htp_status_t htp_tx_state_response_headers(htp_tx_t *tx) {
if (tx == NULL) return HTP_ERROR;
// Check for compression.
// Determine content encoding.
tx->response_content_encoding = HTP_COMPRESSION_NONE;
htp_header_t *ce = htp_table_get_c(tx->response_headers, "content-encoding");
if (ce != NULL) {
if ((bstr_cmp_c(ce->value, "gzip") == 0) || (bstr_cmp_c(ce->value, "x-gzip") == 0)) {
tx->response_content_encoding = HTP_COMPRESSION_GZIP;
} else if ((bstr_cmp_c(ce->value, "deflate") == 0) || (bstr_cmp_c(ce->value, "x-deflate") == 0)) {
tx->response_content_encoding = HTP_COMPRESSION_DEFLATE;
}
}
// Configure decompression, if enabled in the configuration.
if (tx->connp->cfg->response_decompression_enabled) {
tx->response_content_encoding_processing = tx->response_content_encoding;
} else {
tx->response_content_encoding_processing = HTP_COMPRESSION_NONE;
}
// Finalize sending raw header data.
htp_status_t rc = htp_connp_res_receiver_finalize_clear(tx->connp);
if (rc != HTP_OK) return rc;
// Run hook RESPONSE_HEADERS.
rc = htp_hook_run_all(tx->connp->cfg->hook_response_headers, tx);
if (rc != HTP_OK) return rc;
// Initialize the decompression engine as necessary. We can deal with three
// scenarios:
//
// 1. Decompression is enabled, compression indicated in headers, and we decompress.
//
// 2. As above, but the user disables decompression by setting response_content_encoding
// to COMPRESSION_NONE.
//
// 3. Decompression is disabled and we do not attempt to enable it, but the user
// forces decompression by setting response_content_encoding to one of the
// supported algorithms.
if ((tx->response_content_encoding_processing == HTP_COMPRESSION_GZIP) || (tx->response_content_encoding_processing == HTP_COMPRESSION_DEFLATE)) {
if (tx->connp->out_decompressor != NULL) {
tx->connp->out_decompressor->destroy(tx->connp->out_decompressor);
tx->connp->out_decompressor = NULL;
}
tx->connp->out_decompressor = (htp_decompressor_t *) htp_gzip_decompressor_create(tx->connp,
tx->response_content_encoding_processing);
if (tx->connp->out_decompressor == NULL) return HTP_ERROR;
tx->connp->out_decompressor->callback = htp_tx_res_process_body_data_decompressor_callback;
} else if (tx->response_content_encoding_processing != HTP_COMPRESSION_NONE) {
return HTP_ERROR;
}
return HTP_OK;
}
TEST(BstrTest, CmpC) {
bstr *p1;
p1 = bstr_dup_c("arfarf");
EXPECT_EQ(0, bstr_cmp_c(p1, "arfarf"));
EXPECT_EQ(-1, bstr_cmp_c(p1, "arfarf2"));
EXPECT_EQ(1, bstr_cmp_c(p1, "arf"));
EXPECT_EQ(-1, bstr_cmp_c(p1, "not equal"));
bstr_free(p1);
}
TEST(BstrTest, ToLowercase) {
bstr *p1;
bstr *p2;
p1 = bstr_dup_c("aRf3ArF");
p2 = bstr_to_lowercase(p1);
EXPECT_EQ(p1, p2);
EXPECT_EQ(1, bstr_cmp_c(p1, "aRf3ArF"));
EXPECT_EQ(0, bstr_cmp_c(p1, "arf3arf"));
bstr_free(p1);
}
/**
* Determine if the request has a URLENCODED body, then
* create and attach the URLENCODED parser if it does.
