1 const char parsers_rcs[] = "$Id: parsers.c,v 1.285 2014/04/21 12:04:01 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.
758 * 1 : header = A header with linear white space to reduce.
762 *********************************************************************/
763 static void normalize_lws(char *header)
769 if (privoxy_isspace(*p) && privoxy_isspace(*(p+1)))
773 while (privoxy_isspace(*q))
777 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
783 log_error(LOG_LEVEL_HEADER,
784 "Converting tab to space in '%s'", header);
789 char *end_of_token = strstr(p+1, "\"");
791 if (NULL != end_of_token)
793 /* Don't mess with quoted text. */
798 log_error(LOG_LEVEL_HEADER,
799 "Ignoring single quote in '%s'", header);
805 p = strchr(header, ':');
806 if ((p != NULL) && (p != header) && privoxy_isspace(*(p-1)))
809 * There's still space before the colon.
817 /*********************************************************************
819 * Function : get_header
821 * Description : This (odd) routine will parse the csp->iob
822 * to get the next complete header.
825 * 1 : iob = The I/O buffer to parse, usually csp->iob.
827 * Returns : Any one of the following:
829 * 1) a pointer to a dynamically allocated string that contains a header line
830 * 2) NULL indicating that the end of the header was reached
831 * 3) "" indicating that the end of the iob was reached before finding
832 * a complete header line.
834 *********************************************************************/
835 char *get_header(struct iob *iob)
839 header = get_header_line(iob);
841 if ((header == NULL) || (*header == '\0'))
844 * No complete header read yet, tell the client.
849 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
852 * Header spans multiple lines, append the next one.
854 char *continued_header;
856 continued_header = get_header_line(iob);
857 if ((continued_header == NULL) || (*continued_header == '\0'))
860 * No complete header read yet, return what we got.
861 * XXX: Should "unread" header instead.
863 log_error(LOG_LEVEL_INFO,
864 "Failed to read a multi-line header properly: '%s'",
869 if (JB_ERR_OK != string_join(&header, continued_header))
871 log_error(LOG_LEVEL_FATAL,
872 "Out of memory while appending multiple headers.");
876 /* XXX: remove before next stable release. */
877 log_error(LOG_LEVEL_HEADER,
878 "Merged multiple header lines to: '%s'",
883 normalize_lws(header);
890 /*********************************************************************
892 * Function : get_header_line
894 * Description : This (odd) routine will parse the csp->iob
895 * to get the next header line.
898 * 1 : iob = The I/O buffer to parse, usually csp->iob.
900 * Returns : Any one of the following:
902 * 1) a pointer to a dynamically allocated string that contains a header line
903 * 2) NULL indicating that the end of the header was reached
904 * 3) "" indicating that the end of the iob was reached before finding
905 * a complete header line.
907 *********************************************************************/
908 static char *get_header_line(struct iob *iob)
912 if ((iob->cur == NULL)
913 || ((p = strchr(iob->cur, '\n')) == NULL))
915 return(""); /* couldn't find a complete header */
920 ret = strdup(iob->cur);
923 /* FIXME No way to handle error properly */
924 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
930 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
932 /* is this a blank line (i.e. the end of the header) ? */
944 /*********************************************************************
946 * Function : get_header_value
948 * Description : Get the value of a given header from a chained list
949 * of header lines or return NULL if no such header is
950 * present in the list.
953 * 1 : header_list = pointer to list
954 * 2 : header_name = string with name of header to look for.
955 * Trailing colon required, capitalization
958 * Returns : NULL if not found, else value of header
960 *********************************************************************/
961 char *get_header_value(const struct list *header_list, const char *header_name)
963 struct list_entry *cur_entry;
969 length = strlen(header_name);
971 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
975 if (!strncmpic(cur_entry->str, header_name, length))
978 * Found: return pointer to start of value
980 ret = cur_entry->str + length;
981 while (*ret && privoxy_isspace(*ret)) ret++;
995 /*********************************************************************
997 * Function : scan_headers
999 * Description : Scans headers, applies tags and updates action bits.
1002 * 1 : csp = Current client state (buffers, headers, etc...)
1004 * Returns : JB_ERR_OK
1006 *********************************************************************/
1007 static jb_err scan_headers(struct client_state *csp)
1009 struct list_entry *h; /* Header */
1010 jb_err err = JB_ERR_OK;
1012 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1014 /* Header crunch()ed in previous run? -> ignore */
1015 if (h->str == NULL) continue;
1016 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1017 err = header_tagger(csp, h->str);
1024 /*********************************************************************
1026 * Function : enforce_header_order
1028 * Description : Enforces a given header order.
1031 * 1 : headers = List of headers to order.
1032 * 2 : ordered_headers = List of ordered header names.
1036 *********************************************************************/
1037 static void enforce_header_order(struct list *headers, const struct list *ordered_headers)
1039 struct list_entry *sorted_header;
1040 struct list new_headers[1];
1041 struct list_entry *header;
1043 init_list(new_headers);
1045 /* The request line is always the first "header" */
1047 assert(NULL != headers->first->str);
1048 enlist(new_headers, headers->first->str);
1049 freez(headers->first->str)
1051 /* Enlist the specified headers in the given order */
1053 for (sorted_header = ordered_headers->first; sorted_header != NULL;
1054 sorted_header = sorted_header->next)
1056 const size_t sorted_header_length = strlen(sorted_header->str);
1057 for (header = headers->first; header != NULL; header = header->next)
1059 /* Header enlisted in previous run? -> ignore */
1060 if (header->str == NULL) continue;
1062 if (0 == strncmpic(sorted_header->str, header->str, sorted_header_length)
1063 && (header->str[sorted_header_length] == ':'))
1065 log_error(LOG_LEVEL_HEADER, "Enlisting sorted header %s", header->str);
1066 if (JB_ERR_OK != enlist(new_headers, header->str))
1068 log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
1075 /* Enlist the rest of the headers behind the ordered ones */
1076 for (header = headers->first; header != NULL; header = header->next)
1078 /* Header enlisted in previous run? -> ignore */
1079 if (header->str == NULL) continue;
1081 log_error(LOG_LEVEL_HEADER,
1082 "Enlisting left-over header %s", header->str);
1083 if (JB_ERR_OK != enlist(new_headers, header->str))
1085 log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
1090 list_remove_all(headers);
1091 list_duplicate(headers, new_headers);
1092 list_remove_all(new_headers);
1098 /*********************************************************************
1102 * Description : add, delete or modify lines in the HTTP header streams.
1103 * On entry, it receives a linked list of headers space
1104 * that was allocated dynamically (both the list nodes
1105 * and the header contents).
1107 * As a side effect it frees the space used by the original
1111 * 1 : csp = Current client state (buffers, headers, etc...)
1112 * 2 : filter_server_headers = Boolean to switch between
1113 * server and header filtering.
1115 * Returns : JB_ERR_OK in case off success, or
1116 * JB_ERR_MEMORY on out-of-memory error.
1118 *********************************************************************/
1119 jb_err sed(struct client_state *csp, int filter_server_headers)
1121 /* XXX: use more descriptive names. */
1122 struct list_entry *p;
1123 const struct parsers *v;
1124 const add_header_func_ptr *f;
1125 jb_err err = JB_ERR_OK;
1129 if (filter_server_headers)
1131 v = server_patterns;
1132 f = add_server_headers;
1133 check_negative_tag_patterns(csp, PATTERN_SPEC_NO_RESPONSE_TAG_PATTERN);
1137 v = client_patterns;
1138 f = add_client_headers;
1139 check_negative_tag_patterns(csp, PATTERN_SPEC_NO_REQUEST_TAG_PATTERN);
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 enum filter_type wanted_filter_type;
1278 int multi_action_index;
1281 struct re_filterfile_spec *b;
1282 struct list_entry *tag_name;
1284 const size_t header_length = strlen(header);
1286 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1288 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1289 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1293 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1294 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1297 if (list_is_empty(csp->action->multi[multi_action_index])
1298 || filters_available(csp) == FALSE)
1300 /* Return early if no taggers apply or if none are available. */
1304 /* Execute all applying taggers */
1305 for (tag_name = csp->action->multi[multi_action_index]->first;
1306 NULL != tag_name; tag_name = tag_name->next)
1308 char *modified_tag = NULL;
1310 size_t size = header_length;
1313 b = get_filter(csp, tag_name->str, wanted_filter_type);
1319 joblist = b->joblist;
1321 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1323 if (NULL == joblist)
1325 log_error(LOG_LEVEL_RE_FILTER,
1326 "Tagger %s has empty joblist. Nothing to do.", b->name);
1330 /* execute their pcrs_joblist on the header. */
1331 for (job = joblist; NULL != job; job = job->next)
1333 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1337 /* Success, continue with the modified version. */
1346 /* Tagger doesn't match */
1349 /* Regex failure, log it but continue anyway. */
1350 assert(NULL != header);
1351 log_error(LOG_LEVEL_ERROR,
1352 "Problems with tagger \'%s\' and header \'%s\': %s",
1353 b->name, *header, pcrs_strerror(hits));
1355 freez(modified_tag);
1359 if (b->dynamic) pcrs_free_joblist(joblist);
1361 /* If this tagger matched */
1367 * There is no technical limitation which makes
1368 * it impossible to use empty tags, but I assume
1369 * no one would do it intentionally.
1372 log_error(LOG_LEVEL_INFO,
1373 "Tagger \'%s\' created an empty tag. Ignored.", b->name);
1377 if (!list_contains_item(csp->tags, tag))
1379 if (JB_ERR_OK != enlist(csp->tags, tag))
1381 log_error(LOG_LEVEL_ERROR,
1382 "Insufficient memory to add tag \'%s\', "
1383 "based on tagger \'%s\' and header \'%s\'",
1384 tag, b->name, *header);
1388 char *action_message;
1390 * update the action bits right away, to make
1391 * tagging based on tags set by earlier taggers
1392 * of the same kind possible.
