-const char parsers_rcs[] = "$Id: parsers.c,v 1.225 2011/07/08 13:30:08 fabiankeil Exp $";
+const char parsers_rcs[] = "$Id: parsers.c,v 1.244 2012/03/09 16:23:50 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
*
* Purpose : Declares functions to parse/crunch headers and pages.
- * Functions declared include:
- * `add_to_iob', `client_cookie_adder', `client_from',
- * `client_referrer', `client_send_cookie', `client_ua',
- * `client_uagent', `client_x_forwarded',
- * `client_x_forwarded_adder', `client_xtra_adder',
- * `content_type', `crumble', `destroy_list', `enlist',
- * `flush_socket', ``get_header', `sed', `filter_header'
- * `server_content_encoding', `server_content_disposition',
- * `server_last_modified', `client_accept_language',
- * `crunch_client_header', `client_if_modified_since',
- * `client_if_none_match', `get_destination_from_headers',
- * `parse_header_time', `decompress_iob' and `server_set_cookie'.
- *
- * Copyright : Written by and Copyright (C) 2001-2009 the
+ *
+ * Copyright : Written by and Copyright (C) 2001-2011 the
* Privoxy team. http://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
const char parsers_h_rcs[] = PARSERS_H_VERSION;
-/* Fix a problem with Solaris. There should be no effect on other
- * platforms.
- * Solaris's isspace() is a macro which uses its argument directly
- * as an array index. Therefore we need to make sure that high-bit
- * characters generate +ve values, and ideally we also want to make
- * the argument match the declared parameter type of "int".
- *
- * Why did they write a character function that can't take a simple
- * "char" argument? Doh!
- */
-#define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
-#define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
-
static char *get_header_line(struct iob *iob);
static jb_err scan_headers(struct client_state *csp);
static jb_err header_tagger(struct client_state *csp, char *header);
{
/** The header prefix to match */
const char *str;
-
+
/** The length of the prefix to match */
const size_t len;
-
+
/** The function to apply to this line */
const parser_func_ptr parser;
};
client_host_adder,
client_x_forwarded_for_adder,
client_xtra_adder,
- /* Temporarily disabled: client_accept_encoding_adder, */
client_connection_header_adder,
NULL
};
{
want *= 2;
}
-
+
if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
{
iob->size = want;
jb_err decompress_iob(struct client_state *csp)
{
char *buf; /* new, uncompressed buffer */
- char *cur; /* Current iob position (to keep the original
+ 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 */
*
* Fortunately, add_to_iob() has thoughtfully null-terminated
* the buffer; we can just increment the end pointer to include
- * the dummy byte.
+ * the dummy byte.
*/
csp->iob->eod++;
}
{
bufsize = csp->config->buffer_limit;
}
-
+
/* Try to allocate the new buffer. */
tmpbuf = realloc(buf, bufsize);
if (NULL == tmpbuf)
csp->iob->cur = csp->iob->buf + skip_size;
csp->iob->eod = (char *)zstr.next_out;
csp->iob->size = bufsize;
-
+
/*
* Make sure the new uncompressed iob obeys some minimal
* consistency conditions.
while (*p != '\0')
{
- if (ijb_isspace(*p) && ijb_isspace(*(p+1)))
+ if (privoxy_isspace(*p) && privoxy_isspace(*(p+1)))
{
char *q = p+1;
- while (ijb_isspace(*q))
+ while (privoxy_isspace(*q))
{
q++;
}
}
p = strchr(header, ':');
- if ((p != NULL) && (p != header) && ijb_isspace(*(p-1)))
+ if ((p != NULL) && (p != header) && privoxy_isspace(*(p-1)))
{
/*
* There's still space before the colon.
* Header spans multiple lines, append the next one.
