1 const char parsers_rcs[] = "$Id: parsers.c,v 1.185 2009/06/27 11:25:33 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `add_to_iob', `client_cookie_adder', `client_from',
9 * `client_referrer', `client_send_cookie', `client_ua',
10 * `client_uagent', `client_x_forwarded',
11 * `client_x_forwarded_adder', `client_xtra_adder',
12 * `content_type', `crumble', `destroy_list', `enlist',
13 * `flush_socket', ``get_header', `sed', `filter_header'
14 * `server_content_encoding', `server_content_disposition',
15 * `server_last_modified', `client_accept_language',
16 * `crunch_client_header', `client_if_modified_since',
17 * `client_if_none_match', `get_destination_from_headers',
18 * `parse_header_time', `decompress_iob' and `server_set_cookie'.
20 * Copyright : Written by and Copyright (C) 2001-2009 the
21 * Privoxy team. http://www.privoxy.org/
23 * Based on the Internet Junkbuster originally written
24 * by and Copyright (C) 1997 Anonymous Coders and
25 * Junkbusters Corporation. http://www.junkbusters.com
27 * This program is free software; you can redistribute it
28 * and/or modify it under the terms of the GNU General
29 * Public License as published by the Free Software
30 * Foundation; either version 2 of the License, or (at
31 * your option) any later version.
33 * This program is distributed in the hope that it will
34 * be useful, but WITHOUT ANY WARRANTY; without even the
35 * implied warranty of MERCHANTABILITY or FITNESS FOR A
36 * PARTICULAR PURPOSE. See the GNU General Public
37 * License for more details.
39 * The GNU General Public License should be included with
40 * this file. If not, you can view it at
41 * http://www.gnu.org/copyleft/gpl.html
42 * or write to the Free Software Foundation, Inc., 59
43 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
45 *********************************************************************/
52 #include <sys/types.h>
62 * Convince GNU's libc to provide a strptime prototype.
71 #define GZIP_IDENTIFIER_1 0x1f
72 #define GZIP_IDENTIFIER_2 0x8b
74 #define GZIP_FLAG_CHECKSUM 0x02
75 #define GZIP_FLAG_EXTRA_FIELDS 0x04
76 #define GZIP_FLAG_FILE_NAME 0x08
77 #define GZIP_FLAG_COMMENT 0x10
78 #define GZIP_FLAG_RESERVED_BITS 0xe0
81 #if !defined(_WIN32) && !defined(__OS2__)
87 #ifdef FEATURE_PTHREAD
89 /* jcc.h is for mutex semapores only */
90 #endif /* def FEATURE_PTHREAD */
95 #include "jbsockets.h"
101 #ifndef HAVE_STRPTIME
102 #include "strptime.h"
105 const char parsers_h_rcs[] = PARSERS_H_VERSION;
107 /* Fix a problem with Solaris. There should be no effect on other
109 * Solaris's isspace() is a macro which uses its argument directly
110 * as an array index. Therefore we need to make sure that high-bit
111 * characters generate +ve values, and ideally we also want to make
112 * the argument match the declared parameter type of "int".
114 * Why did they write a character function that can't take a simple
115 * "char" argument? Doh!
117 #define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
118 #define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
120 static char *get_header_line(struct iob *iob);
121 static jb_err scan_headers(struct client_state *csp);
122 static jb_err header_tagger(struct client_state *csp, char *header);
123 static jb_err parse_header_time(const char *header_time, time_t *result);
125 static jb_err crumble (struct client_state *csp, char **header);
126 static jb_err filter_header (struct client_state *csp, char **header);
127 static jb_err client_connection (struct client_state *csp, char **header);
128 static jb_err client_referrer (struct client_state *csp, char **header);
129 static jb_err client_uagent (struct client_state *csp, char **header);
130 static jb_err client_ua (struct client_state *csp, char **header);
131 static jb_err client_from (struct client_state *csp, char **header);
132 static jb_err client_send_cookie (struct client_state *csp, char **header);
133 static jb_err client_x_forwarded (struct client_state *csp, char **header);
134 static jb_err client_accept_encoding (struct client_state *csp, char **header);
135 static jb_err client_te (struct client_state *csp, char **header);
136 static jb_err client_max_forwards (struct client_state *csp, char **header);
137 static jb_err client_host (struct client_state *csp, char **header);
138 static jb_err client_if_modified_since (struct client_state *csp, char **header);
139 static jb_err client_accept_language (struct client_state *csp, char **header);
140 static jb_err client_if_none_match (struct client_state *csp, char **header);
141 static jb_err crunch_client_header (struct client_state *csp, char **header);
142 static jb_err client_x_filter (struct client_state *csp, char **header);
143 static jb_err client_range (struct client_state *csp, char **header);
144 static jb_err server_set_cookie (struct client_state *csp, char **header);
145 static jb_err server_connection (struct client_state *csp, char **header);
146 static jb_err server_content_type (struct client_state *csp, char **header);
147 static jb_err server_adjust_content_length(struct client_state *csp, char **header);
148 static jb_err server_content_md5 (struct client_state *csp, char **header);
149 static jb_err server_content_encoding (struct client_state *csp, char **header);
150 static jb_err server_transfer_coding (struct client_state *csp, char **header);
151 static jb_err server_http (struct client_state *csp, char **header);
152 static jb_err crunch_server_header (struct client_state *csp, char **header);
153 static jb_err server_last_modified (struct client_state *csp, char **header);
154 static jb_err server_content_disposition(struct client_state *csp, char **header);
156 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
157 static jb_err server_save_content_length(struct client_state *csp, char **header);
158 static jb_err server_keep_alive(struct client_state *csp, char **header);
159 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
161 static jb_err client_host_adder (struct client_state *csp);
162 static jb_err client_xtra_adder (struct client_state *csp);
163 static jb_err client_x_forwarded_for_adder(struct client_state *csp);
164 static jb_err client_connection_header_adder(struct client_state *csp);
165 static jb_err server_connection_adder(struct client_state *csp);
166 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
167 static jb_err server_proxy_connection_adder(struct client_state *csp);
168 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
170 static jb_err create_forged_referrer(char **header, const char *hostport);
171 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
172 static jb_err handle_conditional_hide_referrer_parameter(char **header,
173 const char *host, const int parameter_conditional_block);
174 static const char *get_appropiate_connection_header(const struct client_state *csp);
175 static void create_content_length_header(unsigned long long content_length,
176 char *header, size_t buffer_length);
179 * List of functions to run on a list of headers.
183 /** The header prefix to match */
186 /** The length of the prefix to match */
189 /** The function to apply to this line */
190 const parser_func_ptr parser;
193 static const struct parsers client_patterns[] = {
194 { "referer:", 8, client_referrer },
195 { "user-agent:", 11, client_uagent },
196 { "ua-", 3, client_ua },
197 { "from:", 5, client_from },
198 { "cookie:", 7, client_send_cookie },
199 { "x-forwarded-for:", 16, client_x_forwarded },
200 { "Accept-Encoding:", 16, client_accept_encoding },
201 { "TE:", 3, client_te },
202 { "Host:", 5, client_host },
203 { "if-modified-since:", 18, client_if_modified_since },
204 #ifndef FEATURE_CONNECTION_KEEP_ALIVE
205 { "Keep-Alive:", 11, crumble },
207 { "connection:", 11, client_connection },
208 { "proxy-connection:", 17, crumble },
209 { "max-forwards:", 13, client_max_forwards },
210 { "Accept-Language:", 16, client_accept_language },
211 { "if-none-match:", 14, client_if_none_match },
212 { "Range:", 6, client_range },
213 { "Request-Range:", 14, client_range },
214 { "If-Range:", 9, client_range },
215 { "X-Filter:", 9, client_x_filter },
216 { "*", 0, crunch_client_header },
217 { "*", 0, filter_header },
221 static const struct parsers server_patterns[] = {
222 { "HTTP/", 5, server_http },
223 { "set-cookie:", 11, server_set_cookie },
224 { "connection:", 11, server_connection },
225 { "Content-Type:", 13, server_content_type },
226 { "Content-MD5:", 12, server_content_md5 },
227 { "Content-Encoding:", 17, server_content_encoding },
228 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
229 { "Content-Length:", 15, server_save_content_length },
230 { "Keep-Alive:", 11, server_keep_alive },
232 { "Keep-Alive:", 11, crumble },
233 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
234 { "Transfer-Encoding:", 18, server_transfer_coding },
235 { "content-disposition:", 20, server_content_disposition },
236 { "Last-Modified:", 14, server_last_modified },
237 { "*", 0, crunch_server_header },
238 { "*", 0, filter_header },
242 static const add_header_func_ptr add_client_headers[] = {
244 client_x_forwarded_for_adder,
246 /* Temporarily disabled: client_accept_encoding_adder, */
247 client_connection_header_adder,
251 static const add_header_func_ptr add_server_headers[] = {
252 server_connection_adder,
253 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
254 server_proxy_connection_adder,
255 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
259 /*********************************************************************
261 * Function : flush_socket
263 * Description : Write any pending "buffered" content.
266 * 1 : fd = file descriptor of the socket to read
267 * 2 : iob = The I/O buffer to flush, usually csp->iob.
269 * Returns : On success, the number of bytes written are returned (zero
270 * indicates nothing was written). On error, -1 is returned,
271 * and errno is set appropriately. If count is zero and the
272 * file descriptor refers to a regular file, 0 will be
273 * returned without causing any other effect. For a special
274 * file, the results are not portable.
276 *********************************************************************/
277 long flush_socket(jb_socket fd, struct iob *iob)
279 long len = iob->eod - iob->cur;
286 if (write_socket(fd, iob->cur, (size_t)len))
290 iob->eod = iob->cur = iob->buf;
296 /*********************************************************************
298 * Function : add_to_iob
300 * Description : Add content to the buffered page, expanding the
301 * buffer if necessary.
304 * 1 : csp = Current client state (buffers, headers, etc...)
305 * 2 : buf = holds the content to be added to the page
306 * 3 : n = number of bytes to be added
308 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
309 * or buffer limit reached.
311 *********************************************************************/
312 jb_err add_to_iob(struct client_state *csp, char *buf, long n)
314 struct iob *iob = csp->iob;
315 size_t used, offset, need, want;
318 if (n <= 0) return JB_ERR_OK;
320 used = (size_t)(iob->eod - iob->buf);
321 offset = (size_t)(iob->cur - iob->buf);
322 need = used + (size_t)n + 1;
325 * If the buffer can't hold the new data, extend it first.
326 * Use the next power of two if possible, else use the actual need.
328 if (need > csp->config->buffer_limit)
330 log_error(LOG_LEVEL_INFO,
331 "Buffer limit reached while extending the buffer (iob). Needed: %d. Limit: %d",
332 need, csp->config->buffer_limit);
333 return JB_ERR_MEMORY;
336 if (need > iob->size)
338 for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2;
340 if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
344 else if (NULL != (p = (char *)realloc(iob->buf, need)))
350 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
351 return JB_ERR_MEMORY;
354 /* Update the iob pointers */
355 iob->cur = p + offset;
360 /* copy the new data into the iob buffer */
361 memcpy(iob->eod, buf, (size_t)n);
363 /* point to the end of the data */
366 /* null terminate == cheap insurance */
375 /*********************************************************************
377 * Function : decompress_iob
379 * Description : Decompress buffered page, expanding the
380 * buffer as necessary. csp->iob->cur
381 * should point to the the beginning of the
382 * compressed data block.
