-const char parsers_rcs[] = "$Id: parsers.c,v 1.92 2007/03/05 13:25:32 fabiankeil Exp $";
+const char parsers_rcs[] = "$Id: parsers.c,v 1.95 2007/03/25 14:26:40 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
*
* Revisions :
* $Log: parsers.c,v $
+ * Revision 1.95 2007/03/25 14:26:40 fabiankeil
+ * - Fix warnings when compiled with glibc.
+ * - Don't use crumble() for cookie crunching.
+ * - Move cookie time parsing into parse_header_time().
+ * - Let parse_header_time() return a jb_err code
+ * instead of a pointer that can only be used to
+ * check for NULL anyway.
+ *
+ * Revision 1.94 2007/03/21 12:23:53 fabiankeil
+ * - Add better protection against malicious gzip headers.
+ * - Stop logging the first hundred bytes of decompressed content.
+ * It looks like it's working and there is always debug 16.
+ * - Log the content size after decompression in decompress_iob()
+ * instead of pcrs_filter_response().
+ *
+ * Revision 1.93 2007/03/20 15:21:44 fabiankeil
+ * - Use dedicated header filter actions instead of abusing "filter".
+ * Replace "filter-client-headers" and "filter-client-headers"
+ * with "server-header-filter" and "client-header-filter".
+ * - Remove filter_client_header() and filter_client_header(),
+ * filter_header() now checks the shiny new
+ * CSP_FLAG_CLIENT_HEADER_PARSING_DONE flag instead.
+ *
* Revision 1.92 2007/03/05 13:25:32 fabiankeil
* - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
* - Handle "Cookie:" and "Connection:" headers a bit smarter
#include <ctype.h>
#include <assert.h>
#include <string.h>
+
+#ifdef __GLIBC__
+/*
+ * Convince GNU's libc to provide a strptime prototype.
+ */
+#define __USE_XOPEN
+#endif /*__GLIBC__ */
#include <time.h>
#ifdef FEATURE_ZLIB
char *cur; /* Current iob position (to keep the original
* iob->cur unmodified if we return early) */
size_t bufsize; /* allocated size of the new buffer */
+ size_t old_size; /* Content size before decompression */
size_t skip_size; /* Number of bytes at the beginning of the iob
that we should NOT decompress. */
int status; /* return status of the inflate() call */
bufsize = csp->iob->size;
skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
+ old_size = (size_t)(csp->iob->eod - csp->iob->cur);
cur = csp->iob->cur;
skip_bytes = *cur++;
skip_bytes = *cur++ << 8;
- assert(skip_bytes == *csp->iob->cur-2 + ((*csp->iob->cur-1) << 8));
+ assert(skip_bytes == *csp->iob->cur - 2 + ((*csp->iob->cur - 1) << 8));
/*
* The number of bytes to skip should be positive
/* Skip the filename if necessary. */
if (flags & 0x08)
{
- /* A null-terminated string follows. */
- while (*cur++);
+ /* A null-terminated string is supposed to follow. */
+ while (*cur++ && (cur < csp->iob->eod));
+
}
/* Skip the comment if necessary. */
if (flags & 0x10)
{
- while (*cur++);
+ /* A null-terminated string is supposed to follow. */
+ while (*cur++ && (cur < csp->iob->eod));
}
/* Skip the CRC if necessary. */
{
cur += 2;
}
+
+ if (cur >= csp->iob->eod)
+ {
+ /*
+ * If the current position pointer reached or passed
+ * the buffer end, we were obviously tricked to skip
+ * too much.
+ */
+ log_error (LOG_LEVEL_ERROR,
+ "Malformed gzip header detected. Aborting decompression.");
+ return JB_ERR_COMPRESS;
+ }
}
}
else if (csp->content_type & CT_DEFLATE)
&& (csp->iob->cur <= csp->iob->eod)
&& (csp->iob->eod <= csp->iob->buf + csp->iob->size))
{
- char t = csp->iob->cur[100];
- csp->iob->cur[100] = '\0';
- /*
- * XXX: The debug level should be lowered
- * before the next stable release.