*/
int htp_ch_urlencoded_callback_request_headers(htp_connp_t *connp) {
// Check the request content type to see if it matches our MIME type
if ((connp->in_tx->request_content_type == NULL) || (bstr_cmp_c(connp->in_tx->request_content_type, HTP_URLENCODED_MIME_TYPE) != 0)) {
#ifdef HTP_DEBUG
fprintf(stderr, "htp_ch_urlencoded_callback_request_headers: Body not URLENCODED\n");
#endif
return HOOK_OK;
}
#ifdef HTP_DEBUG
fprintf(stderr, "htp_ch_urlencoded_callback_request_headers: Parsing URLENCODED body\n");
#endif
// Create parser instance
connp->in_tx->request_urlenp_body = htp_urlenp_create(connp->in_tx);
if (connp->in_tx->request_urlenp_body == NULL) {
return HOOK_ERROR;
}
// Register request body data callbacks
htp_tx_register_request_body_data(connp->in_tx, htp_ch_urlencoded_callback_request_body_data);
return HOOK_OK;
}
bool bstr_equal_c(const bstr *b, const char *c) {
if ((c == NULL) || (b == NULL)) {
return (c == NULL) && (b == NULL);
} else {
return (0 == bstr_cmp_c(b, c));
}
}
TEST(BstrTest, AdjustLen) {
bstr *p1 = bstr_dup_c("abcdef");
bstr_adjust_len(p1, 3);
EXPECT_EQ(3, bstr_len(p1));
EXPECT_EQ(0, bstr_cmp_c(p1,"abc"));
bstr_free(p1);
}
TEST(BstrTest, AddMem) {
bstr *p1;
bstr *p2;
p1 = bstr_dup_c("testtest");
p2 = bstr_add_mem(p1, "12345678", 4);
EXPECT_EQ(0, bstr_cmp_c(p2, "testtest1234"));
bstr_free(p2);
}
TEST(BstrTest, Chop) {
bstr *p1 = bstr_dup_c("abcdef");
bstr *p2 = bstr_alloc(10);
bstr_chop(p1);
EXPECT_EQ(0, bstr_cmp_c(p1,"abcde"));
bstr_chop(p2);
EXPECT_EQ(0, bstr_len(p2));
bstr_free(p1);
bstr_free(p2);
}
TEST(BstrTest, AddMemNoex) {
bstr *p1;
bstr *p2;
p1 = bstr_alloc(10);
p1 = bstr_add_c(p1, "12345");
p2 = bstr_add_mem_noex(p1,"abcdefghijklmnop",6);
EXPECT_EQ(p1,p2);
EXPECT_EQ(0,bstr_cmp_c(p2,"12345abcde"));
bstr_free(p1);
}
TEST(BstrTest, AddNoex) {
bstr *p1;
bstr *p2;
bstr *p3;
p1 = bstr_alloc(10);
p1 = bstr_add_c(p1, "12345");
p2 = bstr_dup_c("abcdef");
p3 = bstr_add_noex(p1,p2);
EXPECT_EQ(p1,p3);
EXPECT_EQ(0,bstr_cmp_c(p3,"12345abcde"));
bstr_free(p1);
bstr_free(p2);
}
TEST(BstrTest, Add) {
bstr *src1;
bstr *src2;
bstr *dest;
src1 = bstr_dup_c("testtest");
src2 = bstr_dup_c("0123456789abcdefghijklmnopqrstuvwxyz");
dest = bstr_add(src1, src2);
EXPECT_EQ(0, bstr_cmp_c(dest, "testtest0123456789abcdefghijklmnopqrstuvwxyz"));
// src1 is either invalid or the same as dest after bstr_add
bstr_free(src2);
bstr_free(dest);
}
/**
* Inspect request headers and register the MULTIPART request data hook
* if it contains a multipart/form-data body.
