1 const char parsers_rcs[] = "$Id: parsers.c,v 1.279 2013/08/06 12:59:34 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;
1285 struct file_list *fl;
1286 struct re_filterfile_spec *b;
1287 struct list_entry *tag_name;
1289 const size_t header_length = strlen(header);
1291 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1293 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1294 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1298 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1299 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1302 if (list_is_empty(csp->action->multi[multi_action_index])
1303 || filters_available(csp) == FALSE)
1305 /* Return early if no taggers apply or if none are available. */
1309 for (i = 0; i < MAX_AF_FILES; i++)
1312 if ((NULL == fl) || (NULL == fl->f))
1315 * Either there are no filter files
1316 * left, or this filter file just
1317 * contains no valid filters.
1319 * Continue to be sure we don't miss
1320 * valid filter files that are chained
1321 * after empty or invalid ones.
1326 /* For all filters, */
1327 for (b = fl->f; b; b = b->next)
1329 if (b->type != wanted_filter_type)
1331 /* skip the ones we don't care about, */
1334 /* leaving only taggers that could apply, of which we use the ones, */
1335 for (tag_name = csp->action->multi[multi_action_index]->first;
1336 NULL != tag_name; tag_name = tag_name->next)
1338 /* that do apply, and */
1339 if (strcmp(b->name, tag_name->str) == 0)
1341 char *modified_tag = NULL;
1343 size_t size = header_length;
1344 pcrs_job *joblist = b->joblist;
1346 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1348 if (NULL == joblist)
1350 log_error(LOG_LEVEL_RE_FILTER,
1351 "Tagger %s has empty joblist. Nothing to do.", b->name);
1355 /* execute their pcrs_joblist on the header. */
1356 for (job = joblist; NULL != job; job = job->next)
1358 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1362 /* Success, continue with the modified version. */
1371 /* Tagger doesn't match */
1374 /* Regex failure, log it but continue anyway. */
1375 assert(NULL != header);
1376 log_error(LOG_LEVEL_ERROR,
1377 "Problems with tagger \'%s\' and header \'%s\': %s",
1378 b->name, *header, pcrs_strerror(hits));
1380 freez(modified_tag);
1384 if (b->dynamic) pcrs_free_joblist(joblist);
1386 /* If this tagger matched */
1392 * There is no technical limitation which makes
1393 * it impossible to use empty tags, but I assume
1394 * no one would do it intentionally.
1397 log_error(LOG_LEVEL_INFO,
1398 "Tagger \'%s\' created an empty tag. Ignored.",
1403 if (!list_contains_item(csp->tags, tag))
1405 if (JB_ERR_OK != enlist(csp->tags, tag))
1407 log_error(LOG_LEVEL_ERROR,
1408 "Insufficient memory to add tag \'%s\', "
1409 "based on tagger \'%s\' and header \'%s\'",
1410 tag, b->name, *header);
1414 char *action_message;
1416 * update the action bits right away, to make
1417 * tagging based on tags set by earlier taggers
1418 * of the same kind possible.
1420 if (update_action_bits_for_tag(csp, tag))
1422 action_message = "Action bits updated accordingly.";
1426 action_message = "No action bits update necessary.";
1429 log_error(LOG_LEVEL_HEADER,
1430 "Tagger \'%s\' added tag \'%s\'. %s",
1431 b->name, tag, action_message);
1436 /* XXX: Is this log-worthy? */
1437 log_error(LOG_LEVEL_HEADER,
1438 "Tagger \'%s\' didn't add tag \'%s\'. "
1439 "Tag already present", b->name, tag);
1442 } /* if the tagger matched */
1443 } /* if the tagger applies */
1444 } /* for every tagger that could apply */
1445 } /* for all filters */
1446 } /* for all filter files */
1451 /* here begins the family of parser functions that reformat header lines */
1453 /*********************************************************************
1455 * Function : filter_header
1457 * Description : Executes all text substitutions from all applying
1458 * +(server|client)-header-filter actions on the header.
1459 * Most of the code was copied from pcrs_filter_response,
1460 * including the rather short variable names
1463 * 1 : csp = Current client state (buffers, headers, etc...)
1464 * 2 : header = On input, pointer to header to modify.
1465 * On output, pointer to the modified header, or NULL
1466 * to remove the header. This function frees the
1467 * original string if necessary.
1469 * Returns : JB_ERR_OK on success and always succeeds
1471 *********************************************************************/
1472 static jb_err filter_header(struct client_state *csp, char **header)
1476 size_t size = strlen(*header);
1478 char *newheader = NULL;
1481 struct file_list *fl;
1482 struct re_filterfile_spec *b;
1483 struct list_entry *filtername;
1486 int wanted_filter_type;
1487 int multi_action_index;
1489 if (csp->flags & CSP_FLAG_NO_FILTERING)
1494 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1496 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1497 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1501 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1502 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1505 if (list_is_empty(csp->action->multi[multi_action_index])
1506 || filters_available(csp) == FALSE)
1508 /* Return early if no filters apply or if none are available. */
1512 for (i = 0; i < MAX_AF_FILES; i++)
1515 if ((NULL == fl) || (NULL == fl->f))
1518 * Either there are no filter files
1519 * left, or this filter file just
1520 * contains no valid filters.
1522 * Continue to be sure we don't miss
1523 * valid filter files that are chained
1524 * after empty or invalid ones.
1529 * For all applying +filter actions, look if a filter by that
1530 * name exists and if yes, execute its pcrs_joblist on the
1533 for (b = fl->f; b; b = b->next)
1535 if (b->type != wanted_filter_type)
1537 /* Skip other filter types */
1541 for (filtername = csp->action->multi[multi_action_index]->first;
1542 filtername ; filtername = filtername->next)
1544 if (strcmp(b->name, filtername->str) == 0)
1546 int current_hits = 0;
1547 pcrs_job *joblist = b->joblist;
1549 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1551 if (NULL == joblist)
1553 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1557 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1558 *header, size, b->name);
1560 /* Apply all jobs from the joblist */
1561 for (job = joblist; NULL != job; job = job->next)
1563 matches = pcrs_execute(job, *header, size, &newheader, &size);
1566 current_hits += matches;
1567 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1569 *header = newheader;
1571 else if (0 == matches)
1573 /* Filter doesn't change header */
1579 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1580 *header, b->name, pcrs_strerror(matches));
1581 if (newheader != NULL)
1583 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1589 if (b->dynamic) pcrs_free_joblist(joblist);
1591 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1592 hits += current_hits;
1599 * Additionally checking for hits is important because if
1600 * the continue hack is triggered, server headers can
1601 * arrive empty to separate multiple heads from each other.
1603 if ((0 == size) && hits)
1605 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
1613 /*********************************************************************
1615 * Function : server_connection
1617 * Description : Makes sure a proper "Connection:" header is
1618 * set and signals connection_header_adder to
1622 * 1 : csp = Current client state (buffers, headers, etc...)
1623 * 2 : header = On input, pointer to header to modify.
1624 * On output, pointer to the modified header, or NULL
1625 * to remove the header. This function frees the
1626 * original string if necessary.
1628 * Returns : JB_ERR_OK on success.
1630 *********************************************************************/
1631 static jb_err server_connection(struct client_state *csp, char **header)
1633 if (!strcmpic(*header, "Connection: keep-alive")
1634 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1635 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
1639 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1640 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1642 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
1645 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE))
1647 log_error(LOG_LEVEL_HEADER,
1648 "Keeping the server header '%s' around.", *header);
1651 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
1653 char *old_header = *header;
1655 *header = strdup_or_die("Connection: close");
1656 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
1661 /* Signal server_connection_adder() to return early. */
1662 csp->flags |= CSP_FLAG_SERVER_CONNECTION_HEADER_SET;
1668 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1669 /*********************************************************************
1671 * Function : server_keep_alive
1673 * Description : Stores the server's keep alive timeout.
1676 * 1 : csp = Current client state (buffers, headers, etc...)
1677 * 2 : header = On input, pointer to header to modify.
1678 * On output, pointer to the modified header, or NULL
1679 * to remove the header. This function frees the
1680 * original string if necessary.
1682 * Returns : JB_ERR_OK.
