1 const char parsers_rcs[] = "$Id: parsers.c,v 1.164 2009/05/25 15:42:40 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `add_to_iob', `client_cookie_adder', `client_from',
9 * `client_referrer', `client_send_cookie', `client_ua',
10 * `client_uagent', `client_x_forwarded',
11 * `client_x_forwarded_adder', `client_xtra_adder',
12 * `content_type', `crumble', `destroy_list', `enlist',
13 * `flush_socket', ``get_header', `sed', `filter_header'
14 * `server_content_encoding', `server_content_disposition',
15 * `server_last_modified', `client_accept_language',
16 * `crunch_client_header', `client_if_modified_since',
17 * `client_if_none_match', `get_destination_from_headers',
18 * `parse_header_time', `decompress_iob' and `server_set_cookie'.
20 * Copyright : Written by and Copyright (C) 2001-2009 the
21 * Privoxy team. http://www.privoxy.org/
23 * Based on the Internet Junkbuster originally written
24 * by and Copyright (C) 1997 Anonymous Coders and
25 * Junkbusters Corporation. http://www.junkbusters.com
27 * This program is free software; you can redistribute it
28 * and/or modify it under the terms of the GNU General
29 * Public License as published by the Free Software
30 * Foundation; either version 2 of the License, or (at
31 * your option) any later version.
33 * This program is distributed in the hope that it will
34 * be useful, but WITHOUT ANY WARRANTY; without even the
35 * implied warranty of MERCHANTABILITY or FITNESS FOR A
36 * PARTICULAR PURPOSE. See the GNU General Public
37 * License for more details.
39 * The GNU General Public License should be included with
40 * this file. If not, you can view it at
41 * http://www.gnu.org/copyleft/gpl.html
42 * or write to the Free Software Foundation, Inc., 59
43 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
45 *********************************************************************/
52 #include <sys/types.h>
62 * Convince GNU's libc to provide a strptime prototype.
72 #if !defined(_WIN32) && !defined(__OS2__)
78 #ifdef FEATURE_PTHREAD
80 /* jcc.h is for mutex semapores only */
81 #endif /* def FEATURE_PTHREAD */
86 #include "jbsockets.h"
96 const char parsers_h_rcs[] = PARSERS_H_VERSION;
98 /* Fix a problem with Solaris. There should be no effect on other
100 * Solaris's isspace() is a macro which uses its argument directly
101 * as an array index. Therefore we need to make sure that high-bit
102 * characters generate +ve values, and ideally we also want to make
103 * the argument match the declared parameter type of "int".
105 * Why did they write a character function that can't take a simple
106 * "char" argument? Doh!
108 #define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
109 #define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
111 static char *get_header_line(struct iob *iob);
112 static jb_err scan_headers(struct client_state *csp);
113 static jb_err header_tagger(struct client_state *csp, char *header);
114 static jb_err parse_header_time(const char *header_time, time_t *result);
116 static jb_err crumble (struct client_state *csp, char **header);
117 static jb_err filter_header (struct client_state *csp, char **header);
118 static jb_err client_connection (struct client_state *csp, char **header);
119 static jb_err client_referrer (struct client_state *csp, char **header);
120 static jb_err client_uagent (struct client_state *csp, char **header);
121 static jb_err client_ua (struct client_state *csp, char **header);
122 static jb_err client_from (struct client_state *csp, char **header);
123 static jb_err client_send_cookie (struct client_state *csp, char **header);
124 static jb_err client_x_forwarded (struct client_state *csp, char **header);
125 static jb_err client_accept_encoding (struct client_state *csp, char **header);
126 static jb_err client_te (struct client_state *csp, char **header);
127 static jb_err client_max_forwards (struct client_state *csp, char **header);
128 static jb_err client_host (struct client_state *csp, char **header);
129 static jb_err client_if_modified_since (struct client_state *csp, char **header);
130 static jb_err client_accept_language (struct client_state *csp, char **header);
131 static jb_err client_if_none_match (struct client_state *csp, char **header);
132 static jb_err crunch_client_header (struct client_state *csp, char **header);
133 static jb_err client_x_filter (struct client_state *csp, char **header);
134 static jb_err client_range (struct client_state *csp, char **header);
135 static jb_err server_set_cookie (struct client_state *csp, char **header);
136 static jb_err server_connection (struct client_state *csp, char **header);
137 static jb_err server_content_type (struct client_state *csp, char **header);
138 static jb_err server_adjust_content_length(struct client_state *csp, char **header);
139 static jb_err server_content_md5 (struct client_state *csp, char **header);
140 static jb_err server_content_encoding (struct client_state *csp, char **header);
141 static jb_err server_transfer_coding (struct client_state *csp, char **header);
142 static jb_err server_http (struct client_state *csp, char **header);
143 static jb_err crunch_server_header (struct client_state *csp, char **header);
144 static jb_err server_last_modified (struct client_state *csp, char **header);
145 static jb_err server_content_disposition(struct client_state *csp, char **header);
147 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
148 static jb_err server_save_content_length(struct client_state *csp, char **header);
149 static jb_err server_keep_alive(struct client_state *csp, char **header);
150 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
152 static jb_err client_host_adder (struct client_state *csp);
153 static jb_err client_xtra_adder (struct client_state *csp);
154 static jb_err client_x_forwarded_for_adder(struct client_state *csp);
155 static jb_err client_connection_header_adder(struct client_state *csp);
156 static jb_err server_connection_adder(struct client_state *csp);
157 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
158 static jb_err server_proxy_connection_adder(struct client_state *csp);
159 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
161 static jb_err create_forged_referrer(char **header, const char *hostport);
162 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
163 static jb_err handle_conditional_hide_referrer_parameter(char **header,
164 const char *host, const int parameter_conditional_block);
165 static const char *get_appropiate_connection_header(const struct client_state *csp);
168 * List of functions to run on a list of headers.
172 /** The header prefix to match */
175 /** The length of the prefix to match */
178 /** The function to apply to this line */
179 const parser_func_ptr parser;
182 static const struct parsers client_patterns[] = {
183 { "referer:", 8, client_referrer },
184 { "user-agent:", 11, client_uagent },
185 { "ua-", 3, client_ua },
186 { "from:", 5, client_from },
187 { "cookie:", 7, client_send_cookie },
188 { "x-forwarded-for:", 16, client_x_forwarded },
189 { "Accept-Encoding:", 16, client_accept_encoding },
190 { "TE:", 3, client_te },
191 { "Host:", 5, client_host },
192 { "if-modified-since:", 18, client_if_modified_since },
193 #ifndef FEATURE_CONNECTION_KEEP_ALIVE
194 { "Keep-Alive:", 11, crumble },
196 { "connection:", 11, client_connection },
197 { "proxy-connection:", 17, crumble },
198 { "max-forwards:", 13, client_max_forwards },
199 { "Accept-Language:", 16, client_accept_language },
200 { "if-none-match:", 14, client_if_none_match },
201 { "Range:", 6, client_range },
202 { "Request-Range:", 14, client_range },
203 { "If-Range:", 9, client_range },
204 { "X-Filter:", 9, client_x_filter },
205 { "*", 0, crunch_client_header },
206 { "*", 0, filter_header },
210 static const struct parsers server_patterns[] = {
211 { "HTTP/", 5, server_http },
212 { "set-cookie:", 11, server_set_cookie },
213 { "connection:", 11, server_connection },
214 { "Content-Type:", 13, server_content_type },
215 { "Content-MD5:", 12, server_content_md5 },
216 { "Content-Encoding:", 17, server_content_encoding },
217 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
218 { "Content-Length:", 15, server_save_content_length },
219 { "Keep-Alive:", 11, server_keep_alive },
221 { "Keep-Alive:", 11, crumble },
222 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
223 { "Transfer-Encoding:", 18, server_transfer_coding },
224 { "content-disposition:", 20, server_content_disposition },
225 { "Last-Modified:", 14, server_last_modified },
226 { "*", 0, crunch_server_header },
227 { "*", 0, filter_header },
231 static const add_header_func_ptr add_client_headers[] = {
233 client_x_forwarded_for_adder,
235 /* Temporarily disabled: client_accept_encoding_adder, */
236 client_connection_header_adder,
240 static const add_header_func_ptr add_server_headers[] = {
241 server_connection_adder,
242 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
243 server_proxy_connection_adder,
244 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
248 /*********************************************************************
250 * Function : flush_socket
252 * Description : Write any pending "buffered" content.
255 * 1 : fd = file descriptor of the socket to read
256 * 2 : iob = The I/O buffer to flush, usually csp->iob.
258 * Returns : On success, the number of bytes written are returned (zero
259 * indicates nothing was written). On error, -1 is returned,
260 * and errno is set appropriately. If count is zero and the
261 * file descriptor refers to a regular file, 0 will be
262 * returned without causing any other effect. For a special
263 * file, the results are not portable.
265 *********************************************************************/
266 long flush_socket(jb_socket fd, struct iob *iob)
268 long len = iob->eod - iob->cur;
275 if (write_socket(fd, iob->cur, (size_t)len))
279 iob->eod = iob->cur = iob->buf;
285 /*********************************************************************
287 * Function : add_to_iob
289 * Description : Add content to the buffered page, expanding the
290 * buffer if necessary.
293 * 1 : csp = Current client state (buffers, headers, etc...)
294 * 2 : buf = holds the content to be added to the page
295 * 3 : n = number of bytes to be added
297 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
298 * or buffer limit reached.
300 *********************************************************************/
301 jb_err add_to_iob(struct client_state *csp, char *buf, long n)
303 struct iob *iob = csp->iob;
304 size_t used, offset, need, want;
307 if (n <= 0) return JB_ERR_OK;
309 used = (size_t)(iob->eod - iob->buf);
310 offset = (size_t)(iob->cur - iob->buf);
311 need = used + (size_t)n + 1;
314 * If the buffer can't hold the new data, extend it first.
315 * Use the next power of two if possible, else use the actual need.
317 if (need > csp->config->buffer_limit)
319 log_error(LOG_LEVEL_INFO,
320 "Buffer limit reached while extending the buffer (iob). Needed: %d. Limit: %d",
321 need, csp->config->buffer_limit);
322 return JB_ERR_MEMORY;
325 if (need > iob->size)
327 for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2;
329 if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
333 else if (NULL != (p = (char *)realloc(iob->buf, need)))
339 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
340 return JB_ERR_MEMORY;
343 /* Update the iob pointers */
344 iob->cur = p + offset;
349 /* copy the new data into the iob buffer */
350 memcpy(iob->eod, buf, (size_t)n);
352 /* point to the end of the data */
355 /* null terminate == cheap insurance */
364 /*********************************************************************
366 * Function : decompress_iob
368 * Description : Decompress buffered page, expanding the
369 * buffer as necessary. csp->iob->cur
370 * should point to the the beginning of the
371 * compressed data block.
374 * 1 : csp = Current client state (buffers, headers, etc...)
376 * Returns : JB_ERR_OK on success,
377 * JB_ERR_MEMORY if out-of-memory limit reached, and
378 * JB_ERR_COMPRESS if error decompressing buffer.
