1 const char parsers_rcs[] = "$Id: parsers.c,v 1.281 2013/12/24 13:34:22 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
8 * Copyright : Written by and Copyright (C) 2001-2012 the
9 * Privoxy team. http://www.privoxy.org/
11 * Based on the Internet Junkbuster originally written
12 * by and Copyright (C) 1997 Anonymous Coders and
13 * Junkbusters Corporation. http://www.junkbusters.com
15 * This program is free software; you can redistribute it
16 * and/or modify it under the terms of the GNU General
17 * Public License as published by the Free Software
18 * Foundation; either version 2 of the License, or (at
19 * your option) any later version.
21 * This program is distributed in the hope that it will
22 * be useful, but WITHOUT ANY WARRANTY; without even the
23 * implied warranty of MERCHANTABILITY or FITNESS FOR A
24 * PARTICULAR PURPOSE. See the GNU General Public
25 * License for more details.
27 * The GNU General Public License should be included with
28 * this file. If not, you can view it at
29 * http://www.gnu.org/copyleft/gpl.html
30 * or write to the Free Software Foundation, Inc., 59
31 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
33 *********************************************************************/
40 #include <sys/types.h>
50 * Convince GNU's libc to provide a strptime prototype.
59 #define GZIP_IDENTIFIER_1 0x1f
60 #define GZIP_IDENTIFIER_2 0x8b
62 #define GZIP_FLAG_CHECKSUM 0x02
63 #define GZIP_FLAG_EXTRA_FIELDS 0x04
64 #define GZIP_FLAG_FILE_NAME 0x08
65 #define GZIP_FLAG_COMMENT 0x10
66 #define GZIP_FLAG_RESERVED_BITS 0xe0
69 #if !defined(_WIN32) && !defined(__OS2__)
75 #ifdef FEATURE_PTHREAD
77 /* jcc.h is for mutex semapores only */
78 #endif /* def FEATURE_PTHREAD */
83 #include "jbsockets.h"
93 const char parsers_h_rcs[] = PARSERS_H_VERSION;
95 static char *get_header_line(struct iob *iob);
96 static jb_err scan_headers(struct client_state *csp);
97 static jb_err header_tagger(struct client_state *csp, char *header);
98 static jb_err parse_header_time(const char *header_time, time_t *result);
100 static jb_err crumble (struct client_state *csp, char **header);
101 static jb_err filter_header (struct client_state *csp, char **header);
102 static jb_err client_connection (struct client_state *csp, char **header);
103 static jb_err client_referrer (struct client_state *csp, char **header);
104 static jb_err client_uagent (struct client_state *csp, char **header);
105 static jb_err client_ua (struct client_state *csp, char **header);
106 static jb_err client_from (struct client_state *csp, char **header);
107 static jb_err client_send_cookie (struct client_state *csp, char **header);
108 static jb_err client_x_forwarded (struct client_state *csp, char **header);
109 static jb_err client_accept_encoding (struct client_state *csp, char **header);
110 static jb_err client_te (struct client_state *csp, char **header);
111 static jb_err client_max_forwards (struct client_state *csp, char **header);
112 static jb_err client_host (struct client_state *csp, char **header);
113 static jb_err client_if_modified_since (struct client_state *csp, char **header);
114 static jb_err client_accept_language (struct client_state *csp, char **header);
115 static jb_err client_if_none_match (struct client_state *csp, char **header);
116 static jb_err crunch_client_header (struct client_state *csp, char **header);
117 static jb_err client_x_filter (struct client_state *csp, char **header);
118 static jb_err client_range (struct client_state *csp, char **header);
119 static jb_err server_set_cookie (struct client_state *csp, char **header);
120 static jb_err server_connection (struct client_state *csp, char **header);
121 static jb_err server_content_type (struct client_state *csp, char **header);
122 static jb_err server_adjust_content_length(struct client_state *csp, char **header);
123 static jb_err server_content_md5 (struct client_state *csp, char **header);
124 static jb_err server_content_encoding (struct client_state *csp, char **header);
125 static jb_err server_transfer_coding (struct client_state *csp, char **header);
126 static jb_err server_http (struct client_state *csp, char **header);
127 static jb_err crunch_server_header (struct client_state *csp, char **header);
128 static jb_err server_last_modified (struct client_state *csp, char **header);
129 static jb_err server_content_disposition(struct client_state *csp, char **header);
131 static jb_err server_adjust_content_encoding(struct client_state *csp, char **header);
134 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
135 static jb_err server_save_content_length(struct client_state *csp, char **header);
136 static jb_err server_keep_alive(struct client_state *csp, char **header);
137 static jb_err server_proxy_connection(struct client_state *csp, char **header);
138 static jb_err client_keep_alive(struct client_state *csp, char **header);
139 static jb_err client_save_content_length(struct client_state *csp, char **header);
140 static jb_err client_proxy_connection(struct client_state *csp, char **header);
141 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
143 static jb_err client_host_adder (struct client_state *csp);
144 static jb_err client_xtra_adder (struct client_state *csp);
145 static jb_err client_x_forwarded_for_adder(struct client_state *csp);
146 static jb_err client_connection_header_adder(struct client_state *csp);
147 static jb_err server_connection_adder(struct client_state *csp);
148 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
149 static jb_err server_proxy_connection_adder(struct client_state *csp);
150 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
151 static jb_err proxy_authentication(struct client_state *csp, char **header);
153 static jb_err create_forged_referrer(char **header, const char *hostport);
154 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
155 static jb_err handle_conditional_hide_referrer_parameter(char **header,
156 const char *host, const int parameter_conditional_block);
157 static void create_content_length_header(unsigned long long content_length,
158 char *header, size_t buffer_length);
161 * List of functions to run on a list of headers.
165 /** The header prefix to match */
168 /** The length of the prefix to match */
171 /** The function to apply to this line */
172 const parser_func_ptr parser;
175 static const struct parsers client_patterns[] = {
176 { "referer:", 8, client_referrer },
177 { "user-agent:", 11, client_uagent },
178 { "ua-", 3, client_ua },
179 { "from:", 5, client_from },
180 { "cookie:", 7, client_send_cookie },
181 { "x-forwarded-for:", 16, client_x_forwarded },
182 { "Accept-Encoding:", 16, client_accept_encoding },
183 { "TE:", 3, client_te },
184 { "Host:", 5, client_host },
185 { "if-modified-since:", 18, client_if_modified_since },
186 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
187 { "Keep-Alive:", 11, client_keep_alive },
188 { "Content-Length:", 15, client_save_content_length },
189 { "Proxy-Connection:", 17, client_proxy_connection },
191 { "Keep-Alive:", 11, crumble },
192 { "Proxy-Connection:", 17, crumble },
194 { "connection:", 11, client_connection },
195 { "max-forwards:", 13, client_max_forwards },
196 { "Accept-Language:", 16, client_accept_language },
197 { "if-none-match:", 14, client_if_none_match },
198 { "Range:", 6, client_range },
199 { "Request-Range:", 14, client_range },
200 { "If-Range:", 9, client_range },
201 { "X-Filter:", 9, client_x_filter },
202 { "Proxy-Authorization:", 20, proxy_authentication },
204 { "Transfer-Encoding:", 18, client_transfer_encoding },
206 { "*", 0, crunch_client_header },
207 { "*", 0, filter_header },
211 static const struct parsers server_patterns[] = {
212 { "HTTP/", 5, server_http },
213 { "set-cookie:", 11, server_set_cookie },
214 { "connection:", 11, server_connection },
215 { "Content-Type:", 13, server_content_type },
216 { "Content-MD5:", 12, server_content_md5 },
217 { "Content-Encoding:", 17, server_content_encoding },
218 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
219 { "Content-Length:", 15, server_save_content_length },
220 { "Keep-Alive:", 11, server_keep_alive },
221 { "Proxy-Connection:", 17, server_proxy_connection },
223 { "Keep-Alive:", 11, crumble },
224 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
225 { "Transfer-Encoding:", 18, server_transfer_coding },
226 { "content-disposition:", 20, server_content_disposition },
227 { "Last-Modified:", 14, server_last_modified },
228 { "Proxy-Authenticate:", 19, proxy_authentication },
229 { "*", 0, crunch_server_header },
230 { "*", 0, filter_header },
234 static const add_header_func_ptr add_client_headers[] = {
236 client_x_forwarded_for_adder,
238 client_connection_header_adder,
242 static const add_header_func_ptr add_server_headers[] = {
243 server_connection_adder,
244 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
245 server_proxy_connection_adder,
246 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
250 /*********************************************************************
252 * Function : flush_socket
254 * Description : Write any pending "buffered" content.
257 * 1 : fd = file descriptor of the socket to read
258 * 2 : iob = The I/O buffer to flush, usually csp->iob.
260 * Returns : On success, the number of bytes written are returned (zero
261 * indicates nothing was written). On error, -1 is returned,
262 * and errno is set appropriately. If count is zero and the
263 * file descriptor refers to a regular file, 0 will be
264 * returned without causing any other effect. For a special
265 * file, the results are not portable.
267 *********************************************************************/
268 long flush_socket(jb_socket fd, struct iob *iob)
270 long len = iob->eod - iob->cur;
277 if (write_socket(fd, iob->cur, (size_t)len))
281 iob->eod = iob->cur = iob->buf;
287 /*********************************************************************
289 * Function : add_to_iob
291 * Description : Add content to the buffer, expanding the
292 * buffer if necessary.
295 * 1 : iob = Destination buffer.
296 * 2 : buffer_limit = Limit to which the destination may grow
297 * 3 : src = holds the content to be added
298 * 4 : n = number of bytes to be added
300 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
301 * or buffer limit reached.
303 *********************************************************************/
304 jb_err add_to_iob(struct iob *iob, const size_t buffer_limit, char *src, long n)
306 size_t used, offset, need;
309 if (n <= 0) return JB_ERR_OK;
311 used = (size_t)(iob->eod - iob->buf);
312 offset = (size_t)(iob->cur - iob->buf);
313 need = used + (size_t)n + 1;
316 * If the buffer can't hold the new data, extend it first.
317 * Use the next power of two if possible, else use the actual need.
319 if (need > buffer_limit)
321 log_error(LOG_LEVEL_INFO,
322 "Buffer limit reached while extending the buffer (iob). Needed: %d. Limit: %d",
324 return JB_ERR_MEMORY;
327 if (need > iob->size)
329 size_t want = iob->size ? iob->size : 512;
336 if (want <= buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
340 else if (NULL != (p = (char *)realloc(iob->buf, need)))
346 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
347 return JB_ERR_MEMORY;
350 /* Update the iob pointers */
351 iob->cur = p + offset;
356 /* copy the new data into the iob buffer */
357 memcpy(iob->eod, src, (size_t)n);
359 /* point to the end of the data */
362 /* null terminate == cheap insurance */
370 /*********************************************************************
372 * Function : clear_iob
374 * Description : Frees the memory allocated for an I/O buffer and
375 * resets the structure.
378 * 1 : iob = I/O buffer to clear.
380 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
381 * or buffer limit reached.
383 *********************************************************************/
384 void clear_iob(struct iob *iob)
387 memset(iob, '\0', sizeof(*iob));;
392 /*********************************************************************
394 * Function : decompress_iob
396 * Description : Decompress buffered page, expanding the
397 * buffer as necessary. csp->iob->cur
398 * should point to the the beginning of the
399 * compressed data block.
402 * 1 : csp = Current client state (buffers, headers, etc...)
404 * Returns : JB_ERR_OK on success,
405 * JB_ERR_MEMORY if out-of-memory limit reached, and
406 * JB_ERR_COMPRESS if error decompressing buffer.
408 *********************************************************************/
409 jb_err decompress_iob(struct client_state *csp)
411 char *buf; /* new, uncompressed buffer */
412 char *cur; /* Current iob position (to keep the original
413 * iob->cur unmodified if we return early) */
414 size_t bufsize; /* allocated size of the new buffer */
415 size_t old_size; /* Content size before decompression */
416 size_t skip_size; /* Number of bytes at the beginning of the iob
417 that we should NOT decompress. */
418 int status; /* return status of the inflate() call */
419 z_stream zstr; /* used by calls to zlib */
421 assert(csp->iob->cur - csp->iob->buf > 0);
422 assert(csp->iob->eod - csp->iob->cur > 0);
424 bufsize = csp->iob->size;
425 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
426 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
430 if (bufsize < (size_t)10)
433 * This is to protect the parsing of gzipped data,
434 * but it should(?) be valid for deflated data also.
