1 const char parsers_rcs[] = "$Id: parsers.c,v 1.277 2013/04/23 09:43:25 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);
1100 /*********************************************************************
1104 * Description : add, delete or modify lines in the HTTP header streams.
1105 * On entry, it receives a linked list of headers space
1106 * that was allocated dynamically (both the list nodes
1107 * and the header contents).
1109 * As a side effect it frees the space used by the original
1113 * 1 : csp = Current client state (buffers, headers, etc...)
1114 * 2 : filter_server_headers = Boolean to switch between
1115 * server and header filtering.
1117 * Returns : JB_ERR_OK in case off success, or
1118 * JB_ERR_MEMORY on out-of-memory error.
1120 *********************************************************************/
1121 jb_err sed(struct client_state *csp, int filter_server_headers)
1123 /* XXX: use more descriptive names. */
1124 struct list_entry *p;
1125 const struct parsers *v;
1126 const add_header_func_ptr *f;
1127 jb_err err = JB_ERR_OK;
1129 if (filter_server_headers)
1131 v = server_patterns;
1132 f = add_server_headers;
1136 v = client_patterns;
1137 f = add_client_headers;
1142 while ((err == JB_ERR_OK) && (v->str != NULL))
1144 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1146 /* Header crunch()ed in previous run? -> ignore */
1147 if (p->str == NULL) continue;
1149 /* Does the current parser handle this header? */
1150 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1151 (v->len == CHECK_EVERY_HEADER_REMAINING))
1153 err = v->parser(csp, &(p->str));
1159 /* place additional headers on the csp->headers list */
1160 while ((err == JB_ERR_OK) && (*f))
1166 if (!filter_server_headers && !list_is_empty(csp->config->ordered_client_headers))
1168 enforce_header_order(csp->headers, csp->config->ordered_client_headers);
1175 /*********************************************************************
1177 * Function : update_server_headers
1179 * Description : Updates server headers after the body has been modified.
1182 * 1 : csp = Current client state (buffers, headers, etc...)
1184 * Returns : JB_ERR_OK in case off success, or
1185 * JB_ERR_MEMORY on out-of-memory error.
1187 *********************************************************************/
1188 jb_err update_server_headers(struct client_state *csp)
1190 jb_err err = JB_ERR_OK;
1192 static const struct parsers server_patterns_light[] = {
1193 { "Content-Length:", 15, server_adjust_content_length },
1194 { "Transfer-Encoding:", 18, server_transfer_coding },
1196 { "Content-Encoding:", 17, server_adjust_content_encoding },
1197 #endif /* def FEATURE_ZLIB */
1201 if (strncmpic(csp->http->cmd, "HEAD", 4))
1203 const struct parsers *v;
1204 struct list_entry *p;
1206 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1208 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1210 /* Header crunch()ed in previous run? -> ignore */
1211 if (p->str == NULL) continue;
1213 /* Does the current parser handle this header? */
1214 if (strncmpic(p->str, v->str, v->len) == 0)
1216 err = v->parser(csp, (char **)&(p->str));
1222 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1223 if ((JB_ERR_OK == err)
1224 && (csp->flags & CSP_FLAG_MODIFIED)
1225 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
1226 && !(csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET))
1230 create_content_length_header(csp->content_length, header, sizeof(header));
1231 err = enlist(csp->headers, header);
1232 if (JB_ERR_OK == err)
1234 log_error(LOG_LEVEL_HEADER,
1235 "Content modified with no Content-Length header set. "
1236 "Created: %s.", header);
1237 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
1240 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1242 #ifdef FEATURE_COMPRESSION
1243 if ((JB_ERR_OK == err)
1244 && (csp->flags & CSP_FLAG_BUFFERED_CONTENT_DEFLATED))
1246 err = enlist_unique_header(csp->headers, "Content-Encoding", "deflate");
1247 if (JB_ERR_OK == err)
1249 log_error(LOG_LEVEL_HEADER, "Added header: Content-Encoding: deflate");
1258 /*********************************************************************
1260 * Function : header_tagger
1262 * Description : Executes all text substitutions from applying
1263 * tag actions and saves the result as tag.
1265 * XXX: Shares enough code with filter_header() and
1266 * pcrs_filter_response() to warrant some helper functions.
1269 * 1 : csp = Current client state (buffers, headers, etc...)
1270 * 2 : header = Header that is used as tagger input
1272 * Returns : JB_ERR_OK on success and always succeeds
1274 *********************************************************************/
1275 static jb_err header_tagger(struct client_state *csp, char *header)
1277 int wanted_filter_type;
1278 int multi_action_index;
1282 struct file_list *fl;
1283 struct re_filterfile_spec *b;
1284 struct list_entry *tag_name;
1286 const size_t header_length = strlen(header);
1288 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1290 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1291 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1295 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1296 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1299 if (list_is_empty(csp->action->multi[multi_action_index]))
1301 /* No appropriate tagger enabled, nothing left to do. */
1305 if (filters_available(csp) == FALSE)
1307 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1308 "tagging enabled, but no taggers available.");
1312 for (i = 0; i < MAX_AF_FILES; i++)
1315 if ((NULL == fl) || (NULL == fl->f))
1318 * Either there are no filter files
1319 * left, or this filter file just
1320 * contains no valid filters.
1322 * Continue to be sure we don't miss
1323 * valid filter files that are chained
1324 * after empty or invalid ones.
1329 /* For all filters, */
1330 for (b = fl->f; b; b = b->next)
1332 if (b->type != wanted_filter_type)
1334 /* skip the ones we don't care about, */
1337 /* leaving only taggers that could apply, of which we use the ones, */
1338 for (tag_name = csp->action->multi[multi_action_index]->first;
1339 NULL != tag_name; tag_name = tag_name->next)
1341 /* that do apply, and */
1342 if (strcmp(b->name, tag_name->str) == 0)
1344 char *modified_tag = NULL;
1346 size_t size = header_length;
1347 pcrs_job *joblist = b->joblist;
1349 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1351 if (NULL == joblist)
1353 log_error(LOG_LEVEL_RE_FILTER,
1354 "Tagger %s has empty joblist. Nothing to do.", b->name);
1358 /* execute their pcrs_joblist on the header. */
1359 for (job = joblist; NULL != job; job = job->next)
1361 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1365 /* Success, continue with the modified version. */
1374 /* Tagger doesn't match */
1377 /* Regex failure, log it but continue anyway. */
1378 assert(NULL != header);
1379 log_error(LOG_LEVEL_ERROR,
1380 "Problems with tagger \'%s\' and header \'%s\': %s",
1381 b->name, *header, pcrs_strerror(hits));
1383 freez(modified_tag);
1387 if (b->dynamic) pcrs_free_joblist(joblist);
1389 /* If this tagger matched */
1395 * There is no technical limitation which makes
1396 * it impossible to use empty tags, but I assume
1397 * no one would do it intentionally.
1400 log_error(LOG_LEVEL_INFO,
1401 "Tagger \'%s\' created an empty tag. Ignored.",
1406 if (!list_contains_item(csp->tags, tag))
1408 if (JB_ERR_OK != enlist(csp->tags, tag))
1410 log_error(LOG_LEVEL_ERROR,
1411 "Insufficient memory to add tag \'%s\', "
1412 "based on tagger \'%s\' and header \'%s\'",
1413 tag, b->name, *header);
1417 char *action_message;
1419 * update the action bits right away, to make
1420 * tagging based on tags set by earlier taggers
1421 * of the same kind possible.
1423 if (update_action_bits_for_tag(csp, tag))
1425 action_message = "Action bits updated accordingly.";
1429 action_message = "No action bits update necessary.";
1432 log_error(LOG_LEVEL_HEADER,
1433 "Tagger \'%s\' added tag \'%s\'. %s",
1434 b->name, tag, action_message);
1439 /* XXX: Is this log-worthy? */
1440 log_error(LOG_LEVEL_HEADER,
1441 "Tagger \'%s\' didn't add tag \'%s\'. "
1442 "Tag already present", b->name, tag);
1445 } /* if the tagger matched */
1446 } /* if the tagger applies */
1447 } /* for every tagger that could apply */
1448 } /* for all filters */
1449 } /* for all filter files */
1454 /* here begins the family of parser functions that reformat header lines */
1456 /*********************************************************************
1458 * Function : filter_header
1460 * Description : Executes all text substitutions from all applying
1461 * +(server|client)-header-filter actions on the header.
1462 * Most of the code was copied from pcrs_filter_response,
1463 * including the rather short variable names
1466 * 1 : csp = Current client state (buffers, headers, etc...)
1467 * 2 : header = On input, pointer to header to modify.
1468 * On output, pointer to the modified header, or NULL
1469 * to remove the header. This function frees the
1470 * original string if necessary.