*/
char *continued_header;
-
+
continued_header = get_header_line(iob);
if ((continued_header == NULL) || (*continued_header == '\0'))
{
* Found: return pointer to start of value
*/
ret = cur_entry->str + length;
- while (*ret && ijb_isspace(*ret)) ret++;
+ while (*ret && privoxy_isspace(*ret)) ret++;
return ret;
}
}
}
- /*
+ /*
* Not found
*/
return NULL;
*
* Function : scan_headers
*
- * Description : Scans headers, applies tags and updates action bits.
+ * Description : Scans headers, applies tags and updates action bits.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
b->name);
continue;
}
-
+
if (!list_contains_item(csp->tags, tag))
{
if (JB_ERR_OK != enlist(csp->tags, tag))
for (job = joblist; NULL != job; job = job->next)
{
matches = pcrs_execute(job, *header, size, &newheader, &size);
- if ( 0 < matches )
+ if (0 < matches)
{
- current_hits += matches;
+ current_hits += matches;
log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
freez(*header);
*header = newheader;
}
- else if ( 0 == matches )
+ else if (0 == matches)
{
/* Filter doesn't change header */
freez(newheader);
* to remove the header. This function frees the
* original string if necessary.
*
- * Returns : JB_ERR_OK on success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * Returns : JB_ERR_OK on success.
*
*********************************************************************/
static jb_err server_connection(struct client_state *csp, char **header)
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
&& !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
#endif
- )
+ )
{
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
{
char *old_header = *header;
- *header = strdup("Connection: close");
- if (header == NULL)
- {
- return JB_ERR_MEMORY;
- }
+ *header = strdup_or_die("Connection: close");
log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
freez(old_header);
}
* Function : client_connection
*
* Description : Makes sure a proper "Connection:" header is
- * set and signals connection_header_adder
+ * set and signals connection_header_adder
* to do nothing.
*
* Parameters :
* to remove the header. This function frees the
* original string if necessary.
*
- * Returns : JB_ERR_OK on success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * Returns : JB_ERR_OK on success.
*
*********************************************************************/
static jb_err client_connection(struct client_state *csp, char **header)
log_error(LOG_LEVEL_HEADER,
"Removing \'%s\' to imply keep-alive.", *header);
freez(*header);
+ /*
+ * While we imply keep-alive to the server,
+ * we have to remember that the client didn't.
+ */
+ csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
}
else
{
char *old_header = *header;
- *header = strdup("Connection: keep-alive");
- if (header == NULL)
- {
- return JB_ERR_MEMORY;
- }
+ *header = strdup_or_die("Connection: keep-alive");
log_error(LOG_LEVEL_HEADER,
"Replaced: \'%s\' with \'%s\'", old_header, *header);
freez(old_header);
{
char *old_header = *header;
- *header = strdup(connection_close);
- if (header == NULL)
- {
- return JB_ERR_MEMORY;
- }
+ *header = strdup_or_die(connection_close);
log_error(LOG_LEVEL_HEADER,
"Replaced: \'%s\' with \'%s\'", old_header, *header);
freez(old_header);
/* Is the current header the lucky one? */
if (strstr(*header, crunch_pattern))
{
- log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
+ log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
freez(*header);
}
}
{
/*
* Make sure the user doesn't accidentally
- * change the content type of binary documents.
+ * change the content type of binary documents.
*/
if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
{
freez(*header);
- *header = strdup("Content-Type: ");
+ *header = strdup_or_die("Content-Type: ");
string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
if (header == NULL)
*
* If FEATURE_ZLIB is enabled and the compression type
* supported, the content is marked for decompression.
- *
+ *
* XXX: Doesn't properly deal with multiple or with
* unsupported but unknown encodings.
* Is case-sensitive but shouldn't be.
return JB_ERR_OK;
}
else
- {
+ {
/*
* Replacing Content-Disposition header
*/
static jb_err server_last_modified(struct client_state *csp, char **header)
{
const char *newval;
- char buf[BUFFER_SIZE];
time_t last_modified;
char newheader[50];
newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
- if (0 == strcmpic(newval, "block") )
+ if (0 == strcmpic(newval, "block"))
{
/*
* Blocking Last-Modified header. Useless but why not.
return JB_ERR_OK;
}
else if (0 == strcmpic(newval, "reset-to-request-time"))
- {
+ {
/*
* Setting Last-Modified Header to now.