- */
- log_error(LOG_LEVEL_INFO, "Sucessfully decompressed: %s", csp->iob->cur);
- csp->iob->cur[100] = t;
- return JB_ERR_OK;
+ const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
+ if (new_size > 0)
+ {
+ log_error(LOG_LEVEL_RE_FILTER,
+ "Decompression successful. Old size: %d, new size: %d.",
+ old_size, new_size);
+ }
+ else
+ {
+ /* zlib thinks this is OK, so lets do the same. */
+ log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
+ }
}
else
{
return JB_ERR_COMPRESS;
}
+ return JB_ERR_OK;
+
}
#endif /* defined(FEATURE_ZLIB) */
* Body is compressed, turn off pcrs and gif filtering.
*/
csp->content_type |= CT_TABOO;
+
+ /*
+ * Log a warning if the user expects the content to be filtered.
+ */
+ if ((csp->rlist != NULL) &&
+ (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
+ {
+ log_error(LOG_LEVEL_INFO,
+ "Compressed content detected, content filtering disabled. "
+ "Consider recompiling Privoxy with zlib support or "
+ "enable the prevent-compression action.");
+ }
}
-#endif /* !defined(FEATURE_ZLIB) */
+#endif /* defined(FEATURE_ZLIB) */
return JB_ERR_OK;
}
else if (0 == strcmpic(newval, "randomize"))
{
+ const char *header_time = *header + sizeof("Last-Modified:");
+
log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
now = time(NULL);
#ifdef HAVE_GMTIME_R
#else
timeptr = gmtime(&now);
#endif
- if ((timeptr = parse_header_time(*header, &last_modified)) == NULL)
+ if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
{
- log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s (crunching!)", *header);
+ log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
freez(*header);
}
else
}
else /* add random value */
{
- if ((timeptr = parse_header_time(*header, &tm)) == NULL)
+ const char *header_time = *header + sizeof("If-Modified-Since:");
+
+ if (JB_ERR_OK != parse_header_time(header_time, &tm))
{
- log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s (crunching!)", *header);
+ log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
freez(*header);
}
else
time_t now;
time_t cookie_time;
struct tm tm_now;
- struct tm tm_cookie;
time(&now);
#ifdef FEATURE_COOKIE_JAR
if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
{
- log_error(LOG_LEVEL_HEADER, "Crunched incoming cookie -- yum!");
- return crumble(csp, header);
+ log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
+ freez(*header);
}
else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
{
*/
if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
{
- char *match;
- const char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
- memset(&tm_cookie, 0, sizeof(tm_cookie));
- /*
- * Try the valid time formats we know about.
- *
- * XXX: This should be moved to parse_header_time().
- *
- * XXX: Maybe the log messages should be removed
- * for the next stable release. They just exist to
- * see which time format gets the most hits and
- * should be checked for first.
- */
- if (NULL != (match = strptime(expiration_date, "%a, %e-%b-%y %H:%M:%S ", &tm_cookie)))
- {
- /* 22-Feb-2008 12:01:18 GMT */
- log_error(LOG_LEVEL_HEADER,
- "cookie \'%s\' send by %s appears to be using time format 1.",
- *header, csp->http->url);
- }
- else if (NULL != (match = strptime(expiration_date, "%A, %e-%b-%Y %H:%M:%S ", &tm_cookie)))
- {
- /* Tue, 02-Jun-2037 20:00:00 GMT */
- log_error(LOG_LEVEL_HEADER,
- "cookie \'%s\' send by %s appears to be using time format 2.",
- *header, csp->http->url);
- }
- else if (NULL != (match = strptime(expiration_date, "%a, %e-%b-%Y %H:%M:%S ", &tm_cookie)))
- {
- /* Tuesday, 02-Jun-2037 20:00:00 GMT */
- /*
- * On FreeBSD this is never reached because it's handled
- * by "format 2" as well. I am, however, not sure if all
- * strptime() implementations behave that way.