380 *********************************************************************/
381 jb_err decompress_iob(struct client_state *csp)
383 char *buf; /* new, uncompressed buffer */
384 char *cur; /* Current iob position (to keep the original
385 * iob->cur unmodified if we return early) */
386 size_t bufsize; /* allocated size of the new buffer */
387 size_t old_size; /* Content size before decompression */
388 size_t skip_size; /* Number of bytes at the beginning of the iob
389 that we should NOT decompress. */
390 int status; /* return status of the inflate() call */
391 z_stream zstr; /* used by calls to zlib */
393 assert(csp->iob->cur - csp->iob->buf > 0);
394 assert(csp->iob->eod - csp->iob->cur > 0);
396 bufsize = csp->iob->size;
397 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
398 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
402 if (bufsize < (size_t)10)
405 * This is to protect the parsing of gzipped data,
406 * but it should(?) be valid for deflated data also.
408 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
409 return JB_ERR_COMPRESS;
412 if (csp->content_type & CT_GZIP)
415 * Our task is slightly complicated by the facts that data
416 * compressed by gzip does not include a zlib header, and
417 * that there is no easily accessible interface in zlib to
418 * handle a gzip header. We strip off the gzip header by
419 * hand, and later inform zlib not to expect a header.
423 * Strip off the gzip header. Please see RFC 1952 for more
424 * explanation of the appropriate fields.
426 if ((*cur++ != (char)0x1f)
427 || (*cur++ != (char)0x8b)
428 || (*cur++ != Z_DEFLATED))
430 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
431 return JB_ERR_COMPRESS;
437 * XXX: These magic numbers should be replaced
438 * with macros to give a better idea what they do.
442 /* The gzip header has reserved bits set; bail out. */
443 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
444 return JB_ERR_COMPRESS;
448 /* Skip extra fields if necessary. */
452 * Skip a given number of bytes, specified
453 * as a 16-bit little-endian value.
456 * XXX: This code used to be:
458 * csp->iob->cur += *csp->iob->cur++ + (*csp->iob->cur++ << 8);
460 * which I had to change into:
462 * cur += *cur++ + (*cur++ << 8);
464 * at which point gcc43 finally noticed that the value
465 * of cur is undefined (it depends on which of the
466 * summands is evaluated first).
468 * I haven't come across a site where this
469 * code is actually executed yet, but I hope
474 skip_bytes += *cur++ << 8;
476 assert(skip_bytes == *csp->iob->cur - 2 + ((*csp->iob->cur - 1) << 8));
479 * The number of bytes to skip should be positive
480 * and we'd like to stay in the buffer.
482 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
484 log_error(LOG_LEVEL_ERROR,
485 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
487 return JB_ERR_COMPRESS;
489 log_error(LOG_LEVEL_INFO,
490 "Skipping %d bytes for gzip compression. Does this sound right?",
495 /* Skip the filename if necessary. */
498 /* A null-terminated string is supposed to follow. */
499 while (*cur++ && (cur < csp->iob->eod));
503 /* Skip the comment if necessary. */
506 /* A null-terminated string is supposed to follow. */
507 while (*cur++ && (cur < csp->iob->eod));
510 /* Skip the CRC if necessary. */
516 if (cur >= csp->iob->eod)
519 * If the current position pointer reached or passed
520 * the buffer end, we were obviously tricked to skip
523 log_error(LOG_LEVEL_ERROR,
524 "Malformed gzip header detected. Aborting decompression.");
525 return JB_ERR_COMPRESS;
529 else if (csp->content_type & CT_DEFLATE)
532 * XXX: The debug level should be lowered
533 * before the next stable release.
535 log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
537 * In theory (that is, according to RFC 1950), deflate-compressed
538 * data should begin with a two-byte zlib header and have an
539 * adler32 checksum at the end. It seems that in practice only
540 * the raw compressed data is sent. Note that this means that
541 * we are not RFC 1950-compliant here, but the advantage is that
542 * this actually works. :)
544 * We add a dummy null byte to tell zlib where the data ends,
545 * and later inform it not to expect a header.
547 * Fortunately, add_to_iob() has thoughtfully null-terminated
548 * the buffer; we can just increment the end pointer to include
555 log_error(LOG_LEVEL_ERROR,
556 "Unable to determine compression format for decompression");
557 return JB_ERR_COMPRESS;
560 /* Set up the fields required by zlib. */
561 zstr.next_in = (Bytef *)cur;
562 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
563 zstr.zalloc = Z_NULL;
565 zstr.opaque = Z_NULL;
568 * Passing -MAX_WBITS to inflateInit2 tells the library
569 * that there is no zlib header.
571 if (inflateInit2 (&zstr, -MAX_WBITS) != Z_OK)
573 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
574 return JB_ERR_COMPRESS;
578 * Next, we allocate new storage for the inflated data.
579 * We don't modify the existing iob yet, so in case there
580 * is error in decompression we can recover gracefully.
582 buf = zalloc(bufsize);
585 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
586 return JB_ERR_MEMORY;
589 assert(bufsize >= skip_size);
590 memcpy(buf, csp->iob->buf, skip_size);
591 zstr.avail_out = bufsize - skip_size;
592 zstr.next_out = (Bytef *)buf + skip_size;
594 /* Try to decompress the whole stream in one shot. */
595 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
597 /* We need to allocate more memory for the output buffer. */
599 char *tmpbuf; /* used for realloc'ing the buffer */
600 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
603 * If zlib wants more data then there's a problem, because
604 * the complete compressed file should have been buffered.
606 if (0 == zstr.avail_in)
608 log_error(LOG_LEVEL_ERROR, "Unexpected end of compressed iob");
609 return JB_ERR_COMPRESS;
613 * If we tried the limit and still didn't have enough
614 * memory, just give up.
616 if (bufsize == csp->config->buffer_limit)
618 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
619 return JB_ERR_MEMORY;
622 /* Try doubling the buffer size each time. */
625 /* Don't exceed the buffer limit. */
626 if (bufsize > csp->config->buffer_limit)
628 bufsize = csp->config->buffer_limit;
631 /* Try to allocate the new buffer. */
632 tmpbuf = realloc(buf, bufsize);
635 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
637 return JB_ERR_MEMORY;
641 char *oldnext_out = (char *)zstr.next_out;
644 * Update the fields for inflate() to use the new
645 * buffer, which may be in a location different from
648 zstr.avail_out += bufsize - oldbufsize;
649 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
652 * Compare with an uglier method of calculating these values
653 * that doesn't require the extra oldbufsize variable.
655 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
656 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
657 assert(zstr.avail_out > 0U);
663 if (Z_STREAM_ERROR == inflateEnd(&zstr))
665 log_error(LOG_LEVEL_ERROR,
666 "Inconsistent stream state after decompression: %s", zstr.msg);
668 * XXX: Intentionally no return.
670 * According to zlib.h, Z_STREAM_ERROR is returned
671 * "if the stream state was inconsistent".
673 * I assume in this case inflate()'s status
674 * would also be something different than Z_STREAM_END
675 * so this check should be redundant, but lets see.
679 if (status != Z_STREAM_END)
681 /* We failed to decompress the stream. */
682 log_error(LOG_LEVEL_ERROR,
683 "Error in decompressing to the buffer (iob): %s", zstr.msg);
684 return JB_ERR_COMPRESS;
688 * Finally, we can actually update the iob, since the
689 * decompression was successful. First, free the old
692 freez(csp->iob->buf);
694 /* Now, update the iob to use the new buffer. */
696 csp->iob->cur = csp->iob->buf + skip_size;
697 csp->iob->eod = (char *)zstr.next_out;
698 csp->iob->size = bufsize;
701 * Make sure the new uncompressed iob obeys some minimal
702 * consistency conditions.
704 if ((csp->iob->buf < csp->iob->cur)
705 && (csp->iob->cur <= csp->iob->eod)
706 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
708 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
709 if (new_size > (size_t)0)
711 log_error(LOG_LEVEL_RE_FILTER,
712 "Decompression successful. Old size: %d, new size: %d.",
717 /* zlib thinks this is OK, so lets do the same. */
718 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
723 /* It seems that zlib did something weird. */
724 log_error(LOG_LEVEL_ERROR,
725 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
726 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
727 csp->iob->eod, csp->iob->buf + csp->iob->size);
728 return JB_ERR_COMPRESS;
734 #endif /* defined(FEATURE_ZLIB) */
737 /*********************************************************************
739 * Function : string_move
741 * Description : memmove wrapper to move the last part of a string
742 * towards the beginning, overwriting the part in
743 * the middle. strlcpy() can't be used here as the
747 * 1 : dst = Destination to overwrite
748 * 2 : src = Source to move.
752 *********************************************************************/
753 static void string_move(char *dst, char *src)
757 /* +1 to copy the terminating nul as well. */
758 memmove(dst, src, strlen(src)+1);
762 /*********************************************************************
764 * Function : normalize_lws
766 * Description : Reduces unquoted linear white space in headers
767 * to a single space in accordance with RFC 2616 2.2.
768 * This simplifies parsing and filtering later on.
770 * XXX: Remove log messages before
771 * the next stable release?
774 * 1 : header = A header with linear white space to reduce.
778 *********************************************************************/
779 static void normalize_lws(char *header)
785 if (ijb_isspace(*p) && ijb_isspace(*(p+1)))
789 while (ijb_isspace(*q))
793 log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
799 log_error(LOG_LEVEL_HEADER,
800 "Converting tab to space in '%s'", header);
805 char *end_of_token = strstr(p+1, "\"");
807 if (NULL != end_of_token)
809 /* Don't mess with quoted text. */
814 log_error(LOG_LEVEL_HEADER,
815 "Ignoring single quote in '%s'", header);
821 p = strchr(header, ':');
822 if ((p != NULL) && (p != header) && ijb_isspace(*(p-1)))
825 * There's still space before the colon.
833 /*********************************************************************
835 * Function : get_header
837 * Description : This (odd) routine will parse the csp->iob
838 * to get the next complete header.
841 * 1 : iob = The I/O buffer to parse, usually csp->iob.
843 * Returns : Any one of the following:
845 * 1) a pointer to a dynamically allocated string that contains a header line
846 * 2) NULL indicating that the end of the header was reached
847 * 3) "" indicating that the end of the iob was reached before finding
848 * a complete header line.
850 *********************************************************************/
851 char *get_header(struct iob *iob)
855 header = get_header_line(iob);
857 if ((header == NULL) || (*header == '\0'))
860 * No complete header read yet, tell the client.
865 while ((iob->cur[0] == ' ') || (iob->cur[0] == '\t'))
868 * Header spans multiple lines, append the next one.
870 char *continued_header;
872 continued_header = get_header_line(iob);
873 if ((continued_header == NULL) || (*continued_header == '\0'))
876 * No complete header read yet, return what we got.
877 * XXX: Should "unread" header instead.