*/
+ char buf[30];
get_http_time(0, buf, sizeof(buf));
freez(*header);
*header = strdup("Last-Modified: ");
- string_append(header, buf);
+ string_append(header, buf);
if (*header == NULL)
{
- log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
+ log_error(LOG_LEVEL_HEADER, "Insufficient memory. Last-Modified header got lost, boohoo.");
}
else
{
if (negative_delta)
{
- rtime *= -1;
+ rtime *= -1;
log_error(LOG_LEVEL_HEADER, "Server time in the future.");
}
rtime = pick_from_range(rtime);
if (*header == NULL)
{
log_error(LOG_LEVEL_ERROR, "Insufficient memory, header crunched without replacement.");
- return JB_ERR_MEMORY;
+ return JB_ERR_MEMORY;
}
days = rtime / (3600 * 24);
/* booleans for parameters we have to check multiple times */
int parameter_conditional_block;
int parameter_conditional_forge;
-
+
#ifdef FEATURE_FORCE_LOAD
/*
* Since the referrer can include the prefix even
newval = csp->action->string[ACTION_STRING_LANGUAGE];
- if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
+ if ((newval == NULL) || (0 == strcmpic(newval, "block")))
{
/*
* Blocking Accept-Language header
return JB_ERR_OK;
}
else
- {
+ {
/*
* Replacing Accept-Language header
*/
freez(*header);
*header = strdup("Accept-Language: ");
- string_append(header, newval);
+ string_append(header, newval);
if (*header == NULL)
{
log_error(LOG_LEVEL_ERROR,
- "Insufficient memory. Accept-Language header crunched without replacement.");
+ "Insufficient memory. Accept-Language header crunched without replacement.");
}
else
{
/* Is the current header the lucky one? */
if (strstr(*header, crunch_pattern))
{
- log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
+ log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
freez(*header);
}
}
/*
* Are we blocking the e-mail address?
*/
- if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
+ if ((newval == NULL) || (0 == strcmpic(newval, "block")))
{
log_error(LOG_LEVEL_HEADER, "crunched From!");
return JB_ERR_OK;
return JB_ERR_OK;
}
- if (!csp->http->hostport || (*csp->http->hostport == '*') ||
+ if (!csp->http->hostport || (*csp->http->hostport == '*') ||
*csp->http->hostport == ' ' || *csp->http->hostport == '\0')
{
-
- if (NULL == (p = strdup((*header)+6)))
- {
- return JB_ERR_MEMORY;
- }
+
+ p = strdup_or_die((*header)+6);
chomp(p);
- if (NULL == (q = strdup(p)))
- {
- freez(p);
- return JB_ERR_MEMORY;
- }
+ q = strdup_or_die(p);
freez(csp->http->hostport);
csp->http->hostport = p;
struct tm gmt;
#endif
struct tm *timeptr = NULL;
- time_t tm = 0;
+ time_t tm = 0;
const char *newval;
char * endptr;
-
- if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
+
+ if (0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
{
- /*
+ /*
* The client got an error message because of a temporary problem,
* the problem is gone and the client now tries to revalidate our
* error message on the real server. The revalidation would always
*header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
if (negative_range)
{
- rtime *= -1;
+ rtime *= -1;
}
rtime *= 60;
rtime = pick_from_range(rtime);
if (*header == NULL)
{
log_error(LOG_LEVEL_HEADER, "Insufficient memory, header crunched without replacement.");
- return JB_ERR_MEMORY;
+ return JB_ERR_MEMORY;
}
hours = rtime / 3600;
static jb_err client_if_none_match(struct client_state *csp, char **header)
{
if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
- {
+ {
log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
freez(*header);
}
*********************************************************************/
jb_err client_x_filter(struct client_state *csp, char **header)
{
- if ( 0 == strcmpic(*header, "X-Filter: No"))
+ if (0 == strcmpic(*header, "X-Filter: No"))
{
if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
{
"force-text-mode overruled the client's request to fetch without filtering!");
}
else
- {
+ {
csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
csp->flags |= CSP_FLAG_NO_FILTERING;
log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
freez(*header);
}
}
- return JB_ERR_OK;
+ return JB_ERR_OK;
}
* Function : client_range
*
* Description : Removes Range, Request-Range and If-Range headers if
- * content filtering is enabled. If the client's version
- * of the document has been altered by Privoxy, the server
- * could interpret the range differently than the client
- * intended in which case the user could end up with
- * corrupted content.
