1 const char parsers_rcs[] = "$Id: parsers.c,v 1.280 2013/11/24 14:24:17 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
8 * Copyright : Written by and Copyright (C) 2001-2012 the
9 * Privoxy team. http://www.privoxy.org/
11 * Based on the Internet Junkbuster originally written
12 * by and Copyright (C) 1997 Anonymous Coders and
13 * Junkbusters Corporation. http://www.junkbusters.com
15 * This program is free software; you can redistribute it
16 * and/or modify it under the terms of the GNU General
17 * Public License as published by the Free Software
18 * Foundation; either version 2 of the License, or (at
19 * your option) any later version.
21 * This program is distributed in the hope that it will
22 * be useful, but WITHOUT ANY WARRANTY; without even the
23 * implied warranty of MERCHANTABILITY or FITNESS FOR A
24 * PARTICULAR PURPOSE. See the GNU General Public
25 * License for more details.
27 * The GNU General Public License should be included with
28 * this file. If not, you can view it at
29 * http://www.gnu.org/copyleft/gpl.html
30 * or write to the Free Software Foundation, Inc., 59
31 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
33 *********************************************************************/
40 #include <sys/types.h>
50 * Convince GNU's libc to provide a strptime prototype.
59 #define GZIP_IDENTIFIER_1 0x1f
60 #define GZIP_IDENTIFIER_2 0x8b
62 #define GZIP_FLAG_CHECKSUM 0x02
63 #define GZIP_FLAG_EXTRA_FIELDS 0x04
64 #define GZIP_FLAG_FILE_NAME 0x08
65 #define GZIP_FLAG_COMMENT 0x10
66 #define GZIP_FLAG_RESERVED_BITS 0xe0
69 #if !defined(_WIN32) && !defined(__OS2__)
75 #ifdef FEATURE_PTHREAD
77 /* jcc.h is for mutex semapores only */
78 #endif /* def FEATURE_PTHREAD */
83 #include "jbsockets.h"
93 const char parsers_h_rcs[] = PARSERS_H_VERSION;
95 static char *get_header_line(struct iob *iob);
96 static jb_err scan_headers(struct client_state *csp);
97 static jb_err header_tagger(struct client_state *csp, char *header);
98 static jb_err parse_header_time(const char *header_time, time_t *result);
100 static jb_err crumble (struct client_state *csp, char **header);
101 static jb_err filter_header (struct client_state *csp, char **header);
102 static jb_err client_connection (struct client_state *csp, char **header);
103 static jb_err client_referrer (struct client_state *csp, char **header);
104 static jb_err client_uagent (struct client_state *csp, char **header);
105 static jb_err client_ua (struct client_state *csp, char **header);
106 static jb_err client_from (struct client_state *csp, char **header);
107 static jb_err client_send_cookie (struct client_state *csp, char **header);
108 static jb_err client_x_forwarded (struct client_state *csp, char **header);
109 static jb_err client_accept_encoding (struct client_state *csp, char **header);
110 static jb_err client_te (struct client_state *csp, char **header);
111 static jb_err client_max_forwards (struct client_state *csp, char **header);
112 static jb_err client_host (struct client_state *csp, char **header);
113 static jb_err client_if_modified_since (struct client_state *csp, char **header);
114 static jb_err client_accept_language (struct client_state *csp, char **header);
115 static jb_err client_if_none_match (struct client_state *csp, char **header);
116 static jb_err crunch_client_header (struct client_state *csp, char **header);
117 static jb_err client_x_filter (struct client_state *csp, char **header);
118 static jb_err client_range (struct client_state *csp, char **header);
119 static jb_err server_set_cookie (struct client_state *csp, char **header);
120 static jb_err server_connection (struct client_state *csp, char **header);
121 static jb_err server_content_type (struct client_state *csp, char **header);
122 static jb_err server_adjust_content_length(struct client_state *csp, char **header);
123 static jb_err server_content_md5 (struct client_state *csp, char **header);
124 static jb_err server_content_encoding (struct client_state *csp, char **header);
125 static jb_err server_transfer_coding (struct client_state *csp, char **header);
126 static jb_err server_http (struct client_state *csp, char **header);
127 static jb_err crunch_server_header (struct client_state *csp, char **header);
128 static jb_err server_last_modified (struct client_state *csp, char **header);
129 static jb_err server_content_disposition(struct client_state *csp, char **header);
131 static jb_err server_adjust_content_encoding(struct client_state *csp, char **header);
134 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
135 static jb_err server_save_content_length(struct client_state *csp, char **header);
136 static jb_err server_keep_alive(struct client_state *csp, char **header);
137 static jb_err server_proxy_connection(struct client_state *csp, char **header);
138 static jb_err client_keep_alive(struct client_state *csp, char **header);
139 static jb_err client_save_content_length(struct client_state *csp, char **header);
140 static jb_err client_proxy_connection(struct client_state *csp, char **header);
141 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
143 static jb_err client_host_adder (struct client_state *csp);
144 static jb_err client_xtra_adder (struct client_state *csp);
145 static jb_err client_x_forwarded_for_adder(struct client_state *csp);
146 static jb_err client_connection_header_adder(struct client_state *csp);
147 static jb_err server_connection_adder(struct client_state *csp);
148 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
149 static jb_err server_proxy_connection_adder(struct client_state *csp);
150 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
151 static jb_err proxy_authentication(struct client_state *csp, char **header);
153 static jb_err create_forged_referrer(char **header, const char *hostport);
154 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
155 static jb_err handle_conditional_hide_referrer_parameter(char **header,
156 const char *host, const int parameter_conditional_block);
157 static void create_content_length_header(unsigned long long content_length,
158 char *header, size_t buffer_length);
161 * List of functions to run on a list of headers.
165 /** The header prefix to match */
168 /** The length of the prefix to match */
171 /** The function to apply to this line */
172 const parser_func_ptr parser;
175 static const struct parsers client_patterns[] = {
176 { "referer:", 8, client_referrer },
177 { "user-agent:", 11, client_uagent },
178 { "ua-", 3, client_ua },
179 { "from:", 5, client_from },
180 { "cookie:", 7, client_send_cookie },
181 { "x-forwarded-for:", 16, client_x_forwarded },
182 { "Accept-Encoding:", 16, client_accept_encoding },
183 { "TE:", 3, client_te },
184 { "Host:", 5, client_host },
185 { "if-modified-since:", 18, client_if_modified_since },
186 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
187 { "Keep-Alive:", 11, client_keep_alive },
188 { "Content-Length:", 15, client_save_content_length },
189 { "Proxy-Connection:", 17, client_proxy_connection },
191 { "Keep-Alive:", 11, crumble },
192 { "Proxy-Connection:", 17, crumble },
194 { "connection:", 11, client_connection },
195 { "max-forwards:", 13, client_max_forwards },
196 { "Accept-Language:", 16, client_accept_language },
197 { "if-none-match:", 14, client_if_none_match },
198 { "Range:", 6, client_range },
199 { "Request-Range:", 14, client_range },
200 { "If-Range:", 9, client_range },
201 { "X-Filter:", 9, client_x_filter },
202 { "Proxy-Authorization:", 20, proxy_authentication },
204 { "Transfer-Encoding:", 18, client_transfer_encoding },
206 { "*", 0, crunch_client_header },
207 { "*", 0, filter_header },
211 static const struct parsers server_patterns[] = {
212 { "HTTP/", 5, server_http },
213 { "set-cookie:", 11, server_set_cookie },
214 { "connection:", 11, server_connection },
215 { "Content-Type:", 13, server_content_type },
216 { "Content-MD5:", 12, server_content_md5 },
217 { "Content-Encoding:", 17, server_content_encoding },
218 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
219 { "Content-Length:", 15, server_save_content_length },
220 { "Keep-Alive:", 11, server_keep_alive },
221 { "Proxy-Connection:", 17, server_proxy_connection },
223 { "Keep-Alive:", 11, crumble },
224 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
225 { "Transfer-Encoding:", 18, server_transfer_coding },
226 { "content-disposition:", 20, server_content_disposition },
227 { "Last-Modified:", 14, server_last_modified },
228 { "Proxy-Authenticate:", 19, proxy_authentication },
229 { "*", 0, crunch_server_header },
230 { "*", 0, filter_header },
234 static const add_header_func_ptr add_client_headers[] = {
236 client_x_forwarded_for_adder,
238 client_connection_header_adder,
242 static const add_header_func_ptr add_server_headers[] = {
243 server_connection_adder,
244 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
245 server_proxy_connection_adder,
246 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
250 /*********************************************************************
252 * Function : flush_socket
254 * Description : Write any pending "buffered" content.
257 * 1 : fd = file descriptor of the socket to read
258 * 2 : iob = The I/O buffer to flush, usually csp->iob.
260 * Returns : On success, the number of bytes written are returned (zero
261 * indicates nothing was written). On error, -1 is returned,
262 * and errno is set appropriately. If count is zero and the
263 * file descriptor refers to a regular file, 0 will be
264 * returned without causing any other effect. For a special
265 * file, the results are not portable.
267 *********************************************************************/
268 long flush_socket(jb_socket fd, struct iob *iob)
270 long len = iob->eod - iob->cur;
277 if (write_socket(fd, iob->cur, (size_t)len))
281 iob->eod = iob->cur = iob->buf;
287 /*********************************************************************
289 * Function : add_to_iob
291 * Description : Add content to the buffer, expanding the
292 * buffer if necessary.
295 * 1 : iob = Destination buffer.
296 * 2 : buffer_limit = Limit to which the destination may grow
297 * 3 : src = holds the content to be added
298 * 4 : n = number of bytes to be added
300 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
301 * or buffer limit reached.
303 *********************************************************************/
304 jb_err add_to_iob(struct iob *iob, const size_t buffer_limit, char *src, long n)
306 size_t used, offset, need;
309 if (n <= 0) return JB_ERR_OK;
311 used = (size_t)(iob->eod - iob->buf);
312 offset = (size_t)(iob->cur - iob->buf);
313 need = used + (size_t)n + 1;
316 * If the buffer can't hold the new data, extend it first.
317 * Use the next power of two if possible, else use the actual need.
319 if (need > buffer_limit)
321 log_error(LOG_LEVEL_INFO,
322 "Buffer limit reached while extending the buffer (iob). Needed: %d. Limit: %d",
324 return JB_ERR_MEMORY;
327 if (need > iob->size)
329 size_t want = iob->size ? iob->size : 512;
336 if (want <= buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
340 else if (NULL != (p = (char *)realloc(iob->buf, need)))
346 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
347 return JB_ERR_MEMORY;
350 /* Update the iob pointers */
351 iob->cur = p + offset;
356 /* copy the new data into the iob buffer */
357 memcpy(iob->eod, src, (size_t)n);
359 /* point to the end of the data */
362 /* null terminate == cheap insurance */
370 /*********************************************************************
372 * Function : clear_iob
374 * Description : Frees the memory allocated for an I/O buffer and
375 * resets the structure.
378 * 1 : iob = I/O buffer to clear.
380 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
381 * or buffer limit reached.
383 *********************************************************************/
384 void clear_iob(struct iob *iob)
387 memset(iob, '\0', sizeof(*iob));;
392 /*********************************************************************
394 * Function : decompress_iob
396 * Description : Decompress buffered page, expanding the
397 * buffer as necessary. csp->iob->cur
398 * should point to the the beginning of the
399 * compressed data block.
402 * 1 : csp = Current client state (buffers, headers, etc...)
404 * Returns : JB_ERR_OK on success,
405 * JB_ERR_MEMORY if out-of-memory limit reached, and
406 * JB_ERR_COMPRESS if error decompressing buffer.
408 *********************************************************************/
409 jb_err decompress_iob(struct client_state *csp)
411 char *buf; /* new, uncompressed buffer */
412 char *cur; /* Current iob position (to keep the original
413 * iob->cur unmodified if we return early) */
414 size_t bufsize; /* allocated size of the new buffer */
415 size_t old_size; /* Content size before decompression */
416 size_t skip_size; /* Number of bytes at the beginning of the iob
417 that we should NOT decompress. */
418 int status; /* return status of the inflate() call */
419 z_stream zstr; /* used by calls to zlib */
421 assert(csp->iob->cur - csp->iob->buf > 0);
422 assert(csp->iob->eod - csp->iob->cur > 0);
424 bufsize = csp->iob->size;
425 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
426 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
430 if (bufsize < (size_t)10)
433 * This is to protect the parsing of gzipped data,
434 * but it should(?) be valid for deflated data also.