1684 *********************************************************************/
1685 static jb_err server_keep_alive(struct client_state *csp, char **header)
1687 unsigned int keep_alive_timeout;
1688 const char *timeout_position = strstr(*header, "timeout=");
1690 if ((NULL == timeout_position)
1691 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1693 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1697 if (keep_alive_timeout < csp->server_connection.keep_alive_timeout)
1699 log_error(LOG_LEVEL_HEADER,
1700 "Reducing keep-alive timeout from %u to %u.",
1701 csp->server_connection.keep_alive_timeout, keep_alive_timeout);
1702 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1706 /* XXX: Is this log worthy? */
1707 log_error(LOG_LEVEL_HEADER,
1708 "Server keep-alive timeout is %u. Sticking with %u.",
1709 keep_alive_timeout, csp->server_connection.keep_alive_timeout);
1711 csp->flags |= CSP_FLAG_SERVER_KEEP_ALIVE_TIMEOUT_SET;
1718 /*********************************************************************
1720 * Function : server_proxy_connection
1722 * Description : Figures out whether or not we should add a
1723 * Proxy-Connection header.
1726 * 1 : csp = Current client state (buffers, headers, etc...)
1727 * 2 : header = On input, pointer to header to modify.
1728 * On output, pointer to the modified header, or NULL
1729 * to remove the header. This function frees the
1730 * original string if necessary.
1732 * Returns : JB_ERR_OK.
1734 *********************************************************************/
1735 static jb_err server_proxy_connection(struct client_state *csp, char **header)
1737 csp->flags |= CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET;
1742 /*********************************************************************
1744 * Function : proxy_authentication
1746 * Description : Removes headers that are relevant for proxy
1747 * authentication unless forwarding them has
1748 * been explicitly requested.
1751 * 1 : csp = Current client state (buffers, headers, etc...)
1752 * 2 : header = On input, pointer to header to modify.
1753 * On output, pointer to the modified header, or NULL
1754 * to remove the header. This function frees the
1755 * original string if necessary.
1757 * Returns : JB_ERR_OK.
1759 *********************************************************************/
1760 static jb_err proxy_authentication(struct client_state *csp, char **header)
1762 if ((csp->config->feature_flags &
1763 RUNTIME_FEATURE_FORWARD_PROXY_AUTHENTICATION_HEADERS) == 0) {
1764 log_error(LOG_LEVEL_HEADER,
1765 "Forwarding proxy authentication headers is disabled. Crunching: %s", *header);
1772 /*********************************************************************
1774 * Function : client_keep_alive
1776 * Description : Stores the client's keep alive timeout.
1779 * 1 : csp = Current client state (buffers, headers, etc...)
1780 * 2 : header = On input, pointer to header to modify.
1781 * On output, pointer to the modified header, or NULL
1782 * to remove the header. This function frees the
1783 * original string if necessary.
1785 * Returns : JB_ERR_OK.
1787 *********************************************************************/
1788 static jb_err client_keep_alive(struct client_state *csp, char **header)
1790 unsigned int keep_alive_timeout;
1791 const char *timeout_position = strstr(*header, ": ");
1793 if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1795 log_error(LOG_LEVEL_HEADER,
1796 "keep-alive support is disabled. Crunching: %s.", *header);
1801 if ((NULL == timeout_position)
1802 || (1 != sscanf(timeout_position, ": %u", &keep_alive_timeout)))
1804 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1808 if (keep_alive_timeout < csp->config->keep_alive_timeout)
1810 log_error(LOG_LEVEL_HEADER,
1811 "Reducing keep-alive timeout from %u to %u.",
1812 csp->config->keep_alive_timeout, keep_alive_timeout);
1813 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1817 /* XXX: Is this log worthy? */
1818 log_error(LOG_LEVEL_HEADER,
1819 "Client keep-alive timeout is %u. Sticking with %u.",
1820 keep_alive_timeout, csp->config->keep_alive_timeout);
1828 /*********************************************************************
1830 * Function : get_content_length
1832 * Description : Gets the content length specified in a
1833 * Content-Length header.
1836 * 1 : header_value = The Content-Length header value.
1837 * 2 : length = Storage to return the value.
1839 * Returns : JB_ERR_OK on success, or
1840 * JB_ERR_PARSE if no value is recognized.
1842 *********************************************************************/
1843 static jb_err get_content_length(const char *header_value, unsigned long long *length)
1846 assert(sizeof(unsigned long long) > 4);
1847 if (1 != sscanf(header_value, "%I64u", length))
1849 if (1 != sscanf(header_value, "%llu", length))
1852 return JB_ERR_PARSE;
1859 /*********************************************************************
1861 * Function : client_save_content_length
1863 * Description : Save the Content-Length sent by the client.
1866 * 1 : csp = Current client state (buffers, headers, etc...)
1867 * 2 : header = On input, pointer to header to modify.
1868 * On output, pointer to the modified header, or NULL
1869 * to remove the header. This function frees the
1870 * original string if necessary.
1872 * Returns : JB_ERR_OK on success, or
1873 * JB_ERR_MEMORY on out-of-memory error.
1875 *********************************************************************/
1876 static jb_err client_save_content_length(struct client_state *csp, char **header)
1878 unsigned long long content_length = 0;
1879 const char *header_value;
1881 assert(*(*header+14) == ':');
1883 header_value = *header + 15;
1884 if (JB_ERR_OK != get_content_length(header_value, &content_length))
1886 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
1891 csp->expected_client_content_length = content_length;
1896 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1900 /*********************************************************************
1902 * Function : client_connection
1904 * Description : Makes sure a proper "Connection:" header is
1905 * set and signals connection_header_adder
1909 * 1 : csp = Current client state (buffers, headers, etc...)
1910 * 2 : header = On input, pointer to header to modify.
1911 * On output, pointer to the modified header, or NULL
1912 * to remove the header. This function frees the
1913 * original string if necessary.
1915 * Returns : JB_ERR_OK on success.
1917 *********************************************************************/
1918 static jb_err client_connection(struct client_state *csp, char **header)
1920 static const char connection_close[] = "Connection: close";
1922 if (!strcmpic(*header, connection_close))
1924 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1925 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING)
1926 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1928 if (!strcmpic(csp->http->ver, "HTTP/1.1"))
1930 log_error(LOG_LEVEL_HEADER,
1931 "Removing \'%s\' to imply keep-alive.", *header);
1934 * While we imply keep-alive to the server,
1935 * we have to remember that the client didn't.
1937 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1941 char *old_header = *header;
1943 *header = strdup_or_die("Connection: keep-alive");
1944 log_error(LOG_LEVEL_HEADER,
1945 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1951 log_error(LOG_LEVEL_HEADER,
1952 "Keeping the client header '%s' around. "
1953 "The connection will not be kept alive.",
1955 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1958 else if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
1959 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1961 log_error(LOG_LEVEL_HEADER,
1962 "Keeping the client header '%s' around. "
1963 "The server connection will be kept alive if possible.",
1965 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1966 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1970 char *old_header = *header;
1972 *header = strdup_or_die(connection_close);
1973 log_error(LOG_LEVEL_HEADER,
1974 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1978 /* Signal client_connection_header_adder() to return early. */
1979 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
1985 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1986 /*********************************************************************
1988 * Function : client_proxy_connection
1990 * Description : Sets the CLIENT_CONNECTION_KEEP_ALIVE flag when
1991 * appropriate and removes the Proxy-Connection
1995 * 1 : csp = Current client state (buffers, headers, etc...)
1996 * 2 : header = On input, pointer to header to modify.
1997 * On output, pointer to the modified header, or NULL
1998 * to remove the header. This function frees the
1999 * original string if necessary.
2001 * Returns : JB_ERR_OK
2003 *********************************************************************/
2004 static jb_err client_proxy_connection(struct client_state *csp, char **header)
2006 if (0 == (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
2007 && (csp->http->ssl == 0)
2008 && (NULL == strstr(*header, "close")))
2010 log_error(LOG_LEVEL_HEADER,
2011 "The client connection can be kept alive due to: %s", *header);
2012 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
2014 crumble(csp, header);
2018 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2021 /*********************************************************************
2023 * Function : client_transfer_encoding
2025 * Description : Raise the CSP_FLAG_CHUNKED_CLIENT_BODY flag if
2026 * the request body is "chunked"
2028 * XXX: Currently not called through sed() as we
2029 * need the flag earlier on. Should be fixed.
2032 * 1 : csp = Current client state (buffers, headers, etc...)
2033 * 2 : header = On input, pointer to header to modify.
2034 * On output, pointer to the modified header, or NULL
2035 * to remove the header. This function frees the
2036 * original string if necessary.
2038 * Returns : JB_ERR_OK on success, or
2040 *********************************************************************/
2041 jb_err client_transfer_encoding(struct client_state *csp, char **header)
2043 if (strstr(*header, "chunked"))
2045 csp->flags |= CSP_FLAG_CHUNKED_CLIENT_BODY;
2046 log_error(LOG_LEVEL_HEADER, "Expecting chunked client body");
2053 /*********************************************************************
2055 * Function : crumble
2057 * Description : This is called if a header matches a pattern to "crunch"
2060 * 1 : csp = Current client state (buffers, headers, etc...)