436 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
437 return JB_ERR_COMPRESS;
440 if (csp->content_type & CT_GZIP)
443 * Our task is slightly complicated by the facts that data
444 * compressed by gzip does not include a zlib header, and
445 * that there is no easily accessible interface in zlib to
446 * handle a gzip header. We strip off the gzip header by
447 * hand, and later inform zlib not to expect a header.
451 * Strip off the gzip header. Please see RFC 1952 for more
452 * explanation of the appropriate fields.
454 if (((*cur++ & 0xff) != GZIP_IDENTIFIER_1)
455 || ((*cur++ & 0xff) != GZIP_IDENTIFIER_2)
456 || (*cur++ != Z_DEFLATED))
458 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
459 return JB_ERR_COMPRESS;
464 if (flags & GZIP_FLAG_RESERVED_BITS)
466 /* The gzip header has reserved bits set; bail out. */
467 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
468 return JB_ERR_COMPRESS;
472 * Skip mtime (4 bytes), extra flags (1 byte)
473 * and OS type (1 byte).
477 /* Skip extra fields if necessary. */
478 if (flags & GZIP_FLAG_EXTRA_FIELDS)
481 * Skip a given number of bytes, specified
482 * as a 16-bit little-endian value.
484 * XXX: this code is untested and should probably be removed.
488 skip_bytes += *cur++ << 8;
491 * The number of bytes to skip should be positive
492 * and we'd like to stay in the buffer.
494 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
496 log_error(LOG_LEVEL_ERROR,
497 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
499 return JB_ERR_COMPRESS;
501 log_error(LOG_LEVEL_INFO,
502 "Skipping %d bytes for gzip compression. Does this sound right?",
507 /* Skip the filename if necessary. */
508 if (flags & GZIP_FLAG_FILE_NAME)
510 /* A null-terminated string is supposed to follow. */
511 while (*cur++ && (cur < csp->iob->eod));
514 /* Skip the comment if necessary. */
515 if (flags & GZIP_FLAG_COMMENT)
517 /* A null-terminated string is supposed to follow. */
518 while (*cur++ && (cur < csp->iob->eod));
521 /* Skip the CRC if necessary. */
522 if (flags & GZIP_FLAG_CHECKSUM)
527 if (cur >= csp->iob->eod)
530 * If the current position pointer reached or passed
531 * the buffer end, we were obviously tricked to skip
534 log_error(LOG_LEVEL_ERROR,
535 "Malformed gzip header detected. Aborting decompression.");
536 return JB_ERR_COMPRESS;
540 else if (csp->content_type & CT_DEFLATE)
543 * In theory (that is, according to RFC 1950), deflate-compressed
544 * data should begin with a two-byte zlib header and have an
545 * adler32 checksum at the end. It seems that in practice only
546 * the raw compressed data is sent. Note that this means that
547 * we are not RFC 1950-compliant here, but the advantage is that
548 * this actually works. :)
550 * We add a dummy null byte to tell zlib where the data ends,
551 * and later inform it not to expect a header.
553 * Fortunately, add_to_iob() has thoughtfully null-terminated
554 * the buffer; we can just increment the end pointer to include
561 log_error(LOG_LEVEL_ERROR,
562 "Unable to determine compression format for decompression");
563 return JB_ERR_COMPRESS;
566 /* Set up the fields required by zlib. */
567 zstr.next_in = (Bytef *)cur;
568 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
569 zstr.zalloc = Z_NULL;
571 zstr.opaque = Z_NULL;
574 * Passing -MAX_WBITS to inflateInit2 tells the library
575 * that there is no zlib header.
577 if (inflateInit2(&zstr, -MAX_WBITS) != Z_OK)
579 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
580 return JB_ERR_COMPRESS;
584 * Next, we allocate new storage for the inflated data.
585 * We don't modify the existing iob yet, so in case there
586 * is error in decompression we can recover gracefully.
588 buf = zalloc(bufsize);
591 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
592 return JB_ERR_MEMORY;
595 assert(bufsize >= skip_size);
596 memcpy(buf, csp->iob->buf, skip_size);
597 zstr.avail_out = (uInt)(bufsize - skip_size);
598 zstr.next_out = (Bytef *)buf + skip_size;
600 /* Try to decompress the whole stream in one shot. */
601 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
603 /* We need to allocate more memory for the output buffer. */
605 char *tmpbuf; /* used for realloc'ing the buffer */
606 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
608 if (0 == zstr.avail_in)
611 * If zlib wants more data then there's a problem, because
612 * the complete compressed file should have been buffered.
614 log_error(LOG_LEVEL_ERROR,
615 "Unexpected end of compressed iob. Using what we got so far.");
620 * If we reached the buffer limit and still didn't have enough
621 * memory, just give up. Due to the ceiling enforced by the next
622 * if block we could actually check for equality here, but as it
623 * can be easily mistaken for a bug we don't.
625 if (bufsize >= csp->config->buffer_limit)
627 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
628 return JB_ERR_MEMORY;
631 /* Try doubling the buffer size each time. */
634 /* Don't exceed the buffer limit. */
635 if (bufsize > csp->config->buffer_limit)
637 bufsize = csp->config->buffer_limit;
640 /* Try to allocate the new buffer. */
641 tmpbuf = realloc(buf, bufsize);
644 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
646 return JB_ERR_MEMORY;
650 char *oldnext_out = (char *)zstr.next_out;
653 * Update the fields for inflate() to use the new
654 * buffer, which may be in a location different from
657 zstr.avail_out += (uInt)(bufsize - oldbufsize);
658 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
661 * Compare with an uglier method of calculating these values
662 * that doesn't require the extra oldbufsize variable.
664 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
665 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
671 if (Z_STREAM_ERROR == inflateEnd(&zstr))
673 log_error(LOG_LEVEL_ERROR,
674 "Inconsistent stream state after decompression: %s", zstr.msg);
676 * XXX: Intentionally no return.
678 * According to zlib.h, Z_STREAM_ERROR is returned
679 * "if the stream state was inconsistent".
681 * I assume in this case inflate()'s status
682 * would also be something different than Z_STREAM_END
683 * so this check should be redundant, but lets see.
687 if ((status != Z_STREAM_END) && (0 != zstr.avail_in))
690 * We failed to decompress the stream and it's
691 * not simply because of missing data.
693 log_error(LOG_LEVEL_ERROR,
694 "Unexpected error while decompressing to the buffer (iob): %s",
696 return JB_ERR_COMPRESS;
700 * Finally, we can actually update the iob, since the
701 * decompression was successful. First, free the old
704 freez(csp->iob->buf);
706 /* Now, update the iob to use the new buffer. */
708 csp->iob->cur = csp->iob->buf + skip_size;
709 csp->iob->eod = (char *)zstr.next_out;
710 csp->iob->size = bufsize;
713 * Make sure the new uncompressed iob obeys some minimal
714 * consistency conditions.
716 if ((csp->iob->buf < csp->iob->cur)
717 && (csp->iob->cur <= csp->iob->eod)
718 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
720 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
721 if (new_size > (size_t)0)
723 log_error(LOG_LEVEL_RE_FILTER,
724 "Decompression successful. Old size: %d, new size: %d.",
729 /* zlib thinks this is OK, so lets do the same. */
730 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
735 /* It seems that zlib did something weird. */
736 log_error(LOG_LEVEL_ERROR,
737 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
738 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
739 csp->iob->eod, csp->iob->buf + csp->iob->size);
740 return JB_ERR_COMPRESS;
746 #endif /* defined(FEATURE_ZLIB) */
749 /*********************************************************************
751 * Function : normalize_lws
753 * Description : Reduces unquoted linear white space in headers
754 * to a single space in accordance with RFC 2616 2.2.
755 * This simplifies parsing and filtering later on.
757 * XXX: Remove log messages before
758 * the next stable release?
761 * 1 : header = A header with linear white space to reduce.
765 *********************************************************************/
766 static void normalize_lws(char *header)
772 if (privoxy_isspace(*p) && privoxy_isspace(*(p+1)))
776 while (privoxy_isspace(*q))
780 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
786 log_error(LOG_LEVEL_HEADER,
787 "Converting tab to space in '%s'", header);
792 char *end_of_token = strstr(p+1, "\"");
794 if (NULL != end_of_token)
796 /* Don't mess with quoted text. */
801 log_error(LOG_LEVEL_HEADER,
802 "Ignoring single quote in '%s'", header);
808 p = strchr(header, ':');
809 if ((p != NULL) && (p != header) && privoxy_isspace(*(p-1)))
812 * There's still space before the colon.
820 /*********************************************************************
822 * Function : get_header
824 * Description : This (odd) routine will parse the csp->iob
825 * to get the next complete header.
828 * 1 : iob = The I/O buffer to parse, usually csp->iob.
830 * Returns : Any one of the following:
832 * 1) a pointer to a dynamically allocated string that contains a header line
833 * 2) NULL indicating that the end of the header was reached
834 * 3) "" indicating that the end of the iob was reached before finding
835 * a complete header line.
837 *********************************************************************/
838 char *get_header(struct iob *iob)
842 header = get_header_line(iob);
844 if ((header == NULL) || (*header == '\0'))
847 * No complete header read yet, tell the client.
852 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
855 * Header spans multiple lines, append the next one.
857 char *continued_header;
859 continued_header = get_header_line(iob);
860 if ((continued_header == NULL) || (*continued_header == '\0'))
863 * No complete header read yet, return what we got.
864 * XXX: Should "unread" header instead.
866 log_error(LOG_LEVEL_INFO,
867 "Failed to read a multi-line header properly: '%s'",
872 if (JB_ERR_OK != string_join(&header, continued_header))
874 log_error(LOG_LEVEL_FATAL,
875 "Out of memory while appending multiple headers.");
879 /* XXX: remove before next stable release. */
880 log_error(LOG_LEVEL_HEADER,
881 "Merged multiple header lines to: '%s'",
886 normalize_lws(header);
893 /*********************************************************************
895 * Function : get_header_line
897 * Description : This (odd) routine will parse the csp->iob
898 * to get the next header line.
901 * 1 : iob = The I/O buffer to parse, usually csp->iob.
903 * Returns : Any one of the following:
905 * 1) a pointer to a dynamically allocated string that contains a header line
906 * 2) NULL indicating that the end of the header was reached
907 * 3) "" indicating that the end of the iob was reached before finding
908 * a complete header line.
910 *********************************************************************/
911 static char *get_header_line(struct iob *iob)
915 if ((iob->cur == NULL)
916 || ((p = strchr(iob->cur, '\n')) == NULL))
918 return(""); /* couldn't find a complete header */
923 ret = strdup(iob->cur);
926 /* FIXME No way to handle error properly */
927 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
933 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
935 /* is this a blank line (i.e. the end of the header) ? */
947 /*********************************************************************
949 * Function : get_header_value
951 * Description : Get the value of a given header from a chained list
952 * of header lines or return NULL if no such header is
953 * present in the list.
956 * 1 : header_list = pointer to list
957 * 2 : header_name = string with name of header to look for.
958 * Trailing colon required, capitalization
961 * Returns : NULL if not found, else value of header
963 *********************************************************************/
964 char *get_header_value(const struct list *header_list, const char *header_name)
966 struct list_entry *cur_entry;
972 length = strlen(header_name);
974 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
978 if (!strncmpic(cur_entry->str, header_name, length))
981 * Found: return pointer to start of value
983 ret = cur_entry->str + length;
984 while (*ret && privoxy_isspace(*ret)) ret++;
998 /*********************************************************************
1000 * Function : scan_headers
1002 * Description : Scans headers, applies tags and updates action bits.
1005 * 1 : csp = Current client state (buffers, headers, etc...)
1007 * Returns : JB_ERR_OK
1009 *********************************************************************/
1010 static jb_err scan_headers(struct client_state *csp)
1012 struct list_entry *h; /* Header */
1013 jb_err err = JB_ERR_OK;
1015 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1017 /* Header crunch()ed in previous run? -> ignore */
1018 if (h->str == NULL) continue;
1019 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1020 err = header_tagger(csp, h->str);
1027 /*********************************************************************
1029 * Function : enforce_header_order
1031 * Description : Enforces a given header order.
1034 * 1 : headers = List of headers to order.
1035 * 2 : ordered_headers = List of ordered header names.