436 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
437 return JB_ERR_COMPRESS;
440 if (csp->content_type & CT_GZIP)
443 * Our task is slightly complicated by the facts that data
444 * compressed by gzip does not include a zlib header, and
445 * that there is no easily accessible interface in zlib to
446 * handle a gzip header. We strip off the gzip header by
447 * hand, and later inform zlib not to expect a header.
451 * Strip off the gzip header. Please see RFC 1952 for more
452 * explanation of the appropriate fields.
454 if (((*cur++ & 0xff) != GZIP_IDENTIFIER_1)
455 || ((*cur++ & 0xff) != GZIP_IDENTIFIER_2)
456 || (*cur++ != Z_DEFLATED))
458 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
459 return JB_ERR_COMPRESS;
464 if (flags & GZIP_FLAG_RESERVED_BITS)
466 /* The gzip header has reserved bits set; bail out. */
467 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
468 return JB_ERR_COMPRESS;
472 * Skip mtime (4 bytes), extra flags (1 byte)
473 * and OS type (1 byte).
477 /* Skip extra fields if necessary. */
478 if (flags & GZIP_FLAG_EXTRA_FIELDS)
481 * Skip a given number of bytes, specified
482 * as a 16-bit little-endian value.
484 * XXX: this code is untested and should probably be removed.
488 skip_bytes += *cur++ << 8;
491 * The number of bytes to skip should be positive
492 * and we'd like to stay in the buffer.
494 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
496 log_error(LOG_LEVEL_ERROR,
497 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
499 return JB_ERR_COMPRESS;
501 log_error(LOG_LEVEL_INFO,
502 "Skipping %d bytes for gzip compression. Does this sound right?",
507 /* Skip the filename if necessary. */
508 if (flags & GZIP_FLAG_FILE_NAME)
510 /* A null-terminated string is supposed to follow. */
511 while (*cur++ && (cur < csp->iob->eod));
514 /* Skip the comment if necessary. */
515 if (flags & GZIP_FLAG_COMMENT)
517 /* A null-terminated string is supposed to follow. */
518 while (*cur++ && (cur < csp->iob->eod));
521 /* Skip the CRC if necessary. */
522 if (flags & GZIP_FLAG_CHECKSUM)
527 if (cur >= csp->iob->eod)
530 * If the current position pointer reached or passed
531 * the buffer end, we were obviously tricked to skip
534 log_error(LOG_LEVEL_ERROR,
535 "Malformed gzip header detected. Aborting decompression.");
536 return JB_ERR_COMPRESS;
540 else if (csp->content_type & CT_DEFLATE)
543 * In theory (that is, according to RFC 1950), deflate-compressed
544 * data should begin with a two-byte zlib header and have an
545 * adler32 checksum at the end. It seems that in practice only
546 * the raw compressed data is sent. Note that this means that
547 * we are not RFC 1950-compliant here, but the advantage is that
548 * this actually works. :)
550 * We add a dummy null byte to tell zlib where the data ends,
551 * and later inform it not to expect a header.
553 * Fortunately, add_to_iob() has thoughtfully null-terminated
554 * the buffer; we can just increment the end pointer to include
561 log_error(LOG_LEVEL_ERROR,
562 "Unable to determine compression format for decompression");
563 return JB_ERR_COMPRESS;
566 /* Set up the fields required by zlib. */
567 zstr.next_in = (Bytef *)cur;
568 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
569 zstr.zalloc = Z_NULL;
571 zstr.opaque = Z_NULL;
574 * Passing -MAX_WBITS to inflateInit2 tells the library
575 * that there is no zlib header.
577 if (inflateInit2(&zstr, -MAX_WBITS) != Z_OK)
579 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
580 return JB_ERR_COMPRESS;
584 * Next, we allocate new storage for the inflated data.
585 * We don't modify the existing iob yet, so in case there
586 * is error in decompression we can recover gracefully.
588 buf = zalloc(bufsize);
591 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
592 return JB_ERR_MEMORY;
595 assert(bufsize >= skip_size);
596 memcpy(buf, csp->iob->buf, skip_size);
597 zstr.avail_out = (uInt)(bufsize - skip_size);
598 zstr.next_out = (Bytef *)buf + skip_size;
600 /* Try to decompress the whole stream in one shot. */
601 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
603 /* We need to allocate more memory for the output buffer. */
605 char *tmpbuf; /* used for realloc'ing the buffer */
606 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
608 if (0 == zstr.avail_in)
611 * If zlib wants more data then there's a problem, because
612 * the complete compressed file should have been buffered.
614 log_error(LOG_LEVEL_ERROR,
615 "Unexpected end of compressed iob. Using what we got so far.");
620 * If we reached the buffer limit and still didn't have enough
621 * memory, just give up. Due to the ceiling enforced by the next
622 * if block we could actually check for equality here, but as it
623 * can be easily mistaken for a bug we don't.
625 if (bufsize >= csp->config->buffer_limit)
627 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
628 return JB_ERR_MEMORY;
631 /* Try doubling the buffer size each time. */
634 /* Don't exceed the buffer limit. */
635 if (bufsize > csp->config->buffer_limit)
637 bufsize = csp->config->buffer_limit;
640 /* Try to allocate the new buffer. */
641 tmpbuf = realloc(buf, bufsize);
644 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
646 return JB_ERR_MEMORY;
650 char *oldnext_out = (char *)zstr.next_out;
653 * Update the fields for inflate() to use the new
654 * buffer, which may be in a location different from
657 zstr.avail_out += (uInt)(bufsize - oldbufsize);
658 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
661 * Compare with an uglier method of calculating these values
662 * that doesn't require the extra oldbufsize variable.
664 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
665 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
671 if (Z_STREAM_ERROR == inflateEnd(&zstr))
673 log_error(LOG_LEVEL_ERROR,
674 "Inconsistent stream state after decompression: %s", zstr.msg);
676 * XXX: Intentionally no return.
678 * According to zlib.h, Z_STREAM_ERROR is returned
679 * "if the stream state was inconsistent".
681 * I assume in this case inflate()'s status
682 * would also be something different than Z_STREAM_END
683 * so this check should be redundant, but lets see.
687 if ((status != Z_STREAM_END) && (0 != zstr.avail_in))
690 * We failed to decompress the stream and it's
691 * not simply because of missing data.
693 log_error(LOG_LEVEL_ERROR,
694 "Unexpected error while decompressing to the buffer (iob): %s",
696 return JB_ERR_COMPRESS;
700 * Finally, we can actually update the iob, since the
701 * decompression was successful. First, free the old
704 freez(csp->iob->buf);
706 /* Now, update the iob to use the new buffer. */
708 csp->iob->cur = csp->iob->buf + skip_size;
709 csp->iob->eod = (char *)zstr.next_out;
710 csp->iob->size = bufsize;
713 * Make sure the new uncompressed iob obeys some minimal
714 * consistency conditions.
716 if ((csp->iob->buf < csp->iob->cur)
717 && (csp->iob->cur <= csp->iob->eod)
718 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
720 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
721 if (new_size > (size_t)0)
723 log_error(LOG_LEVEL_RE_FILTER,
724 "Decompression successful. Old size: %d, new size: %d.",
729 /* zlib thinks this is OK, so lets do the same. */
730 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
735 /* It seems that zlib did something weird. */
736 log_error(LOG_LEVEL_ERROR,
737 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
738 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
739 csp->iob->eod, csp->iob->buf + csp->iob->size);
740 return JB_ERR_COMPRESS;
746 #endif /* defined(FEATURE_ZLIB) */
749 /*********************************************************************
751 * Function : normalize_lws
753 * Description : Reduces unquoted linear white space in headers
754 * to a single space in accordance with RFC 2616 2.2.
755 * This simplifies parsing and filtering later on.
757 * XXX: Remove log messages before
758 * the next stable release?
761 * 1 : header = A header with linear white space to reduce.
765 *********************************************************************/
766 static void normalize_lws(char *header)
772 if (privoxy_isspace(*p) && privoxy_isspace(*(p+1)))
776 while (privoxy_isspace(*q))
780 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
786 log_error(LOG_LEVEL_HEADER,
787 "Converting tab to space in '%s'", header);
792 char *end_of_token = strstr(p+1, "\"");
794 if (NULL != end_of_token)
796 /* Don't mess with quoted text. */
801 log_error(LOG_LEVEL_HEADER,
802 "Ignoring single quote in '%s'", header);
808 p = strchr(header, ':');
809 if ((p != NULL) && (p != header) && privoxy_isspace(*(p-1)))
812 * There's still space before the colon.
820 /*********************************************************************
822 * Function : get_header
824 * Description : This (odd) routine will parse the csp->iob
825 * to get the next complete header.
828 * 1 : iob = The I/O buffer to parse, usually csp->iob.
830 * Returns : Any one of the following:
832 * 1) a pointer to a dynamically allocated string that contains a header line
833 * 2) NULL indicating that the end of the header was reached
834 * 3) "" indicating that the end of the iob was reached before finding
835 * a complete header line.
837 *********************************************************************/
838 char *get_header(struct iob *iob)
842 header = get_header_line(iob);
844 if ((header == NULL) || (*header == '\0'))
847 * No complete header read yet, tell the client.
852 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
855 * Header spans multiple lines, append the next one.
857 char *continued_header;
859 continued_header = get_header_line(iob);
860 if ((continued_header == NULL) || (*continued_header == '\0'))
863 * No complete header read yet, return what we got.
864 * XXX: Should "unread" header instead.
866 log_error(LOG_LEVEL_INFO,
867 "Failed to read a multi-line header properly: '%s'",
872 if (JB_ERR_OK != string_join(&header, continued_header))
874 log_error(LOG_LEVEL_FATAL,
875 "Out of memory while appending multiple headers.");
879 /* XXX: remove before next stable release. */
880 log_error(LOG_LEVEL_HEADER,
881 "Merged multiple header lines to: '%s'",
886 normalize_lws(header);
893 /*********************************************************************
895 * Function : get_header_line
897 * Description : This (odd) routine will parse the csp->iob
898 * to get the next header line.
901 * 1 : iob = The I/O buffer to parse, usually csp->iob.
903 * Returns : Any one of the following:
905 * 1) a pointer to a dynamically allocated string that contains a header line
906 * 2) NULL indicating that the end of the header was reached
907 * 3) "" indicating that the end of the iob was reached before finding
908 * a complete header line.
910 *********************************************************************/
911 static char *get_header_line(struct iob *iob)
915 if ((iob->cur == NULL)
916 || ((p = strchr(iob->cur, '\n')) == NULL))
918 return(""); /* couldn't find a complete header */
923 ret = strdup(iob->cur);
926 /* FIXME No way to handle error properly */
927 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
933 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
935 /* is this a blank line (i.e. the end of the header) ? */
947 /*********************************************************************
949 * Function : get_header_value
951 * Description : Get the value of a given header from a chained list
952 * of header lines or return NULL if no such header is
953 * present in the list.
956 * 1 : header_list = pointer to list
957 * 2 : header_name = string with name of header to look for.
958 * Trailing colon required, capitalization
961 * Returns : NULL if not found, else value of header
963 *********************************************************************/
964 char *get_header_value(const struct list *header_list, const char *header_name)
966 struct list_entry *cur_entry;
972 length = strlen(header_name);
974 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
978 if (!strncmpic(cur_entry->str, header_name, length))
981 * Found: return pointer to start of value
983 ret = cur_entry->str + length;
984 while (*ret && privoxy_isspace(*ret)) ret++;
998 /*********************************************************************
1000 * Function : scan_headers
1002 * Description : Scans headers, applies tags and updates action bits.
1005 * 1 : csp = Current client state (buffers, headers, etc...)
1007 * Returns : JB_ERR_OK
1009 *********************************************************************/
1010 static jb_err scan_headers(struct client_state *csp)
1012 struct list_entry *h; /* Header */
1013 jb_err err = JB_ERR_OK;
1015 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1017 /* Header crunch()ed in previous run? -> ignore */
1018 if (h->str == NULL) continue;
1019 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1020 err = header_tagger(csp, h->str);
1027 /*********************************************************************
1029 * Function : enforce_header_order
1031 * Description : Enforces a given header order.
1034 * 1 : headers = List of headers to order.
1035 * 2 : ordered_headers = List of ordered header names.
