,235 +1935,559 @@ char *client_referrer(const struct parsers *v, char *s, struct client_state *csp * * Function : client_uagent * - * Description : Handle the "user-agent" config setting properly. + * Description : Handle the "user-agent" config setting properly + * and remember its original value to enable browser + * bug workarounds. Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_uagent(struct client_state *csp, char **header) +{ + const char *newval; + + if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0) + { + return JB_ERR_OK; + } + + newval = csp->action->string[ACTION_STRING_USER_AGENT]; + if (newval == NULL) + { + return JB_ERR_OK; + } + + freez(*header); + *header = strdup("User-Agent: "); + string_append(header, newval); + + log_error(LOG_LEVEL_HEADER, "Modified: %s", *header); + + return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK; +} + +/********************************************************************* + * + * Function : client_ua + * + * Description : Handle "ua-" headers properly. Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_ua(struct client_state *csp, char **header) +{ + if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0) + { + log_error(LOG_LEVEL_HEADER, "crunched User-Agent!"); + freez(*header); + } + + return JB_ERR_OK; +} + + +/********************************************************************* + * + * Function : client_from + * + * Description : Handle the "from" config setting properly. * Called from `sed'. * * Parameters : - * 1 : v = ignored - * 2 : s = header (from sed) to "crunch" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_from(struct client_state *csp, char **header) +{ + const char *newval; + + if ((csp->action->flags & ACTION_HIDE_FROM) == 0) + { + return JB_ERR_OK; + } + + freez(*header); + + newval = csp->action->string[ACTION_STRING_FROM]; + + /* + * Are we blocking the e-mail address? + */ + if ((newval == NULL) || (0 == strcmpic(newval, "block")) ) + { + log_error(LOG_LEVEL_HEADER, "crunched From!"); + return JB_ERR_OK; + } + + log_error(LOG_LEVEL_HEADER, " modified"); + + *header = strdup("From: "); + string_append(header, newval); + + return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK; +} + + +/********************************************************************* + * + * Function : client_send_cookie + * + * Description : Handle the "cookie" header properly. Called from `sed'. + * If cookie is accepted, add it to the cookie_list, + * else we crunch it. Mmmmmmmmmmm ... cookie ...... + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_send_cookie(struct client_state *csp, char **header) +{ + jb_err result = JB_ERR_OK; + + if ((csp->action->flags & ACTION_NO_COOKIE_READ) == 0) + { + /* strlen("cookie: ") == 8 */ + result = enlist(csp->cookie_list, *header + 8); + } + else + { + log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie -- yum!"); + } + + /* + * Always remove the cookie here. The cookie header + * will be sent at the end of the header. + */ + freez(*header); + + return result; +} + + +/********************************************************************* + * + * Function : client_x_forwarded + * + * Description : Handle the "x-forwarded-for" config setting properly, + * also used in the add_client_headers list. Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : A malloc'ed pointer to the default agent, or - * a malloc'ed string pointer to this header (ie. pass thru). + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_uagent(const struct parsers *v, char *s, struct client_state *csp) +jb_err client_x_forwarded(struct client_state *csp, char **header) { - const char * newval; - -#ifdef DETECT_MSIE_IMAGES - if (strstr (s, "MSIE ")) + if ((csp->action->flags & ACTION_HIDE_FORWARDED) == 0) { - /* This is Microsoft Internet Explorer. - * Enable auto-detect. - */ - csp->accept_types |= ACCEPT_TYPE_IS_MSIE; - } -#endif /* def DETECT_MSIE_IMAGES */ + /* Save it so we can re-add it later */ + freez(csp->x_forwarded); + csp->x_forwarded = *header; - if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0) - { - return(strdup(s)); + /* + * Always set *header = NULL, since this information + * will be sent at the end of the header. + */ + *header = NULL; } - - newval = csp->action->string[ACTION_STRING_USER_AGENT]; - if (newval == NULL) + else { - return(strdup(s)); + freez(*header); + log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!"); } - log_error(LOG_LEVEL_HEADER, "modified"); - - s = strsav( NULL, "User-Agent: " ); - s = strsav( s, newval ); - return(s); - + return