1472 * Returns : JB_ERR_OK on success and always succeeds
1474 *********************************************************************/
1475 static jb_err filter_header(struct client_state *csp, char **header)
1479 size_t size = strlen(*header);
1481 char *newheader = NULL;
1484 struct file_list *fl;
1485 struct re_filterfile_spec *b;
1486 struct list_entry *filtername;
1489 int wanted_filter_type;
1490 int multi_action_index;
1492 if (csp->flags & CSP_FLAG_NO_FILTERING)
1497 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1499 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1500 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1504 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1505 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1508 if (list_is_empty(csp->action->multi[multi_action_index]))
1510 /* No appropriate header filter enabled, nothing left to do. */
1514 if (filters_available(csp) == FALSE)
1516 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1517 "header filtering enabled, but no matching filters available.");
1521 for (i = 0; i < MAX_AF_FILES; i++)
1524 if ((NULL == fl) || (NULL == fl->f))
1527 * Either there are no filter files
1528 * left, or this filter file just
1529 * contains no valid filters.
1531 * Continue to be sure we don't miss
1532 * valid filter files that are chained
1533 * after empty or invalid ones.
1538 * For all applying +filter actions, look if a filter by that
1539 * name exists and if yes, execute its pcrs_joblist on the
1542 for (b = fl->f; b; b = b->next)
1544 if (b->type != wanted_filter_type)
1546 /* Skip other filter types */
1550 for (filtername = csp->action->multi[multi_action_index]->first;
1551 filtername ; filtername = filtername->next)
1553 if (strcmp(b->name, filtername->str) == 0)
1555 int current_hits = 0;
1556 pcrs_job *joblist = b->joblist;
1558 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1560 if (NULL == joblist)
1562 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1566 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1567 *header, size, b->name);
1569 /* Apply all jobs from the joblist */
1570 for (job = joblist; NULL != job; job = job->next)
1572 matches = pcrs_execute(job, *header, size, &newheader, &size);
1575 current_hits += matches;
1576 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1578 *header = newheader;
1580 else if (0 == matches)
1582 /* Filter doesn't change header */
1588 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1589 *header, b->name, pcrs_strerror(matches));
1590 if (newheader != NULL)
1592 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1598 if (b->dynamic) pcrs_free_joblist(joblist);
1600 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1601 hits += current_hits;
1608 * Additionally checking for hits is important because if
1609 * the continue hack is triggered, server headers can
1610 * arrive empty to separate multiple heads from each other.
1612 if ((0 == size) && hits)
1614 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
1622 /*********************************************************************
1624 * Function : server_connection
1626 * Description : Makes sure a proper "Connection:" header is
1627 * set and signals connection_header_adder to
1631 * 1 : csp = Current client state (buffers, headers, etc...)
1632 * 2 : header = On input, pointer to header to modify.
1633 * On output, pointer to the modified header, or NULL
1634 * to remove the header. This function frees the
1635 * original string if necessary.
1637 * Returns : JB_ERR_OK on success.
1639 *********************************************************************/
1640 static jb_err server_connection(struct client_state *csp, char **header)
1642 if (!strcmpic(*header, "Connection: keep-alive")
1643 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1644 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
1648 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1649 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1651 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
1654 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE))
1656 log_error(LOG_LEVEL_HEADER,
1657 "Keeping the server header '%s' around.", *header);
1660 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
1662 char *old_header = *header;
1664 *header = strdup_or_die("Connection: close");
1665 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
1670 /* Signal server_connection_adder() to return early. */
1671 csp->flags |= CSP_FLAG_SERVER_CONNECTION_HEADER_SET;
1677 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1678 /*********************************************************************
1680 * Function : server_keep_alive
1682 * Description : Stores the server's keep alive timeout.
1685 * 1 : csp = Current client state (buffers, headers, etc...)
1686 * 2 : header = On input, pointer to header to modify.
1687 * On output, pointer to the modified header, or NULL
1688 * to remove the header. This function frees the
1689 * original string if necessary.
1691 * Returns : JB_ERR_OK.
1693 *********************************************************************/
1694 static jb_err server_keep_alive(struct client_state *csp, char **header)
1696 unsigned int keep_alive_timeout;
1697 const char *timeout_position = strstr(*header, "timeout=");
1699 if ((NULL == timeout_position)
1700 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1702 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1706 if (keep_alive_timeout < csp->server_connection.keep_alive_timeout)
1708 log_error(LOG_LEVEL_HEADER,
1709 "Reducing keep-alive timeout from %u to %u.",
1710 csp->server_connection.keep_alive_timeout, keep_alive_timeout);
1711 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1715 /* XXX: Is this log worthy? */
1716 log_error(LOG_LEVEL_HEADER,
1717 "Server keep-alive timeout is %u. Sticking with %u.",
1718 keep_alive_timeout, csp->server_connection.keep_alive_timeout);
1720 csp->flags |= CSP_FLAG_SERVER_KEEP_ALIVE_TIMEOUT_SET;
1727 /*********************************************************************
1729 * Function : server_proxy_connection
1731 * Description : Figures out whether or not we should add a
1732 * Proxy-Connection header.
1735 * 1 : csp = Current client state (buffers, headers, etc...)
1736 * 2 : header = On input, pointer to header to modify.
1737 * On output, pointer to the modified header, or NULL
1738 * to remove the header. This function frees the
1739 * original string if necessary.
1741 * Returns : JB_ERR_OK.
1743 *********************************************************************/
1744 static jb_err server_proxy_connection(struct client_state *csp, char **header)
1746 csp->flags |= CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET;
1751 /*********************************************************************
1753 * Function : proxy_authentication
1755 * Description : Removes headers that are relevant for proxy
1756 * authentication unless forwarding them has
1757 * been explicitly requested.
1760 * 1 : csp = Current client state (buffers, headers, etc...)
1761 * 2 : header = On input, pointer to header to modify.
1762 * On output, pointer to the modified header, or NULL
1763 * to remove the header. This function frees the
1764 * original string if necessary.
1766 * Returns : JB_ERR_OK.
1768 *********************************************************************/
1769 static jb_err proxy_authentication(struct client_state *csp, char **header)
1771 if ((csp->config->feature_flags &
1772 RUNTIME_FEATURE_FORWARD_PROXY_AUTHENTICATION_HEADERS) == 0) {
1773 log_error(LOG_LEVEL_HEADER,
1774 "Forwarding proxy authentication headers is disabled. Crunching: %s", *header);
1781 /*********************************************************************
1783 * Function : client_keep_alive
1785 * Description : Stores the client's keep alive timeout.
1788 * 1 : csp = Current client state (buffers, headers, etc...)
1789 * 2 : header = On input, pointer to header to modify.
1790 * On output, pointer to the modified header, or NULL
1791 * to remove the header. This function frees the
1792 * original string if necessary.
1794 * Returns : JB_ERR_OK.
1796 *********************************************************************/
1797 static jb_err client_keep_alive(struct client_state *csp, char **header)
1799 unsigned int keep_alive_timeout;
1800 const char *timeout_position = strstr(*header, ": ");
1802 if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1804 log_error(LOG_LEVEL_HEADER,
1805 "keep-alive support is disabled. Crunching: %s.", *header);
1810 if ((NULL == timeout_position)
1811 || (1 != sscanf(timeout_position, ": %u", &keep_alive_timeout)))
1813 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1817 if (keep_alive_timeout < csp->config->keep_alive_timeout)
1819 log_error(LOG_LEVEL_HEADER,
1820 "Reducing keep-alive timeout from %u to %u.",
1821 csp->config->keep_alive_timeout, keep_alive_timeout);
1822 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1826 /* XXX: Is this log worthy? */
1827 log_error(LOG_LEVEL_HEADER,
1828 "Client keep-alive timeout is %u. Sticking with %u.",
1829 keep_alive_timeout, csp->config->keep_alive_timeout);
1837 /*********************************************************************
1839 * Function : get_content_length
1841 * Description : Gets the content length specified in a
1842 * Content-Length header.
1845 * 1 : header_value = The Content-Length header value.
1846 * 2 : length = Storage to return the value.
1848 * Returns : JB_ERR_OK on success, or
1849 * JB_ERR_PARSE if no value is recognized.
1851 *********************************************************************/
1852 static jb_err get_content_length(const char *header_value, unsigned long long *length)
1855 assert(sizeof(unsigned long long) > 4);
1856 if (1 != sscanf(header_value, "%I64u", length))
1858 if (1 != sscanf(header_value, "%llu", length))
1861 return JB_ERR_PARSE;
1868 /*********************************************************************
1870 * Function : client_save_content_length
1872 * Description : Save the Content-Length sent by the client.
1875 * 1 : csp = Current client state (buffers, headers, etc...)
1876 * 2 : header = On input, pointer to header to modify.
1877 * On output, pointer to the modified header, or NULL
1878 * to remove the header. This function frees the
1879 * original string if necessary.
1881 * Returns : JB_ERR_OK on success, or
1882 * JB_ERR_MEMORY on out-of-memory error.
1884 *********************************************************************/
1885 static jb_err client_save_content_length(struct client_state *csp, char **header)
1887 unsigned long long content_length = 0;
1888 const char *header_value;
1890 assert(*(*header+14) == ':');
1892 header_value = *header + 15;
1893 if (JB_ERR_OK != get_content_length(header_value, &content_length))
1895 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
1900 csp->expected_client_content_length = content_length;
1905 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1909 /*********************************************************************
1911 * Function : client_connection
1913 * Description : Makes sure a proper "Connection:" header is
1914 * set and signals connection_header_adder
1918 * 1 : csp = Current client state (buffers, headers, etc...)
