php_http_message_parser_state_t php_http_message_parser_parse(php_http_message_parser_t *parser, php_http_buffer_t *buffer, unsigned flags, php_http_message_t **message)
{
char *str = NULL;
size_t len = 0;
size_t cut = 0;
while (buffer->used || !php_http_message_parser_states[php_http_message_parser_state_is(parser)].need_data) {
#if DBG_PARSER
fprintf(stderr, "#MP: %s (f: %u, t:%d, l:%zu)\n",
php_http_message_parser_state_name(php_http_message_parser_state_is(parser)),
flags,
message && *message ? (*message)->type : -1,
buffer->used
);
_dpf(0, buffer->data, buffer->used);
#endif
switch (php_http_message_parser_state_pop(parser))
{
case PHP_HTTP_MESSAGE_PARSER_STATE_FAILURE:
return php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_FAILURE);
case PHP_HTTP_MESSAGE_PARSER_STATE_START:
{
char *ptr = buffer->data;
while (ptr - buffer->data < buffer->used && PHP_HTTP_IS_CTYPE(space, *ptr)) {
++ptr;
}
php_http_buffer_cut(buffer, 0, ptr - buffer->data);
if (buffer->used) {
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_HEADER);
}
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_HEADER:
{
unsigned header_parser_flags = (flags & PHP_HTTP_MESSAGE_PARSER_CLEANUP) ? PHP_HTTP_HEADER_PARSER_CLEANUP : 0;
switch (php_http_header_parser_parse(&parser->header, buffer, header_parser_flags, *message ? &(*message)->hdrs : NULL, (php_http_info_callback_t) php_http_message_info_callback, message)) {
case PHP_HTTP_HEADER_PARSER_STATE_FAILURE:
return PHP_HTTP_MESSAGE_PARSER_STATE_FAILURE;
case PHP_HTTP_HEADER_PARSER_STATE_DONE:
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_HEADER_DONE);
break;
default:
if (buffer->used || !(flags & PHP_HTTP_MESSAGE_PARSER_CLEANUP)) {
return php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_HEADER);
} else {
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_HEADER_DONE);
}
}
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_HEADER_DONE:
{
zval h, *h_ptr, *h_loc = NULL, *h_con = NULL, *h_ce;
zend_bool chunked = 0;
zend_long content_length = -1;
zend_string *content_range = NULL;
/* Content-Range has higher precedence than Content-Length,
* and content-length denotes the original length of the entity,
* so let's *NOT* remove CR/CL, because that would fundamentally
* change the meaning of the whole message
*/
if ((h_ptr = php_http_message_header(*message, ZEND_STRL("Transfer-Encoding")))) {
zend_string *zs = zval_get_string(h_ptr);
chunked = zend_string_equals_literal(zs, "chunked");
zend_string_release(zs);
Z_TRY_ADDREF_P(h_ptr);
zend_hash_str_update(&(*message)->hdrs, "X-Original-Transfer-Encoding", lenof("X-Original-Transfer-Encoding"), h_ptr);
zend_hash_str_del(&(*message)->hdrs, "Transfer-Encoding", lenof("Transfer-Encoding"));
/* reset */
ZVAL_LONG(&h, 0);
zend_hash_str_update(&(*message)->hdrs, "Content-Length", lenof("Content-Length"), &h);
} else if ((h_ptr = php_http_message_header(*message, ZEND_STRL("Content-Length")))) {
content_length = zval_get_long(h_ptr);
Z_TRY_ADDREF_P(h_ptr);
zend_hash_str_update(&(*message)->hdrs, "X-Original-Content-Length", lenof("X-Original-Content-Length"), h_ptr);
}
if ((content_range = php_http_message_header_string(*message, ZEND_STRL("Content-Range")))) {
ZVAL_STR_COPY(&h, content_range);
zend_hash_str_update(&(*message)->hdrs, "Content-Range", lenof("Content-Range"), &h);
}
/* so, if curl sees a 3xx code, a Location header and a Connection:close header
* it decides not to read the response body.