1039 *********************************************************************/
1040 static void enforce_header_order(struct list *headers, const struct list *ordered_headers)
1042 struct list_entry *sorted_header;
1043 struct list new_headers[1];
1044 struct list_entry *header;
1046 init_list(new_headers);
1048 /* The request line is always the first "header" */
1050 assert(NULL != headers->first->str);
1051 enlist(new_headers, headers->first->str);
1052 freez(headers->first->str)
1054 /* Enlist the specified headers in the given order */
1056 for (sorted_header = ordered_headers->first; sorted_header != NULL;
1057 sorted_header = sorted_header->next)
1059 const size_t sorted_header_length = strlen(sorted_header->str);
1060 for (header = headers->first; header != NULL; header = header->next)
1062 /* Header enlisted in previous run? -> ignore */
1063 if (header->str == NULL) continue;
1065 if (0 == strncmpic(sorted_header->str, header->str, sorted_header_length)
1066 && (header->str[sorted_header_length] == ':'))
1068 log_error(LOG_LEVEL_HEADER, "Enlisting sorted header %s", header->str);
1069 if (JB_ERR_OK != enlist(new_headers, header->str))
1071 log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
1078 /* Enlist the rest of the headers behind the ordered ones */
1079 for (header = headers->first; header != NULL; header = header->next)
1081 /* Header enlisted in previous run? -> ignore */
1082 if (header->str == NULL) continue;
1084 log_error(LOG_LEVEL_HEADER,
1085 "Enlisting left-over header %s", header->str);
1086 if (JB_ERR_OK != enlist(new_headers, header->str))
1088 log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
1093 list_remove_all(headers);
1094 list_duplicate(headers, new_headers);
1095 list_remove_all(new_headers);
1101 /*********************************************************************
1105 * Description : add, delete or modify lines in the HTTP header streams.
1106 * On entry, it receives a linked list of headers space
1107 * that was allocated dynamically (both the list nodes
1108 * and the header contents).
1110 * As a side effect it frees the space used by the original
1114 * 1 : csp = Current client state (buffers, headers, etc...)
1115 * 2 : filter_server_headers = Boolean to switch between
1116 * server and header filtering.
1118 * Returns : JB_ERR_OK in case off success, or
1119 * JB_ERR_MEMORY on out-of-memory error.
1121 *********************************************************************/
1122 jb_err sed(struct client_state *csp, int filter_server_headers)
1124 /* XXX: use more descriptive names. */
1125 struct list_entry *p;
1126 const struct parsers *v;
1127 const add_header_func_ptr *f;
1128 jb_err err = JB_ERR_OK;
1132 if (filter_server_headers)
1134 v = server_patterns;
1135 f = add_server_headers;
1136 check_negative_tag_patterns(csp, PATTERN_SPEC_NO_RESPONSE_TAG_PATTERN);
1140 v = client_patterns;
1141 f = add_client_headers;
1142 check_negative_tag_patterns(csp, PATTERN_SPEC_NO_REQUEST_TAG_PATTERN);
1145 while ((err == JB_ERR_OK) && (v->str != NULL))
1147 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1149 /* Header crunch()ed in previous run? -> ignore */
1150 if (p->str == NULL) continue;
1152 /* Does the current parser handle this header? */
1153 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1154 (v->len == CHECK_EVERY_HEADER_REMAINING))
1156 err = v->parser(csp, &(p->str));
1162 /* place additional headers on the csp->headers list */
1163 while ((err == JB_ERR_OK) && (*f))
1169 if (!filter_server_headers && !list_is_empty(csp->config->ordered_client_headers))
1171 enforce_header_order(csp->headers, csp->config->ordered_client_headers);
1178 /*********************************************************************
1180 * Function : update_server_headers
1182 * Description : Updates server headers after the body has been modified.
1185 * 1 : csp = Current client state (buffers, headers, etc...)
1187 * Returns : JB_ERR_OK in case off success, or
1188 * JB_ERR_MEMORY on out-of-memory error.
1190 *********************************************************************/
1191 jb_err update_server_headers(struct client_state *csp)
1193 jb_err err = JB_ERR_OK;
1195 static const struct parsers server_patterns_light[] = {
1196 { "Content-Length:", 15, server_adjust_content_length },
1197 { "Transfer-Encoding:", 18, server_transfer_coding },
1199 { "Content-Encoding:", 17, server_adjust_content_encoding },
1200 #endif /* def FEATURE_ZLIB */
1204 if (strncmpic(csp->http->cmd, "HEAD", 4))
1206 const struct parsers *v;
1207 struct list_entry *p;
1209 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1211 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1213 /* Header crunch()ed in previous run? -> ignore */
1214 if (p->str == NULL) continue;
1216 /* Does the current parser handle this header? */
1217 if (strncmpic(p->str, v->str, v->len) == 0)
1219 err = v->parser(csp, (char **)&(p->str));
1225 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1226 if ((JB_ERR_OK == err)
1227 && (csp->flags & CSP_FLAG_MODIFIED)
1228 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
1229 && !(csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET))
1233 create_content_length_header(csp->content_length, header, sizeof(header));
1234 err = enlist(csp->headers, header);
1235 if (JB_ERR_OK == err)
1237 log_error(LOG_LEVEL_HEADER,
1238 "Content modified with no Content-Length header set. "
1239 "Created: %s.", header);
1240 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
1243 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1245 #ifdef FEATURE_COMPRESSION
1246 if ((JB_ERR_OK == err)
1247 && (csp->flags & CSP_FLAG_BUFFERED_CONTENT_DEFLATED))
1249 err = enlist_unique_header(csp->headers, "Content-Encoding", "deflate");
1250 if (JB_ERR_OK == err)
1252 log_error(LOG_LEVEL_HEADER, "Added header: Content-Encoding: deflate");
1261 /*********************************************************************
1263 * Function : header_tagger
1265 * Description : Executes all text substitutions from applying
1266 * tag actions and saves the result as tag.
1268 * XXX: Shares enough code with filter_header() and
1269 * pcrs_filter_response() to warrant some helper functions.
1272 * 1 : csp = Current client state (buffers, headers, etc...)
1273 * 2 : header = Header that is used as tagger input
1275 * Returns : JB_ERR_OK on success and always succeeds
1277 *********************************************************************/
1278 static jb_err header_tagger(struct client_state *csp, char *header)
1280 int wanted_filter_type;
1281 int multi_action_index;
1284 struct re_filterfile_spec *b;
1285 struct list_entry *tag_name;
1287 const size_t header_length = strlen(header);
1289 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1291 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1292 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1296 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1297 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1300 if (list_is_empty(csp->action->multi[multi_action_index])
1301 || filters_available(csp) == FALSE)
1303 /* Return early if no taggers apply or if none are available. */
1307 /* Execute all applying taggers */
1308 for (tag_name = csp->action->multi[multi_action_index]->first;
1309 NULL != tag_name; tag_name = tag_name->next)
1311 char *modified_tag = NULL;
1313 size_t size = header_length;
1316 b = get_filter(csp, tag_name->str, wanted_filter_type);
1322 joblist = b->joblist;
1324 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1326 if (NULL == joblist)
1328 log_error(LOG_LEVEL_RE_FILTER,
1329 "Tagger %s has empty joblist. Nothing to do.", b->name);
1333 /* execute their pcrs_joblist on the header. */
1334 for (job = joblist; NULL != job; job = job->next)
1336 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1340 /* Success, continue with the modified version. */
1349 /* Tagger doesn't match */
1352 /* Regex failure, log it but continue anyway. */
1353 assert(NULL != header);
1354 log_error(LOG_LEVEL_ERROR,
1355 "Problems with tagger \'%s\' and header \'%s\': %s",
1356 b->name, *header, pcrs_strerror(hits));
1358 freez(modified_tag);
1362 if (b->dynamic) pcrs_free_joblist(joblist);
1364 /* If this tagger matched */
1370 * There is no technical limitation which makes
1371 * it impossible to use empty tags, but I assume
1372 * no one would do it intentionally.
1375 log_error(LOG_LEVEL_INFO,
1376 "Tagger \'%s\' created an empty tag. Ignored.", b->name);
1380 if (!list_contains_item(csp->tags, tag))
1382 if (JB_ERR_OK != enlist(csp->tags, tag))
1384 log_error(LOG_LEVEL_ERROR,
1385 "Insufficient memory to add tag \'%s\', "
1386 "based on tagger \'%s\' and header \'%s\'",
1387 tag, b->name, *header);
1391 char *action_message;
1393 * update the action bits right away, to make
1394 * tagging based on tags set by earlier taggers
1395 * of the same kind possible.
1397 if (update_action_bits_for_tag(csp, tag))
1399 action_message = "Action bits updated accordingly.";
1403 action_message = "No action bits update necessary.";
1406 log_error(LOG_LEVEL_HEADER,
1407 "Tagger \'%s\' added tag \'%s\'. %s",
1408 b->name, tag, action_message);
1413 /* XXX: Is this log-worthy? */
1414 log_error(LOG_LEVEL_HEADER,
1415 "Tagger \'%s\' didn't add tag \'%s\'. Tag already present",
1425 /* here begins the family of parser functions that reformat header lines */
1427 /*********************************************************************
1429 * Function : filter_header
1431 * Description : Executes all text substitutions from all applying
1432 * +(server|client)-header-filter actions on the header.
1433 * Most of the code was copied from pcrs_filter_response,
1434 * including the rather short variable names
1437 * 1 : csp = Current client state (buffers, headers, etc...)
1438 * 2 : header = On input, pointer to header to modify.
1439 * On output, pointer to the modified header, or NULL
1440 * to remove the header. This function frees the
1441 * original string if necessary.
1443 * Returns : JB_ERR_OK on success and always succeeds
1445 *********************************************************************/
1446 static jb_err filter_header(struct client_state *csp, char **header)
1450 size_t size = strlen(*header);
1452 char *newheader = NULL;
1455 struct re_filterfile_spec *b;
1456 struct list_entry *filtername;
1458 int wanted_filter_type;
1459 int multi_action_index;
1461 if (csp->flags & CSP_FLAG_NO_FILTERING)
1466 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1468 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1469 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1473 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1474 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1477 if (list_is_empty(csp->action->multi[multi_action_index])
1478 || filters_available(csp) == FALSE)
1480 /* Return early if no filters apply or if none are available. */
1484 /* Execute all applying header filters */
1485 for (filtername = csp->action->multi[multi_action_index]->first;
1486 filtername != NULL; filtername = filtername->next)
1488 int current_hits = 0;
1491 b = get_filter(csp, filtername->str, wanted_filter_type);
1497 joblist = b->joblist;
1499 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1501 if (NULL == joblist)
1503 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1507 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1508 *header, size, b->name);
1510 /* Apply all jobs from the joblist */
1511 for (job = joblist; NULL != job; job = job->next)
1513 matches = pcrs_execute(job, *header, size, &newheader, &size);
1516 current_hits += matches;
1517 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1519 *header = newheader;
1521 else if (0 == matches)
1523 /* Filter doesn't change header */
1529 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1530 *header, b->name, pcrs_strerror(matches));
1531 if (newheader != NULL)
1533 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1539 if (b->dynamic) pcrs_free_joblist(joblist);
1541 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1542 hits += current_hits;
1546 * Additionally checking for hits is important because if
1547 * the continue hack is triggered, server headers can
1548 * arrive empty to separate multiple heads from each other.
1550 if ((0 == size) && hits)
1552 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
1560 /*********************************************************************
1562 * Function : server_connection
1564 * Description : Makes sure a proper "Connection:" header is
1565 * set and signals connection_header_adder to
1569 * 1 : csp = Current client state (buffers, headers, etc...)
1570 * 2 : header = On input, pointer to header to modify.
1571 * On output, pointer to the modified header, or NULL
1572 * to remove the header. This function frees the
1573 * original string if necessary.
1575 * Returns : JB_ERR_OK on success.
1577 *********************************************************************/
1578 static jb_err server_connection(struct client_state *csp, char **header)
1580 if (!strcmpic(*header, "Connection: keep-alive")
1581 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1582 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
1586 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1587 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1589 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
1592 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE))
1594 log_error(LOG_LEVEL_HEADER,
1595 "Keeping the server header '%s' around.", *header);
1598 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
1600 char *old_header = *header;
1602 *header = strdup_or_die("Connection: close");
1603 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
1608 /* Signal server_connection_adder() to return early. */
1609 csp->flags |= CSP_FLAG_SERVER_CONNECTION_HEADER_SET;
1615 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1616 /*********************************************************************
1618 * Function : server_keep_alive
1620 * Description : Stores the server's keep alive timeout.
1623 * 1 : csp = Current client state (buffers, headers, etc...)
1624 * 2 : header = On input, pointer to header to modify.
1625 * On output, pointer to the modified header, or NULL
1626 * to remove the header. This function frees the
1627 * original string if necessary.
1629 * Returns : JB_ERR_OK.