1919 * 2 : header = On input, pointer to header to modify.
1920 * On output, pointer to the modified header, or NULL
1921 * to remove the header. This function frees the
1922 * original string if necessary.
1924 * Returns : JB_ERR_OK on success.
1926 *********************************************************************/
1927 static jb_err client_connection(struct client_state *csp, char **header)
1929 static const char connection_close[] = "Connection: close";
1931 if (!strcmpic(*header, connection_close))
1933 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1934 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING)
1935 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1937 if (!strcmpic(csp->http->ver, "HTTP/1.1"))
1939 log_error(LOG_LEVEL_HEADER,
1940 "Removing \'%s\' to imply keep-alive.", *header);
1943 * While we imply keep-alive to the server,
1944 * we have to remember that the client didn't.
1946 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1950 char *old_header = *header;
1952 *header = strdup_or_die("Connection: keep-alive");
1953 log_error(LOG_LEVEL_HEADER,
1954 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1960 log_error(LOG_LEVEL_HEADER,
1961 "Keeping the client header '%s' around. "
1962 "The connection will not be kept alive.",
1964 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1967 else if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
1968 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1970 log_error(LOG_LEVEL_HEADER,
1971 "Keeping the client header '%s' around. "
1972 "The server connection will be kept alive if possible.",
1974 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1975 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1979 char *old_header = *header;
1981 *header = strdup_or_die(connection_close);
1982 log_error(LOG_LEVEL_HEADER,
1983 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1987 /* Signal client_connection_header_adder() to return early. */
1988 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
1994 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1995 /*********************************************************************
1997 * Function : client_proxy_connection
1999 * Description : Sets the CLIENT_CONNECTION_KEEP_ALIVE flag when
2000 * appropriate and removes the Proxy-Connection
2004 * 1 : csp = Current client state (buffers, headers, etc...)
2005 * 2 : header = On input, pointer to header to modify.
2006 * On output, pointer to the modified header, or NULL
2007 * to remove the header. This function frees the
2008 * original string if necessary.
2010 * Returns : JB_ERR_OK
2012 *********************************************************************/
2013 static jb_err client_proxy_connection(struct client_state *csp, char **header)
2015 if (0 == (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
2016 && (csp->http->ssl == 0)
2017 && (NULL == strstr(*header, "close")))
2019 log_error(LOG_LEVEL_HEADER,
2020 "The client connection can be kept alive due to: %s", *header);
2021 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
2023 crumble(csp, header);
2027 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2030 /*********************************************************************
2032 * Function : client_transfer_encoding
2034 * Description : Raise the CSP_FLAG_CHUNKED_CLIENT_BODY flag if
2035 * the request body is "chunked"
2037 * XXX: Currently not called through sed() as we
2038 * need the flag earlier on. Should be fixed.
2041 * 1 : csp = Current client state (buffers, headers, etc...)
2042 * 2 : header = On input, pointer to header to modify.
2043 * On output, pointer to the modified header, or NULL
2044 * to remove the header. This function frees the
2045 * original string if necessary.
2047 * Returns : JB_ERR_OK on success, or
2049 *********************************************************************/
2050 jb_err client_transfer_encoding(struct client_state *csp, char **header)
2052 if (strstr(*header, "chunked"))
2054 csp->flags |= CSP_FLAG_CHUNKED_CLIENT_BODY;
2055 log_error(LOG_LEVEL_HEADER, "Expecting chunked client body");
2062 /*********************************************************************
2064 * Function : crumble
2066 * Description : This is called if a header matches a pattern to "crunch"
2069 * 1 : csp = Current client state (buffers, headers, etc...)
2070 * 2 : header = On input, pointer to header to modify.
2071 * On output, pointer to the modified header, or NULL
2072 * to remove the header. This function frees the
2073 * original string if necessary.
2075 * Returns : JB_ERR_OK on success, or
2076 * JB_ERR_MEMORY on out-of-memory error.
2078 *********************************************************************/
2079 static jb_err crumble(struct client_state *csp, char **header)
2082 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2088 /*********************************************************************
2090 * Function : crunch_server_header
2092 * Description : Crunch server header if it matches a string supplied by the
2093 * user. Called from `sed'.
2096 * 1 : csp = Current client state (buffers, headers, etc...)
2097 * 2 : header = On input, pointer to header to modify.
2098 * On output, pointer to the modified header, or NULL
2099 * to remove the header. This function frees the
2100 * original string if necessary.
2102 * Returns : JB_ERR_OK on success and always succeeds
2104 *********************************************************************/
2105 static jb_err crunch_server_header(struct client_state *csp, char **header)
2107 const char *crunch_pattern;
2109 /* Do we feel like crunching? */
2110 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2112 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2114 /* Is the current header the lucky one? */
2115 if (strstr(*header, crunch_pattern))
2117 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2126 /*********************************************************************
2128 * Function : server_content_type
2130 * Description : Set the content-type for filterable types (text/.*,
2131 * .*xml.*, .*script.* and image/gif) unless filtering has been
2132 * forbidden (CT_TABOO) while parsing earlier headers.
2133 * NOTE: Since text/plain is commonly used by web servers
2134 * for files whose correct type is unknown, we don't
2135 * set CT_TEXT for it.
2138 * 1 : csp = Current client state (buffers, headers, etc...)
2139 * 2 : header = On input, pointer to header to modify.
2140 * On output, pointer to the modified header, or NULL
2141 * to remove the header. This function frees the
2142 * original string if necessary.
2144 * Returns : JB_ERR_OK on success, or
2145 * JB_ERR_MEMORY on out-of-memory error.
2147 *********************************************************************/
2148 static jb_err server_content_type(struct client_state *csp, char **header)
2150 /* Remove header if it isn't the first Content-Type header */
2151 if ((csp->content_type & CT_DECLARED))
2153 if (content_filters_enabled(csp->action))
2156 * Making sure the client interprets the content the same way
2157 * Privoxy did is only relevant if Privoxy modified it.
2159 * Checking for this is "hard" as it's not yet known when
2160 * this function is called, thus go shopping and and just
2161 * check if Privoxy could filter it.
2163 * The main thing is that we don't mess with the headers
2164 * unless the user signalled that it's acceptable.
2166 log_error(LOG_LEVEL_HEADER,
2167 "Multiple Content-Type headers detected. "
2168 "Removing and ignoring: %s",
2176 * Signal that the Content-Type has been set.
2178 csp->content_type |= CT_DECLARED;
2180 if (!(csp->content_type & CT_TABOO))
2183 * XXX: The assumption that text/plain is a sign of
2184 * binary data seems to be somewhat unreasonable nowadays
2185 * and should be dropped after 3.0.8 is out.
2187 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2188 || strstr(*header, "xml")
2189 || strstr(*header, "script"))
2191 csp->content_type |= CT_TEXT;
2193 else if (strstr(*header, "image/gif"))
2195 csp->content_type |= CT_GIF;
2200 * Are we messing with the content type?
2202 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2205 * Make sure the user doesn't accidentally
2206 * change the content type of binary documents.
2208 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2211 *header = strdup_or_die("Content-Type: ");
2212 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2216 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2217 return JB_ERR_MEMORY;
2219 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2223 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2224 "It doesn't look like a content type that should be filtered. "
2225 "Enable force-text-mode if you know what you're doing.", *header);
2233 /*********************************************************************
2235 * Function : server_transfer_coding
2237 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2238 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2239 * - Remove header if body was chunked but has been
2240 * de-chunked for filtering.
2243 * 1 : csp = Current client state (buffers, headers, etc...)
2244 * 2 : header = On input, pointer to header to modify.
2245 * On output, pointer to the modified header, or NULL
2246 * to remove the header. This function frees the
2247 * original string if necessary.
2249 * Returns : JB_ERR_OK on success, or
2250 * JB_ERR_MEMORY on out-of-memory error.
2252 *********************************************************************/
2253 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2256 * Turn off pcrs and gif filtering if body compressed
2258 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2262 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2264 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2265 csp->http->cmd, *header);
2266 #endif /* def FEATURE_ZLIB */
2267 csp->content_type = CT_TABOO;
2271 * Raise flag if body chunked
2273 if (strstr(*header, "chunked"))
2275 csp->flags |= CSP_FLAG_CHUNKED;
2278 * If the body was modified, it has been de-chunked first
2279 * and the header must be removed.
2281 * FIXME: If there is more than one transfer encoding,
2282 * only the "chunked" part should be removed here.
2284 if (csp->flags & CSP_FLAG_MODIFIED)
2286 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2295 /*********************************************************************
2297 * Function : server_content_encoding
2299 * Description : Used to check if the content is compressed, and if
2300 * FEATURE_ZLIB is disabled, filtering is disabled as
2303 * If FEATURE_ZLIB is enabled and the compression type
2304 * supported, the content is marked for decompression.
2306 * XXX: Doesn't properly deal with multiple or with
2307 * unsupported but unknown encodings.
2308 * Is case-sensitive but shouldn't be.
2311 * 1 : csp = Current client state (buffers, headers, etc...)
2312 * 2 : header = On input, pointer to header to modify.