*/
if ((flags & PHP_HTTP_MESSAGE_PARSER_EMPTY_REDIRECTS)
&& (*message)->type == PHP_HTTP_RESPONSE
&& (*message)->http.info.response.code/100 == 3
&& (h_loc = php_http_message_header(*message, ZEND_STRL("Location")))
&& (h_con = php_http_message_header(*message, ZEND_STRL("Connection")))
) {
zend_string *con = zval_get_string(h_con);
if (php_http_match(con->val, "close", PHP_HTTP_MATCH_WORD)) {
zend_string_release(con);
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_DONE);
break;
}
zend_string_release(con);
}
if ((h_ce = php_http_message_header(*message, ZEND_STRL("Content-Encoding")))) {
zend_string *ce = zval_get_string(h_ce);
if (php_http_match(ce->val, "gzip", PHP_HTTP_MATCH_WORD)
|| php_http_match(ce->val, "x-gzip", PHP_HTTP_MATCH_WORD)
|| php_http_match(ce->val, "deflate", PHP_HTTP_MATCH_WORD)
) {
if (parser->inflate) {
php_http_encoding_stream_reset(&parser->inflate);
} else {
parser->inflate = php_http_encoding_stream_init(NULL, php_http_encoding_stream_get_inflate_ops(), 0);
}
Z_TRY_ADDREF_P(h_ce);
zend_hash_str_update(&(*message)->hdrs, "X-Original-Content-Encoding", lenof("X-Original-Content-Encoding"), h_ce);
zend_hash_str_del(&(*message)->hdrs, "Content-Encoding", lenof("Content-Encoding"));
}
zend_string_release(ce);
}
if ((flags & PHP_HTTP_MESSAGE_PARSER_DUMB_BODIES)) {
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DUMB);
} else {
if (chunked) {
parser->dechunk = php_http_encoding_stream_init(parser->dechunk, php_http_encoding_stream_get_dechunk_ops(), 0);
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_BODY_CHUNKED);
break;
}
if (content_range) {
ulong total = 0, start = 0, end = 0;
if (!strncasecmp(content_range->val, "bytes", lenof("bytes"))
&& ( content_range->val[lenof("bytes")] == ':'
|| content_range->val[lenof("bytes")] == ' '
|| content_range->val[lenof("bytes")] == '='
)
) {
char *total_at = NULL, *end_at = NULL;
char *start_at = content_range->val + sizeof("bytes");
start = strtoul(start_at, &end_at, 10);
if (end_at) {
end = strtoul(end_at + 1, &total_at, 10);
if (total_at && strncmp(total_at + 1, "*", 1)) {
total = strtoul(total_at + 1, NULL, 10);
}
if (end >= start && (!total || end <= total)) {
parser->body_length = end + 1 - start;
php_http_message_parser_state_push(parser, 1, !parser->body_length?PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DONE:PHP_HTTP_MESSAGE_PARSER_STATE_BODY_LENGTH);
zend_string_release(content_range);
break;
}
}
}
zend_string_release(content_range);
}
if (content_length >= 0) {
parser->body_length = content_length;
php_http_message_parser_state_push(parser, 1, !parser->body_length?PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DONE:PHP_HTTP_MESSAGE_PARSER_STATE_BODY_LENGTH);
break;
}
if ((*message)->type == PHP_HTTP_REQUEST) {
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_DONE);
} else {
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DUMB);
}
}
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_BODY:
{
if (len) {
if (parser->inflate) {
char *dec_str = NULL;
size_t dec_len;
if (SUCCESS != php_http_encoding_stream_update(parser->inflate, str, len, &dec_str, &dec_len)) {
return php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_FAILURE);
}
if (str != buffer->data) {