879 log_error(LOG_LEVEL_INFO,
880 "Failed to read a multi-line header properly: '%s'",
885 if (JB_ERR_OK != string_join(&header, continued_header))
887 log_error(LOG_LEVEL_FATAL,
888 "Out of memory while appending multiple headers.");
892 /* XXX: remove before next stable release. */
893 log_error(LOG_LEVEL_HEADER,
894 "Merged multiple header lines to: '%s'",
899 normalize_lws(header);
906 /*********************************************************************
908 * Function : get_header_line
910 * Description : This (odd) routine will parse the csp->iob
911 * to get the next header line.
914 * 1 : iob = The I/O buffer to parse, usually csp->iob.
916 * Returns : Any one of the following:
918 * 1) a pointer to a dynamically allocated string that contains a header line
919 * 2) NULL indicating that the end of the header was reached
920 * 3) "" indicating that the end of the iob was reached before finding
921 * a complete header line.
923 *********************************************************************/
924 static char *get_header_line(struct iob *iob)
928 if ((iob->cur == NULL)
929 || ((p = strchr(iob->cur, '\n')) == NULL))
931 return(""); /* couldn't find a complete header */
936 ret = strdup(iob->cur);
939 /* FIXME No way to handle error properly */
940 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header_line()");
946 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
948 /* is this a blank line (i.e. the end of the header) ? */
960 /*********************************************************************
962 * Function : get_header_value
964 * Description : Get the value of a given header from a chained list
965 * of header lines or return NULL if no such header is
966 * present in the list.
969 * 1 : header_list = pointer to list
970 * 2 : header_name = string with name of header to look for.
971 * Trailing colon required, capitalization
974 * Returns : NULL if not found, else value of header
976 *********************************************************************/
977 char *get_header_value(const struct list *header_list, const char *header_name)
979 struct list_entry *cur_entry;
985 length = strlen(header_name);
987 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
991 if (!strncmpic(cur_entry->str, header_name, length))
994 * Found: return pointer to start of value
996 ret = cur_entry->str + length;
997 while (*ret && ijb_isspace(*ret)) ret++;
1011 /*********************************************************************
1013 * Function : scan_headers
1015 * Description : Scans headers, applies tags and updates action bits.
1018 * 1 : csp = Current client state (buffers, headers, etc...)
1020 * Returns : JB_ERR_OK
1022 *********************************************************************/
1023 static jb_err scan_headers(struct client_state *csp)
1025 struct list_entry *h; /* Header */
1026 jb_err err = JB_ERR_OK;
1028 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1030 /* Header crunch()ed in previous run? -> ignore */
1031 if (h->str == NULL) continue;
1032 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1033 err = header_tagger(csp, h->str);
1040 /*********************************************************************
1044 * Description : add, delete or modify lines in the HTTP header streams.
1045 * On entry, it receives a linked list of headers space
1046 * that was allocated dynamically (both the list nodes
1047 * and the header contents).
1049 * As a side effect it frees the space used by the original
1053 * 1 : csp = Current client state (buffers, headers, etc...)
1054 * 2 : filter_server_headers = Boolean to switch between
1055 * server and header filtering.
1057 * Returns : JB_ERR_OK in case off success, or
1058 * JB_ERR_MEMORY on out-of-memory error.
1060 *********************************************************************/
1061 jb_err sed(struct client_state *csp, int filter_server_headers)
1063 /* XXX: use more descriptive names. */
1064 struct list_entry *p;
1065 const struct parsers *v;
1066 const add_header_func_ptr *f;
1067 jb_err err = JB_ERR_OK;
1069 if (filter_server_headers)
1071 v = server_patterns;
1072 f = add_server_headers;
1076 v = client_patterns;
1077 f = add_client_headers;
1082 while ((err == JB_ERR_OK) && (v->str != NULL))
1084 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1086 /* Header crunch()ed in previous run? -> ignore */
1087 if (p->str == NULL) continue;
1089 /* Does the current parser handle this header? */
1090 if ((strncmpic(p->str, v->str, v->len) == 0) ||
1091 (v->len == CHECK_EVERY_HEADER_REMAINING))
1093 err = v->parser(csp, &(p->str));
1099 /* place additional headers on the csp->headers list */
1100 while ((err == JB_ERR_OK) && (*f))
1110 /*********************************************************************
1112 * Function : update_server_headers
1114 * Description : Updates server headers after the body has been modified.
1117 * 1 : csp = Current client state (buffers, headers, etc...)
1119 * Returns : JB_ERR_OK in case off success, or
1120 * JB_ERR_MEMORY on out-of-memory error.
1122 *********************************************************************/
1123 jb_err update_server_headers(struct client_state *csp)
1125 jb_err err = JB_ERR_OK;
1127 static const struct parsers server_patterns_light[] = {
1128 { "Content-Length:", 15, server_adjust_content_length },
1129 { "Transfer-Encoding:", 18, server_transfer_coding },
1131 { "Content-Encoding:", 17, server_content_encoding },
1132 #endif /* def FEATURE_ZLIB */
1136 if (strncmpic(csp->http->cmd, "HEAD", 4))
1138 const struct parsers *v;
1139 struct list_entry *p;
1141 for (v = server_patterns_light; (err == JB_ERR_OK) && (v->str != NULL); v++)
1143 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
1145 /* Header crunch()ed in previous run? -> ignore */
1146 if (p->str == NULL) continue;
1148 /* Does the current parser handle this header? */
1149 if (strncmpic(p->str, v->str, v->len) == 0)
1151 err = v->parser(csp, (char **)&(p->str));
1161 /*********************************************************************
1163 * Function : header_tagger
1165 * Description : Executes all text substitutions from applying
1166 * tag actions and saves the result as tag.
1168 * XXX: Shares enough code with filter_header() and
1169 * pcrs_filter_response() to warrant some helper functions.
1172 * 1 : csp = Current client state (buffers, headers, etc...)
1173 * 2 : header = Header that is used as tagger input
1175 * Returns : JB_ERR_OK on success and always succeeds
1177 *********************************************************************/
1178 static jb_err header_tagger(struct client_state *csp, char *header)
1180 int wanted_filter_type;
1181 int multi_action_index;
1185 struct file_list *fl;
1186 struct re_filterfile_spec *b;
1187 struct list_entry *tag_name;
1189 int found_filters = 0;
1190 const size_t header_length = strlen(header);
1192 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1194 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1195 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1199 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1200 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1203 /* Check if there are any filters */
1204 for (i = 0; i < MAX_AF_FILES; i++)
1217 if (0 == found_filters)
1219 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1220 "tagging enabled, but no taggers available.");
1224 for (i = 0; i < MAX_AF_FILES; i++)
1227 if ((NULL == fl) || (NULL == fl->f))
1230 * Either there are no filter files
1231 * left, or this filter file just
1232 * contains no valid filters.
1234 * Continue to be sure we don't miss
1235 * valid filter files that are chained
1236 * after empty or invalid ones.
1241 /* For all filters, */
1242 for (b = fl->f; b; b = b->next)
1244 if (b->type != wanted_filter_type)
1246 /* skip the ones we don't care about, */
1249 /* leaving only taggers that could apply, of which we use the ones, */
1250 for (tag_name = csp->action->multi[multi_action_index]->first;
1251 NULL != tag_name; tag_name = tag_name->next)
1253 /* that do apply, and */
1254 if (strcmp(b->name, tag_name->str) == 0)
1256 char *modified_tag = NULL;
1258 size_t size = header_length;
1259 pcrs_job *joblist = b->joblist;
1261 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1263 if (NULL == joblist)
1265 log_error(LOG_LEVEL_RE_FILTER,
1266 "Tagger %s has empty joblist. Nothing to do.", b->name);
1270 /* execute their pcrs_joblist on the header. */
1271 for (job = joblist; NULL != job; job = job->next)
1273 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1277 /* Success, continue with the modified version. */
1286 /* Tagger doesn't match */
1289 /* Regex failure, log it but continue anyway. */
1290 assert(NULL != header);
1291 log_error(LOG_LEVEL_ERROR,
1292 "Problems with tagger \'%s\' and header \'%s\': %s",
1293 b->name, *header, pcrs_strerror(hits));
1295 freez(modified_tag);
1299 if (b->dynamic) pcrs_free_joblist(joblist);
1301 /* If this tagger matched */
1307 * There is to technical limitation which makes
1308 * it impossible to use empty tags, but I assume
1309 * no one would do it intentionally.
1312 log_error(LOG_LEVEL_INFO,
1313 "Tagger \'%s\' created an empty tag. Ignored.",
1318 if (!list_contains_item(csp->tags, tag))
1320 if (JB_ERR_OK != enlist(csp->tags, tag))
1322 log_error(LOG_LEVEL_ERROR,
1323 "Insufficient memory to add tag \'%s\', "
1324 "based on tagger \'%s\' and header \'%s\'",
1325 tag, b->name, *header);
1329 char *action_message;
1331 * update the action bits right away, to make
1332 * tagging based on tags set by earlier taggers
1333 * of the same kind possible.
1335 if (update_action_bits_for_tag(csp, tag))
1337 action_message = "Action bits updated accordingly.";
1341 action_message = "No action bits update necessary.";
1344 log_error(LOG_LEVEL_HEADER,
1345 "Tagger \'%s\' added tag \'%s\'. %s",
1346 b->name, tag, action_message);
1351 /* XXX: Is this log-worthy? */
1352 log_error(LOG_LEVEL_HEADER,
1353 "Tagger \'%s\' didn't add tag \'%s\'. "
1354 "Tag already present", b->name, tag);
1357 } /* if the tagger matched */
1358 } /* if the tagger applies */
1359 } /* for every tagger that could apply */
1360 } /* for all filters */
1361 } /* for all filter files */
1366 /* here begins the family of parser functions that reformat header lines */
1368 /*********************************************************************
1370 * Function : filter_header
1372 * Description : Executes all text substitutions from all applying
1373 * +(server|client)-header-filter actions on the header.
1374 * Most of the code was copied from pcrs_filter_response,
1375 * including the rather short variable names
1378 * 1 : csp = Current client state (buffers, headers, etc...)
1379 * 2 : header = On input, pointer to header to modify.
1380 * On output, pointer to the modified header, or NULL
1381 * to remove the header. This function frees the
1382 * original string if necessary.
1384 * Returns : JB_ERR_OK on success and always succeeds
1386 *********************************************************************/
1387 static jb_err filter_header(struct client_state *csp, char **header)
1391 size_t size = strlen(*header);
1393 char *newheader = NULL;
1396 struct file_list *fl;
1397 struct re_filterfile_spec *b;
1398 struct list_entry *filtername;
1400 int i, found_filters = 0;
1401 int wanted_filter_type;
1402 int multi_action_index;
1404 if (csp->flags & CSP_FLAG_NO_FILTERING)
1409 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1411 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1412 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1416 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1417 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1421 * Need to check the set of re_filterfiles...
1423 for (i = 0; i < MAX_AF_FILES; i++)
1436 if (0 == found_filters)
1438 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1439 "header filtering enabled, but no matching filters available.");
1443 for (i = 0; i < MAX_AF_FILES; i++)
1446 if ((NULL == fl) || (NULL == fl->f))
1449 * Either there are no filter files
1450 * left, or this filter file just
1451 * contains no valid filters.