1631 *********************************************************************/
1632 static jb_err server_keep_alive(struct client_state *csp, char **header)
1634 unsigned int keep_alive_timeout;
1635 const char *timeout_position = strstr(*header, "timeout=");
1637 if ((NULL == timeout_position)
1638 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1640 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1644 if (keep_alive_timeout < csp->server_connection.keep_alive_timeout)
1646 log_error(LOG_LEVEL_HEADER,
1647 "Reducing keep-alive timeout from %u to %u.",
1648 csp->server_connection.keep_alive_timeout, keep_alive_timeout);
1649 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1653 /* XXX: Is this log worthy? */
1654 log_error(LOG_LEVEL_HEADER,
1655 "Server keep-alive timeout is %u. Sticking with %u.",
1656 keep_alive_timeout, csp->server_connection.keep_alive_timeout);
1658 csp->flags |= CSP_FLAG_SERVER_KEEP_ALIVE_TIMEOUT_SET;
1665 /*********************************************************************
1667 * Function : server_proxy_connection
1669 * Description : Figures out whether or not we should add a
1670 * Proxy-Connection header.
1673 * 1 : csp = Current client state (buffers, headers, etc...)
1674 * 2 : header = On input, pointer to header to modify.
1675 * On output, pointer to the modified header, or NULL
1676 * to remove the header. This function frees the
1677 * original string if necessary.
1679 * Returns : JB_ERR_OK.
1681 *********************************************************************/
1682 static jb_err server_proxy_connection(struct client_state *csp, char **header)
1684 csp->flags |= CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET;
1689 /*********************************************************************
1691 * Function : proxy_authentication
1693 * Description : Removes headers that are relevant for proxy
1694 * authentication unless forwarding them has
1695 * been explicitly requested.
1698 * 1 : csp = Current client state (buffers, headers, etc...)
1699 * 2 : header = On input, pointer to header to modify.
1700 * On output, pointer to the modified header, or NULL
1701 * to remove the header. This function frees the
1702 * original string if necessary.
1704 * Returns : JB_ERR_OK.
1706 *********************************************************************/
1707 static jb_err proxy_authentication(struct client_state *csp, char **header)
1709 if ((csp->config->feature_flags &
1710 RUNTIME_FEATURE_FORWARD_PROXY_AUTHENTICATION_HEADERS) == 0) {
1711 log_error(LOG_LEVEL_HEADER,
1712 "Forwarding proxy authentication headers is disabled. Crunching: %s", *header);
1719 /*********************************************************************
1721 * Function : client_keep_alive
1723 * Description : Stores the client's keep alive timeout.
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 client_keep_alive(struct client_state *csp, char **header)
1737 unsigned int keep_alive_timeout;
1738 const char *timeout_position = strstr(*header, ": ");
1740 if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1742 log_error(LOG_LEVEL_HEADER,
1743 "keep-alive support is disabled. Crunching: %s.", *header);
1748 if ((NULL == timeout_position)
1749 || (1 != sscanf(timeout_position, ": %u", &keep_alive_timeout)))
1751 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1755 if (keep_alive_timeout < csp->config->keep_alive_timeout)
1757 log_error(LOG_LEVEL_HEADER,
1758 "Reducing keep-alive timeout from %u to %u.",
1759 csp->config->keep_alive_timeout, keep_alive_timeout);
1760 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1764 /* XXX: Is this log worthy? */
1765 log_error(LOG_LEVEL_HEADER,
1766 "Client keep-alive timeout is %u. Sticking with %u.",
1767 keep_alive_timeout, csp->config->keep_alive_timeout);
1775 /*********************************************************************
1777 * Function : get_content_length
1779 * Description : Gets the content length specified in a
1780 * Content-Length header.
1783 * 1 : header_value = The Content-Length header value.
1784 * 2 : length = Storage to return the value.
1786 * Returns : JB_ERR_OK on success, or
1787 * JB_ERR_PARSE if no value is recognized.
1789 *********************************************************************/
1790 static jb_err get_content_length(const char *header_value, unsigned long long *length)
1793 assert(sizeof(unsigned long long) > 4);
1794 if (1 != sscanf(header_value, "%I64u", length))
1796 if (1 != sscanf(header_value, "%llu", length))
1799 return JB_ERR_PARSE;
1806 /*********************************************************************
1808 * Function : client_save_content_length
1810 * Description : Save the Content-Length sent by the client.
1813 * 1 : csp = Current client state (buffers, headers, etc...)
1814 * 2 : header = On input, pointer to header to modify.
1815 * On output, pointer to the modified header, or NULL
1816 * to remove the header. This function frees the
1817 * original string if necessary.
1819 * Returns : JB_ERR_OK on success, or
1820 * JB_ERR_MEMORY on out-of-memory error.
1822 *********************************************************************/
1823 static jb_err client_save_content_length(struct client_state *csp, char **header)
1825 unsigned long long content_length = 0;
1826 const char *header_value;
1828 assert(*(*header+14) == ':');
1830 header_value = *header + 15;
1831 if (JB_ERR_OK != get_content_length(header_value, &content_length))
1833 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
1838 csp->expected_client_content_length = content_length;
1843 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1847 /*********************************************************************
1849 * Function : client_connection
1851 * Description : Makes sure a proper "Connection:" header is
1852 * set and signals connection_header_adder
1856 * 1 : csp = Current client state (buffers, headers, etc...)
1857 * 2 : header = On input, pointer to header to modify.
1858 * On output, pointer to the modified header, or NULL
1859 * to remove the header. This function frees the
1860 * original string if necessary.
1862 * Returns : JB_ERR_OK on success.
1864 *********************************************************************/
1865 static jb_err client_connection(struct client_state *csp, char **header)
1867 static const char connection_close[] = "Connection: close";
1869 if (!strcmpic(*header, connection_close))
1871 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1872 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING)
1873 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1875 if (!strcmpic(csp->http->ver, "HTTP/1.1"))
1877 log_error(LOG_LEVEL_HEADER,
1878 "Removing \'%s\' to imply keep-alive.", *header);
1881 * While we imply keep-alive to the server,
1882 * we have to remember that the client didn't.
1884 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1888 char *old_header = *header;
1890 *header = strdup_or_die("Connection: keep-alive");
1891 log_error(LOG_LEVEL_HEADER,
1892 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1898 log_error(LOG_LEVEL_HEADER,
1899 "Keeping the client header '%s' around. "
1900 "The connection will not be kept alive.",
1902 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1905 else if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
1906 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1908 log_error(LOG_LEVEL_HEADER,
1909 "Keeping the client header '%s' around. "
1910 "The server connection will be kept alive if possible.",
1912 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1913 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1917 char *old_header = *header;
1919 *header = strdup_or_die(connection_close);
1920 log_error(LOG_LEVEL_HEADER,
1921 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1925 /* Signal client_connection_header_adder() to return early. */
1926 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
1932 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1933 /*********************************************************************
1935 * Function : client_proxy_connection
1937 * Description : Sets the CLIENT_CONNECTION_KEEP_ALIVE flag when
1938 * appropriate and removes the Proxy-Connection
1942 * 1 : csp = Current client state (buffers, headers, etc...)
1943 * 2 : header = On input, pointer to header to modify.
1944 * On output, pointer to the modified header, or NULL
1945 * to remove the header. This function frees the
1946 * original string if necessary.
1948 * Returns : JB_ERR_OK
1950 *********************************************************************/
1951 static jb_err client_proxy_connection(struct client_state *csp, char **header)
1953 if (0 == (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
1954 && (csp->http->ssl == 0)
1955 && (NULL == strstr(*header, "close")))
1957 log_error(LOG_LEVEL_HEADER,
1958 "The client connection can be kept alive due to: %s", *header);
1959 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1961 crumble(csp, header);
1965 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1968 /*********************************************************************
1970 * Function : client_transfer_encoding
1972 * Description : Raise the CSP_FLAG_CHUNKED_CLIENT_BODY flag if
1973 * the request body is "chunked"
1975 * XXX: Currently not called through sed() as we
1976 * need the flag earlier on. Should be fixed.
1979 * 1 : csp = Current client state (buffers, headers, etc...)
1980 * 2 : header = On input, pointer to header to modify.
1981 * On output, pointer to the modified header, or NULL
1982 * to remove the header. This function frees the
1983 * original string if necessary.
1985 * Returns : JB_ERR_OK on success, or
1987 *********************************************************************/
1988 jb_err client_transfer_encoding(struct client_state *csp, char **header)
1990 if (strstr(*header, "chunked"))
1992 csp->flags |= CSP_FLAG_CHUNKED_CLIENT_BODY;
1993 log_error(LOG_LEVEL_HEADER, "Expecting chunked client body");
2000 /*********************************************************************
2002 * Function : crumble
2004 * Description : This is called if a header matches a pattern to "crunch"
2007 * 1 : csp = Current client state (buffers, headers, etc...)
2008 * 2 : header = On input, pointer to header to modify.
2009 * On output, pointer to the modified header, or NULL
2010 * to remove the header. This function frees the
2011 * original string if necessary.
2013 * Returns : JB_ERR_OK on success, or
2014 * JB_ERR_MEMORY on out-of-memory error.
2016 *********************************************************************/
2017 static jb_err crumble(struct client_state *csp, char **header)
2020 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2026 /*********************************************************************
2028 * Function : crunch_server_header
2030 * Description : Crunch server header if it matches a string supplied by the
2031 * user. Called from `sed'.
2034 * 1 : csp = Current client state (buffers, headers, etc...)
2035 * 2 : header = On input, pointer to header to modify.
2036 * On output, pointer to the modified header, or NULL
2037 * to remove the header. This function frees the
2038 * original string if necessary.
2040 * Returns : JB_ERR_OK on success and always succeeds
2042 *********************************************************************/
2043 static jb_err crunch_server_header(struct client_state *csp, char **header)
2045 const char *crunch_pattern;
2047 /* Do we feel like crunching? */
2048 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2050 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2052 /* Is the current header the lucky one? */
2053 if (strstr(*header, crunch_pattern))
2055 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2064 /*********************************************************************
2066 * Function : server_content_type
2068 * Description : Set the content-type for filterable types (text/.*,
2069 * .*xml.*, .*script.* and image/gif) unless filtering has been
2070 * forbidden (CT_TABOO) while parsing earlier headers.
2071 * NOTE: Since text/plain is commonly used by web servers
2072 * for files whose correct type is unknown, we don't
2073 * set CT_TEXT for it.
2076 * 1 : csp = Current client state (buffers, headers, etc...)
2077 * 2 : header = On input, pointer to header to modify.
2078 * On output, pointer to the modified header, or NULL
2079 * to remove the header. This function frees the
2080 * original string if necessary.
2082 * Returns : JB_ERR_OK on success, or
2083 * JB_ERR_MEMORY on out-of-memory error.
2085 *********************************************************************/
2086 static jb_err server_content_type(struct client_state *csp, char **header)
2088 /* Remove header if it isn't the first Content-Type header */
2089 if ((csp->content_type & CT_DECLARED))
2091 if (content_filters_enabled(csp->action))
2094 * Making sure the client interprets the content the same way
2095 * Privoxy did is only relevant if Privoxy modified it.
2097 * Checking for this is "hard" as it's not yet known when
2098 * this function is called, thus go shopping and and just
2099 * check if Privoxy could filter it.
2101 * The main thing is that we don't mess with the headers
2102 * unless the user signalled that it's acceptable.
2104 log_error(LOG_LEVEL_HEADER,
2105 "Multiple Content-Type headers detected. "
2106 "Removing and ignoring: %s",
2114 * Signal that the Content-Type has been set.
2116 csp->content_type |= CT_DECLARED;
2118 if (!(csp->content_type & CT_TABOO))
2121 * XXX: The assumption that text/plain is a sign of
2122 * binary data seems to be somewhat unreasonable nowadays
2123 * and should be dropped after 3.0.8 is out.
2125 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2126 || strstr(*header, "xml")
2127 || strstr(*header, "script"))
2129 csp->content_type |= CT_TEXT;
2131 else if (strstr(*header, "image/gif"))
2133 csp->content_type |= CT_GIF;
2138 * Are we messing with the content type?
2140 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2143 * Make sure the user doesn't accidentally
2144 * change the content type of binary documents.
2146 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2149 *header = strdup_or_die("Content-Type: ");
2150 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2154 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2155 return JB_ERR_MEMORY;
2157 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2161 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2162 "It doesn't look like a content type that should be filtered. "
2163 "Enable force-text-mode if you know what you're doing.", *header);
2171 /*********************************************************************
2173 * Function : server_transfer_coding
2175 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2176 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2177 * - Remove header if body was chunked but has been
2178 * de-chunked for filtering.