2061 * 2 : header = On input, pointer to header to modify.
2062 * On output, pointer to the modified header, or NULL
2063 * to remove the header. This function frees the
2064 * original string if necessary.
2066 * Returns : JB_ERR_OK on success, or
2067 * JB_ERR_MEMORY on out-of-memory error.
2069 *********************************************************************/
2070 static jb_err crumble(struct client_state *csp, char **header)
2073 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2079 /*********************************************************************
2081 * Function : crunch_server_header
2083 * Description : Crunch server header if it matches a string supplied by the
2084 * user. Called from `sed'.
2087 * 1 : csp = Current client state (buffers, headers, etc...)
2088 * 2 : header = On input, pointer to header to modify.
2089 * On output, pointer to the modified header, or NULL
2090 * to remove the header. This function frees the
2091 * original string if necessary.
2093 * Returns : JB_ERR_OK on success and always succeeds
2095 *********************************************************************/
2096 static jb_err crunch_server_header(struct client_state *csp, char **header)
2098 const char *crunch_pattern;
2100 /* Do we feel like crunching? */
2101 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2103 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2105 /* Is the current header the lucky one? */
2106 if (strstr(*header, crunch_pattern))
2108 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2117 /*********************************************************************
2119 * Function : server_content_type
2121 * Description : Set the content-type for filterable types (text/.*,
2122 * .*xml.*, .*script.* and image/gif) unless filtering has been
2123 * forbidden (CT_TABOO) while parsing earlier headers.
2124 * NOTE: Since text/plain is commonly used by web servers
2125 * for files whose correct type is unknown, we don't
2126 * set CT_TEXT for it.
2129 * 1 : csp = Current client state (buffers, headers, etc...)
2130 * 2 : header = On input, pointer to header to modify.
2131 * On output, pointer to the modified header, or NULL
2132 * to remove the header. This function frees the
2133 * original string if necessary.
2135 * Returns : JB_ERR_OK on success, or
2136 * JB_ERR_MEMORY on out-of-memory error.
2138 *********************************************************************/
2139 static jb_err server_content_type(struct client_state *csp, char **header)
2141 /* Remove header if it isn't the first Content-Type header */
2142 if ((csp->content_type & CT_DECLARED))
2144 if (content_filters_enabled(csp->action))
2147 * Making sure the client interprets the content the same way
2148 * Privoxy did is only relevant if Privoxy modified it.
2150 * Checking for this is "hard" as it's not yet known when
2151 * this function is called, thus go shopping and and just
2152 * check if Privoxy could filter it.
2154 * The main thing is that we don't mess with the headers
2155 * unless the user signalled that it's acceptable.
2157 log_error(LOG_LEVEL_HEADER,
2158 "Multiple Content-Type headers detected. "
2159 "Removing and ignoring: %s",
2167 * Signal that the Content-Type has been set.
2169 csp->content_type |= CT_DECLARED;
2171 if (!(csp->content_type & CT_TABOO))
2174 * XXX: The assumption that text/plain is a sign of
2175 * binary data seems to be somewhat unreasonable nowadays
2176 * and should be dropped after 3.0.8 is out.
2178 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2179 || strstr(*header, "xml")
2180 || strstr(*header, "script"))
2182 csp->content_type |= CT_TEXT;
2184 else if (strstr(*header, "image/gif"))
2186 csp->content_type |= CT_GIF;
2191 * Are we messing with the content type?
2193 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2196 * Make sure the user doesn't accidentally
2197 * change the content type of binary documents.
2199 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2202 *header = strdup_or_die("Content-Type: ");
2203 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2207 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2208 return JB_ERR_MEMORY;
2210 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2214 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2215 "It doesn't look like a content type that should be filtered. "
2216 "Enable force-text-mode if you know what you're doing.", *header);
2224 /*********************************************************************
2226 * Function : server_transfer_coding
2228 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2229 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2230 * - Remove header if body was chunked but has been
2231 * de-chunked for filtering.
2234 * 1 : csp = Current client state (buffers, headers, etc...)
2235 * 2 : header = On input, pointer to header to modify.
2236 * On output, pointer to the modified header, or NULL
2237 * to remove the header. This function frees the
2238 * original string if necessary.
2240 * Returns : JB_ERR_OK on success, or
2241 * JB_ERR_MEMORY on out-of-memory error.
2243 *********************************************************************/
2244 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2247 * Turn off pcrs and gif filtering if body compressed
2249 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2253 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2255 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2256 csp->http->cmd, *header);
2257 #endif /* def FEATURE_ZLIB */
2258 csp->content_type = CT_TABOO;
2262 * Raise flag if body chunked
2264 if (strstr(*header, "chunked"))
2266 csp->flags |= CSP_FLAG_CHUNKED;
2269 * If the body was modified, it has been de-chunked first
2270 * and the header must be removed.
2272 * FIXME: If there is more than one transfer encoding,
2273 * only the "chunked" part should be removed here.
2275 if (csp->flags & CSP_FLAG_MODIFIED)
2277 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2286 /*********************************************************************
2288 * Function : server_content_encoding
2290 * Description : Used to check if the content is compressed, and if
2291 * FEATURE_ZLIB is disabled, filtering is disabled as
2294 * If FEATURE_ZLIB is enabled and the compression type
2295 * supported, the content is marked for decompression.
2297 * XXX: Doesn't properly deal with multiple or with
2298 * unsupported but unknown encodings.
2299 * Is case-sensitive but shouldn't be.
2302 * 1 : csp = Current client state (buffers, headers, etc...)
2303 * 2 : header = On input, pointer to header to modify.
2304 * On output, pointer to the modified header, or NULL
2305 * to remove the header. This function frees the
2306 * original string if necessary.
2308 * Returns : JB_ERR_OK on success, or
2309 * JB_ERR_MEMORY on out-of-memory error.
2311 *********************************************************************/
2312 static jb_err server_content_encoding(struct client_state *csp, char **header)
2315 if (strstr(*header, "sdch"))
2318 * Shared Dictionary Compression over HTTP isn't supported,
2319 * filtering it anyway is pretty much guaranteed to mess up
2322 csp->content_type |= CT_TABOO;
2325 * Log a warning if the user expects the content to be filtered.
2327 if ((csp->rlist != NULL) &&
2328 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2330 log_error(LOG_LEVEL_INFO,
2331 "SDCH-compressed content detected, content filtering disabled. "
2332 "Consider suppressing SDCH offers made by the client.");
2335 else if (strstr(*header, "gzip"))
2337 /* Mark for gzip decompression */
2338 csp->content_type |= CT_GZIP;
2340 else if (strstr(*header, "deflate"))
2342 /* Mark for zlib decompression */
2343 csp->content_type |= CT_DEFLATE;
2345 else if (strstr(*header, "compress"))
2348 * We can't decompress this; therefore we can't filter
2351 csp->content_type |= CT_TABOO;
2353 #else /* !defined(FEATURE_ZLIB) */
2355 * XXX: Using a black list here isn't the right approach.
2357 * In case of SDCH, building with zlib support isn't
2360 if (strstr(*header, "gzip") ||
2361 strstr(*header, "compress") ||
2362 strstr(*header, "deflate") ||
2363 strstr(*header, "sdch"))
2366 * Body is compressed, turn off pcrs and gif filtering.
2368 csp->content_type |= CT_TABOO;
2371 * Log a warning if the user expects the content to be filtered.
2373 if ((csp->rlist != NULL) &&
2374 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2376 log_error(LOG_LEVEL_INFO,
2377 "Compressed content detected, content filtering disabled. "
2378 "Consider recompiling Privoxy with zlib support or "
2379 "enable the prevent-compression action.");
2382 #endif /* defined(FEATURE_ZLIB) */
2390 /*********************************************************************
2392 * Function : server_adjust_content_encoding
2394 * Description : Remove the Content-Encoding header if the
2395 * decompression was successful and the content
2399 * 1 : csp = Current client state (buffers, headers, etc...)
2400 * 2 : header = On input, pointer to header to modify.
2401 * On output, pointer to the modified header, or NULL
2402 * to remove the header. This function frees the
2403 * original string if necessary.
2405 * Returns : JB_ERR_OK on success, or
2406 * JB_ERR_MEMORY on out-of-memory error.
2408 *********************************************************************/
2409 static jb_err server_adjust_content_encoding(struct client_state *csp, char **header)
2411 if ((csp->flags & CSP_FLAG_MODIFIED)
2412 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2415 * We successfully decompressed the content,
2416 * and have to clean the header now, so the
2417 * client no longer expects compressed data.
2419 * XXX: There is a difference between cleaning
2420 * and removing it completely.