+ * content filtering is enabled and the range doesn't
+ * start at byte 0.
+ *
+ * If the client's version of the document has been
+ * altered by Privoxy, the server could interpret the
+ * range differently than the client intended in which
+ * case the user could end up with corrupted content.
+ *
+ * If the range starts at byte 0 this isn't an issue
+ * so the header can pass. Partial requests like this
+ * are used to render preview images for videos without
+ * downloading the whole video.
+ *
+ * While HTTP doesn't require that range requests are
+ * honoured and the client could simply abort the download
+ * after receiving a sufficient amount of data, various
+ * clients don't handle complete responses to range
+ * requests gracefully and emit misleading error messages
+ * instead.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
*********************************************************************/
static jb_err client_range(struct client_state *csp, char **header)
{
- if (content_filters_enabled(csp->action))
+ if (content_filters_enabled(csp->action)
+ && (0 != strncmpic(strstr(*header, ":"), ": bytes=0-", 10)))
{
log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
" Crunching: \'%s\' to prevent range-mismatch problems.", *header);
freez(*header);
}
- return JB_ERR_OK;
+ return JB_ERR_OK;
}
/* the following functions add headers directly to the header list */
return JB_ERR_OK;
}
- if ( !csp->http->hostport || !*(csp->http->hostport))
+ if (!csp->http->hostport || !*(csp->http->hostport))
{
/* XXX: When does this happen and why is it OK? */
log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
/*
* remove 'user:pass@' from 'proto://user:pass@host'
*/
- if ( (p = strchr( csp->http->hostport, '@')) != NULL )
+ if ((p = strchr( csp->http->hostport, '@')) != NULL)
{
p++;
}
}
-#if 0
-/*********************************************************************
- *
- * Function : client_accept_encoding_adder
- *
- * Description : Add an Accept-Encoding header to the client's request
- * that disables compression if the action applies, and
- * the header is not already there. Called from `sed'.
- * Note: For HTTP/1.0, the absence of the header is enough.
- *
- * Parameters :
- * 1 : csp = Current client state (buffers, headers, etc...)
- *
- * Returns : JB_ERR_OK on success, or
- * JB_ERR_MEMORY on out-of-memory error.
- *
- *********************************************************************/
-static jb_err client_accept_encoding_adder(struct client_state *csp)
-{
- if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
- && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
- {
- return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
- }
-
- return JB_ERR_OK;
-}
-#endif
-
-
/*********************************************************************
*
* Function : client_xtra_adder
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
&& !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
#endif
- )
+ )
{
log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
"without Connection header implies keep-alive.");
static jb_err server_set_cookie(struct client_state *csp, char **header)
{
time_t now;
- time_t cookie_time;
+ time_t cookie_time;
time(&now);
cur_tag = *header + 11;
/* skip whitespace between "Set-Cookie:" and value */
- while (*cur_tag && ijb_isspace(*cur_tag))
+ while (*cur_tag && privoxy_isspace(*cur_tag))
{
cur_tag++;
}
next_tag++;
/* skip whitespace ";" and start of tag */
- while (*next_tag && ijb_isspace(*next_tag))
+ while (*next_tag && privoxy_isspace(*next_tag))
{
next_tag++;
}
* to get the numerical respresentation.
