static htp_status_t htp_tx_process_request_headers(htp_tx_t *tx) {
// Remember how many header lines there were before trailers.
tx->request_header_lines_no_trailers = htp_list_size(tx->request_header_lines);
// Determine if we have a request body, and how it is packaged.
htp_header_t *cl = htp_table_get_c(tx->request_headers, "content-length");
htp_header_t *te = htp_table_get_c(tx->request_headers, "transfer-encoding");
// Check for the Transfer-Encoding header, which would indicate a chunked request body.
if (te != NULL) {
// Make sure it contains "chunked" only.
if (bstr_cmp_c(te->value, "chunked") != 0) {
// Invalid T-E header value.
tx->flags |= HTP_INVALID_CHUNKING;
htp_log(tx->connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0,
"Invalid T-E value in request");
}
// Chunked encoding is a HTTP/1.1 feature. Check that some other protocol is not
// used. The flag will also be set if the protocol could not be parsed.
//
// TODO IIS 7.0, for example, would ignore the T-E header when it
// it is used with a protocol below HTTP 1.1.
if (tx->request_protocol_number < HTP_PROTOCOL_1_1) {
tx->flags |= HTP_INVALID_CHUNKING;
}
// If the T-E header is present we are going to use it.
tx->request_transfer_coding = HTP_CODING_CHUNKED;
// We are still going to check for the presence of C-L.
if (cl != NULL) {
// This is a violation of the RFC.
tx->flags |= HTP_REQUEST_SMUGGLING;
}
} else if (cl != NULL) {
// We have a request body of known length.
tx->request_transfer_coding = HTP_CODING_IDENTITY;
// Check for a folded C-L header.
if (cl->flags & HTP_FIELD_FOLDED) {
tx->flags |= HTP_REQUEST_SMUGGLING;
}
// Check for multiple C-L headers.
if (cl->flags & HTP_FIELD_REPEATED) {
tx->flags |= HTP_REQUEST_SMUGGLING;
}
// Get body length.
tx->request_content_length = htp_parse_content_length(cl->value);
if (tx->request_content_length < 0) {
htp_log(tx->connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0, "Invalid C-L field in request");
return HTP_ERROR;
}
} else {
// No body.
tx->request_transfer_coding = HTP_CODING_NO_BODY;
}
// Check for PUT requests, which we need to treat as file uploads.
if (tx->request_method_number == HTP_M_PUT) {
if (htp_tx_req_has_body(tx)) {
// Prepare to treat PUT request body as a file.
tx->connp->put_file = calloc(1, sizeof (htp_file_t));
if (tx->connp->put_file == NULL) return HTP_ERROR;
tx->connp->put_file->source = HTP_FILE_PUT;
} else {
// TODO Warn about PUT request without a body.
}
return HTP_OK;
}
// Host resolution
htp_header_t *h = htp_table_get_c(tx->request_headers, "host");
if (h == NULL) {
// No host information in the headers.
// HTTP/1.1 requires host information in the headers.
if (tx->request_protocol_number >= HTP_PROTOCOL_1_1) {
tx->flags |= HTP_HOST_MISSING;
htp_log(tx->connp, HTP_LOG_MARK, HTP_LOG_WARNING, 0,
"Host information in request headers required by HTTP/1.1");
}
} else {
// Host information available in the headers.
bstr *hostname;
int port;
if (htp_parse_hostport(h->value, &hostname, &port, &(tx->flags)) != HTP_OK) return HTP_ERROR;
// Is there host information in the URI?
if (tx->parsed_uri->hostname == NULL) {
// There is no host information in the URI. Place the
// hostname from the headers into the parsed_uri structure.
tx->parsed_uri->hostname = hostname;
tx->parsed_uri->port_number = port;
} else {
if ((bstr_cmp_nocase(hostname, tx->parsed_uri->hostname) != 0) || (port != tx->parsed_uri->port_number)) {
// The host information is different in the
// headers and the URI. The HTTP RFC states that
// we should ignore the header copy.
tx->flags |= HTP_HOST_AMBIGUOUS;
htp_log(tx->connp, HTP_LOG_MARK, HTP_LOG_WARNING, 0, "Host information ambiguous");
}
bstr_free(hostname);
}
}
// Parse the Content-Type header.
htp_header_t *ct = htp_table_get_c(tx->request_headers, "content-type");
if (ct != NULL) {
if (htp_parse_ct_header(ct->value, &tx->request_content_type) != HTP_OK) return HTP_ERROR;
}
// Parse cookies.