1394 if (update_action_bits_for_tag(csp, tag))
1396 action_message = "Action bits updated accordingly.";
1400 action_message = "No action bits update necessary.";
1403 log_error(LOG_LEVEL_HEADER,
1404 "Tagger \'%s\' added tag \'%s\'. %s",
1405 b->name, tag, action_message);
1410 /* XXX: Is this log-worthy? */
1411 log_error(LOG_LEVEL_HEADER,
1412 "Tagger \'%s\' didn't add tag \'%s\'. Tag already present",
1422 /* here begins the family of parser functions that reformat header lines */
1424 /*********************************************************************
1426 * Function : filter_header
1428 * Description : Executes all text substitutions from all applying
1429 * +(server|client)-header-filter actions on the header.
1430 * Most of the code was copied from pcrs_filter_response,
1431 * including the rather short variable names
1434 * 1 : csp = Current client state (buffers, headers, etc...)
1435 * 2 : header = On input, pointer to header to modify.
1436 * On output, pointer to the modified header, or NULL
1437 * to remove the header. This function frees the
1438 * original string if necessary.
1440 * Returns : JB_ERR_OK on success and always succeeds
1442 *********************************************************************/
1443 static jb_err filter_header(struct client_state *csp, char **header)
1447 size_t size = strlen(*header);
1449 char *newheader = NULL;
1452 struct re_filterfile_spec *b;
1453 struct list_entry *filtername;
1455 enum filter_type wanted_filter_type;
1456 int multi_action_index;
1458 if (csp->flags & CSP_FLAG_NO_FILTERING)
1463 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1465 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1466 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1470 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1471 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1474 if (list_is_empty(csp->action->multi[multi_action_index])
1475 || filters_available(csp) == FALSE)
1477 /* Return early if no filters apply or if none are available. */
1481 /* Execute all applying header filters */
1482 for (filtername = csp->action->multi[multi_action_index]->first;
1483 filtername != NULL; filtername = filtername->next)
1485 int current_hits = 0;
1488 b = get_filter(csp, filtername->str, wanted_filter_type);
1494 joblist = b->joblist;
1496 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1498 if (NULL == joblist)
1500 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1504 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1505 *header, size, b->name);
1507 /* Apply all jobs from the joblist */
1508 for (job = joblist; NULL != job; job = job->next)
1510 matches = pcrs_execute(job, *header, size, &newheader, &size);
1513 current_hits += matches;
1514 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1516 *header = newheader;
1518 else if (0 == matches)
1520 /* Filter doesn't change header */
1526 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1527 *header, b->name, pcrs_strerror(matches));
1528 if (newheader != NULL)
1530 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1536 if (b->dynamic) pcrs_free_joblist(joblist);
1538 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1539 hits += current_hits;
1543 * Additionally checking for hits is important because if
1544 * the continue hack is triggered, server headers can
1545 * arrive empty to separate multiple heads from each other.
1547 if ((0 == size) && hits)
1549 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
1557 /*********************************************************************
1559 * Function : server_connection
1561 * Description : Makes sure a proper "Connection:" header is
1562 * set and signals connection_header_adder to
1566 * 1 : csp = Current client state (buffers, headers, etc...)
1567 * 2 : header = On input, pointer to header to modify.
1568 * On output, pointer to the modified header, or NULL
1569 * to remove the header. This function frees the
1570 * original string if necessary.
1572 * Returns : JB_ERR_OK on success.
1574 *********************************************************************/
1575 static jb_err server_connection(struct client_state *csp, char **header)
1577 if (!strcmpic(*header, "Connection: keep-alive")
1578 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1579 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
1583 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1584 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1586 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
1589 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE))
1591 log_error(LOG_LEVEL_HEADER,
1592 "Keeping the server header '%s' around.", *header);
1595 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
1597 char *old_header = *header;
1599 *header = strdup_or_die("Connection: close");
1600 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
1605 /* Signal server_connection_adder() to return early. */
1606 csp->flags |= CSP_FLAG_SERVER_CONNECTION_HEADER_SET;
1612 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1613 /*********************************************************************
1615 * Function : server_keep_alive
1617 * Description : Stores the server's keep alive timeout.
1620 * 1 : csp = Current client state (buffers, headers, etc...)
1621 * 2 : header = On input, pointer to header to modify.
1622 * On output, pointer to the modified header, or NULL
1623 * to remove the header. This function frees the
1624 * original string if necessary.
1626 * Returns : JB_ERR_OK.
1628 *********************************************************************/
1629 static jb_err server_keep_alive(struct client_state *csp, char **header)
1631 unsigned int keep_alive_timeout;
1632 const char *timeout_position = strstr(*header, "timeout=");
1634 if ((NULL == timeout_position)
1635 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1637 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1641 if (keep_alive_timeout < csp->server_connection.keep_alive_timeout)
1643 log_error(LOG_LEVEL_HEADER,
1644 "Reducing keep-alive timeout from %u to %u.",
1645 csp->server_connection.keep_alive_timeout, keep_alive_timeout);
1646 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1650 /* XXX: Is this log worthy? */
1651 log_error(LOG_LEVEL_HEADER,
1652 "Server keep-alive timeout is %u. Sticking with %u.",
1653 keep_alive_timeout, csp->server_connection.keep_alive_timeout);
1655 csp->flags |= CSP_FLAG_SERVER_KEEP_ALIVE_TIMEOUT_SET;
1662 /*********************************************************************
1664 * Function : server_proxy_connection
1666 * Description : Figures out whether or not we should add a
1667 * Proxy-Connection header.
1670 * 1 : csp = Current client state (buffers, headers, etc...)
1671 * 2 : header = On input, pointer to header to modify.
1672 * On output, pointer to the modified header, or NULL
1673 * to remove the header. This function frees the
1674 * original string if necessary.
1676 * Returns : JB_ERR_OK.
1678 *********************************************************************/
1679 static jb_err server_proxy_connection(struct client_state *csp, char **header)
1681 csp->flags |= CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET;
1686 /*********************************************************************
1688 * Function : proxy_authentication
1690 * Description : Removes headers that are relevant for proxy
1691 * authentication unless forwarding them has
1692 * been explicitly requested.
1695 * 1 : csp = Current client state (buffers, headers, etc...)
1696 * 2 : header = On input, pointer to header to modify.
1697 * On output, pointer to the modified header, or NULL
1698 * to remove the header. This function frees the
1699 * original string if necessary.
1701 * Returns : JB_ERR_OK.
1703 *********************************************************************/
1704 static jb_err proxy_authentication(struct client_state *csp, char **header)
1706 if ((csp->config->feature_flags &
1707 RUNTIME_FEATURE_FORWARD_PROXY_AUTHENTICATION_HEADERS) == 0) {
1708 log_error(LOG_LEVEL_HEADER,
1709 "Forwarding proxy authentication headers is disabled. Crunching: %s", *header);
1716 /*********************************************************************
1718 * Function : client_keep_alive
1720 * Description : Stores the client's keep alive timeout.
1723 * 1 : csp = Current client state (buffers, headers, etc...)
1724 * 2 : header = On input, pointer to header to modify.
1725 * On output, pointer to the modified header, or NULL
1726 * to remove the header. This function frees the
1727 * original string if necessary.
1729 * Returns : JB_ERR_OK.
1731 *********************************************************************/
1732 static jb_err client_keep_alive(struct client_state *csp, char **header)
1734 unsigned int keep_alive_timeout;
1735 char *timeout_position;
1737 if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1739 log_error(LOG_LEVEL_HEADER,
1740 "keep-alive support is disabled. Crunching: %s.", *header);
1745 /* Check for parameter-less format "Keep-Alive: 100" */
1746 timeout_position = strstr(*header, ": ");
1747 if ((NULL == timeout_position)
1748 || (1 != sscanf(timeout_position, ": %u", &keep_alive_timeout)))
1750 /* Assume parameter format "Keep-Alive: timeout=100" */
1751 timeout_position = strstr(*header, "timeout=");
1752 if ((NULL == timeout_position)
1753 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1755 log_error(LOG_LEVEL_HEADER,
1756 "Couldn't parse: '%s'. Using default timeout %u",
1757 *header, csp->config->keep_alive_timeout);
1764 if (keep_alive_timeout < csp->config->keep_alive_timeout)
1766 log_error(LOG_LEVEL_HEADER,
1767 "Reducing keep-alive timeout from %u to %u.",
1768 csp->config->keep_alive_timeout, keep_alive_timeout);
1769 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1773 /* XXX: Is this log worthy? */
1774 log_error(LOG_LEVEL_HEADER,
1775 "Client keep-alive timeout is %u. Sticking with %u.",
1776 keep_alive_timeout, csp->config->keep_alive_timeout);
1784 /*********************************************************************
1786 * Function : get_content_length
1788 * Description : Gets the content length specified in a
1789 * Content-Length header.
1792 * 1 : header_value = The Content-Length header value.
1793 * 2 : length = Storage to return the value.
1795 * Returns : JB_ERR_OK on success, or
1796 * JB_ERR_PARSE if no value is recognized.
1798 *********************************************************************/
1799 static jb_err get_content_length(const char *header_value, unsigned long long *length)
1802 assert(sizeof(unsigned long long) > 4);
1803 if (1 != sscanf(header_value, "%I64u", length))
1805 if (1 != sscanf(header_value, "%llu", length))
1808 return JB_ERR_PARSE;
1815 /*********************************************************************
1817 * Function : client_save_content_length
1819 * Description : Save the Content-Length sent by the client.
1822 * 1 : csp = Current client state (buffers, headers, etc...)
1823 * 2 : header = On input, pointer to header to modify.