*
* @param connp
*/
int htp_ch_multipart_callback_request_headers(htp_connp_t *connp) {
// Check the request content type to see if it matches our MIME type
if ((connp->in_tx->request_content_type == NULL) || (bstr_cmp_c(connp->in_tx->request_content_type, HTP_MULTIPART_MIME_TYPE) != 0)) {
#ifdef HTP_DEBUG
fprintf(stderr, "htp_ch_multipart_callback_request_headers: Body not MULTIPART\n");
#endif
return HOOK_OK;
}
#ifdef HTP_DEBUG
fprintf(stderr, "htp_ch_multipart_callback_request_headers: Parsing MULTIPART body\n");
#endif
htp_header_t *ct = table_get_c(connp->in_tx->request_headers, "content-type");
if (ct == NULL) return HOOK_OK;
char *boundary = NULL;
int rc = htp_mpartp_extract_boundary(ct->value, &boundary);
if (rc != HTP_OK) {
// TODO Invalid boundary
return HOOK_OK;
}
// Create parser instance
connp->in_tx->request_mpartp = htp_mpartp_create(connp->cfg, boundary);
if (connp->in_tx->request_mpartp == NULL) {
free(boundary);
return HOOK_ERROR;
}
if (connp->cfg->extract_request_files) {
connp->in_tx->request_mpartp->extract_files = 1;
connp->in_tx->request_mpartp->extract_dir = connp->cfg->tmpdir;
}
free(boundary);
// Register request body data callbacks
htp_tx_register_request_body_data(connp->in_tx, htp_ch_multipart_callback_request_body_data);
return HOOK_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;
}
TEST(Base64, Decode) {
const char *input ="dGhpcyBpcyBhIHRlc3QuLg==";
bstr *out = htp_base64_decode_mem(input, strlen(input));
EXPECT_EQ(0, bstr_cmp_c(out, "this is a test.."));
bstr_free(out);
}
/**
* 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;
}
/**
* Parses request line.
*
* @param connp
* @returns HTP_OK on state change, HTTP_ERROR on error, or HTP_DATA when more data is needed.
*/
int htp_connp_REQ_LINE(htp_connp_t *connp) {
for (;;) {
// Get one byte
IN_COPY_BYTE_OR_RETURN(connp);
// Keep track of NUL bytes
if (connp->in_next_byte == 0) {
// Remember how many NULs there were
connp->in_tx->request_line_nul++;
// Store the offset of the first NUL byte
if (connp->in_tx->request_line_nul_offset == -1) {
connp->in_tx->request_line_nul_offset = connp->in_line_len;
}
}
// Have we reached the end of the line?
if (connp->in_next_byte == LF) {
#ifdef HTP_DEBUG
fprint_raw_data(stderr, __FUNCTION__, connp->in_line, connp->in_line_len);
#endif
// Is this a line that should be ignored?
if (htp_connp_is_line_ignorable(connp, connp->in_line, connp->in_line_len)) {
// We have an empty/whitespace line, which we'll note, ignore and move on
connp->in_tx->request_ignored_lines++;
// TODO How many empty lines are we willing to accept?
// Start again
connp->in_line_len = 0;
return HTP_OK;
}
// Process request line
connp->in_tx->request_line_raw = bstr_dup_mem((char *) connp->in_line, connp->in_line_len);
if (connp->in_tx->request_line_raw == NULL) {
return HTP_ERROR;
}
/// @todo Would be nice to reference request_line_raw data
htp_chomp(connp->in_line, &connp->in_line_len);
connp->in_tx->request_line = bstr_dup_ex(connp->in_tx->request_line_raw, 0, connp->in_line_len);
if (connp->in_tx->request_line == NULL) {
return HTP_ERROR;
}
// Parse request line
if (connp->cfg->parse_request_line(connp) != HTP_OK) {
// Note: downstream responsible for error logging
return HTP_ERROR;
}
if (connp->in_tx->request_method_number == M_CONNECT) {
// Parse authority
if (htp_parse_authority(connp, connp->in_tx->request_uri, &(connp->in_tx->parsed_uri_incomplete)) != HTP_OK) {
// Note: downstream responsible for error logging
return HTP_ERROR;
}
} else {
// Parse the request URI
if (htp_parse_uri(connp->in_tx->request_uri, &(connp->in_tx->parsed_uri_incomplete)) != HTP_OK) {
// Note: downstream responsible for error logging
return HTP_ERROR;
}
// Keep the original URI components, but
// create a copy which we can normalize and use internally
if (htp_normalize_parsed_uri(connp, connp->in_tx->parsed_uri_incomplete, connp->in_tx->parsed_uri)) {
// Note: downstream responsible for error logging
return HTP_ERROR;
}
// Run hook REQUEST_URI_NORMALIZE
int rc = hook_run_all(connp->cfg->hook_request_uri_normalize, 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;
}
}
// Now is a good time to generate request_uri_normalized, before we finalize
// parsed_uri (and lose the information which parts were provided in the request and
// which parts we added).