JB_ERR_OK; } /********************************************************************* * - * Function : client_ua + * Function : client_max_forwards * - * Description : Handle "ua-" headers properly. Called from `sed'. + * Description : If the HTTP method is OPTIONS or TRACE, subtract one + * from the value of the Max-Forwards header field. * * Parameters : - * 1 : v = ignored - * 2 : s = header (from sed) to "crunch" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : NULL if crunched, or a malloc'ed string to original header + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_ua(const struct parsers *v, char *s, struct client_state *csp) +jb_err client_max_forwards(struct client_state *csp, char **header) { - if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0) - { - return(strdup(s)); - } - else + unsigned int max_forwards; + + if ((0 == strcmpic(csp->http->gpc, "trace")) + || (0 == strcmpic(csp->http->gpc, "options"))) { - log_error(LOG_LEVEL_HEADER, "crunch!"); - return(NULL); + if (1 == sscanf(*header, "Max-Forwards: %u", &max_forwards)) + { + if (max_forwards-- >= 1) + { + sprintf(*header, "Max-Forwards: %u", max_forwards); + log_error(LOG_LEVEL_HEADER, "Max forwards of %s request now %d", csp->http->gpc, max_forwards); + } + else + { + log_error(LOG_LEVEL_ERROR, "Non-intercepted %s request with Max-Forwards zero!", csp->http->gpc); + } + } } + + return JB_ERR_OK; } /********************************************************************* * - * Function : client_from + * Function : client_host * - * Description : Handle the "from" config setting properly. - * Called from `sed'. + * Description : If the request URI did not contain host and + * port information, parse and evaluate the Host + * header field. + * + * Also, kill ill-formed HOST: headers as sent by + * Apple's iTunes software when used with a proxy. * * Parameters : - * 1 : v = ignored - * 2 : s = header (from sed) to "crunch" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : NULL if crunched, or a malloc'ed string to - * modified/original header. + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_from(const struct parsers *v, char *s, struct client_state *csp) +jb_err client_host(struct client_state *csp, char **header) { - const char * newval; - - if ((csp->action->flags & ACTION_HIDE_FROM) == 0) - { - return(strdup(s)); - } - - newval = csp->action->string[ACTION_STRING_FROM]; + char *p, *q; /* - * Are we blocking referer? + * If the header field name is all upper-case, chances are that it's + * an ill-formed one from iTunes. BTW, killing innocent headers here is + * not a problem -- they are regenerated later. */ - if ((newval == NULL) || (0 == strcmpic(newval, "block")) ) + if ((*header)[1] == 'O') { - log_error(LOG_LEVEL_HEADER, "crunch!"); - return(NULL); + log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header); + freez(*header); + return JB_ERR_OK; } - log_error(LOG_LEVEL_HEADER, " modified"); + if (!csp->http->hostport || (*csp->http->hostport == '*') || + *csp->http->hostport == ' ' || *csp->http->hostport == '\0') + { + + if (NULL == (p = strdup((*header)+6))) + { + return JB_ERR_MEMORY; + } + chomp(p); + if (NULL == (q = strdup(p))) + { + freez(p); + return JB_ERR_MEMORY; + } - s = strsav( NULL, "From: " ); - s = strsav( s, newval ); - return(s); + freez(csp->http->hostport); + csp->http->hostport = p; + freez(csp->http->host); + csp->http->host = q; + q = strchr(csp->http->host, ':'); + if (q != NULL) + { + /* Terminate hostname and evaluate port string */ + *q++ = '\0'; + csp->http->port = atoi(q); + } + else + { + csp->http->port = csp->http->ssl ? 443 : 80; + } -} + log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d", + csp->http->hostport, csp->http->host, csp->http->port); + } + return JB_ERR_OK; +} /********************************************************************* * - * Function : client_send_cookie + * Function : client_if_modified_since * - * Description : Handle the "cookie" header properly. Called from `sed'. - * If cookie is accepted, add it to the cookie_list, - * else we crunch it. Mmmmmmmmmmm ... cookie ...... + * Description : Remove or modify the If-Modified-Since header. * * Parameters : - * 1 : v = pattern of cookie `sed' found matching - * 2 : s = header (from sed) to "crunch" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : Always NULL. + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_send_cookie(const struct parsers *v, char *s, struct client_state *csp) +jb_err client_if_modified_since(struct client_state *csp, char **header) { - if ((csp->action->flags & ACTION_NO_COOKIE_READ) == 0) + char newheader[50]; +#ifdef HAVE_GMTIME_R + struct tm gmt; +#endif + struct tm *timeptr = NULL; + time_t tm = 0; + const char *newval; + long int rtime; + long int hours, minutes, seconds; + int negative = 0; + char * endptr; + + if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT")) { - enlist(csp->cookie_list, s + v->len + 1); + /* + * The client got an error message because of a temporary problem, + * the problem is gone and the client now tries to revalidate our + * error message on the real server. The revalidation would always + * end with the transmission of the whole document and there is + * no need to expose the bogus If-Modified-Since header. + */ + log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header."); + freez(*header); } - else + else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE) { - log_error(LOG_LEVEL_HEADER, " crunch!"); - } + newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE]; - /* - * Always return NULL here. The cookie header - * will be sent at the end of the header. - */ - return(NULL); + if ((0 == strcmpic(newval, "block"))) + { + log_error(LOG_LEVEL_HEADER, "Crunching %s", *header); + freez(*header); + } + else /* add random value */ + { + if ((timeptr = parse_header_time(*header, &tm)) == NULL) + { + log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s (crunching!)", *header); + freez(*header); + } + else + { + rtime = strtol(newval, &endptr, 0); + if(rtime) + { + log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)", + *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es"); + if(rtime < 0) + { + rtime *= -1; + negative = 1; + } + rtime *= 60; + rtime = pick_from_range(rtime); + } + else + { + log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.", + *header); + } + tm += rtime * (negative ? -1 : 1); +#ifdef HAVE_GMTIME_R + timeptr = gmtime_r(&tm, &gmt); +#elif OSX_DARWIN + pthread_mutex_lock(&gmtime_mutex); + timeptr = gmtime(&tm); + pthread_mutex_unlock(&gmtime_mutex); +#else + timeptr = gmtime(&tm); +#endif + strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr); + + freez(*header); + *header = strdup("If-Modified-Since: "); + string_append(header, newheader); + + if (*header == NULL) + { + log_error(LOG_LEVEL_HEADER, "Insufficent memory, header crunched without replacement."); + return JB_ERR_MEMORY; + } + + if(LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */ + { + hours = rtime / 3600; + minutes = rtime / 60 % 60; + seconds = rtime % 60; + + log_error(LOG_LEVEL_HEADER, "Randomized: %s (%s %d hou%s %d minut%s %d second%s", + *header, (negative) ? "subtracted" : "added", hours, (hours == 1) ? "r" : "rs", + minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)"); + } + } + } + } + return JB_ERR_OK; } - /********************************************************************* * - * Function : client_x_forwarded + * Function : client_if_none_match * - * Description : Handle the "x-forwarded-for" config setting properly, - * also used in the add_client_headers list. Called from `sed'. + * Description : Remove the If-None-Match header. * * Parameters : - * 1 : v = ignored - * 2 : s = header (from sed) to "crunch" - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : Always NULL. + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_x_forwarded(const struct parsers *v, char *s, struct client_state *csp) +jb_err client_if_none_match(struct client_state *csp, char **header) { - if ((csp->action->flags & ACTION_HIDE_FORWARDED) == 0) - { - /* Save it so we can re-add it later */ - csp->x_forwarded = strdup(s); + if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH) + { + log_error(LOG_LEVEL_HEADER, "Crunching %s", *header); + freez(*header); } - /* - * Always return NULL, since this information - * will be sent at the end of the header. - */ - - return(NULL); + return JB_ERR_OK; +} +/********************************************************************* + * + * Function : client_x_filter + * + * Description : Disables filtering if the client set "X-Filter: No". + * Called from `sed'. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success + * + *********************************************************************/ +jb_err client_x_filter(struct client_state *csp, char **header) +{ + if ( 0 == strcmpic(*header, "X-Filter: No")) + { + if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE)) + { + log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering."); + } + else + { + if (csp->action->flags & ACTION_FORCE_TEXT_MODE) + { + log_error(LOG_LEVEL_HEADER, "force-text-mode overruled the client's request to fetch without filtering!"); + } + else + { + csp->content_type = CT_TABOO; + log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering."); + } + log_error(LOG_LEVEL_HEADER, "Crunching %s", *header); + freez(*header); + } + } + return JB_ERR_OK; } -#if defined(DETECT_MSIE_IMAGES) +/* the following functions add headers directly to the header list */ + /********************************************************************* * - * Function : client_accept + * Function : client_host_adder * - * Description : Detect whether the client wants HTML or an image. - * Clients do not always make this information available - * in a sane way. Always passes the header through - * the proxy unchanged. + * Description : Adds the Host: header field if it is missing. + * Called from `sed'. * * Parameters : - * 1 : v = Ignored. - * 2 : s = Header string. Null terminated. - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : Duplicate of argument s. + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_accept(const struct parsers *v, char *s, struct client_state *csp) +jb_err client_host_adder(struct client_state *csp) { -#ifdef DETECT_MSIE_IMAGES - if (strstr (s, "image/gif")) + char *p; + jb_err err; + + if ( !csp->http->hostport || !*(csp->http->hostport)) { - /* Client will accept HTML. If this seems counterintuitive, - * blame Microsoft. - */ - csp->accept_types |= ACCEPT_TYPE_MSIE_HTML; + return JB_ERR_OK; + } + + /* + * remove 'user:pass@' from 'proto://user:pass@host' + */ + if ( (p = strchr( csp->http->hostport, '@')) != NULL ) + { + p++; } else { - csp->accept_types |= ACCEPT_TYPE_MSIE_IMAGE; + p = csp->http->hostport; } -#endif /* def DETECT_MSIE_IMAGES */ - return(strdup(s)); + log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p); + err = enlist_unique_header(csp->headers, "Host", p); + return err; } -#endif /* defined(DETECT_MSIE_IMAGES) */ - - - -/* the following functions add headers directly to the header list */ /********************************************************************* @@ -1106,50 +2499,90 @@ char *client_accept(const struct parsers *v, char *s, struct client_state *csp) * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : N/A + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -void client_cookie_adder(struct client_state *csp) +jb_err client_cookie_adder(struct client_state *csp) { - struct list *lst; - char *tmp = NULL; - char *e; + struct list_entry *lst; + char *tmp; + struct list_entry *list1 = csp->cookie_list->first; + struct list_entry *list2 = csp->action->multi[ACTION_MULTI_WAFER]->first; + int first_cookie = 1; + jb_err err; + + if ((list1 == NULL) && (list2 == NULL)) + { + /* Nothing to do */ + return JB_ERR_OK; + } - for (lst = csp->cookie_list->next; lst ; lst = lst->next) + tmp = strdup("Cookie: "); + + for (lst = list1; lst ; lst = lst->next) { - if (tmp) + if (first_cookie) + { + first_cookie = 0; + } + else { - tmp = strsav(tmp, "; "); + string_append(&tmp, "; "); } - tmp = strsav(tmp, lst->str); + string_append(&tmp, lst->str); } - for (lst = csp->action->multi[ACTION_MULTI_WAFER]->next; lst ; lst = lst->next) + for (lst = list2; lst ; lst = lst->next) { - if (tmp) + if (first_cookie) { - tmp = strsav(tmp, "; "); + first_cookie = 0; } - - if ((e = cookie_encode(lst->str))) + else { - tmp = strsav(tmp, e); - freez(e); + string_append(&tmp, "; "); } + string_join(&tmp, cookie_encode(lst->str)); } - if (tmp) + if (tmp == NULL) { - char *ret; + return JB_ERR_MEMORY; + } - ret = strdup("Cookie: "); - ret = strsav(ret, tmp); - log_error(LOG_LEVEL_HEADER, "addh: %s", ret); - enlist(csp->headers, ret); - freez(tmp); - freez(ret); + log_error(LOG_LEVEL_HEADER, "addh: %s", tmp); + err = enlist(csp->headers, tmp); + free(tmp); + return err; +} + + +/********************************************************************* + * + * Function : client_accept_encoding_adder + * + * Description : Add an Accept-Encoding header to the client's request + * that disables compression if the action applies, and + * the header is not already there. Called from `sed'. + * Note: For HTTP/1.0, the absence of the header is enough. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err client_accept_encoding_adder(struct client_state *csp) +{ + if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0) + && (!strcmpic(csp->http->ver, "HTTP/1.1")) ) + { + return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16); } + return JB_ERR_OK; } @@ -1162,19 +2595,28 @@ void client_cookie_adder(struct client_state *csp) * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : N/A + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -void client_xtra_adder(struct client_state *csp) +jb_err client_xtra_adder(struct client_state *csp) { - struct list *lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]; + struct list_entry *lst; + jb_err err; - for (lst = lst->next; lst ; lst = lst->next) + for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first; + lst ; lst = lst->next) { log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str); - enlist(csp->headers, lst->str); + err = enlist(csp->headers, lst->str); + if (err) + { + return err; + } + } + return JB_ERR_OK; } @@ -1187,105 +2629,240 @@ void client_xtra_adder(struct client_state *csp) * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : N/A + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -void client_x_forwarded_adder(struct client_state *csp) +jb_err client_x_forwarded_adder(struct client_state *csp) { char *p = NULL; + jb_err err; if ((csp->action->flags & ACTION_HIDE_FORWARDED) != 0) { - return; + return JB_ERR_OK; } if (csp->x_forwarded) { - p = strsav(p, csp->x_forwarded); - p = strsav(p, ", "); - p = strsav(p, csp->ip_addr_str); + p = strdup(csp->x_forwarded); + string_append(&p, ", "); } else { - p = strsav(p, "X-Forwarded-For: "); - p = strsav(p, csp->ip_addr_str); + p = strdup("X-Forwarded-For: "); + } + string_append(&p, csp->ip_addr_str); + + if (p == NULL) + { + return JB_ERR_MEMORY; } log_error(LOG_LEVEL_HEADER, "addh: %s", p); - enlist(csp->headers, p); + err = enlist(csp->headers, p); + free(p); + return err; } /********************************************************************* * - * Function : server_set_cookie + * Function : connection_close_adder * - * Description : Handle the server "cookie" header properly. - * Log cookie to the jar file. Then "crunch" it, - * or accept it. Called from `sed'. + * Description : Adds a "Connection: close" header to csp->headers + * as a temporary fix for the needed but missing HTTP/1.1 + * support. Called from `sed'. + * FIXME: This whole function shouldn't be neccessary! * * Parameters : - * 1 : v = parser pattern that matched this header - * 2 : s = header that matched this pattern - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : `crumble' or a newly malloc'ed string. + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *server_set_cookie(const struct parsers *v, char *s, struct client_state *csp) +jb_err connection_close_adder(struct client_state *csp) { -#ifdef JAR_FILES - if (csp->config->jar) + log_error(LOG_LEVEL_HEADER, "Adding: Connection: close"); + return enlist(csp->headers, "Connection: close"); +} + + +/********************************************************************* + * + * Function : server_http + * + * Description : - Save the HTTP Status into csp->http->status + * - Set CT_TABOO to prevent filtering if the answer + * is a partial range (HTTP status 206) + * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade + * action applies. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. + * + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. + * + *********************************************************************/ +jb_err server_http(struct client_state *csp, char **header) +{ + sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status)); + if (csp->http->status == 206) { - fprintf(csp->config->jar, "%s\t%s\n", csp->http->host, (s + v->len + 1)); + csp->content_type = CT_TABOO; } -#endif /* def JAR_FILES */ - if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0) + if ((csp->action->flags & ACTION_DOWNGRADE) != 0) { - return(crumble(v, s, csp)); + (*header)[7] = '0'; + log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0"); } - return(strdup(s)); - + return JB_ERR_OK; } -#ifdef FORCE_LOAD /********************************************************************* * - * Function : client_host + * Function : server_set_cookie * - * Description : Clean the FORCE_PREFIX out of the 'host' http - * header, if we use force + * Description : Handle the server "cookie" header properly. + * Log cookie to the jar file. Then "crunch" it, + * or accept it. Called from `sed'. * * Parameters : - * 1 : v = ignored - * 2 : s = header (from sed) to clean - * 3 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : header = On input, pointer to header to modify. + * On output, pointer to the modified header, or NULL + * to remove the header. This function frees the + * original string if necessary. * - * Returns : A malloc'ed pointer to the cleaned