1824 * On output, pointer to the modified header, or NULL
1825 * to remove the header. This function frees the
1826 * original string if necessary.
1828 * Returns : JB_ERR_OK on success, or
1829 * JB_ERR_MEMORY on out-of-memory error.
1831 *********************************************************************/
1832 static jb_err client_save_content_length(struct client_state *csp, char **header)
1834 unsigned long long content_length = 0;
1835 const char *header_value;
1837 assert(*(*header+14) == ':');
1839 header_value = *header + 15;
1840 if (JB_ERR_OK != get_content_length(header_value, &content_length))
1842 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
1847 csp->expected_client_content_length = content_length;
1852 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1856 /*********************************************************************
1858 * Function : client_connection
1860 * Description : Makes sure a proper "Connection:" header is
1861 * set and signals connection_header_adder
1865 * 1 : csp = Current client state (buffers, headers, etc...)
1866 * 2 : header = On input, pointer to header to modify.
1867 * On output, pointer to the modified header, or NULL
1868 * to remove the header. This function frees the
1869 * original string if necessary.
1871 * Returns : JB_ERR_OK on success.
1873 *********************************************************************/
1874 static jb_err client_connection(struct client_state *csp, char **header)
1876 static const char connection_close[] = "Connection: close";
1878 if (!strcmpic(*header, connection_close))
1880 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1881 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING)
1882 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1884 if (!strcmpic(csp->http->ver, "HTTP/1.1"))
1886 log_error(LOG_LEVEL_HEADER,
1887 "Removing \'%s\' to imply keep-alive.", *header);
1890 * While we imply keep-alive to the server,
1891 * we have to remember that the client didn't.
1893 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1897 char *old_header = *header;
1899 *header = strdup_or_die("Connection: keep-alive");
1900 log_error(LOG_LEVEL_HEADER,
1901 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1907 log_error(LOG_LEVEL_HEADER,
1908 "Keeping the client header '%s' around. "
1909 "The connection will not be kept alive.",
1911 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1914 else if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
1915 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
1917 log_error(LOG_LEVEL_HEADER,
1918 "Keeping the client header '%s' around. "
1919 "The server connection will be kept alive if possible.",
1921 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1922 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1926 char *old_header = *header;
1928 *header = strdup_or_die(connection_close);
1929 log_error(LOG_LEVEL_HEADER,
1930 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1934 /* Signal client_connection_header_adder() to return early. */
1935 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
1941 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1942 /*********************************************************************
1944 * Function : client_proxy_connection
1946 * Description : Sets the CLIENT_CONNECTION_KEEP_ALIVE flag when
1947 * appropriate and removes the Proxy-Connection
1951 * 1 : csp = Current client state (buffers, headers, etc...)
1952 * 2 : header = On input, pointer to header to modify.
1953 * On output, pointer to the modified header, or NULL
1954 * to remove the header. This function frees the
1955 * original string if necessary.
1957 * Returns : JB_ERR_OK
1959 *********************************************************************/
1960 static jb_err client_proxy_connection(struct client_state *csp, char **header)
1962 if (0 == (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
1963 && (csp->http->ssl == 0)
1964 && (NULL == strstr(*header, "close")))
1966 log_error(LOG_LEVEL_HEADER,
1967 "The client connection can be kept alive due to: %s", *header);
1968 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1970 crumble(csp, header);
1974 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1977 /*********************************************************************
1979 * Function : client_transfer_encoding
1981 * Description : Raise the CSP_FLAG_CHUNKED_CLIENT_BODY flag if
1982 * the request body is "chunked"
1984 * XXX: Currently not called through sed() as we
1985 * need the flag earlier on. Should be fixed.
1988 * 1 : csp = Current client state (buffers, headers, etc...)
1989 * 2 : header = On input, pointer to header to modify.
1990 * On output, pointer to the modified header, or NULL
1991 * to remove the header. This function frees the
1992 * original string if necessary.
1994 * Returns : JB_ERR_OK on success, or
1996 *********************************************************************/
1997 jb_err client_transfer_encoding(struct client_state *csp, char **header)
1999 if (strstr(*header, "chunked"))
2001 csp->flags |= CSP_FLAG_CHUNKED_CLIENT_BODY;
2002 log_error(LOG_LEVEL_HEADER, "Expecting chunked client body");
2009 /*********************************************************************
2011 * Function : crumble
2013 * Description : This is called if a header matches a pattern to "crunch"
2016 * 1 : csp = Current client state (buffers, headers, etc...)
2017 * 2 : header = On input, pointer to header to modify.
2018 * On output, pointer to the modified header, or NULL
2019 * to remove the header. This function frees the
2020 * original string if necessary.
2022 * Returns : JB_ERR_OK on success, or
2023 * JB_ERR_MEMORY on out-of-memory error.
2025 *********************************************************************/
2026 static jb_err crumble(struct client_state *csp, char **header)
2029 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2035 /*********************************************************************
2037 * Function : crunch_server_header
2039 * Description : Crunch server header if it matches a string supplied by the
2040 * user. Called from `sed'.
2043 * 1 : csp = Current client state (buffers, headers, etc...)
2044 * 2 : header = On input, pointer to header to modify.
2045 * On output, pointer to the modified header, or NULL
2046 * to remove the header. This function frees the
2047 * original string if necessary.
2049 * Returns : JB_ERR_OK on success and always succeeds
2051 *********************************************************************/
2052 static jb_err crunch_server_header(struct client_state *csp, char **header)
2054 const char *crunch_pattern;
2056 /* Do we feel like crunching? */
2057 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2059 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2061 /* Is the current header the lucky one? */
2062 if (strstr(*header, crunch_pattern))
2064 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2073 /*********************************************************************
2075 * Function : server_content_type
2077 * Description : Set the content-type for filterable types (text/.*,
2078 * .*xml.*, .*script.* and image/gif) unless filtering has been
2079 * forbidden (CT_TABOO) while parsing earlier headers.
2080 * NOTE: Since text/plain is commonly used by web servers
2081 * for files whose correct type is unknown, we don't
2082 * set CT_TEXT for it.
2085 * 1 : csp = Current client state (buffers, headers, etc...)
2086 * 2 : header = On input, pointer to header to modify.
2087 * On output, pointer to the modified header, or NULL
2088 * to remove the header. This function frees the
2089 * original string if necessary.
2091 * Returns : JB_ERR_OK on success, or
2092 * JB_ERR_MEMORY on out-of-memory error.
2094 *********************************************************************/
2095 static jb_err server_content_type(struct client_state *csp, char **header)
2097 /* Remove header if it isn't the first Content-Type header */
2098 if ((csp->content_type & CT_DECLARED))
2100 if (content_filters_enabled(csp->action))
2103 * Making sure the client interprets the content the same way
2104 * Privoxy did is only relevant if Privoxy modified it.
2106 * Checking for this is "hard" as it's not yet known when
2107 * this function is called, thus go shopping and and just
2108 * check if Privoxy could filter it.
2110 * The main thing is that we don't mess with the headers
2111 * unless the user signalled that it's acceptable.
2113 log_error(LOG_LEVEL_HEADER,
2114 "Multiple Content-Type headers detected. "
2115 "Removing and ignoring: %s",
2123 * Signal that the Content-Type has been set.
2125 csp->content_type |= CT_DECLARED;
2127 if (!(csp->content_type & CT_TABOO))
2130 * XXX: The assumption that text/plain is a sign of
2131 * binary data seems to be somewhat unreasonable nowadays
2132 * and should be dropped after 3.0.8 is out.
2134 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2135 || strstr(*header, "xml")
2136 || strstr(*header, "script"))
2138 csp->content_type |= CT_TEXT;
2140 else if (strstr(*header, "image/gif"))
2142 csp->content_type |= CT_GIF;
2147 * Are we messing with the content type?
2149 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2152 * Make sure the user doesn't accidentally
2153 * change the content type of binary documents.
2155 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2158 *header = strdup_or_die("Content-Type: ");
2159 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2163 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2164 return JB_ERR_MEMORY;
2166 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2170 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2171 "It doesn't look like a content type that should be filtered. "
2172 "Enable force-text-mode if you know what you're doing.", *header);
2180 /*********************************************************************
2182 * Function : server_transfer_coding
2184 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2185 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2186 * - Remove header if body was chunked but has been
2187 * de-chunked for filtering.
2190 * 1 : csp = Current client state (buffers, headers, etc...)
2191 * 2 : header = On input, pointer to header to modify.
2192 * On output, pointer to the modified header, or NULL
2193 * to remove the header. This function frees the
2194 * original string if necessary.
2196 * Returns : JB_ERR_OK on success, or
2197 * JB_ERR_MEMORY on out-of-memory error.
2199 *********************************************************************/
2200 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2203 * Turn off pcrs and gif filtering if body compressed
2205 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2209 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2211 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2212 csp->http->cmd, *header);
2213 #endif /* def FEATURE_ZLIB */
2214 csp->content_type = CT_TABOO;
2218 * Raise flag if body chunked
2220 if (strstr(*header, "chunked"))
2222 csp->flags |= CSP_FLAG_CHUNKED;
2225 * If the body was modified, it has been de-chunked first
2226 * and the header must be removed.
2228 * FIXME: If there is more than one transfer encoding,
2229 * only the "chunked" part should be removed here.
2231 if (csp->flags & CSP_FLAG_MODIFIED)
2233 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2242 /*********************************************************************
2244 * Function : server_content_encoding
2246 * Description : Used to check if the content is compressed, and if
2247 * FEATURE_ZLIB is disabled, filtering is disabled as
2250 * If FEATURE_ZLIB is enabled and the compression type
2251 * supported, the content is marked for decompression.
2253 * XXX: Doesn't properly deal with multiple or with
2254 * unsupported but unknown encodings.
2255 * Is case-sensitive but shouldn't be.