if (connp->cfg->generate_request_uri_normalized) {
connp->in_tx->request_uri_normalized = htp_unparse_uri_noencode(connp->in_tx->parsed_uri);
if (connp->in_tx->request_uri_normalized == NULL) {
// There's no sense in logging anything on a memory allocation failure
return HTP_ERROR;
}
#ifdef HTP_DEBUG
fprint_raw_data(stderr, "request_uri_normalized",
(unsigned char *) bstr_ptr(connp->in_tx->request_uri_normalized),
bstr_len(connp->in_tx->request_uri_normalized));
#endif
}
// Finalize parsed_uri
// Scheme
if (connp->in_tx->parsed_uri->scheme != NULL) {
if (bstr_cmp_c(connp->in_tx->parsed_uri->scheme, "http") != 0) {
// TODO Invalid scheme
}
} else {
connp->in_tx->parsed_uri->scheme = bstr_dup_c("http");
if (connp->in_tx->parsed_uri->scheme == NULL) {
return HTP_ERROR;
}
}
// Port
if (connp->in_tx->parsed_uri->port != NULL) {
if (connp->in_tx->parsed_uri->port_number != -1) {
// Check that the port in the URI is the same
// as the port on which the client is talking
// to the server
if (connp->conn->use_local_port) {
if (connp->in_tx->parsed_uri->port_number != connp->conn->local_port) {
// Incorrect port; use the real port instead
connp->in_tx->parsed_uri->port_number = connp->conn->local_port;
// TODO Log
}
} else {
connp->in_tx->parsed_uri->port_number = connp->conn->remote_port;
}
} else {
// Invalid port; use the real port instead
if (connp->conn->use_local_port) {
connp->in_tx->parsed_uri->port_number = connp->conn->local_port;
} else {
connp->in_tx->parsed_uri->port_number = connp->conn->remote_port;
}
// TODO Log
}
} else {
if (connp->conn->use_local_port) {
connp->in_tx->parsed_uri->port_number = connp->conn->local_port;
} else {
connp->in_tx->parsed_uri->port_number = connp->conn->remote_port;
}
}
// Path
if (connp->in_tx->parsed_uri->path == NULL) {
connp->in_tx->parsed_uri->path = bstr_dup_c("/");
if (connp->in_tx->parsed_uri->path == NULL) {
return HTP_ERROR;
}
}
}
// Run hook REQUEST_LINE
int rc = hook_run_all(connp->cfg->hook_request_line, 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;
}
}
// Clean up.
connp->in_line_len = 0;
// Move on to the next phase.
connp->in_state = htp_connp_REQ_PROTOCOL;
return HTP_OK;
}
}
}
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;
}
htp_status_t htp_tx_state_request_line(htp_tx_t *tx) {
htp_connp_t *connp = tx->connp;
if (connp->in_tx->request_method_number == HTP_M_CONNECT) {
// Parse authority
if (htp_parse_uri_hostport(connp, connp->in_tx->request_uri, &(connp->in_tx->parsed_uri_incomplete)) != HTP_OK) {
// Note: downstream responsible for error logging.