1039 *********************************************************************/
1040 static void enforce_header_order(struct list *headers, const struct list *ordered_headers)
1042 struct list_entry *sorted_header;
1043 struct list new_headers[1];
1044 struct list_entry *header;
1046 init_list(new_headers);
1048 /* The request line is always the first "header" */
1050 assert(NULL != headers->first->str);
1051 enlist(new_headers, headers->first->str);
1052 freez(headers->first->str)
1054 /* Enlist the specified headers in the given order */
1056 for (sorted_header = ordered_headers->first; sorted_header != NULL;
1057 sorted_header = sorted_header->next)
1059 const size_t sorted_header_length = strlen(sorted_header->str);
1060 for (header = headers->first; header != NULL; header = header->next)
1062 /* Header enlisted in previous run? -> ignore */
1063 if (header->str == NULL) continue;
1065 if (0 == strncmpic(sorted_header->str, header->str, sorted_header_length)
1066 && (header->str[sorted_header_length] == ':'))
1068 log_error(LOG_LEVEL_HEADER, "Enlisting sorted header %s", header->str);
1069 if (JB_ERR_OK != enlist(new_headers, header->str))
1071 log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
1078 /* Enlist the rest of the headers behind the ordered ones */
1079 for (header = headers->first; header != NULL; header = header->next)
1081 /* Header enlisted in previous run? -> ignore */
1082 if (header->str == NULL) continue;
1084 log_error(LOG_LEVEL_HEADER,
1085 "Enlisting left-over header %s", header->str);
1086 if (JB_ERR_OK != enlist(new_headers, header->str))
1088 log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
1093 list_remove_all(headers);
1094 list_duplicate(headers, new_headers);
1095 list_remove_all(new_headers);
1101 /*********************************************************************
1105 * Description : add, delete or modify lines in the HTTP header streams.
1106 * On entry, it receives a linked list of headers space
1107 * that was allocated dynamically (both the list nodes
1108 * and the header contents).
1110 * As a side effect it frees the space used by the original
1114 * 1 : csp = Current client state (buffers, headers, etc...)
1115 * 2 : filter_server_headers = Boolean to switch between
1116 * server and header filtering.
1118 * Returns : JB_ERR_OK in case off success, or
1119 * JB_ERR_MEMORY on out-of-memory error.
1121 *********************************************************************/
1122 jb_err sed(struct client_state *csp, int filter_server_headers)
1124 /* XXX: use more descriptive names. */
1125 struct list_entry *p;
1126 const struct parsers *v;
1127 const add_header_func_ptr *f;
1128 jb_err err = JB_ERR_OK;
1132 if (filter_server_headers)
1134 v = server_patterns;
1135 f = add_server_headers;
1136 check_negative_tag_patterns(csp, PATTERN_SPEC_NO_RESPONSE_TAG_PATTERN);
1140 v = client_patterns;
1141 f = add_client_headers;
1142 check_negative_tag_patterns(csp, PATTERN_SPEC_NO_REQUEST_TAG_PATTERN);
1145 while ((err == JB_ERR_OK) && (v->str != NULL))
1147 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1149 /* Header crunch()ed in previous run? -> ignore */
1150 if (p->str == NULL) continue;
1152 /* Does the current parser handle this header? */
1153 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1154 (v->len == CHECK_EVERY_HEADER_REMAINING))
1156 err = v->parser(csp, &(p->str));
1162 /* place additional headers on the csp->headers list */
1163 while ((err == JB_ERR_OK) && (*f))
1169 if (!filter_server_headers && !list_is_empty(csp->config->ordered_client_headers))
1171 enforce_header_order(csp->headers, csp->config->ordered_client_headers);
1178 /*********************************************************************
1180 * Function : update_server_headers
1182 * Description : Updates server headers after the body has been modified.
1185 * 1 : csp = Current client state (buffers, headers, etc...)
1187 * Returns : JB_ERR_OK in case off success, or
1188 * JB_ERR_MEMORY on out-of-memory error.
1190 *********************************************************************/
1191 jb_err update_server_headers(struct client_state *csp)
1193 jb_err err = JB_ERR_OK;
1195 static const struct parsers server_patterns_light[] = {
1196 { "Content-Length:", 15, server_adjust_content_length },
1197 { "Transfer-Encoding:", 18, server_transfer_coding },
1199 { "Content-Encoding:", 17, server_adjust_content_encoding },
1200 #endif /* def FEATURE_ZLIB */
1204 if (strncmpic(csp->http->cmd, "HEAD", 4))
1206 const struct parsers *v;
1207 struct list_entry *p;
1209 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1211 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1213 /* Header crunch()ed in previous run? -> ignore */
1214 if (p->str == NULL) continue;
1216 /* Does the current parser handle this header? */
1217 if (strncmpic(p->str, v->str, v->len) == 0)
1219 err = v->parser(csp, (char **)&(p->str));
1225 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1226 if ((JB_ERR_OK == err)
1227 && (csp->flags & CSP_FLAG_MODIFIED)
1228 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
1229 && !(csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET))
1233 create_content_length_header(csp->content_length, header, sizeof(header));
1234 err = enlist(csp->headers, header);
1235 if (JB_ERR_OK == err)
1237 log_error(LOG_LEVEL_HEADER,
1238 "Content modified with no Content-Length header set. "
1239 "Created: %s.", header);
1240 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
1243 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1245 #ifdef FEATURE_COMPRESSION
1246 if ((JB_ERR_OK == err)
1247 && (csp->flags & CSP_FLAG_BUFFERED_CONTENT_DEFLATED))
1249 err = enlist_unique_header(csp->headers, "Content-Encoding", "deflate");
1250 if (JB_ERR_OK == err)
1252 log_error(LOG_LEVEL_HEADER, "Added header: Content-Encoding: deflate");
1261 /*********************************************************************
1263 * Function : header_tagger
1265 * Description : Executes all text substitutions from applying
1266 * tag actions and saves the result as tag.
1268 * XXX: Shares enough code with filter_header() and
1269 * pcrs_filter_response() to warrant some helper functions.
1272 * 1 : csp = Current client state (buffers, headers, etc...)
1273 * 2 : header = Header that is used as tagger input
1275 * Returns : JB_ERR_OK on success and always succeeds
1277 *********************************************************************/
1278 static jb_err header_tagger(struct client_state *csp, char *header)
1280 int wanted_filter_type;
1281 int multi_action_index;
1284 struct re_filterfile_spec *b;
1285 struct list_entry *tag_name;
1287 const size_t header_length = strlen(header);
1289 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1291 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1292 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1296 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1297 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1300 if (list_is_empty(csp->action->multi[multi_action_index])
1301 || filters_available(csp) == FALSE)
1303 /* Return early if no taggers apply or if none are available. */
1307 /* Execute all applying taggers */
1308 for (tag_name = csp->action->multi[multi_action_index]->first;
1309 NULL != tag_name; tag_name = tag_name->next)
1311 b = get_filter(csp, tag_name->str, wanted_filter_type);
1314 char *modified_tag = NULL;
1316 size_t size = header_length;
1317 pcrs_job *joblist = b->joblist;
1319 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1321 if (NULL == joblist)
1323 log_error(LOG_LEVEL_RE_FILTER,
1324 "Tagger %s has empty joblist. Nothing to do.", b->name);
1328 /* execute their pcrs_joblist on the header. */
1329 for (job = joblist; NULL != job; job = job->next)
1331 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1335 /* Success, continue with the modified version. */
1344 /* Tagger doesn't match */
1347 /* Regex failure, log it but continue anyway. */
1348 assert(NULL != header);
1349 log_error(LOG_LEVEL_ERROR,
1350 "Problems with tagger \'%s\' and header \'%s\': %s",
1351 b->name, *header, pcrs_strerror(hits));
1353 freez(modified_tag);
1357 if (b->dynamic) pcrs_free_joblist(joblist);
1359 /* If this tagger matched */
1365 * There is no technical limitation which makes
1366 * it impossible to use empty tags, but I assume
1367 * no one would do it intentionally.
1370 log_error(LOG_LEVEL_INFO,
1371 "Tagger \'%s\' created an empty tag. Ignored.",
1376 if (!list_contains_item(csp->tags, tag))
1378 if (JB_ERR_OK != enlist(csp->tags, tag))
1380 log_error(LOG_LEVEL_ERROR,
1381 "Insufficient memory to add tag \'%s\', "
1382 "based on tagger \'%s\' and header \'%s\'",
1383 tag, b->name, *header);
1387 char *action_message;
1389 * update the action bits right away, to make
1390 * tagging based on tags set by earlier taggers
1391 * of the same kind possible.
1393 if (update_action_bits_for_tag(csp, tag))
1395 action_message = "Action bits updated accordingly.";
1399 action_message = "No action bits update necessary.";
1402 log_error(LOG_LEVEL_HEADER,
1403 "Tagger \'%s\' added tag \'%s\'. %s",
1404 b->name, tag, action_message);
1409 /* XXX: Is this log-worthy? */
1410 log_error(LOG_LEVEL_HEADER,
1411 "Tagger \'%s\' didn't add tag \'%s\'. "
1412 "Tag already present", b->name, tag);
1421 /* here begins the family of parser functions that reformat header lines */
1423 /*********************************************************************
1425 * Function : filter_header
1427 * Description : Executes all text substitutions from all applying
1428 * +(server|client)-header-filter actions on the header.
1429 * Most of the code was copied from pcrs_filter_response,
1430 * including the rather short variable names
1433 * 1 : csp = Current client state (buffers, headers, etc...)
1434 * 2 : header = On input, pointer to header to modify.
1435 * On output, pointer to the modified header, or NULL
1436 * to remove the header. This function frees the
1437 * original string if necessary.
1439 * Returns : JB_ERR_OK on success and always succeeds
1441 *********************************************************************/
1442 static jb_err filter_header(struct client_state *csp, char **header)
1446 size_t size = strlen(*header);
1448 char *newheader = NULL;
1451 struct re_filterfile_spec *b;
1452 struct list_entry *filtername;
1454 int wanted_filter_type;
1455 int multi_action_index;
1457 if (csp->flags & CSP_FLAG_NO_FILTERING)
1462 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1464 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1465 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1469 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1470 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1473 if (list_is_empty(csp->action->multi[multi_action_index])
1474 || filters_available(csp) == FALSE)
1476 /* Return early if no filters apply or if none are available. */
1480 /* Execute all applying header filters */
1481 for (filtername = csp->action->multi[multi_action_index]->first;
1482 filtername ; filtername = filtername->next)
1484 b = get_filter(csp, filtername->str, wanted_filter_type);
1487 int current_hits = 0;
1488 pcrs_job *joblist = b->joblist;
1490 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1492 if (NULL == joblist)
1494 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1498 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1499 *header, size, b->name);
1501 /* Apply all jobs from the joblist */
1502 for (job = joblist; NULL != job; job = job->next)
1504 matches = pcrs_execute(job, *header, size, &newheader, &size);
1507 current_hits += matches;
1508 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1510 *header = newheader;
1512 else if (0 == matches)
1514 /* Filter doesn't change header */
1520 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1521 *header, b->name, pcrs_strerror(matches));
1522 if (newheader != NULL)
1524 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1530 if (b->dynamic) pcrs_free_joblist(joblist);
1532 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1533 hits += current_hits;
1538 * Additionally checking for hits is important because if
1539 * the continue hack is triggered, server headers can
1540 * arrive empty to separate multiple heads from each other.
1542 if ((0 == size) && hits)
1544 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
1552 /*********************************************************************
1554 * Function : server_connection
1556 * Description : Makes sure a proper "Connection:" header is
1557 * set and signals connection_header_adder to
1561 * 1 : csp = Current client state (buffers, headers, etc...)
1562 * 2 : header = On input, pointer to header to modify.
1563 * On output, pointer to the modified header, or NULL
1564 * to remove the header. This function frees the
1565 * original string if necessary.
1567 * Returns : JB_ERR_OK on success.
1569 *********************************************************************/
1570 static jb_err server_connection(struct client_state *csp, char **header)
1572 if (!strcmpic(*header, "Connection: keep-alive")
1573 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1574 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
1578 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1579 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1581 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
1584 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE))
1586 log_error(LOG_LEVEL_HEADER,
1587 "Keeping the server header '%s' around.", *header);
1590 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
1592 char *old_header = *header;
1594 *header = strdup_or_die("Connection: close");
1595 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
1600 /* Signal server_connection_adder() to return early. */
1601 csp->flags |= CSP_FLAG_SERVER_CONNECTION_HEADER_SET;
1607 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1608 /*********************************************************************
1610 * Function : server_keep_alive
1612 * Description : Stores the server's keep alive timeout.