385 * 1 : csp = Current client state (buffers, headers, etc...)
387 * Returns : JB_ERR_OK on success,
388 * JB_ERR_MEMORY if out-of-memory limit reached, and
389 * JB_ERR_COMPRESS if error decompressing buffer.
391 *********************************************************************/
392 jb_err decompress_iob(struct client_state *csp)
394 char *buf; /* new, uncompressed buffer */
395 char *cur; /* Current iob position (to keep the original
396 * iob->cur unmodified if we return early) */
397 size_t bufsize; /* allocated size of the new buffer */
398 size_t old_size; /* Content size before decompression */
399 size_t skip_size; /* Number of bytes at the beginning of the iob
400 that we should NOT decompress. */
401 int status; /* return status of the inflate() call */
402 z_stream zstr; /* used by calls to zlib */
404 assert(csp->iob->cur - csp->iob->buf > 0);
405 assert(csp->iob->eod - csp->iob->cur > 0);
407 bufsize = csp->iob->size;
408 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
409 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
413 if (bufsize < (size_t)10)
416 * This is to protect the parsing of gzipped data,
417 * but it should(?) be valid for deflated data also.
419 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
420 return JB_ERR_COMPRESS;
423 if (csp->content_type & CT_GZIP)
426 * Our task is slightly complicated by the facts that data
427 * compressed by gzip does not include a zlib header, and
428 * that there is no easily accessible interface in zlib to
429 * handle a gzip header. We strip off the gzip header by
430 * hand, and later inform zlib not to expect a header.
434 * Strip off the gzip header. Please see RFC 1952 for more
435 * explanation of the appropriate fields.
437 if (((*cur++ & 0xff) != GZIP_IDENTIFIER_1)
438 || ((*cur++ & 0xff) != GZIP_IDENTIFIER_2)
439 || (*cur++ != Z_DEFLATED))
441 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
442 return JB_ERR_COMPRESS;
447 if (flags & GZIP_FLAG_RESERVED_BITS)
449 /* The gzip header has reserved bits set; bail out. */
450 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
451 return JB_ERR_COMPRESS;
455 * Skip mtime (4 bytes), extra flags (1 byte)
456 * and OS type (1 byte).
460 /* Skip extra fields if necessary. */
461 if (flags & GZIP_FLAG_EXTRA_FIELDS)
464 * Skip a given number of bytes, specified
465 * as a 16-bit little-endian value.
467 * XXX: this code is untested and should probably be removed.
471 skip_bytes += *cur++ << 8;
474 * The number of bytes to skip should be positive
475 * and we'd like to stay in the buffer.
477 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
479 log_error(LOG_LEVEL_ERROR,
480 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
482 return JB_ERR_COMPRESS;
484 log_error(LOG_LEVEL_INFO,
485 "Skipping %d bytes for gzip compression. Does this sound right?",
490 /* Skip the filename if necessary. */
491 if (flags & GZIP_FLAG_FILE_NAME)
493 /* A null-terminated string is supposed to follow. */
494 while (*cur++ && (cur < csp->iob->eod));
497 /* Skip the comment if necessary. */
498 if (flags & GZIP_FLAG_COMMENT)
500 /* A null-terminated string is supposed to follow. */
501 while (*cur++ && (cur < csp->iob->eod));
504 /* Skip the CRC if necessary. */
505 if (flags & GZIP_FLAG_CHECKSUM)
510 if (cur >= csp->iob->eod)
513 * If the current position pointer reached or passed
514 * the buffer end, we were obviously tricked to skip
517 log_error(LOG_LEVEL_ERROR,
518 "Malformed gzip header detected. Aborting decompression.");
519 return JB_ERR_COMPRESS;
523 else if (csp->content_type & CT_DEFLATE)
526 * XXX: The debug level should be lowered
527 * before the next stable release.
529 log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
531 * In theory (that is, according to RFC 1950), deflate-compressed
532 * data should begin with a two-byte zlib header and have an
533 * adler32 checksum at the end. It seems that in practice only
534 * the raw compressed data is sent. Note that this means that
535 * we are not RFC 1950-compliant here, but the advantage is that
536 * this actually works. :)
538 * We add a dummy null byte to tell zlib where the data ends,
539 * and later inform it not to expect a header.
541 * Fortunately, add_to_iob() has thoughtfully null-terminated
542 * the buffer; we can just increment the end pointer to include
549 log_error(LOG_LEVEL_ERROR,
550 "Unable to determine compression format for decompression");
551 return JB_ERR_COMPRESS;
554 /* Set up the fields required by zlib. */
555 zstr.next_in = (Bytef *)cur;
556 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
557 zstr.zalloc = Z_NULL;
559 zstr.opaque = Z_NULL;
562 * Passing -MAX_WBITS to inflateInit2 tells the library
563 * that there is no zlib header.
565 if (inflateInit2(&zstr, -MAX_WBITS) != Z_OK)
567 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
568 return JB_ERR_COMPRESS;
572 * Next, we allocate new storage for the inflated data.
573 * We don't modify the existing iob yet, so in case there
574 * is error in decompression we can recover gracefully.
576 buf = zalloc(bufsize);
579 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
580 return JB_ERR_MEMORY;
583 assert(bufsize >= skip_size);
584 memcpy(buf, csp->iob->buf, skip_size);
585 zstr.avail_out = bufsize - skip_size;
586 zstr.next_out = (Bytef *)buf + skip_size;
588 /* Try to decompress the whole stream in one shot. */
589 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
591 /* We need to allocate more memory for the output buffer. */
593 char *tmpbuf; /* used for realloc'ing the buffer */
594 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
596 if (0 == zstr.avail_in)
599 * If zlib wants more data then there's a problem, because
600 * the complete compressed file should have been buffered.
602 log_error(LOG_LEVEL_ERROR,
603 "Unexpected end of compressed iob. Using what we got so far.");
608 * If we tried the limit and still didn't have enough
609 * memory, just give up.
611 if (bufsize == csp->config->buffer_limit)
613 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
614 return JB_ERR_MEMORY;
617 /* Try doubling the buffer size each time. */
620 /* Don't exceed the buffer limit. */
621 if (bufsize > csp->config->buffer_limit)
623 bufsize = csp->config->buffer_limit;
626 /* Try to allocate the new buffer. */
627 tmpbuf = realloc(buf, bufsize);
630 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
632 return JB_ERR_MEMORY;
636 char *oldnext_out = (char *)zstr.next_out;
639 * Update the fields for inflate() to use the new
640 * buffer, which may be in a location different from
643 zstr.avail_out += bufsize - oldbufsize;
644 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
647 * Compare with an uglier method of calculating these values
648 * that doesn't require the extra oldbufsize variable.
650 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
651 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
652 assert(zstr.avail_out > 0U);
658 if (Z_STREAM_ERROR == inflateEnd(&zstr))
660 log_error(LOG_LEVEL_ERROR,
661 "Inconsistent stream state after decompression: %s", zstr.msg);
663 * XXX: Intentionally no return.
665 * According to zlib.h, Z_STREAM_ERROR is returned
666 * "if the stream state was inconsistent".
668 * I assume in this case inflate()'s status
669 * would also be something different than Z_STREAM_END
670 * so this check should be redundant, but lets see.
674 if ((status != Z_STREAM_END) && (0 != zstr.avail_in))
677 * We failed to decompress the stream and it's
678 * not simply because of missing data.
680 log_error(LOG_LEVEL_ERROR,
681 "Unexpected error while decompressing to the buffer (iob): %s",
683 return JB_ERR_COMPRESS;
687 * Finally, we can actually update the iob, since the
688 * decompression was successful. First, free the old
691 freez(csp->iob->buf);
693 /* Now, update the iob to use the new buffer. */
695 csp->iob->cur = csp->iob->buf + skip_size;
696 csp->iob->eod = (char *)zstr.next_out;
697 csp->iob->size = bufsize;
700 * Make sure the new uncompressed iob obeys some minimal
701 * consistency conditions.
703 if ((csp->iob->buf < csp->iob->cur)
704 && (csp->iob->cur <= csp->iob->eod)
705 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
707 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
708 if (new_size > (size_t)0)
710 log_error(LOG_LEVEL_RE_FILTER,
711 "Decompression successful. Old size: %d, new size: %d.",
716 /* zlib thinks this is OK, so lets do the same. */
717 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
722 /* It seems that zlib did something weird. */
723 log_error(LOG_LEVEL_ERROR,
724 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
725 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
726 csp->iob->eod, csp->iob->buf + csp->iob->size);
727 return JB_ERR_COMPRESS;
733 #endif /* defined(FEATURE_ZLIB) */
736 /*********************************************************************
738 * Function : string_move
740 * Description : memmove wrapper to move the last part of a string
741 * towards the beginning, overwriting the part in
742 * the middle. strlcpy() can't be used here as the
746 * 1 : dst = Destination to overwrite
747 * 2 : src = Source to move.
751 *********************************************************************/
752 static void string_move(char *dst, char *src)
756 /* +1 to copy the terminating nul as well. */
757 memmove(dst, src, strlen(src)+1);
761 /*********************************************************************
763 * Function : normalize_lws
765 * Description : Reduces unquoted linear white space in headers
766 * to a single space in accordance with RFC 2616 2.2.
767 * This simplifies parsing and filtering later on.
769 * XXX: Remove log messages before
770 * the next stable release?
773 * 1 : header = A header with linear white space to reduce.
777 *********************************************************************/
778 static void normalize_lws(char *header)
784 if (ijb_isspace(*p) && ijb_isspace(*(p+1)))
788 while (ijb_isspace(*q))
792 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
798 log_error(LOG_LEVEL_HEADER,
799 "Converting tab to space in '%s'", header);
804 char *end_of_token = strstr(p+1, "\"");
806 if (NULL != end_of_token)
808 /* Don't mess with quoted text. */
813 log_error(LOG_LEVEL_HEADER,
814 "Ignoring single quote in '%s'", header);
820 p = strchr(header, ':');
821 if ((p != NULL) && (p != header) && ijb_isspace(*(p-1)))
824 * There's still space before the colon.
832 /*********************************************************************
834 * Function : get_header
836 * Description : This (odd) routine will parse the csp->iob
837 * to get the next complete header.
840 * 1 : iob = The I/O buffer to parse, usually csp->iob.
842 * Returns : Any one of the following:
844 * 1) a pointer to a dynamically allocated string that contains a header line
845 * 2) NULL indicating that the end of the header was reached
846 * 3) "" indicating that the end of the iob was reached before finding
847 * a complete header line.
849 *********************************************************************/
850 char *get_header(struct iob *iob)
854 header = get_header_line(iob);
856 if ((header == NULL) || (*header == '\0'))
859 * No complete header read yet, tell the client.
864 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
867 * Header spans multiple lines, append the next one.