if (tx->connp->cfg->parse_request_cookies) {
htp_parse_cookies_v0(tx->connp);
}
// Parse authentication information.
if (tx->connp->cfg->parse_request_http_authentication) {
htp_parse_authorization(tx->connp);
}
// Run hook REQUEST_HEADERS.
int rc = htp_hook_run_all(tx->connp->cfg->hook_request_headers, tx->connp);
if (rc != HTP_OK) return rc;
return HTP_OK;
}
static htp_status_t htp_tx_process_request_headers(htp_tx_t *tx) {
if (tx == NULL) return HTP_ERROR;
// Determine if we have a request body, and how it is packaged.
htp_status_t rc = HTP_OK;
htp_header_t *cl = htp_table_get_c(tx->request_headers, "content-length");
htp_header_t *te = htp_table_get_c(tx->request_headers, "transfer-encoding");
// Check for the Transfer-Encoding header, which would indicate a chunked request body.
if (te != NULL) {
// Make sure it contains "chunked" only.
// TODO The HTTP/1.1 RFC also allows the T-E header to contain "identity", which
// presumably should have the same effect as T-E header absence. However, Apache
// (2.2.22 on Ubuntu 12.04 LTS) instead errors out with "Unknown Transfer-Encoding: identity".
// And it behaves strangely, too, sending a 501 and proceeding to process the request
// (e.g., PHP is run), but without the body. It then closes the connection.
if (bstr_cmp_c(te->value, "chunked") != 0) {
// Invalid T-E header value.
tx->request_transfer_coding = HTP_CODING_INVALID;
tx->flags |= HTP_REQUEST_INVALID_T_E;
tx->flags |= HTP_REQUEST_INVALID;
} else {
// Chunked encoding is a HTTP/1.1 feature, so check that an earlier protocol
// version is not used. The flag will also be set if the protocol could not be parsed.
//
// TODO IIS 7.0, for example, would ignore the T-E header when it
// it is used with a protocol below HTTP 1.1. This should be a
// personality trait.
if (tx->request_protocol_number < HTP_PROTOCOL_1_1) {
tx->flags |= HTP_REQUEST_INVALID_T_E;
tx->flags |= HTP_REQUEST_SMUGGLING;
}
// If the T-E header is present we are going to use it.
tx->request_transfer_coding = HTP_CODING_CHUNKED;
// We are still going to check for the presence of C-L.
if (cl != NULL) {
// According to the HTTP/1.1 RFC (section 4.4):
//
// "The Content-Length header field MUST NOT be sent
// if these two lengths are different (i.e., if a Transfer-Encoding
// header field is present). If a message is received with both a
// Transfer-Encoding header field and a Content-Length header field,
// the latter MUST be ignored."
//
tx->flags |= HTP_REQUEST_SMUGGLING;
}
}
} else if (cl != NULL) {
// Check for a folded C-L header.
if (cl->flags & HTP_FIELD_FOLDED) {
tx->flags |= HTP_REQUEST_SMUGGLING;
}
// Check for multiple C-L headers.
if (cl->flags & HTP_FIELD_REPEATED) {
tx->flags |= HTP_REQUEST_SMUGGLING;
// TODO Personality trait to determine which C-L header to parse.
// At the moment we're parsing the combination of all instances,
// which is bound to fail (because it will contain commas).
}
// Get the body length.
tx->request_content_length = htp_parse_content_length(cl->value);
if (tx->request_content_length < 0) {
tx->request_transfer_coding = HTP_CODING_INVALID;
tx->flags |= HTP_REQUEST_INVALID_C_L;
tx->flags |= HTP_REQUEST_INVALID;
} else {
// We have a request body of known length.
tx->request_transfer_coding = HTP_CODING_IDENTITY;
}
} else {
// No body.
tx->request_transfer_coding = HTP_CODING_NO_BODY;
}
// If we could not determine the correct body handling,
// consider the request invalid.
if (tx->request_transfer_coding == HTP_CODING_UNKNOWN) {
tx->request_transfer_coding = HTP_CODING_INVALID;
tx->flags |= HTP_REQUEST_INVALID;
}
// Check for PUT requests, which we need to treat as file uploads.
if (tx->request_method_number == HTP_M_PUT) {
if (htp_tx_req_has_body(tx)) {
// Prepare to treat PUT request body as a file.
tx->connp->put_file = calloc(1, sizeof (htp_file_t));
if (tx->connp->put_file == NULL) return HTP_ERROR;
tx->connp->put_file->source = HTP_FILE_PUT;
} else {
// TODO Warn about PUT request without a body.