1453 * Continue to be sure we don't miss
1454 * valid filter files that are chained
1455 * after empty or invalid ones.
1460 * For all applying +filter actions, look if a filter by that
1461 * name exists and if yes, execute its pcrs_joblist on the
1464 for (b = fl->f; b; b = b->next)
1466 if (b->type != wanted_filter_type)
1468 /* Skip other filter types */
1472 for (filtername = csp->action->multi[multi_action_index]->first;
1473 filtername ; filtername = filtername->next)
1475 if (strcmp(b->name, filtername->str) == 0)
1477 int current_hits = 0;
1478 pcrs_job *joblist = b->joblist;
1480 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1482 if (NULL == joblist)
1484 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1488 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1489 *header, size, b->name);
1491 /* Apply all jobs from the joblist */
1492 for (job = joblist; NULL != job; job = job->next)
1494 matches = pcrs_execute(job, *header, size, &newheader, &size);
1497 current_hits += matches;
1498 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1500 *header = newheader;
1502 else if ( 0 == matches )
1504 /* Filter doesn't change header */
1510 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1511 *header, b->name, pcrs_strerror(matches));
1512 if (newheader != NULL)
1514 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1520 if (b->dynamic) pcrs_free_joblist(joblist);
1522 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1523 hits += current_hits;
1530 * Additionally checking for hits is important because if
1531 * the continue hack is triggered, server headers can
1532 * arrive empty to separate multiple heads from each other.
1534 if ((0 == size) && hits)
1536 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
1544 /*********************************************************************
1546 * Function : server_connection
1548 * Description : Makes sure a proper "Connection:" header is
1549 * set and signals connection_header_adder to
1553 * 1 : csp = Current client state (buffers, headers, etc...)
1554 * 2 : header = On input, pointer to header to modify.
1555 * On output, pointer to the modified header, or NULL
1556 * to remove the header. This function frees the
1557 * original string if necessary.
1559 * Returns : JB_ERR_OK on success, or
1560 * JB_ERR_MEMORY on out-of-memory error.
1562 *********************************************************************/
1563 static jb_err server_connection(struct client_state *csp, char **header)
1565 if (!strcmpic(*header, "Connection: keep-alive"))
1567 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1568 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
1570 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
1573 if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE))
1575 log_error(LOG_LEVEL_HEADER,
1576 "Keeping the server header '%s' around.", *header);
1579 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
1581 char *old_header = *header;
1583 *header = strdup("Connection: close");
1586 return JB_ERR_MEMORY;
1588 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
1593 /* Signal server_connection_adder() to return early. */
1594 csp->flags |= CSP_FLAG_SERVER_CONNECTION_HEADER_SET;
1600 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1601 /*********************************************************************
1603 * Function : server_keep_alive
1605 * Description : Stores the servers keep alive timeout.
1608 * 1 : csp = Current client state (buffers, headers, etc...)
1609 * 2 : header = On input, pointer to header to modify.
1610 * On output, pointer to the modified header, or NULL
1611 * to remove the header. This function frees the
1612 * original string if necessary.
1614 * Returns : JB_ERR_OK.
1616 *********************************************************************/
1617 static jb_err server_keep_alive(struct client_state *csp, char **header)
1619 unsigned int keep_alive_timeout;
1620 const char *timeout_position = strstr(*header, "timeout=");
1622 if ((NULL == timeout_position)
1623 || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
1625 log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
1629 if (keep_alive_timeout < csp->server_connection.keep_alive_timeout)
1631 log_error(LOG_LEVEL_HEADER,
1632 "Reducing keep-alive timeout from %u to %u.",
1633 csp->server_connection.keep_alive_timeout, keep_alive_timeout);
1634 csp->server_connection.keep_alive_timeout = keep_alive_timeout;
1638 /* XXX: Is this log worthy? */
1639 log_error(LOG_LEVEL_HEADER,
1640 "Server keep-alive timeout is %u. Sticking with %u.",
1641 keep_alive_timeout, csp->server_connection.keep_alive_timeout);
1647 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1651 /*********************************************************************
1653 * Function : client_connection
1655 * Description : Makes sure a proper "Connection:" header is
1656 * set and signals connection_header_adder
1660 * 1 : csp = Current client state (buffers, headers, etc...)
1661 * 2 : header = On input, pointer to header to modify.
1662 * On output, pointer to the modified header, or NULL
1663 * to remove the header. This function frees the
1664 * original string if necessary.
1666 * Returns : JB_ERR_OK on success, or
1667 * JB_ERR_MEMORY on out-of-memory error.
1669 *********************************************************************/
1670 static jb_err client_connection(struct client_state *csp, char **header)
1672 const char *wanted_header = get_appropiate_connection_header(csp);
1674 if (strcmpic(*header, wanted_header))
1676 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1677 if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING))
1679 log_error(LOG_LEVEL_HEADER,
1680 "Keeping the client header '%s' around. "
1681 "The connection will not be kept alive.",
1685 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1687 char *old_header = *header;
1689 *header = strdup(wanted_header);
1692 return JB_ERR_MEMORY;
1694 log_error(LOG_LEVEL_HEADER,
1695 "Replaced: \'%s\' with \'%s\'", old_header, *header);
1699 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
1702 log_error(LOG_LEVEL_HEADER,
1703 "Keeping the client header '%s' around. "
1704 "The server connection will be kept alive if possible.",
1706 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
1708 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
1710 /* Signal client_connection_adder() to return early. */
1711 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
1717 /*********************************************************************
1719 * Function : crumble
1721 * Description : This is called if a header matches a pattern to "crunch"
1724 * 1 : csp = Current client state (buffers, headers, etc...)
1725 * 2 : header = On input, pointer to header to modify.
1726 * On output, pointer to the modified header, or NULL
1727 * to remove the header. This function frees the
1728 * original string if necessary.
1730 * Returns : JB_ERR_OK on success, or
1731 * JB_ERR_MEMORY on out-of-memory error.
1733 *********************************************************************/
1734 static jb_err crumble(struct client_state *csp, char **header)
1737 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
1743 /*********************************************************************
1745 * Function : crunch_server_header
1747 * Description : Crunch server header if it matches a string supplied by the
1748 * user. Called from `sed'.
1751 * 1 : csp = Current client state (buffers, headers, etc...)
1752 * 2 : header = On input, pointer to header to modify.
1753 * On output, pointer to the modified header, or NULL
1754 * to remove the header. This function frees the
1755 * original string if necessary.
1757 * Returns : JB_ERR_OK on success and always succeeds
1759 *********************************************************************/
1760 static jb_err crunch_server_header(struct client_state *csp, char **header)
1762 const char *crunch_pattern;
1764 /* Do we feel like crunching? */
1765 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
1767 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
1769 /* Is the current header the lucky one? */
1770 if (strstr(*header, crunch_pattern))
1772 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
1781 /*********************************************************************
1783 * Function : server_content_type
1785 * Description : Set the content-type for filterable types (text/.*,
1786 * .*xml.*, javascript and image/gif) unless filtering has been
1787 * forbidden (CT_TABOO) while parsing earlier headers.
1788 * NOTE: Since text/plain is commonly used by web servers
1789 * for files whose correct type is unknown, we don't
1790 * set CT_TEXT for it.
1793 * 1 : csp = Current client state (buffers, headers, etc...)
1794 * 2 : header = On input, pointer to header to modify.
1795 * On output, pointer to the modified header, or NULL
1796 * to remove the header. This function frees the
1797 * original string if necessary.
1799 * Returns : JB_ERR_OK on success, or
1800 * JB_ERR_MEMORY on out-of-memory error.
1802 *********************************************************************/
1803 static jb_err server_content_type(struct client_state *csp, char **header)
1805 /* Remove header if it isn't the first Content-Type header */
1806 if ((csp->content_type & CT_DECLARED))
1809 * Another, slightly slower, way to see if
1810 * we already parsed another Content-Type header.
1812 assert(NULL != get_header_value(csp->headers, "Content-Type:"));
1814 log_error(LOG_LEVEL_ERROR,
1815 "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
1823 * Signal that the Content-Type has been set.
1825 csp->content_type |= CT_DECLARED;
1827 if (!(csp->content_type & CT_TABOO))
1830 * XXX: The assumption that text/plain is a sign of
1831 * binary data seems to be somewhat unreasonable nowadays
1832 * and should be dropped after 3.0.8 is out.
1834 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
1835 || strstr(*header, "xml")
1836 || strstr(*header, "application/x-javascript"))
1838 csp->content_type |= CT_TEXT;
1840 else if (strstr(*header, "image/gif"))
1842 csp->content_type |= CT_GIF;
1847 * Are we messing with the content type?
1849 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
1852 * Make sure the user doesn't accidently
1853 * change the content type of binary documents.
1855 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
1858 *header = strdup("Content-Type: ");
1859 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
1863 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
1864 return JB_ERR_MEMORY;
1866 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
1870 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
1871 "It doesn't look like a content type that should be filtered. "
1872 "Enable force-text-mode if you know what you're doing.", *header);
1880 /*********************************************************************
1882 * Function : server_transfer_coding
1884 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
1885 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
1886 * - Remove header if body was chunked but has been
1887 * de-chunked for filtering.
1890 * 1 : csp = Current client state (buffers, headers, etc...)
1891 * 2 : header = On input, pointer to header to modify.
1892 * On output, pointer to the modified header, or NULL
1893 * to remove the header. This function frees the
1894 * original string if necessary.
1896 * Returns : JB_ERR_OK on success, or
1897 * JB_ERR_MEMORY on out-of-memory error.
1899 *********************************************************************/
1900 static jb_err server_transfer_coding(struct client_state *csp, char **header)
1903 * Turn off pcrs and gif filtering if body compressed
1905 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
1909 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
1911 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
1912 csp->http->cmd, *header);
1913 #endif /* def FEATURE_ZLIB */
1914 csp->content_type = CT_TABOO;
1918 * Raise flag if body chunked
1920 if (strstr(*header, "chunked"))
1922 csp->flags |= CSP_FLAG_CHUNKED;
1925 * If the body was modified, it has been de-chunked first
1926 * and the header must be removed.
1928 * FIXME: If there is more than one transfer encoding,
1929 * only the "chunked" part should be removed here.
1931 if (csp->flags & CSP_FLAG_MODIFIED)
1933 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
1942 /*********************************************************************
1944 * Function : server_content_encoding
1946 * Description : This function is run twice for each request,
1947 * unless FEATURE_ZLIB and filtering are disabled.
1949 * The first run is used to check if the content
1950 * is compressed, if FEATURE_ZLIB is disabled
1951 * filtering is then disabled as well, if FEATURE_ZLIB
1952 * is enabled the content is marked for decompression.
1954 * The second run is used to remove the Content-Encoding
1955 * header if the decompression was successful.
1958 * 1 : csp = Current client state (buffers, headers, etc...)
1959 * 2 : header = On input, pointer to header to modify.