PTR_FREE(str);
}
str = dec_str;
len = dec_len;
}
php_stream_write(php_http_message_body_stream((*message)->body), str, len);
}
if (cut) {
php_http_buffer_cut(buffer, 0, cut);
}
if (str != buffer->data) {
PTR_FREE(str);
}
str = NULL;
len = 0;
cut = 0;
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DUMB:
{
str = buffer->data;
len = buffer->used;
cut = len;
php_http_message_parser_state_push(parser, 2, PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DONE, PHP_HTTP_MESSAGE_PARSER_STATE_BODY);
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_BODY_LENGTH:
{
len = MIN(parser->body_length, buffer->used);
str = buffer->data;
cut = len;
parser->body_length -= len;
php_http_message_parser_state_push(parser, 2, !parser->body_length?PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DONE:PHP_HTTP_MESSAGE_PARSER_STATE_BODY_LENGTH, PHP_HTTP_MESSAGE_PARSER_STATE_BODY);
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_BODY_CHUNKED:
{
/*
* - pass available data through the dechunk stream
* - pass decoded data along
* - if stream zeroed:
* Y: - cut processed string out of buffer, but leave length of unprocessed dechunk stream data untouched
* - body done
* N: - parse ahaed
*/
char *dec_str = NULL;
size_t dec_len;
if (SUCCESS != php_http_encoding_stream_update(parser->dechunk, buffer->data, buffer->used, &dec_str, &dec_len)) {
return PHP_HTTP_MESSAGE_PARSER_STATE_FAILURE;
}
str = dec_str;
len = dec_len;
if (php_http_encoding_stream_done(parser->dechunk)) {
cut = buffer->used - PHP_HTTP_BUFFER(parser->dechunk->ctx)->used;
php_http_message_parser_state_push(parser, 2, PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DONE, PHP_HTTP_MESSAGE_PARSER_STATE_BODY);
} else {
cut = buffer->used;
php_http_message_parser_state_push(parser, 2, PHP_HTTP_MESSAGE_PARSER_STATE_BODY_CHUNKED, PHP_HTTP_MESSAGE_PARSER_STATE_BODY);
}
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_BODY_DONE:
{
php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_DONE);
if (parser->dechunk && parser->dechunk->ctx) {
char *dec_str = NULL;
size_t dec_len;
if (SUCCESS != php_http_encoding_stream_finish(parser->dechunk, &dec_str, &dec_len)) {
return php_http_message_parser_state_push(parser, 1, PHP_HTTP_MESSAGE_PARSER_STATE_FAILURE);
}
php_http_encoding_stream_dtor(parser->dechunk);
if (dec_str && dec_len) {
str = dec_str;
len = dec_len;
cut = 0;
php_http_message_parser_state_push(parser, 2, PHP_HTTP_MESSAGE_PARSER_STATE_UPDATE_CL, PHP_HTTP_MESSAGE_PARSER_STATE_BODY);
}
}
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_UPDATE_CL:
{
zval zcl;
ZVAL_LONG(&zcl, php_http_message_body_size((*message)->body));
zend_hash_str_update(&(*message)->hdrs, "Content-Length", lenof("Content-Length"), &zcl);
break;
}
case PHP_HTTP_MESSAGE_PARSER_STATE_DONE:
{
char *ptr = buffer->data;
while (ptr - buffer->data < buffer->used && PHP_HTTP_IS_CTYPE(space, *ptr)) {
++ptr;
}
php_http_buffer_cut(buffer, 0, ptr - buffer->data);
if (!(flags & PHP_HTTP_MESSAGE_PARSER_GREEDY)) {
return PHP_HTTP_MESSAGE_PARSER_STATE_DONE;
}
break;
}
}
}
return php_http_message_parser_state_is(parser);
}
static PHP_HTTP_FILTER_FUNCTION(chunked_decode)
{
int out_avail = 0;
php_stream_bucket *ptr, *nxt;
PHP_HTTP_FILTER_BUFFER(chunked_decode) *buffer = Z_PTR(this->abstract);