}
return HTP_OK;
}
// Determine hostname.
// Use the hostname from the URI, when available.
if (tx->parsed_uri->hostname != NULL) {
tx->request_hostname = bstr_dup(tx->parsed_uri->hostname);
if (tx->request_hostname == NULL) return HTP_ERROR;
}
tx->request_port_number = tx->parsed_uri->port_number;
// Examine the Host header.
htp_header_t *h = htp_table_get_c(tx->request_headers, "host");
if (h == NULL) {
// No host information in the headers.
// HTTP/1.1 requires host information in the headers.
if (tx->request_protocol_number >= HTP_PROTOCOL_1_1) {
tx->flags |= HTP_HOST_MISSING;
}
} else {
// Host information available in the headers.
bstr *hostname;
int port;
rc = htp_parse_header_hostport(h->value, &hostname, &port, &(tx->flags));
if (rc != HTP_OK) return rc;
// Is there host information in the URI?
if (tx->request_hostname == NULL) {
// There is no host information in the URI. Place the
// hostname from the headers into the parsed_uri structure.
tx->request_hostname = hostname;
tx->request_port_number = port;
} else {
// The host information appears in the URI and in the headers. It's
// OK if both have the same thing, but we want to check for differences.
if ((bstr_cmp_nocase(hostname, tx->request_hostname) != 0) || (port != tx->request_port_number)) {
// The host information is different in the headers and the URI. The
// HTTP RFC states that we should ignore the header copy.
tx->flags |= HTP_HOST_AMBIGUOUS;
}
bstr_free(hostname);
}
}
// Determine Content-Type.
htp_header_t *ct = htp_table_get_c(tx->request_headers, "content-type");
if (ct != NULL) {
rc = htp_parse_ct_header(ct->value, &tx->request_content_type);
if (rc != HTP_OK) return rc;
}
// Parse cookies.
if (tx->connp->cfg->parse_request_cookies) {
rc = htp_parse_cookies_v0(tx->connp);
if (rc != HTP_OK) return rc;
}
// Parse authentication information.
if (tx->connp->cfg->parse_request_auth) {
rc = htp_parse_authorization(tx->connp);
if (rc == HTP_DECLINED) {
// Don't fail the stream if an authorization header is invalid, just set a flag.
tx->flags |= HTP_AUTH_INVALID;
} else {
if (rc != HTP_OK) return rc;
}
}
// Finalize sending raw header data.
rc = htp_connp_req_receiver_finalize_clear(tx->connp);
if (rc != HTP_OK) return rc;
// Run hook REQUEST_HEADERS.
rc = htp_hook_run_all(tx->connp->cfg->hook_request_headers, tx);
if (rc != HTP_OK) return rc;
// We cannot proceed if the request is invalid.
if (tx->flags & HTP_REQUEST_INVALID) {
return HTP_ERROR;
}
return HTP_OK;
}
/**
* Determines presence (and encoding) of a request body.
*
* @param connp
* @returns HTP_OK on state change, HTTP_ERROR on error, or HTP_DATA when more data is needed.
*/
int htp_connp_REQ_BODY_DETERMINE(htp_connp_t *connp) {
htp_header_t *cl = table_get_c(connp->in_tx->request_headers, "content-length");
htp_header_t *te = table_get_c(connp->in_tx->request_headers, "transfer-encoding");
// Check for the Transfer-Encoding header, which
// would indicate a chunked request body
if (te != NULL) {
// Make sure it contains "chunked" only
if (bstr_cmp_c(te->value, "chunked") != 0) {
// Invalid T-E header value
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0,
"Invalid T-E value in request");
}
// Chunked encoding is a HTTP/1.1 feature. Check
// that some other protocol is not used. The flag will
// also be set if the protocol could not be parsed.
//
// TODO IIS 7.0, for example, would ignore the T-E header when it
// it is used with a protocol below HTTP 1.1.
if (connp->in_tx->request_protocol_number < HTTP_1_1) {
connp->in_tx->flags |= HTP_INVALID_CHUNKING;
// TODO Log
}
// If the T-E header is present we are going to use it.
connp->in_tx->request_transfer_coding = CHUNKED;
// We are still going to check for the presence of C-L
if (cl != NULL) {
// This is a violation of the RFC
connp->in_tx->flags |= HTP_REQUEST_SMUGGLING;
// TODO Log
}
connp->in_state = htp_connp_REQ_BODY_CHUNKED_LENGTH;
connp->in_tx->progress = TX_PROGRESS_REQ_BODY;
} else
// Next check for the presence of the Content-Length header
if (cl != NULL) {
// It seems that we have a request body.
