-const char parsers_rcs[] = "$Id: parsers.c,v 1.288 2014/07/25 11:55:27 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
*
* Purpose : Declares functions to parse/crunch headers and pages.
*
- * Copyright : Written by and Copyright (C) 2001-2014 the
- * Privoxy team. http://www.privoxy.org/
+ * Copyright : Written by and Copyright (C) 2001-2020 the
+ * Privoxy team. https://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
* by and Copyright (C) 1997 Anonymous Coders and
#include "strptime.h"
#endif
-const char parsers_h_rcs[] = PARSERS_H_VERSION;
-
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);
static jb_err parse_header_time(const char *header_time, time_t *result);
+static jb_err parse_time_header(const char *header, time_t *result);
static jb_err crumble (struct client_state *csp, char **header);
static jb_err filter_header (struct client_state *csp, char **header);
/*********************************************************************
*
- * Function : flush_socket
+ * Function : flush_iob
*
* Description : Write any pending "buffered" content.
*
* Parameters :
* 1 : fd = file descriptor of the socket to read
* 2 : iob = The I/O buffer to flush, usually csp->iob.
+ * 3 : delay = Number of milliseconds to delay the writes
*
* Returns : On success, the number of bytes written are returned (zero
* indicates nothing was written). On error, -1 is returned,
* file, the results are not portable.
*
*********************************************************************/
-long flush_socket(jb_socket fd, struct iob *iob)
+long flush_iob(jb_socket fd, struct iob *iob, unsigned int delay)
{
long len = iob->eod - iob->cur;
return(0);
}
- if (write_socket(fd, iob->cur, (size_t)len))
+ if (write_socket_delayed(fd, iob->cur, (size_t)len, delay))
{
return(-1);
}
void clear_iob(struct iob *iob)
{
free(iob->buf);
- memset(iob, '\0', sizeof(*iob));;
+ memset(iob, '\0', sizeof(*iob));
}
int status; /* return status of the inflate() call */
z_stream zstr; /* used by calls to zlib */
+#ifdef FUZZ
+ assert(csp->iob->cur - csp->iob->buf >= 0);
+ assert(csp->iob->eod - csp->iob->cur >= 0);
+#else
assert(csp->iob->cur - csp->iob->buf > 0);
assert(csp->iob->eod - csp->iob->cur > 0);
+#endif
bufsize = csp->iob->size;
skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
* This is to protect the parsing of gzipped data,
* but it should(?) be valid for deflated data also.
*/
- log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
+ log_error(LOG_LEVEL_ERROR,
+ "Insufficient data to start decompression. Bytes in buffer: %d",
+ csp->iob->eod - csp->iob->cur);
return JB_ERR_COMPRESS;
}
if (bufsize >= csp->config->buffer_limit)
{
log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
+ freez(buf);
return JB_ERR_MEMORY;
}
* Make sure the new uncompressed iob obeys some minimal
* consistency conditions.
*/
- if ((csp->iob->buf < csp->iob->cur)
+ if ((csp->iob->buf <= csp->iob->cur)
&& (csp->iob->cur <= csp->iob->eod)
&& (csp->iob->eod <= csp->iob->buf + csp->iob->size))
{
*
* Function : normalize_lws
*
- * Description : Reduces unquoted linear whitespace in headers
- * to a single space in accordance with RFC 2616 2.2.
+ * Description : Reduces unquoted linear whitespace in headers to
+ * a single space in accordance with RFC 7230 3.2.4.
* This simplifies parsing and filtering later on.
*
* Parameters :
* server and header filtering.
*
* Returns : JB_ERR_OK in case off success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * JB_ERR_MEMORY on some out-of-memory errors, or
+ * JB_ERR_PARSE in case of fatal parse errors.
*
*********************************************************************/
jb_err sed(struct client_state *csp, int filter_server_headers)
check_negative_tag_patterns(csp, PATTERN_SPEC_NO_REQUEST_TAG_PATTERN);
}
- while ((err == JB_ERR_OK) && (v->str != NULL))
+ while (v->str != NULL)
{
- for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
+ for (p = csp->headers->first; p != NULL; p = p->next)
{
/* Header crunch()ed in previous run? -> ignore */
if (p->str == NULL) continue;
(v->len == CHECK_EVERY_HEADER_REMAINING))
{
err = v->parser(csp, &(p->str));
+ if (err != JB_ERR_OK)
+ {
+ return err;
+ }
}
}
v++;
}
+#ifdef FEATURE_HTTPS_INSPECTION
+/*********************************************************************
+ *
+ * Function : sed_https
+ *
+ * Description : add, delete or modify lines in the HTTPS client
+ * header streams. Wrapper around sed().