2422 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2429 #endif /* defined(FEATURE_ZLIB) */
2432 /*********************************************************************
2434 * Function : server_adjust_content_length
2436 * Description : Adjust Content-Length header if we modified
2440 * 1 : csp = Current client state (buffers, headers, etc...)
2441 * 2 : header = On input, pointer to header to modify.
2442 * On output, pointer to the modified header, or NULL
2443 * to remove the header. This function frees the
2444 * original string if necessary.
2446 * Returns : JB_ERR_OK on success, or
2447 * JB_ERR_MEMORY on out-of-memory error.
2449 *********************************************************************/
2450 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2452 /* Regenerate header if the content was modified. */
2453 if (csp->flags & CSP_FLAG_MODIFIED)
2455 const size_t header_length = 50;
2457 *header = malloc(header_length);
2458 if (*header == NULL)
2460 return JB_ERR_MEMORY;
2462 create_content_length_header(csp->content_length, *header, header_length);
2463 log_error(LOG_LEVEL_HEADER,
2464 "Adjusted Content-Length to %llu", csp->content_length);
2471 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2472 /*********************************************************************
2474 * Function : server_save_content_length
2476 * Description : Save the Content-Length sent by the server.
2479 * 1 : csp = Current client state (buffers, headers, etc...)
2480 * 2 : header = On input, pointer to header to modify.
2481 * On output, pointer to the modified header, or NULL
2482 * to remove the header. This function frees the
2483 * original string if necessary.
2485 * Returns : JB_ERR_OK on success, or
2486 * JB_ERR_MEMORY on out-of-memory error.
2488 *********************************************************************/
2489 static jb_err server_save_content_length(struct client_state *csp, char **header)
2491 unsigned long long content_length = 0;
2492 const char *header_value;
2494 assert(*(*header+14) == ':');
2496 header_value = *header + 15;
2497 if (JB_ERR_OK != get_content_length(header_value, &content_length))
2499 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2504 csp->expected_content_length = content_length;
2505 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
2506 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2511 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2514 /*********************************************************************
2516 * Function : server_content_md5
2518 * Description : Crumble any Content-MD5 headers if the document was
2519 * modified. FIXME: Should we re-compute instead?
2522 * 1 : csp = Current client state (buffers, headers, etc...)
2523 * 2 : header = On input, pointer to header to modify.
2524 * On output, pointer to the modified header, or NULL
2525 * to remove the header. This function frees the
2526 * original string if necessary.
2528 * Returns : JB_ERR_OK on success, or
2529 * JB_ERR_MEMORY on out-of-memory error.
2531 *********************************************************************/
2532 static jb_err server_content_md5(struct client_state *csp, char **header)
2534 if (csp->flags & CSP_FLAG_MODIFIED)
2536 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2544 /*********************************************************************
2546 * Function : server_content_disposition
2548 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2549 * Called from `sed'.
2552 * 1 : csp = Current client state (buffers, headers, etc...)
2553 * 2 : header = On input, pointer to header to modify.
2554 * On output, pointer to the modified header, or NULL
2555 * to remove the header. This function frees the
2556 * original string if necessary.
2558 * Returns : JB_ERR_OK on success, or
2559 * JB_ERR_MEMORY on out-of-memory error.
2561 *********************************************************************/
2562 static jb_err server_content_disposition(struct client_state *csp, char **header)
2567 * Are we messing with the Content-Disposition header?
2569 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2575 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2577 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2580 * Blocking content-disposition header
2582 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2589 * Replacing Content-Disposition header
2592 *header = strdup("Content-Disposition: ");
2593 string_append(header, newval);
2595 if (*header != NULL)
2597 log_error(LOG_LEVEL_HEADER,
2598 "Content-Disposition header crunched and replaced with: %s", *header);
2601 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2605 /*********************************************************************
2607 * Function : server_last_modified
2609 * Description : Changes Last-Modified header to the actual date
2610 * to help hide-if-modified-since.
2611 * Called from `sed'.
2614 * 1 : csp = Current client state (buffers, headers, etc...)
2615 * 2 : header = On input, pointer to header to modify.
2616 * On output, pointer to the modified header, or NULL
2617 * to remove the header. This function frees the
2618 * original string if necessary.
2620 * Returns : JB_ERR_OK on success, or
2621 * JB_ERR_MEMORY on out-of-memory error.
2623 *********************************************************************/
2624 static jb_err server_last_modified(struct client_state *csp, char **header)
2627 time_t last_modified;
2631 * Are we messing with the Last-Modified header?
2633 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2639 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2641 if (0 == strcmpic(newval, "block"))
2644 * Blocking Last-Modified header. Useless but why not.
2646 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2650 else if (0 == strcmpic(newval, "reset-to-request-time"))
2653 * Setting Last-Modified Header to now.
2656 get_http_time(0, buf, sizeof(buf));
2658 *header = strdup("Last-Modified: ");
2659 string_append(header, buf);
2661 if (*header == NULL)
2663 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2667 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2670 else if (0 == strcmpic(newval, "randomize"))
2672 const char *header_time = *header + sizeof("Last-Modified:");
2674 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2676 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2678 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2684 struct tm *timeptr = NULL;
2686 #ifdef HAVE_GMTIME_R
2690 rtime = (long int)difftime(now, last_modified);
2693 long int days, hours, minutes, seconds;
2694 const int negative_delta = (rtime < 0);
2699 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2701 rtime = pick_from_range(rtime);
2706 last_modified += rtime;
2707 #ifdef HAVE_GMTIME_R
2708 timeptr = gmtime_r(&last_modified, &gmt);
2709 #elif defined(MUTEX_LOCKS_AVAILABLE)
2710 privoxy_mutex_lock(&gmtime_mutex);
2711 timeptr = gmtime(&last_modified);
2712 privoxy_mutex_unlock(&gmtime_mutex);
2714 timeptr = gmtime(&last_modified);
2716 if ((NULL == timeptr) || !strftime(newheader,
2717 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
2719 log_error(LOG_LEVEL_ERROR,
2720 "Randomizing '%s' failed. Crunching the header without replacement.",
2727 *header = strdup("Last-Modified: ");
2728 string_append(header, newheader);
2730 if (*header == NULL)
2732 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2733 return JB_ERR_MEMORY;
2736 days = rtime / (3600 * 24);
2737 hours = rtime / 3600 % 24;
2738 minutes = rtime / 60 % 60;
2739 seconds = rtime % 60;
2741 log_error(LOG_LEVEL_HEADER,
2742 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2743 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2744 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2748 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2757 /*********************************************************************
2759 * Function : client_accept_encoding
2761 * Description : Rewrite the client's Accept-Encoding header so that
2762 * if doesn't allow compression, if the action applies.
2763 * Note: For HTTP/1.0 the absence of the header is enough.
2766 * 1 : csp = Current client state (buffers, headers, etc...)
2767 * 2 : header = On input, pointer to header to modify.
2768 * On output, pointer to the modified header, or NULL
2769 * to remove the header. This function frees the
2770 * original string if necessary.
2772 * Returns : JB_ERR_OK on success, or
2773 * JB_ERR_MEMORY on out-of-memory error.
2775 *********************************************************************/
2776 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2778 #ifdef FEATURE_COMPRESSION
2779 if ((csp->config->feature_flags & RUNTIME_FEATURE_COMPRESSION)
2780 && strstr(*header, "deflate"))
2782 csp->flags |= CSP_FLAG_CLIENT_SUPPORTS_DEFLATE;
2785 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2787 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2795 /*********************************************************************
2797 * Function : client_te
2799 * Description : Rewrite the client's TE header so that
2800 * if doesn't allow compression, if the action applies.
2803 * 1 : csp = Current client state (buffers, headers, etc...)
2804 * 2 : header = On input, pointer to header to modify.
2805 * On output, pointer to the modified header, or NULL
2806 * to remove the header. This function frees the
2807 * original string if necessary.
2809 * Returns : JB_ERR_OK on success, or
2810 * JB_ERR_MEMORY on out-of-memory error.
2812 *********************************************************************/
2813 static jb_err client_te(struct client_state *csp, char **header)
2815 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2818 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2825 /*********************************************************************
2827 * Function : client_referrer
2829 * Description : Handle the "referer" config setting properly.
2830 * Called from `sed'.
2833 * 1 : csp = Current client state (buffers, headers, etc...)
2834 * 2 : header = On input, pointer to header to modify.
2835 * On output, pointer to the modified header, or NULL
2836 * to remove the header. This function frees the
2837 * original string if necessary.
2839 * Returns : JB_ERR_OK on success, or
2840 * JB_ERR_MEMORY on out-of-memory error.