2181 * 1 : csp = Current client state (buffers, headers, etc...)
2182 * 2 : header = On input, pointer to header to modify.
2183 * On output, pointer to the modified header, or NULL
2184 * to remove the header. This function frees the
2185 * original string if necessary.
2187 * Returns : JB_ERR_OK on success, or
2188 * JB_ERR_MEMORY on out-of-memory error.
2190 *********************************************************************/
2191 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2194 * Turn off pcrs and gif filtering if body compressed
2196 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2200 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2202 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2203 csp->http->cmd, *header);
2204 #endif /* def FEATURE_ZLIB */
2205 csp->content_type = CT_TABOO;
2209 * Raise flag if body chunked
2211 if (strstr(*header, "chunked"))
2213 csp->flags |= CSP_FLAG_CHUNKED;
2216 * If the body was modified, it has been de-chunked first
2217 * and the header must be removed.
2219 * FIXME: If there is more than one transfer encoding,
2220 * only the "chunked" part should be removed here.
2222 if (csp->flags & CSP_FLAG_MODIFIED)
2224 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2233 /*********************************************************************
2235 * Function : server_content_encoding
2237 * Description : Used to check if the content is compressed, and if
2238 * FEATURE_ZLIB is disabled, filtering is disabled as
2241 * If FEATURE_ZLIB is enabled and the compression type
2242 * supported, the content is marked for decompression.
2244 * XXX: Doesn't properly deal with multiple or with
2245 * unsupported but unknown encodings.
2246 * Is case-sensitive but shouldn't be.
2249 * 1 : csp = Current client state (buffers, headers, etc...)
2250 * 2 : header = On input, pointer to header to modify.
2251 * On output, pointer to the modified header, or NULL
2252 * to remove the header. This function frees the
2253 * original string if necessary.
2255 * Returns : JB_ERR_OK on success, or
2256 * JB_ERR_MEMORY on out-of-memory error.
2258 *********************************************************************/
2259 static jb_err server_content_encoding(struct client_state *csp, char **header)
2262 if (strstr(*header, "sdch"))
2265 * Shared Dictionary Compression over HTTP isn't supported,
2266 * filtering it anyway is pretty much guaranteed to mess up
2269 csp->content_type |= CT_TABOO;
2272 * Log a warning if the user expects the content to be filtered.
2274 if ((csp->rlist != NULL) &&
2275 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2277 log_error(LOG_LEVEL_INFO,
2278 "SDCH-compressed content detected, content filtering disabled. "
2279 "Consider suppressing SDCH offers made by the client.");
2282 else if (strstr(*header, "gzip"))
2284 /* Mark for gzip decompression */
2285 csp->content_type |= CT_GZIP;
2287 else if (strstr(*header, "deflate"))
2289 /* Mark for zlib decompression */
2290 csp->content_type |= CT_DEFLATE;
2292 else if (strstr(*header, "compress"))
2295 * We can't decompress this; therefore we can't filter
2298 csp->content_type |= CT_TABOO;
2300 #else /* !defined(FEATURE_ZLIB) */
2302 * XXX: Using a black list here isn't the right approach.
2304 * In case of SDCH, building with zlib support isn't
2307 if (strstr(*header, "gzip") ||
2308 strstr(*header, "compress") ||
2309 strstr(*header, "deflate") ||
2310 strstr(*header, "sdch"))
2313 * Body is compressed, turn off pcrs and gif filtering.
2315 csp->content_type |= CT_TABOO;
2318 * Log a warning if the user expects the content to be filtered.
2320 if ((csp->rlist != NULL) &&
2321 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2323 log_error(LOG_LEVEL_INFO,
2324 "Compressed content detected, content filtering disabled. "
2325 "Consider recompiling Privoxy with zlib support or "
2326 "enable the prevent-compression action.");
2329 #endif /* defined(FEATURE_ZLIB) */
2337 /*********************************************************************
2339 * Function : server_adjust_content_encoding
2341 * Description : Remove the Content-Encoding header if the
2342 * decompression was successful and the content
2346 * 1 : csp = Current client state (buffers, headers, etc...)
2347 * 2 : header = On input, pointer to header to modify.
2348 * On output, pointer to the modified header, or NULL
2349 * to remove the header. This function frees the
2350 * original string if necessary.
2352 * Returns : JB_ERR_OK on success, or
2353 * JB_ERR_MEMORY on out-of-memory error.
2355 *********************************************************************/
2356 static jb_err server_adjust_content_encoding(struct client_state *csp, char **header)
2358 if ((csp->flags & CSP_FLAG_MODIFIED)
2359 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2362 * We successfully decompressed the content,
2363 * and have to clean the header now, so the
2364 * client no longer expects compressed data.
2366 * XXX: There is a difference between cleaning
2367 * and removing it completely.
2369 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2376 #endif /* defined(FEATURE_ZLIB) */
2379 /*********************************************************************
2381 * Function : server_adjust_content_length
2383 * Description : Adjust Content-Length header if we modified
2387 * 1 : csp = Current client state (buffers, headers, etc...)
2388 * 2 : header = On input, pointer to header to modify.
2389 * On output, pointer to the modified header, or NULL
2390 * to remove the header. This function frees the
2391 * original string if necessary.
2393 * Returns : JB_ERR_OK on success, or
2394 * JB_ERR_MEMORY on out-of-memory error.
2396 *********************************************************************/
2397 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2399 /* Regenerate header if the content was modified. */
2400 if (csp->flags & CSP_FLAG_MODIFIED)
2402 const size_t header_length = 50;
2404 *header = malloc(header_length);
2405 if (*header == NULL)
2407 return JB_ERR_MEMORY;
2409 create_content_length_header(csp->content_length, *header, header_length);
2410 log_error(LOG_LEVEL_HEADER,
2411 "Adjusted Content-Length to %llu", csp->content_length);
2418 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2419 /*********************************************************************
2421 * Function : server_save_content_length
2423 * Description : Save the Content-Length sent by the server.
2426 * 1 : csp = Current client state (buffers, headers, etc...)
2427 * 2 : header = On input, pointer to header to modify.
2428 * On output, pointer to the modified header, or NULL
2429 * to remove the header. This function frees the
2430 * original string if necessary.
2432 * Returns : JB_ERR_OK on success, or
2433 * JB_ERR_MEMORY on out-of-memory error.
2435 *********************************************************************/
2436 static jb_err server_save_content_length(struct client_state *csp, char **header)
2438 unsigned long long content_length = 0;
2439 const char *header_value;
2441 assert(*(*header+14) == ':');
2443 header_value = *header + 15;
2444 if (JB_ERR_OK != get_content_length(header_value, &content_length))
2446 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2451 csp->expected_content_length = content_length;
2452 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
2453 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2458 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2461 /*********************************************************************
2463 * Function : server_content_md5
2465 * Description : Crumble any Content-MD5 headers if the document was
2466 * modified. FIXME: Should we re-compute instead?
2469 * 1 : csp = Current client state (buffers, headers, etc...)
2470 * 2 : header = On input, pointer to header to modify.
2471 * On output, pointer to the modified header, or NULL
2472 * to remove the header. This function frees the
2473 * original string if necessary.
2475 * Returns : JB_ERR_OK on success, or
2476 * JB_ERR_MEMORY on out-of-memory error.
2478 *********************************************************************/
2479 static jb_err server_content_md5(struct client_state *csp, char **header)
2481 if (csp->flags & CSP_FLAG_MODIFIED)
2483 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2491 /*********************************************************************
2493 * Function : server_content_disposition
2495 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2496 * Called from `sed'.
2499 * 1 : csp = Current client state (buffers, headers, etc...)
2500 * 2 : header = On input, pointer to header to modify.
2501 * On output, pointer to the modified header, or NULL
2502 * to remove the header. This function frees the
2503 * original string if necessary.
2505 * Returns : JB_ERR_OK on success, or
2506 * JB_ERR_MEMORY on out-of-memory error.
2508 *********************************************************************/
2509 static jb_err server_content_disposition(struct client_state *csp, char **header)
2514 * Are we messing with the Content-Disposition header?
2516 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2522 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2524 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2527 * Blocking content-disposition header
2529 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2536 * Replacing Content-Disposition header
2539 *header = strdup("Content-Disposition: ");
2540 string_append(header, newval);
2542 if (*header != NULL)
2544 log_error(LOG_LEVEL_HEADER,
2545 "Content-Disposition header crunched and replaced with: %s", *header);
2548 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2552 /*********************************************************************
2554 * Function : server_last_modified
2556 * Description : Changes Last-Modified header to the actual date
2557 * to help hide-if-modified-since.
2558 * Called from `sed'.
2561 * 1 : csp = Current client state (buffers, headers, etc...)
2562 * 2 : header = On input, pointer to header to modify.
2563 * On output, pointer to the modified header, or NULL
2564 * to remove the header. This function frees the
2565 * original string if necessary.
2567 * Returns : JB_ERR_OK on success, or
2568 * JB_ERR_MEMORY on out-of-memory error.
2570 *********************************************************************/
2571 static jb_err server_last_modified(struct client_state *csp, char **header)
2574 time_t last_modified;
2578 * Are we messing with the Last-Modified header?
2580 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2586 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2588 if (0 == strcmpic(newval, "block"))
2591 * Blocking Last-Modified header. Useless but why not.
2593 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2597 else if (0 == strcmpic(newval, "reset-to-request-time"))
2600 * Setting Last-Modified Header to now.
2603 get_http_time(0, buf, sizeof(buf));
2605 *header = strdup("Last-Modified: ");
2606 string_append(header, buf);
2608 if (*header == NULL)
2610 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2614 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2617 else if (0 == strcmpic(newval, "randomize"))
2619 const char *header_time = *header + sizeof("Last-Modified:");
2621 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2623 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2625 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2631 struct tm *timeptr = NULL;
2633 #ifdef HAVE_GMTIME_R
2637 rtime = (long int)difftime(now, last_modified);
2640 long int days, hours, minutes, seconds;
2641 const int negative_delta = (rtime < 0);
2646 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2648 rtime = pick_from_range(rtime);
2653 last_modified += rtime;
2654 #ifdef HAVE_GMTIME_R
2655 timeptr = gmtime_r(&last_modified, &gmt);
2656 #elif defined(MUTEX_LOCKS_AVAILABLE)
2657 privoxy_mutex_lock(&gmtime_mutex);
2658 timeptr = gmtime(&last_modified);
2659 privoxy_mutex_unlock(&gmtime_mutex);
2661 timeptr = gmtime(&last_modified);
2663 if ((NULL == timeptr) || !strftime(newheader,
2664 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
2666 log_error(LOG_LEVEL_ERROR,
2667 "Randomizing '%s' failed. Crunching the header without replacement.",
2674 *header = strdup("Last-Modified: ");
2675 string_append(header, newheader);
2677 if (*header == NULL)
2679 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2680 return JB_ERR_MEMORY;
2683 days = rtime / (3600 * 24);
2684 hours = rtime / 3600 % 24;
2685 minutes = rtime / 60 % 60;
2686 seconds = rtime % 60;
2688 log_error(LOG_LEVEL_HEADER,
2689 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2690 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2691 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2695 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2704 /*********************************************************************
2706 * Function : client_accept_encoding
2708 * Description : Rewrite the client's Accept-Encoding header so that
2709 * if doesn't allow compression, if the action applies.
2710 * Note: For HTTP/1.0 the absence of the header is enough.
2713 * 1 : csp = Current client state (buffers, headers, etc...)
2714 * 2 : header = On input, pointer to header to modify.
2715 * On output, pointer to the modified header, or NULL
2716 * to remove the header. This function frees the
2717 * original string if necessary.
2719 * Returns : JB_ERR_OK on success, or
2720 * JB_ERR_MEMORY on out-of-memory error.
2722 *********************************************************************/
2723 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2725 #ifdef FEATURE_COMPRESSION
2726 if ((csp->config->feature_flags & RUNTIME_FEATURE_COMPRESSION)
2727 && strstr(*header, "deflate"))
2729 csp->flags |= CSP_FLAG_CLIENT_SUPPORTS_DEFLATE;
2732 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2734 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2742 /*********************************************************************
2744 * Function : client_te
2746 * Description : Rewrite the client's TE header so that
2747 * if doesn't allow compression, if the action applies.
2750 * 1 : csp = Current client state (buffers, headers, etc...)
2751 * 2 : header = On input, pointer to header to modify.