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : JB_ERR_OK in case off success, or
+ * JB_ERR_MEMORY on some out-of-memory errors, or
+ * JB_ERR_PARSE in case of fatal parse errors.
+ *
+ *********************************************************************/
+jb_err sed_https(struct client_state *csp)
+{
+ jb_err err;
+ struct list headers;
+
+ /*
+ * Temporarly replace csp->headers with csp->https_headers
+ * to trick sed() into filtering the https headers.
+ */
+ headers.first = csp->headers->first;
+ headers.last = csp->headers->last;
+ csp->headers->first = csp->https_headers->first;
+ csp->headers->last = csp->https_headers->last;
+
+ /*
+ * Start with fresh tags. Already existing tags may
+ * be set again. This is necessary to overrule
+ * URL-based patterns.
+ */
+ destroy_list(csp->tags);
+
+ /*
+ * We want client header filters and taggers
+ * so temporarly remove the flag.
+ */
+ csp->flags &= ~CSP_FLAG_CLIENT_HEADER_PARSING_DONE;
+ err = sed(csp, FILTER_CLIENT_HEADERS);
+ csp->flags |= CSP_FLAG_CLIENT_HEADER_PARSING_DONE;
+
+ csp->headers->first = headers.first;
+ csp->headers->last = headers.last;
+
+ return err;
+}
+#endif /* def FEATURE_HTTPS_INSPECTION */
+
+
/*********************************************************************
*
* Function : update_server_headers
csp->flags |= CSP_FLAG_SERVER_KEEP_ALIVE_TIMEOUT_SET;
}
+ freez(*header);
+
return JB_ERR_OK;
}
static jb_err get_content_length(const char *header_value, unsigned long long *length)
{
#ifdef _WIN32
- assert(sizeof(unsigned long long) > 4);
+#if SIZEOF_LONG_LONG < 8
+#error sizeof(unsigned long long) too small
+#endif
if (1 != sscanf(header_value, "%I64u", length))
#else
if (1 != sscanf(header_value, "%llu", length))
*/
if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
{
+ jb_err err;
freez(*header);
*header = strdup_or_die("Content-Type: ");
- string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
- if (header == NULL)
+ err = string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
+ if (JB_ERR_OK != err)
{
log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
return JB_ERR_MEMORY;
/*
* 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])))
+ if (content_filters_enabled(csp->action))
{
log_error(LOG_LEVEL_INFO,
"SDCH-compressed content detected, content filtering disabled. "
/*
* 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])))
+ if (content_filters_enabled(csp->action))
{
log_error(LOG_LEVEL_INFO,
"Compressed content detected, content filtering disabled. "
}
else if (0 == strcmpic(newval, "randomize"))
{
- const char *header_time = *header + sizeof("Last-Modified:");
-
log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
- if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
+ if (JB_ERR_OK != parse_time_header(*header, &last_modified))
{
- log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
+ log_error(LOG_LEVEL_HEADER,
+ "Couldn't parse time in %s (crunching!)", *header);
freez(*header);
}
else
{
char *p, *q;
+ if (strlen(*header) < 7)
+ {
+ log_error(LOG_LEVEL_HEADER, "Removing empty Host header");
+ freez(*header);
+ return JB_ERR_OK;
+ }
+
if (!csp->http->hostport || (*csp->http->hostport == '*') ||
*csp->http->hostport == ' ' || *csp->http->hostport == '\0')
{
}
else /* add random value */
{
- const char *header_time = *header + sizeof("If-Modified-Since:");
-
- if (JB_ERR_OK != parse_header_time(header_time, &tm))
+ if (JB_ERR_OK != parse_time_header(*header, &tm))
{
- log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
+ log_error(LOG_LEVEL_HEADER,
+ "Couldn't parse time in %s (crunching!)", *header);
freez(*header);
}
else
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.");
- return JB_ERR_OK;
+ log_error(LOG_LEVEL_ERROR, "Destination host unknown.");
+ return JB_ERR_PARSE;
}
/*
* Function : client_connection_header_adder
*
* Description : Adds a proper "Connection:" header to csp->headers
- * unless the header was already present. Called from `sed'.