1615 * 1 : csp = Current client state (buffers, headers, etc...)
1616 * 2 : header = On input, pointer to header to modify.
1617 * On output, pointer to the modified header, or NULL
1618 * to remove the header. This function frees the
1619 * original string if necessary.
1621 * Returns : JB_ERR_OK.
1623 *********************************************************************/
1624 static jb_err server_keep_alive(struct client_state *csp, char **header)
1626 unsigned int keep_alive_timeout;
1627 const char *timeout_position = strstr(*header, "timeout=");
1629 if ((NULL == timeout_position)
1630 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1632 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1636 if (keep_alive_timeout < csp->server_connection.keep_alive_timeout)
1638 log_error(LOG_LEVEL_HEADER,
1639 "Reducing keep-alive timeout from %u to %u.",
1640 csp->server_connection.keep_alive_timeout, keep_alive_timeout);
1641 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1645 /* XXX: Is this log worthy? */
1646 log_error(LOG_LEVEL_HEADER,
1647 "Server keep-alive timeout is %u. Sticking with %u.",
1648 keep_alive_timeout, csp->server_connection.keep_alive_timeout);
1650 csp->flags |= CSP_FLAG_SERVER_KEEP_ALIVE_TIMEOUT_SET;
1657 /*********************************************************************
1659 * Function : server_proxy_connection
1661 * Description : Figures out whether or not we should add a
1662 * Proxy-Connection header.
1665 * 1 : csp = Current client state (buffers, headers, etc...)
1666 * 2 : header = On input, pointer to header to modify.
1667 * On output, pointer to the modified header, or NULL
1668 * to remove the header. This function frees the
1669 * original string if necessary.
1671 * Returns : JB_ERR_OK.
1673 *********************************************************************/
1674 static jb_err server_proxy_connection(struct client_state *csp, char **header)
1676 csp->flags |= CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET;
1681 /*********************************************************************
1683 * Function : proxy_authentication
1685 * Description : Removes headers that are relevant for proxy
1686 * authentication unless forwarding them has
1687 * been explicitly requested.
1690 * 1 : csp = Current client state (buffers, headers, etc...)
1691 * 2 : header = On input, pointer to header to modify.
1692 * On output, pointer to the modified header, or NULL
1693 * to remove the header. This function frees the
1694 * original string if necessary.
1696 * Returns : JB_ERR_OK.
1698 *********************************************************************/
1699 static jb_err proxy_authentication(struct client_state *csp, char **header)
1701 if ((csp->config->feature_flags &
1702 RUNTIME_FEATURE_FORWARD_PROXY_AUTHENTICATION_HEADERS) == 0) {
1703 log_error(LOG_LEVEL_HEADER,
1704 "Forwarding proxy authentication headers is disabled. Crunching: %s", *header);
1711 /*********************************************************************
1713 * Function : client_keep_alive
1715 * Description : Stores the client's keep alive timeout.
1718 * 1 : csp = Current client state (buffers, headers, etc...)
1719 * 2 : header = On input, pointer to header to modify.
1720 * On output, pointer to the modified header, or NULL
1721 * to remove the header. This function frees the
1722 * original string if necessary.
1724 * Returns : JB_ERR_OK.
1726 *********************************************************************/
1727 static jb_err client_keep_alive(struct client_state *csp, char **header)
1729 unsigned int keep_alive_timeout;
1730 const char *timeout_position = strstr(*header, ": ");
1732 if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1734 log_error(LOG_LEVEL_HEADER,
1735 "keep-alive support is disabled. Crunching: %s.", *header);
1740 if ((NULL == timeout_position)
1741 || (1 != sscanf(timeout_position, ": %u", &keep_alive_timeout)))
1743 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1747 if (keep_alive_timeout < csp->config->keep_alive_timeout)
1749 log_error(LOG_LEVEL_HEADER,
1750 "Reducing keep-alive timeout from %u to %u.",
1751 csp->config->keep_alive_timeout, keep_alive_timeout);
1752 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1756 /* XXX: Is this log worthy? */
1757 log_error(LOG_LEVEL_HEADER,
1758 "Client keep-alive timeout is %u. Sticking with %u.",
1759 keep_alive_timeout, csp->config->keep_alive_timeout);
1767 /*********************************************************************
1769 * Function : get_content_length
1771 * Description : Gets the content length specified in a
1772 * Content-Length header.
1775 * 1 : header_value = The Content-Length header value.
1776 * 2 : length = Storage to return the value.
1778 * Returns : JB_ERR_OK on success, or
1779 * JB_ERR_PARSE if no value is recognized.
1781 *********************************************************************/
1782 static jb_err get_content_length(const char *header_value, unsigned long long *length)
1785 assert(sizeof(unsigned long long) > 4);
1786 if (1 != sscanf(header_value, "%I64u", length))
1788 if (1 != sscanf(header_value, "%llu", length))
1791 return JB_ERR_PARSE;
1798 /*********************************************************************
1800 * Function : client_save_content_length
1802 * Description : Save the Content-Length sent by the client.
1805 * 1 : csp = Current client state (buffers, headers, etc...)
1806 * 2 : header = On input, pointer to header to modify.
1807 * On output, pointer to the modified header, or NULL
1808 * to remove the header. This function frees the
1809 * original string if necessary.
1811 * Returns : JB_ERR_OK on success, or
1812 * JB_ERR_MEMORY on out-of-memory error.
1814 *********************************************************************/
1815 static jb_err client_save_content_length(struct client_state *csp, char **header)
1817 unsigned long long content_length = 0;
1818 const char *header_value;
1820 assert(*(*header+14) == ':');
1822 header_value = *header + 15;
1823 if (JB_ERR_OK != get_content_length(header_value, &content_length))
1825 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
1830 csp->expected_client_content_length = content_length;
1835 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1839 /*********************************************************************
1841 * Function : client_connection
1843 * Description : Makes sure a proper "Connection:" header is
1844 * set and signals connection_header_adder
1848 * 1 : csp = Current client state (buffers, headers, etc...)
1849 * 2 : header = On input, pointer to header to modify.
1850 * On output, pointer to the modified header, or NULL
1851 * to remove the header. This function frees the
1852 * original string if necessary.
1854 * Returns : JB_ERR_OK on success.
1856 *********************************************************************/
1857 static jb_err client_connection(struct client_state *csp, char **header)
1859 static const char connection_close[] = "Connection: close";
1861 if (!strcmpic(*header, connection_close))
1863 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1864 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING)
1865 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1867 if (!strcmpic(csp->http->ver, "HTTP/1.1"))
1869 log_error(LOG_LEVEL_HEADER,
1870 "Removing \'%s\' to imply keep-alive.", *header);
1873 * While we imply keep-alive to the server,
1874 * we have to remember that the client didn't.
1876 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1880 char *old_header = *header;
1882 *header = strdup_or_die("Connection: keep-alive");
1883 log_error(LOG_LEVEL_HEADER,
1884 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1890 log_error(LOG_LEVEL_HEADER,
1891 "Keeping the client header '%s' around. "
1892 "The connection will not be kept alive.",
1894 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1897 else if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
1898 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1900 log_error(LOG_LEVEL_HEADER,
1901 "Keeping the client header '%s' around. "
1902 "The server connection will be kept alive if possible.",
1904 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1905 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1909 char *old_header = *header;
1911 *header = strdup_or_die(connection_close);
1912 log_error(LOG_LEVEL_HEADER,
1913 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1917 /* Signal client_connection_header_adder() to return early. */
1918 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
1924 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1925 /*********************************************************************
1927 * Function : client_proxy_connection
1929 * Description : Sets the CLIENT_CONNECTION_KEEP_ALIVE flag when
1930 * appropriate and removes the Proxy-Connection
1934 * 1 : csp = Current client state (buffers, headers, etc...)
1935 * 2 : header = On input, pointer to header to modify.
1936 * On output, pointer to the modified header, or NULL
1937 * to remove the header. This function frees the
1938 * original string if necessary.
1940 * Returns : JB_ERR_OK
1942 *********************************************************************/
1943 static jb_err client_proxy_connection(struct client_state *csp, char **header)
1945 if (0 == (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
1946 && (csp->http->ssl == 0)
1947 && (NULL == strstr(*header, "close")))
1949 log_error(LOG_LEVEL_HEADER,
1950 "The client connection can be kept alive due to: %s", *header);
1951 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1953 crumble(csp, header);
1957 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1960 /*********************************************************************
1962 * Function : client_transfer_encoding
1964 * Description : Raise the CSP_FLAG_CHUNKED_CLIENT_BODY flag if
1965 * the request body is "chunked"
1967 * XXX: Currently not called through sed() as we
1968 * need the flag earlier on. Should be fixed.
1971 * 1 : csp = Current client state (buffers, headers, etc...)
1972 * 2 : header = On input, pointer to header to modify.
1973 * On output, pointer to the modified header, or NULL
1974 * to remove the header. This function frees the
1975 * original string if necessary.
1977 * Returns : JB_ERR_OK on success, or
1979 *********************************************************************/
1980 jb_err client_transfer_encoding(struct client_state *csp, char **header)
1982 if (strstr(*header, "chunked"))
1984 csp->flags |= CSP_FLAG_CHUNKED_CLIENT_BODY;
1985 log_error(LOG_LEVEL_HEADER, "Expecting chunked client body");
1992 /*********************************************************************
1994 * Function : crumble
1996 * Description : This is called if a header matches a pattern to "crunch"
1999 * 1 : csp = Current client state (buffers, headers, etc...)
2000 * 2 : header = On input, pointer to header to modify.
2001 * On output, pointer to the modified header, or NULL
2002 * to remove the header. This function frees the
2003 * original string if necessary.
2005 * Returns : JB_ERR_OK on success, or
2006 * JB_ERR_MEMORY on out-of-memory error.
2008 *********************************************************************/
2009 static jb_err crumble(struct client_state *csp, char **header)
2012 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2018 /*********************************************************************
2020 * Function : crunch_server_header
2022 * Description : Crunch server header if it matches a string supplied by the
2023 * user. Called from `sed'.
2026 * 1 : csp = Current client state (buffers, headers, etc...)
2027 * 2 : header = On input, pointer to header to modify.
2028 * On output, pointer to the modified header, or NULL
2029 * to remove the header. This function frees the
2030 * original string if necessary.
2032 * Returns : JB_ERR_OK on success and always succeeds
2034 *********************************************************************/
2035 static jb_err crunch_server_header(struct client_state *csp, char **header)
2037 const char *crunch_pattern;
2039 /* Do we feel like crunching? */
2040 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2042 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2044 /* Is the current header the lucky one? */
2045 if (strstr(*header, crunch_pattern))
2047 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2056 /*********************************************************************
2058 * Function : server_content_type
2060 * Description : Set the content-type for filterable types (text/.*,
2061 * .*xml.*, .*script.* and image/gif) unless filtering has been
2062 * forbidden (CT_TABOO) while parsing earlier headers.
2063 * NOTE: Since text/plain is commonly used by web servers
2064 * for files whose correct type is unknown, we don't
2065 * set CT_TEXT for it.
2068 * 1 : csp = Current client state (buffers, headers, etc...)
2069 * 2 : header = On input, pointer to header to modify.
2070 * On output, pointer to the modified header, or NULL
2071 * to remove the header. This function frees the
2072 * original string if necessary.
2074 * Returns : JB_ERR_OK on success, or
2075 * JB_ERR_MEMORY on out-of-memory error.
2077 *********************************************************************/
2078 static jb_err server_content_type(struct client_state *csp, char **header)
2080 /* Remove header if it isn't the first Content-Type header */
2081 if ((csp->content_type & CT_DECLARED))
2083 if (content_filters_enabled(csp->action))
2086 * Making sure the client interprets the content the same way
2087 * Privoxy did is only relevant if Privoxy modified it.
2089 * Checking for this is "hard" as it's not yet known when
2090 * this function is called, thus go shopping and and just
2091 * check if Privoxy could filter it.
2093 * The main thing is that we don't mess with the headers
2094 * unless the user signalled that it's acceptable.
2096 log_error(LOG_LEVEL_HEADER,
2097 "Multiple Content-Type headers detected. "
2098 "Removing and ignoring: %s",
2106 * Signal that the Content-Type has been set.