if (bytes_consumed) {
*bytes_consumed = 0;
}
/* fetch available bucket data */
for (ptr = buckets_in->head; ptr; ptr = nxt) {
if (bytes_consumed) {
*bytes_consumed += ptr->buflen;
}
if (PHP_HTTP_BUFFER_NOMEM == php_http_buffer_append(PHP_HTTP_BUFFER(buffer), ptr->buf, ptr->buflen)) {
return PSFS_ERR_FATAL;
}
nxt = ptr->next;
php_stream_bucket_unlink(ptr);
php_stream_bucket_delref(ptr);
}
if (!php_http_buffer_fix(PHP_HTTP_BUFFER(buffer))) {
return PSFS_ERR_FATAL;
}
/* we have data in our buffer */
while (PHP_HTTP_BUFFER(buffer)->used) {
/* we already know the size of the chunk and are waiting for data */
if (buffer->hexlen) {
/* not enough data buffered */
if (PHP_HTTP_BUFFER(buffer)->used < buffer->hexlen) {
/* flush anyway? */
if (flags & PSFS_FLAG_FLUSH_INC) {
/* flush all data (should only be chunk data) */
out_avail = 1;
NEW_BUCKET(PHP_HTTP_BUFFER(buffer)->data, PHP_HTTP_BUFFER(buffer)->used);
/* waiting for less data now */
buffer->hexlen -= PHP_HTTP_BUFFER(buffer)->used;
/* no more buffered data */
php_http_buffer_reset(PHP_HTTP_BUFFER(buffer));
/* break */
}
/* we have too less data and don't need to flush */
else {
break;
}
}
/* we seem to have all data of the chunk */
else {
out_avail = 1;
NEW_BUCKET(PHP_HTTP_BUFFER(buffer)->data, buffer->hexlen);
/* remove outgoing data from the buffer */
php_http_buffer_cut(PHP_HTTP_BUFFER(buffer), 0, buffer->hexlen);
/* reset hexlen */
buffer->hexlen = 0;
/* continue */
}
}
/* we don't know the length of the chunk yet */
else {
size_t off = 0;
/* ignore preceeding CRLFs (too loose?) */
while (off < PHP_HTTP_BUFFER(buffer)->used && (
PHP_HTTP_BUFFER(buffer)->data[off] == '\n' ||
PHP_HTTP_BUFFER(buffer)->data[off] == '\r')) {
++off;
}
if (off) {
php_http_buffer_cut(PHP_HTTP_BUFFER(buffer), 0, off);
}
/* still data there? */
if (PHP_HTTP_BUFFER(buffer)->used) {
int eollen;
const char *eolstr;
/* we need eol, so we can be sure we have all hex digits */
php_http_buffer_fix(PHP_HTTP_BUFFER(buffer));
if ((eolstr = php_http_locate_bin_eol(PHP_HTTP_BUFFER(buffer)->data, PHP_HTTP_BUFFER(buffer)->used, &eollen))) {
char *stop = NULL;
/* read in chunk size */
buffer->hexlen = strtoul(PHP_HTTP_BUFFER(buffer)->data, &stop, 16);
/* if strtoul() stops at the beginning of the buffered data
there's something oddly wrong, i.e. bad input */
if (stop == PHP_HTTP_BUFFER(buffer)->data) {
return PSFS_ERR_FATAL;
}
/* cut out <chunk size hex><chunk extension><eol> */
php_http_buffer_cut(PHP_HTTP_BUFFER(buffer), 0, eolstr + eollen - PHP_HTTP_BUFFER(buffer)->data);
/* buffer->hexlen is 0 now or contains the size of the next chunk */
if (!buffer->hexlen) {
php_stream_notify_info(PHP_STREAM_CONTEXT(stream), PHP_STREAM_NOTIFY_COMPLETED, NULL, 0);
break;
}
/* continue */
} else {
/* we have not enough data buffered to read in chunk size */
break;
}
}
/* break */
}
}
/* flush before close, but only if we are already waiting for more data */
if (PHP_HTTP_FILTER_IS_CLOSING(stream, flags) && buffer->hexlen && PHP_HTTP_BUFFER(buffer)->used) {
out_avail = 1;
NEW_BUCKET(PHP_HTTP_BUFFER(buffer)->data, PHP_HTTP_BUFFER(buffer)->used);
php_http_buffer_reset(PHP_HTTP_BUFFER(buffer));
buffer->hexlen = 0;
}
return out_avail ? PSFS_PASS_ON : PSFS_FEED_ME;
}