return HTP_ERROR;
}
} else {
// Parse the request URI
if (htp_parse_uri(connp->in_tx->request_uri, &(connp->in_tx->parsed_uri_incomplete)) != HTP_OK) {
// Note: downstream responsible for error logging.
return HTP_ERROR;
}
// Keep the original URI components, but
// create a copy which we can normalize and use internally.
if (htp_normalize_parsed_uri(connp, connp->in_tx->parsed_uri_incomplete, connp->in_tx->parsed_uri) != HTP_OK) {
// Note: downstream responsible for error logging.
return HTP_ERROR;
}
// Run hook REQUEST_URI_NORMALIZE.
int rc = htp_hook_run_all(connp->cfg->hook_request_uri_normalize, connp);
if (rc != HTP_OK) return rc;
// Now is a good time to generate request_uri_normalized, before we finalize
// parsed_uri (and lose the information which parts were provided in the request and
// which parts we added).
if (connp->cfg->generate_request_uri_normalized) {
connp->in_tx->request_uri_normalized = htp_unparse_uri_noencode(connp->in_tx->parsed_uri);
if (connp->in_tx->request_uri_normalized == NULL) return HTP_ERROR;
#ifdef HTP_DEBUG
fprint_raw_data(stderr, "request_uri_normalized",
(unsigned char *) bstr_ptr(connp->in_tx->request_uri_normalized),
bstr_len(connp->in_tx->request_uri_normalized));
#endif
}
// Finalize parsed_uri.
// Scheme.
if (connp->in_tx->parsed_uri->scheme != NULL) {
if (bstr_cmp_c(connp->in_tx->parsed_uri->scheme, "http") != 0) {
// TODO Invalid scheme.
}
} else {
connp->in_tx->parsed_uri->scheme = bstr_dup_c("http");
if (connp->in_tx->parsed_uri->scheme == NULL) {
return HTP_ERROR;
}
}
// Path.
if (connp->in_tx->parsed_uri->path == NULL) {
connp->in_tx->parsed_uri->path = bstr_dup_c("/");
if (connp->in_tx->parsed_uri->path == NULL) {
return HTP_ERROR;
}
}
}
// Run hook REQUEST_LINE.
int rc = htp_hook_run_all(connp->cfg->hook_request_line, connp);
if (rc != HTP_OK) return rc;
// Move on to the next phase.
connp->in_state = htp_connp_REQ_PROTOCOL;
return HTP_OK;
}
/**
* Determines presence (and encoding) of a response body.
*
* @param[in] connp
* @returns HTP_OK on state change, HTP_ERROR on error, or HTP_DATA when more data is needed.
*/
htp_status_t htp_connp_RES_BODY_DETERMINE(htp_connp_t *connp) {
// If the request uses the CONNECT method, then not only are we
// to assume there's no body, but we need to ignore all
// subsequent data in the stream.
if (connp->out_tx->request_method_number == HTP_M_CONNECT) {
if ((connp->out_tx->response_status_number >= 200)
&& (connp->out_tx->response_status_number <= 299)) {
// This is a successful CONNECT stream, which means
// we need to switch into tunnelling mode.
connp->in_status = HTP_STREAM_TUNNEL;
connp->out_status = HTP_STREAM_TUNNEL;
connp->out_state = htp_connp_RES_FINALIZE;
return HTP_OK;
} else {
// This is a failed CONNECT stream, which means that
// we can unblock request parsing
connp->in_status = HTP_STREAM_DATA;
// We are going to continue processing this transaction,
// adding a note for ourselves to stop at the end (because
// we don't want to see the beginning of a new transaction).
connp->out_data_other_at_tx_end = 1;
}
}
// Check for an interim "100 Continue" response. Ignore it if found, and revert back to RES_FIRST_LINE.