869 char *continued_header;
871 continued_header = get_header_line(iob);
872 if ((continued_header == NULL) || (*continued_header == '\0'))
875 * No complete header read yet, return what we got.
876 * XXX: Should "unread" header instead.
878 log_error(LOG_LEVEL_INFO,
879 "Failed to read a multi-line header properly: '%s'",
884 if (JB_ERR_OK != string_join(&header, continued_header))
886 log_error(LOG_LEVEL_FATAL,
887 "Out of memory while appending multiple headers.");
891 /* XXX: remove before next stable release. */
892 log_error(LOG_LEVEL_HEADER,
893 "Merged multiple header lines to: '%s'",
898 normalize_lws(header);
905 /*********************************************************************
907 * Function : get_header_line
909 * Description : This (odd) routine will parse the csp->iob
910 * to get the next header line.
913 * 1 : iob = The I/O buffer to parse, usually csp->iob.
915 * Returns : Any one of the following:
917 * 1) a pointer to a dynamically allocated string that contains a header line
918 * 2) NULL indicating that the end of the header was reached
919 * 3) "" indicating that the end of the iob was reached before finding
920 * a complete header line.
922 *********************************************************************/
923 static char *get_header_line(struct iob *iob)
927 if ((iob->cur == NULL)
928 || ((p = strchr(iob->cur, '\n')) == NULL))
930 return(""); /* couldn't find a complete header */
935 ret = strdup(iob->cur);
938 /* FIXME No way to handle error properly */
939 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
945 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
947 /* is this a blank line (i.e. the end of the header) ? */
959 /*********************************************************************
961 * Function : get_header_value
963 * Description : Get the value of a given header from a chained list
964 * of header lines or return NULL if no such header is
965 * present in the list.
968 * 1 : header_list = pointer to list
969 * 2 : header_name = string with name of header to look for.
970 * Trailing colon required, capitalization
973 * Returns : NULL if not found, else value of header
975 *********************************************************************/
976 char *get_header_value(const struct list *header_list, const char *header_name)
978 struct list_entry *cur_entry;
984 length = strlen(header_name);
986 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
990 if (!strncmpic(cur_entry->str, header_name, length))
993 * Found: return pointer to start of value
995 ret = cur_entry->str + length;
996 while (*ret && ijb_isspace(*ret)) ret++;
1010 /*********************************************************************
1012 * Function : scan_headers
1014 * Description : Scans headers, applies tags and updates action bits.
1017 * 1 : csp = Current client state (buffers, headers, etc...)
1019 * Returns : JB_ERR_OK
1021 *********************************************************************/
1022 static jb_err scan_headers(struct client_state *csp)
1024 struct list_entry *h; /* Header */
1025 jb_err err = JB_ERR_OK;
1027 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1029 /* Header crunch()ed in previous run? -> ignore */
1030 if (h->str == NULL) continue;
1031 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1032 err = header_tagger(csp, h->str);
1039 /*********************************************************************
1043 * Description : add, delete or modify lines in the HTTP header streams.
1044 * On entry, it receives a linked list of headers space
1045 * that was allocated dynamically (both the list nodes
1046 * and the header contents).
1048 * As a side effect it frees the space used by the original
1052 * 1 : csp = Current client state (buffers, headers, etc...)
1053 * 2 : filter_server_headers = Boolean to switch between
1054 * server and header filtering.
1056 * Returns : JB_ERR_OK in case off success, or
1057 * JB_ERR_MEMORY on out-of-memory error.
1059 *********************************************************************/
1060 jb_err sed(struct client_state *csp, int filter_server_headers)
1062 /* XXX: use more descriptive names. */
1063 struct list_entry *p;
1064 const struct parsers *v;
1065 const add_header_func_ptr *f;
1066 jb_err err = JB_ERR_OK;
1068 if (filter_server_headers)
1070 v = server_patterns;
1071 f = add_server_headers;
1075 v = client_patterns;
1076 f = add_client_headers;
1081 while ((err == JB_ERR_OK) && (v->str != NULL))
1083 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1085 /* Header crunch()ed in previous run? -> ignore */
1086 if (p->str == NULL) continue;
1088 /* Does the current parser handle this header? */
1089 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1090 (v->len == CHECK_EVERY_HEADER_REMAINING))
1092 err = v->parser(csp, &(p->str));
1098 /* place additional headers on the csp->headers list */
1099 while ((err == JB_ERR_OK) && (*f))
1109 /*********************************************************************
1111 * Function : update_server_headers
1113 * Description : Updates server headers after the body has been modified.
1116 * 1 : csp = Current client state (buffers, headers, etc...)
1118 * Returns : JB_ERR_OK in case off success, or
1119 * JB_ERR_MEMORY on out-of-memory error.
1121 *********************************************************************/
1122 jb_err update_server_headers(struct client_state *csp)
1124 jb_err err = JB_ERR_OK;
1126 static const struct parsers server_patterns_light[] = {
1127 { "Content-Length:", 15, server_adjust_content_length },
1128 { "Transfer-Encoding:", 18, server_transfer_coding },
1130 { "Content-Encoding:", 17, server_content_encoding },
1131 #endif /* def FEATURE_ZLIB */
1135 if (strncmpic(csp->http->cmd, "HEAD", 4))
1137 const struct parsers *v;
1138 struct list_entry *p;
1140 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1142 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1144 /* Header crunch()ed in previous run? -> ignore */
1145 if (p->str == NULL) continue;
1147 /* Does the current parser handle this header? */
1148 if (strncmpic(p->str, v->str, v->len) == 0)
1150 err = v->parser(csp, (char **)&(p->str));
1156 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1157 if ((JB_ERR_OK == err)
1158 && (csp->flags & CSP_FLAG_MODIFIED)
1159 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
1160 && !(csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET))
1164 create_content_length_header(csp->content_length, header, sizeof(header));
1165 err = enlist(csp->headers, header);
1166 if (JB_ERR_OK == err)
1168 log_error(LOG_LEVEL_HEADER,
1169 "Content modified with no Content-Length header set. "
1170 "Created: %s.", header);
1173 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1179 /*********************************************************************
1181 * Function : header_tagger
1183 * Description : Executes all text substitutions from applying
1184 * tag actions and saves the result as tag.
1186 * XXX: Shares enough code with filter_header() and
1187 * pcrs_filter_response() to warrant some helper functions.
1190 * 1 : csp = Current client state (buffers, headers, etc...)
1191 * 2 : header = Header that is used as tagger input
1193 * Returns : JB_ERR_OK on success and always succeeds
1195 *********************************************************************/
1196 static jb_err header_tagger(struct client_state *csp, char *header)
1198 int wanted_filter_type;
1199 int multi_action_index;
1203 struct file_list *fl;
1204 struct re_filterfile_spec *b;
1205 struct list_entry *tag_name;
1207 int found_filters = 0;
1208 const size_t header_length = strlen(header);
1210 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1212 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1213 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1217 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1218 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1221 /* Check if there are any filters */
1222 for (i = 0; i < MAX_AF_FILES; i++)
1235 if (0 == found_filters)
1237 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1238 "tagging enabled, but no taggers available.");
1242 for (i = 0; i < MAX_AF_FILES; i++)
1245 if ((NULL == fl) || (NULL == fl->f))
1248 * Either there are no filter files
1249 * left, or this filter file just
1250 * contains no valid filters.
1252 * Continue to be sure we don't miss
1253 * valid filter files that are chained
1254 * after empty or invalid ones.
1259 /* For all filters, */
1260 for (b = fl->f; b; b = b->next)
1262 if (b->type != wanted_filter_type)
1264 /* skip the ones we don't care about, */
1267 /* leaving only taggers that could apply, of which we use the ones, */
1268 for (tag_name = csp->action->multi[multi_action_index]->first;
1269 NULL != tag_name; tag_name = tag_name->next)
1271 /* that do apply, and */
1272 if (strcmp(b->name, tag_name->str) == 0)
1274 char *modified_tag = NULL;
1276 size_t size = header_length;
1277 pcrs_job *joblist = b->joblist;
1279 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1281 if (NULL == joblist)
1283 log_error(LOG_LEVEL_RE_FILTER,
1284 "Tagger %s has empty joblist. Nothing to do.", b->name);
1288 /* execute their pcrs_joblist on the header. */
1289 for (job = joblist; NULL != job; job = job->next)
1291 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1295 /* Success, continue with the modified version. */
1304 /* Tagger doesn't match */
1307 /* Regex failure, log it but continue anyway. */
1308 assert(NULL != header);
1309 log_error(LOG_LEVEL_ERROR,
1310 "Problems with tagger \'%s\' and header \'%s\': %s",
1311 b->name, *header, pcrs_strerror(hits));
1313 freez(modified_tag);
1317 if (b->dynamic) pcrs_free_joblist(joblist);
1319 /* If this tagger matched */
1325 * There is to technical limitation which makes
1326 * it impossible to use empty tags, but I assume
1327 * no one would do it intentionally.
1330 log_error(LOG_LEVEL_INFO,
1331 "Tagger \'%s\' created an empty tag. Ignored.",
1336 if (!list_contains_item(csp->tags, tag))
1338 if (JB_ERR_OK != enlist(csp->tags, tag))
1340 log_error(LOG_LEVEL_ERROR,
1341 "Insufficient memory to add tag \'%s\', "
1342 "based on tagger \'%s\' and header \'%s\'",
1343 tag, b->name, *header);
1347 char *action_message;
1349 * update the action bits right away, to make
1350 * tagging based on tags set by earlier taggers
1351 * of the same kind possible.
1353 if (update_action_bits_for_tag(csp, tag))
1355 action_message = "Action bits updated accordingly.";
1359 action_message = "No action bits update necessary.";
1362 log_error(LOG_LEVEL_HEADER,
1363 "Tagger \'%s\' added tag \'%s\'. %s",
1364 b->name, tag, action_message);
1369 /* XXX: Is this log-worthy? */
1370 log_error(LOG_LEVEL_HEADER,
1371 "Tagger \'%s\' didn't add tag \'%s\'. "
1372 "Tag already present", b->name, tag);
1375 } /* if the tagger matched */
1376 } /* if the tagger applies */
1377 } /* for every tagger that could apply */
1378 } /* for all filters */
1379 } /* for all filter files */
1384 /* here begins the family of parser functions that reformat header lines */
1386 /*********************************************************************
1388 * Function : filter_header
1390 * Description : Executes all text substitutions from all applying
1391 * +(server|client)-header-filter actions on the header.
1392 * Most of the code was copied from pcrs_filter_response,
1393 * including the rather short variable names
1396 * 1 : csp = Current client state (buffers, headers, etc...)
1397 * 2 : header = On input, pointer to header to modify.
1398 * On output, pointer to the modified header, or NULL
1399 * to remove the header. This function frees the
1400 * original string if necessary.