2313 * On output, pointer to the modified header, or NULL
2314 * to remove the header. This function frees the
2315 * original string if necessary.
2317 * Returns : JB_ERR_OK on success, or
2318 * JB_ERR_MEMORY on out-of-memory error.
2320 *********************************************************************/
2321 static jb_err server_content_encoding(struct client_state *csp, char **header)
2324 if (strstr(*header, "sdch"))
2327 * Shared Dictionary Compression over HTTP isn't supported,
2328 * filtering it anyway is pretty much guaranteed to mess up
2331 csp->content_type |= CT_TABOO;
2334 * Log a warning if the user expects the content to be filtered.
2336 if ((csp->rlist != NULL) &&
2337 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2339 log_error(LOG_LEVEL_INFO,
2340 "SDCH-compressed content detected, content filtering disabled. "
2341 "Consider suppressing SDCH offers made by the client.");
2344 else if (strstr(*header, "gzip"))
2346 /* Mark for gzip decompression */
2347 csp->content_type |= CT_GZIP;
2349 else if (strstr(*header, "deflate"))
2351 /* Mark for zlib decompression */
2352 csp->content_type |= CT_DEFLATE;
2354 else if (strstr(*header, "compress"))
2357 * We can't decompress this; therefore we can't filter
2360 csp->content_type |= CT_TABOO;
2362 #else /* !defined(FEATURE_ZLIB) */
2364 * XXX: Using a black list here isn't the right approach.
2366 * In case of SDCH, building with zlib support isn't
2369 if (strstr(*header, "gzip") ||
2370 strstr(*header, "compress") ||
2371 strstr(*header, "deflate") ||
2372 strstr(*header, "sdch"))
2375 * Body is compressed, turn off pcrs and gif filtering.
2377 csp->content_type |= CT_TABOO;
2380 * Log a warning if the user expects the content to be filtered.
2382 if ((csp->rlist != NULL) &&
2383 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2385 log_error(LOG_LEVEL_INFO,
2386 "Compressed content detected, content filtering disabled. "
2387 "Consider recompiling Privoxy with zlib support or "
2388 "enable the prevent-compression action.");
2391 #endif /* defined(FEATURE_ZLIB) */
2399 /*********************************************************************
2401 * Function : server_adjust_content_encoding
2403 * Description : Remove the Content-Encoding header if the
2404 * decompression was successful and the content
2408 * 1 : csp = Current client state (buffers, headers, etc...)
2409 * 2 : header = On input, pointer to header to modify.
2410 * On output, pointer to the modified header, or NULL
2411 * to remove the header. This function frees the
2412 * original string if necessary.
2414 * Returns : JB_ERR_OK on success, or
2415 * JB_ERR_MEMORY on out-of-memory error.
2417 *********************************************************************/
2418 static jb_err server_adjust_content_encoding(struct client_state *csp, char **header)
2420 if ((csp->flags & CSP_FLAG_MODIFIED)
2421 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2424 * We successfully decompressed the content,
2425 * and have to clean the header now, so the
2426 * client no longer expects compressed data.
2428 * XXX: There is a difference between cleaning
2429 * and removing it completely.
2431 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2438 #endif /* defined(FEATURE_ZLIB) */
2441 /*********************************************************************
2443 * Function : server_adjust_content_length
2445 * Description : Adjust Content-Length header if we modified
2449 * 1 : csp = Current client state (buffers, headers, etc...)
2450 * 2 : header = On input, pointer to header to modify.
2451 * On output, pointer to the modified header, or NULL
2452 * to remove the header. This function frees the
2453 * original string if necessary.
2455 * Returns : JB_ERR_OK on success, or
2456 * JB_ERR_MEMORY on out-of-memory error.
2458 *********************************************************************/
2459 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2461 /* Regenerate header if the content was modified. */
2462 if (csp->flags & CSP_FLAG_MODIFIED)
2464 const size_t header_length = 50;
2466 *header = malloc(header_length);
2467 if (*header == NULL)
2469 return JB_ERR_MEMORY;
2471 create_content_length_header(csp->content_length, *header, header_length);
2472 log_error(LOG_LEVEL_HEADER,
2473 "Adjusted Content-Length to %llu", csp->content_length);
2480 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2481 /*********************************************************************
2483 * Function : server_save_content_length
2485 * Description : Save the Content-Length sent by the server.
2488 * 1 : csp = Current client state (buffers, headers, etc...)
2489 * 2 : header = On input, pointer to header to modify.
2490 * On output, pointer to the modified header, or NULL
2491 * to remove the header. This function frees the
2492 * original string if necessary.
2494 * Returns : JB_ERR_OK on success, or
2495 * JB_ERR_MEMORY on out-of-memory error.
2497 *********************************************************************/
2498 static jb_err server_save_content_length(struct client_state *csp, char **header)
2500 unsigned long long content_length = 0;
2501 const char *header_value;
2503 assert(*(*header+14) == ':');
2505 header_value = *header + 15;
2506 if (JB_ERR_OK != get_content_length(header_value, &content_length))
2508 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2513 csp->expected_content_length = content_length;
2514 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
2515 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2520 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2523 /*********************************************************************
2525 * Function : server_content_md5
2527 * Description : Crumble any Content-MD5 headers if the document was
2528 * modified. FIXME: Should we re-compute instead?
2531 * 1 : csp = Current client state (buffers, headers, etc...)
2532 * 2 : header = On input, pointer to header to modify.
2533 * On output, pointer to the modified header, or NULL
2534 * to remove the header. This function frees the
2535 * original string if necessary.
2537 * Returns : JB_ERR_OK on success, or
2538 * JB_ERR_MEMORY on out-of-memory error.
2540 *********************************************************************/
2541 static jb_err server_content_md5(struct client_state *csp, char **header)
2543 if (csp->flags & CSP_FLAG_MODIFIED)
2545 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2553 /*********************************************************************
2555 * Function : server_content_disposition
2557 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
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_content_disposition(struct client_state *csp, char **header)
2576 * Are we messing with the Content-Disposition header?
2578 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2584 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2586 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2589 * Blocking content-disposition header
2591 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2598 * Replacing Content-Disposition header
2601 *header = strdup("Content-Disposition: ");
2602 string_append(header, newval);
2604 if (*header != NULL)
2606 log_error(LOG_LEVEL_HEADER,
2607 "Content-Disposition header crunched and replaced with: %s", *header);
2610 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2614 /*********************************************************************
2616 * Function : server_last_modified
2618 * Description : Changes Last-Modified header to the actual date
2619 * to help hide-if-modified-since.
2620 * Called from `sed'.
2623 * 1 : csp = Current client state (buffers, headers, etc...)
2624 * 2 : header = On input, pointer to header to modify.
2625 * On output, pointer to the modified header, or NULL
2626 * to remove the header. This function frees the
2627 * original string if necessary.
2629 * Returns : JB_ERR_OK on success, or
2630 * JB_ERR_MEMORY on out-of-memory error.
2632 *********************************************************************/
2633 static jb_err server_last_modified(struct client_state *csp, char **header)
2636 time_t last_modified;
2640 * Are we messing with the Last-Modified header?
2642 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2648 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2650 if (0 == strcmpic(newval, "block"))
2653 * Blocking Last-Modified header. Useless but why not.
2655 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2659 else if (0 == strcmpic(newval, "reset-to-request-time"))
2662 * Setting Last-Modified Header to now.
2665 get_http_time(0, buf, sizeof(buf));
2667 *header = strdup("Last-Modified: ");
2668 string_append(header, buf);
2670 if (*header == NULL)
2672 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2676 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2679 else if (0 == strcmpic(newval, "randomize"))
2681 const char *header_time = *header + sizeof("Last-Modified:");
2683 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2685 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2687 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2693 struct tm *timeptr = NULL;
2695 #ifdef HAVE_GMTIME_R
2699 rtime = (long int)difftime(now, last_modified);
2702 long int days, hours, minutes, seconds;
2703 const int negative_delta = (rtime < 0);
2708 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2710 rtime = pick_from_range(rtime);
2715 last_modified += rtime;
2716 #ifdef HAVE_GMTIME_R
2717 timeptr = gmtime_r(&last_modified, &gmt);
2718 #elif defined(MUTEX_LOCKS_AVAILABLE)
2719 privoxy_mutex_lock(&gmtime_mutex);
2720 timeptr = gmtime(&last_modified);
2721 privoxy_mutex_unlock(&gmtime_mutex);
2723 timeptr = gmtime(&last_modified);
2725 if ((NULL == timeptr) || !strftime(newheader,
2726 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
2728 log_error(LOG_LEVEL_ERROR,
2729 "Randomizing '%s' failed. Crunching the header without replacement.",
2736 *header = strdup("Last-Modified: ");
2737 string_append(header, newheader);
2739 if (*header == NULL)
2741 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2742 return JB_ERR_MEMORY;
2745 days = rtime / (3600 * 24);
2746 hours = rtime / 3600 % 24;
2747 minutes = rtime / 60 % 60;
2748 seconds = rtime % 60;
2750 log_error(LOG_LEVEL_HEADER,
2751 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2752 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2753 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2757 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2766 /*********************************************************************
2768 * Function : client_accept_encoding
2770 * Description : Rewrite the client's Accept-Encoding header so that
2771 * if doesn't allow compression, if the action applies.