2842 *********************************************************************/
2843 static jb_err client_referrer(struct client_state *csp, char **header)
2845 const char *parameter;
2846 /* booleans for parameters we have to check multiple times */
2847 int parameter_conditional_block;
2848 int parameter_conditional_forge;
2850 #ifdef FEATURE_FORCE_LOAD
2852 * Since the referrer can include the prefix even
2853 * if the request itself is non-forced, we must
2854 * clean it unconditionally.
2856 * XXX: strclean is too broad
2858 strclean(*header, FORCE_PREFIX);
2859 #endif /* def FEATURE_FORCE_LOAD */
2861 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2863 /* Nothing left to do */
2867 parameter = csp->action->string[ACTION_STRING_REFERER];
2868 assert(parameter != NULL);
2869 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2870 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2872 if (!parameter_conditional_block && !parameter_conditional_forge)
2875 * As conditional-block and conditional-forge are the only
2876 * parameters that rely on the original referrer, we can
2877 * remove it now for all the others.
2882 if (0 == strcmpic(parameter, "block"))
2884 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2887 else if (parameter_conditional_block || parameter_conditional_forge)
2889 return handle_conditional_hide_referrer_parameter(header,
2890 csp->http->hostport, parameter_conditional_block);
2892 else if (0 == strcmpic(parameter, "forge"))
2894 return create_forged_referrer(header, csp->http->hostport);
2898 /* interpret parameter as user-supplied referer to fake */
2899 return create_fake_referrer(header, parameter);
2904 /*********************************************************************
2906 * Function : client_accept_language
2908 * Description : Handle the "Accept-Language" config setting properly.
2909 * Called from `sed'.
2912 * 1 : csp = Current client state (buffers, headers, etc...)
2913 * 2 : header = On input, pointer to header to modify.
2914 * On output, pointer to the modified header, or NULL
2915 * to remove the header. This function frees the
2916 * original string if necessary.
2918 * Returns : JB_ERR_OK on success, or
2919 * JB_ERR_MEMORY on out-of-memory error.
2921 *********************************************************************/
2922 static jb_err client_accept_language(struct client_state *csp, char **header)
2927 * Are we messing with the Accept-Language?
2929 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2931 /*I don't think so*/
2935 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2937 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2940 * Blocking Accept-Language header
2942 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2949 * Replacing Accept-Language header
2952 *header = strdup("Accept-Language: ");
2953 string_append(header, newval);
2955 if (*header == NULL)
2957 log_error(LOG_LEVEL_ERROR,
2958 "Insufficient memory. Accept-Language header crunched without replacement.");
2962 log_error(LOG_LEVEL_HEADER,
2963 "Accept-Language header crunched and replaced with: %s", *header);
2966 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2970 /*********************************************************************
2972 * Function : crunch_client_header
2974 * Description : Crunch client header if it matches a string supplied by the
2975 * user. Called from `sed'.
2978 * 1 : csp = Current client state (buffers, headers, etc...)
2979 * 2 : header = On input, pointer to header to modify.
2980 * On output, pointer to the modified header, or NULL
2981 * to remove the header. This function frees the
2982 * original string if necessary.
2984 * Returns : JB_ERR_OK on success and always succeeds
2986 *********************************************************************/
2987 static jb_err crunch_client_header(struct client_state *csp, char **header)
2989 const char *crunch_pattern;
2991 /* Do we feel like crunching? */
2992 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2994 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2996 /* Is the current header the lucky one? */
2997 if (strstr(*header, crunch_pattern))
2999 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
3007 /*********************************************************************
3009 * Function : client_uagent
3011 * Description : Handle the "user-agent" config setting properly
3012 * and remember its original value to enable browser
3013 * bug workarounds. Called from `sed'.
3016 * 1 : csp = Current client state (buffers, headers, etc...)
3017 * 2 : header = On input, pointer to header to modify.
3018 * On output, pointer to the modified header, or NULL
3019 * to remove the header. This function frees the
3020 * original string if necessary.
3022 * Returns : JB_ERR_OK on success, or
3023 * JB_ERR_MEMORY on out-of-memory error.
3025 *********************************************************************/
3026 static jb_err client_uagent(struct client_state *csp, char **header)
3030 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
3035 newval = csp->action->string[ACTION_STRING_USER_AGENT];
3042 *header = strdup("User-Agent: ");
3043 string_append(header, newval);
3045 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
3047 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3051 /*********************************************************************
3053 * Function : client_ua
3055 * Description : Handle "ua-" headers properly. Called from `sed'.
3058 * 1 : csp = Current client state (buffers, headers, etc...)
3059 * 2 : header = On input, pointer to header to modify.
3060 * On output, pointer to the modified header, or NULL
3061 * to remove the header. This function frees the
3062 * original string if necessary.
3064 * Returns : JB_ERR_OK on success, or
3065 * JB_ERR_MEMORY on out-of-memory error.
3067 *********************************************************************/
3068 static jb_err client_ua(struct client_state *csp, char **header)
3070 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3072 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3080 /*********************************************************************
3082 * Function : client_from
3084 * Description : Handle the "from" config setting properly.
3085 * Called from `sed'.
3088 * 1 : csp = Current client state (buffers, headers, etc...)
3089 * 2 : header = On input, pointer to header to modify.
3090 * On output, pointer to the modified header, or NULL
3091 * to remove the header. This function frees the
3092 * original string if necessary.
3094 * Returns : JB_ERR_OK on success, or
3095 * JB_ERR_MEMORY on out-of-memory error.
3097 *********************************************************************/
3098 static jb_err client_from(struct client_state *csp, char **header)
3102 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3109 newval = csp->action->string[ACTION_STRING_FROM];
3112 * Are we blocking the e-mail address?
3114 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
3116 log_error(LOG_LEVEL_HEADER, "crunched From!");
3120 log_error(LOG_LEVEL_HEADER, " modified");
3122 *header = strdup("From: ");
3123 string_append(header, newval);
3125 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3129 /*********************************************************************
3131 * Function : client_send_cookie
3133 * Description : Crunches the "cookie" header if necessary.
3134 * Called from `sed'.
3136 * XXX: Stupid name, doesn't send squat.
3139 * 1 : csp = Current client state (buffers, headers, etc...)
3140 * 2 : header = On input, pointer to header to modify.
3141 * On output, pointer to the modified header, or NULL
3142 * to remove the header. This function frees the
3143 * original string if necessary.
3145 * Returns : JB_ERR_OK on success, or
3146 * JB_ERR_MEMORY on out-of-memory error.
3148 *********************************************************************/
3149 static jb_err client_send_cookie(struct client_state *csp, char **header)
3151 if (csp->action->flags & ACTION_CRUNCH_OUTGOING_COOKIES)
3153 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3161 /*********************************************************************
3163 * Function : client_x_forwarded
3165 * Description : Handle the "x-forwarded-for" config setting properly,
3166 * also used in the add_client_headers list. Called from `sed'.
3169 * 1 : csp = Current client state (buffers, headers, etc...)
3170 * 2 : header = On input, pointer to header to modify.
3171 * On output, pointer to the modified header, or NULL
3172 * to remove the header. This function frees the
3173 * original string if necessary.
3175 * Returns : JB_ERR_OK on success, or
3176 * JB_ERR_MEMORY on out-of-memory error.
3178 *********************************************************************/
3179 jb_err client_x_forwarded(struct client_state *csp, char **header)
3181 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
3183 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
3185 if (0 == strcmpic(parameter, "block"))
3188 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3190 else if (0 == strcmpic(parameter, "add"))
3192 string_append(header, ", ");
3193 string_append(header, csp->ip_addr_str);
3195 if (*header == NULL)
3197 return JB_ERR_MEMORY;
3199 log_error(LOG_LEVEL_HEADER,
3200 "Appended client IP address to %s", *header);
3201 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
3205 log_error(LOG_LEVEL_FATAL,
3206 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
3214 /*********************************************************************
3216 * Function : client_max_forwards
3218 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3219 * from the value of the Max-Forwards header field.
3222 * 1 : csp = Current client state (buffers, headers, etc...)
3223 * 2 : header = On input, pointer to header to modify.
3224 * On output, pointer to the modified header, or NULL
3225 * to remove the header. This function frees the
3226 * original string if necessary.
3228 * Returns : JB_ERR_OK on success, or
3229 * JB_ERR_MEMORY on out-of-memory error.