1402 * Returns : JB_ERR_OK on success and always succeeds
1404 *********************************************************************/
1405 static jb_err filter_header(struct client_state *csp, char **header)
1409 size_t size = strlen(*header);
1411 char *newheader = NULL;
1414 struct file_list *fl;
1415 struct re_filterfile_spec *b;
1416 struct list_entry *filtername;
1418 int i, found_filters = 0;
1419 int wanted_filter_type;
1420 int multi_action_index;
1422 if (csp->flags & CSP_FLAG_NO_FILTERING)
1427 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1429 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1430 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1434 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1435 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1439 * Need to check the set of re_filterfiles...
1441 for (i = 0; i < MAX_AF_FILES; i++)
1454 if (0 == found_filters)
1456 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1457 "header filtering enabled, but no matching filters available.");
1461 for (i = 0; i < MAX_AF_FILES; i++)
1464 if ((NULL == fl) || (NULL == fl->f))
1467 * Either there are no filter files
1468 * left, or this filter file just
1469 * contains no valid filters.
1471 * Continue to be sure we don't miss
1472 * valid filter files that are chained
1473 * after empty or invalid ones.
1478 * For all applying +filter actions, look if a filter by that
1479 * name exists and if yes, execute its pcrs_joblist on the
1482 for (b = fl->f; b; b = b->next)
1484 if (b->type != wanted_filter_type)
1486 /* Skip other filter types */
1490 for (filtername = csp->action->multi[multi_action_index]->first;
1491 filtername ; filtername = filtername->next)
1493 if (strcmp(b->name, filtername->str) == 0)
1495 int current_hits = 0;
1496 pcrs_job *joblist = b->joblist;
1498 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1500 if (NULL == joblist)
1502 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1506 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1507 *header, size, b->name);
1509 /* Apply all jobs from the joblist */
1510 for (job = joblist; NULL != job; job = job->next)
1512 matches = pcrs_execute(job, *header, size, &newheader, &size);
1515 current_hits += matches;
1516 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1518 *header = newheader;
1520 else if ( 0 == matches )
1522 /* Filter doesn't change header */
1528 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1529 *header, b->name, pcrs_strerror(matches));
1530 if (newheader != NULL)
1532 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1538 if (b->dynamic) pcrs_free_joblist(joblist);
1540 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1541 hits += current_hits;
1548 * Additionally checking for hits is important because if
1549 * the continue hack is triggered, server headers can
1550 * arrive empty to separate multiple heads from each other.
1552 if ((0 == size) && hits)
1554 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
1562 /*********************************************************************
1564 * Function : server_connection
1566 * Description : Makes sure a proper "Connection:" header is
1567 * set and signals connection_header_adder to
1571 * 1 : csp = Current client state (buffers, headers, etc...)
1572 * 2 : header = On input, pointer to header to modify.
1573 * On output, pointer to the modified header, or NULL
1574 * to remove the header. This function frees the
1575 * original string if necessary.
1577 * Returns : JB_ERR_OK on success, or
1578 * JB_ERR_MEMORY on out-of-memory error.
1580 *********************************************************************/
1581 static jb_err server_connection(struct client_state *csp, char **header)
1583 if (!strcmpic(*header, "Connection: keep-alive")
1584 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1585 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
1589 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1590 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1592 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
1595 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE))
1597 log_error(LOG_LEVEL_HEADER,
1598 "Keeping the server header '%s' around.", *header);
1601 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
1603 char *old_header = *header;
1605 *header = strdup("Connection: close");
1608 return JB_ERR_MEMORY;
1610 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
1615 /* Signal server_connection_adder() to return early. */
1616 csp->flags |= CSP_FLAG_SERVER_CONNECTION_HEADER_SET;
1622 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1623 /*********************************************************************
1625 * Function : server_keep_alive
1627 * Description : Stores the servers keep alive timeout.
1630 * 1 : csp = Current client state (buffers, headers, etc...)
1631 * 2 : header = On input, pointer to header to modify.
1632 * On output, pointer to the modified header, or NULL
1633 * to remove the header. This function frees the
1634 * original string if necessary.
1636 * Returns : JB_ERR_OK.
1638 *********************************************************************/
1639 static jb_err server_keep_alive(struct client_state *csp, char **header)
1641 unsigned int keep_alive_timeout;
1642 const char *timeout_position = strstr(*header, "timeout=");
1644 if ((NULL == timeout_position)
1645 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1647 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1651 if (keep_alive_timeout < csp->server_connection.keep_alive_timeout)
1653 log_error(LOG_LEVEL_HEADER,
1654 "Reducing keep-alive timeout from %u to %u.",
1655 csp->server_connection.keep_alive_timeout, keep_alive_timeout);
1656 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1660 /* XXX: Is this log worthy? */
1661 log_error(LOG_LEVEL_HEADER,
1662 "Server keep-alive timeout is %u. Sticking with %u.",
1663 keep_alive_timeout, csp->server_connection.keep_alive_timeout);
1669 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1673 /*********************************************************************
1675 * Function : client_connection
1677 * Description : Makes sure a proper "Connection:" header is
1678 * set and signals connection_header_adder
1682 * 1 : csp = Current client state (buffers, headers, etc...)
1683 * 2 : header = On input, pointer to header to modify.
1684 * On output, pointer to the modified header, or NULL
1685 * to remove the header. This function frees the
1686 * original string if necessary.
1688 * Returns : JB_ERR_OK on success, or
1689 * JB_ERR_MEMORY on out-of-memory error.
1691 *********************************************************************/
1692 static jb_err client_connection(struct client_state *csp, char **header)
1694 const char *wanted_header = get_appropiate_connection_header(csp);
1696 if (strcmpic(*header, wanted_header))
1698 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1699 if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING))
1701 log_error(LOG_LEVEL_HEADER,
1702 "Keeping the client header '%s' around. "
1703 "The connection will not be kept alive.",
1707 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1709 char *old_header = *header;
1711 *header = strdup(wanted_header);
1714 return JB_ERR_MEMORY;
1716 log_error(LOG_LEVEL_HEADER,
1717 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1721 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1724 log_error(LOG_LEVEL_HEADER,
1725 "Keeping the client header '%s' around. "
1726 "The server connection will be kept alive if possible.",
1728 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1730 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1732 /* Signal client_connection_adder() to return early. */
1733 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
1739 /*********************************************************************
1741 * Function : crumble
1743 * Description : This is called if a header matches a pattern to "crunch"
1746 * 1 : csp = Current client state (buffers, headers, etc...)
1747 * 2 : header = On input, pointer to header to modify.
1748 * On output, pointer to the modified header, or NULL
1749 * to remove the header. This function frees the
1750 * original string if necessary.
1752 * Returns : JB_ERR_OK on success, or
1753 * JB_ERR_MEMORY on out-of-memory error.
1755 *********************************************************************/
1756 static jb_err crumble(struct client_state *csp, char **header)
1759 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
1765 /*********************************************************************
1767 * Function : crunch_server_header
1769 * Description : Crunch server header if it matches a string supplied by the
1770 * user. Called from `sed'.
1773 * 1 : csp = Current client state (buffers, headers, etc...)
1774 * 2 : header = On input, pointer to header to modify.
1775 * On output, pointer to the modified header, or NULL
1776 * to remove the header. This function frees the
1777 * original string if necessary.
1779 * Returns : JB_ERR_OK on success and always succeeds
1781 *********************************************************************/
1782 static jb_err crunch_server_header(struct client_state *csp, char **header)
1784 const char *crunch_pattern;
1786 /* Do we feel like crunching? */
1787 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
1789 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
1791 /* Is the current header the lucky one? */
1792 if (strstr(*header, crunch_pattern))
1794 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
1803 /*********************************************************************
1805 * Function : server_content_type
1807 * Description : Set the content-type for filterable types (text/.*,
1808 * .*xml.*, javascript and image/gif) unless filtering has been
1809 * forbidden (CT_TABOO) while parsing earlier headers.
1810 * NOTE: Since text/plain is commonly used by web servers
1811 * for files whose correct type is unknown, we don't
1812 * set CT_TEXT for it.
1815 * 1 : csp = Current client state (buffers, headers, etc...)
1816 * 2 : header = On input, pointer to header to modify.
1817 * On output, pointer to the modified header, or NULL
1818 * to remove the header. This function frees the
1819 * original string if necessary.
1821 * Returns : JB_ERR_OK on success, or
1822 * JB_ERR_MEMORY on out-of-memory error.
1824 *********************************************************************/
1825 static jb_err server_content_type(struct client_state *csp, char **header)
1827 /* Remove header if it isn't the first Content-Type header */
1828 if ((csp->content_type & CT_DECLARED))
1831 * Another, slightly slower, way to see if
1832 * we already parsed another Content-Type header.
1834 assert(NULL != get_header_value(csp->headers, "Content-Type:"));
1836 log_error(LOG_LEVEL_ERROR,
1837 "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
1845 * Signal that the Content-Type has been set.
1847 csp->content_type |= CT_DECLARED;
1849 if (!(csp->content_type & CT_TABOO))
1852 * XXX: The assumption that text/plain is a sign of
1853 * binary data seems to be somewhat unreasonable nowadays
1854 * and should be dropped after 3.0.8 is out.
1856 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
1857 || strstr(*header, "xml")
1858 || strstr(*header, "application/x-javascript"))
1860 csp->content_type |= CT_TEXT;
1862 else if (strstr(*header, "image/gif"))
1864 csp->content_type |= CT_GIF;
1869 * Are we messing with the content type?
1871 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
1874 * Make sure the user doesn't accidently
1875 * change the content type of binary documents.
1877 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
1880 *header = strdup("Content-Type: ");
1881 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
1885 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
1886 return JB_ERR_MEMORY;
1888 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
1892 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
1893 "It doesn't look like a content type that should be filtered. "
1894 "Enable force-text-mode if you know what you're doing.", *header);
1902 /*********************************************************************
1904 * Function : server_transfer_coding
1906 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
1907 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
1908 * - Remove header if body was chunked but has been
1909 * de-chunked for filtering.
1912 * 1 : csp = Current client state (buffers, headers, etc...)
1913 * 2 : header = On input, pointer to header to modify.
1914 * On output, pointer to the modified header, or NULL
1915 * to remove the header. This function frees the
1916 * original string if necessary.
1918 * Returns : JB_ERR_OK on success, or
1919 * JB_ERR_MEMORY on out-of-memory error.
1921 *********************************************************************/
1922 static jb_err server_transfer_coding(struct client_state *csp, char **header)
1925 * Turn off pcrs and gif filtering if body compressed
1927 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
1931 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
1933 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
1934 csp->http->cmd, *header);
1935 #endif /* def FEATURE_ZLIB */
1936 csp->content_type = CT_TABOO;
1940 * Raise flag if body chunked
1942 if (strstr(*header, "chunked"))
1944 csp->flags |= CSP_FLAG_CHUNKED;
1947 * If the body was modified, it has been de-chunked first
1948 * and the header must be removed.
1950 * FIXME: If there is more than one transfer encoding,
1951 * only the "chunked" part should be removed here.
1953 if (csp->flags & CSP_FLAG_MODIFIED)
1955 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
1964 /*********************************************************************
1966 * Function : server_content_encoding
1968 * Description : This function is run twice for each request,
1969 * unless FEATURE_ZLIB and filtering are disabled.