2258 * 1 : csp = Current client state (buffers, headers, etc...)
2259 * 2 : header = On input, pointer to header to modify.
2260 * On output, pointer to the modified header, or NULL
2261 * to remove the header. This function frees the
2262 * original string if necessary.
2264 * Returns : JB_ERR_OK on success, or
2265 * JB_ERR_MEMORY on out-of-memory error.
2267 *********************************************************************/
2268 static jb_err server_content_encoding(struct client_state *csp, char **header)
2271 if (strstr(*header, "sdch"))
2274 * Shared Dictionary Compression over HTTP isn't supported,
2275 * filtering it anyway is pretty much guaranteed to mess up
2278 csp->content_type |= CT_TABOO;
2281 * Log a warning if the user expects the content to be filtered.
2283 if ((csp->rlist != NULL) &&
2284 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2286 log_error(LOG_LEVEL_INFO,
2287 "SDCH-compressed content detected, content filtering disabled. "
2288 "Consider suppressing SDCH offers made by the client.");
2291 else if (strstr(*header, "gzip"))
2293 /* Mark for gzip decompression */
2294 csp->content_type |= CT_GZIP;
2296 else if (strstr(*header, "deflate"))
2298 /* Mark for zlib decompression */
2299 csp->content_type |= CT_DEFLATE;
2301 else if (strstr(*header, "compress"))
2304 * We can't decompress this; therefore we can't filter
2307 csp->content_type |= CT_TABOO;
2309 #else /* !defined(FEATURE_ZLIB) */
2311 * XXX: Using a black list here isn't the right approach.
2313 * In case of SDCH, building with zlib support isn't
2316 if (strstr(*header, "gzip") ||
2317 strstr(*header, "compress") ||
2318 strstr(*header, "deflate") ||
2319 strstr(*header, "sdch"))
2322 * Body is compressed, turn off pcrs and gif filtering.
2324 csp->content_type |= CT_TABOO;
2327 * Log a warning if the user expects the content to be filtered.
2329 if ((csp->rlist != NULL) &&
2330 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2332 log_error(LOG_LEVEL_INFO,
2333 "Compressed content detected, content filtering disabled. "
2334 "Consider recompiling Privoxy with zlib support or "
2335 "enable the prevent-compression action.");
2338 #endif /* defined(FEATURE_ZLIB) */
2346 /*********************************************************************
2348 * Function : server_adjust_content_encoding
2350 * Description : Remove the Content-Encoding header if the
2351 * decompression was successful and the content
2355 * 1 : csp = Current client state (buffers, headers, etc...)
2356 * 2 : header = On input, pointer to header to modify.
2357 * On output, pointer to the modified header, or NULL
2358 * to remove the header. This function frees the
2359 * original string if necessary.
2361 * Returns : JB_ERR_OK on success, or
2362 * JB_ERR_MEMORY on out-of-memory error.
2364 *********************************************************************/
2365 static jb_err server_adjust_content_encoding(struct client_state *csp, char **header)
2367 if ((csp->flags & CSP_FLAG_MODIFIED)
2368 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2371 * We successfully decompressed the content,
2372 * and have to clean the header now, so the
2373 * client no longer expects compressed data.
2375 * XXX: There is a difference between cleaning
2376 * and removing it completely.
2378 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2385 #endif /* defined(FEATURE_ZLIB) */
2388 /*********************************************************************
2390 * Function : server_adjust_content_length
2392 * Description : Adjust Content-Length header if we modified
2396 * 1 : csp = Current client state (buffers, headers, etc...)
2397 * 2 : header = On input, pointer to header to modify.
2398 * On output, pointer to the modified header, or NULL
2399 * to remove the header. This function frees the
2400 * original string if necessary.
2402 * Returns : JB_ERR_OK on success, or
2403 * JB_ERR_MEMORY on out-of-memory error.
2405 *********************************************************************/
2406 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2408 /* Regenerate header if the content was modified. */
2409 if (csp->flags & CSP_FLAG_MODIFIED)
2411 const size_t header_length = 50;
2413 *header = malloc(header_length);
2414 if (*header == NULL)
2416 return JB_ERR_MEMORY;
2418 create_content_length_header(csp->content_length, *header, header_length);
2419 log_error(LOG_LEVEL_HEADER,
2420 "Adjusted Content-Length to %llu", csp->content_length);
2427 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2428 /*********************************************************************
2430 * Function : server_save_content_length
2432 * Description : Save the Content-Length sent by the server.
2435 * 1 : csp = Current client state (buffers, headers, etc...)
2436 * 2 : header = On input, pointer to header to modify.
2437 * On output, pointer to the modified header, or NULL
2438 * to remove the header. This function frees the
2439 * original string if necessary.
2441 * Returns : JB_ERR_OK on success, or
2442 * JB_ERR_MEMORY on out-of-memory error.
2444 *********************************************************************/
2445 static jb_err server_save_content_length(struct client_state *csp, char **header)
2447 unsigned long long content_length = 0;
2448 const char *header_value;
2450 assert(*(*header+14) == ':');
2452 header_value = *header + 15;
2453 if (JB_ERR_OK != get_content_length(header_value, &content_length))
2455 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2460 csp->expected_content_length = content_length;
2461 csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
2462 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2467 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2470 /*********************************************************************
2472 * Function : server_content_md5
2474 * Description : Crumble any Content-MD5 headers if the document was
2475 * modified. FIXME: Should we re-compute instead?
2478 * 1 : csp = Current client state (buffers, headers, etc...)
2479 * 2 : header = On input, pointer to header to modify.
2480 * On output, pointer to the modified header, or NULL
2481 * to remove the header. This function frees the
2482 * original string if necessary.
2484 * Returns : JB_ERR_OK on success, or
2485 * JB_ERR_MEMORY on out-of-memory error.
2487 *********************************************************************/
2488 static jb_err server_content_md5(struct client_state *csp, char **header)
2490 if (csp->flags & CSP_FLAG_MODIFIED)
2492 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2500 /*********************************************************************
2502 * Function : server_content_disposition
2504 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2505 * Called from `sed'.
2508 * 1 : csp = Current client state (buffers, headers, etc...)
2509 * 2 : header = On input, pointer to header to modify.
2510 * On output, pointer to the modified header, or NULL
2511 * to remove the header. This function frees the
2512 * original string if necessary.
2514 * Returns : JB_ERR_OK on success, or
2515 * JB_ERR_MEMORY on out-of-memory error.
2517 *********************************************************************/
2518 static jb_err server_content_disposition(struct client_state *csp, char **header)
2523 * Are we messing with the Content-Disposition header?
2525 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2531 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2533 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2536 * Blocking content-disposition header
2538 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2545 * Replacing Content-Disposition header
2548 *header = strdup("Content-Disposition: ");
2549 string_append(header, newval);
2551 if (*header != NULL)
2553 log_error(LOG_LEVEL_HEADER,
2554 "Content-Disposition header crunched and replaced with: %s", *header);
2557 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2561 /*********************************************************************
2563 * Function : server_last_modified
2565 * Description : Changes Last-Modified header to the actual date
2566 * to help hide-if-modified-since.
2567 * Called from `sed'.
2570 * 1 : csp = Current client state (buffers, headers, etc...)
2571 * 2 : header = On input, pointer to header to modify.
2572 * On output, pointer to the modified header, or NULL
2573 * to remove the header. This function frees the
2574 * original string if necessary.
2576 * Returns : JB_ERR_OK on success, or
2577 * JB_ERR_MEMORY on out-of-memory error.
2579 *********************************************************************/
2580 static jb_err server_last_modified(struct client_state *csp, char **header)
2583 time_t last_modified;
2587 * Are we messing with the Last-Modified header?
2589 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2595 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2597 if (0 == strcmpic(newval, "block"))
2600 * Blocking Last-Modified header. Useless but why not.
2602 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2606 else if (0 == strcmpic(newval, "reset-to-request-time"))
2609 * Setting Last-Modified Header to now.
2612 get_http_time(0, buf, sizeof(buf));
2614 *header = strdup("Last-Modified: ");
2615 string_append(header, buf);
2617 if (*header == NULL)
2619 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2623 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2626 else if (0 == strcmpic(newval, "randomize"))
2628 const char *header_time = *header + sizeof("Last-Modified:");
2630 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2632 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2634 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2640 struct tm *timeptr = NULL;
2642 #ifdef HAVE_GMTIME_R
2646 rtime = (long int)difftime(now, last_modified);
2649 long int days, hours, minutes, seconds;
2650 const int negative_delta = (rtime < 0);
2655 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2657 rtime = pick_from_range(rtime);
2662 last_modified += rtime;
2663 #ifdef HAVE_GMTIME_R
2664 timeptr = gmtime_r(&last_modified, &gmt);
2665 #elif defined(MUTEX_LOCKS_AVAILABLE)
2666 privoxy_mutex_lock(&gmtime_mutex);
2667 timeptr = gmtime(&last_modified);
2668 privoxy_mutex_unlock(&gmtime_mutex);
2670 timeptr = gmtime(&last_modified);
2672 if ((NULL == timeptr) || !strftime(newheader,
2673 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
2675 log_error(LOG_LEVEL_ERROR,
2676 "Randomizing '%s' failed. Crunching the header without replacement.",
2683 *header = strdup("Last-Modified: ");
2684 string_append(header, newheader);
2686 if (*header == NULL)
2688 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2689 return JB_ERR_MEMORY;
2692 days = rtime / (3600 * 24);
2693 hours = rtime / 3600 % 24;
2694 minutes = rtime / 60 % 60;
2695 seconds = rtime % 60;
2697 log_error(LOG_LEVEL_HEADER,
2698 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2699 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2700 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2704 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2713 /*********************************************************************
2715 * Function : client_accept_encoding
2717 * Description : Rewrite the client's Accept-Encoding header so that
2718 * if doesn't allow compression, if the action applies.