if (connp->out_tx->response_status_number == 100) {
if (connp->out_tx->seen_100continue != 0) {
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0, "Already seen 100-Continue.");
return HTP_ERROR;
}
// Ignore any response headers seen so far.
htp_header_t *h = NULL;
for (int i = 0, n = htp_table_size(connp->out_tx->response_headers); i < n; i++) {
h = htp_table_get_index(connp->out_tx->response_headers, i, NULL);
bstr_free(h->name);
bstr_free(h->value);
free(h);
}
htp_table_clear(connp->out_tx->response_headers);
// Expecting to see another response line next.
connp->out_state = htp_connp_RES_LINE;
connp->out_tx->progress = HTP_RESPONSE_LINE;
connp->out_tx->seen_100continue++;
return HTP_OK;
}
// 1. Any response message which MUST NOT include a message-body
// (such as the 1xx, 204, and 304 responses and any response to a HEAD
// request) is always terminated by the first empty line after the
// header fields, regardless of the entity-header fields present in the
// message.
if (((connp->out_tx->response_status_number >= 100) && (connp->out_tx->response_status_number <= 199))
|| (connp->out_tx->response_status_number == 204) || (connp->out_tx->response_status_number == 304)
|| (connp->out_tx->request_method_number == HTP_M_HEAD)) {
// There's no response body
connp->out_tx->response_transfer_coding = HTP_CODING_NO_BODY;
connp->out_state = htp_connp_RES_FINALIZE;
} else {
// We have a response body
htp_header_t *ct = htp_table_get_c(connp->out_tx->response_headers, "content-type");
htp_header_t *cl = htp_table_get_c(connp->out_tx->response_headers, "content-length");
htp_header_t *te = htp_table_get_c(connp->out_tx->response_headers, "transfer-encoding");
if (ct != NULL) {
connp->out_tx->response_content_type = bstr_dup_lower(ct->value);
if (connp->out_tx->response_content_type == NULL) return HTP_ERROR;
// Ignore parameters
unsigned char *data = bstr_ptr(connp->out_tx->response_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_adjust_len(connp->out_tx->response_content_type, newlen);
break;
}
newlen++;
}
}
// 2. If a Transfer-Encoding header field (section 14.40) is present and
// indicates that the "chunked" transfer coding has been applied, then
// the length is defined by the chunked encoding (section 3.6).
if ((te != NULL) && (bstr_cmp_c(te->value, "chunked") == 0)) {
// If the T-E header is present we are going to use it.
connp->out_tx->response_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
connp->out_tx->flags |= HTP_REQUEST_SMUGGLING;
}
connp->out_state = htp_connp_RES_BODY_CHUNKED_LENGTH;
connp->out_tx->progress = HTP_RESPONSE_BODY;
}// 3. If a Content-Length header field (section 14.14) is present, its
// value in bytes represents the length of the message-body.
else if (cl != NULL) {
// We know the exact length
connp->out_tx->response_transfer_coding = HTP_CODING_IDENTITY;
// Check for multiple C-L headers
if (cl->flags & HTP_FIELD_REPEATED) {
connp->out_tx->flags |= HTP_REQUEST_SMUGGLING;
}
// Get body length
connp->out_tx->response_content_length = htp_parse_content_length(cl->value);
if (connp->out_tx->response_content_length < 0) {
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0, "Invalid C-L field in response: %d",
connp->out_tx->response_content_length);
return HTP_ERROR;
} else {
connp->out_content_length = connp->out_tx->response_content_length;
connp->out_body_data_left = connp->out_content_length;
if (connp->out_content_length != 0) {
connp->out_state = htp_connp_RES_BODY_IDENTITY_CL_KNOWN;
connp->out_tx->progress = HTP_RESPONSE_BODY;
} else {
connp->out_state = htp_connp_RES_FINALIZE;
}
}
} else {
// 4. If the message uses the media type "multipart/byteranges", which is
// self-delimiting, then that defines the length. This media type MUST
// NOT be used unless the sender knows that the recipient can parse it;
// the presence in a request of a Range header with multiple byte-range
// specifiers implies that the client can parse multipart/byteranges
// responses.