2752 * On output, pointer to the modified header, or NULL
2753 * to remove the header. This function frees the
2754 * original string if necessary.
2756 * Returns : JB_ERR_OK on success, or
2757 * JB_ERR_MEMORY on out-of-memory error.
2759 *********************************************************************/
2760 static jb_err client_te(struct client_state *csp, char **header)
2762 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2765 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2772 /*********************************************************************
2774 * Function : client_referrer
2776 * Description : Handle the "referer" config setting properly.
2777 * Called from `sed'.
2780 * 1 : csp = Current client state (buffers, headers, etc...)
2781 * 2 : header = On input, pointer to header to modify.
2782 * On output, pointer to the modified header, or NULL
2783 * to remove the header. This function frees the
2784 * original string if necessary.
2786 * Returns : JB_ERR_OK on success, or
2787 * JB_ERR_MEMORY on out-of-memory error.
2789 *********************************************************************/
2790 static jb_err client_referrer(struct client_state *csp, char **header)
2792 const char *parameter;
2793 /* booleans for parameters we have to check multiple times */
2794 int parameter_conditional_block;
2795 int parameter_conditional_forge;
2797 #ifdef FEATURE_FORCE_LOAD
2799 * Since the referrer can include the prefix even
2800 * if the request itself is non-forced, we must
2801 * clean it unconditionally.
2803 * XXX: strclean is too broad
2805 strclean(*header, FORCE_PREFIX);
2806 #endif /* def FEATURE_FORCE_LOAD */
2808 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2810 /* Nothing left to do */
2814 parameter = csp->action->string[ACTION_STRING_REFERER];
2815 assert(parameter != NULL);
2816 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2817 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2819 if (!parameter_conditional_block && !parameter_conditional_forge)
2822 * As conditional-block and conditional-forge are the only
2823 * parameters that rely on the original referrer, we can
2824 * remove it now for all the others.
2829 if (0 == strcmpic(parameter, "block"))
2831 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2834 else if (parameter_conditional_block || parameter_conditional_forge)
2836 return handle_conditional_hide_referrer_parameter(header,
2837 csp->http->hostport, parameter_conditional_block);
2839 else if (0 == strcmpic(parameter, "forge"))
2841 return create_forged_referrer(header, csp->http->hostport);
2845 /* interpret parameter as user-supplied referer to fake */
2846 return create_fake_referrer(header, parameter);
2851 /*********************************************************************
2853 * Function : client_accept_language
2855 * Description : Handle the "Accept-Language" config setting properly.
2856 * Called from `sed'.
2859 * 1 : csp = Current client state (buffers, headers, etc...)
2860 * 2 : header = On input, pointer to header to modify.
2861 * On output, pointer to the modified header, or NULL
2862 * to remove the header. This function frees the
2863 * original string if necessary.
2865 * Returns : JB_ERR_OK on success, or
2866 * JB_ERR_MEMORY on out-of-memory error.
2868 *********************************************************************/
2869 static jb_err client_accept_language(struct client_state *csp, char **header)
2874 * Are we messing with the Accept-Language?
2876 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2878 /*I don't think so*/
2882 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2884 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2887 * Blocking Accept-Language header
2889 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2896 * Replacing Accept-Language header
2899 *header = strdup("Accept-Language: ");
2900 string_append(header, newval);
2902 if (*header == NULL)
2904 log_error(LOG_LEVEL_ERROR,
2905 "Insufficient memory. Accept-Language header crunched without replacement.");
2909 log_error(LOG_LEVEL_HEADER,
2910 "Accept-Language header crunched and replaced with: %s", *header);
2913 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2917 /*********************************************************************
2919 * Function : crunch_client_header
2921 * Description : Crunch client header if it matches a string supplied by the
2922 * user. Called from `sed'.
2925 * 1 : csp = Current client state (buffers, headers, etc...)
2926 * 2 : header = On input, pointer to header to modify.
2927 * On output, pointer to the modified header, or NULL
2928 * to remove the header. This function frees the
2929 * original string if necessary.
2931 * Returns : JB_ERR_OK on success and always succeeds
2933 *********************************************************************/
2934 static jb_err crunch_client_header(struct client_state *csp, char **header)
2936 const char *crunch_pattern;
2938 /* Do we feel like crunching? */
2939 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2941 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2943 /* Is the current header the lucky one? */
2944 if (strstr(*header, crunch_pattern))
2946 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
2954 /*********************************************************************
2956 * Function : client_uagent
2958 * Description : Handle the "user-agent" config setting properly
2959 * and remember its original value to enable browser
2960 * bug workarounds. Called from `sed'.
2963 * 1 : csp = Current client state (buffers, headers, etc...)
2964 * 2 : header = On input, pointer to header to modify.
2965 * On output, pointer to the modified header, or NULL
2966 * to remove the header. This function frees the
2967 * original string if necessary.
2969 * Returns : JB_ERR_OK on success, or
2970 * JB_ERR_MEMORY on out-of-memory error.
2972 *********************************************************************/
2973 static jb_err client_uagent(struct client_state *csp, char **header)
2977 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
2982 newval = csp->action->string[ACTION_STRING_USER_AGENT];
2989 *header = strdup("User-Agent: ");
2990 string_append(header, newval);
2992 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
2994 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2998 /*********************************************************************
3000 * Function : client_ua
3002 * Description : Handle "ua-" headers properly. Called from `sed'.
3005 * 1 : csp = Current client state (buffers, headers, etc...)
3006 * 2 : header = On input, pointer to header to modify.
3007 * On output, pointer to the modified header, or NULL
3008 * to remove the header. This function frees the
3009 * original string if necessary.
3011 * Returns : JB_ERR_OK on success, or
3012 * JB_ERR_MEMORY on out-of-memory error.
3014 *********************************************************************/
3015 static jb_err client_ua(struct client_state *csp, char **header)
3017 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3019 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3027 /*********************************************************************
3029 * Function : client_from
3031 * Description : Handle the "from" config setting properly.
3032 * Called from `sed'.
3035 * 1 : csp = Current client state (buffers, headers, etc...)
3036 * 2 : header = On input, pointer to header to modify.
3037 * On output, pointer to the modified header, or NULL
3038 * to remove the header. This function frees the
3039 * original string if necessary.
3041 * Returns : JB_ERR_OK on success, or
3042 * JB_ERR_MEMORY on out-of-memory error.
3044 *********************************************************************/
3045 static jb_err client_from(struct client_state *csp, char **header)
3049 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3056 newval = csp->action->string[ACTION_STRING_FROM];
3059 * Are we blocking the e-mail address?
3061 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
3063 log_error(LOG_LEVEL_HEADER, "crunched From!");
3067 log_error(LOG_LEVEL_HEADER, " modified");
3069 *header = strdup("From: ");
3070 string_append(header, newval);
3072 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3076 /*********************************************************************
3078 * Function : client_send_cookie
3080 * Description : Crunches the "cookie" header if necessary.
3081 * Called from `sed'.
3083 * XXX: Stupid name, doesn't send squat.
3086 * 1 : csp = Current client state (buffers, headers, etc...)
3087 * 2 : header = On input, pointer to header to modify.
3088 * On output, pointer to the modified header, or NULL
3089 * to remove the header. This function frees the
3090 * original string if necessary.
3092 * Returns : JB_ERR_OK on success, or
3093 * JB_ERR_MEMORY on out-of-memory error.
3095 *********************************************************************/
3096 static jb_err client_send_cookie(struct client_state *csp, char **header)
3098 if (csp->action->flags & ACTION_CRUNCH_OUTGOING_COOKIES)
3100 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3108 /*********************************************************************
3110 * Function : client_x_forwarded
3112 * Description : Handle the "x-forwarded-for" config setting properly,
3113 * also used in the add_client_headers list. Called from `sed'.
3116 * 1 : csp = Current client state (buffers, headers, etc...)
3117 * 2 : header = On input, pointer to header to modify.
3118 * On output, pointer to the modified header, or NULL
3119 * to remove the header. This function frees the
3120 * original string if necessary.
3122 * Returns : JB_ERR_OK on success, or
3123 * JB_ERR_MEMORY on out-of-memory error.
3125 *********************************************************************/
3126 jb_err client_x_forwarded(struct client_state *csp, char **header)
3128 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
3130 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
3132 if (0 == strcmpic(parameter, "block"))
3135 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3137 else if (0 == strcmpic(parameter, "add"))
3139 string_append(header, ", ");
3140 string_append(header, csp->ip_addr_str);
3142 if (*header == NULL)
3144 return JB_ERR_MEMORY;
3146 log_error(LOG_LEVEL_HEADER,
3147 "Appended client IP address to %s", *header);
3148 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
3152 log_error(LOG_LEVEL_FATAL,
3153 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
3161 /*********************************************************************
3163 * Function : client_max_forwards
3165 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3166 * from the value of the Max-Forwards header field.
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 static jb_err client_max_forwards(struct client_state *csp, char **header)
3183 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3184 (0 == strcmpic(csp->http->gpc, "options")))
3186 assert(*(*header+12) == ':');
3187 if (1 == sscanf(*header+12, ": %d", &max_forwards))
3189 if (max_forwards > 0)
3191 snprintf(*header, strlen(*header)+1, "Max-Forwards: %d", --max_forwards);
3192 log_error(LOG_LEVEL_HEADER,
3193 "Max-Forwards value for %s request reduced to %d.",
3194 csp->http->gpc, max_forwards);
3196 else if (max_forwards < 0)
3198 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3204 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3213 /*********************************************************************
3215 * Function : client_host
3217 * Description : If the request URI did not contain host and
3218 * port information, parse and evaluate the Host
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_host(struct client_state *csp, char **header)
3236 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3237 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3240 p = strdup_or_die((*header)+6);
3242 q = strdup_or_die(p);
3244 freez(csp->http->hostport);
3245 csp->http->hostport = p;
3246 freez(csp->http->host);
3247 csp->http->host = q;
3248 q = strchr(csp->http->host, ':');
3251 /* Terminate hostname and evaluate port string */
3253 csp->http->port = atoi(q);
3257 csp->http->port = csp->http->ssl ? 443 : 80;
3260 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3261 csp->http->hostport, csp->http->host, csp->http->port);
3264 /* Signal client_host_adder() to return right away */
3265 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3271 /*********************************************************************
3273 * Function : client_if_modified_since
3275 * Description : Remove or modify the If-Modified-Since header.
3278 * 1 : csp = Current client state (buffers, headers, etc...)
3279 * 2 : header = On input, pointer to header to modify.
3280 * On output, pointer to the modified header, or NULL
3281 * to remove the header. This function frees the
3282 * original string if necessary.
3284 * Returns : JB_ERR_OK on success, or
3285 * JB_ERR_MEMORY on out-of-memory error.
3287 *********************************************************************/
3288 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3291 #ifdef HAVE_GMTIME_R
3294 struct tm *timeptr = NULL;
3299 if (0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3302 * The client got an error message because of a temporary problem,
3303 * the problem is gone and the client now tries to revalidate our
3304 * error message on the real server. The revalidation would always
3305 * end with the transmission of the whole document and there is
3306 * no need to expose the bogus If-Modified-Since header.
3308 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3311 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3313 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3315 if ((0 == strcmpic(newval, "block")))
3317 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3320 else /* add random value */
3322 const char *header_time = *header + sizeof("If-Modified-Since:");
3324 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3326 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3331 long int hours, minutes, seconds;
3332 long int rtime = strtol(newval, &endptr, 0);
3333 const int negative_range = (rtime < 0);
3337 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3338 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3344 rtime = pick_from_range(rtime);
3348 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3351 tm += rtime * (negative_range ? -1 : 1);
3352 #ifdef HAVE_GMTIME_R
3353 timeptr = gmtime_r(&tm, &gmt);
3354 #elif defined(MUTEX_LOCKS_AVAILABLE)
3355 privoxy_mutex_lock(&gmtime_mutex);
3356 timeptr = gmtime(&tm);
3357 privoxy_mutex_unlock(&gmtime_mutex);
3359 timeptr = gmtime(&tm);
3361 if ((NULL == timeptr) || !strftime(newheader,
3362 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
3364 log_error(LOG_LEVEL_ERROR,
3365 "Randomizing '%s' failed. Crunching the header without replacement.",
3372 *header = strdup("If-Modified-Since: ");
3373 string_append(header, newheader);
3375 if (*header == NULL)
3377 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3378 return JB_ERR_MEMORY;
3381 hours = rtime / 3600;
3382 minutes = rtime / 60 % 60;
3383 seconds = rtime % 60;
3385 log_error(LOG_LEVEL_HEADER,
3386 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3387 *header, (negative_range) ? "subtracted" : "added", hours,
3388 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3389 seconds, (seconds == 1) ? ")" : "s)");
3398 /*********************************************************************
3400 * Function : client_if_none_match
3402 * Description : Remove the If-None-Match header.