1971 * The first run is used to check if the content
1972 * is compressed, if FEATURE_ZLIB is disabled
1973 * filtering is then disabled as well, if FEATURE_ZLIB
1974 * is enabled the content is marked for decompression.
1976 * The second run is used to remove the Content-Encoding
1977 * header if the decompression was successful.
1980 * 1 : csp = Current client state (buffers, headers, etc...)
1981 * 2 : header = On input, pointer to header to modify.
1982 * On output, pointer to the modified header, or NULL
1983 * to remove the header. This function frees the
1984 * original string if necessary.
1986 * Returns : JB_ERR_OK on success, or
1987 * JB_ERR_MEMORY on out-of-memory error.
1989 *********************************************************************/
1990 static jb_err server_content_encoding(struct client_state *csp, char **header)
1993 if ((csp->flags & CSP_FLAG_MODIFIED)
1994 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
1997 * We successfully decompressed the content,
1998 * and have to clean the header now, so the
1999 * client no longer expects compressed data..
2001 * XXX: There is a difference between cleaning
2002 * and removing it completely.
2004 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2007 else if (strstr(*header, "gzip"))
2009 /* Mark for gzip decompression */
2010 csp->content_type |= CT_GZIP;
2012 else if (strstr(*header, "deflate"))
2014 /* Mark for zlib decompression */
2015 csp->content_type |= CT_DEFLATE;
2017 else if (strstr(*header, "compress"))
2020 * We can't decompress this; therefore we can't filter
2023 csp->content_type |= CT_TABOO;
2025 #else /* !defined(FEATURE_ZLIB) */
2026 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2029 * Body is compressed, turn off pcrs and gif filtering.
2031 csp->content_type |= CT_TABOO;
2034 * Log a warning if the user expects the content to be filtered.
2036 if ((csp->rlist != NULL) &&
2037 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2039 log_error(LOG_LEVEL_INFO,
2040 "Compressed content detected, content filtering disabled. "
2041 "Consider recompiling Privoxy with zlib support or "
2042 "enable the prevent-compression action.");
2045 #endif /* defined(FEATURE_ZLIB) */
2052 /*********************************************************************
2054 * Function : server_adjust_content_length
2056 * Description : Adjust Content-Length header if we modified
2060 * 1 : csp = Current client state (buffers, headers, etc...)
2061 * 2 : header = On input, pointer to header to modify.
2062 * On output, pointer to the modified header, or NULL
2063 * to remove the header. This function frees the
2064 * original string if necessary.
2066 * Returns : JB_ERR_OK on success, or
2067 * JB_ERR_MEMORY on out-of-memory error.
2069 *********************************************************************/
2070 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2072 /* Regenerate header if the content was modified. */
2073 if (csp->flags & CSP_FLAG_MODIFIED)
2075 const size_t header_length = 50;
2077 *header = malloc(header_length);
2078 if (*header == NULL)
2080 return JB_ERR_MEMORY;
2082 create_content_length_header(csp->content_length, *header, header_length);
2083 log_error(LOG_LEVEL_HEADER,
2084 "Adjusted Content-Length to %llu", csp->content_length);
2091 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2092 /*********************************************************************
2094 * Function : server_save_content_length
2096 * Description : Save the Content-Length sent by the server.
2099 * 1 : csp = Current client state (buffers, headers, etc...)
2100 * 2 : header = On input, pointer to header to modify.
2101 * On output, pointer to the modified header, or NULL
2102 * to remove the header. This function frees the
2103 * original string if necessary.
2105 * Returns : JB_ERR_OK on success, or
2106 * JB_ERR_MEMORY on out-of-memory error.
2108 *********************************************************************/
2109 static jb_err server_save_content_length(struct client_state *csp, char **header)
2111 unsigned long long content_length = 0;
2113 assert(*(*header+14) == ':');
2115 if (1 != sscanf(*header+14, ": %llu", &content_length))
2117 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2122 csp->expected_content_length = content_length;
2123 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
2124 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2129 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2132 /*********************************************************************
2134 * Function : server_content_md5
2136 * Description : Crumble any Content-MD5 headers if the document was
2137 * modified. FIXME: Should we re-compute instead?
2140 * 1 : csp = Current client state (buffers, headers, etc...)
2141 * 2 : header = On input, pointer to header to modify.
2142 * On output, pointer to the modified header, or NULL
2143 * to remove the header. This function frees the
2144 * original string if necessary.
2146 * Returns : JB_ERR_OK on success, or
2147 * JB_ERR_MEMORY on out-of-memory error.
2149 *********************************************************************/
2150 static jb_err server_content_md5(struct client_state *csp, char **header)
2152 if (csp->flags & CSP_FLAG_MODIFIED)
2154 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2162 /*********************************************************************
2164 * Function : server_content_disposition
2166 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2167 * Called from `sed'.
2170 * 1 : csp = Current client state (buffers, headers, etc...)
2171 * 2 : header = On input, pointer to header to modify.
2172 * On output, pointer to the modified header, or NULL
2173 * to remove the header. This function frees the
2174 * original string if necessary.
2176 * Returns : JB_ERR_OK on success, or
2177 * JB_ERR_MEMORY on out-of-memory error.
2179 *********************************************************************/
2180 static jb_err server_content_disposition(struct client_state *csp, char **header)
2185 * Are we messing with the Content-Disposition header?
2187 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2193 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2195 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2198 * Blocking content-disposition header
2200 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2207 * Replacing Content-Disposition header
2210 *header = strdup("Content-Disposition: ");
2211 string_append(header, newval);
2213 if (*header != NULL)
2215 log_error(LOG_LEVEL_HEADER,
2216 "Content-Disposition header crunched and replaced with: %s", *header);
2219 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2223 /*********************************************************************
2225 * Function : server_last_modified
2227 * Description : Changes Last-Modified header to the actual date
2228 * to help hide-if-modified-since.
2229 * Called from `sed'.
2232 * 1 : csp = Current client state (buffers, headers, etc...)
2233 * 2 : header = On input, pointer to header to modify.
2234 * On output, pointer to the modified header, or NULL
2235 * to remove the header. This function frees the
2236 * original string if necessary.
2238 * Returns : JB_ERR_OK on success, or
2239 * JB_ERR_MEMORY on out-of-memory error.
2241 *********************************************************************/
2242 static jb_err server_last_modified(struct client_state *csp, char **header)
2245 char buf[BUFFER_SIZE];
2248 #ifdef HAVE_GMTIME_R
2251 struct tm *timeptr = NULL;
2252 time_t now, last_modified;
2253 long int days, hours, minutes, seconds;
2256 * Are we messing with the Last-Modified header?
2258 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2264 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2266 if (0 == strcmpic(newval, "block") )
2269 * Blocking Last-Modified header. Useless but why not.
2271 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2275 else if (0 == strcmpic(newval, "reset-to-request-time"))
2278 * Setting Last-Modified Header to now.
2280 get_http_time(0, buf, sizeof(buf));
2282 *header = strdup("Last-Modified: ");
2283 string_append(header, buf);
2285 if (*header == NULL)
2287 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2291 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2294 else if (0 == strcmpic(newval, "randomize"))
2296 const char *header_time = *header + sizeof("Last-Modified:");
2298 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2300 #ifdef HAVE_GMTIME_R
2301 gmtime_r(&now, &gmt);
2302 #elif defined(MUTEX_LOCKS_AVAILABLE)
2303 privoxy_mutex_lock(&gmtime_mutex);
2305 privoxy_mutex_unlock(&gmtime_mutex);
2309 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2311 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2316 long int rtime = (long int)difftime(now, last_modified);
2319 const int negative_delta = (rtime < 0);
2324 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2326 rtime = pick_from_range(rtime);
2331 last_modified += rtime;
2332 #ifdef HAVE_GMTIME_R
2333 timeptr = gmtime_r(&last_modified, &gmt);
2334 #elif defined(MUTEX_LOCKS_AVAILABLE)
2335 privoxy_mutex_lock(&gmtime_mutex);
2336 timeptr = gmtime(&last_modified);
2337 privoxy_mutex_unlock(&gmtime_mutex);
2339 timeptr = gmtime(&last_modified);
2341 if ((NULL == timeptr) || !strftime(newheader,
2342 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
2344 log_error(LOG_LEVEL_ERROR,
2345 "Randomizing '%s' failed. Crunching the header without replacement.",
2352 *header = strdup("Last-Modified: ");
2353 string_append(header, newheader);
2355 if (*header == NULL)
2357 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2358 return JB_ERR_MEMORY;
2361 days = rtime / (3600 * 24);
2362 hours = rtime / 3600 % 24;
2363 minutes = rtime / 60 % 60;
2364 seconds = rtime % 60;
2366 log_error(LOG_LEVEL_HEADER,
2367 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2368 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2369 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2373 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2382 /*********************************************************************
2384 * Function : client_accept_encoding
2386 * Description : Rewrite the client's Accept-Encoding header so that
2387 * if doesn't allow compression, if the action applies.
2388 * Note: For HTTP/1.0 the absence of the header is enough.
2391 * 1 : csp = Current client state (buffers, headers, etc...)
2392 * 2 : header = On input, pointer to header to modify.
2393 * On output, pointer to the modified header, or NULL
2394 * to remove the header. This function frees the
2395 * original string if necessary.
2397 * Returns : JB_ERR_OK on success, or
2398 * JB_ERR_MEMORY on out-of-memory error.
2400 *********************************************************************/
2401 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2403 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2405 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2409 /* Temporarily disable the correct behaviour to
2410 * work around a PHP bug.
2412 * if (!strcmpic(csp->http->ver, "HTTP/1.1"))
2414 * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0");
2415 * if (*header == NULL)
2417 * return JB_ERR_MEMORY;
2428 /*********************************************************************
2430 * Function : client_te
2432 * Description : Rewrite the client's TE header so that
2433 * if doesn't allow compression, if the action applies.
2436 * 1 : csp = Current client state (buffers, headers, etc...)
2437 * 2 : header = On input, pointer to header to modify.
2438 * On output, pointer to the modified header, or NULL
2439 * to remove the header. This function frees the
2440 * original string if necessary.
2442 * Returns : JB_ERR_OK on success, or
2443 * JB_ERR_MEMORY on out-of-memory error.
2445 *********************************************************************/
2446 static jb_err client_te(struct client_state *csp, char **header)
2448 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2451 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2458 /*********************************************************************
2460 * Function : client_referrer
2462 * Description : Handle the "referer" config setting properly.
2463 * Called from `sed'.
2466 * 1 : csp = Current client state (buffers, headers, etc...)
2467 * 2 : header = On input, pointer to header to modify.
2468 * On output, pointer to the modified header, or NULL
2469 * to remove the header. This function frees the
2470 * original string if necessary.
2472 * Returns : JB_ERR_OK on success, or
2473 * JB_ERR_MEMORY on out-of-memory error.
2475 *********************************************************************/
2476 static jb_err client_referrer(struct client_state *csp, char **header)
2478 const char *parameter;
2479 /* booleans for parameters we have to check multiple times */
2480 int parameter_conditional_block;
2481 int parameter_conditional_forge;
2483 #ifdef FEATURE_FORCE_LOAD
2485 * Since the referrer can include the prefix even
2486 * if the request itself is non-forced, we must
2487 * clean it unconditionally.