2772 * Note: For HTTP/1.0 the absence of the header is enough.
2775 * 1 : csp = Current client state (buffers, headers, etc...)
2776 * 2 : header = On input, pointer to header to modify.
2777 * On output, pointer to the modified header, or NULL
2778 * to remove the header. This function frees the
2779 * original string if necessary.
2781 * Returns : JB_ERR_OK on success, or
2782 * JB_ERR_MEMORY on out-of-memory error.
2784 *********************************************************************/
2785 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2787 #ifdef FEATURE_COMPRESSION
2788 if ((csp->config->feature_flags & RUNTIME_FEATURE_COMPRESSION)
2789 && strstr(*header, "deflate"))
2791 csp->flags |= CSP_FLAG_CLIENT_SUPPORTS_DEFLATE;
2794 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2796 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2804 /*********************************************************************
2806 * Function : client_te
2808 * Description : Rewrite the client's TE header so that
2809 * if doesn't allow compression, if the action applies.
2812 * 1 : csp = Current client state (buffers, headers, etc...)
2813 * 2 : header = On input, pointer to header to modify.
2814 * On output, pointer to the modified header, or NULL
2815 * to remove the header. This function frees the
2816 * original string if necessary.
2818 * Returns : JB_ERR_OK on success, or
2819 * JB_ERR_MEMORY on out-of-memory error.
2821 *********************************************************************/
2822 static jb_err client_te(struct client_state *csp, char **header)
2824 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2827 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2834 /*********************************************************************
2836 * Function : client_referrer
2838 * Description : Handle the "referer" config setting properly.
2839 * Called from `sed'.
2842 * 1 : csp = Current client state (buffers, headers, etc...)
2843 * 2 : header = On input, pointer to header to modify.
2844 * On output, pointer to the modified header, or NULL
2845 * to remove the header. This function frees the
2846 * original string if necessary.
2848 * Returns : JB_ERR_OK on success, or
2849 * JB_ERR_MEMORY on out-of-memory error.
2851 *********************************************************************/
2852 static jb_err client_referrer(struct client_state *csp, char **header)
2854 const char *parameter;
2855 /* booleans for parameters we have to check multiple times */
2856 int parameter_conditional_block;
2857 int parameter_conditional_forge;
2859 #ifdef FEATURE_FORCE_LOAD
2861 * Since the referrer can include the prefix even
2862 * if the request itself is non-forced, we must
2863 * clean it unconditionally.
2865 * XXX: strclean is too broad
2867 strclean(*header, FORCE_PREFIX);
2868 #endif /* def FEATURE_FORCE_LOAD */
2870 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2872 /* Nothing left to do */
2876 parameter = csp->action->string[ACTION_STRING_REFERER];
2877 assert(parameter != NULL);
2878 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2879 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2881 if (!parameter_conditional_block && !parameter_conditional_forge)
2884 * As conditional-block and conditional-forge are the only
2885 * parameters that rely on the original referrer, we can
2886 * remove it now for all the others.
2891 if (0 == strcmpic(parameter, "block"))
2893 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2896 else if (parameter_conditional_block || parameter_conditional_forge)
2898 return handle_conditional_hide_referrer_parameter(header,
2899 csp->http->hostport, parameter_conditional_block);
2901 else if (0 == strcmpic(parameter, "forge"))
2903 return create_forged_referrer(header, csp->http->hostport);
2907 /* interpret parameter as user-supplied referer to fake */
2908 return create_fake_referrer(header, parameter);
2913 /*********************************************************************
2915 * Function : client_accept_language
2917 * Description : Handle the "Accept-Language" config setting properly.
2918 * Called from `sed'.
2921 * 1 : csp = Current client state (buffers, headers, etc...)
2922 * 2 : header = On input, pointer to header to modify.
2923 * On output, pointer to the modified header, or NULL
2924 * to remove the header. This function frees the
2925 * original string if necessary.
2927 * Returns : JB_ERR_OK on success, or
2928 * JB_ERR_MEMORY on out-of-memory error.
2930 *********************************************************************/
2931 static jb_err client_accept_language(struct client_state *csp, char **header)
2936 * Are we messing with the Accept-Language?
2938 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2940 /*I don't think so*/
2944 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2946 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2949 * Blocking Accept-Language header
2951 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2958 * Replacing Accept-Language header
2961 *header = strdup("Accept-Language: ");
2962 string_append(header, newval);
2964 if (*header == NULL)
2966 log_error(LOG_LEVEL_ERROR,
2967 "Insufficient memory. Accept-Language header crunched without replacement.");
2971 log_error(LOG_LEVEL_HEADER,
2972 "Accept-Language header crunched and replaced with: %s", *header);
2975 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2979 /*********************************************************************
2981 * Function : crunch_client_header
2983 * Description : Crunch client header if it matches a string supplied by the
2984 * user. Called from `sed'.
2987 * 1 : csp = Current client state (buffers, headers, etc...)
2988 * 2 : header = On input, pointer to header to modify.
2989 * On output, pointer to the modified header, or NULL
2990 * to remove the header. This function frees the
2991 * original string if necessary.
2993 * Returns : JB_ERR_OK on success and always succeeds
2995 *********************************************************************/
2996 static jb_err crunch_client_header(struct client_state *csp, char **header)
2998 const char *crunch_pattern;
3000 /* Do we feel like crunching? */
3001 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
3003 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
3005 /* Is the current header the lucky one? */
3006 if (strstr(*header, crunch_pattern))
3008 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
3016 /*********************************************************************
3018 * Function : client_uagent
3020 * Description : Handle the "user-agent" config setting properly
3021 * and remember its original value to enable browser
3022 * bug workarounds. Called from `sed'.
3025 * 1 : csp = Current client state (buffers, headers, etc...)
3026 * 2 : header = On input, pointer to header to modify.
3027 * On output, pointer to the modified header, or NULL
3028 * to remove the header. This function frees the
3029 * original string if necessary.
3031 * Returns : JB_ERR_OK on success, or
3032 * JB_ERR_MEMORY on out-of-memory error.
3034 *********************************************************************/
3035 static jb_err client_uagent(struct client_state *csp, char **header)
3039 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
3044 newval = csp->action->string[ACTION_STRING_USER_AGENT];
3051 *header = strdup("User-Agent: ");
3052 string_append(header, newval);
3054 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
3056 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3060 /*********************************************************************
3062 * Function : client_ua
3064 * Description : Handle "ua-" headers properly. Called from `sed'.
3067 * 1 : csp = Current client state (buffers, headers, etc...)
3068 * 2 : header = On input, pointer to header to modify.
3069 * On output, pointer to the modified header, or NULL
3070 * to remove the header. This function frees the
3071 * original string if necessary.
3073 * Returns : JB_ERR_OK on success, or
3074 * JB_ERR_MEMORY on out-of-memory error.
3076 *********************************************************************/
3077 static jb_err client_ua(struct client_state *csp, char **header)
3079 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3081 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3089 /*********************************************************************
3091 * Function : client_from
3093 * Description : Handle the "from" config setting properly.
3094 * Called from `sed'.
3097 * 1 : csp = Current client state (buffers, headers, etc...)
3098 * 2 : header = On input, pointer to header to modify.
3099 * On output, pointer to the modified header, or NULL
3100 * to remove the header. This function frees the
3101 * original string if necessary.
3103 * Returns : JB_ERR_OK on success, or
3104 * JB_ERR_MEMORY on out-of-memory error.
3106 *********************************************************************/
3107 static jb_err client_from(struct client_state *csp, char **header)
3111 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3118 newval = csp->action->string[ACTION_STRING_FROM];
3121 * Are we blocking the e-mail address?
3123 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
3125 log_error(LOG_LEVEL_HEADER, "crunched From!");
3129 log_error(LOG_LEVEL_HEADER, " modified");
3131 *header = strdup("From: ");
3132 string_append(header, newval);
3134 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3138 /*********************************************************************
3140 * Function : client_send_cookie
3142 * Description : Crunches the "cookie" header if necessary.
3143 * Called from `sed'.
3145 * XXX: Stupid name, doesn't send squat.
3148 * 1 : csp = Current client state (buffers, headers, etc...)
3149 * 2 : header = On input, pointer to header to modify.
3150 * On output, pointer to the modified header, or NULL
3151 * to remove the header. This function frees the
3152 * original string if necessary.
3154 * Returns : JB_ERR_OK on success, or
3155 * JB_ERR_MEMORY on out-of-memory error.
3157 *********************************************************************/
3158 static jb_err client_send_cookie(struct client_state *csp, char **header)
3160 if (csp->action->flags & ACTION_CRUNCH_OUTGOING_COOKIES)
3162 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3170 /*********************************************************************
3172 * Function : client_x_forwarded
3174 * Description : Handle the "x-forwarded-for" config setting properly,
3175 * also used in the add_client_headers list. Called from `sed'.
3178 * 1 : csp = Current client state (buffers, headers, etc...)
3179 * 2 : header = On input, pointer to header to modify.