2108 csp->content_type |= CT_DECLARED;
2110 if (!(csp->content_type & CT_TABOO))
2113 * XXX: The assumption that text/plain is a sign of
2114 * binary data seems to be somewhat unreasonable nowadays
2115 * and should be dropped after 3.0.8 is out.
2117 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2118 || strstr(*header, "xml")
2119 || strstr(*header, "script"))
2121 csp->content_type |= CT_TEXT;
2123 else if (strstr(*header, "image/gif"))
2125 csp->content_type |= CT_GIF;
2130 * Are we messing with the content type?
2132 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2135 * Make sure the user doesn't accidentally
2136 * change the content type of binary documents.
2138 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2141 *header = strdup_or_die("Content-Type: ");
2142 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2146 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2147 return JB_ERR_MEMORY;
2149 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2153 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2154 "It doesn't look like a content type that should be filtered. "
2155 "Enable force-text-mode if you know what you're doing.", *header);
2163 /*********************************************************************
2165 * Function : server_transfer_coding
2167 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2168 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2169 * - Remove header if body was chunked but has been
2170 * de-chunked for filtering.
2173 * 1 : csp = Current client state (buffers, headers, etc...)
2174 * 2 : header = On input, pointer to header to modify.
2175 * On output, pointer to the modified header, or NULL
2176 * to remove the header. This function frees the
2177 * original string if necessary.
2179 * Returns : JB_ERR_OK on success, or
2180 * JB_ERR_MEMORY on out-of-memory error.
2182 *********************************************************************/
2183 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2186 * Turn off pcrs and gif filtering if body compressed
2188 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2192 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2194 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2195 csp->http->cmd, *header);
2196 #endif /* def FEATURE_ZLIB */
2197 csp->content_type = CT_TABOO;
2201 * Raise flag if body chunked
2203 if (strstr(*header, "chunked"))
2205 csp->flags |= CSP_FLAG_CHUNKED;
2208 * If the body was modified, it has been de-chunked first
2209 * and the header must be removed.
2211 * FIXME: If there is more than one transfer encoding,
2212 * only the "chunked" part should be removed here.
2214 if (csp->flags & CSP_FLAG_MODIFIED)
2216 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2225 /*********************************************************************
2227 * Function : server_content_encoding
2229 * Description : Used to check if the content is compressed, and if
2230 * FEATURE_ZLIB is disabled, filtering is disabled as
2233 * If FEATURE_ZLIB is enabled and the compression type
2234 * supported, the content is marked for decompression.
2236 * XXX: Doesn't properly deal with multiple or with
2237 * unsupported but unknown encodings.
2238 * Is case-sensitive but shouldn't be.
2241 * 1 : csp = Current client state (buffers, headers, etc...)
2242 * 2 : header = On input, pointer to header to modify.
2243 * On output, pointer to the modified header, or NULL
2244 * to remove the header. This function frees the
2245 * original string if necessary.
2247 * Returns : JB_ERR_OK on success, or
2248 * JB_ERR_MEMORY on out-of-memory error.
2250 *********************************************************************/
2251 static jb_err server_content_encoding(struct client_state *csp, char **header)
2254 if (strstr(*header, "sdch"))
2257 * Shared Dictionary Compression over HTTP isn't supported,
2258 * filtering it anyway is pretty much guaranteed to mess up
2261 csp->content_type |= CT_TABOO;
2264 * Log a warning if the user expects the content to be filtered.
2266 if ((csp->rlist != NULL) &&
2267 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2269 log_error(LOG_LEVEL_INFO,
2270 "SDCH-compressed content detected, content filtering disabled. "
2271 "Consider suppressing SDCH offers made by the client.");
2274 else if (strstr(*header, "gzip"))
2276 /* Mark for gzip decompression */
2277 csp->content_type |= CT_GZIP;
2279 else if (strstr(*header, "deflate"))
2281 /* Mark for zlib decompression */
2282 csp->content_type |= CT_DEFLATE;
2284 else if (strstr(*header, "compress"))
2287 * We can't decompress this; therefore we can't filter
2290 csp->content_type |= CT_TABOO;
2292 #else /* !defined(FEATURE_ZLIB) */
2294 * XXX: Using a black list here isn't the right approach.
2296 * In case of SDCH, building with zlib support isn't
2299 if (strstr(*header, "gzip") ||
2300 strstr(*header, "compress") ||
2301 strstr(*header, "deflate") ||
2302 strstr(*header, "sdch"))
2305 * Body is compressed, turn off pcrs and gif filtering.
2307 csp->content_type |= CT_TABOO;
2310 * Log a warning if the user expects the content to be filtered.
2312 if ((csp->rlist != NULL) &&
2313 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2315 log_error(LOG_LEVEL_INFO,
2316 "Compressed content detected, content filtering disabled. "
2317 "Consider recompiling Privoxy with zlib support or "
2318 "enable the prevent-compression action.");
2321 #endif /* defined(FEATURE_ZLIB) */
2329 /*********************************************************************
2331 * Function : server_adjust_content_encoding
2333 * Description : Remove the Content-Encoding header if the
2334 * decompression was successful and the content
2338 * 1 : csp = Current client state (buffers, headers, etc...)
2339 * 2 : header = On input, pointer to header to modify.
2340 * On output, pointer to the modified header, or NULL
2341 * to remove the header. This function frees the
2342 * original string if necessary.
2344 * Returns : JB_ERR_OK on success, or
2345 * JB_ERR_MEMORY on out-of-memory error.
2347 *********************************************************************/
2348 static jb_err server_adjust_content_encoding(struct client_state *csp, char **header)
2350 if ((csp->flags & CSP_FLAG_MODIFIED)
2351 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2354 * We successfully decompressed the content,
2355 * and have to clean the header now, so the
2356 * client no longer expects compressed data.
2358 * XXX: There is a difference between cleaning
2359 * and removing it completely.
2361 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2368 #endif /* defined(FEATURE_ZLIB) */
2371 /*********************************************************************
2373 * Function : server_adjust_content_length
2375 * Description : Adjust Content-Length header if we modified
2379 * 1 : csp = Current client state (buffers, headers, etc...)
2380 * 2 : header = On input, pointer to header to modify.
2381 * On output, pointer to the modified header, or NULL
2382 * to remove the header. This function frees the
2383 * original string if necessary.
2385 * Returns : JB_ERR_OK on success, or
2386 * JB_ERR_MEMORY on out-of-memory error.
2388 *********************************************************************/
2389 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2391 /* Regenerate header if the content was modified. */
2392 if (csp->flags & CSP_FLAG_MODIFIED)
2394 const size_t header_length = 50;
2396 *header = malloc(header_length);
2397 if (*header == NULL)
2399 return JB_ERR_MEMORY;
2401 create_content_length_header(csp->content_length, *header, header_length);
2402 log_error(LOG_LEVEL_HEADER,
2403 "Adjusted Content-Length to %llu", csp->content_length);
2410 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2411 /*********************************************************************
2413 * Function : server_save_content_length
2415 * Description : Save the Content-Length sent by the server.
2418 * 1 : csp = Current client state (buffers, headers, etc...)
2419 * 2 : header = On input, pointer to header to modify.
2420 * On output, pointer to the modified header, or NULL
2421 * to remove the header. This function frees the
2422 * original string if necessary.
2424 * Returns : JB_ERR_OK on success, or
2425 * JB_ERR_MEMORY on out-of-memory error.
2427 *********************************************************************/
2428 static jb_err server_save_content_length(struct client_state *csp, char **header)
2430 unsigned long long content_length = 0;
2431 const char *header_value;
2433 assert(*(*header+14) == ':');
2435 header_value = *header + 15;
2436 if (JB_ERR_OK != get_content_length(header_value, &content_length))
2438 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2443 csp->expected_content_length = content_length;
2444 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
2445 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2450 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2453 /*********************************************************************
2455 * Function : server_content_md5
2457 * Description : Crumble any Content-MD5 headers if the document was
2458 * modified. FIXME: Should we re-compute instead?
2461 * 1 : csp = Current client state (buffers, headers, etc...)
2462 * 2 : header = On input, pointer to header to modify.
2463 * On output, pointer to the modified header, or NULL
2464 * to remove the header. This function frees the
2465 * original string if necessary.
2467 * Returns : JB_ERR_OK on success, or
2468 * JB_ERR_MEMORY on out-of-memory error.
2470 *********************************************************************/
2471 static jb_err server_content_md5(struct client_state *csp, char **header)
2473 if (csp->flags & CSP_FLAG_MODIFIED)
2475 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2483 /*********************************************************************
2485 * Function : server_content_disposition
2487 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2488 * Called from `sed'.
2491 * 1 : csp = Current client state (buffers, headers, etc...)
2492 * 2 : header = On input, pointer to header to modify.
2493 * On output, pointer to the modified header, or NULL
2494 * to remove the header. This function frees the
2495 * original string if necessary.
2497 * Returns : JB_ERR_OK on success, or
2498 * JB_ERR_MEMORY on out-of-memory error.
2500 *********************************************************************/
2501 static jb_err server_content_disposition(struct client_state *csp, char **header)
2506 * Are we messing with the Content-Disposition header?
2508 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2514 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2516 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2519 * Blocking content-disposition header
2521 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2528 * Replacing Content-Disposition header
2531 *header = strdup("Content-Disposition: ");
2532 string_append(header, newval);
2534 if (*header != NULL)
2536 log_error(LOG_LEVEL_HEADER,
2537 "Content-Disposition header crunched and replaced with: %s", *header);
2540 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2544 /*********************************************************************
2546 * Function : server_last_modified
2548 * Description : Changes Last-Modified header to the actual date
2549 * to help hide-if-modified-since.
2550 * Called from `sed'.
2553 * 1 : csp = Current client state (buffers, headers, etc...)
2554 * 2 : header = On input, pointer to header to modify.
2555 * On output, pointer to the modified header, or NULL
2556 * to remove the header. This function frees the
2557 * original string if necessary.
2559 * Returns : JB_ERR_OK on success, or
2560 * JB_ERR_MEMORY on out-of-memory error.
2562 *********************************************************************/
2563 static jb_err server_last_modified(struct client_state *csp, char **header)
2566 time_t last_modified;
2570 * Are we messing with the Last-Modified header?
2572 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2578 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2580 if (0 == strcmpic(newval, "block"))
2583 * Blocking Last-Modified header. Useless but why not.
2585 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2589 else if (0 == strcmpic(newval, "reset-to-request-time"))
2592 * Setting Last-Modified Header to now.
2595 get_http_time(0, buf, sizeof(buf));
2597 *header = strdup("Last-Modified: ");
2598 string_append(header, buf);
2600 if (*header == NULL)
2602 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2606 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2609 else if (0 == strcmpic(newval, "randomize"))
2611 const char *header_time = *header + sizeof("Last-Modified:");
2613 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2615 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2617 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2623 struct tm *timeptr = NULL;
2625 #ifdef HAVE_GMTIME_R
2629 rtime = (long int)difftime(now, last_modified);
2632 long int days, hours, minutes, seconds;
2633 const int negative_delta = (rtime < 0);
2638 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2640 rtime = pick_from_range(rtime);
2645 last_modified += rtime;
2646 #ifdef HAVE_GMTIME_R
2647 timeptr = gmtime_r(&last_modified, &gmt);
2648 #elif defined(MUTEX_LOCKS_AVAILABLE)
2649 privoxy_mutex_lock(&gmtime_mutex);
2650 timeptr = gmtime(&last_modified);
2651 privoxy_mutex_unlock(&gmtime_mutex);
2653 timeptr = gmtime(&last_modified);
2655 if ((NULL == timeptr) || !strftime(newheader,
2656 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
2658 log_error(LOG_LEVEL_ERROR,
2659 "Randomizing '%s' failed. Crunching the header without replacement.",
2666 *header = strdup("Last-Modified: ");
2667 string_append(header, newheader);
2669 if (*header == NULL)
2671 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2672 return JB_ERR_MEMORY;
2675 days = rtime / (3600 * 24);
2676 hours = rtime / 3600 % 24;
2677 minutes = rtime / 60 % 60;
2678 seconds = rtime % 60;
2680 log_error(LOG_LEVEL_HEADER,
2681 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2682 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2683 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2687 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2696 /*********************************************************************
2698 * Function : client_accept_encoding
2700 * Description : Rewrite the client's Accept-Encoding header so that
2701 * if doesn't allow compression, if the action applies.
2702 * Note: For HTTP/1.0 the absence of the header is enough.
2705 * 1 : csp = Current client state (buffers, headers, etc...)