+ * unless the header was already present or it's a
+ * CONNECT request. Called from `sed'.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
return JB_ERR_OK;
}
+ /*
+ * In case of CONNECT requests "Connection: close" is implied,
+ * but actually setting the header has been reported to cause
+ * problems with some forwarding proxies that close the
+ * connection prematurely.
+ */
+ if (csp->http->ssl != 0)
+ {
+ return JB_ERR_OK;
+ }
+
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
&& !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
- && (csp->http->ssl == 0)
&& !strcmpic(csp->http->ver, "HTTP/1.1"))
{
csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
* is a partial range (HTTP status 206)
* - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
* action applies.
+ * - Normalize the HTTP-version.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
* original string if necessary.
*
* Returns : JB_ERR_OK on success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * JB_ERR_PARSE on fatal parse errors.
*
*********************************************************************/
static jb_err server_http(struct client_state *csp, char **header)
{
- sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
+ char *reason_phrase = NULL;
+ char *new_response_line;
+ char *p;
+ size_t length;
+ unsigned int major_version;
+ unsigned int minor_version;
+
+ /* Get the reason phrase which start after the second whitespace */
+ p = strchr(*header, ' ');
+ if (NULL != p)
+ {
+ p++;
+ reason_phrase = strchr(p, ' ');
+ }
+
+ if (reason_phrase != NULL)
+ {
+ reason_phrase++;
+ }
+ else
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Response line lacks reason phrase: %s", *header);
+ reason_phrase="";
+ }
+
+ if (3 != sscanf(*header, "HTTP/%u.%u %d", &major_version,
+ &minor_version, &(csp->http->status)))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Failed to parse the response line: %s", *header);
+ return JB_ERR_PARSE;
+ }
+
if (csp->http->status == 206)
{
csp->content_type = CT_TABOO;
}
- if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
+ if (major_version != 1 || (minor_version != 0 && minor_version != 1))
{
- /* XXX: Should we do a real validity check here? */
- if (strlen(*header) > 8)
- {
- (*header)[7] = '0';
- log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
- }
- else
- {
- /*
- * XXX: Should we block the request or
- * enlist a valid status code line here?
- */
- log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
- "Downgrading to HTTP/1.0 impossible.");
- }
+ /*
+ * According to RFC 7230 2.6 intermediaries MUST send
+ * their own HTTP-version in forwarded messages.
+ */
+ log_error(LOG_LEVEL_ERROR,
+ "Unsupported HTTP version. Downgrading to 1.1.");
+ major_version = 1;
+ minor_version = 1;
+ }
+
+ if (((csp->action->flags & ACTION_DOWNGRADE) != 0) && (minor_version == 1))
+ {
+ log_error(LOG_LEVEL_HEADER, "Downgrading answer to HTTP/1.0");
+ minor_version = 0;
+ }
+
+ /* Rebuild response line. */
+ length = sizeof("HTTP/1.1 200 ") + strlen(reason_phrase) + 1;
+ new_response_line = malloc_or_die(length);
+
+ snprintf(new_response_line, length, "HTTP/%u.%u %d %s",
+ major_version, minor_version, csp->http->status, reason_phrase);
+
+ if (0 != strcmp(*header, new_response_line))
+ {
+ log_error(LOG_LEVEL_HEADER, "Response line '%s' changed to '%s'",
+ *header, new_response_line);
}
+ freez(*header);
+ *header = new_response_line;
+
return JB_ERR_OK;
}
time_t result2;
tm = gmtime(result);
- strftime(recreated_date, sizeof(recreated_date), time_formats[i], tm);
+ if (!strftime(recreated_date, sizeof(recreated_date),
+ time_formats[i], tm))
+ {
+ log_error(LOG_LEVEL_ERROR, "Failed to recreate date '%s' with '%s'.",
+ header_time, time_formats[i]);
+ continue;
+ }
memset(&gmt, 0, sizeof(gmt));
if (NULL == strptime(recreated_date, time_formats[i], &gmt))
{
}
+/*********************************************************************
+ *
+ * Function : parse_time_header
+ *
+ * Description : Parses the time in an HTTP time header to get
+ * the numerical respresentation.
+ *
+ * Parameters :
+ * 1 : header = HTTP header with a time value
+ * 2 : result = storage for header_time in seconds
+ *
+ * Returns : JB_ERR_OK if the time format was recognized, or
+ * JB_ERR_PARSE otherwise.