3180 * On output, pointer to the modified header, or NULL
3181 * to remove the header. This function frees the
3182 * original string if necessary.
3184 * Returns : JB_ERR_OK on success, or
3185 * JB_ERR_MEMORY on out-of-memory error.
3187 *********************************************************************/
3188 jb_err client_x_forwarded(struct client_state *csp, char **header)
3190 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
3192 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
3194 if (0 == strcmpic(parameter, "block"))
3197 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3199 else if (0 == strcmpic(parameter, "add"))
3201 string_append(header, ", ");
3202 string_append(header, csp->ip_addr_str);
3204 if (*header == NULL)
3206 return JB_ERR_MEMORY;
3208 log_error(LOG_LEVEL_HEADER,
3209 "Appended client IP address to %s", *header);
3210 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
3214 log_error(LOG_LEVEL_FATAL,
3215 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
3223 /*********************************************************************
3225 * Function : client_max_forwards
3227 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3228 * from the value of the Max-Forwards header field.
3231 * 1 : csp = Current client state (buffers, headers, etc...)
3232 * 2 : header = On input, pointer to header to modify.
3233 * On output, pointer to the modified header, or NULL
3234 * to remove the header. This function frees the
3235 * original string if necessary.
3237 * Returns : JB_ERR_OK on success, or
3238 * JB_ERR_MEMORY on out-of-memory error.
3240 *********************************************************************/
3241 static jb_err client_max_forwards(struct client_state *csp, char **header)
3245 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3246 (0 == strcmpic(csp->http->gpc, "options")))
3248 assert(*(*header+12) == ':');
3249 if (1 == sscanf(*header+12, ": %d", &max_forwards))
3251 if (max_forwards > 0)
3253 snprintf(*header, strlen(*header)+1, "Max-Forwards: %d", --max_forwards);
3254 log_error(LOG_LEVEL_HEADER,
3255 "Max-Forwards value for %s request reduced to %d.",
3256 csp->http->gpc, max_forwards);
3258 else if (max_forwards < 0)
3260 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3266 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3275 /*********************************************************************
3277 * Function : client_host
3279 * Description : If the request URI did not contain host and
3280 * port information, parse and evaluate the Host
3284 * 1 : csp = Current client state (buffers, headers, etc...)
3285 * 2 : header = On input, pointer to header to modify.
3286 * On output, pointer to the modified header, or NULL
3287 * to remove the header. This function frees the
3288 * original string if necessary.
3290 * Returns : JB_ERR_OK on success, or
3291 * JB_ERR_MEMORY on out-of-memory error.
3293 *********************************************************************/
3294 static jb_err client_host(struct client_state *csp, char **header)
3298 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3299 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3302 p = strdup_or_die((*header)+6);
3304 q = strdup_or_die(p);
3306 freez(csp->http->hostport);
3307 csp->http->hostport = p;
3308 freez(csp->http->host);
3309 csp->http->host = q;
3310 q = strchr(csp->http->host, ':');
3313 /* Terminate hostname and evaluate port string */
3315 csp->http->port = atoi(q);
3319 csp->http->port = csp->http->ssl ? 443 : 80;
3322 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3323 csp->http->hostport, csp->http->host, csp->http->port);
3326 /* Signal client_host_adder() to return right away */
3327 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3333 /*********************************************************************
3335 * Function : client_if_modified_since
3337 * Description : Remove or modify the If-Modified-Since header.
3340 * 1 : csp = Current client state (buffers, headers, etc...)
3341 * 2 : header = On input, pointer to header to modify.
3342 * On output, pointer to the modified header, or NULL
3343 * to remove the header. This function frees the
3344 * original string if necessary.
3346 * Returns : JB_ERR_OK on success, or
3347 * JB_ERR_MEMORY on out-of-memory error.
3349 *********************************************************************/
3350 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3353 #ifdef HAVE_GMTIME_R
3356 struct tm *timeptr = NULL;
3361 if (0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3364 * The client got an error message because of a temporary problem,
3365 * the problem is gone and the client now tries to revalidate our
3366 * error message on the real server. The revalidation would always
3367 * end with the transmission of the whole document and there is
3368 * no need to expose the bogus If-Modified-Since header.
3370 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3373 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3375 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3377 if ((0 == strcmpic(newval, "block")))
3379 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3382 else /* add random value */
3384 const char *header_time = *header + sizeof("If-Modified-Since:");
3386 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3388 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3393 long int hours, minutes, seconds;
3394 long int rtime = strtol(newval, &endptr, 0);
3395 const int negative_range = (rtime < 0);
3399 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3400 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3406 rtime = pick_from_range(rtime);
3410 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3413 tm += rtime * (negative_range ? -1 : 1);
3414 #ifdef HAVE_GMTIME_R
3415 timeptr = gmtime_r(&tm, &gmt);
3416 #elif defined(MUTEX_LOCKS_AVAILABLE)
3417 privoxy_mutex_lock(&gmtime_mutex);
3418 timeptr = gmtime(&tm);
3419 privoxy_mutex_unlock(&gmtime_mutex);
3421 timeptr = gmtime(&tm);
3423 if ((NULL == timeptr) || !strftime(newheader,
3424 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
3426 log_error(LOG_LEVEL_ERROR,
3427 "Randomizing '%s' failed. Crunching the header without replacement.",
3434 *header = strdup("If-Modified-Since: ");
3435 string_append(header, newheader);
3437 if (*header == NULL)
3439 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3440 return JB_ERR_MEMORY;
3443 hours = rtime / 3600;
3444 minutes = rtime / 60 % 60;
3445 seconds = rtime % 60;
3447 log_error(LOG_LEVEL_HEADER,
3448 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3449 *header, (negative_range) ? "subtracted" : "added", hours,
3450 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3451 seconds, (seconds == 1) ? ")" : "s)");
3460 /*********************************************************************
3462 * Function : client_if_none_match
3464 * Description : Remove the If-None-Match header.
3467 * 1 : csp = Current client state (buffers, headers, etc...)
3468 * 2 : header = On input, pointer to header to modify.
3469 * On output, pointer to the modified header, or NULL
3470 * to remove the header. This function frees the
3471 * original string if necessary.
3473 * Returns : JB_ERR_OK on success, or
3474 * JB_ERR_MEMORY on out-of-memory error.
3476 *********************************************************************/
3477 static jb_err client_if_none_match(struct client_state *csp, char **header)
3479 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3481 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3489 /*********************************************************************
3491 * Function : client_x_filter
3493 * Description : Disables filtering if the client set "X-Filter: No".
3494 * Called from `sed'.
3497 * 1 : csp = Current client state (buffers, headers, etc...)
3498 * 2 : header = On input, pointer to header to modify.
3499 * On output, pointer to the modified header, or NULL
3500 * to remove the header. This function frees the
3501 * original string if necessary.
3503 * Returns : JB_ERR_OK on success
3505 *********************************************************************/
3506 jb_err client_x_filter(struct client_state *csp, char **header)
3508 if (0 == strcmpic(*header, "X-Filter: No"))
3510 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3512 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3516 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3518 log_error(LOG_LEVEL_HEADER,
3519 "force-text-mode overruled the client's request to fetch without filtering!");
3523 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3524 csp->flags |= CSP_FLAG_NO_FILTERING;
3525 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3527 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3535 /*********************************************************************
3537 * Function : client_range
3539 * Description : Removes Range, Request-Range and If-Range headers if
3540 * content filtering is enabled and the range doesn't
3543 * If the client's version of the document has been
3544 * altered by Privoxy, the server could interpret the
3545 * range differently than the client intended in which
3546 * case the user could end up with corrupted content.
3548 * If the range starts at byte 0 this isn't an issue
3549 * so the header can pass. Partial requests like this
3550 * are used to render preview images for videos without
3551 * downloading the whole video.
3553 * While HTTP doesn't require that range requests are
3554 * honoured and the client could simply abort the download
3555 * after receiving a sufficient amount of data, various
3556 * clients don't handle complete responses to range
3557 * requests gracefully and emit misleading error messages
3561 * 1 : csp = Current client state (buffers, headers, etc...)
3562 * 2 : header = On input, pointer to header to modify.
3563 * On output, pointer to the modified header, or NULL
3564 * to remove the header. This function frees the
3565 * original string if necessary.
3567 * Returns : JB_ERR_OK
3569 *********************************************************************/
3570 static jb_err client_range(struct client_state *csp, char **header)
3572 if (content_filters_enabled(csp->action)
3573 && (0 != strncmpic(strstr(*header, ":"), ": bytes=0-", 10)))
3575 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3576 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3583 /* the following functions add headers directly to the header list */
3585 /*********************************************************************
3587 * Function : client_host_adder
3589 * Description : Adds the Host: header field if it is missing.
3590 * Called from `sed'.
3593 * 1 : csp = Current client state (buffers, headers, etc...)
3595 * Returns : JB_ERR_OK on success, or
3596 * JB_ERR_MEMORY on out-of-memory error.