2489 * XXX: strclean is too broad
2491 strclean(*header, FORCE_PREFIX);
2492 #endif /* def FEATURE_FORCE_LOAD */
2494 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2496 /* Nothing left to do */
2500 parameter = csp->action->string[ACTION_STRING_REFERER];
2501 assert(parameter != NULL);
2502 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2503 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2505 if (!parameter_conditional_block && !parameter_conditional_forge)
2508 * As conditional-block and conditional-forge are the only
2509 * parameters that rely on the original referrer, we can
2510 * remove it now for all the others.
2515 if (0 == strcmpic(parameter, "block"))
2517 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2520 else if (parameter_conditional_block || parameter_conditional_forge)
2522 return handle_conditional_hide_referrer_parameter(header,
2523 csp->http->hostport, parameter_conditional_block);
2525 else if (0 == strcmpic(parameter, "forge"))
2527 return create_forged_referrer(header, csp->http->hostport);
2531 /* interpret parameter as user-supplied referer to fake */
2532 return create_fake_referrer(header, parameter);
2537 /*********************************************************************
2539 * Function : client_accept_language
2541 * Description : Handle the "Accept-Language" config setting properly.
2542 * Called from `sed'.
2545 * 1 : csp = Current client state (buffers, headers, etc...)
2546 * 2 : header = On input, pointer to header to modify.
2547 * On output, pointer to the modified header, or NULL
2548 * to remove the header. This function frees the
2549 * original string if necessary.
2551 * Returns : JB_ERR_OK on success, or
2552 * JB_ERR_MEMORY on out-of-memory error.
2554 *********************************************************************/
2555 static jb_err client_accept_language(struct client_state *csp, char **header)
2560 * Are we messing with the Accept-Language?
2562 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2564 /*I don't think so*/
2568 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2570 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
2573 * Blocking Accept-Language header
2575 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2582 * Replacing Accept-Language header
2585 *header = strdup("Accept-Language: ");
2586 string_append(header, newval);
2588 if (*header == NULL)
2590 log_error(LOG_LEVEL_ERROR,
2591 "Insufficient memory. Accept-Language header crunched without replacement.");
2595 log_error(LOG_LEVEL_HEADER,
2596 "Accept-Language header crunched and replaced with: %s", *header);
2599 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2603 /*********************************************************************
2605 * Function : crunch_client_header
2607 * Description : Crunch client header if it matches a string supplied by the
2608 * user. Called from `sed'.
2611 * 1 : csp = Current client state (buffers, headers, etc...)
2612 * 2 : header = On input, pointer to header to modify.
2613 * On output, pointer to the modified header, or NULL
2614 * to remove the header. This function frees the
2615 * original string if necessary.
2617 * Returns : JB_ERR_OK on success and always succeeds
2619 *********************************************************************/
2620 static jb_err crunch_client_header(struct client_state *csp, char **header)
2622 const char *crunch_pattern;
2624 /* Do we feel like crunching? */
2625 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2627 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2629 /* Is the current header the lucky one? */
2630 if (strstr(*header, crunch_pattern))
2632 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
2640 /*********************************************************************
2642 * Function : client_uagent
2644 * Description : Handle the "user-agent" config setting properly
2645 * and remember its original value to enable browser
2646 * bug workarounds. Called from `sed'.
2649 * 1 : csp = Current client state (buffers, headers, etc...)
2650 * 2 : header = On input, pointer to header to modify.
2651 * On output, pointer to the modified header, or NULL
2652 * to remove the header. This function frees the
2653 * original string if necessary.
2655 * Returns : JB_ERR_OK on success, or
2656 * JB_ERR_MEMORY on out-of-memory error.
2658 *********************************************************************/
2659 static jb_err client_uagent(struct client_state *csp, char **header)
2663 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
2668 newval = csp->action->string[ACTION_STRING_USER_AGENT];
2675 *header = strdup("User-Agent: ");
2676 string_append(header, newval);
2678 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
2680 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2684 /*********************************************************************
2686 * Function : client_ua
2688 * Description : Handle "ua-" headers properly. Called from `sed'.
2691 * 1 : csp = Current client state (buffers, headers, etc...)
2692 * 2 : header = On input, pointer to header to modify.
2693 * On output, pointer to the modified header, or NULL
2694 * to remove the header. This function frees the
2695 * original string if necessary.
2697 * Returns : JB_ERR_OK on success, or
2698 * JB_ERR_MEMORY on out-of-memory error.
2700 *********************************************************************/
2701 static jb_err client_ua(struct client_state *csp, char **header)
2703 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
2705 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
2713 /*********************************************************************
2715 * Function : client_from
2717 * Description : Handle the "from" config setting properly.
2718 * Called from `sed'.
2721 * 1 : csp = Current client state (buffers, headers, etc...)
2722 * 2 : header = On input, pointer to header to modify.
2723 * On output, pointer to the modified header, or NULL
2724 * to remove the header. This function frees the
2725 * original string if necessary.
2727 * Returns : JB_ERR_OK on success, or
2728 * JB_ERR_MEMORY on out-of-memory error.
2730 *********************************************************************/
2731 static jb_err client_from(struct client_state *csp, char **header)
2735 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
2742 newval = csp->action->string[ACTION_STRING_FROM];
2745 * Are we blocking the e-mail address?
2747 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
2749 log_error(LOG_LEVEL_HEADER, "crunched From!");
2753 log_error(LOG_LEVEL_HEADER, " modified");
2755 *header = strdup("From: ");
2756 string_append(header, newval);
2758 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2762 /*********************************************************************
2764 * Function : client_send_cookie
2766 * Description : Crunches the "cookie" header if necessary.
2767 * Called from `sed'.
2769 * XXX: Stupid name, doesn't send squat.
2772 * 1 : csp = Current client state (buffers, headers, etc...)
2773 * 2 : header = On input, pointer to header to modify.
2774 * On output, pointer to the modified header, or NULL
2775 * to remove the header. This function frees the
2776 * original string if necessary.
2778 * Returns : JB_ERR_OK on success, or
2779 * JB_ERR_MEMORY on out-of-memory error.
2781 *********************************************************************/
2782 static jb_err client_send_cookie(struct client_state *csp, char **header)
2784 if (csp->action->flags & ACTION_NO_COOKIE_READ)
2786 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
2794 /*********************************************************************
2796 * Function : client_x_forwarded
2798 * Description : Handle the "x-forwarded-for" config setting properly,
2799 * also used in the add_client_headers list. Called from `sed'.
2802 * 1 : csp = Current client state (buffers, headers, etc...)
2803 * 2 : header = On input, pointer to header to modify.
2804 * On output, pointer to the modified header, or NULL
2805 * to remove the header. This function frees the
2806 * original string if necessary.
2808 * Returns : JB_ERR_OK on success, or
2809 * JB_ERR_MEMORY on out-of-memory error.
2811 *********************************************************************/
2812 jb_err client_x_forwarded(struct client_state *csp, char **header)
2814 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
2816 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
2818 if (0 == strcmpic(parameter, "block"))
2821 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
2823 else if (0 == strcmpic(parameter, "add"))
2825 string_append(header, ", ");
2826 string_append(header, csp->ip_addr_str);
2828 if (*header == NULL)
2830 return JB_ERR_MEMORY;
2832 log_error(LOG_LEVEL_HEADER,
2833 "Appended client IP address to %s", *header);
2834 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
2838 log_error(LOG_LEVEL_FATAL,
2839 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
2847 /*********************************************************************
2849 * Function : client_max_forwards
2851 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
2852 * from the value of the Max-Forwards header field.
2855 * 1 : csp = Current client state (buffers, headers, etc...)
2856 * 2 : header = On input, pointer to header to modify.
2857 * On output, pointer to the modified header, or NULL
2858 * to remove the header. This function frees the
2859 * original string if necessary.
2861 * Returns : JB_ERR_OK on success, or
2862 * JB_ERR_MEMORY on out-of-memory error.
2864 *********************************************************************/
2865 static jb_err client_max_forwards(struct client_state *csp, char **header)
2869 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
2870 (0 == strcmpic(csp->http->gpc, "options")))
2872 assert(*(*header+12) == ':');
2873 if (1 == sscanf(*header+12, ": %d", &max_forwards))
2875 if (max_forwards > 0)
2877 snprintf(*header, strlen(*header)+1, "Max-Forwards: %d", --max_forwards);
2878 log_error(LOG_LEVEL_HEADER,
2879 "Max-Forwards value for %s request reduced to %d.",
2880 csp->http->gpc, max_forwards);
2882 else if (max_forwards < 0)
2884 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2890 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2899 /*********************************************************************
2901 * Function : client_host
2903 * Description : If the request URI did not contain host and
2904 * port information, parse and evaluate the Host
2907 * Also, kill ill-formed HOST: headers as sent by
2908 * Apple's iTunes software when used with a proxy.
2911 * 1 : csp = Current client state (buffers, headers, etc...)
2912 * 2 : header = On input, pointer to header to modify.
2913 * On output, pointer to the modified header, or NULL
2914 * to remove the header. This function frees the
2915 * original string if necessary.
2917 * Returns : JB_ERR_OK on success, or
2918 * JB_ERR_MEMORY on out-of-memory error.
2920 *********************************************************************/
2921 static jb_err client_host(struct client_state *csp, char **header)
2926 * If the header field name is all upper-case, chances are that it's
2927 * an ill-formed one from iTunes. BTW, killing innocent headers here is
2928 * not a problem -- they are regenerated later.
2930 if ((*header)[1] == 'O')
2932 log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
2937 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
2938 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
2941 if (NULL == (p = strdup((*header)+6)))
2943 return JB_ERR_MEMORY;
2946 if (NULL == (q = strdup(p)))
2949 return JB_ERR_MEMORY;
2952 freez(csp->http->hostport);
2953 csp->http->hostport = p;
2954 freez(csp->http->host);
2955 csp->http->host = q;
2956 q = strchr(csp->http->host, ':');
2959 /* Terminate hostname and evaluate port string */
2961 csp->http->port = atoi(q);
2965 csp->http->port = csp->http->ssl ? 443 : 80;
2968 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
2969 csp->http->hostport, csp->http->host, csp->http->port);
2972 /* Signal client_host_adder() to return right away */
2973 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
2979 /*********************************************************************
2981 * Function : client_if_modified_since
2983 * Description : Remove or modify the If-Modified-Since header.
2986 * 1 : csp = Current client state (buffers, headers, etc...)
2987 * 2 : header = On input, pointer to header to modify.
2988 * On output, pointer to the modified header, or NULL
2989 * to remove the header. This function frees the
2990 * original string if necessary.
2992 * Returns : JB_ERR_OK on success, or
2993 * JB_ERR_MEMORY on out-of-memory error.