2706 * 2 : header = On input, pointer to header to modify.
2707 * On output, pointer to the modified header, or NULL
2708 * to remove the header. This function frees the
2709 * original string if necessary.
2711 * Returns : JB_ERR_OK on success, or
2712 * JB_ERR_MEMORY on out-of-memory error.
2714 *********************************************************************/
2715 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2717 #ifdef FEATURE_COMPRESSION
2718 if ((csp->config->feature_flags & RUNTIME_FEATURE_COMPRESSION)
2719 && strstr(*header, "deflate"))
2721 csp->flags |= CSP_FLAG_CLIENT_SUPPORTS_DEFLATE;
2724 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2726 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2734 /*********************************************************************
2736 * Function : client_te
2738 * Description : Rewrite the client's TE header so that
2739 * if doesn't allow compression, if the action applies.
2742 * 1 : csp = Current client state (buffers, headers, etc...)
2743 * 2 : header = On input, pointer to header to modify.
2744 * On output, pointer to the modified header, or NULL
2745 * to remove the header. This function frees the
2746 * original string if necessary.
2748 * Returns : JB_ERR_OK on success, or
2749 * JB_ERR_MEMORY on out-of-memory error.
2751 *********************************************************************/
2752 static jb_err client_te(struct client_state *csp, char **header)
2754 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2757 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2764 /*********************************************************************
2766 * Function : client_referrer
2768 * Description : Handle the "referer" config setting properly.
2769 * Called from `sed'.
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_referrer(struct client_state *csp, char **header)
2784 const char *parameter;
2785 /* booleans for parameters we have to check multiple times */
2786 int parameter_conditional_block;
2787 int parameter_conditional_forge;
2789 #ifdef FEATURE_FORCE_LOAD
2791 * Since the referrer can include the prefix even
2792 * if the request itself is non-forced, we must
2793 * clean it unconditionally.
2795 * XXX: strclean is too broad
2797 strclean(*header, FORCE_PREFIX);
2798 #endif /* def FEATURE_FORCE_LOAD */
2800 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2802 /* Nothing left to do */
2806 parameter = csp->action->string[ACTION_STRING_REFERER];
2807 assert(parameter != NULL);
2808 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2809 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2811 if (!parameter_conditional_block && !parameter_conditional_forge)
2814 * As conditional-block and conditional-forge are the only
2815 * parameters that rely on the original referrer, we can
2816 * remove it now for all the others.
2821 if (0 == strcmpic(parameter, "block"))
2823 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2826 else if (parameter_conditional_block || parameter_conditional_forge)
2828 return handle_conditional_hide_referrer_parameter(header,
2829 csp->http->hostport, parameter_conditional_block);
2831 else if (0 == strcmpic(parameter, "forge"))
2833 return create_forged_referrer(header, csp->http->hostport);
2837 /* interpret parameter as user-supplied referer to fake */
2838 return create_fake_referrer(header, parameter);
2843 /*********************************************************************
2845 * Function : client_accept_language
2847 * Description : Handle the "Accept-Language" config setting properly.
2848 * Called from `sed'.
2851 * 1 : csp = Current client state (buffers, headers, etc...)
2852 * 2 : header = On input, pointer to header to modify.
2853 * On output, pointer to the modified header, or NULL
2854 * to remove the header. This function frees the
2855 * original string if necessary.
2857 * Returns : JB_ERR_OK on success, or
2858 * JB_ERR_MEMORY on out-of-memory error.
2860 *********************************************************************/
2861 static jb_err client_accept_language(struct client_state *csp, char **header)
2866 * Are we messing with the Accept-Language?
2868 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2870 /*I don't think so*/
2874 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2876 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2879 * Blocking Accept-Language header
2881 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2888 * Replacing Accept-Language header
2891 *header = strdup("Accept-Language: ");
2892 string_append(header, newval);
2894 if (*header == NULL)
2896 log_error(LOG_LEVEL_ERROR,
2897 "Insufficient memory. Accept-Language header crunched without replacement.");
2901 log_error(LOG_LEVEL_HEADER,
2902 "Accept-Language header crunched and replaced with: %s", *header);
2905 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2909 /*********************************************************************
2911 * Function : crunch_client_header
2913 * Description : Crunch client header if it matches a string supplied by the
2914 * user. Called from `sed'.
2917 * 1 : csp = Current client state (buffers, headers, etc...)
2918 * 2 : header = On input, pointer to header to modify.
2919 * On output, pointer to the modified header, or NULL
2920 * to remove the header. This function frees the
2921 * original string if necessary.
2923 * Returns : JB_ERR_OK on success and always succeeds
2925 *********************************************************************/
2926 static jb_err crunch_client_header(struct client_state *csp, char **header)
2928 const char *crunch_pattern;
2930 /* Do we feel like crunching? */
2931 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2933 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2935 /* Is the current header the lucky one? */
2936 if (strstr(*header, crunch_pattern))
2938 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
2946 /*********************************************************************
2948 * Function : client_uagent
2950 * Description : Handle the "user-agent" config setting properly
2951 * and remember its original value to enable browser
2952 * bug workarounds. Called from `sed'.
2955 * 1 : csp = Current client state (buffers, headers, etc...)
2956 * 2 : header = On input, pointer to header to modify.
2957 * On output, pointer to the modified header, or NULL
2958 * to remove the header. This function frees the
2959 * original string if necessary.
2961 * Returns : JB_ERR_OK on success, or
2962 * JB_ERR_MEMORY on out-of-memory error.
2964 *********************************************************************/
2965 static jb_err client_uagent(struct client_state *csp, char **header)
2969 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
2974 newval = csp->action->string[ACTION_STRING_USER_AGENT];
2981 *header = strdup("User-Agent: ");
2982 string_append(header, newval);
2984 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
2986 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2990 /*********************************************************************
2992 * Function : client_ua
2994 * Description : Handle "ua-" headers properly. Called from `sed'.
2997 * 1 : csp = Current client state (buffers, headers, etc...)
2998 * 2 : header = On input, pointer to header to modify.
2999 * On output, pointer to the modified header, or NULL
3000 * to remove the header. This function frees the
3001 * original string if necessary.
3003 * Returns : JB_ERR_OK on success, or
3004 * JB_ERR_MEMORY on out-of-memory error.
3006 *********************************************************************/
3007 static jb_err client_ua(struct client_state *csp, char **header)
3009 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3011 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3019 /*********************************************************************
3021 * Function : client_from
3023 * Description : Handle the "from" config setting properly.
3024 * Called from `sed'.
3027 * 1 : csp = Current client state (buffers, headers, etc...)
3028 * 2 : header = On input, pointer to header to modify.
3029 * On output, pointer to the modified header, or NULL
3030 * to remove the header. This function frees the
3031 * original string if necessary.
3033 * Returns : JB_ERR_OK on success, or
3034 * JB_ERR_MEMORY on out-of-memory error.
3036 *********************************************************************/
3037 static jb_err client_from(struct client_state *csp, char **header)
3041 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3048 newval = csp->action->string[ACTION_STRING_FROM];
3051 * Are we blocking the e-mail address?
3053 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
3055 log_error(LOG_LEVEL_HEADER, "crunched From!");
3059 log_error(LOG_LEVEL_HEADER, " modified");
3061 *header = strdup("From: ");
3062 string_append(header, newval);
3064 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3068 /*********************************************************************
3070 * Function : client_send_cookie
3072 * Description : Crunches the "cookie" header if necessary.
3073 * Called from `sed'.
3075 * XXX: Stupid name, doesn't send squat.
3078 * 1 : csp = Current client state (buffers, headers, etc...)
3079 * 2 : header = On input, pointer to header to modify.
3080 * On output, pointer to the modified header, or NULL
3081 * to remove the header. This function frees the
3082 * original string if necessary.
3084 * Returns : JB_ERR_OK on success, or
3085 * JB_ERR_MEMORY on out-of-memory error.
3087 *********************************************************************/
3088 static jb_err client_send_cookie(struct client_state *csp, char **header)
3090 if (csp->action->flags & ACTION_CRUNCH_OUTGOING_COOKIES)
3092 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3100 /*********************************************************************
3102 * Function : client_x_forwarded
3104 * Description : Handle the "x-forwarded-for" config setting properly,
3105 * also used in the add_client_headers list. Called from `sed'.
3108 * 1 : csp = Current client state (buffers, headers, etc...)
3109 * 2 : header = On input, pointer to header to modify.
3110 * On output, pointer to the modified header, or NULL
3111 * to remove the header. This function frees the
3112 * original string if necessary.
3114 * Returns : JB_ERR_OK on success, or
3115 * JB_ERR_MEMORY on out-of-memory error.
3117 *********************************************************************/
3118 jb_err client_x_forwarded(struct client_state *csp, char **header)
3120 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
3122 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
3124 if (0 == strcmpic(parameter, "block"))
3127 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3129 else if (0 == strcmpic(parameter, "add"))
3131 string_append(header, ", ");
3132 string_append(header, csp->ip_addr_str);
3134 if (*header == NULL)
3136 return JB_ERR_MEMORY;
3138 log_error(LOG_LEVEL_HEADER,
3139 "Appended client IP address to %s", *header);
3140 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
3144 log_error(LOG_LEVEL_FATAL,
3145 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
3153 /*********************************************************************
3155 * Function : client_max_forwards
3157 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3158 * from the value of the Max-Forwards header field.
3161 * 1 : csp = Current client state (buffers, headers, etc...)
3162 * 2 : header = On input, pointer to header to modify.
3163 * On output, pointer to the modified header, or NULL
3164 * to remove the header. This function frees the
3165 * original string if necessary.
3167 * Returns : JB_ERR_OK on success, or
3168 * JB_ERR_MEMORY on out-of-memory error.
3170 *********************************************************************/
3171 static jb_err client_max_forwards(struct client_state *csp, char **header)
3175 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3176 (0 == strcmpic(csp->http->gpc, "options")))
3178 assert(*(*header+12) == ':');
3179 if (1 == sscanf(*header+12, ": %d", &max_forwards))
3181 if (max_forwards > 0)
3183 snprintf(*header, strlen(*header)+1, "Max-Forwards: %d", --max_forwards);
3184 log_error(LOG_LEVEL_HEADER,
3185 "Max-Forwards value for %s request reduced to %d.",
3186 csp->http->gpc, max_forwards);
3188 else if (max_forwards < 0)
3190 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3196 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3205 /*********************************************************************
3207 * Function : client_host
3209 * Description : If the request URI did not contain host and
3210 * port information, parse and evaluate the Host
3214 * 1 : csp = Current client state (buffers, headers, etc...)
3215 * 2 : header = On input, pointer to header to modify.
3216 * On output, pointer to the modified header, or NULL
3217 * to remove the header. This function frees the
3218 * original string if necessary.
3220 * Returns : JB_ERR_OK on success, or
3221 * JB_ERR_MEMORY on out-of-memory error.
3223 *********************************************************************/
3224 static jb_err client_host(struct client_state *csp, char **header)
3228 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3229 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3232 p = strdup_or_die((*header)+6);
3234 q = strdup_or_die(p);
3236 freez(csp->http->hostport);
3237 csp->http->hostport = p;
3238 freez(csp->http->host);
3239 csp->http->host = q;
3240 q = strchr(csp->http->host, ':');
3243 /* Terminate hostname and evaluate port string */
3245 csp->http->port = atoi(q);
3249 csp->http->port = csp->http->ssl ? 443 : 80;
3252 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3253 csp->http->hostport, csp->http->host, csp->http->port);
3256 /* Signal client_host_adder() to return right away */
3257 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3263 /*********************************************************************
3265 * Function : client_if_modified_since
3267 * Description : Remove or modify the If-Modified-Since header.
3270 * 1 : csp = Current client state (buffers, headers, etc...)
3271 * 2 : header = On input, pointer to header to modify.
3272 * On output, pointer to the modified header, or NULL
3273 * to remove the header. This function frees the
3274 * original string if necessary.
3276 * Returns : JB_ERR_OK on success, or
3277 * JB_ERR_MEMORY on out-of-memory error.
3279 *********************************************************************/
3280 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3283 #ifdef HAVE_GMTIME_R
3286 struct tm *timeptr = NULL;
3291 if (0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3294 * The client got an error message because of a temporary problem,
3295 * the problem is gone and the client now tries to revalidate our
3296 * error message on the real server. The revalidation would always
3297 * end with the transmission of the whole document and there is
3298 * no need to expose the bogus If-Modified-Since header.