1960 * On output, pointer to the modified header, or NULL
1961 * to remove the header. This function frees the
1962 * original string if necessary.
1964 * Returns : JB_ERR_OK on success, or
1965 * JB_ERR_MEMORY on out-of-memory error.
1967 *********************************************************************/
1968 static jb_err server_content_encoding(struct client_state *csp, char **header)
1971 if ((csp->flags & CSP_FLAG_MODIFIED)
1972 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
1975 * We successfully decompressed the content,
1976 * and have to clean the header now, so the
1977 * client no longer expects compressed data..
1979 * XXX: There is a difference between cleaning
1980 * and removing it completely.
1982 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
1985 else if (strstr(*header, "gzip"))
1987 /* Mark for gzip decompression */
1988 csp->content_type |= CT_GZIP;
1990 else if (strstr(*header, "deflate"))
1992 /* Mark for zlib decompression */
1993 csp->content_type |= CT_DEFLATE;
1995 else if (strstr(*header, "compress"))
1998 * We can't decompress this; therefore we can't filter
2001 csp->content_type |= CT_TABOO;
2003 #else /* !defined(FEATURE_ZLIB) */
2004 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2007 * Body is compressed, turn off pcrs and gif filtering.
2009 csp->content_type |= CT_TABOO;
2012 * Log a warning if the user expects the content to be filtered.
2014 if ((csp->rlist != NULL) &&
2015 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2017 log_error(LOG_LEVEL_INFO,
2018 "Compressed content detected, content filtering disabled. "
2019 "Consider recompiling Privoxy with zlib support or "
2020 "enable the prevent-compression action.");
2023 #endif /* defined(FEATURE_ZLIB) */
2030 /*********************************************************************
2032 * Function : server_adjust_content_length
2034 * Description : Adjust Content-Length header if we modified
2038 * 1 : csp = Current client state (buffers, headers, etc...)
2039 * 2 : header = On input, pointer to header to modify.
2040 * On output, pointer to the modified header, or NULL
2041 * to remove the header. This function frees the
2042 * original string if necessary.
2044 * Returns : JB_ERR_OK on success, or
2045 * JB_ERR_MEMORY on out-of-memory error.
2047 *********************************************************************/
2048 static jb_err server_adjust_content_length(struct client_state *csp, char **header)
2050 const size_t max_header_length = 80;
2052 /* Regenerate header if the content was modified. */
2053 if (csp->flags & CSP_FLAG_MODIFIED)
2056 *header = (char *) zalloc(max_header_length);
2057 if (*header == NULL)
2059 return JB_ERR_MEMORY;
2062 snprintf(*header, max_header_length, "Content-Length: %d",
2063 (int)csp->content_length);
2064 log_error(LOG_LEVEL_HEADER, "Adjusted Content-Length to %d",
2065 (int)csp->content_length);
2072 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
2073 /*********************************************************************
2075 * Function : server_save_content_length
2077 * Description : Save the Content-Length sent by the server.
2080 * 1 : csp = Current client state (buffers, headers, etc...)
2081 * 2 : header = On input, pointer to header to modify.
2082 * On output, pointer to the modified header, or NULL
2083 * to remove the header. This function frees the
2084 * original string if necessary.
2086 * Returns : JB_ERR_OK on success, or
2087 * JB_ERR_MEMORY on out-of-memory error.
2089 *********************************************************************/
2090 static jb_err server_save_content_length(struct client_state *csp, char **header)
2092 unsigned long long content_length = 0;
2094 assert(*(*header+14) == ':');
2096 if (1 != sscanf(*header+14, ": %llu", &content_length))
2098 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2103 csp->expected_content_length = content_length;
2104 csp->flags |= CSP_FLAG_CONTENT_LENGTH_SET;
2109 #endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
2112 /*********************************************************************
2114 * Function : server_content_md5
2116 * Description : Crumble any Content-MD5 headers if the document was
2117 * modified. FIXME: Should we re-compute instead?
2120 * 1 : csp = Current client state (buffers, headers, etc...)
2121 * 2 : header = On input, pointer to header to modify.
2122 * On output, pointer to the modified header, or NULL
2123 * to remove the header. This function frees the
2124 * original string if necessary.
2126 * Returns : JB_ERR_OK on success, or
2127 * JB_ERR_MEMORY on out-of-memory error.
2129 *********************************************************************/
2130 static jb_err server_content_md5(struct client_state *csp, char **header)
2132 if (csp->flags & CSP_FLAG_MODIFIED)
2134 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2142 /*********************************************************************
2144 * Function : server_content_disposition
2146 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2147 * Called from `sed'.
2150 * 1 : csp = Current client state (buffers, headers, etc...)
2151 * 2 : header = On input, pointer to header to modify.
2152 * On output, pointer to the modified header, or NULL
2153 * to remove the header. This function frees the
2154 * original string if necessary.
2156 * Returns : JB_ERR_OK on success, or
2157 * JB_ERR_MEMORY on out-of-memory error.
2159 *********************************************************************/
2160 static jb_err server_content_disposition(struct client_state *csp, char **header)
2165 * Are we messing with the Content-Disposition header?
2167 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2173 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2175 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2178 * Blocking content-disposition header
2180 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2187 * Replacing Content-Disposition header
2190 *header = strdup("Content-Disposition: ");
2191 string_append(header, newval);
2193 if (*header != NULL)
2195 log_error(LOG_LEVEL_HEADER,
2196 "Content-Disposition header crunched and replaced with: %s", *header);
2199 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2203 /*********************************************************************
2205 * Function : server_last_modified
2207 * Description : Changes Last-Modified header to the actual date
2208 * to help hide-if-modified-since.
2209 * Called from `sed'.
2212 * 1 : csp = Current client state (buffers, headers, etc...)
2213 * 2 : header = On input, pointer to header to modify.
2214 * On output, pointer to the modified header, or NULL
2215 * to remove the header. This function frees the
2216 * original string if necessary.
2218 * Returns : JB_ERR_OK on success, or
2219 * JB_ERR_MEMORY on out-of-memory error.
2221 *********************************************************************/
2222 static jb_err server_last_modified(struct client_state *csp, char **header)
2225 char buf[BUFFER_SIZE];
2228 #ifdef HAVE_GMTIME_R
2231 struct tm *timeptr = NULL;
2232 time_t now, last_modified;
2234 long int days, hours, minutes, seconds;
2237 * Are we messing with the Last-Modified header?
2239 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2245 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2247 if (0 == strcmpic(newval, "block") )
2250 * Blocking Last-Modified header. Useless but why not.
2252 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2256 else if (0 == strcmpic(newval, "reset-to-request-time"))
2259 * Setting Last-Modified Header to now.
2261 get_http_time(0, buf, sizeof(buf));
2263 *header = strdup("Last-Modified: ");
2264 string_append(header, buf);
2266 if (*header == NULL)
2268 log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
2272 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2275 else if (0 == strcmpic(newval, "randomize"))
2277 const char *header_time = *header + sizeof("Last-Modified:");
2279 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2281 #ifdef HAVE_GMTIME_R
2282 gmtime_r(&now, &gmt);
2283 #elif def MUTEX_LOCKS_AVAILABLE
2284 privoxy_mutex_lock(&gmtime_mutex);
2286 privoxy_mutex_unlock(&gmtime_mutex);
2290 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2292 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2297 rtime = (long int)difftime(now, last_modified);
2306 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2308 rtime = pick_from_range(rtime);
2309 if (negative) rtime *= -1;
2310 last_modified += rtime;
2311 #ifdef HAVE_GMTIME_R
2312 timeptr = gmtime_r(&last_modified, &gmt);
2313 #elif def MUTEX_LOCKS_AVAILABLE
2314 privoxy_mutex_lock(&gmtime_mutex);
2315 timeptr = gmtime(&last_modified);
2316 privoxy_mutex_unlock(&gmtime_mutex);
2318 timeptr = gmtime(&last_modified);
2320 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
2322 *header = strdup("Last-Modified: ");
2323 string_append(header, newheader);
2325 if (*header == NULL)
2327 log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
2328 return JB_ERR_MEMORY;
2331 days = rtime / (3600 * 24);
2332 hours = rtime / 3600 % 24;
2333 minutes = rtime / 60 % 60;
2334 seconds = rtime % 60;
2336 log_error(LOG_LEVEL_HEADER,
2337 "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2338 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2339 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2343 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2352 /*********************************************************************
2354 * Function : client_accept_encoding
2356 * Description : Rewrite the client's Accept-Encoding header so that
2357 * if doesn't allow compression, if the action applies.
2358 * Note: For HTTP/1.0 the absence of the header is enough.
2361 * 1 : csp = Current client state (buffers, headers, etc...)
2362 * 2 : header = On input, pointer to header to modify.
2363 * On output, pointer to the modified header, or NULL
2364 * to remove the header. This function frees the
2365 * original string if necessary.
2367 * Returns : JB_ERR_OK on success, or
2368 * JB_ERR_MEMORY on out-of-memory error.
2370 *********************************************************************/
2371 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2373 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2375 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2379 /* Temporarily disable the correct behaviour to
2380 * work around a PHP bug.
2382 * if (!strcmpic(csp->http->ver, "HTTP/1.1"))
2384 * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0");
2385 * if (*header == NULL)
2387 * return JB_ERR_MEMORY;
2398 /*********************************************************************
2400 * Function : client_te
2402 * Description : Rewrite the client's TE header so that
2403 * if doesn't allow compression, if the action applies.
2406 * 1 : csp = Current client state (buffers, headers, etc...)
2407 * 2 : header = On input, pointer to header to modify.
2408 * On output, pointer to the modified header, or NULL
2409 * to remove the header. This function frees the
2410 * original string if necessary.
2412 * Returns : JB_ERR_OK on success, or
2413 * JB_ERR_MEMORY on out-of-memory error.
2415 *********************************************************************/
2416 static jb_err client_te(struct client_state *csp, char **header)
2418 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2421 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2428 /*********************************************************************
2430 * Function : client_referrer
2432 * Description : Handle the "referer" config setting properly.
2433 * Called from `sed'.
2436 * 1 : csp = Current client state (buffers, headers, etc...)
2437 * 2 : header = On input, pointer to header to modify.
2438 * On output, pointer to the modified header, or NULL
2439 * to remove the header. This function frees the
2440 * original string if necessary.
2442 * Returns : JB_ERR_OK on success, or
2443 * JB_ERR_MEMORY on out-of-memory error.
2445 *********************************************************************/
2446 static jb_err client_referrer(struct client_state *csp, char **header)
2448 const char *parameter;
2449 /* booleans for parameters we have to check multiple times */
2450 int parameter_conditional_block;
2451 int parameter_conditional_forge;
2453 #ifdef FEATURE_FORCE_LOAD
2455 * Since the referrer can include the prefix even
2456 * if the request itself is non-forced, we must
2457 * clean it unconditionally.