connp->in_tx->request_transfer_coding = IDENTITY;
// Check for a folded C-L header
if (cl->flags & HTP_FIELD_FOLDED) {
connp->in_tx->flags |= HTP_REQUEST_SMUGGLING;
// TODO Log
}
// Check for multiple C-L headers
if (cl->flags & HTP_FIELD_REPEATED) {
connp->in_tx->flags |= HTP_REQUEST_SMUGGLING;
// TODO Log
}
// Get body length
int i = htp_parse_content_length(cl->value);
if (i < 0) {
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0, "Invalid C-L field in request");
return HTP_ERROR;
} else {
connp->in_content_length = i;
connp->in_body_data_left = connp->in_content_length;
if (connp->in_content_length != 0) {
connp->in_state = htp_connp_REQ_BODY_IDENTITY;
connp->in_tx->progress = TX_PROGRESS_REQ_BODY;
} else {
connp->in_state = htp_connp_REQ_IDLE;
connp->in_tx->progress = TX_PROGRESS_WAIT;
}
}
} else {
// This request does not have a body, which
// means that we're done with it
connp->in_state = htp_connp_REQ_IDLE;
connp->in_tx->progress = TX_PROGRESS_WAIT;
}
// Check for PUT requests, which we need to treat as file uploads
if (connp->in_tx->request_method_number == M_PUT) {
if (connp->in_tx->connp->in_tx->request_transfer_coding != 0) {
// Prepare to treat PUT request body as a file
connp->put_file = calloc(1, sizeof (htp_file_t));
if (connp->put_file == NULL) return HTP_ERROR;
connp->put_file->source = HTP_FILE_PUT;
} else {
// TODO Warn about PUT request without a body
}
return HTP_OK;
}
// Host resolution
htp_header_t *h = table_get_c(connp->in_tx->request_headers, "host");
if (h == NULL) {
// No host information in the headers
// HTTP/1.1 requires host information in the headers
if (connp->in_tx->request_protocol_number >= HTTP_1_1) {
connp->in_tx->flags |= HTP_HOST_MISSING;
htp_log(connp, HTP_LOG_MARK, HTP_LOG_WARNING, 0,
"Host information in request headers required by HTTP/1.1");
}
} else {
// Host information available in the headers
// Is there host information in the URI?
if (connp->in_tx->parsed_uri->hostname == NULL) {
// There is no host information in the URI. Place the
// hostname from the headers into the parsed_uri structure.
htp_replace_hostname(connp, connp->in_tx->parsed_uri, h->value);
} else if (bstr_cmp_nocase(h->value, connp->in_tx->parsed_uri->hostname) != 0) {
// The host information is different in the
// headers and the URI. The HTTP RFC states that
// we should ignore the headers copy.
connp->in_tx->flags |= HTP_AMBIGUOUS_HOST;
htp_log(connp, HTP_LOG_MARK, HTP_LOG_WARNING, 0, "Host information ambiguous");
}
}
// Parse Content-Type
htp_header_t *ct = table_get_c(connp->in_tx->request_headers, "content-type");
if (ct != NULL) {
connp->in_tx->request_content_type = bstr_dup_lower(ct->value);
if (connp->in_tx->request_content_type == NULL) {
return HTP_ERROR;
}
// Ignore parameters
char *data = bstr_ptr(connp->in_tx->request_content_type);
size_t len = bstr_len(ct->value);
size_t newlen = 0;
while (newlen < len) {
// TODO Some platforms may do things differently here
if (htp_is_space(data[newlen]) || (data[newlen] == ';')) {
bstr_util_adjust_len(connp->in_tx->request_content_type, newlen);
break;
}
newlen++;
}
}
// Parse cookies
if (connp->cfg->parse_request_cookies) {
htp_parse_cookies_v0(connp);
}
// Parse authentication information
if (connp->cfg->parse_request_http_authentication) {
htp_parse_authorization(connp);
}
// Run hook REQUEST_HEADERS
int rc = hook_run_all(connp->cfg->hook_request_headers, connp);
if (rc != HOOK_OK) {
switch (rc) {
case HOOK_STOP:
return HTP_STOP;
case HOOK_ERROR:
case HOOK_DECLINED:
default:
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0,
"Request headers callback returned error (%d)", rc);
return HTP_ERROR;
}
}
return HTP_OK;
}