2719 * Note: For HTTP/1.0 the absence of the header is enough.
2722 * 1 : csp = Current client state (buffers, headers, etc...)
2723 * 2 : header = On input, pointer to header to modify.
2724 * On output, pointer to the modified header, or NULL
2725 * to remove the header. This function frees the
2726 * original string if necessary.
2728 * Returns : JB_ERR_OK on success, or
2729 * JB_ERR_MEMORY on out-of-memory error.
2731 *********************************************************************/
2732 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2734 #ifdef FEATURE_COMPRESSION
2735 if ((csp->config->feature_flags & RUNTIME_FEATURE_COMPRESSION)
2736 && strstr(*header, "deflate"))
2738 csp->flags |= CSP_FLAG_CLIENT_SUPPORTS_DEFLATE;
2741 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2743 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2751 /*********************************************************************
2753 * Function : client_te
2755 * Description : Rewrite the client's TE header so that
2756 * if doesn't allow compression, if the action applies.
2759 * 1 : csp = Current client state (buffers, headers, etc...)
2760 * 2 : header = On input, pointer to header to modify.
2761 * On output, pointer to the modified header, or NULL
2762 * to remove the header. This function frees the
2763 * original string if necessary.
2765 * Returns : JB_ERR_OK on success, or
2766 * JB_ERR_MEMORY on out-of-memory error.
2768 *********************************************************************/
2769 static jb_err client_te(struct client_state *csp, char **header)
2771 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2774 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2781 /*********************************************************************
2783 * Function : client_referrer
2785 * Description : Handle the "referer" config setting properly.
2786 * Called from `sed'.
2789 * 1 : csp = Current client state (buffers, headers, etc...)
2790 * 2 : header = On input, pointer to header to modify.
2791 * On output, pointer to the modified header, or NULL
2792 * to remove the header. This function frees the
2793 * original string if necessary.
2795 * Returns : JB_ERR_OK on success, or
2796 * JB_ERR_MEMORY on out-of-memory error.
2798 *********************************************************************/
2799 static jb_err client_referrer(struct client_state *csp, char **header)
2801 const char *parameter;
2802 /* booleans for parameters we have to check multiple times */
2803 int parameter_conditional_block;
2804 int parameter_conditional_forge;
2806 #ifdef FEATURE_FORCE_LOAD
2808 * Since the referrer can include the prefix even
2809 * if the request itself is non-forced, we must
2810 * clean it unconditionally.
2812 * XXX: strclean is too broad
2814 strclean(*header, FORCE_PREFIX);
2815 #endif /* def FEATURE_FORCE_LOAD */
2817 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2819 /* Nothing left to do */
2823 parameter = csp->action->string[ACTION_STRING_REFERER];
2824 assert(parameter != NULL);
2825 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2826 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2828 if (!parameter_conditional_block && !parameter_conditional_forge)
2831 * As conditional-block and conditional-forge are the only
2832 * parameters that rely on the original referrer, we can
2833 * remove it now for all the others.
2838 if (0 == strcmpic(parameter, "block"))
2840 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2843 else if (parameter_conditional_block || parameter_conditional_forge)
2845 return handle_conditional_hide_referrer_parameter(header,
2846 csp->http->hostport, parameter_conditional_block);
2848 else if (0 == strcmpic(parameter, "forge"))
2850 return create_forged_referrer(header, csp->http->hostport);
2854 /* interpret parameter as user-supplied referer to fake */
2855 return create_fake_referrer(header, parameter);
2860 /*********************************************************************
2862 * Function : client_accept_language
2864 * Description : Handle the "Accept-Language" config setting properly.
2865 * Called from `sed'.
2868 * 1 : csp = Current client state (buffers, headers, etc...)
2869 * 2 : header = On input, pointer to header to modify.
2870 * On output, pointer to the modified header, or NULL
2871 * to remove the header. This function frees the
2872 * original string if necessary.
2874 * Returns : JB_ERR_OK on success, or
2875 * JB_ERR_MEMORY on out-of-memory error.
2877 *********************************************************************/
2878 static jb_err client_accept_language(struct client_state *csp, char **header)
2883 * Are we messing with the Accept-Language?
2885 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2887 /*I don't think so*/
2891 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2893 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2896 * Blocking Accept-Language header
2898 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2905 * Replacing Accept-Language header
2908 *header = strdup("Accept-Language: ");
2909 string_append(header, newval);
2911 if (*header == NULL)
2913 log_error(LOG_LEVEL_ERROR,
2914 "Insufficient memory. Accept-Language header crunched without replacement.");
2918 log_error(LOG_LEVEL_HEADER,
2919 "Accept-Language header crunched and replaced with: %s", *header);
2922 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2926 /*********************************************************************
2928 * Function : crunch_client_header
2930 * Description : Crunch client header if it matches a string supplied by the
2931 * user. Called from `sed'.
2934 * 1 : csp = Current client state (buffers, headers, etc...)
2935 * 2 : header = On input, pointer to header to modify.
2936 * On output, pointer to the modified header, or NULL
2937 * to remove the header. This function frees the
2938 * original string if necessary.
2940 * Returns : JB_ERR_OK on success and always succeeds
2942 *********************************************************************/
2943 static jb_err crunch_client_header(struct client_state *csp, char **header)
2945 const char *crunch_pattern;
2947 /* Do we feel like crunching? */
2948 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2950 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2952 /* Is the current header the lucky one? */
2953 if (strstr(*header, crunch_pattern))
2955 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
2963 /*********************************************************************
2965 * Function : client_uagent
2967 * Description : Handle the "user-agent" config setting properly
2968 * and remember its original value to enable browser
2969 * bug workarounds. Called from `sed'.
2972 * 1 : csp = Current client state (buffers, headers, etc...)
2973 * 2 : header = On input, pointer to header to modify.
2974 * On output, pointer to the modified header, or NULL
2975 * to remove the header. This function frees the
2976 * original string if necessary.
2978 * Returns : JB_ERR_OK on success, or
2979 * JB_ERR_MEMORY on out-of-memory error.
2981 *********************************************************************/
2982 static jb_err client_uagent(struct client_state *csp, char **header)
2986 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
2991 newval = csp->action->string[ACTION_STRING_USER_AGENT];
2998 *header = strdup("User-Agent: ");
2999 string_append(header, newval);
3001 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
3003 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3007 /*********************************************************************
3009 * Function : client_ua
3011 * Description : Handle "ua-" headers properly. Called from `sed'.
3014 * 1 : csp = Current client state (buffers, headers, etc...)
3015 * 2 : header = On input, pointer to header to modify.
3016 * On output, pointer to the modified header, or NULL
3017 * to remove the header. This function frees the
3018 * original string if necessary.
3020 * Returns : JB_ERR_OK on success, or
3021 * JB_ERR_MEMORY on out-of-memory error.
3023 *********************************************************************/
3024 static jb_err client_ua(struct client_state *csp, char **header)
3026 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3028 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3036 /*********************************************************************
3038 * Function : client_from
3040 * Description : Handle the "from" config setting properly.
3041 * Called from `sed'.
3044 * 1 : csp = Current client state (buffers, headers, etc...)
3045 * 2 : header = On input, pointer to header to modify.
3046 * On output, pointer to the modified header, or NULL
3047 * to remove the header. This function frees the
3048 * original string if necessary.
3050 * Returns : JB_ERR_OK on success, or
3051 * JB_ERR_MEMORY on out-of-memory error.
3053 *********************************************************************/
3054 static jb_err client_from(struct client_state *csp, char **header)
3058 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3065 newval = csp->action->string[ACTION_STRING_FROM];
3068 * Are we blocking the e-mail address?
3070 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
3072 log_error(LOG_LEVEL_HEADER, "crunched From!");
3076 log_error(LOG_LEVEL_HEADER, " modified");
3078 *header = strdup("From: ");
3079 string_append(header, newval);
3081 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3085 /*********************************************************************
3087 * Function : client_send_cookie
3089 * Description : Crunches the "cookie" header if necessary.
3090 * Called from `sed'.
3092 * XXX: Stupid name, doesn't send squat.
3095 * 1 : csp = Current client state (buffers, headers, etc...)
3096 * 2 : header = On input, pointer to header to modify.
3097 * On output, pointer to the modified header, or NULL
3098 * to remove the header. This function frees the
3099 * original string if necessary.
3101 * Returns : JB_ERR_OK on success, or
3102 * JB_ERR_MEMORY on out-of-memory error.
3104 *********************************************************************/
3105 static jb_err client_send_cookie(struct client_state *csp, char **header)
3107 if (csp->action->flags & ACTION_CRUNCH_OUTGOING_COOKIES)
3109 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3117 /*********************************************************************
3119 * Function : client_x_forwarded
3121 * Description : Handle the "x-forwarded-for" config setting properly,
3122 * also used in the add_client_headers list. Called from `sed'.
3125 * 1 : csp = Current client state (buffers, headers, etc...)
3126 * 2 : header = On input, pointer to header to modify.
3127 * On output, pointer to the modified header, or NULL
3128 * to remove the header. This function frees the
3129 * original string if necessary.
3131 * Returns : JB_ERR_OK on success, or
3132 * JB_ERR_MEMORY on out-of-memory error.
3134 *********************************************************************/
3135 jb_err client_x_forwarded(struct client_state *csp, char **header)
3137 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
3139 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
3141 if (0 == strcmpic(parameter, "block"))
3144 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3146 else if (0 == strcmpic(parameter, "add"))
3148 string_append(header, ", ");
3149 string_append(header, csp->ip_addr_str);
3151 if (*header == NULL)
3153 return JB_ERR_MEMORY;
3155 log_error(LOG_LEVEL_HEADER,
3156 "Appended client IP address to %s", *header);
3157 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
3161 log_error(LOG_LEVEL_FATAL,
3162 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
3170 /*********************************************************************
3172 * Function : client_max_forwards
3174 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3175 * from the value of the Max-Forwards header field.