2459 * XXX: strclean is too broad
2461 strclean(*header, FORCE_PREFIX);
2462 #endif /* def FEATURE_FORCE_LOAD */
2464 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2466 /* Nothing left to do */
2470 parameter = csp->action->string[ACTION_STRING_REFERER];
2471 assert(parameter != NULL);
2472 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2473 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2475 if (!parameter_conditional_block && !parameter_conditional_forge)
2478 * As conditional-block and conditional-forge are the only
2479 * parameters that rely on the original referrer, we can
2480 * remove it now for all the others.
2485 if (0 == strcmpic(parameter, "block"))
2487 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2490 else if (parameter_conditional_block || parameter_conditional_forge)
2492 return handle_conditional_hide_referrer_parameter(header,
2493 csp->http->hostport, parameter_conditional_block);
2495 else if (0 == strcmpic(parameter, "forge"))
2497 return create_forged_referrer(header, csp->http->hostport);
2501 /* interpret parameter as user-supplied referer to fake */
2502 return create_fake_referrer(header, parameter);
2507 /*********************************************************************
2509 * Function : client_accept_language
2511 * Description : Handle the "Accept-Language" config setting properly.
2512 * Called from `sed'.
2515 * 1 : csp = Current client state (buffers, headers, etc...)
2516 * 2 : header = On input, pointer to header to modify.
2517 * On output, pointer to the modified header, or NULL
2518 * to remove the header. This function frees the
2519 * original string if necessary.
2521 * Returns : JB_ERR_OK on success, or
2522 * JB_ERR_MEMORY on out-of-memory error.
2524 *********************************************************************/
2525 static jb_err client_accept_language(struct client_state *csp, char **header)
2530 * Are we messing with the Accept-Language?
2532 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2534 /*I don't think so*/
2538 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2540 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
2543 * Blocking Accept-Language header
2545 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2552 * Replacing Accept-Language header
2555 *header = strdup("Accept-Language: ");
2556 string_append(header, newval);
2558 if (*header == NULL)
2560 log_error(LOG_LEVEL_ERROR,
2561 "Insufficient memory. Accept-Language header crunched without replacement.");
2565 log_error(LOG_LEVEL_HEADER,
2566 "Accept-Language header crunched and replaced with: %s", *header);
2569 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2573 /*********************************************************************
2575 * Function : crunch_client_header
2577 * Description : Crunch client header if it matches a string supplied by the
2578 * user. Called from `sed'.
2581 * 1 : csp = Current client state (buffers, headers, etc...)
2582 * 2 : header = On input, pointer to header to modify.
2583 * On output, pointer to the modified header, or NULL
2584 * to remove the header. This function frees the
2585 * original string if necessary.
2587 * Returns : JB_ERR_OK on success and always succeeds
2589 *********************************************************************/
2590 static jb_err crunch_client_header(struct client_state *csp, char **header)
2592 const char *crunch_pattern;
2594 /* Do we feel like crunching? */
2595 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2597 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2599 /* Is the current header the lucky one? */
2600 if (strstr(*header, crunch_pattern))
2602 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
2610 /*********************************************************************
2612 * Function : client_uagent
2614 * Description : Handle the "user-agent" config setting properly
2615 * and remember its original value to enable browser
2616 * bug workarounds. Called from `sed'.
2619 * 1 : csp = Current client state (buffers, headers, etc...)
2620 * 2 : header = On input, pointer to header to modify.
2621 * On output, pointer to the modified header, or NULL
2622 * to remove the header. This function frees the
2623 * original string if necessary.
2625 * Returns : JB_ERR_OK on success, or
2626 * JB_ERR_MEMORY on out-of-memory error.
2628 *********************************************************************/
2629 static jb_err client_uagent(struct client_state *csp, char **header)
2633 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
2638 newval = csp->action->string[ACTION_STRING_USER_AGENT];
2645 *header = strdup("User-Agent: ");
2646 string_append(header, newval);
2648 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
2650 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2654 /*********************************************************************
2656 * Function : client_ua
2658 * Description : Handle "ua-" headers properly. Called from `sed'.
2661 * 1 : csp = Current client state (buffers, headers, etc...)
2662 * 2 : header = On input, pointer to header to modify.
2663 * On output, pointer to the modified header, or NULL
2664 * to remove the header. This function frees the
2665 * original string if necessary.
2667 * Returns : JB_ERR_OK on success, or
2668 * JB_ERR_MEMORY on out-of-memory error.
2670 *********************************************************************/
2671 static jb_err client_ua(struct client_state *csp, char **header)
2673 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
2675 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
2683 /*********************************************************************
2685 * Function : client_from
2687 * Description : Handle the "from" config setting properly.
2688 * Called from `sed'.
2691 * 1 : csp = Current client state (buffers, headers, etc...)
2692 * 2 : header = On input, pointer to header to modify.
2693 * On output, pointer to the modified header, or NULL
2694 * to remove the header. This function frees the
2695 * original string if necessary.
2697 * Returns : JB_ERR_OK on success, or
2698 * JB_ERR_MEMORY on out-of-memory error.
2700 *********************************************************************/
2701 static jb_err client_from(struct client_state *csp, char **header)
2705 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
2712 newval = csp->action->string[ACTION_STRING_FROM];
2715 * Are we blocking the e-mail address?
2717 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
2719 log_error(LOG_LEVEL_HEADER, "crunched From!");
2723 log_error(LOG_LEVEL_HEADER, " modified");
2725 *header = strdup("From: ");
2726 string_append(header, newval);
2728 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2732 /*********************************************************************
2734 * Function : client_send_cookie
2736 * Description : Crunches the "cookie" header if necessary.
2737 * Called from `sed'.
2739 * XXX: Stupid name, doesn't send squat.
2742 * 1 : csp = Current client state (buffers, headers, etc...)
2743 * 2 : header = On input, pointer to header to modify.
2744 * On output, pointer to the modified header, or NULL
2745 * to remove the header. This function frees the
2746 * original string if necessary.
2748 * Returns : JB_ERR_OK on success, or
2749 * JB_ERR_MEMORY on out-of-memory error.
2751 *********************************************************************/
2752 static jb_err client_send_cookie(struct client_state *csp, char **header)
2754 if (csp->action->flags & ACTION_NO_COOKIE_READ)
2756 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
2764 /*********************************************************************
2766 * Function : client_x_forwarded
2768 * Description : Handle the "x-forwarded-for" config setting properly,
2769 * also used in the add_client_headers list. Called from `sed'.
2772 * 1 : csp = Current client state (buffers, headers, etc...)
2773 * 2 : header = On input, pointer to header to modify.
2774 * On output, pointer to the modified header, or NULL
2775 * to remove the header. This function frees the
2776 * original string if necessary.
2778 * Returns : JB_ERR_OK on success, or
2779 * JB_ERR_MEMORY on out-of-memory error.
2781 *********************************************************************/
2782 jb_err client_x_forwarded(struct client_state *csp, char **header)
2784 if (0 != (csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR))
2786 const char *parameter = csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR];
2788 if (0 == strcmpic(parameter, "block"))
2791 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
2793 else if (0 == strcmpic(parameter, "add"))
2795 string_append(header, ", ");
2796 string_append(header, csp->ip_addr_str);
2798 if (*header == NULL)
2800 return JB_ERR_MEMORY;
2802 log_error(LOG_LEVEL_HEADER,
2803 "Appended client IP address to %s", *header);
2804 csp->flags |= CSP_FLAG_X_FORWARDED_FOR_APPENDED;
2808 log_error(LOG_LEVEL_FATAL,
2809 "Invalid change-x-forwarded-for parameter: '%s'", parameter);
2817 /*********************************************************************
2819 * Function : client_max_forwards
2821 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
2822 * from the value of the Max-Forwards header field.
2825 * 1 : csp = Current client state (buffers, headers, etc...)
2826 * 2 : header = On input, pointer to header to modify.
2827 * On output, pointer to the modified header, or NULL
2828 * to remove the header. This function frees the
2829 * original string if necessary.
2831 * Returns : JB_ERR_OK on success, or
2832 * JB_ERR_MEMORY on out-of-memory error.
2834 *********************************************************************/
2835 static jb_err client_max_forwards(struct client_state *csp, char **header)
2839 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
2840 (0 == strcmpic(csp->http->gpc, "options")))
2842 assert(*(*header+12) == ':');
2843 if (1 == sscanf(*header+12, ": %d", &max_forwards))
2845 if (max_forwards > 0)
2847 snprintf(*header, strlen(*header)+1, "Max-Forwards: %d", --max_forwards);
2848 log_error(LOG_LEVEL_HEADER,
2849 "Max-Forwards value for %s request reduced to %d.",
2850 csp->http->gpc, max_forwards);
2852 else if (max_forwards < 0)
2854 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2860 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
2869 /*********************************************************************
2871 * Function : client_host
2873 * Description : If the request URI did not contain host and
2874 * port information, parse and evaluate the Host
2877 * Also, kill ill-formed HOST: headers as sent by
2878 * Apple's iTunes software when used with a proxy.
2881 * 1 : csp = Current client state (buffers, headers, etc...)
2882 * 2 : header = On input, pointer to header to modify.
2883 * On output, pointer to the modified header, or NULL
2884 * to remove the header. This function frees the
2885 * original string if necessary.
2887 * Returns : JB_ERR_OK on success, or
2888 * JB_ERR_MEMORY on out-of-memory error.
2890 *********************************************************************/
2891 static jb_err client_host(struct client_state *csp, char **header)
2896 * If the header field name is all upper-case, chances are that it's
2897 * an ill-formed one from iTunes. BTW, killing innocent headers here is
2898 * not a problem -- they are regenerated later.
2900 if ((*header)[1] == 'O')
2902 log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
2907 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
2908 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
2911 if (NULL == (p = strdup((*header)+6)))
2913 return JB_ERR_MEMORY;
2916 if (NULL == (q = strdup(p)))
2919 return JB_ERR_MEMORY;
2922 freez(csp->http->hostport);
2923 csp->http->hostport = p;
2924 freez(csp->http->host);
2925 csp->http->host = q;
2926 q = strchr(csp->http->host, ':');
2929 /* Terminate hostname and evaluate port string */
2931 csp->http->port = atoi(q);
2935 csp->http->port = csp->http->ssl ? 443 : 80;
2938 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
2939 csp->http->hostport, csp->http->host, csp->http->port);
2942 /* Signal client_host_adder() to return right away */
2943 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
2949 /*********************************************************************
2951 * Function : client_if_modified_since
2953 * Description : Remove or modify the If-Modified-Since header.
2956 * 1 : csp = Current client state (buffers, headers, etc...)
2957 * 2 : header = On input, pointer to header to modify.
2958 * On output, pointer to the modified header, or NULL
2959 * to remove the header. This function frees the
2960 * original string if necessary.
2962 * Returns : JB_ERR_OK on success, or
2963 * JB_ERR_MEMORY on out-of-memory error.
2965 *********************************************************************/
2966 static jb_err client_if_modified_since(struct client_state *csp, char **header)
2969 #ifdef HAVE_GMTIME_R
2972 struct tm *timeptr = NULL;
2976 long int hours, minutes, seconds;
2980 if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
2983 * The client got an error message because of a temporary problem,
2984 * the problem is gone and the client now tries to revalidate our
2985 * error message on the real server. The revalidation would always
2986 * end with the transmission of the whole document and there is
2987 * no need to expose the bogus If-Modified-Since header.