3405 * 1 : csp = Current client state (buffers, headers, etc...)
3406 * 2 : header = On input, pointer to header to modify.
3407 * On output, pointer to the modified header, or NULL
3408 * to remove the header. This function frees the
3409 * original string if necessary.
3411 * Returns : JB_ERR_OK on success, or
3412 * JB_ERR_MEMORY on out-of-memory error.
3414 *********************************************************************/
3415 static jb_err client_if_none_match(struct client_state *csp, char **header)
3417 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3419 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3427 /*********************************************************************
3429 * Function : client_x_filter
3431 * Description : Disables filtering if the client set "X-Filter: No".
3432 * Called from `sed'.
3435 * 1 : csp = Current client state (buffers, headers, etc...)
3436 * 2 : header = On input, pointer to header to modify.
3437 * On output, pointer to the modified header, or NULL
3438 * to remove the header. This function frees the
3439 * original string if necessary.
3441 * Returns : JB_ERR_OK on success
3443 *********************************************************************/
3444 jb_err client_x_filter(struct client_state *csp, char **header)
3446 if (0 == strcmpic(*header, "X-Filter: No"))
3448 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3450 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3454 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3456 log_error(LOG_LEVEL_HEADER,
3457 "force-text-mode overruled the client's request to fetch without filtering!");
3461 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3462 csp->flags |= CSP_FLAG_NO_FILTERING;
3463 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3465 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3473 /*********************************************************************
3475 * Function : client_range
3477 * Description : Removes Range, Request-Range and If-Range headers if
3478 * content filtering is enabled and the range doesn't
3481 * If the client's version of the document has been
3482 * altered by Privoxy, the server could interpret the
3483 * range differently than the client intended in which
3484 * case the user could end up with corrupted content.
3486 * If the range starts at byte 0 this isn't an issue
3487 * so the header can pass. Partial requests like this
3488 * are used to render preview images for videos without
3489 * downloading the whole video.
3491 * While HTTP doesn't require that range requests are
3492 * honoured and the client could simply abort the download
3493 * after receiving a sufficient amount of data, various
3494 * clients don't handle complete responses to range
3495 * requests gracefully and emit misleading error messages
3499 * 1 : csp = Current client state (buffers, headers, etc...)
3500 * 2 : header = On input, pointer to header to modify.
3501 * On output, pointer to the modified header, or NULL
3502 * to remove the header. This function frees the
3503 * original string if necessary.
3505 * Returns : JB_ERR_OK
3507 *********************************************************************/
3508 static jb_err client_range(struct client_state *csp, char **header)
3510 if (content_filters_enabled(csp->action)
3511 && (0 != strncmpic(strstr(*header, ":"), ": bytes=0-", 10)))
3513 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3514 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3521 /* the following functions add headers directly to the header list */
3523 /*********************************************************************
3525 * Function : client_host_adder
3527 * Description : Adds the Host: header field if it is missing.
3528 * Called from `sed'.
3531 * 1 : csp = Current client state (buffers, headers, etc...)
3533 * Returns : JB_ERR_OK on success, or
3534 * JB_ERR_MEMORY on out-of-memory error.
3536 *********************************************************************/
3537 static jb_err client_host_adder(struct client_state *csp)
3542 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3544 /* Header already set by the client, nothing to do. */
3548 if (!csp->http->hostport || !*(csp->http->hostport))
3550 /* XXX: When does this happen and why is it OK? */
3551 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3556 * remove 'user:pass@' from 'proto://user:pass@host'
3558 if ((p = strchr( csp->http->hostport, '@')) != NULL)
3564 p = csp->http->hostport;
3567 /* XXX: Just add it, we already made sure that it will be unique */
3568 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3569 err = enlist_unique_header(csp->headers, "Host", p);
3575 /*********************************************************************
3577 * Function : client_xtra_adder
3579 * Description : Used in the add_client_headers list. Called from `sed'.
3582 * 1 : csp = Current client state (buffers, headers, etc...)
3584 * Returns : JB_ERR_OK on success, or
3585 * JB_ERR_MEMORY on out-of-memory error.
3587 *********************************************************************/
3588 static jb_err client_xtra_adder(struct client_state *csp)
3590 struct list_entry *lst;
3593 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3594 lst ; lst = lst->next)
3596 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3597 err = enlist(csp->headers, lst->str);
3609 /*********************************************************************
3611 * Function : client_x_forwarded_for_adder
3613 * Description : Used in the add_client_headers list. Called from `sed'.
3616 * 1 : csp = Current client state (buffers, headers, etc...)
3618 * Returns : JB_ERR_OK on success, or
3619 * JB_ERR_MEMORY on out-of-memory error.
3621 *********************************************************************/
3622 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3624 char *header = NULL;
3627 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3628 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3629 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3632 * If we aren't adding X-Forwarded-For headers,
3633 * or we already appended an existing X-Forwarded-For
3634 * header, there's nothing left to do here.
3639 header = strdup("X-Forwarded-For: ");
3640 string_append(&header, csp->ip_addr_str);
3644 return JB_ERR_MEMORY;
3647 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3648 err = enlist(csp->headers, header);
3655 /*********************************************************************
3657 * Function : server_connection_adder
3659 * Description : Adds an appropriate "Connection:" header to csp->headers
3660 * unless the header was already present. Called from `sed'.
3663 * 1 : csp = Current client state (buffers, headers, etc...)
3665 * Returns : JB_ERR_OK on success, or
3666 * JB_ERR_MEMORY on out-of-memory error.
3668 *********************************************************************/
3669 static jb_err server_connection_adder(struct client_state *csp)
3671 const unsigned int flags = csp->flags;
3672 const char *response_status_line = csp->headers->first->str;
3673 static const char connection_close[] = "Connection: close";
3675 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3676 && (flags & CSP_FLAG_SERVER_CONNECTION_HEADER_SET))
3682 * XXX: if we downgraded the response, this check will fail.
3684 if ((csp->config->feature_flags &
3685 RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3686 && (NULL != response_status_line)
3687 && !strncmpic(response_status_line, "HTTP/1.1", 8)
3688 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3689 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3693 log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
3694 "without Connection header implies keep-alive.");
3695 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
3699 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3701 return enlist(csp->headers, connection_close);
3705 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3706 /*********************************************************************
3708 * Function : server_proxy_connection_adder
3710 * Description : Adds a "Proxy-Connection: keep-alive" header to
3711 * csp->headers when appropriate.
3714 * 1 : csp = Current client state (buffers, headers, etc...)
3716 * Returns : JB_ERR_OK on success, or
3717 * JB_ERR_MEMORY on out-of-memory error.
3719 *********************************************************************/
3720 static jb_err server_proxy_connection_adder(struct client_state *csp)
3722 static const char proxy_connection_header[] = "Proxy-Connection: keep-alive";
3723 jb_err err = JB_ERR_OK;
3725 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
3726 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3727 && !(csp->flags & CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET)
3728 && ((csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET)
3729 || (csp->flags & CSP_FLAG_CHUNKED)))
3731 log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
3732 err = enlist(csp->headers, proxy_connection_header);
3737 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3740 /*********************************************************************
3742 * Function : client_connection_header_adder
3744 * Description : Adds a proper "Connection:" header to csp->headers
3745 * unless the header was already present. Called from `sed'.
3748 * 1 : csp = Current client state (buffers, headers, etc...)
3750 * Returns : JB_ERR_OK on success, or
3751 * JB_ERR_MEMORY on out-of-memory error.
3753 *********************************************************************/
3754 static jb_err client_connection_header_adder(struct client_state *csp)
3756 static const char connection_close[] = "Connection: close";
3758 if (!(csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3759 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
3764 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3765 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3766 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3767 && (csp->http->ssl == 0)
3768 && !strcmpic(csp->http->ver, "HTTP/1.1"))
3770 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
3773 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3775 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3777 return enlist(csp->headers, connection_close);
3781 /*********************************************************************
3783 * Function : server_http
3785 * Description : - Save the HTTP Status into csp->http->status
3786 * - Set CT_TABOO to prevent filtering if the answer
3787 * is a partial range (HTTP status 206)
3788 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3792 * 1 : csp = Current client state (buffers, headers, etc...)
3793 * 2 : header = On input, pointer to header to modify.
3794 * On output, pointer to the modified header, or NULL
3795 * to remove the header. This function frees the
3796 * original string if necessary.
3798 * Returns : JB_ERR_OK on success, or
3799 * JB_ERR_MEMORY on out-of-memory error.
3801 *********************************************************************/
3802 static jb_err server_http(struct client_state *csp, char **header)
3804 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3805 if (csp->http->status == 206)
3807 csp->content_type = CT_TABOO;
3810 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3812 /* XXX: Should we do a real validity check here? */
3813 if (strlen(*header) > 8)
3816 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3821 * XXX: Should we block the request or
3822 * enlist a valid status code line here?
3824 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3825 "Downgrading to HTTP/1.0 impossible.");
3832 /*********************************************************************
3834 * Function : add_cooky_expiry_date
3836 * Description : Adds a cookie expiry date to a string.
3839 * 1 : cookie = On input, pointer to cookie to modify.
3840 * On output, pointer to the modified header.
3841 * The original string is freed.
3842 * 2 : lifetime = Seconds the cookie should be valid
3846 *********************************************************************/
3847 static void add_cookie_expiry_date(char **cookie, time_t lifetime)
3850 struct tm *timeptr = NULL;
3851 time_t expiry_date = time(NULL) + lifetime;
3852 #ifdef HAVE_GMTIME_R
3855 timeptr = gmtime_r(&expiry_date, &gmt);
3856 #elif defined(MUTEX_LOCKS_AVAILABLE)
3857 privoxy_mutex_lock(&gmtime_mutex);
3858 timeptr = gmtime(&expiry_date);
3859 privoxy_mutex_unlock(&gmtime_mutex);
3861 timeptr = gmtime(&expiry_date);
3864 if (NULL == timeptr)
3866 log_error(LOG_LEVEL_FATAL,
3867 "Failed to get the time in add_cooky_expiry_date()");
3869 strftime(tmp, sizeof(tmp), "; expires=%a, %d-%b-%Y %H:%M:%S GMT", timeptr);
3870 if (JB_ERR_OK != string_append(cookie, tmp))
3872 log_error(LOG_LEVEL_FATAL, "Out of memory in add_cooky_expiry()");
3877 /*********************************************************************
3879 * Function : server_set_cookie
3881 * Description : Handle the server "cookie" header properly.
3882 * Crunch, accept or rewrite it to a session cookie.
3883 * Called from `sed'.
3886 * 1 : csp = Current client state (buffers, headers, etc...)
3887 * 2 : header = On input, pointer to header to modify.
3888 * On output, pointer to the modified header, or NULL
3889 * to remove the header. This function frees the
3890 * original string if necessary.
3892 * Returns : JB_ERR_OK on success, or
3893 * JB_ERR_MEMORY on out-of-memory error.
3895 *********************************************************************/
3896 static jb_err server_set_cookie(struct client_state *csp, char **header)
3898 if ((csp->action->flags & ACTION_CRUNCH_INCOMING_COOKIES) != 0)
3900 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3903 else if ((0 != (csp->action->flags & ACTION_SESSION_COOKIES_ONLY))
3904 || (0 != (csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME)))
3908 long cookie_lifetime = 0;
3911 NO_EXPIRY_DATE_SPECIFIED,
3912 EXPIRY_DATE_ACCEPTABLE,
3913 EXPIRY_DATE_UNACCEPTABLE
3914 } expiry_date_status = NO_EXPIRY_DATE_SPECIFIED;
3916 /* A variable to store the tag we're working on */
3919 /* Skip "Set-Cookie:" (11 characters) in header */
3920 cur_tag = *header + 11;
3922 /* skip whitespace between "Set-Cookie:" and value */
3923 while (*cur_tag && privoxy_isspace(*cur_tag))
3930 if ((csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME) != 0)
3932 const char *param = csp->action->string[ACTION_STRING_LIMIT_COOKIE_LIFETIME];
3934 cookie_lifetime = strtol(param, NULL, 0);
3935 if (cookie_lifetime < 0)
3937 log_error(LOG_LEVEL_FATAL, "Invalid cookie lifetime limit: %s", param);
3939 cookie_lifetime *= 60;
3942 /* Loop through each tag in the cookie */
3946 char *next_tag = strchr(cur_tag, ';');
3947 if (next_tag != NULL)
3949 /* Skip the ';' character itself */
3952 /* skip whitespace ";" and start of tag */
3953 while (*next_tag && privoxy_isspace(*next_tag))
3960 /* "Next tag" is the end of the string */
3961 next_tag = cur_tag + strlen(cur_tag);
3965 * Check the expiration date to see
3966 * if the cookie is still valid, if yes,
3967 * rewrite it to a session cookie.