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 static jb_err client_max_forwards(struct client_state *csp, char **header)
3192 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3193 (0 == strcmpic(csp->http->gpc, "options")))
3195 assert(*(*header+12) == ':');
3196 if (1 == sscanf(*header+12, ": %d", &max_forwards))
3198 if (max_forwards > 0)
3200 snprintf(*header, strlen(*header)+1, "Max-Forwards: %d", --max_forwards);
3201 log_error(LOG_LEVEL_HEADER,
3202 "Max-Forwards value for %s request reduced to %d.",
3203 csp->http->gpc, max_forwards);
3205 else if (max_forwards < 0)
3207 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3213 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3222 /*********************************************************************
3224 * Function : client_host
3226 * Description : If the request URI did not contain host and
3227 * port information, parse and evaluate the Host
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_host(struct client_state *csp, char **header)
3245 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3246 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3249 p = strdup_or_die((*header)+6);
3251 q = strdup_or_die(p);
3253 freez(csp->http->hostport);
3254 csp->http->hostport = p;
3255 freez(csp->http->host);
3256 csp->http->host = q;
3257 q = strchr(csp->http->host, ':');
3260 /* Terminate hostname and evaluate port string */
3262 csp->http->port = atoi(q);
3266 csp->http->port = csp->http->ssl ? 443 : 80;
3269 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3270 csp->http->hostport, csp->http->host, csp->http->port);
3273 /* Signal client_host_adder() to return right away */
3274 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3280 /*********************************************************************
3282 * Function : client_if_modified_since
3284 * Description : Remove or modify the If-Modified-Since header.
3287 * 1 : csp = Current client state (buffers, headers, etc...)
3288 * 2 : header = On input, pointer to header to modify.
3289 * On output, pointer to the modified header, or NULL
3290 * to remove the header. This function frees the
3291 * original string if necessary.
3293 * Returns : JB_ERR_OK on success, or
3294 * JB_ERR_MEMORY on out-of-memory error.
3296 *********************************************************************/
3297 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3300 #ifdef HAVE_GMTIME_R
3303 struct tm *timeptr = NULL;
3308 if (0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3311 * The client got an error message because of a temporary problem,
3312 * the problem is gone and the client now tries to revalidate our
3313 * error message on the real server. The revalidation would always
3314 * end with the transmission of the whole document and there is
3315 * no need to expose the bogus If-Modified-Since header.
3317 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3320 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3322 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3324 if ((0 == strcmpic(newval, "block")))
3326 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3329 else /* add random value */
3331 const char *header_time = *header + sizeof("If-Modified-Since:");
3333 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3335 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3340 long int hours, minutes, seconds;
3341 long int rtime = strtol(newval, &endptr, 0);
3342 const int negative_range = (rtime < 0);
3346 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3347 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3353 rtime = pick_from_range(rtime);
3357 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3360 tm += rtime * (negative_range ? -1 : 1);
3361 #ifdef HAVE_GMTIME_R
3362 timeptr = gmtime_r(&tm, &gmt);
3363 #elif defined(MUTEX_LOCKS_AVAILABLE)
3364 privoxy_mutex_lock(&gmtime_mutex);
3365 timeptr = gmtime(&tm);
3366 privoxy_mutex_unlock(&gmtime_mutex);
3368 timeptr = gmtime(&tm);
3370 if ((NULL == timeptr) || !strftime(newheader,
3371 sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr))
3373 log_error(LOG_LEVEL_ERROR,
3374 "Randomizing '%s' failed. Crunching the header without replacement.",
3381 *header = strdup("If-Modified-Since: ");
3382 string_append(header, newheader);
3384 if (*header == NULL)
3386 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3387 return JB_ERR_MEMORY;
3390 hours = rtime / 3600;
3391 minutes = rtime / 60 % 60;
3392 seconds = rtime % 60;
3394 log_error(LOG_LEVEL_HEADER,
3395 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3396 *header, (negative_range) ? "subtracted" : "added", hours,
3397 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3398 seconds, (seconds == 1) ? ")" : "s)");
3407 /*********************************************************************
3409 * Function : client_if_none_match
3411 * Description : Remove the If-None-Match header.
3414 * 1 : csp = Current client state (buffers, headers, etc...)
3415 * 2 : header = On input, pointer to header to modify.
3416 * On output, pointer to the modified header, or NULL
3417 * to remove the header. This function frees the
3418 * original string if necessary.
3420 * Returns : JB_ERR_OK on success, or
3421 * JB_ERR_MEMORY on out-of-memory error.
3423 *********************************************************************/
3424 static jb_err client_if_none_match(struct client_state *csp, char **header)
3426 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3428 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3436 /*********************************************************************
3438 * Function : client_x_filter
3440 * Description : Disables filtering if the client set "X-Filter: No".
3441 * Called from `sed'.
3444 * 1 : csp = Current client state (buffers, headers, etc...)
3445 * 2 : header = On input, pointer to header to modify.
3446 * On output, pointer to the modified header, or NULL
3447 * to remove the header. This function frees the
3448 * original string if necessary.
3450 * Returns : JB_ERR_OK on success
3452 *********************************************************************/
3453 jb_err client_x_filter(struct client_state *csp, char **header)
3455 if (0 == strcmpic(*header, "X-Filter: No"))
3457 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3459 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3463 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3465 log_error(LOG_LEVEL_HEADER,
3466 "force-text-mode overruled the client's request to fetch without filtering!");
3470 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3471 csp->flags |= CSP_FLAG_NO_FILTERING;
3472 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3474 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3482 /*********************************************************************
3484 * Function : client_range
3486 * Description : Removes Range, Request-Range and If-Range headers if
3487 * content filtering is enabled and the range doesn't
3490 * If the client's version of the document has been
3491 * altered by Privoxy, the server could interpret the
3492 * range differently than the client intended in which
3493 * case the user could end up with corrupted content.
3495 * If the range starts at byte 0 this isn't an issue
3496 * so the header can pass. Partial requests like this
3497 * are used to render preview images for videos without
3498 * downloading the whole video.
3500 * While HTTP doesn't require that range requests are
3501 * honoured and the client could simply abort the download
3502 * after receiving a sufficient amount of data, various
3503 * clients don't handle complete responses to range
3504 * requests gracefully and emit misleading error messages
3508 * 1 : csp = Current client state (buffers, headers, etc...)
3509 * 2 : header = On input, pointer to header to modify.
3510 * On output, pointer to the modified header, or NULL
3511 * to remove the header. This function frees the
3512 * original string if necessary.
3514 * Returns : JB_ERR_OK
3516 *********************************************************************/
3517 static jb_err client_range(struct client_state *csp, char **header)
3519 if (content_filters_enabled(csp->action)
3520 && (0 != strncmpic(strstr(*header, ":"), ": bytes=0-", 10)))
3522 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3523 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3530 /* the following functions add headers directly to the header list */
3532 /*********************************************************************
3534 * Function : client_host_adder
3536 * Description : Adds the Host: header field if it is missing.
3537 * Called from `sed'.
3540 * 1 : csp = Current client state (buffers, headers, etc...)
3542 * Returns : JB_ERR_OK on success, or
3543 * JB_ERR_MEMORY on out-of-memory error.
3545 *********************************************************************/
3546 static jb_err client_host_adder(struct client_state *csp)
3551 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3553 /* Header already set by the client, nothing to do. */
3557 if (!csp->http->hostport || !*(csp->http->hostport))
3559 /* XXX: When does this happen and why is it OK? */
3560 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3565 * remove 'user:pass@' from 'proto://user:pass@host'
3567 if ((p = strchr( csp->http->hostport, '@')) != NULL)
3573 p = csp->http->hostport;
3576 /* XXX: Just add it, we already made sure that it will be unique */
3577 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3578 err = enlist_unique_header(csp->headers, "Host", p);
3584 /*********************************************************************
3586 * Function : client_xtra_adder
3588 * Description : Used in the add_client_headers list. Called from `sed'.
3591 * 1 : csp = Current client state (buffers, headers, etc...)
3593 * Returns : JB_ERR_OK on success, or
3594 * JB_ERR_MEMORY on out-of-memory error.
3596 *********************************************************************/
3597 static jb_err client_xtra_adder(struct client_state *csp)
3599 struct list_entry *lst;
3602 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3603 lst ; lst = lst->next)
3605 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3606 err = enlist(csp->headers, lst->str);
3618 /*********************************************************************
3620 * Function : client_x_forwarded_for_adder
3622 * Description : Used in the add_client_headers list. Called from `sed'.
3625 * 1 : csp = Current client state (buffers, headers, etc...)
3627 * Returns : JB_ERR_OK on success, or
3628 * JB_ERR_MEMORY on out-of-memory error.
3630 *********************************************************************/
3631 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3633 char *header = NULL;
3636 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3637 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3638 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3641 * If we aren't adding X-Forwarded-For headers,
3642 * or we already appended an existing X-Forwarded-For
3643 * header, there's nothing left to do here.
3648 header = strdup("X-Forwarded-For: ");
3649 string_append(&header, csp->ip_addr_str);
3653 return JB_ERR_MEMORY;
3656 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3657 err = enlist(csp->headers, header);
3664 /*********************************************************************
3666 * Function : server_connection_adder
3668 * Description : Adds an appropriate "Connection:" header to csp->headers
3669 * unless the header was already present. Called from `sed'.
3672 * 1 : csp = Current client state (buffers, headers, etc...)
3674 * Returns : JB_ERR_OK on success, or
3675 * JB_ERR_MEMORY on out-of-memory error.