2989 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
2992 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
2994 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
2996 if ((0 == strcmpic(newval, "block")))
2998 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3001 else /* add random value */
3003 const char *header_time = *header + sizeof("If-Modified-Since:");
3005 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3007 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3012 rtime = strtol(newval, &endptr, 0);
3015 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3016 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3023 rtime = pick_from_range(rtime);
3027 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3030 tm += rtime * (negative ? -1 : 1);
3031 #ifdef HAVE_GMTIME_R
3032 timeptr = gmtime_r(&tm, &gmt);
3033 #elif def MUTEX_LOCKS_AVAILABLE
3034 privoxy_mutex_lock(&gmtime_mutex);
3035 timeptr = gmtime(&tm);
3036 privoxy_mutex_unlock(&gmtime_mutex);
3038 timeptr = gmtime(&tm);
3040 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
3043 *header = strdup("If-Modified-Since: ");
3044 string_append(header, newheader);
3046 if (*header == NULL)
3048 log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
3049 return JB_ERR_MEMORY;
3052 hours = rtime / 3600;
3053 minutes = rtime / 60 % 60;
3054 seconds = rtime % 60;
3056 log_error(LOG_LEVEL_HEADER,
3057 "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3058 *header, (negative) ? "subtracted" : "added", hours,
3059 (hours == 1) ? "r" : "rs", minutes, (minutes == 1) ? "e" : "es",
3060 seconds, (seconds == 1) ? ")" : "s)");
3069 /*********************************************************************
3071 * Function : client_if_none_match
3073 * Description : Remove the If-None-Match header.
3076 * 1 : csp = Current client state (buffers, headers, etc...)
3077 * 2 : header = On input, pointer to header to modify.
3078 * On output, pointer to the modified header, or NULL
3079 * to remove the header. This function frees the
3080 * original string if necessary.
3082 * Returns : JB_ERR_OK on success, or
3083 * JB_ERR_MEMORY on out-of-memory error.
3085 *********************************************************************/
3086 static jb_err client_if_none_match(struct client_state *csp, char **header)
3088 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3090 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3098 /*********************************************************************
3100 * Function : client_x_filter
3102 * Description : Disables filtering if the client set "X-Filter: No".
3103 * Called from `sed'.
3106 * 1 : csp = Current client state (buffers, headers, etc...)
3107 * 2 : header = On input, pointer to header to modify.
3108 * On output, pointer to the modified header, or NULL
3109 * to remove the header. This function frees the
3110 * original string if necessary.
3112 * Returns : JB_ERR_OK on success
3114 *********************************************************************/
3115 jb_err client_x_filter(struct client_state *csp, char **header)
3117 if ( 0 == strcmpic(*header, "X-Filter: No"))
3119 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3121 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3125 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3127 log_error(LOG_LEVEL_HEADER,
3128 "force-text-mode overruled the client's request to fetch without filtering!");
3132 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3133 csp->flags |= CSP_FLAG_NO_FILTERING;
3134 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3136 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3144 /*********************************************************************
3146 * Function : client_range
3148 * Description : Removes Range, Request-Range and If-Range headers if
3149 * content filtering is enabled. If the client's version
3150 * of the document has been altered by Privoxy, the server
3151 * could interpret the range differently than the client
3152 * intended in which case the user could end up with
3153 * corrupted content.
3156 * 1 : csp = Current client state (buffers, headers, etc...)
3157 * 2 : header = On input, pointer to header to modify.
3158 * On output, pointer to the modified header, or NULL
3159 * to remove the header. This function frees the
3160 * original string if necessary.
3162 * Returns : JB_ERR_OK
3164 *********************************************************************/
3165 static jb_err client_range(struct client_state *csp, char **header)
3167 if (content_filters_enabled(csp->action))
3169 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3170 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3177 /* the following functions add headers directly to the header list */
3179 /*********************************************************************
3181 * Function : client_host_adder
3183 * Description : Adds the Host: header field if it is missing.
3184 * Called from `sed'.
3187 * 1 : csp = Current client state (buffers, headers, etc...)
3189 * Returns : JB_ERR_OK on success, or
3190 * JB_ERR_MEMORY on out-of-memory error.
3192 *********************************************************************/
3193 static jb_err client_host_adder(struct client_state *csp)
3198 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3200 /* Header already set by the client, nothing to do. */
3204 if ( !csp->http->hostport || !*(csp->http->hostport))
3206 /* XXX: When does this happen and why is it OK? */
3207 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3212 * remove 'user:pass@' from 'proto://user:pass@host'
3214 if ( (p = strchr( csp->http->hostport, '@')) != NULL )
3220 p = csp->http->hostport;
3223 /* XXX: Just add it, we already made sure that it will be unique */
3224 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3225 err = enlist_unique_header(csp->headers, "Host", p);
3232 /*********************************************************************
3234 * Function : client_accept_encoding_adder
3236 * Description : Add an Accept-Encoding header to the client's request
3237 * that disables compression if the action applies, and
3238 * the header is not already there. Called from `sed'.
3239 * Note: For HTTP/1.0, the absence of the header is enough.
3242 * 1 : csp = Current client state (buffers, headers, etc...)
3244 * Returns : JB_ERR_OK on success, or
3245 * JB_ERR_MEMORY on out-of-memory error.
3247 *********************************************************************/
3248 static jb_err client_accept_encoding_adder(struct client_state *csp)
3250 if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3251 && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
3253 return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
3261 /*********************************************************************
3263 * Function : client_xtra_adder
3265 * Description : Used in the add_client_headers list. Called from `sed'.
3268 * 1 : csp = Current client state (buffers, headers, etc...)
3270 * Returns : JB_ERR_OK on success, or
3271 * JB_ERR_MEMORY on out-of-memory error.
3273 *********************************************************************/
3274 static jb_err client_xtra_adder(struct client_state *csp)
3276 struct list_entry *lst;
3279 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3280 lst ; lst = lst->next)
3282 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3283 err = enlist(csp->headers, lst->str);
3295 /*********************************************************************
3297 * Function : client_x_forwarded_for_adder
3299 * Description : Used in the add_client_headers list. Called from `sed'.
3302 * 1 : csp = Current client state (buffers, headers, etc...)
3304 * Returns : JB_ERR_OK on success, or
3305 * JB_ERR_MEMORY on out-of-memory error.
3307 *********************************************************************/
3308 static jb_err client_x_forwarded_for_adder(struct client_state *csp)
3310 char *header = NULL;
3313 if (!((csp->action->flags & ACTION_CHANGE_X_FORWARDED_FOR)
3314 && (0 == strcmpic(csp->action->string[ACTION_STRING_CHANGE_X_FORWARDED_FOR], "add")))
3315 || (csp->flags & CSP_FLAG_X_FORWARDED_FOR_APPENDED))
3318 * If we aren't adding X-Forwarded-For headers,
3319 * or we already appended an existing X-Forwarded-For
3320 * header, there's nothing left to do here.
3325 header = strdup("X-Forwarded-For: ");
3326 string_append(&header, csp->ip_addr_str);
3330 return JB_ERR_MEMORY;
3333 log_error(LOG_LEVEL_HEADER, "addh: %s", header);
3334 err = enlist(csp->headers, header);
3341 /*********************************************************************
3343 * Function : server_connection_adder
3345 * Description : Adds an appropiate "Connection:" header to csp->headers
3346 * unless the header was already present. Called from `sed'.
3349 * 1 : csp = Current client state (buffers, headers, etc...)
3351 * Returns : JB_ERR_OK on success, or
3352 * JB_ERR_MEMORY on out-of-memory error.
3354 *********************************************************************/
3355 static jb_err server_connection_adder(struct client_state *csp)
3357 const unsigned int flags = csp->flags;
3358 const char *response_status_line = csp->headers->first->str;
3359 const char *wanted_header = get_appropiate_connection_header(csp);
3361 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3362 && (flags & CSP_FLAG_SERVER_CONNECTION_HEADER_SET))
3368 * XXX: if we downgraded the response, this check will fail.
3370 if ((csp->config->feature_flags &
3371 RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3372 && (NULL != response_status_line)
3373 && !strncmpic(response_status_line, "HTTP/1.1", 8))
3375 log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
3376 "without Connection header implies keep-alive.");
3377 csp->flags |= CSP_FLAG_SERVER_CONNECTION_KEEP_ALIVE;
3380 log_error(LOG_LEVEL_HEADER, "Adding: %s", wanted_header);
3382 return enlist(csp->headers, wanted_header);
3386 #ifdef FEATURE_CONNECTION_KEEP_ALIVE
3387 /*********************************************************************
3389 * Function : server_proxy_connection_adder
3391 * Description : Adds a "Proxy-Connection: keep-alive" header to
3392 * csp->headers. XXX: We should reuse existant ones.
3395 * 1 : csp = Current client state (buffers, headers, etc...)
3397 * Returns : JB_ERR_OK on success, or
3398 * JB_ERR_MEMORY on out-of-memory error.
3400 *********************************************************************/
3401 static jb_err server_proxy_connection_adder(struct client_state *csp)
3403 static const char proxy_connection_header[] = "Proxy-Connection: keep-alive";
3404 log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
3405 return enlist(csp->headers, proxy_connection_header);
3407 #endif /* FEATURE_CONNECTION_KEEP_ALIVE */
3410 /*********************************************************************
3412 * Function : client_connection_header_adder
3414 * Description : Adds a proper "Connection:" header to csp->headers
3415 * unless the header was already present. Called from `sed'.
3418 * 1 : csp = Current client state (buffers, headers, etc...)
3420 * Returns : JB_ERR_OK on success, or
3421 * JB_ERR_MEMORY on out-of-memory error.
3423 *********************************************************************/
3424 static jb_err client_connection_header_adder(struct client_state *csp)
3426 const char *wanted_header = get_appropiate_connection_header(csp);
3428 if (!(csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3429 && (csp->flags & CSP_FLAG_CLIENT_CONNECTION_HEADER_SET))
3434 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
3435 && (csp->http->ssl == 0))
3437 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
3440 log_error(LOG_LEVEL_HEADER, "Adding: %s", wanted_header);
3442 return enlist(csp->headers, wanted_header);
3446 /*********************************************************************
3448 * Function : server_http
3450 * Description : - Save the HTTP Status into csp->http->status
3451 * - Set CT_TABOO to prevent filtering if the answer
3452 * is a partial range (HTTP status 206)
3453 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3457 * 1 : csp = Current client state (buffers, headers, etc...)
3458 * 2 : header = On input, pointer to header to modify.
3459 * On output, pointer to the modified header, or NULL
3460 * to remove the header. This function frees the
3461 * original string if necessary.
3463 * Returns : JB_ERR_OK on success, or
3464 * JB_ERR_MEMORY on out-of-memory error.