3598 *********************************************************************/
3599 static jb_err client_host_adder(struct client_state *csp)
3604 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3606 /* Header already set by the client, nothing to do. */
3610 if (!csp->http->hostport || !*(csp->http->hostport))
3612 /* XXX: When does this happen and why is it OK? */
3613 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3618 * remove 'user:pass@' from 'proto://user:pass@host'
3620 if ((p = strchr( csp->http->hostport, '@')) != NULL)
3626 p = csp->http->hostport;
3629 /* XXX: Just add it, we already made sure that it will be unique */
3630 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3631 err = enlist_unique_header(csp->headers, "Host", p);
3637 /*********************************************************************
3639 * Function : client_xtra_adder
3641 * Description : Used in the add_client_headers list. Called from `sed'.
3644 * 1 : csp = Current client state (buffers, headers, etc...)
3646 * Returns : JB_ERR_OK on success, or
3647 * JB_ERR_MEMORY on out-of-memory error.
3649 *********************************************************************/
3650 static jb_err client_xtra_adder(struct client_state *csp)
3652 struct list_entry *lst;
3655 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3656 lst ; lst = lst->next)
3658 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3659 err = enlist(csp->headers, lst->str);
3671 /*********************************************************************
3673 * Function : client_x_forwarded_for_adder
3675 * Description : Used in the add_client_headers list. Called from `sed'.
3678 * 1 : csp = Current client state (buffers, headers, etc...)
3680 * Returns : JB_ERR_OK on success, or
3681 * JB_ERR_MEMORY on out-of-memory error.
3683 *********************************************************************/
3684 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3686 char *header = NULL;
3689 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3690 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3691 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3694 * If we aren't adding X-Forwarded-For headers,
3695 * or we already appended an existing X-Forwarded-For
3696 * header, there's nothing left to do here.
3701 header = strdup("X-Forwarded-For: ");
3702 string_append(&header, csp->ip_addr_str);
3706 return JB_ERR_MEMORY;
3709 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3710 err = enlist(csp->headers, header);
3717 /*********************************************************************
3719 * Function : server_connection_adder
3721 * Description : Adds an appropriate "Connection:" header to csp->headers
3722 * unless the header was already present. Called from `sed'.
3725 * 1 : csp = Current client state (buffers, headers, etc...)
3727 * Returns : JB_ERR_OK on success, or
3728 * JB_ERR_MEMORY on out-of-memory error.
3730 *********************************************************************/
3731 static jb_err server_connection_adder(struct client_state *csp)
3733 const unsigned int flags = csp->flags;
3734 const char *response_status_line = csp->headers->first->str;
3735 static const char connection_close[] = "Connection: close";
3737 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3738 && (flags & CSP_FLAG_SERVER_CONNECTION_HEADER_SET))
3744 * XXX: if we downgraded the response, this check will fail.
3746 if ((csp->config->feature_flags &
3747 RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3748 && (NULL != response_status_line)
3749 && !strncmpic(response_status_line, "HTTP/1.1", 8)
3750 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3751 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3755 log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
3756 "without Connection header implies keep-alive.");
3757 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
3761 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3763 return enlist(csp->headers, connection_close);
3767 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3768 /*********************************************************************
3770 * Function : server_proxy_connection_adder
3772 * Description : Adds a "Proxy-Connection: keep-alive" header to
3773 * csp->headers when appropriate.
3776 * 1 : csp = Current client state (buffers, headers, etc...)
3778 * Returns : JB_ERR_OK on success, or
3779 * JB_ERR_MEMORY on out-of-memory error.
3781 *********************************************************************/
3782 static jb_err server_proxy_connection_adder(struct client_state *csp)
3784 static const char proxy_connection_header[] = "Proxy-Connection: keep-alive";
3785 jb_err err = JB_ERR_OK;
3787 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
3788 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3789 && !(csp->flags & CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET)
3790 && ((csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET)
3791 || (csp->flags & CSP_FLAG_CHUNKED)))
3793 log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
3794 err = enlist(csp->headers, proxy_connection_header);
3799 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3802 /*********************************************************************
3804 * Function : client_connection_header_adder
3806 * Description : Adds a proper "Connection:" header to csp->headers
3807 * unless the header was already present. Called from `sed'.
3810 * 1 : csp = Current client state (buffers, headers, etc...)
3812 * Returns : JB_ERR_OK on success, or
3813 * JB_ERR_MEMORY on out-of-memory error.
3815 *********************************************************************/
3816 static jb_err client_connection_header_adder(struct client_state *csp)
3818 static const char connection_close[] = "Connection: close";
3820 if (!(csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3821 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
3826 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3827 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3828 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3829 && (csp->http->ssl == 0)
3830 && !strcmpic(csp->http->ver, "HTTP/1.1"))
3832 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
3835 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3837 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3839 return enlist(csp->headers, connection_close);
3843 /*********************************************************************
3845 * Function : server_http
3847 * Description : - Save the HTTP Status into csp->http->status
3848 * - Set CT_TABOO to prevent filtering if the answer
3849 * is a partial range (HTTP status 206)
3850 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3854 * 1 : csp = Current client state (buffers, headers, etc...)
3855 * 2 : header = On input, pointer to header to modify.
3856 * On output, pointer to the modified header, or NULL
3857 * to remove the header. This function frees the
3858 * original string if necessary.
3860 * Returns : JB_ERR_OK on success, or
3861 * JB_ERR_MEMORY on out-of-memory error.
3863 *********************************************************************/
3864 static jb_err server_http(struct client_state *csp, char **header)
3866 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3867 if (csp->http->status == 206)
3869 csp->content_type = CT_TABOO;
3872 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3874 /* XXX: Should we do a real validity check here? */
3875 if (strlen(*header) > 8)
3878 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3883 * XXX: Should we block the request or
3884 * enlist a valid status code line here?
3886 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3887 "Downgrading to HTTP/1.0 impossible.");
3894 /*********************************************************************
3896 * Function : add_cooky_expiry_date
3898 * Description : Adds a cookie expiry date to a string.
3901 * 1 : cookie = On input, pointer to cookie to modify.
3902 * On output, pointer to the modified header.
3903 * The original string is freed.
3904 * 2 : lifetime = Seconds the cookie should be valid
3908 *********************************************************************/
3909 static void add_cookie_expiry_date(char **cookie, time_t lifetime)
3912 struct tm *timeptr = NULL;
3913 time_t expiry_date = time(NULL) + lifetime;
3914 #ifdef HAVE_GMTIME_R
3917 timeptr = gmtime_r(&expiry_date, &gmt);
3918 #elif defined(MUTEX_LOCKS_AVAILABLE)
3919 privoxy_mutex_lock(&gmtime_mutex);
3920 timeptr = gmtime(&expiry_date);
3921 privoxy_mutex_unlock(&gmtime_mutex);
3923 timeptr = gmtime(&expiry_date);
3926 if (NULL == timeptr)
3928 log_error(LOG_LEVEL_FATAL,
3929 "Failed to get the time in add_cooky_expiry_date()");
3931 strftime(tmp, sizeof(tmp), "; expires=%a, %d-%b-%Y %H:%M:%S GMT", timeptr);
3932 if (JB_ERR_OK != string_append(cookie, tmp))
3934 log_error(LOG_LEVEL_FATAL, "Out of memory in add_cooky_expiry()");
3939 /*********************************************************************
3941 * Function : server_set_cookie
3943 * Description : Handle the server "cookie" header properly.
3944 * Crunch, accept or rewrite it to a session cookie.
3945 * Called from `sed'.
3948 * 1 : csp = Current client state (buffers, headers, etc...)
3949 * 2 : header = On input, pointer to header to modify.
3950 * On output, pointer to the modified header, or NULL
3951 * to remove the header. This function frees the
3952 * original string if necessary.
3954 * Returns : JB_ERR_OK on success, or
3955 * JB_ERR_MEMORY on out-of-memory error.
3957 *********************************************************************/
3958 static jb_err server_set_cookie(struct client_state *csp, char **header)
3960 if ((csp->action->flags & ACTION_CRUNCH_INCOMING_COOKIES) != 0)
3962 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3965 else if ((0 != (csp->action->flags & ACTION_SESSION_COOKIES_ONLY))
3966 || (0 != (csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME)))
3970 long cookie_lifetime = 0;
3973 NO_EXPIRY_DATE_SPECIFIED,
3974 EXPIRY_DATE_ACCEPTABLE,
3975 EXPIRY_DATE_UNACCEPTABLE
3976 } expiry_date_status = NO_EXPIRY_DATE_SPECIFIED;
3978 /* A variable to store the tag we're working on */
3981 /* Skip "Set-Cookie:" (11 characters) in header */
3982 cur_tag = *header + 11;
3984 /* skip whitespace between "Set-Cookie:" and value */
3985 while (*cur_tag && privoxy_isspace(*cur_tag))
3992 if ((csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME) != 0)
3994 const char *param = csp->action->string[ACTION_STRING_LIMIT_COOKIE_LIFETIME];
3996 cookie_lifetime = strtol(param, NULL, 0);
3997 if (cookie_lifetime < 0)
3999 log_error(LOG_LEVEL_FATAL, "Invalid cookie lifetime limit: %s", param);
4001 cookie_lifetime *= 60;
4004 /* Loop through each tag in the cookie */
4008 char *next_tag = strchr(cur_tag, ';');
4009 if (next_tag != NULL)
4011 /* Skip the ';' character itself */
4014 /* skip whitespace ";" and start of tag */
4015 while (*next_tag && privoxy_isspace(*next_tag))
4022 /* "Next tag" is the end of the string */
4023 next_tag = cur_tag + strlen(cur_tag);
4027 * Check the expiration date to see
4028 * if the cookie is still valid, if yes,
4029 * rewrite it to a session cookie.