3677 *********************************************************************/
3678 static jb_err server_connection_adder(struct client_state *csp)
3680 const unsigned int flags = csp->flags;
3681 const char *response_status_line = csp->headers->first->str;
3682 static const char connection_close[] = "Connection: close";
3684 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3685 && (flags & CSP_FLAG_SERVER_CONNECTION_HEADER_SET))
3691 * XXX: if we downgraded the response, this check will fail.
3693 if ((csp->config->feature_flags &
3694 RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3695 && (NULL != response_status_line)
3696 && !strncmpic(response_status_line, "HTTP/1.1", 8)
3697 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3698 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3702 log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
3703 "without Connection header implies keep-alive.");
3704 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
3708 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3710 return enlist(csp->headers, connection_close);
3714 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3715 /*********************************************************************
3717 * Function : server_proxy_connection_adder
3719 * Description : Adds a "Proxy-Connection: keep-alive" header to
3720 * csp->headers when appropriate.
3723 * 1 : csp = Current client state (buffers, headers, etc...)
3725 * Returns : JB_ERR_OK on success, or
3726 * JB_ERR_MEMORY on out-of-memory error.
3728 *********************************************************************/
3729 static jb_err server_proxy_connection_adder(struct client_state *csp)
3731 static const char proxy_connection_header[] = "Proxy-Connection: keep-alive";
3732 jb_err err = JB_ERR_OK;
3734 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
3735 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3736 && !(csp->flags & CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET)
3737 && ((csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET)
3738 || (csp->flags & CSP_FLAG_CHUNKED)))
3740 log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
3741 err = enlist(csp->headers, proxy_connection_header);
3746 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3749 /*********************************************************************
3751 * Function : client_connection_header_adder
3753 * Description : Adds a proper "Connection:" header to csp->headers
3754 * unless the header was already present. Called from `sed'.
3757 * 1 : csp = Current client state (buffers, headers, etc...)
3759 * Returns : JB_ERR_OK on success, or
3760 * JB_ERR_MEMORY on out-of-memory error.
3762 *********************************************************************/
3763 static jb_err client_connection_header_adder(struct client_state *csp)
3765 static const char connection_close[] = "Connection: close";
3767 if (!(csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3768 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
3773 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3774 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3775 && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
3776 && (csp->http->ssl == 0)
3777 && !strcmpic(csp->http->ver, "HTTP/1.1"))
3779 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
3782 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3784 log_error(LOG_LEVEL_HEADER, "Adding: %s", connection_close);
3786 return enlist(csp->headers, connection_close);
3790 /*********************************************************************
3792 * Function : server_http
3794 * Description : - Save the HTTP Status into csp->http->status
3795 * - Set CT_TABOO to prevent filtering if the answer
3796 * is a partial range (HTTP status 206)
3797 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3801 * 1 : csp = Current client state (buffers, headers, etc...)
3802 * 2 : header = On input, pointer to header to modify.
3803 * On output, pointer to the modified header, or NULL
3804 * to remove the header. This function frees the
3805 * original string if necessary.
3807 * Returns : JB_ERR_OK on success, or
3808 * JB_ERR_MEMORY on out-of-memory error.
3810 *********************************************************************/
3811 static jb_err server_http(struct client_state *csp, char **header)
3813 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3814 if (csp->http->status == 206)
3816 csp->content_type = CT_TABOO;
3819 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3821 /* XXX: Should we do a real validity check here? */
3822 if (strlen(*header) > 8)
3825 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3830 * XXX: Should we block the request or
3831 * enlist a valid status code line here?
3833 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3834 "Downgrading to HTTP/1.0 impossible.");
3841 /*********************************************************************
3843 * Function : add_cooky_expiry_date
3845 * Description : Adds a cookie expiry date to a string.
3848 * 1 : cookie = On input, pointer to cookie to modify.
3849 * On output, pointer to the modified header.
3850 * The original string is freed.
3851 * 2 : lifetime = Seconds the cookie should be valid
3855 *********************************************************************/
3856 static void add_cookie_expiry_date(char **cookie, time_t lifetime)
3859 struct tm *timeptr = NULL;
3860 time_t expiry_date = time(NULL) + lifetime;
3861 #ifdef HAVE_GMTIME_R
3864 timeptr = gmtime_r(&expiry_date, &gmt);
3865 #elif defined(MUTEX_LOCKS_AVAILABLE)
3866 privoxy_mutex_lock(&gmtime_mutex);
3867 timeptr = gmtime(&expiry_date);
3868 privoxy_mutex_unlock(&gmtime_mutex);
3870 timeptr = gmtime(&expiry_date);
3873 if (NULL == timeptr)
3875 log_error(LOG_LEVEL_FATAL,
3876 "Failed to get the time in add_cooky_expiry_date()");
3878 strftime(tmp, sizeof(tmp), "; expires=%a, %d-%b-%Y %H:%M:%S GMT", timeptr);
3879 if (JB_ERR_OK != string_append(cookie, tmp))
3881 log_error(LOG_LEVEL_FATAL, "Out of memory in add_cooky_expiry()");
3886 /*********************************************************************
3888 * Function : server_set_cookie
3890 * Description : Handle the server "cookie" header properly.
3891 * Crunch, accept or rewrite it to a session cookie.
3892 * Called from `sed'.
3895 * 1 : csp = Current client state (buffers, headers, etc...)
3896 * 2 : header = On input, pointer to header to modify.
3897 * On output, pointer to the modified header, or NULL
3898 * to remove the header. This function frees the
3899 * original string if necessary.
3901 * Returns : JB_ERR_OK on success, or
3902 * JB_ERR_MEMORY on out-of-memory error.
3904 *********************************************************************/
3905 static jb_err server_set_cookie(struct client_state *csp, char **header)
3907 if ((csp->action->flags & ACTION_CRUNCH_INCOMING_COOKIES) != 0)
3909 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3912 else if ((0 != (csp->action->flags & ACTION_SESSION_COOKIES_ONLY))
3913 || (0 != (csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME)))
3917 long cookie_lifetime = 0;
3920 NO_EXPIRY_DATE_SPECIFIED,
3921 EXPIRY_DATE_ACCEPTABLE,
3922 EXPIRY_DATE_UNACCEPTABLE
3923 } expiry_date_status = NO_EXPIRY_DATE_SPECIFIED;
3925 /* A variable to store the tag we're working on */
3928 /* Skip "Set-Cookie:" (11 characters) in header */
3929 cur_tag = *header + 11;
3931 /* skip whitespace between "Set-Cookie:" and value */
3932 while (*cur_tag && privoxy_isspace(*cur_tag))
3939 if ((csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME) != 0)
3941 const char *param = csp->action->string[ACTION_STRING_LIMIT_COOKIE_LIFETIME];
3943 cookie_lifetime = strtol(param, NULL, 0);
3944 if (cookie_lifetime < 0)
3946 log_error(LOG_LEVEL_FATAL, "Invalid cookie lifetime limit: %s", param);
3948 cookie_lifetime *= 60;
3951 /* Loop through each tag in the cookie */
3955 char *next_tag = strchr(cur_tag, ';');
3956 if (next_tag != NULL)
3958 /* Skip the ';' character itself */
3961 /* skip whitespace ";" and start of tag */
3962 while (*next_tag && privoxy_isspace(*next_tag))
3969 /* "Next tag" is the end of the string */
3970 next_tag = cur_tag + strlen(cur_tag);
3974 * Check the expiration date to see
3975 * if the cookie is still valid, if yes,
3976 * rewrite it to a session cookie.
3978 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
3980 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
3982 if ((expiration_date[0] == '"')
3983 && (expiration_date[1] != '\0'))
3986 * Skip quotation mark. RFC 2109 10.1.2 seems to hint
3987 * that the expiration date isn't supposed to be quoted,
3988 * but some servers do it anyway.
3993 /* Did we detect the date properly? */
3994 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
3997 * Nope, treat it as if it was still valid.
3999 * XXX: Should we remove the whole cookie instead?
4001 log_error(LOG_LEVEL_ERROR,
4002 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
4003 string_move(cur_tag, next_tag);
4004 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4009 * Yes. Check if the cookie is still valid.
4011 * If the cookie is already expired it's probably
4012 * a delete cookie and even if it isn't, the browser
4013 * will discard it anyway.
4017 * XXX: timegm() isn't available on some AmigaOS
4018 * versions and our replacement doesn't work.
4020 * Our options are to either:
4022 * - disable session-cookies-only completely if timegm
4025 * - to simply remove all expired tags, like it has
4026 * been done until Privoxy 3.0.6 and to live with
4027 * the consequence that it can cause login/logout
4028 * problems on servers that don't validate their
4029 * input properly, or
4031 * - to replace it with mktime in which
4032 * case there is a slight chance of valid cookies
4033 * passing as already expired.
4035 * This is the way it's currently done and it's not
4036 * as bad as it sounds. If the missing GMT offset is
4037 * enough to change the result of the expiration check
4038 * the cookie will be only valid for a few hours
4039 * anyway, which in many cases will be shorter
4040 * than a browser session.
4042 if (cookie_time < now)
4044 log_error(LOG_LEVEL_HEADER,
4045 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
4046 /* Just in case some clown sets more then one expiration date */
4048 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4050 else if ((cookie_lifetime != 0) && (cookie_time < (now + cookie_lifetime)))
4052 log_error(LOG_LEVEL_HEADER, "Cookie \'%s\' can pass unmodified. "
4053 "Its lifetime is below the limit.", *header);
4054 /* Just in case some clown sets more then one expiration date */
4056 expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
4061 * Still valid, delete expiration date by copying
4062 * the rest of the string over it.
4064 string_move(cur_tag, next_tag);
4066 /* That changed the header, need to issue a log message */
4067 expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
4070 * Note that the next tag has now been moved to *cur_tag,
4071 * so we do not need to update the cur_tag pointer.