2995 *********************************************************************/
2996 static jb_err client_if_modified_since(struct client_state *csp, char **header)
2999 #ifdef HAVE_GMTIME_R
3002 struct tm *timeptr = NULL;
3005 long int hours, minutes, seconds;
3008 if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3011 * The client got an error message because of a temporary problem,
3012 * the problem is gone and the client now tries to revalidate our
3013 * error message on the real server. The revalidation would always
3014 * end with the transmission of the whole document and there is
3015 * no need to expose the bogus If-Modified-Since header.
3017 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3020 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3022 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3024 if ((0 == strcmpic(newval, "block")))
3026 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3029 else /* add random value */
3031 const char *header_time = *header + sizeof("If-Modified-Since:");
3033 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3035 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3040 long int rtime = strtol(newval, &endptr, 0);
3041 const int negative_range = (rtime < 0);
3045 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3046 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3052 rtime = pick_from_range(rtime);
3056 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3059 tm += rtime * (negative_range ? -1 : 1);
3060 #ifdef HAVE_GMTIME_R
3061 timeptr = gmtime_r(&tm, &gmt);
3062 #elif defined(MUTEX_LOCKS_AVAILABLE)
3063 privoxy_mutex_lock(&gmtime_mutex);
3064 timeptr = gmtime(&tm);
3065 privoxy_mutex_unlock(&gmtime_mutex);
3067 timeptr = gmtime(&tm);
3069 if ((NULL == timeptr) || !strftime(newheader,
3070 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
3072 log_error(LOG_LEVEL_ERROR,
3073 "Randomizing '%s' failed. Crunching the header without replacement.",
3080 *header = strdup("If-Modified-Since: ");
3081 string_append(header, newheader);
3083 if (*header == NULL)
3085 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3086 return JB_ERR_MEMORY;
3089 hours = rtime / 3600;
3090 minutes = rtime / 60 % 60;
3091 seconds = rtime % 60;
3093 log_error(LOG_LEVEL_HEADER,
3094 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3095 *header, (negative_range) ? "subtracted" : "added", hours,
3096 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3097 seconds, (seconds == 1) ? ")" : "s)");
3106 /*********************************************************************
3108 * Function : client_if_none_match
3110 * Description : Remove the If-None-Match header.
3113 * 1 : csp = Current client state (buffers, headers, etc...)
3114 * 2 : header = On input, pointer to header to modify.
3115 * On output, pointer to the modified header, or NULL
3116 * to remove the header. This function frees the
3117 * original string if necessary.
3119 * Returns : JB_ERR_OK on success, or
3120 * JB_ERR_MEMORY on out-of-memory error.
3122 *********************************************************************/
3123 static jb_err client_if_none_match(struct client_state *csp, char **header)
3125 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3127 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3135 /*********************************************************************
3137 * Function : client_x_filter
3139 * Description : Disables filtering if the client set "X-Filter: No".
3140 * Called from `sed'.
3143 * 1 : csp = Current client state (buffers, headers, etc...)
3144 * 2 : header = On input, pointer to header to modify.
3145 * On output, pointer to the modified header, or NULL
3146 * to remove the header. This function frees the
3147 * original string if necessary.
3149 * Returns : JB_ERR_OK on success
3151 *********************************************************************/
3152 jb_err client_x_filter(struct client_state *csp, char **header)
3154 if ( 0 == strcmpic(*header, "X-Filter: No"))
3156 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3158 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3162 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3164 log_error(LOG_LEVEL_HEADER,
3165 "force-text-mode overruled the client's request to fetch without filtering!");
3169 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3170 csp->flags |= CSP_FLAG_NO_FILTERING;
3171 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3173 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3181 /*********************************************************************
3183 * Function : client_range
3185 * Description : Removes Range, Request-Range and If-Range headers if
3186 * content filtering is enabled. If the client's version
3187 * of the document has been altered by Privoxy, the server
3188 * could interpret the range differently than the client
3189 * intended in which case the user could end up with
3190 * corrupted content.
3193 * 1 : csp = Current client state (buffers, headers, etc...)
3194 * 2 : header = On input, pointer to header to modify.
3195 * On output, pointer to the modified header, or NULL
3196 * to remove the header. This function frees the
3197 * original string if necessary.
3199 * Returns : JB_ERR_OK
3201 *********************************************************************/
3202 static jb_err client_range(struct client_state *csp, char **header)
3204 if (content_filters_enabled(csp->action))
3206 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3207 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3214 /* the following functions add headers directly to the header list */
3216 /*********************************************************************
3218 * Function : client_host_adder
3220 * Description : Adds the Host: header field if it is missing.
3221 * Called from `sed'.
3224 * 1 : csp = Current client state (buffers, headers, etc...)
3226 * Returns : JB_ERR_OK on success, or
3227 * JB_ERR_MEMORY on out-of-memory error.
3229 *********************************************************************/
3230 static jb_err client_host_adder(struct client_state *csp)
3235 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3237 /* Header already set by the client, nothing to do. */
3241 if ( !csp->http->hostport || !*(csp->http->hostport))
3243 /* XXX: When does this happen and why is it OK? */
3244 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3249 * remove 'user:pass@' from 'proto://user:pass@host'
3251 if ( (p = strchr( csp->http->hostport, '@')) != NULL )
3257 p = csp->http->hostport;
3260 /* XXX: Just add it, we already made sure that it will be unique */
3261 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3262 err = enlist_unique_header(csp->headers, "Host", p);
3269 /*********************************************************************
3271 * Function : client_accept_encoding_adder
3273 * Description : Add an Accept-Encoding header to the client's request
3274 * that disables compression if the action applies, and
3275 * the header is not already there. Called from `sed'.
3276 * Note: For HTTP/1.0, the absence of the header is enough.
3279 * 1 : csp = Current client state (buffers, headers, etc...)
3281 * Returns : JB_ERR_OK on success, or
3282 * JB_ERR_MEMORY on out-of-memory error.
3284 *********************************************************************/
3285 static jb_err client_accept_encoding_adder(struct client_state *csp)
3287 if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3288 && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
3290 return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
3298 /*********************************************************************
3300 * Function : client_xtra_adder
3302 * Description : Used in the add_client_headers list. Called from `sed'.
3305 * 1 : csp = Current client state (buffers, headers, etc...)
3307 * Returns : JB_ERR_OK on success, or
3308 * JB_ERR_MEMORY on out-of-memory error.
3310 *********************************************************************/
3311 static jb_err client_xtra_adder(struct client_state *csp)
3313 struct list_entry *lst;
3316 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3317 lst ; lst = lst->next)
3319 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3320 err = enlist(csp->headers, lst->str);
3332 /*********************************************************************
3334 * Function : client_x_forwarded_for_adder
3336 * Description : Used in the add_client_headers list. Called from `sed'.
3339 * 1 : csp = Current client state (buffers, headers, etc...)
3341 * Returns : JB_ERR_OK on success, or
3342 * JB_ERR_MEMORY on out-of-memory error.
3344 *********************************************************************/
3345 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3347 char *header = NULL;
3350 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3351 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3352 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3355 * If we aren't adding X-Forwarded-For headers,
3356 * or we already appended an existing X-Forwarded-For
3357 * header, there's nothing left to do here.
3362 header = strdup("X-Forwarded-For: ");
3363 string_append(&header, csp->ip_addr_str);
3367 return JB_ERR_MEMORY;
3370 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3371 err = enlist(csp->headers, header);
3378 /*********************************************************************
3380 * Function : server_connection_adder
3382 * Description : Adds an appropiate "Connection:" header to csp->headers
3383 * unless the header was already present. Called from `sed'.
3386 * 1 : csp = Current client state (buffers, headers, etc...)
3388 * Returns : JB_ERR_OK on success, or
3389 * JB_ERR_MEMORY on out-of-memory error.
3391 *********************************************************************/
3392 static jb_err server_connection_adder(struct client_state *csp)
3394 const unsigned int flags = csp->flags;
3395 const char *response_status_line = csp->headers->first->str;
3396 const char *wanted_header = get_appropiate_connection_header(csp);
3398 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3399 && (flags & CSP_FLAG_SERVER_CONNECTION_HEADER_SET))
3405 * XXX: if we downgraded the response, this check will fail.
3407 if ((csp->config->feature_flags &
3408 RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3409 && (NULL != response_status_line)
3410 && !strncmpic(response_status_line, "HTTP/1.1", 8)
3411 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3412 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3416 log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
3417 "without Connection header implies keep-alive.");
3418 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
3421 log_error(LOG_LEVEL_HEADER, "Adding: %s", wanted_header);
3423 return enlist(csp->headers, wanted_header);
3427 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3428 /*********************************************************************
3430 * Function : server_proxy_connection_adder
3432 * Description : Adds a "Proxy-Connection: keep-alive" header to
3433 * csp->headers if the client asked for keep-alive.
3434 * XXX: We should reuse existant ones.
3437 * 1 : csp = Current client state (buffers, headers, etc...)
3439 * Returns : JB_ERR_OK on success, or
3440 * JB_ERR_MEMORY on out-of-memory error.
3442 *********************************************************************/
3443 static jb_err server_proxy_connection_adder(struct client_state *csp)
3445 static const char proxy_connection_header[] = "Proxy-Connection: keep-alive";
3446 jb_err err = JB_ERR_OK;
3448 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
3449 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
3451 log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
3452 err = enlist(csp->headers, proxy_connection_header);
3457 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3460 /*********************************************************************
3462 * Function : client_connection_header_adder
3464 * Description : Adds a proper "Connection:" header to csp->headers
3465 * unless the header was already present. Called from `sed'.
3468 * 1 : csp = Current client state (buffers, headers, etc...)
3470 * Returns : JB_ERR_OK on success, or
3471 * JB_ERR_MEMORY on out-of-memory error.
3473 *********************************************************************/
3474 static jb_err client_connection_header_adder(struct client_state *csp)
3476 const char *wanted_header = get_appropiate_connection_header(csp);
3478 if (!(csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3479 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
3484 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3485 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3486 && (csp->http->ssl == 0)
3487 && !strcmpic(csp->http->ver, "HTTP/1.1"))
3489 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
3491 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3493 log_error(LOG_LEVEL_HEADER, "Adding: %s", wanted_header);
3495 return enlist(csp->headers, wanted_header);
3499 /*********************************************************************
3501 * Function : server_http
3503 * Description : - Save the HTTP Status into csp->http->status
3504 * - Set CT_TABOO to prevent filtering if the answer
3505 * is a partial range (HTTP status 206)
3506 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3510 * 1 : csp = Current client state (buffers, headers, etc...)
3511 * 2 : header = On input, pointer to header to modify.
3512 * On output, pointer to the modified header, or NULL
3513 * to remove the header. This function frees the
3514 * original string if necessary.
3516 * Returns : JB_ERR_OK on success, or
3517 * JB_ERR_MEMORY on out-of-memory error.
3519 *********************************************************************/
3520 static jb_err server_http(struct client_state *csp, char **header)
3522 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3523 if (csp->http->status == 206)
3525 csp->content_type = CT_TABOO;
3528 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3530 /* XXX: Should we do a real validity check here? */
3531 if (strlen(*header) > 8)
3534 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3539 * XXX: Should we block the request or
3540 * enlist a valid status code line here?
3542 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3543 "Downgrading to HTTP/1.0 impossible.");
3551 /*********************************************************************
3553 * Function : server_set_cookie
3555 * Description : Handle the server "cookie" header properly.