*
* Parameters :
- * 1 : header_time = HTTP header time as string.
+ * 1 : header_time = HTTP header time as string.
* 2 : result = storage for header_time in seconds
*
* Returns : JB_ERR_OK if the time format was recognized, or
static jb_err parse_header_time(const char *header_time, time_t *result)
{
struct tm gmt;
-
/*
- * 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.
+ * Checking for two-digit years first in an
+ * attempt to work around GNU libc's strptime()
+ * reporting negative year values when using %Y.
*/
- 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)))
+ static const char * const time_formats[] = {
+ /* Tue, 02-Jun-37 20:00:00 */
+ "%a, %d-%b-%y %H:%M:%S",
+ /* Tue, 02 Jun 2037 20:00:00 */
+ "%a, %d %b %Y %H:%M:%S",
+ /* Tue, 02-Jun-2037 20:00:00 */
+ "%a, %d-%b-%Y %H:%M:%S",
+ /* Tuesday, 02-Jun-2037 20:00:00 */
+ "%A, %d-%b-%Y %H:%M:%S",
+ /* Tuesday Jun 02 20:00:00 2037 */
+ "%A %b %d %H:%M:%S %Y"
+ };
+ unsigned int i;
+
+ for (i = 0; i < SZ(time_formats); i++)
{
- return JB_ERR_PARSE;
- }
+ /*
+ * Zero out gmt to prevent time zone offsets.
+ * Documented to be required for GNU libc.
+ */
+ memset(&gmt, 0, sizeof(gmt));
- *result = timegm(&gmt);
+ if (NULL != strptime(header_time, time_formats[i], &gmt))
+ {
+ /* Sanity check for GNU libc. */
+ if (gmt.tm_year < 0)
+ {
+ log_error(LOG_LEVEL_HEADER,
+ "Failed to parse '%s' using '%s'. Moving on.",
+ header_time, time_formats[i]);
+ continue;
+ }
+ *result = timegm(&gmt);
+ return JB_ERR_OK;
+ }
+ }
- return JB_ERR_OK;
+ return JB_ERR_PARSE;
}
* Parameters :
* 1 : headers = List of headers (one of them hopefully being
* the "Host:" header)
- * 2 : http = storage for the result (host, port and hostport).
+ * 2 : http = storage for the result (host, port and hostport).
*
- * Returns : JB_ERR_MEMORY in case of memory problems,
+ * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
* JB_ERR_PARSE if the host header couldn't be found,
* JB_ERR_OK otherwise.
*
return JB_ERR_PARSE;
}
- p = strdup(host);
- if (NULL == p)
- {
- log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
- return JB_ERR_MEMORY;
- }
+ p = strdup_or_die(host);
chomp(p);
- if (NULL == (q = strdup(p)))
- {
- freez(p);
- log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
- return JB_ERR_MEMORY;
- }
+ q = strdup_or_die(p);
freez(http->hostport);
http->hostport = p;
* Function : create_forged_referrer
*
* Description : Helper for client_referrer to forge a referer as
- * 'http://[hostname:port/' to fool stupid
- * checks for in-site links
+ * 'http://hostname[:port]/' to fool stupid
+ * checks for in-site links
*
* Parameters :
* 1 : header = Pointer to header pointer
static jb_err handle_conditional_hide_referrer_parameter(char **header,
const char *host, const int parameter_conditional_block)
{
- char *referer = strdup(*header);
+ char *referer = strdup_or_die(*header);
const size_t hostlength = strlen(host);
const char *referer_url = NULL;
- if (NULL == referer)
- {
- freez(*header);
- return JB_ERR_MEMORY;
- }
-
/* referer begins with 'Referer: http[s]://' */
if ((hostlength+17) < strlen(referer))
{
projects / privoxy.git / blobdiff