4079 /* Move on to next cookie tag */
4084 if (expiry_date_status != EXPIRY_DATE_ACCEPTABLE)
4086 assert(NULL != *header);
4087 if (cookie_lifetime != 0)
4089 add_cookie_expiry_date(header, cookie_lifetime);
4090 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to: %s", *header);
4092 else if (expiry_date_status != NO_EXPIRY_DATE_SPECIFIED)
4094 log_error(LOG_LEVEL_HEADER,
4095 "Cookie rewritten to a temporary one: %s", *header);
4104 #ifdef FEATURE_FORCE_LOAD
4105 /*********************************************************************
4107 * Function : strclean
4109 * Description : In-Situ-Eliminate all occurrences of substring in
4113 * 1 : string = string to clean
4114 * 2 : substring = substring to eliminate
4116 * Returns : Number of eliminations
4118 *********************************************************************/
4119 int strclean(char *string, const char *substring)
4125 len = strlen(substring);
4127 while((pos = strstr(string, substring)) != NULL)
4134 while (*p++ != '\0');
4141 #endif /* def FEATURE_FORCE_LOAD */
4144 /*********************************************************************
4146 * Function : parse_header_time
4148 * Description : Parses time formats used in HTTP header strings
4149 * to get the numerical respresentation.
4152 * 1 : header_time = HTTP header time as string.
4153 * 2 : result = storage for header_time in seconds
4155 * Returns : JB_ERR_OK if the time format was recognized, or
4156 * JB_ERR_PARSE otherwise.
4158 *********************************************************************/
4159 static jb_err parse_header_time(const char *header_time, time_t *result)
4163 * Checking for two-digit years first in an
4164 * attempt to work around GNU libc's strptime()
4165 * reporting negative year values when using %Y.
4167 static const char time_formats[][22] = {
4168 /* Tue, 02-Jun-37 20:00:00 */
4169 "%a, %d-%b-%y %H:%M:%S",
4170 /* Tue, 02 Jun 2037 20:00:00 */
4171 "%a, %d %b %Y %H:%M:%S",
4172 /* Tue, 02-Jun-2037 20:00:00 */
4173 "%a, %d-%b-%Y %H:%M:%S",
4174 /* Tuesday, 02-Jun-2037 20:00:00 */
4175 "%A, %d-%b-%Y %H:%M:%S",
4176 /* Tuesday Jun 02 20:00:00 2037 */
4177 "%A %b %d %H:%M:%S %Y"
4181 for (i = 0; i < SZ(time_formats); i++)
4184 * Zero out gmt to prevent time zone offsets.
4185 * Documented to be required for GNU libc.
4187 memset(&gmt, 0, sizeof(gmt));
4189 if (NULL != strptime(header_time, time_formats[i], &gmt))
4191 /* Sanity check for GNU libc. */
4192 if (gmt.tm_year < 0)
4194 log_error(LOG_LEVEL_HEADER,
4195 "Failed to parse '%s' using '%s'. Moving on.",
4196 header_time, time_formats[i]);
4199 *result = timegm(&gmt);
4201 #ifdef FEATURE_STRPTIME_SANITY_CHECKS
4203 * Verify that parsing the date recreated from the first
4204 * parse operation gets the previous result. If it doesn't,
4205 * either strptime() or strftime() are malfunctioning.
4207 * We could string-compare the recreated date with the original
4208 * header date, but this leads to false positives as strptime()
4209 * may let %a accept all day formats while strftime() will only
4213 char recreated_date[100];
4217 tm = gmtime(result);
4218 strftime(recreated_date, sizeof(recreated_date), time_formats[i], tm);
4219 memset(&gmt, 0, sizeof(gmt));
4220 if (NULL == strptime(recreated_date, time_formats[i], &gmt))
4222 log_error(LOG_LEVEL_ERROR,
4223 "Failed to parse '%s' generated with '%s' to recreate '%s'.",
4224 recreated_date, time_formats[i], header_time);
4227 result2 = timegm(&gmt);
4228 if (*result != result2)
4230 log_error(LOG_LEVEL_ERROR, "strftime() and strptime() disagree. "
4231 "Format: '%s'. In: '%s', out: '%s'. %d != %d. Rejecting.",
4232 time_formats[i], header_time, recreated_date, *result, result2);
4242 return JB_ERR_PARSE;
4247 /*********************************************************************
4249 * Function : get_destination_from_headers
4251 * Description : Parse the "Host:" header to get the request's destination.
4252 * Only needed if the client's request was forcefully
4253 * redirected into Privoxy.
4255 * Code mainly copied from client_host() which is currently
4256 * run too late for this purpose.
4259 * 1 : headers = List of headers (one of them hopefully being
4260 * the "Host:" header)
4261 * 2 : http = storage for the result (host, port and hostport).
4263 * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
4264 * JB_ERR_PARSE if the host header couldn't be found,
4265 * JB_ERR_OK otherwise.
4267 *********************************************************************/
4268 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4274 host = get_header_value(headers, "Host:");
4278 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4279 return JB_ERR_PARSE;
4282 p = strdup_or_die(host);
4284 q = strdup_or_die(p);
4286 freez(http->hostport);
4290 q = strchr(http->host, ':');
4293 /* Terminate hostname and evaluate port string */
4295 http->port = atoi(q);
4299 http->port = http->ssl ? 443 : 80;
4302 /* Rebuild request URL */
4304 http->url = strdup(http->ssl ? "https://" : "http://");
4305 string_append(&http->url, http->hostport);
4306 string_append(&http->url, http->path);
4307 if (http->url == NULL)
4309 return JB_ERR_MEMORY;
4312 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4320 /*********************************************************************
4322 * Function : create_forged_referrer
4324 * Description : Helper for client_referrer to forge a referer as
4325 * 'http://hostname[:port]/' to fool stupid
4326 * checks for in-site links
4329 * 1 : header = Pointer to header pointer
4330 * 2 : hostport = Host and optionally port as string
4332 * Returns : JB_ERR_OK in case of success, or
4333 * JB_ERR_MEMORY in case of memory problems.
4335 *********************************************************************/
4336 static jb_err create_forged_referrer(char **header, const char *hostport)
4338 assert(NULL == *header);
4340 *header = strdup("Referer: http://");
4341 string_append(header, hostport);
4342 string_append(header, "/");
4344 if (NULL == *header)
4346 return JB_ERR_MEMORY;
4349 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4356 /*********************************************************************
4358 * Function : create_fake_referrer
4360 * Description : Helper for client_referrer to create a fake referrer
4361 * based on a string supplied by the user.
4364 * 1 : header = Pointer to header pointer
4365 * 2 : hosthost = Referrer to fake
4367 * Returns : JB_ERR_OK in case of success, or
4368 * JB_ERR_MEMORY in case of memory problems.
4370 *********************************************************************/
4371 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4373 assert(NULL == *header);
4375 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4377 log_error(LOG_LEVEL_HEADER,
4378 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4380 *header = strdup("Referer: ");
4381 string_append(header, fake_referrer);
4383 if (NULL == *header)
4385 return JB_ERR_MEMORY;
4388 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4395 /*********************************************************************
4397 * Function : handle_conditional_hide_referrer_parameter
4399 * Description : Helper for client_referrer to crunch or forge
4400 * the referrer header if the host has changed.
4403 * 1 : header = Pointer to header pointer
4404 * 2 : host = The target host (may include the port)
4405 * 3 : parameter_conditional_block = Boolean to signal
4406 * if we're in conditional-block mode. If not set,
4407 * we're in conditional-forge mode.
4409 * Returns : JB_ERR_OK in case of success, or
4410 * JB_ERR_MEMORY in case of memory problems.
4412 *********************************************************************/
4413 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4414 const char *host, const int parameter_conditional_block)
4416 char *referer = strdup_or_die(*header);
4417 const size_t hostlength = strlen(host);
4418 const char *referer_url = NULL;
4420 /* referer begins with 'Referer: http[s]://' */
4421 if ((hostlength+17) < strlen(referer))
4424 * Shorten referer to make sure the referer is blocked
4425 * if www.example.org/www.example.com-shall-see-the-referer/
4426 * links to www.example.com/
4428 referer[hostlength+17] = '\0';
4430 referer_url = strstr(referer, "http://");
4431 if ((NULL == referer_url) || (NULL == strstr(referer_url, host)))
4433 /* Host has changed, Referer is invalid or a https URL. */
4434 if (parameter_conditional_block)
4436 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4443 return create_forged_referrer(header, host);
4453 /*********************************************************************
4455 * Function : create_content_length_header
4457 * Description : Creates a Content-Length header.
4460 * 1 : content_length = The content length to be used in the header.
4461 * 2 : header = Allocated space to safe the header.
4462 * 3 : buffer_length = The length of the allocated space.
4466 *********************************************************************/
4467 static void create_content_length_header(unsigned long long content_length,
4468 char *header, size_t buffer_length)
4470 snprintf(header, buffer_length, "Content-Length: %llu", content_length);
4474 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
4475 /*********************************************************************
4477 * Function : get_expected_content_length
4479 * Description : Figures out the content length from a list of headers.
4482 * 1 : headers = List of headers
4484 * Returns : Number of bytes to expect
4486 *********************************************************************/
4487 unsigned long long get_expected_content_length(struct list *headers)
4489 const char *content_length_header;
4490 unsigned long long content_length = 0;
4492 content_length_header = get_header_value(headers, "Content-Length:");
4493 if (content_length_header != NULL)
4495 if (JB_ERR_OK != get_content_length(content_length_header, &content_length))
4497 log_error(LOG_LEVEL_ERROR,
4498 "Failed to get the Content-Length in %s", content_length_header);
4499 /* XXX: The header will be removed later on */
4504 return content_length;