3300 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3303 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3305 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3307 if ((0 == strcmpic(newval, "block")))
3309 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3312 else /* add random value */
3314 const char *header_time = *header + sizeof("If-Modified-Since:");
3316 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3318 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3323 long int hours, minutes, seconds;
3324 long int rtime = strtol(newval, &endptr, 0);
3325 const int negative_range = (rtime < 0);
3329 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3330 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3336 rtime = pick_from_range(rtime);
3340 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3343 tm += rtime * (negative_range ? -1 : 1);
3344 #ifdef HAVE_GMTIME_R
3345 timeptr = gmtime_r(&tm, &gmt);
3346 #elif defined(MUTEX_LOCKS_AVAILABLE)
3347 privoxy_mutex_lock(&gmtime_mutex);
3348 timeptr = gmtime(&tm);
3349 privoxy_mutex_unlock(&gmtime_mutex);
3351 timeptr = gmtime(&tm);
3353 if ((NULL == timeptr) || !strftime(newheader,
3354 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
3356 log_error(LOG_LEVEL_ERROR,
3357 "Randomizing '%s' failed. Crunching the header without replacement.",
3364 *header = strdup("If-Modified-Since: ");
3365 string_append(header, newheader);
3367 if (*header == NULL)
3369 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3370 return JB_ERR_MEMORY;
3373 hours = rtime / 3600;
3374 minutes = rtime / 60 % 60;
3375 seconds = rtime % 60;
3377 log_error(LOG_LEVEL_HEADER,
3378 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3379 *header, (negative_range) ? "subtracted" : "added", hours,
3380 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3381 seconds, (seconds == 1) ? ")" : "s)");
3390 /*********************************************************************
3392 * Function : client_if_none_match
3394 * Description : Remove the If-None-Match header.
3397 * 1 : csp = Current client state (buffers, headers, etc...)
3398 * 2 : header = On input, pointer to header to modify.
3399 * On output, pointer to the modified header, or NULL
3400 * to remove the header. This function frees the
3401 * original string if necessary.
3403 * Returns : JB_ERR_OK on success, or
3404 * JB_ERR_MEMORY on out-of-memory error.
3406 *********************************************************************/
3407 static jb_err client_if_none_match(struct client_state *csp, char **header)
3409 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3411 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3419 /*********************************************************************
3421 * Function : client_x_filter
3423 * Description : Disables filtering if the client set "X-Filter: No".
3424 * Called from `sed'.
3427 * 1 : csp = Current client state (buffers, headers, etc...)
3428 * 2 : header = On input, pointer to header to modify.
3429 * On output, pointer to the modified header, or NULL
3430 * to remove the header. This function frees the
3431 * original string if necessary.
3433 * Returns : JB_ERR_OK on success
3435 *********************************************************************/
3436 jb_err client_x_filter(struct client_state *csp, char **header)
3438 if (0 == strcmpic(*header, "X-Filter: No"))
3440 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3442 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3446 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3448 log_error(LOG_LEVEL_HEADER,
3449 "force-text-mode overruled the client's request to fetch without filtering!");
3453 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3454 csp->flags |= CSP_FLAG_NO_FILTERING;
3455 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3457 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3465 /*********************************************************************
3467 * Function : client_range
3469 * Description : Removes Range, Request-Range and If-Range headers if
3470 * content filtering is enabled and the range doesn't
3473 * If the client's version of the document has been
3474 * altered by Privoxy, the server could interpret the
3475 * range differently than the client intended in which
3476 * case the user could end up with corrupted content.
3478 * If the range starts at byte 0 this isn't an issue
3479 * so the header can pass. Partial requests like this
3480 * are used to render preview images for videos without
3481 * downloading the whole video.
3483 * While HTTP doesn't require that range requests are
3484 * honoured and the client could simply abort the download
3485 * after receiving a sufficient amount of data, various
3486 * clients don't handle complete responses to range
3487 * requests gracefully and emit misleading error messages
3491 * 1 : csp = Current client state (buffers, headers, etc...)
3492 * 2 : header = On input, pointer to header to modify.
3493 * On output, pointer to the modified header, or NULL
3494 * to remove the header. This function frees the
3495 * original string if necessary.
3497 * Returns : JB_ERR_OK
3499 *********************************************************************/
3500 static jb_err client_range(struct client_state *csp, char **header)
3502 if (content_filters_enabled(csp->action)
3503 && (0 != strncmpic(strstr(*header, ":"), ": bytes=0-", 10)))
3505 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3506 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3513 /* the following functions add headers directly to the header list */
3515 /*********************************************************************
3517 * Function : client_host_adder
3519 * Description : Adds the Host: header field if it is missing.
3520 * Called from `sed'.
3523 * 1 : csp = Current client state (buffers, headers, etc...)
3525 * Returns : JB_ERR_OK on success, or
3526 * JB_ERR_MEMORY on out-of-memory error.
3528 *********************************************************************/
3529 static jb_err client_host_adder(struct client_state *csp)
3534 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3536 /* Header already set by the client, nothing to do. */
3540 if (!csp->http->hostport || !*(csp->http->hostport))
3542 /* XXX: When does this happen and why is it OK? */
3543 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3548 * remove 'user:pass@' from 'proto://user:pass@host'
3550 if ((p = strchr( csp->http->hostport, '@')) != NULL)
3556 p = csp->http->hostport;
3559 /* XXX: Just add it, we already made sure that it will be unique */
3560 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3561 err = enlist_unique_header(csp->headers, "Host", p);
3567 /*********************************************************************
3569 * Function : client_xtra_adder
3571 * Description : Used in the add_client_headers list. Called from `sed'.
3574 * 1 : csp = Current client state (buffers, headers, etc...)
3576 * Returns : JB_ERR_OK on success, or
3577 * JB_ERR_MEMORY on out-of-memory error.
3579 *********************************************************************/
3580 static jb_err client_xtra_adder(struct client_state *csp)
3582 struct list_entry *lst;
3585 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3586 lst ; lst = lst->next)
3588 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3589 err = enlist(csp->headers, lst->str);
3601 /*********************************************************************
3603 * Function : client_x_forwarded_for_adder
3605 * Description : Used in the add_client_headers list. Called from `sed'.
3608 * 1 : csp = Current client state (buffers, headers, etc...)
3610 * Returns : JB_ERR_OK on success, or
3611 * JB_ERR_MEMORY on out-of-memory error.
3613 *********************************************************************/
3614 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3616 char *header = NULL;
3619 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3620 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3621 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3624 * If we aren't adding X-Forwarded-For headers,
3625 * or we already appended an existing X-Forwarded-For
3626 * header, there's nothing left to do here.
3631 header = strdup("X-Forwarded-For: ");
3632 string_append(&header, csp->ip_addr_str);
3636 return JB_ERR_MEMORY;
3639 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3640 err = enlist(csp->headers, header);
3647 /*********************************************************************
3649 * Function : server_connection_adder
3651 * Description : Adds an appropriate "Connection:" header to csp->headers
3652 * unless the header was already present. Called from `sed'.
3655 * 1 : csp = Current client state (buffers, headers, etc...)
3657 * Returns : JB_ERR_OK on success, or
3658 * JB_ERR_MEMORY on out-of-memory error.
3660 *********************************************************************/
3661 static jb_err server_connection_adder(struct client_state *csp)
3663 const unsigned int flags = csp->flags;
3664 const char *response_status_line = csp->headers->first->str;
3665 static const char connection_close[] = "Connection: close";
3667 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3668 && (flags & CSP_FLAG_SERVER_CONNECTION_HEADER_SET))
3674 * XXX: if we downgraded the response, this check will fail.
3676 if ((csp->config->feature_flags &
3677 RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3678 && (NULL != response_status_line)
3679 && !strncmpic(response_status_line, "HTTP/1.1", 8)
3680 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3681 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3685 log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
3686 "without Connection header implies keep-alive.");
3687 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
3691 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3693 return enlist(csp->headers, connection_close);
3697 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3698 /*********************************************************************
3700 * Function : server_proxy_connection_adder
3702 * Description : Adds a "Proxy-Connection: keep-alive" header to
3703 * csp->headers when appropriate.
3706 * 1 : csp = Current client state (buffers, headers, etc...)
3708 * Returns : JB_ERR_OK on success, or
3709 * JB_ERR_MEMORY on out-of-memory error.
3711 *********************************************************************/
3712 static jb_err server_proxy_connection_adder(struct client_state *csp)
3714 static const char proxy_connection_header[] = "Proxy-Connection: keep-alive";
3715 jb_err err = JB_ERR_OK;
3717 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
3718 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3719 && !(csp->flags & CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET)
3720 && ((csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET)
3721 || (csp->flags & CSP_FLAG_CHUNKED)))
3723 log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
3724 err = enlist(csp->headers, proxy_connection_header);
3729 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3732 /*********************************************************************
3734 * Function : client_connection_header_adder
3736 * Description : Adds a proper "Connection:" header to csp->headers
3737 * unless the header was already present. Called from `sed'.
3740 * 1 : csp = Current client state (buffers, headers, etc...)
3742 * Returns : JB_ERR_OK on success, or
3743 * JB_ERR_MEMORY on out-of-memory error.
3745 *********************************************************************/
3746 static jb_err client_connection_header_adder(struct client_state *csp)
3748 static const char connection_close[] = "Connection: close";
3750 if (!(csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3751 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
3756 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3757 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3758 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3759 && (csp->http->ssl == 0)
3760 && !strcmpic(csp->http->ver, "HTTP/1.1"))
3762 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
3765 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3767 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3769 return enlist(csp->headers, connection_close);
3773 /*********************************************************************
3775 * Function : server_http
3777 * Description : - Save the HTTP Status into csp->http->status
3778 * - Set CT_TABOO to prevent filtering if the answer
3779 * is a partial range (HTTP status 206)
3780 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3784 * 1 : csp = Current client state (buffers, headers, etc...)
3785 * 2 : header = On input, pointer to header to modify.
3786 * On output, pointer to the modified header, or NULL
3787 * to remove the header. This function frees the
3788 * original string if necessary.
3790 * Returns : JB_ERR_OK on success, or
3791 * JB_ERR_MEMORY on out-of-memory error.
3793 *********************************************************************/
3794 static jb_err server_http(struct client_state *csp, char **header)
3796 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3797 if (csp->http->status == 206)
3799 csp->content_type = CT_TABOO;
3802 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3804 /* XXX: Should we do a real validity check here? */
3805 if (strlen(*header) > 8)
3808 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3813 * XXX: Should we block the request or
3814 * enlist a valid status code line here?
3816 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3817 "Downgrading to HTTP/1.0 impossible.");
3824 /*********************************************************************
3826 * Function : add_cooky_expiry_date
3828 * Description : Adds a cookie expiry date to a string.
3831 * 1 : cookie = On input, pointer to cookie to modify.
3832 * On output, pointer to the modified header.
3833 * The original string is freed.
3834 * 2 : lifetime = Seconds the cookie should be valid
3838 *********************************************************************/
3839 static void add_cookie_expiry_date(char **cookie, time_t lifetime)
3842 struct tm *timeptr = NULL;
3843 time_t expiry_date = time(NULL) + lifetime;
3844 #ifdef HAVE_GMTIME_R
3847 timeptr = gmtime_r(&expiry_date, &gmt);
3848 #elif defined(MUTEX_LOCKS_AVAILABLE)
3849 privoxy_mutex_lock(&gmtime_mutex);
3850 timeptr = gmtime(&expiry_date);
3851 privoxy_mutex_unlock(&gmtime_mutex);
3853 timeptr = gmtime(&expiry_date);
3856 if (NULL == timeptr)
3858 log_error(LOG_LEVEL_FATAL,
3859 "Failed to get the time in add_cooky_expiry_date()");
3861 strftime(tmp, sizeof(tmp), "; expires=%a, %d-%b-%Y %H:%M:%S GMT", timeptr);
3862 if (JB_ERR_OK != string_append(cookie, tmp))
3864 log_error(LOG_LEVEL_FATAL, "Out of memory in add_cooky_expiry()");
3869 /*********************************************************************
3871 * Function : server_set_cookie
3873 * Description : Handle the server "cookie" header properly.
3874 * Crunch, accept or rewrite it to a session cookie.
3875 * Called from `sed'.
3878 * 1 : csp = Current client state (buffers, headers, etc...)
3879 * 2 : header = On input, pointer to header to modify.