3466 *********************************************************************/
3467 static jb_err server_http(struct client_state *csp, char **header)
3469 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3470 if (csp->http->status == 206)
3472 csp->content_type = CT_TABOO;
3475 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3477 /* XXX: Should we do a real validity check here? */
3478 if (strlen(*header) > 8)
3481 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3486 * XXX: Should we block the request or
3487 * enlist a valid status code line here?
3489 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3490 "Downgrading to HTTP/1.0 impossible.");
3498 /*********************************************************************
3500 * Function : server_set_cookie
3502 * Description : Handle the server "cookie" header properly.
3503 * Log cookie to the jar file. Then "crunch",
3504 * accept or rewrite it to a session cookie.
3505 * Called from `sed'.
3507 * TODO: Allow the user to specify a new expiration
3508 * time to cause the cookie to expire even before the
3509 * browser is closed.
3512 * 1 : csp = Current client state (buffers, headers, etc...)
3513 * 2 : header = On input, pointer to header to modify.
3514 * On output, pointer to the modified header, or NULL
3515 * to remove the header. This function frees the
3516 * original string if necessary.
3518 * Returns : JB_ERR_OK on success, or
3519 * JB_ERR_MEMORY on out-of-memory error.
3521 *********************************************************************/
3522 static jb_err server_set_cookie(struct client_state *csp, char **header)
3529 if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
3531 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3534 else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
3536 /* Flag whether or not to log a message */
3539 /* A variable to store the tag we're working on */
3542 /* Skip "Set-Cookie:" (11 characters) in header */
3543 cur_tag = *header + 11;
3545 /* skip whitespace between "Set-Cookie:" and value */
3546 while (*cur_tag && ijb_isspace(*cur_tag))
3551 /* Loop through each tag in the cookie */
3555 char *next_tag = strchr(cur_tag, ';');
3556 if (next_tag != NULL)
3558 /* Skip the ';' character itself */
3561 /* skip whitespace ";" and start of tag */
3562 while (*next_tag && ijb_isspace(*next_tag))
3569 /* "Next tag" is the end of the string */
3570 next_tag = cur_tag + strlen(cur_tag);
3574 * Check the expiration date to see
3575 * if the cookie is still valid, if yes,
3576 * rewrite it to a session cookie.
3578 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
3580 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
3582 /* Did we detect the date properly? */
3583 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
3586 * Nope, treat it as if it was still valid.
3588 * XXX: Should we remove the whole cookie instead?
3590 log_error(LOG_LEVEL_ERROR,
3591 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
3592 string_move(cur_tag, next_tag);
3598 * Yes. Check if the cookie is still valid.
3600 * If the cookie is already expired it's probably
3601 * a delete cookie and even if it isn't, the browser
3602 * will discard it anyway.
3606 * XXX: timegm() isn't available on some AmigaOS
3607 * versions and our replacement doesn't work.
3609 * Our options are to either:
3611 * - disable session-cookies-only completely if timegm
3614 * - to simply remove all expired tags, like it has
3615 * been done until Privoxy 3.0.6 and to live with
3616 * the consequence that it can cause login/logout
3617 * problems on servers that don't validate their
3618 * input properly, or
3620 * - to replace it with mktime in which
3621 * case there is a slight chance of valid cookies
3622 * passing as already expired.
3624 * This is the way it's currently done and it's not
3625 * as bad as it sounds. If the missing GMT offset is
3626 * enough to change the result of the expiration check
3627 * the cookie will be only valid for a few hours
3628 * anyway, which in many cases will be shorter
3629 * than a browser session.
3631 if (cookie_time - now < 0)
3633 log_error(LOG_LEVEL_HEADER,
3634 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
3635 /* Just in case some clown sets more then one expiration date */
3641 * Still valid, delete expiration date by copying
3642 * the rest of the string over it.
3644 string_move(cur_tag, next_tag);
3646 /* That changed the header, need to issue a log message */
3650 * Note that the next tag has now been moved to *cur_tag,
3651 * so we do not need to update the cur_tag pointer.
3659 /* Move on to next cookie tag */
3666 assert(NULL != *header);
3667 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
3676 #ifdef FEATURE_FORCE_LOAD
3677 /*********************************************************************
3679 * Function : strclean
3681 * Description : In-Situ-Eliminate all occurances of substring in
3685 * 1 : string = string to clean
3686 * 2 : substring = substring to eliminate
3688 * Returns : Number of eliminations
3690 *********************************************************************/
3691 int strclean(char *string, const char *substring)
3697 len = strlen(substring);
3699 while((pos = strstr(string, substring)) != NULL)
3706 while (*p++ != '\0');
3713 #endif /* def FEATURE_FORCE_LOAD */
3716 /*********************************************************************
3718 * Function : parse_header_time
3720 * Description : Parses time formats used in HTTP header strings
3721 * to get the numerical respresentation.
3724 * 1 : header_time = HTTP header time as string.
3725 * 2 : result = storage for header_time in seconds
3727 * Returns : JB_ERR_OK if the time format was recognized, or
3728 * JB_ERR_PARSE otherwise.
3730 *********************************************************************/
3731 static jb_err parse_header_time(const char *header_time, time_t *result)
3736 * Zero out gmt to prevent time zone offsets.
3738 * While this is only necessary on some platforms
3739 * (mingw32 for example), I don't know how to
3740 * detect these automatically and doing it everywhere
3743 memset(&gmt, 0, sizeof(gmt));
3745 /* Tue, 02 Jun 2037 20:00:00 */
3746 if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
3747 /* Tue, 02-Jun-2037 20:00:00 */
3748 && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
3749 /* Tue, 02-Jun-37 20:00:00 */
3750 && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
3751 /* Tuesday, 02-Jun-2037 20:00:00 */
3752 && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
3753 /* Tuesday Jun 02 20:00:00 2037 */
3754 && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
3756 return JB_ERR_PARSE;
3759 *result = timegm(&gmt);
3766 /*********************************************************************
3768 * Function : get_destination_from_headers
3770 * Description : Parse the "Host:" header to get the request's destination.
3771 * Only needed if the client's request was forcefully
3772 * redirected into Privoxy.
3774 * Code mainly copied from client_host() which is currently
3775 * run too late for this purpose.
3778 * 1 : headers = List of headers (one of them hopefully being
3779 * the "Host:" header)
3780 * 2 : http = storage for the result (host, port and hostport).
3782 * Returns : JB_ERR_MEMORY in case of memory problems,
3783 * JB_ERR_PARSE if the host header couldn't be found,
3784 * JB_ERR_OK otherwise.
3786 *********************************************************************/
3787 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
3793 host = get_header_value(headers, "Host:");
3797 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
3798 return JB_ERR_PARSE;
3801 if (NULL == (p = strdup((host))))
3803 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
3804 return JB_ERR_MEMORY;
3807 if (NULL == (q = strdup(p)))
3810 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
3811 return JB_ERR_MEMORY;
3814 freez(http->hostport);
3818 q = strchr(http->host, ':');
3821 /* Terminate hostname and evaluate port string */
3823 http->port = atoi(q);
3827 http->port = http->ssl ? 443 : 80;
3830 /* Rebuild request URL */
3832 http->url = strdup(http->ssl ? "https://" : "http://");
3833 string_append(&http->url, http->hostport);
3834 string_append(&http->url, http->path);
3835 if (http->url == NULL)
3837 return JB_ERR_MEMORY;
3840 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
3848 /*********************************************************************
3850 * Function : create_forged_referrer
3852 * Description : Helper for client_referrer to forge a referer as
3853 * 'http://[hostname:port/' to fool stupid
3854 * checks for in-site links
3857 * 1 : header = Pointer to header pointer
3858 * 2 : hostport = Host and optionally port as string
3860 * Returns : JB_ERR_OK in case of success, or
3861 * JB_ERR_MEMORY in case of memory problems.
3863 *********************************************************************/
3864 static jb_err create_forged_referrer(char **header, const char *hostport)
3866 assert(NULL == *header);
3868 *header = strdup("Referer: http://");
3869 string_append(header, hostport);
3870 string_append(header, "/");
3872 if (NULL == *header)
3874 return JB_ERR_MEMORY;
3877 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
3884 /*********************************************************************
3886 * Function : create_fake_referrer
3888 * Description : Helper for client_referrer to create a fake referrer
3889 * based on a string supplied by the user.
3892 * 1 : header = Pointer to header pointer
3893 * 2 : hosthost = Referrer to fake
3895 * Returns : JB_ERR_OK in case of success, or
3896 * JB_ERR_MEMORY in case of memory problems.
3898 *********************************************************************/
3899 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
3901 assert(NULL == *header);
3903 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
3905 log_error(LOG_LEVEL_HEADER,
3906 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
3908 *header = strdup("Referer: ");
3909 string_append(header, fake_referrer);
3911 if (NULL == *header)
3913 return JB_ERR_MEMORY;
3916 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
3923 /*********************************************************************
3925 * Function : handle_conditional_hide_referrer_parameter
3927 * Description : Helper for client_referrer to crunch or forge
3928 * the referrer header if the host has changed.
3931 * 1 : header = Pointer to header pointer
3932 * 2 : host = The target host (may include the port)
3933 * 3 : parameter_conditional_block = Boolean to signal
3934 * if we're in conditional-block mode. If not set,
3935 * we're in conditional-forge mode.
3937 * Returns : JB_ERR_OK in case of success, or
3938 * JB_ERR_MEMORY in case of memory problems.
3940 *********************************************************************/
3941 static jb_err handle_conditional_hide_referrer_parameter(char **header,
3942 const char *host, const int parameter_conditional_block)
3944 char *referer = strdup(*header);
3945 const size_t hostlenght = strlen(host);
3946 const char *referer_url = NULL;
3948 if (NULL == referer)
3951 return JB_ERR_MEMORY;
3954 /* referer begins with 'Referer: http[s]://' */
3955 if ((hostlenght+17) < strlen(referer))
3958 * Shorten referer to make sure the referer is blocked
3959 * if www.example.org/www.example.com-shall-see-the-referer/
3960 * links to www.example.com/
3962 referer[hostlenght+17] = '\0';
3964 referer_url = strstr(referer, "http://");
3965 if ((NULL == referer_url) || (NULL == strstr(referer_url, host)))
3967 /* Host has changed, Referer is invalid or a https URL. */
3968 if (parameter_conditional_block)
3970 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
3977 return create_forged_referrer(header, host);
3987 /*********************************************************************
3989 * Function : get_appropiate_connection_header
3991 * Description : Returns an appropiate Connection header
3992 * depending on whether or not we try to keep
3993 * the connection to the server alive.
3996 * 1 : csp = Current client state (buffers, headers, etc...)
3998 * Returns : Pointer to statically allocated header buffer.
4000 *********************************************************************/
4001 static const char *get_appropiate_connection_header(const struct client_state *csp)
4003 static const char connection_keep_alive[] = "Connection: keep-alive";
4004 static const char connection_close[] = "Connection: close";
4006 if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
4007 && (csp->http->ssl == 0))
4009 return connection_keep_alive;
4011 return connection_close;