3969 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
3971 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
3973 if ((expiration_date[0] == '"')
3974 && (expiration_date[1] != '\0'))
3977 * Skip quotation mark. RFC 2109 10.1.2 seems to hint
3978 * that the expiration date isn't supposed to be quoted,
3979 * but some servers do it anyway.
3984 /* Did we detect the date properly? */
3985 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
3988 * Nope, treat it as if it was still valid.
3990 * XXX: Should we remove the whole cookie instead?
3992 log_error(LOG_LEVEL_ERROR,
3993 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
3994 string_move(cur_tag, next_tag);
3995 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4000 * Yes. Check if the cookie is still valid.
4002 * If the cookie is already expired it's probably
4003 * a delete cookie and even if it isn't, the browser
4004 * will discard it anyway.
4008 * XXX: timegm() isn't available on some AmigaOS
4009 * versions and our replacement doesn't work.
4011 * Our options are to either:
4013 * - disable session-cookies-only completely if timegm
4016 * - to simply remove all expired tags, like it has
4017 * been done until Privoxy 3.0.6 and to live with
4018 * the consequence that it can cause login/logout
4019 * problems on servers that don't validate their
4020 * input properly, or
4022 * - to replace it with mktime in which
4023 * case there is a slight chance of valid cookies
4024 * passing as already expired.
4026 * This is the way it's currently done and it's not
4027 * as bad as it sounds. If the missing GMT offset is
4028 * enough to change the result of the expiration check
4029 * the cookie will be only valid for a few hours
4030 * anyway, which in many cases will be shorter
4031 * than a browser session.
4033 if (cookie_time < now)
4035 log_error(LOG_LEVEL_HEADER,
4036 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4037 /* Just in case some clown sets more then one expiration date */
4039 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4041 else if ((cookie_lifetime != 0) && (cookie_time < (now + cookie_lifetime)))
4043 log_error(LOG_LEVEL_HEADER, "Cookie \'%s\' can pass unmodified. "
4044 "Its lifetime is below the limit.", *header);
4045 /* Just in case some clown sets more then one expiration date */
4047 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4052 * Still valid, delete expiration date by copying
4053 * the rest of the string over it.
4055 string_move(cur_tag, next_tag);
4057 /* That changed the header, need to issue a log message */
4058 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4061 * Note that the next tag has now been moved to *cur_tag,
4062 * so we do not need to update the cur_tag pointer.
4070 /* Move on to next cookie tag */
4075 if (expiry_date_status != EXPIRY_DATE_ACCEPTABLE)
4077 assert(NULL != *header);
4078 if (cookie_lifetime != 0)
4080 add_cookie_expiry_date(header, cookie_lifetime);
4081 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to: %s", *header);
4083 else if (expiry_date_status != NO_EXPIRY_DATE_SPECIFIED)
4085 log_error(LOG_LEVEL_HEADER,
4086 "Cookie rewritten to a temporary one: %s", *header);
4095 #ifdef FEATURE_FORCE_LOAD
4096 /*********************************************************************
4098 * Function : strclean
4100 * Description : In-Situ-Eliminate all occurrences of substring in
4104 * 1 : string = string to clean
4105 * 2 : substring = substring to eliminate
4107 * Returns : Number of eliminations
4109 *********************************************************************/
4110 int strclean(char *string, const char *substring)
4116 len = strlen(substring);
4118 while((pos = strstr(string, substring)) != NULL)
4125 while (*p++ != '\0');
4132 #endif /* def FEATURE_FORCE_LOAD */
4135 /*********************************************************************
4137 * Function : parse_header_time
4139 * Description : Parses time formats used in HTTP header strings
4140 * to get the numerical respresentation.
4143 * 1 : header_time = HTTP header time as string.
4144 * 2 : result = storage for header_time in seconds
4146 * Returns : JB_ERR_OK if the time format was recognized, or
4147 * JB_ERR_PARSE otherwise.
4149 *********************************************************************/
4150 static jb_err parse_header_time(const char *header_time, time_t *result)
4154 * Checking for two-digit years first in an
4155 * attempt to work around GNU libc's strptime()
4156 * reporting negative year values when using %Y.
4158 static const char time_formats[][22] = {
4159 /* Tue, 02-Jun-37 20:00:00 */
4160 "%a, %d-%b-%y %H:%M:%S",
4161 /* Tue, 02 Jun 2037 20:00:00 */
4162 "%a, %d %b %Y %H:%M:%S",
4163 /* Tue, 02-Jun-2037 20:00:00 */
4164 "%a, %d-%b-%Y %H:%M:%S",
4165 /* Tuesday, 02-Jun-2037 20:00:00 */
4166 "%A, %d-%b-%Y %H:%M:%S",
4167 /* Tuesday Jun 02 20:00:00 2037 */
4168 "%A %b %d %H:%M:%S %Y"
4172 for (i = 0; i < SZ(time_formats); i++)
4175 * Zero out gmt to prevent time zone offsets.
4176 * Documented to be required for GNU libc.
4178 memset(&gmt, 0, sizeof(gmt));
4180 if (NULL != strptime(header_time, time_formats[i], &gmt))
4182 /* Sanity check for GNU libc. */
4183 if (gmt.tm_year < 0)
4185 log_error(LOG_LEVEL_HEADER,
4186 "Failed to parse '%s' using '%s'. Moving on.",
4187 header_time, time_formats[i]);
4190 *result = timegm(&gmt);
4192 #ifdef FEATURE_STRPTIME_SANITY_CHECKS
4194 * Verify that parsing the date recreated from the first
4195 * parse operation gets the previous result. If it doesn't,
4196 * either strptime() or strftime() are malfunctioning.
4198 * We could string-compare the recreated date with the original
4199 * header date, but this leads to false positives as strptime()
4200 * may let %a accept all day formats while strftime() will only
4204 char recreated_date[100];
4208 tm = gmtime(result);
4209 strftime(recreated_date, sizeof(recreated_date), time_formats[i], tm);
4210 memset(&gmt, 0, sizeof(gmt));
4211 if (NULL == strptime(recreated_date, time_formats[i], &gmt))
4213 log_error(LOG_LEVEL_ERROR,
4214 "Failed to parse '%s' generated with '%s' to recreate '%s'.",
4215 recreated_date, time_formats[i], header_time);
4218 result2 = timegm(&gmt);
4219 if (*result != result2)
4221 log_error(LOG_LEVEL_ERROR, "strftime() and strptime() disagree. "
4222 "Format: '%s'. In: '%s', out: '%s'. %d != %d. Rejecting.",
4223 time_formats[i], header_time, recreated_date, *result, result2);
4233 return JB_ERR_PARSE;
4238 /*********************************************************************
4240 * Function : get_destination_from_headers
4242 * Description : Parse the "Host:" header to get the request's destination.
4243 * Only needed if the client's request was forcefully
4244 * redirected into Privoxy.
4246 * Code mainly copied from client_host() which is currently
4247 * run too late for this purpose.
4250 * 1 : headers = List of headers (one of them hopefully being
4251 * the "Host:" header)
4252 * 2 : http = storage for the result (host, port and hostport).
4254 * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
4255 * JB_ERR_PARSE if the host header couldn't be found,
4256 * JB_ERR_OK otherwise.
4258 *********************************************************************/
4259 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4265 host = get_header_value(headers, "Host:");
4269 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4270 return JB_ERR_PARSE;
4273 p = strdup_or_die(host);
4275 q = strdup_or_die(p);
4277 freez(http->hostport);
4281 q = strchr(http->host, ':');
4284 /* Terminate hostname and evaluate port string */
4286 http->port = atoi(q);
4290 http->port = http->ssl ? 443 : 80;
4293 /* Rebuild request URL */
4295 http->url = strdup(http->ssl ? "https://" : "http://");
4296 string_append(&http->url, http->hostport);
4297 string_append(&http->url, http->path);
4298 if (http->url == NULL)
4300 return JB_ERR_MEMORY;
4303 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4311 /*********************************************************************
4313 * Function : create_forged_referrer
4315 * Description : Helper for client_referrer to forge a referer as
4316 * 'http://hostname[:port]/' to fool stupid
4317 * checks for in-site links
4320 * 1 : header = Pointer to header pointer
4321 * 2 : hostport = Host and optionally port as string
4323 * Returns : JB_ERR_OK in case of success, or
4324 * JB_ERR_MEMORY in case of memory problems.
4326 *********************************************************************/
4327 static jb_err create_forged_referrer(char **header, const char *hostport)
4329 assert(NULL == *header);
4331 *header = strdup("Referer: http://");
4332 string_append(header, hostport);
4333 string_append(header, "/");
4335 if (NULL == *header)
4337 return JB_ERR_MEMORY;
4340 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4347 /*********************************************************************
4349 * Function : create_fake_referrer
4351 * Description : Helper for client_referrer to create a fake referrer
4352 * based on a string supplied by the user.
4355 * 1 : header = Pointer to header pointer
4356 * 2 : hosthost = Referrer to fake
4358 * Returns : JB_ERR_OK in case of success, or
4359 * JB_ERR_MEMORY in case of memory problems.
4361 *********************************************************************/
4362 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4364 assert(NULL == *header);
4366 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4368 log_error(LOG_LEVEL_HEADER,
4369 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4371 *header = strdup("Referer: ");
4372 string_append(header, fake_referrer);
4374 if (NULL == *header)
4376 return JB_ERR_MEMORY;
4379 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4386 /*********************************************************************
4388 * Function : handle_conditional_hide_referrer_parameter
4390 * Description : Helper for client_referrer to crunch or forge
4391 * the referrer header if the host has changed.
4394 * 1 : header = Pointer to header pointer
4395 * 2 : host = The target host (may include the port)
4396 * 3 : parameter_conditional_block = Boolean to signal
4397 * if we're in conditional-block mode. If not set,
4398 * we're in conditional-forge mode.
4400 * Returns : JB_ERR_OK in case of success, or
4401 * JB_ERR_MEMORY in case of memory problems.
4403 *********************************************************************/
4404 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4405 const char *host, const int parameter_conditional_block)
4407 char *referer = strdup_or_die(*header);
4408 const size_t hostlength = strlen(host);
4409 const char *referer_url = NULL;
4411 /* referer begins with 'Referer: http[s]://' */
4412 if ((hostlength+17) < strlen(referer))
4415 * Shorten referer to make sure the referer is blocked
4416 * if www.example.org/www.example.com-shall-see-the-referer/
4417 * links to www.example.com/
4419 referer[hostlength+17] = '\0';
4421 referer_url = strstr(referer, "http://");
4422 if ((NULL == referer_url) || (NULL == strstr(referer_url, host)))
4424 /* Host has changed, Referer is invalid or a https URL. */
4425 if (parameter_conditional_block)
4427 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4434 return create_forged_referrer(header, host);
4444 /*********************************************************************
4446 * Function : create_content_length_header
4448 * Description : Creates a Content-Length header.
4451 * 1 : content_length = The content length to be used in the header.
4452 * 2 : header = Allocated space to safe the header.
4453 * 3 : buffer_length = The length of the allocated space.
4457 *********************************************************************/
4458 static void create_content_length_header(unsigned long long content_length,
4459 char *header, size_t buffer_length)
4461 snprintf(header, buffer_length, "Content-Length: %llu", content_length);
4465 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
4466 /*********************************************************************
4468 * Function : get_expected_content_length
4470 * Description : Figures out the content length from a list of headers.
4473 * 1 : headers = List of headers
4475 * Returns : Number of bytes to expect
4477 *********************************************************************/
4478 unsigned long long get_expected_content_length(struct list *headers)
4480 const char *content_length_header;
4481 unsigned long long content_length = 0;
4483 content_length_header = get_header_value(headers, "Content-Length:");
4484 if (content_length_header != NULL)
4486 if (JB_ERR_OK != get_content_length(content_length_header, &content_length))
4488 log_error(LOG_LEVEL_ERROR,
4489 "Failed to get the Content-Length in %s", content_length_header);
4490 /* XXX: The header will be removed later on */
4495 return content_length;