3880 * On output, pointer to the modified header, or NULL
3881 * to remove the header. This function frees the
3882 * original string if necessary.
3884 * Returns : JB_ERR_OK on success, or
3885 * JB_ERR_MEMORY on out-of-memory error.
3887 *********************************************************************/
3888 static jb_err server_set_cookie(struct client_state *csp, char **header)
3890 if ((csp->action->flags & ACTION_CRUNCH_INCOMING_COOKIES) != 0)
3892 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3895 else if ((0 != (csp->action->flags & ACTION_SESSION_COOKIES_ONLY))
3896 || (0 != (csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME)))
3900 long cookie_lifetime = 0;
3903 NO_EXPIRY_DATE_SPECIFIED,
3904 EXPIRY_DATE_ACCEPTABLE,
3905 EXPIRY_DATE_UNACCEPTABLE
3906 } expiry_date_status = NO_EXPIRY_DATE_SPECIFIED;
3908 /* A variable to store the tag we're working on */
3911 /* Skip "Set-Cookie:" (11 characters) in header */
3912 cur_tag = *header + 11;
3914 /* skip whitespace between "Set-Cookie:" and value */
3915 while (*cur_tag && privoxy_isspace(*cur_tag))
3922 if ((csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME) != 0)
3924 const char *param = csp->action->string[ACTION_STRING_LIMIT_COOKIE_LIFETIME];
3926 cookie_lifetime = strtol(param, NULL, 0);
3927 if (cookie_lifetime < 0)
3929 log_error(LOG_LEVEL_FATAL, "Invalid cookie lifetime limit: %s", param);
3931 cookie_lifetime *= 60;
3934 /* Loop through each tag in the cookie */
3938 char *next_tag = strchr(cur_tag, ';');
3939 if (next_tag != NULL)
3941 /* Skip the ';' character itself */
3944 /* skip whitespace ";" and start of tag */
3945 while (*next_tag && privoxy_isspace(*next_tag))
3952 /* "Next tag" is the end of the string */
3953 next_tag = cur_tag + strlen(cur_tag);
3957 * Check the expiration date to see
3958 * if the cookie is still valid, if yes,
3959 * rewrite it to a session cookie.
3961 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
3963 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
3965 if ((expiration_date[0] == '"')
3966 && (expiration_date[1] != '\0'))
3969 * Skip quotation mark. RFC 2109 10.1.2 seems to hint
3970 * that the expiration date isn't supposed to be quoted,
3971 * but some servers do it anyway.
3976 /* Did we detect the date properly? */
3977 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
3980 * Nope, treat it as if it was still valid.
3982 * XXX: Should we remove the whole cookie instead?
3984 log_error(LOG_LEVEL_ERROR,
3985 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
3986 string_move(cur_tag, next_tag);
3987 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
3992 * Yes. Check if the cookie is still valid.
3994 * If the cookie is already expired it's probably
3995 * a delete cookie and even if it isn't, the browser
3996 * will discard it anyway.
4000 * XXX: timegm() isn't available on some AmigaOS
4001 * versions and our replacement doesn't work.
4003 * Our options are to either:
4005 * - disable session-cookies-only completely if timegm
4008 * - to simply remove all expired tags, like it has
4009 * been done until Privoxy 3.0.6 and to live with
4010 * the consequence that it can cause login/logout
4011 * problems on servers that don't validate their
4012 * input properly, or
4014 * - to replace it with mktime in which
4015 * case there is a slight chance of valid cookies
4016 * passing as already expired.
4018 * This is the way it's currently done and it's not
4019 * as bad as it sounds. If the missing GMT offset is
4020 * enough to change the result of the expiration check
4021 * the cookie will be only valid for a few hours
4022 * anyway, which in many cases will be shorter
4023 * than a browser session.
4025 if (cookie_time < now)
4027 log_error(LOG_LEVEL_HEADER,
4028 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4029 /* Just in case some clown sets more then one expiration date */
4031 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4033 else if ((cookie_lifetime != 0) && (cookie_time < (now + cookie_lifetime)))
4035 log_error(LOG_LEVEL_HEADER, "Cookie \'%s\' can pass unmodified. "
4036 "Its lifetime is below the limit.", *header);
4037 /* Just in case some clown sets more then one expiration date */
4039 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4044 * Still valid, delete expiration date by copying
4045 * the rest of the string over it.
4047 string_move(cur_tag, next_tag);
4049 /* That changed the header, need to issue a log message */
4050 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4053 * Note that the next tag has now been moved to *cur_tag,
4054 * so we do not need to update the cur_tag pointer.
4062 /* Move on to next cookie tag */
4067 if (expiry_date_status != EXPIRY_DATE_ACCEPTABLE)
4069 assert(NULL != *header);
4070 if (cookie_lifetime != 0)
4072 add_cookie_expiry_date(header, cookie_lifetime);
4073 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to: %s", *header);
4075 else if (expiry_date_status != NO_EXPIRY_DATE_SPECIFIED)
4077 log_error(LOG_LEVEL_HEADER,
4078 "Cookie rewritten to a temporary one: %s", *header);
4087 #ifdef FEATURE_FORCE_LOAD
4088 /*********************************************************************
4090 * Function : strclean
4092 * Description : In-Situ-Eliminate all occurrences of substring in
4096 * 1 : string = string to clean
4097 * 2 : substring = substring to eliminate
4099 * Returns : Number of eliminations
4101 *********************************************************************/
4102 int strclean(char *string, const char *substring)
4108 len = strlen(substring);
4110 while((pos = strstr(string, substring)) != NULL)
4117 while (*p++ != '\0');
4124 #endif /* def FEATURE_FORCE_LOAD */
4127 /*********************************************************************
4129 * Function : parse_header_time
4131 * Description : Parses time formats used in HTTP header strings
4132 * to get the numerical respresentation.
4135 * 1 : header_time = HTTP header time as string.
4136 * 2 : result = storage for header_time in seconds
4138 * Returns : JB_ERR_OK if the time format was recognized, or
4139 * JB_ERR_PARSE otherwise.
4141 *********************************************************************/
4142 static jb_err parse_header_time(const char *header_time, time_t *result)
4146 * Checking for two-digit years first in an
4147 * attempt to work around GNU libc's strptime()
4148 * reporting negative year values when using %Y.
4150 static const char time_formats[][22] = {
4151 /* Tue, 02-Jun-37 20:00:00 */
4152 "%a, %d-%b-%y %H:%M:%S",
4153 /* Tue, 02 Jun 2037 20:00:00 */
4154 "%a, %d %b %Y %H:%M:%S",
4155 /* Tue, 02-Jun-2037 20:00:00 */
4156 "%a, %d-%b-%Y %H:%M:%S",
4157 /* Tuesday, 02-Jun-2037 20:00:00 */
4158 "%A, %d-%b-%Y %H:%M:%S",
4159 /* Tuesday Jun 02 20:00:00 2037 */
4160 "%A %b %d %H:%M:%S %Y"
4164 for (i = 0; i < SZ(time_formats); i++)
4167 * Zero out gmt to prevent time zone offsets.
4168 * Documented to be required for GNU libc.
4170 memset(&gmt, 0, sizeof(gmt));
4172 if (NULL != strptime(header_time, time_formats[i], &gmt))
4174 /* Sanity check for GNU libc. */
4175 if (gmt.tm_year < 0)
4177 log_error(LOG_LEVEL_HEADER,
4178 "Failed to parse '%s' using '%s'. Moving on.",
4179 header_time, time_formats[i]);
4182 *result = timegm(&gmt);
4184 #ifdef FEATURE_STRPTIME_SANITY_CHECKS
4186 * Verify that parsing the date recreated from the first
4187 * parse operation gets the previous result. If it doesn't,
4188 * either strptime() or strftime() are malfunctioning.
4190 * We could string-compare the recreated date with the original
4191 * header date, but this leads to false positives as strptime()
4192 * may let %a accept all day formats while strftime() will only
4196 char recreated_date[100];
4200 tm = gmtime(result);
4201 strftime(recreated_date, sizeof(recreated_date), time_formats[i], tm);
4202 memset(&gmt, 0, sizeof(gmt));
4203 if (NULL == strptime(recreated_date, time_formats[i], &gmt))
4205 log_error(LOG_LEVEL_ERROR,
4206 "Failed to parse '%s' generated with '%s' to recreate '%s'.",
4207 recreated_date, time_formats[i], header_time);
4210 result2 = timegm(&gmt);
4211 if (*result != result2)
4213 log_error(LOG_LEVEL_ERROR, "strftime() and strptime() disagree. "
4214 "Format: '%s'. In: '%s', out: '%s'. %d != %d. Rejecting.",
4215 time_formats[i], header_time, recreated_date, *result, result2);
4225 return JB_ERR_PARSE;
4230 /*********************************************************************
4232 * Function : get_destination_from_headers
4234 * Description : Parse the "Host:" header to get the request's destination.
4235 * Only needed if the client's request was forcefully
4236 * redirected into Privoxy.
4238 * Code mainly copied from client_host() which is currently
4239 * run too late for this purpose.
4242 * 1 : headers = List of headers (one of them hopefully being
4243 * the "Host:" header)
4244 * 2 : http = storage for the result (host, port and hostport).
4246 * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
4247 * JB_ERR_PARSE if the host header couldn't be found,
4248 * JB_ERR_OK otherwise.
4250 *********************************************************************/
4251 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4257 host = get_header_value(headers, "Host:");
4261 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4262 return JB_ERR_PARSE;
4265 p = strdup_or_die(host);
4267 q = strdup_or_die(p);
4269 freez(http->hostport);
4273 q = strchr(http->host, ':');
4276 /* Terminate hostname and evaluate port string */
4278 http->port = atoi(q);
4282 http->port = http->ssl ? 443 : 80;
4285 /* Rebuild request URL */
4287 http->url = strdup(http->ssl ? "https://" : "http://");
4288 string_append(&http->url, http->hostport);
4289 string_append(&http->url, http->path);
4290 if (http->url == NULL)
4292 return JB_ERR_MEMORY;
4295 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4303 /*********************************************************************
4305 * Function : create_forged_referrer
4307 * Description : Helper for client_referrer to forge a referer as
4308 * 'http://hostname[:port]/' to fool stupid
4309 * checks for in-site links
4312 * 1 : header = Pointer to header pointer
4313 * 2 : hostport = Host and optionally port as string
4315 * Returns : JB_ERR_OK in case of success, or
4316 * JB_ERR_MEMORY in case of memory problems.
4318 *********************************************************************/
4319 static jb_err create_forged_referrer(char **header, const char *hostport)
4321 assert(NULL == *header);
4323 *header = strdup("Referer: http://");
4324 string_append(header, hostport);
4325 string_append(header, "/");
4327 if (NULL == *header)
4329 return JB_ERR_MEMORY;
4332 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4339 /*********************************************************************
4341 * Function : create_fake_referrer
4343 * Description : Helper for client_referrer to create a fake referrer
4344 * based on a string supplied by the user.
4347 * 1 : header = Pointer to header pointer
4348 * 2 : hosthost = Referrer to fake
4350 * Returns : JB_ERR_OK in case of success, or
4351 * JB_ERR_MEMORY in case of memory problems.
4353 *********************************************************************/
4354 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4356 assert(NULL == *header);
4358 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4360 log_error(LOG_LEVEL_HEADER,
4361 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4363 *header = strdup("Referer: ");
4364 string_append(header, fake_referrer);
4366 if (NULL == *header)
4368 return JB_ERR_MEMORY;
4371 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4378 /*********************************************************************
4380 * Function : handle_conditional_hide_referrer_parameter
4382 * Description : Helper for client_referrer to crunch or forge
4383 * the referrer header if the host has changed.
4386 * 1 : header = Pointer to header pointer
4387 * 2 : host = The target host (may include the port)
4388 * 3 : parameter_conditional_block = Boolean to signal
4389 * if we're in conditional-block mode. If not set,
4390 * we're in conditional-forge mode.
4392 * Returns : JB_ERR_OK in case of success, or
4393 * JB_ERR_MEMORY in case of memory problems.
4395 *********************************************************************/
4396 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4397 const char *host, const int parameter_conditional_block)
4399 char *referer = strdup_or_die(*header);
4400 const size_t hostlength = strlen(host);
4401 const char *referer_url = NULL;
4403 /* referer begins with 'Referer: http[s]://' */
4404 if ((hostlength+17) < strlen(referer))
4407 * Shorten referer to make sure the referer is blocked