3231 *********************************************************************/
3232 static jb_err client_max_forwards(struct client_state *csp, char **header)
3236 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3237 (0 == strcmpic(csp->http->gpc, "options")))
3239 assert(*(*header+12) == ':');
3240 if (1 == sscanf(*header+12, ": %d", &max_forwards))
3242 if (max_forwards > 0)
3244 snprintf(*header, strlen(*header)+1, "Max-Forwards: %d", --max_forwards);
3245 log_error(LOG_LEVEL_HEADER,
3246 "Max-Forwards value for %s request reduced to %d.",
3247 csp->http->gpc, max_forwards);
3249 else if (max_forwards < 0)
3251 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3257 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3266 /*********************************************************************
3268 * Function : client_host
3270 * Description : If the request URI did not contain host and
3271 * port information, parse and evaluate the Host
3275 * 1 : csp = Current client state (buffers, headers, etc...)
3276 * 2 : header = On input, pointer to header to modify.
3277 * On output, pointer to the modified header, or NULL
3278 * to remove the header. This function frees the
3279 * original string if necessary.
3281 * Returns : JB_ERR_OK on success, or
3282 * JB_ERR_MEMORY on out-of-memory error.
3284 *********************************************************************/
3285 static jb_err client_host(struct client_state *csp, char **header)
3289 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3290 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3293 p = strdup_or_die((*header)+6);
3295 q = strdup_or_die(p);
3297 freez(csp->http->hostport);
3298 csp->http->hostport = p;
3299 freez(csp->http->host);
3300 csp->http->host = q;
3301 q = strchr(csp->http->host, ':');
3304 /* Terminate hostname and evaluate port string */
3306 csp->http->port = atoi(q);
3310 csp->http->port = csp->http->ssl ? 443 : 80;
3313 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3314 csp->http->hostport, csp->http->host, csp->http->port);
3317 /* Signal client_host_adder() to return right away */
3318 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3324 /*********************************************************************
3326 * Function : client_if_modified_since
3328 * Description : Remove or modify the If-Modified-Since header.
3331 * 1 : csp = Current client state (buffers, headers, etc...)
3332 * 2 : header = On input, pointer to header to modify.
3333 * On output, pointer to the modified header, or NULL
3334 * to remove the header. This function frees the
3335 * original string if necessary.
3337 * Returns : JB_ERR_OK on success, or
3338 * JB_ERR_MEMORY on out-of-memory error.
3340 *********************************************************************/
3341 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3344 #ifdef HAVE_GMTIME_R
3347 struct tm *timeptr = NULL;
3352 if (0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3355 * The client got an error message because of a temporary problem,
3356 * the problem is gone and the client now tries to revalidate our
3357 * error message on the real server. The revalidation would always
3358 * end with the transmission of the whole document and there is
3359 * no need to expose the bogus If-Modified-Since header.
3361 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3364 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3366 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3368 if ((0 == strcmpic(newval, "block")))
3370 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3373 else /* add random value */
3375 const char *header_time = *header + sizeof("If-Modified-Since:");
3377 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3379 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3384 long int hours, minutes, seconds;
3385 long int rtime = strtol(newval, &endptr, 0);
3386 const int negative_range = (rtime < 0);
3390 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3391 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3397 rtime = pick_from_range(rtime);
3401 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3404 tm += rtime * (negative_range ? -1 : 1);
3405 #ifdef HAVE_GMTIME_R
3406 timeptr = gmtime_r(&tm, &gmt);
3407 #elif defined(MUTEX_LOCKS_AVAILABLE)
3408 privoxy_mutex_lock(&gmtime_mutex);
3409 timeptr = gmtime(&tm);
3410 privoxy_mutex_unlock(&gmtime_mutex);
3412 timeptr = gmtime(&tm);
3414 if ((NULL == timeptr) || !strftime(newheader,
3415 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
3417 log_error(LOG_LEVEL_ERROR,
3418 "Randomizing '%s' failed. Crunching the header without replacement.",
3425 *header = strdup("If-Modified-Since: ");
3426 string_append(header, newheader);
3428 if (*header == NULL)
3430 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3431 return JB_ERR_MEMORY;
3434 hours = rtime / 3600;
3435 minutes = rtime / 60 % 60;
3436 seconds = rtime % 60;
3438 log_error(LOG_LEVEL_HEADER,
3439 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3440 *header, (negative_range) ? "subtracted" : "added", hours,
3441 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3442 seconds, (seconds == 1) ? ")" : "s)");
3451 /*********************************************************************
3453 * Function : client_if_none_match
3455 * Description : Remove the If-None-Match header.
3458 * 1 : csp = Current client state (buffers, headers, etc...)
3459 * 2 : header = On input, pointer to header to modify.
3460 * On output, pointer to the modified header, or NULL
3461 * to remove the header. This function frees the
3462 * original string if necessary.
3464 * Returns : JB_ERR_OK on success, or
3465 * JB_ERR_MEMORY on out-of-memory error.
3467 *********************************************************************/
3468 static jb_err client_if_none_match(struct client_state *csp, char **header)
3470 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3472 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3480 /*********************************************************************
3482 * Function : client_x_filter
3484 * Description : Disables filtering if the client set "X-Filter: No".
3485 * Called from `sed'.
3488 * 1 : csp = Current client state (buffers, headers, etc...)
3489 * 2 : header = On input, pointer to header to modify.
3490 * On output, pointer to the modified header, or NULL
3491 * to remove the header. This function frees the
3492 * original string if necessary.
3494 * Returns : JB_ERR_OK on success
3496 *********************************************************************/
3497 jb_err client_x_filter(struct client_state *csp, char **header)
3499 if (0 == strcmpic(*header, "X-Filter: No"))
3501 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3503 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3507 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3509 log_error(LOG_LEVEL_HEADER,
3510 "force-text-mode overruled the client's request to fetch without filtering!");
3514 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3515 csp->flags |= CSP_FLAG_NO_FILTERING;
3516 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3518 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3526 /*********************************************************************
3528 * Function : client_range
3530 * Description : Removes Range, Request-Range and If-Range headers if
3531 * content filtering is enabled and the range doesn't
3534 * If the client's version of the document has been
3535 * altered by Privoxy, the server could interpret the
3536 * range differently than the client intended in which
3537 * case the user could end up with corrupted content.
3539 * If the range starts at byte 0 this isn't an issue
3540 * so the header can pass. Partial requests like this
3541 * are used to render preview images for videos without
3542 * downloading the whole video.
3544 * While HTTP doesn't require that range requests are
3545 * honoured and the client could simply abort the download
3546 * after receiving a sufficient amount of data, various
3547 * clients don't handle complete responses to range
3548 * requests gracefully and emit misleading error messages
3552 * 1 : csp = Current client state (buffers, headers, etc...)
3553 * 2 : header = On input, pointer to header to modify.
3554 * On output, pointer to the modified header, or NULL
3555 * to remove the header. This function frees the
3556 * original string if necessary.
3558 * Returns : JB_ERR_OK
3560 *********************************************************************/
3561 static jb_err client_range(struct client_state *csp, char **header)
3563 if (content_filters_enabled(csp->action)
3564 && (0 != strncmpic(strstr(*header, ":"), ": bytes=0-", 10)))
3566 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3567 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3574 /* the following functions add headers directly to the header list */
3576 /*********************************************************************
3578 * Function : client_host_adder
3580 * Description : Adds the Host: header field if it is missing.
3581 * Called from `sed'.
3584 * 1 : csp = Current client state (buffers, headers, etc...)
3586 * Returns : JB_ERR_OK on success, or
3587 * JB_ERR_MEMORY on out-of-memory error.
3589 *********************************************************************/
3590 static jb_err client_host_adder(struct client_state *csp)
3595 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3597 /* Header already set by the client, nothing to do. */
3601 if (!csp->http->hostport || !*(csp->http->hostport))
3603 /* XXX: When does this happen and why is it OK? */
3604 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3609 * remove 'user:pass@' from 'proto://user:pass@host'
3611 if ((p = strchr( csp->http->hostport, '@')) != NULL)
3617 p = csp->http->hostport;
3620 /* XXX: Just add it, we already made sure that it will be unique */
3621 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3622 err = enlist_unique_header(csp->headers, "Host", p);
3628 /*********************************************************************
3630 * Function : client_xtra_adder
3632 * Description : Used in the add_client_headers list. Called from `sed'.
3635 * 1 : csp = Current client state (buffers, headers, etc...)
3637 * Returns : JB_ERR_OK on success, or
3638 * JB_ERR_MEMORY on out-of-memory error.
3640 *********************************************************************/
3641 static jb_err client_xtra_adder(struct client_state *csp)
3643 struct list_entry *lst;
3646 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3647 lst ; lst = lst->next)
3649 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3650 err = enlist(csp->headers, lst->str);
3662 /*********************************************************************
3664 * Function : client_x_forwarded_for_adder
3666 * Description : Used in the add_client_headers list. Called from `sed'.