host header + * Returns : JB_ERR_OK on success, or + * JB_ERR_MEMORY on out-of-memory error. * *********************************************************************/ -char *client_host(const struct parsers *v, char *s, struct client_state *csp) +jb_err server_set_cookie(struct client_state *csp, char **header) { - char *cleanhost = strdup(s); - - if(csp->force) - strclean(cleanhost, FORCE_PREFIX); - - return(cleanhost); +#ifdef FEATURE_COOKIE_JAR + if (csp->config->jar) + { + /* + * Write timestamp into outbuf. + * + * Complex because not all OSs have tm_gmtoff or + * the %z field in strftime() + */ + char tempbuf[ BUFFER_SIZE ]; + time_t now; + struct tm tm_now; + time (&now); +#ifdef HAVE_LOCALTIME_R + tm_now = *localtime_r(&now, &tm_now); +#elif OSX_DARWIN + pthread_mutex_lock(&localtime_mutex); + tm_now = *localtime (&now); + pthread_mutex_unlock(&localtime_mutex); +#else + tm_now = *localtime (&now); +#endif + strftime(tempbuf, BUFFER_SIZE-6, "%b %d %H:%M:%S ", &tm_now); + + /* strlen("set-cookie: ") = 12 */ + fprintf(csp->config->jar, "%s %s\t%s\n", tempbuf, csp->http->host, *header + 12); + } +#endif /* def FEATURE_COOKIE_JAR */ + + if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0) + { + log_error(LOG_LEVEL_HEADER, "Crunched incoming cookie -- yum!"); + return crumble(csp, header); + } + else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0) + { + /* Flag whether or not to log a message */ + int changed = 0; + + /* A variable to store the tag we're working on */ + char *cur_tag; + + /* Skip "Set-Cookie:" (11 characters) in header */ + cur_tag = *header + 11; + + /* skip whitespace between "Set-Cookie:" and value */ + while (*cur_tag && ijb_isspace(*cur_tag)) + { + cur_tag++; + } + + /* Loop through each tag in the cookie */ + while (*cur_tag) + { + /* Find next tag */ + char *next_tag = strchr(cur_tag, ';'); + if (next_tag != NULL) + { + /* Skip the ';' character itself */ + next_tag++; + + /* skip whitespace ";" and start of tag */ + while (*next_tag && ijb_isspace(*next_tag)) + { + next_tag++; + } + } + else + { + /* "Next tag" is the end of the string */ + next_tag = cur_tag + strlen(cur_tag); + } + + /* Is this the "Expires" tag? */ + if (strncmpic(cur_tag, "expires=", 8) == 0) + { + /* Delete the tag by copying the rest of the string over it. + * (Note that we cannot just use "strcpy(cur_tag, next_tag)", + * since the behaviour of strcpy is undefined for overlapping + * strings.) + */ + memmove(cur_tag, next_tag, strlen(next_tag) + 1); + + /* That changed the header, need to issue a log message */ + changed = 1; + + /* Note that the next tag has now been moved to *cur_tag, + * so we do not need to update the cur_tag pointer. + */ + } + else + { + /* Move on to next cookie tag */ + cur_tag = next_tag; + } + } + + if (changed) + { + log_error(LOG_LEVEL_HEADER, "Changed cookie to a temporary one."); + } + } + + return JB_ERR_OK; } -#endif /* def FORCE_LOAD */ - - -#ifdef FORCE_LOAD + + +#ifdef FEATURE_FORCE_LOAD /********************************************************************* * * Function : strclean * - * Description : In-Situ-Eliminate all occurances of substring in + * Description : In-Situ-Eliminate all occurances of substring in * string * * Parameters : @@ -1300,12 +2877,12 @@ int strclean(const char *string, const char *substring) int hits = 0, len = strlen(substring); char *pos, *p; - while((pos = strstr(string, substring))) + while((pos = strstr(string, substring)) != NULL) { p = pos + len; do { - *(p - len) = *p; + *(p - len) = *p; } while (*p++ != '\0'); @@ -1314,7 +2891,42 @@ int strclean(const char *string, const char *substring) return(hits); } -#endif /* def FORCE_LOAD */ +#endif /* def FEATURE_FORCE_LOAD */ + +/********************************************************************* + * + * Function : parse_header_time + * + * Description : Transforms time inside a HTTP header into + * the usual time format. + * + * Parameters : + * 1 : header = header to parse + * 2 : tm = storage for the resulting time in seconds + * + * Returns : Time struct containing the header time, or + * NULL in case of a parsing problem. + * + *********************************************************************/ +struct tm *parse_header_time(char *header, time_t *tm) { + + char * timestring; + struct tm gmt; + struct tm * timeptr; + + /* Skipping header name */ + timestring = strstr(header, ": "); + if (strptime(timestring, ": %a, %d %b %Y %H:%M:%S", &gmt) == NULL) + { + timeptr = NULL; + } + else + { + *tm = timegm(&gmt); + timeptr=&gmt; + } + return(timeptr); +} /*