4031 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
4033 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
4035 if ((expiration_date[0] == '"')
4036 && (expiration_date[1] != '\0'))
4039 * Skip quotation mark. RFC 2109 10.1.2 seems to hint
4040 * that the expiration date isn't supposed to be quoted,
4041 * but some servers do it anyway.
4046 /* Did we detect the date properly? */
4047 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
4050 * Nope, treat it as if it was still valid.
4052 * XXX: Should we remove the whole cookie instead?
4054 log_error(LOG_LEVEL_ERROR,
4055 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
4056 string_move(cur_tag, next_tag);
4057 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4062 * Yes. Check if the cookie is still valid.
4064 * If the cookie is already expired it's probably
4065 * a delete cookie and even if it isn't, the browser
4066 * will discard it anyway.
4070 * XXX: timegm() isn't available on some AmigaOS
4071 * versions and our replacement doesn't work.
4073 * Our options are to either:
4075 * - disable session-cookies-only completely if timegm
4078 * - to simply remove all expired tags, like it has
4079 * been done until Privoxy 3.0.6 and to live with
4080 * the consequence that it can cause login/logout
4081 * problems on servers that don't validate their
4082 * input properly, or
4084 * - to replace it with mktime in which
4085 * case there is a slight chance of valid cookies
4086 * passing as already expired.
4088 * This is the way it's currently done and it's not
4089 * as bad as it sounds. If the missing GMT offset is
4090 * enough to change the result of the expiration check
4091 * the cookie will be only valid for a few hours
4092 * anyway, which in many cases will be shorter
4093 * than a browser session.
4095 if (cookie_time < now)
4097 log_error(LOG_LEVEL_HEADER,
4098 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4099 /* Just in case some clown sets more then one expiration date */
4101 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4103 else if ((cookie_lifetime != 0) && (cookie_time < (now + cookie_lifetime)))
4105 log_error(LOG_LEVEL_HEADER, "Cookie \'%s\' can pass unmodified. "
4106 "Its lifetime is below the limit.", *header);
4107 /* Just in case some clown sets more then one expiration date */
4109 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4114 * Still valid, delete expiration date by copying
4115 * the rest of the string over it.
4117 string_move(cur_tag, next_tag);
4119 /* That changed the header, need to issue a log message */
4120 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4123 * Note that the next tag has now been moved to *cur_tag,
4124 * so we do not need to update the cur_tag pointer.
4132 /* Move on to next cookie tag */
4137 if (expiry_date_status != EXPIRY_DATE_ACCEPTABLE)
4139 assert(NULL != *header);
4140 if (cookie_lifetime != 0)
4142 add_cookie_expiry_date(header, cookie_lifetime);
4143 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to: %s", *header);
4145 else if (expiry_date_status != NO_EXPIRY_DATE_SPECIFIED)
4147 log_error(LOG_LEVEL_HEADER,
4148 "Cookie rewritten to a temporary one: %s", *header);
4157 #ifdef FEATURE_FORCE_LOAD
4158 /*********************************************************************
4160 * Function : strclean
4162 * Description : In-Situ-Eliminate all occurrences of substring in
4166 * 1 : string = string to clean
4167 * 2 : substring = substring to eliminate
4169 * Returns : Number of eliminations
4171 *********************************************************************/
4172 int strclean(char *string, const char *substring)
4178 len = strlen(substring);
4180 while((pos = strstr(string, substring)) != NULL)
4187 while (*p++ != '\0');
4194 #endif /* def FEATURE_FORCE_LOAD */
4197 /*********************************************************************
4199 * Function : parse_header_time
4201 * Description : Parses time formats used in HTTP header strings
4202 * to get the numerical respresentation.
4205 * 1 : header_time = HTTP header time as string.
4206 * 2 : result = storage for header_time in seconds
4208 * Returns : JB_ERR_OK if the time format was recognized, or
4209 * JB_ERR_PARSE otherwise.
4211 *********************************************************************/
4212 static jb_err parse_header_time(const char *header_time, time_t *result)
4216 * Checking for two-digit years first in an
4217 * attempt to work around GNU libc's strptime()
4218 * reporting negative year values when using %Y.
4220 static const char time_formats[][22] = {
4221 /* Tue, 02-Jun-37 20:00:00 */
4222 "%a, %d-%b-%y %H:%M:%S",
4223 /* Tue, 02 Jun 2037 20:00:00 */
4224 "%a, %d %b %Y %H:%M:%S",
4225 /* Tue, 02-Jun-2037 20:00:00 */
4226 "%a, %d-%b-%Y %H:%M:%S",
4227 /* Tuesday, 02-Jun-2037 20:00:00 */
4228 "%A, %d-%b-%Y %H:%M:%S",
4229 /* Tuesday Jun 02 20:00:00 2037 */
4230 "%A %b %d %H:%M:%S %Y"
4234 for (i = 0; i < SZ(time_formats); i++)
4237 * Zero out gmt to prevent time zone offsets.
4238 * Documented to be required for GNU libc.
4240 memset(&gmt, 0, sizeof(gmt));
4242 if (NULL != strptime(header_time, time_formats[i], &gmt))
4244 /* Sanity check for GNU libc. */
4245 if (gmt.tm_year < 0)
4247 log_error(LOG_LEVEL_HEADER,
4248 "Failed to parse '%s' using '%s'. Moving on.",
4249 header_time, time_formats[i]);
4252 *result = timegm(&gmt);
4254 #ifdef FEATURE_STRPTIME_SANITY_CHECKS
4256 * Verify that parsing the date recreated from the first
4257 * parse operation gets the previous result. If it doesn't,
4258 * either strptime() or strftime() are malfunctioning.
4260 * We could string-compare the recreated date with the original
4261 * header date, but this leads to false positives as strptime()
4262 * may let %a accept all day formats while strftime() will only
4266 char recreated_date[100];
4270 tm = gmtime(result);
4271 strftime(recreated_date, sizeof(recreated_date), time_formats[i], tm);
4272 memset(&gmt, 0, sizeof(gmt));
4273 if (NULL == strptime(recreated_date, time_formats[i], &gmt))
4275 log_error(LOG_LEVEL_ERROR,
4276 "Failed to parse '%s' generated with '%s' to recreate '%s'.",
4277 recreated_date, time_formats[i], header_time);
4280 result2 = timegm(&gmt);
4281 if (*result != result2)
4283 log_error(LOG_LEVEL_ERROR, "strftime() and strptime() disagree. "
4284 "Format: '%s'. In: '%s', out: '%s'. %d != %d. Rejecting.",
4285 time_formats[i], header_time, recreated_date, *result, result2);
4295 return JB_ERR_PARSE;
4300 /*********************************************************************
4302 * Function : get_destination_from_headers
4304 * Description : Parse the "Host:" header to get the request's destination.
4305 * Only needed if the client's request was forcefully
4306 * redirected into Privoxy.
4308 * Code mainly copied from client_host() which is currently
4309 * run too late for this purpose.
4312 * 1 : headers = List of headers (one of them hopefully being
4313 * the "Host:" header)
4314 * 2 : http = storage for the result (host, port and hostport).
4316 * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
4317 * JB_ERR_PARSE if the host header couldn't be found,
4318 * JB_ERR_OK otherwise.
4320 *********************************************************************/
4321 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4327 host = get_header_value(headers, "Host:");
4331 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4332 return JB_ERR_PARSE;
4335 p = strdup_or_die(host);
4337 q = strdup_or_die(p);
4339 freez(http->hostport);
4343 q = strchr(http->host, ':');
4346 /* Terminate hostname and evaluate port string */
4348 http->port = atoi(q);
4352 http->port = http->ssl ? 443 : 80;
4355 /* Rebuild request URL */
4357 http->url = strdup(http->ssl ? "https://" : "http://");
4358 string_append(&http->url, http->hostport);
4359 string_append(&http->url, http->path);
4360 if (http->url == NULL)
4362 return JB_ERR_MEMORY;
4365 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4373 /*********************************************************************
4375 * Function : create_forged_referrer
4377 * Description : Helper for client_referrer to forge a referer as
4378 * 'http://hostname[:port]/' to fool stupid
4379 * checks for in-site links
4382 * 1 : header = Pointer to header pointer
4383 * 2 : hostport = Host and optionally port as string
4385 * Returns : JB_ERR_OK in case of success, or
4386 * JB_ERR_MEMORY in case of memory problems.
4388 *********************************************************************/
4389 static jb_err create_forged_referrer(char **header, const char *hostport)
4391 assert(NULL == *header);
4393 *header = strdup("Referer: http://");
4394 string_append(header, hostport);
4395 string_append(header, "/");
4397 if (NULL == *header)
4399 return JB_ERR_MEMORY;
4402 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);