3669 * 1 : csp = Current client state (buffers, headers, etc...)
3671 * Returns : JB_ERR_OK on success, or
3672 * JB_ERR_MEMORY on out-of-memory error.
3674 *********************************************************************/
3675 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3677 char *header = NULL;
3680 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3681 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3682 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3685 * If we aren't adding X-Forwarded-For headers,
3686 * or we already appended an existing X-Forwarded-For
3687 * header, there's nothing left to do here.
3692 header = strdup("X-Forwarded-For: ");
3693 string_append(&header, csp->ip_addr_str);
3697 return JB_ERR_MEMORY;
3700 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3701 err = enlist(csp->headers, header);
3708 /*********************************************************************
3710 * Function : server_connection_adder
3712 * Description : Adds an appropriate "Connection:" header to csp->headers
3713 * unless the header was already present. Called from `sed'.
3716 * 1 : csp = Current client state (buffers, headers, etc...)
3718 * Returns : JB_ERR_OK on success, or
3719 * JB_ERR_MEMORY on out-of-memory error.
3721 *********************************************************************/
3722 static jb_err server_connection_adder(struct client_state *csp)
3724 const unsigned int flags = csp->flags;
3725 const char *response_status_line = csp->headers->first->str;
3726 static const char connection_close[] = "Connection: close";
3728 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3729 && (flags & CSP_FLAG_SERVER_CONNECTION_HEADER_SET))
3735 * XXX: if we downgraded the response, this check will fail.
3737 if ((csp->config->feature_flags &
3738 RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3739 && (NULL != response_status_line)
3740 && !strncmpic(response_status_line, "HTTP/1.1", 8)
3741 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3742 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3746 log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
3747 "without Connection header implies keep-alive.");
3748 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
3752 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3754 return enlist(csp->headers, connection_close);
3758 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3759 /*********************************************************************
3761 * Function : server_proxy_connection_adder
3763 * Description : Adds a "Proxy-Connection: keep-alive" header to
3764 * csp->headers when appropriate.
3767 * 1 : csp = Current client state (buffers, headers, etc...)
3769 * Returns : JB_ERR_OK on success, or
3770 * JB_ERR_MEMORY on out-of-memory error.
3772 *********************************************************************/
3773 static jb_err server_proxy_connection_adder(struct client_state *csp)
3775 static const char proxy_connection_header[] = "Proxy-Connection: keep-alive";
3776 jb_err err = JB_ERR_OK;
3778 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
3779 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3780 && !(csp->flags & CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET)
3781 && ((csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET)
3782 || (csp->flags & CSP_FLAG_CHUNKED)))
3784 log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
3785 err = enlist(csp->headers, proxy_connection_header);
3790 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3793 /*********************************************************************
3795 * Function : client_connection_header_adder
3797 * Description : Adds a proper "Connection:" header to csp->headers
3798 * unless the header was already present. Called from `sed'.
3801 * 1 : csp = Current client state (buffers, headers, etc...)
3803 * Returns : JB_ERR_OK on success, or
3804 * JB_ERR_MEMORY on out-of-memory error.
3806 *********************************************************************/
3807 static jb_err client_connection_header_adder(struct client_state *csp)
3809 static const char connection_close[] = "Connection: close";
3811 if (!(csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3812 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
3817 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3818 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3819 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3820 && (csp->http->ssl == 0)
3821 && !strcmpic(csp->http->ver, "HTTP/1.1"))
3823 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
3826 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3828 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3830 return enlist(csp->headers, connection_close);
3834 /*********************************************************************
3836 * Function : server_http
3838 * Description : - Save the HTTP Status into csp->http->status
3839 * - Set CT_TABOO to prevent filtering if the answer
3840 * is a partial range (HTTP status 206)
3841 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3845 * 1 : csp = Current client state (buffers, headers, etc...)
3846 * 2 : header = On input, pointer to header to modify.
3847 * On output, pointer to the modified header, or NULL
3848 * to remove the header. This function frees the
3849 * original string if necessary.
3851 * Returns : JB_ERR_OK on success, or
3852 * JB_ERR_MEMORY on out-of-memory error.
3854 *********************************************************************/
3855 static jb_err server_http(struct client_state *csp, char **header)
3857 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3858 if (csp->http->status == 206)
3860 csp->content_type = CT_TABOO;
3863 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3865 /* XXX: Should we do a real validity check here? */
3866 if (strlen(*header) > 8)
3869 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3874 * XXX: Should we block the request or
3875 * enlist a valid status code line here?
3877 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3878 "Downgrading to HTTP/1.0 impossible.");
3885 /*********************************************************************
3887 * Function : add_cooky_expiry_date
3889 * Description : Adds a cookie expiry date to a string.
3892 * 1 : cookie = On input, pointer to cookie to modify.
3893 * On output, pointer to the modified header.
3894 * The original string is freed.
3895 * 2 : lifetime = Seconds the cookie should be valid
3899 *********************************************************************/
3900 static void add_cookie_expiry_date(char **cookie, time_t lifetime)
3903 struct tm *timeptr = NULL;
3904 time_t expiry_date = time(NULL) + lifetime;
3905 #ifdef HAVE_GMTIME_R
3908 timeptr = gmtime_r(&expiry_date, &gmt);
3909 #elif defined(MUTEX_LOCKS_AVAILABLE)
3910 privoxy_mutex_lock(&gmtime_mutex);
3911 timeptr = gmtime(&expiry_date);
3912 privoxy_mutex_unlock(&gmtime_mutex);
3914 timeptr = gmtime(&expiry_date);
3917 if (NULL == timeptr)
3919 log_error(LOG_LEVEL_FATAL,
3920 "Failed to get the time in add_cooky_expiry_date()");
3922 strftime(tmp, sizeof(tmp), "; expires=%a, %d-%b-%Y %H:%M:%S GMT", timeptr);
3923 if (JB_ERR_OK != string_append(cookie, tmp))
3925 log_error(LOG_LEVEL_FATAL, "Out of memory in add_cooky_expiry()");
3930 /*********************************************************************
3932 * Function : server_set_cookie
3934 * Description : Handle the server "cookie" header properly.
3935 * Crunch, accept or rewrite it to a session cookie.
3936 * Called from `sed'.
3939 * 1 : csp = Current client state (buffers, headers, etc...)
3940 * 2 : header = On input, pointer to header to modify.
3941 * On output, pointer to the modified header, or NULL
3942 * to remove the header. This function frees the
3943 * original string if necessary.
3945 * Returns : JB_ERR_OK on success, or
3946 * JB_ERR_MEMORY on out-of-memory error.
3948 *********************************************************************/
3949 static jb_err server_set_cookie(struct client_state *csp, char **header)
3951 if ((csp->action->flags & ACTION_CRUNCH_INCOMING_COOKIES) != 0)
3953 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3956 else if ((0 != (csp->action->flags & ACTION_SESSION_COOKIES_ONLY))
3957 || (0 != (csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME)))
3961 long cookie_lifetime = 0;
3964 NO_EXPIRY_DATE_SPECIFIED,
3965 EXPIRY_DATE_ACCEPTABLE,
3966 EXPIRY_DATE_UNACCEPTABLE
3967 } expiry_date_status = NO_EXPIRY_DATE_SPECIFIED;
3969 /* A variable to store the tag we're working on */
3972 /* Skip "Set-Cookie:" (11 characters) in header */
3973 cur_tag = *header + 11;
3975 /* skip whitespace between "Set-Cookie:" and value */
3976 while (*cur_tag && privoxy_isspace(*cur_tag))
3983 if ((csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME) != 0)
3985 const char *param = csp->action->string[ACTION_STRING_LIMIT_COOKIE_LIFETIME];
3987 cookie_lifetime = strtol(param, NULL, 0);
3988 if (cookie_lifetime < 0)
3990 log_error(LOG_LEVEL_FATAL, "Invalid cookie lifetime limit: %s", param);
3992 cookie_lifetime *= 60;
3995 /* Loop through each tag in the cookie */
3999 char *next_tag = strchr(cur_tag, ';');
4000 if (next_tag != NULL)
4002 /* Skip the ';' character itself */
4005 /* skip whitespace ";" and start of tag */
4006 while (*next_tag && privoxy_isspace(*next_tag))
4013 /* "Next tag" is the end of the string */
4014 next_tag = cur_tag + strlen(cur_tag);
4018 * Check the expiration date to see
4019 * if the cookie is still valid, if yes,
4020 * rewrite it to a session cookie.
4022 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
4024 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
4026 if ((expiration_date[0] == '"')
4027 && (expiration_date[1] != '\0'))
4030 * Skip quotation mark. RFC 2109 10.1.2 seems to hint
4031 * that the expiration date isn't supposed to be quoted,
4032 * but some servers do it anyway.