if (ct != NULL) {
// TODO Handle multipart/byteranges
if (bstr_index_of_c_nocase(ct->value, "multipart/byteranges") != -1) {
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0,
"C-T multipart/byteranges in responses not supported");
return HTP_ERROR;
}
}
// 5. By the server closing the connection. (Closing the connection
// cannot be used to indicate the end of a request body, since that
// would leave no possibility for the server to send back a response.)
connp->out_state = htp_connp_RES_BODY_IDENTITY_STREAM_CLOSE;
connp->out_tx->response_transfer_coding = HTP_CODING_IDENTITY;
connp->out_tx->progress = HTP_RESPONSE_BODY;
connp->out_body_data_left = -1;
}
}
// NOTE We do not need to check for short-style HTTP/0.9 requests here because
// that is done earlier, before response line parsing begins
int rc = htp_tx_state_response_headers(connp->out_tx);
if (rc != HTP_OK) return rc;
return HTP_OK;
}
/**
* Determines presence (and encoding) of a response body.
*
* @param connp
* @returns HTP_OK on state change, HTTP_ERROR on error, or HTP_DATA when more data is needed.
*/
int htp_connp_RES_BODY_DETERMINE(htp_connp_t *connp) {
// If the request uses the CONNECT method, then not only are we
// to assume there's no body, but we need to ignore all
// subsequent data in the stream.
if (connp->out_tx->request_method_number == M_CONNECT) {
if ((connp->out_tx->response_status_number >= 200)
&& (connp->out_tx->response_status_number <= 299))
{
// This is a successful CONNECT stream, which means
// we need to switch into tunnelling mode.
connp->in_status = STREAM_STATE_TUNNEL;
connp->out_status = STREAM_STATE_TUNNEL;
connp->out_state = htp_connp_RES_IDLE;
connp->out_tx->progress = TX_PROGRESS_DONE;
return HTP_OK;
} else {
// This is a failed CONNECT stream, which means that
// we can unblock request parsing
connp->in_status = STREAM_STATE_DATA;
// We are going to continue processing this transaction,
// adding a note for ourselves to stop at the end (because
// we don't want to see the beginning of a new transaction).
connp->out_data_other_at_tx_end = 1;
}
}
// Check for an interim "100 Continue"
// response. Ignore it if found, and revert back to RES_FIRST_LINE.
if (connp->out_tx->response_status_number == 100) {
if (connp->out_tx->seen_100continue != 0) {
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0, "Already seen 100-Continue");
return HTP_ERROR;
}
// Ignore any response headers set
table_clear(connp->out_tx->response_headers);
connp->out_state = htp_connp_RES_LINE;
connp->out_tx->progress = TX_PROGRESS_RES_LINE;
connp->out_tx->seen_100continue++;
return HTP_OK;
}
// Check for compression
if (connp->cfg->response_decompression_enabled) {
htp_header_t *ce = table_get_c(connp->out_tx->response_headers, "content-encoding");
if (ce != NULL) {
if ((bstr_cmp_c(ce->value, "gzip") == 0) || (bstr_cmp_c(ce->value, "x-gzip") == 0)) {
connp->out_tx->response_content_encoding = COMPRESSION_GZIP;
} else if ((bstr_cmp_c(ce->value, "deflate") == 0) || (bstr_cmp_c(ce->value, "x-deflate") == 0)) {
connp->out_tx->response_content_encoding = COMPRESSION_DEFLATE;
}
if (connp->out_tx->response_content_encoding != COMPRESSION_NONE) {
connp->out_decompressor = (htp_decompressor_t *) htp_gzip_decompressor_create(connp,
connp->out_tx->response_content_encoding);
if (connp->out_decompressor != NULL) {
connp->out_decompressor->callback = htp_connp_RES_BODY_DECOMPRESSOR_CALLBACK;
} else {
// No need to do anything; the error will have already
// been reported by the failed decompressor.