3556 * Log cookie to the jar file. Then "crunch",
3557 * accept or rewrite it to a session cookie.
3558 * Called from `sed'.
3560 * TODO: Allow the user to specify a new expiration
3561 * time to cause the cookie to expire even before the
3562 * browser is closed.
3565 * 1 : csp = Current client state (buffers, headers, etc...)
3566 * 2 : header = On input, pointer to header to modify.
3567 * On output, pointer to the modified header, or NULL
3568 * to remove the header. This function frees the
3569 * original string if necessary.
3571 * Returns : JB_ERR_OK on success, or
3572 * JB_ERR_MEMORY on out-of-memory error.
3574 *********************************************************************/
3575 static jb_err server_set_cookie(struct client_state *csp, char **header)
3582 if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
3584 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3587 else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
3589 /* Flag whether or not to log a message */
3592 /* A variable to store the tag we're working on */
3595 /* Skip "Set-Cookie:" (11 characters) in header */
3596 cur_tag = *header + 11;
3598 /* skip whitespace between "Set-Cookie:" and value */
3599 while (*cur_tag && ijb_isspace(*cur_tag))
3604 /* Loop through each tag in the cookie */
3608 char *next_tag = strchr(cur_tag, ';');
3609 if (next_tag != NULL)
3611 /* Skip the ';' character itself */
3614 /* skip whitespace ";" and start of tag */
3615 while (*next_tag && ijb_isspace(*next_tag))
3622 /* "Next tag" is the end of the string */
3623 next_tag = cur_tag + strlen(cur_tag);
3627 * Check the expiration date to see
3628 * if the cookie is still valid, if yes,
3629 * rewrite it to a session cookie.
3631 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
3633 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
3635 /* Did we detect the date properly? */
3636 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
3639 * Nope, treat it as if it was still valid.
3641 * XXX: Should we remove the whole cookie instead?
3643 log_error(LOG_LEVEL_ERROR,
3644 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
3645 string_move(cur_tag, next_tag);
3651 * Yes. Check if the cookie is still valid.
3653 * If the cookie is already expired it's probably
3654 * a delete cookie and even if it isn't, the browser
3655 * will discard it anyway.
3659 * XXX: timegm() isn't available on some AmigaOS
3660 * versions and our replacement doesn't work.
3662 * Our options are to either:
3664 * - disable session-cookies-only completely if timegm
3667 * - to simply remove all expired tags, like it has
3668 * been done until Privoxy 3.0.6 and to live with
3669 * the consequence that it can cause login/logout
3670 * problems on servers that don't validate their
3671 * input properly, or
3673 * - to replace it with mktime in which
3674 * case there is a slight chance of valid cookies
3675 * passing as already expired.
3677 * This is the way it's currently done and it's not
3678 * as bad as it sounds. If the missing GMT offset is
3679 * enough to change the result of the expiration check
3680 * the cookie will be only valid for a few hours
3681 * anyway, which in many cases will be shorter
3682 * than a browser session.
3684 if (cookie_time - now < 0)
3686 log_error(LOG_LEVEL_HEADER,
3687 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
3688 /* Just in case some clown sets more then one expiration date */
3694 * Still valid, delete expiration date by copying
3695 * the rest of the string over it.
3697 string_move(cur_tag, next_tag);
3699 /* That changed the header, need to issue a log message */
3703 * Note that the next tag has now been moved to *cur_tag,
3704 * so we do not need to update the cur_tag pointer.
3712 /* Move on to next cookie tag */
3719 assert(NULL != *header);
3720 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
3729 #ifdef FEATURE_FORCE_LOAD
3730 /*********************************************************************
3732 * Function : strclean
3734 * Description : In-Situ-Eliminate all occurances of substring in
3738 * 1 : string = string to clean
3739 * 2 : substring = substring to eliminate
3741 * Returns : Number of eliminations
3743 *********************************************************************/
3744 int strclean(char *string, const char *substring)
3750 len = strlen(substring);
3752 while((pos = strstr(string, substring)) != NULL)
3759 while (*p++ != '\0');
3766 #endif /* def FEATURE_FORCE_LOAD */
3769 /*********************************************************************
3771 * Function : parse_header_time
3773 * Description : Parses time formats used in HTTP header strings
3774 * to get the numerical respresentation.
3777 * 1 : header_time = HTTP header time as string.
3778 * 2 : result = storage for header_time in seconds
3780 * Returns : JB_ERR_OK if the time format was recognized, or
3781 * JB_ERR_PARSE otherwise.
3783 *********************************************************************/
3784 static jb_err parse_header_time(const char *header_time, time_t *result)
3789 * Zero out gmt to prevent time zone offsets.
3791 * While this is only necessary on some platforms
3792 * (mingw32 for example), I don't know how to
3793 * detect these automatically and doing it everywhere
3796 memset(&gmt, 0, sizeof(gmt));
3798 /* Tue, 02 Jun 2037 20:00:00 */
3799 if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
3800 /* Tue, 02-Jun-2037 20:00:00 */
3801 && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
3802 /* Tue, 02-Jun-37 20:00:00 */
3803 && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
3804 /* Tuesday, 02-Jun-2037 20:00:00 */
3805 && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
3806 /* Tuesday Jun 02 20:00:00 2037 */
3807 && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
3809 return JB_ERR_PARSE;
3812 *result = timegm(&gmt);
3819 /*********************************************************************
3821 * Function : get_destination_from_headers
3823 * Description : Parse the "Host:" header to get the request's destination.
3824 * Only needed if the client's request was forcefully
3825 * redirected into Privoxy.
3827 * Code mainly copied from client_host() which is currently
3828 * run too late for this purpose.
3831 * 1 : headers = List of headers (one of them hopefully being
3832 * the "Host:" header)
3833 * 2 : http = storage for the result (host, port and hostport).
3835 * Returns : JB_ERR_MEMORY in case of memory problems,
3836 * JB_ERR_PARSE if the host header couldn't be found,
3837 * JB_ERR_OK otherwise.
3839 *********************************************************************/
3840 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
3846 host = get_header_value(headers, "Host:");
3850 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
3851 return JB_ERR_PARSE;
3854 if (NULL == (p = strdup((host))))
3856 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
3857 return JB_ERR_MEMORY;
3860 if (NULL == (q = strdup(p)))
3863 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
3864 return JB_ERR_MEMORY;
3867 freez(http->hostport);
3871 q = strchr(http->host, ':');
3874 /* Terminate hostname and evaluate port string */
3876 http->port = atoi(q);
3880 http->port = http->ssl ? 443 : 80;
3883 /* Rebuild request URL */
3885 http->url = strdup(http->ssl ? "https://" : "http://");
3886 string_append(&http->url, http->hostport);
3887 string_append(&http->url, http->path);
3888 if (http->url == NULL)
3890 return JB_ERR_MEMORY;
3893 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
3901 /*********************************************************************
3903 * Function : create_forged_referrer
3905 * Description : Helper for client_referrer to forge a referer as
3906 * 'http://[hostname:port/' to fool stupid
3907 * checks for in-site links
3910 * 1 : header = Pointer to header pointer
3911 * 2 : hostport = Host and optionally port as string
3913 * Returns : JB_ERR_OK in case of success, or
3914 * JB_ERR_MEMORY in case of memory problems.
3916 *********************************************************************/
3917 static jb_err create_forged_referrer(char **header, const char *hostport)
3919 assert(NULL == *header);
3921 *header = strdup("Referer: http://");
3922 string_append(header, hostport);
3923 string_append(header, "/");
3925 if (NULL == *header)
3927 return JB_ERR_MEMORY;
3930 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
3937 /*********************************************************************
3939 * Function : create_fake_referrer
3941 * Description : Helper for client_referrer to create a fake referrer
3942 * based on a string supplied by the user.
3945 * 1 : header = Pointer to header pointer
3946 * 2 : hosthost = Referrer to fake
3948 * Returns : JB_ERR_OK in case of success, or
3949 * JB_ERR_MEMORY in case of memory problems.
3951 *********************************************************************/
3952 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
3954 assert(NULL == *header);
3956 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
3958 log_error(LOG_LEVEL_HEADER,
3959 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
3961 *header = strdup("Referer: ");
3962 string_append(header, fake_referrer);
3964 if (NULL == *header)
3966 return JB_ERR_MEMORY;
3969 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
3976 /*********************************************************************
3978 * Function : handle_conditional_hide_referrer_parameter
3980 * Description : Helper for client_referrer to crunch or forge
3981 * the referrer header if the host has changed.
3984 * 1 : header = Pointer to header pointer
3985 * 2 : host = The target host (may include the port)
3986 * 3 : parameter_conditional_block = Boolean to signal
3987 * if we're in conditional-block mode. If not set,
3988 * we're in conditional-forge mode.
3990 * Returns : JB_ERR_OK in case of success, or
3991 * JB_ERR_MEMORY in case of memory problems.
3993 *********************************************************************/
3994 static jb_err handle_conditional_hide_referrer_parameter(char **header,
3995 const char *host, const int parameter_conditional_block)
3997 char *referer = strdup(*header);
3998 const size_t hostlenght = strlen(host);
3999 const char *referer_url = NULL;
4001 if (NULL == referer)
4004 return JB_ERR_MEMORY;
4007 /* referer begins with 'Referer: http[s]://' */
4008 if ((hostlenght+17) < strlen(referer))
4011 * Shorten referer to make sure the referer is blocked
4012 * if www.example.org/www.example.com-shall-see-the-referer/
4013 * links to www.example.com/
4015 referer[hostlenght+17] = '\0';
4017 referer_url = strstr(referer, "http://");
4018 if ((NULL == referer_url) || (NULL == strstr(referer_url, host)))
4020 /* Host has changed, Referer is invalid or a https URL. */
4021 if (parameter_conditional_block)
4023 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4030 return create_forged_referrer(header, host);
4040 /*********************************************************************
4042 * Function : get_appropiate_connection_header
4044 * Description : Returns an appropiate Connection header
4045 * depending on whether or not we try to keep
4046 * the connection to the server alive.
4049 * 1 : csp = Current client state (buffers, headers, etc...)
4051 * Returns : Pointer to statically allocated header buffer.
4053 *********************************************************************/
4054 static const char *get_appropiate_connection_header(const struct client_state *csp)
4056 static const char connection_keep_alive[] = "Connection: keep-alive";
4057 static const char connection_close[] = "Connection: close";
4059 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
4060 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
4061 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
4063 && (csp->http->ssl == 0))
4065 return connection_keep_alive;
4067 return connection_close;
4071 /*********************************************************************
4073 * Function : create_content_length_header
4075 * Description : Creates a Content-Length header.
4078 * 1 : content_length = The content length to be used in the header.
4079 * 2 : header = Allocated space to safe the header.
4080 * 3 : buffer_length = The length of the allocated space.
4084 *********************************************************************/
4085 static void create_content_length_header(unsigned long long content_length,
4086 char *header, size_t buffer_length)
4088 snprintf(header, buffer_length, "Content-Length: %llu", content_length);