4037 /* Did we detect the date properly? */
4038 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
4041 * Nope, treat it as if it was still valid.
4043 * XXX: Should we remove the whole cookie instead?
4045 log_error(LOG_LEVEL_ERROR,
4046 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
4047 string_move(cur_tag, next_tag);
4048 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4053 * Yes. Check if the cookie is still valid.
4055 * If the cookie is already expired it's probably
4056 * a delete cookie and even if it isn't, the browser
4057 * will discard it anyway.
4061 * XXX: timegm() isn't available on some AmigaOS
4062 * versions and our replacement doesn't work.
4064 * Our options are to either:
4066 * - disable session-cookies-only completely if timegm
4069 * - to simply remove all expired tags, like it has
4070 * been done until Privoxy 3.0.6 and to live with
4071 * the consequence that it can cause login/logout
4072 * problems on servers that don't validate their
4073 * input properly, or
4075 * - to replace it with mktime in which
4076 * case there is a slight chance of valid cookies
4077 * passing as already expired.
4079 * This is the way it's currently done and it's not
4080 * as bad as it sounds. If the missing GMT offset is
4081 * enough to change the result of the expiration check
4082 * the cookie will be only valid for a few hours
4083 * anyway, which in many cases will be shorter
4084 * than a browser session.
4086 if (cookie_time < now)
4088 log_error(LOG_LEVEL_HEADER,
4089 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4090 /* Just in case some clown sets more then one expiration date */
4092 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4094 else if ((cookie_lifetime != 0) && (cookie_time < (now + cookie_lifetime)))
4096 log_error(LOG_LEVEL_HEADER, "Cookie \'%s\' can pass unmodified. "
4097 "Its lifetime is below the limit.", *header);
4098 /* Just in case some clown sets more then one expiration date */
4100 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4105 * Still valid, delete expiration date by copying
4106 * the rest of the string over it.
4108 string_move(cur_tag, next_tag);
4110 /* That changed the header, need to issue a log message */
4111 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4114 * Note that the next tag has now been moved to *cur_tag,
4115 * so we do not need to update the cur_tag pointer.
4123 /* Move on to next cookie tag */
4128 if (expiry_date_status != EXPIRY_DATE_ACCEPTABLE)
4130 assert(NULL != *header);
4131 if (cookie_lifetime != 0)
4133 add_cookie_expiry_date(header, cookie_lifetime);
4134 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to: %s", *header);
4136 else if (expiry_date_status != NO_EXPIRY_DATE_SPECIFIED)
4138 log_error(LOG_LEVEL_HEADER,
4139 "Cookie rewritten to a temporary one: %s", *header);
4148 #ifdef FEATURE_FORCE_LOAD
4149 /*********************************************************************
4151 * Function : strclean
4153 * Description : In-Situ-Eliminate all occurrences of substring in
4157 * 1 : string = string to clean
4158 * 2 : substring = substring to eliminate
4160 * Returns : Number of eliminations
4162 *********************************************************************/
4163 int strclean(char *string, const char *substring)
4169 len = strlen(substring);
4171 while((pos = strstr(string, substring)) != NULL)
4178 while (*p++ != '\0');
4185 #endif /* def FEATURE_FORCE_LOAD */
4188 /*********************************************************************
4190 * Function : parse_header_time
4192 * Description : Parses time formats used in HTTP header strings
4193 * to get the numerical respresentation.
4196 * 1 : header_time = HTTP header time as string.
4197 * 2 : result = storage for header_time in seconds
4199 * Returns : JB_ERR_OK if the time format was recognized, or
4200 * JB_ERR_PARSE otherwise.
4202 *********************************************************************/
4203 static jb_err parse_header_time(const char *header_time, time_t *result)
4207 * Checking for two-digit years first in an
4208 * attempt to work around GNU libc's strptime()
4209 * reporting negative year values when using %Y.
4211 static const char time_formats[][22] = {
4212 /* Tue, 02-Jun-37 20:00:00 */
4213 "%a, %d-%b-%y %H:%M:%S",
4214 /* Tue, 02 Jun 2037 20:00:00 */
4215 "%a, %d %b %Y %H:%M:%S",
4216 /* Tue, 02-Jun-2037 20:00:00 */
4217 "%a, %d-%b-%Y %H:%M:%S",
4218 /* Tuesday, 02-Jun-2037 20:00:00 */
4219 "%A, %d-%b-%Y %H:%M:%S",
4220 /* Tuesday Jun 02 20:00:00 2037 */
4221 "%A %b %d %H:%M:%S %Y"
4225 for (i = 0; i < SZ(time_formats); i++)
4228 * Zero out gmt to prevent time zone offsets.
4229 * Documented to be required for GNU libc.
4231 memset(&gmt, 0, sizeof(gmt));
4233 if (NULL != strptime(header_time, time_formats[i], &gmt))
4235 /* Sanity check for GNU libc. */
4236 if (gmt.tm_year < 0)
4238 log_error(LOG_LEVEL_HEADER,
4239 "Failed to parse '%s' using '%s'. Moving on.",
4240 header_time, time_formats[i]);
4243 *result = timegm(&gmt);
4245 #ifdef FEATURE_STRPTIME_SANITY_CHECKS
4247 * Verify that parsing the date recreated from the first
4248 * parse operation gets the previous result. If it doesn't,
4249 * either strptime() or strftime() are malfunctioning.
4251 * We could string-compare the recreated date with the original
4252 * header date, but this leads to false positives as strptime()
4253 * may let %a accept all day formats while strftime() will only
4257 char recreated_date[100];
4261 tm = gmtime(result);
4262 strftime(recreated_date, sizeof(recreated_date), time_formats[i], tm);
4263 memset(&gmt, 0, sizeof(gmt));
4264 if (NULL == strptime(recreated_date, time_formats[i], &gmt))
4266 log_error(LOG_LEVEL_ERROR,
4267 "Failed to parse '%s' generated with '%s' to recreate '%s'.",
4268 recreated_date, time_formats[i], header_time);
4271 result2 = timegm(&gmt);
4272 if (*result != result2)
4274 log_error(LOG_LEVEL_ERROR, "strftime() and strptime() disagree. "
4275 "Format: '%s'. In: '%s', out: '%s'. %d != %d. Rejecting.",
4276 time_formats[i], header_time, recreated_date, *result, result2);
4286 return JB_ERR_PARSE;
4291 /*********************************************************************
4293 * Function : get_destination_from_headers
4295 * Description : Parse the "Host:" header to get the request's destination.
4296 * Only needed if the client's request was forcefully
4297 * redirected into Privoxy.
4299 * Code mainly copied from client_host() which is currently
4300 * run too late for this purpose.
4303 * 1 : headers = List of headers (one of them hopefully being
4304 * the "Host:" header)
4305 * 2 : http = storage for the result (host, port and hostport).
4307 * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
4308 * JB_ERR_PARSE if the host header couldn't be found,
4309 * JB_ERR_OK otherwise.
4311 *********************************************************************/
4312 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4318 host = get_header_value(headers, "Host:");
4322 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4323 return JB_ERR_PARSE;
4326 p = strdup_or_die(host);
4328 q = strdup_or_die(p);
4330 freez(http->hostport);
4334 q = strchr(http->host, ':');
4337 /* Terminate hostname and evaluate port string */
4339 http->port = atoi(q);
4343 http->port = http->ssl ? 443 : 80;
4346 /* Rebuild request URL */
4348 http->url = strdup(http->ssl ? "https://" : "http://");
4349 string_append(&http->url, http->hostport);
4350 string_append(&http->url, http->path);
4351 if (http->url == NULL)
4353 return JB_ERR_MEMORY;
4356 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4364 /*********************************************************************
4366 * Function : create_forged_referrer
4368 * Description : Helper for client_referrer to forge a referer as
4369 * 'http://hostname[:port]/' to fool stupid
4370 * checks for in-site links
4373 * 1 : header = Pointer to header pointer
4374 * 2 : hostport = Host and optionally port as string
4376 * Returns : JB_ERR_OK in case of success, or
4377 * JB_ERR_MEMORY in case of memory problems.
4379 *********************************************************************/
4380 static jb_err create_forged_referrer(char **header, const char *hostport)
4382 assert(NULL == *header);
4384 *header = strdup("Referer: http://");
4385 string_append(header, hostport);
4386 string_append(header, "/");
4388 if (NULL == *header)
4390 return JB_ERR_MEMORY;
4393 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4400 /*********************************************************************
4402 * Function : create_fake_referrer
4404 * Description : Helper for client_referrer to create a fake referrer
4405 * based on a string supplied by the user.