}
}
}
}
// 1. Any response message which MUST NOT include a message-body
// (such as the 1xx, 204, and 304 responses and any response to a HEAD
// request) is always terminated by the first empty line after the
// header fields, regardless of the entity-header fields present in the
// message.
if (((connp->out_tx->response_status_number >= 100) && (connp->out_tx->response_status_number <= 199))
|| (connp->out_tx->response_status_number == 204) || (connp->out_tx->response_status_number == 304)
|| (connp->out_tx->request_method_number == M_HEAD)) {
// There's no response body
connp->out_state = htp_connp_RES_IDLE;
} else {
// We have a response body
htp_header_t *cl = table_get_c(connp->out_tx->response_headers, "content-length");
htp_header_t *te = table_get_c(connp->out_tx->response_headers, "transfer-encoding");
// 2. If a Transfer-Encoding header field (section 14.40) is present and
// indicates that the "chunked" transfer coding has been applied, then
// the length is defined by the chunked encoding (section 3.6).
if ((te != NULL) && (bstr_cmp_c(te->value, "chunked") == 0)) {
// If the T-E header is present we are going to use it.
connp->out_tx->response_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->out_tx->flags |= HTP_REQUEST_SMUGGLING;
// TODO
}
connp->out_state = htp_connp_RES_BODY_CHUNKED_LENGTH;
connp->out_tx->progress = TX_PROGRESS_RES_BODY;
}// 3. If a Content-Length header field (section 14.14) is present, its
// value in bytes represents the length of the message-body.
else if (cl != NULL) {
// We know the exact length
connp->out_tx->response_transfer_coding = IDENTITY;
// Check for multiple C-L headers
if (cl->flags & HTP_FIELD_REPEATED) {
connp->out_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 response: %d", i);
return HTP_ERROR;
} else {
connp->out_content_length = i;
connp->out_body_data_left = connp->out_content_length;
if (connp->out_content_length != 0) {
connp->out_state = htp_connp_RES_BODY_IDENTITY;
connp->out_tx->progress = TX_PROGRESS_RES_BODY;
} else {
connp->out_state = htp_connp_RES_IDLE;
connp->out_tx->progress = TX_PROGRESS_DONE;
}
}
} else {
// 4. If the message uses the media type "multipart/byteranges", which is
// self-delimiting, then that defines the length. This media type MUST
// NOT be used unless the sender knows that the recipient can parse it;
// the presence in a request of a Range header with multiple byte-range
// specifiers implies that the client can parse multipart/byteranges
// responses.
htp_header_t *ct = table_get_c(connp->out_tx->response_headers, "content-type");
if (ct != NULL) {
// TODO Handle multipart/byteranges
if (bstr_index_of_c_nocase(ct->value, "multipart/byteranges") != -1) {
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0,
"C-T multipart/byteranges in responses not supported");
return HTP_ERROR;
}
}
// 5. By the server closing the connection. (Closing the connection
// cannot be used to indicate the end of a request body, since that
// would leave no possibility for the server to send back a response.)
connp->out_state = htp_connp_RES_BODY_IDENTITY;
connp->out_tx->progress = TX_PROGRESS_RES_BODY;
}
}
// NOTE We do not need to check for short-style HTTP/0.9 requests here because
// that is done earlier, before response line parsing begins
// Run hook RESPONSE_HEADERS_COMPLETE
int rc = hook_run_all(connp->cfg->hook_response_headers, connp);
if (rc != HOOK_OK) {
htp_log(connp, HTP_LOG_MARK, HTP_LOG_ERROR, 0,
"Response headers callback returned error (%d)", rc);
return HTP_ERROR;
}
return HTP_OK;
}