+ *
+ *********************************************************************/
+static jb_err parse_time_header(const char *header, time_t *result)
+{
+ const char *header_time;
+
+ header_time = strchr(header, ':');
+
+ /*
+ * Currently this can't happen as all callers are called
+ * through sed() which requires a header name followed by
+ * a colon.
+ */
+ assert(header_time != NULL);
+
+ header_time++;
+ if (*header_time == ' ')
+ {
+ header_time++;
+ }
+
+ return parse_header_time(header_time, result);
+
+}
+
/*********************************************************************
*
}
else
{
- http->port = http->ssl ? 443 : 80;
+ http->port = 80;
}
/* Rebuild request URL */
freez(http->url);
- http->url = strdup(http->ssl ? "https://" : "http://");
+ http->url = strdup("http://");
string_append(&http->url, http->hostport);
string_append(&http->url, http->path);
if (http->url == NULL)
return JB_ERR_MEMORY;
}
- log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
+ log_error(LOG_LEVEL_HEADER,
+ "Destination extracted from \"Host\" header. New request URL: %s",
+ http->url);
+
+ /*
+ * Regenerate request line in "proxy format"
+ * to make rewrites more convenient.
+ */
+ assert(http->cmd != NULL);
+ freez(http->cmd);
+ http->cmd = strdup_or_die(http->gpc);
+ string_append(&http->cmd, " ");
+ string_append(&http->cmd, http->url);
+ string_append(&http->cmd, " ");
+ string_append(&http->cmd, http->ver);
+ if (http->cmd == NULL)
+ {
+ return JB_ERR_MEMORY;
+ }
+
+ return JB_ERR_OK;
+
+}
+
+
+#ifdef FEATURE_HTTPS_INSPECTION
+/*********************************************************************
+ *
+ * Function : get_destination_from_https_headers
+ *
+ * Description : Parse the previously encrypted "Host:" header to
+ * get the request's destination.
+ *
+ * Parameters :
+ * 1 : headers = List of headers (one of them hopefully being
+ * the "Host:" header)
+ * 2 : http = storage for the result (host, port and hostport).
+ *
+ * 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.
+ *
+ *********************************************************************/
+jb_err get_destination_from_https_headers(const struct list *headers, struct http_request *http)
+{
+ char *q;
+ char *p;
+ char *host;
+
+ host = get_header_value(headers, "Host:");
+
+ if (NULL == host)
+ {
+ log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
+ return JB_ERR_PARSE;
+ }
+
+ p = strdup_or_die(host);
+ chomp(p);
+ q = strdup_or_die(p);
+
+ freez(http->hostport);
+ http->hostport = p;
+ freez(http->host);
+ http->host = q;
+ q = strchr(http->host, ':');
+ if (q != NULL)
+ {
+ /* Terminate hostname and evaluate port string */
+ *q++ = '\0';
+ http->port = atoi(q);
+ }
+ else
+ {
+ http->port = 443;
+ }
+
+ /* Rebuild request URL */
+ freez(http->url);
+ http->url = strdup_or_die(http->path);
+
+ log_error(LOG_LEVEL_HEADER,
+ "Destination extracted from \"Host\" header. New request URL: %s",
http->url);
+ /*
+ * Regenerate request line in "proxy format"
+ * to make rewrites more convenient.
+ */
+ assert(http->cmd != NULL);
+ freez(http->cmd);
+ http->cmd = strdup_or_die(http->gpc);
+ string_append(&http->cmd, " ");
+ string_append(&http->cmd, http->url);
+ string_append(&http->cmd, " ");
+ string_append(&http->cmd, http->ver);
+ if (http->cmd == NULL)
+ {
+ return JB_ERR_MEMORY;
+ }
+
return JB_ERR_OK;
}
+#endif /* def FEATURE_HTTPS_INSPECTION */
/*********************************************************************
static void create_content_length_header(unsigned long long content_length,
char *header, size_t buffer_length)
{
+#ifdef _WIN32
+#if SIZEOF_LONG_LONG < 8
+#error sizeof(unsigned long long) too small
+#endif
+ snprintf(header, buffer_length, "Content-Length: %I64u", content_length);
+#else
snprintf(header, buffer_length, "Content-Length: %llu", content_length);
+#endif
}
projects / privoxy.git / blobdiff