- */
- log_error(LOG_LEVEL_HEADER,
- "cookie \'%s\' send by %s appears to be using time format 3.",
- *header, csp->http->url);
- }
- else if (NULL != (match = strptime(expiration_date, "%a, %e %b %Y %H:%M:%S ", &tm_cookie)))
- {
- /* Fri, 22 Feb 2008 19:20:05 GMT */
- log_error(LOG_LEVEL_HEADER,
- "cookie \'%s\' send by %s appears to be using time format 4.",
- *header, csp->http->url);
- }
- else if (NULL != (match = strptime(expiration_date, "%A %b %e %H:%M:%S %Y", &tm_cookie)))
- {
- /* Thu Mar 08 23:00:00 2007 GMT */
- log_error(LOG_LEVEL_HEADER,
- "cookie \'%s\' send by %s appears to be using time format 5.",
- *header, csp->http->url);
- }
+ char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
- /* Did any of them match? */
- if (NULL == match)
+ /* Did we detect the date properly? */
+ if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
{
/*
* Nope, treat it as if it was still valid.
* anyway, which in many cases will be shorter
* than a browser session.
*/
- cookie_time = timegm(&tm_cookie);
if (cookie_time - now < 0)
{
log_error(LOG_LEVEL_HEADER,
*
* Function : parse_header_time
*
- * Description : Transforms time inside a HTTP header into
- * the usual time format.
+ * Description : Parses time formats used in HTTP header strings
+ * to get the numerical respresentation.
*
* Parameters :
- * 1 : header = header to parse
- * 2 : tm = storage for the resulting time in seconds
+ * 1 : header_time = HTTP header time as string.
+ * 2 : result = storage for header_time in seconds
*
- * Returns : Time struct containing the header time, or
- * NULL in case of a parsing problems.
+ * Returns : JB_ERR_OK if the time format was recognized, or
+ * JB_ERR_PARSE otherwise.
*
*********************************************************************/
-struct tm *parse_header_time(char *header, time_t *tm) {
-
- char * timestring;
+jb_err parse_header_time(const char *header_time, time_t *result)
+{
struct tm gmt;
- struct tm * timeptr;
/*
- * Initializing gmt to prevent time zone offsets.
+ * Zero out gmt to prevent time zone offsets.
*
* While this is only necessary on some platforms
* (mingw32 for example), I don't know how to
* detect these automatically and doing it everywhere
* shouldn't hurt.
*/
- time(tm);
-#ifdef HAVE_LOCALTIME_R
- gmt = *localtime_r(tm, &gmt);
-#elif FEATURE_PTHREAD
- pthread_mutex_lock(&localtime_mutex);
- gmt = *localtime(tm);
- pthread_mutex_unlock(&localtime_mutex);
-#else
- gmt = *localtime(tm);
-#endif
-
- /* Skipping header name */
- timestring = strstr(header, ": ");
- if (strptime(timestring, ": %a, %d %b %Y %H:%M:%S", &gmt) == NULL)
+ memset(&gmt, 0, sizeof(gmt));
+
+ /* Tue, 02 Jun 2037 20:00:00 */
+ if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
+ /* Tue, 02-Jun-2037 20:00:00 */
+ && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
+ /* Tue, 02-Jun-37 20:00:00 */
+ && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
+ /* Tuesday, 02-Jun-2037 20:00:00 */
+ && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
+ /* Tuesday Jun 02 20:00:00 2037 */
+ && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
{
- timeptr = NULL;
- }
- else
- {
- *tm = timegm(&gmt);
- timeptr = &gmt;
+ return JB_ERR_PARSE;
}
- return(timeptr);
+
+ *result = timegm(&gmt);
+
+ return JB_ERR_OK;
}