size_t http_parser_execute (http_parser *parser,
const http_parser_settings *settings,
const char *data,
size_t len)
{
char c, ch;
const char *p = data, *pe;
int64_t to_read;
enum state state = (enum state) parser->state;
enum header_states header_state = (enum header_states) parser->header_state;
uint64_t index = parser->index;
uint64_t nread = parser->nread;
if (len == 0) {
if (state == s_body_identity_eof) {
CALLBACK2(message_complete);
}
return 0;
}
/* technically we could combine all of these (except for url_mark) into one
variable, saving stack space, but it seems more clear to have them
separated. */
const char *header_field_mark = 0;
const char *header_value_mark = 0;
const char *fragment_mark = 0;
const char *query_string_mark = 0;
const char *path_mark = 0;
const char *url_mark = 0;
if (state == s_header_field)
header_field_mark = data;
if (state == s_header_value)
header_value_mark = data;
if (state == s_req_fragment)
fragment_mark = data;
if (state == s_req_query_string)
query_string_mark = data;
if (state == s_req_path)
path_mark = data;
if (state == s_req_path || state == s_req_schema || state == s_req_schema_slash
|| state == s_req_schema_slash_slash || state == s_req_port
|| state == s_req_query_string_start || state == s_req_query_string
|| state == s_req_host
|| state == s_req_fragment_start || state == s_req_fragment)
url_mark = data;
for (p=data, pe=data+len; p != pe; p++) {
ch = *p;
if (PARSING_HEADER(state)) {
++nread;
/* Buffer overflow attack */
if (nread > HTTP_MAX_HEADER_SIZE) goto error;
}
switch (state) {
case s_dead:
/* this state is used after a 'Connection: close' message
* the parser will error out if it reads another message
*/
goto error;
case s_start_req_or_res:
{
if (ch == CR || ch == LF)
break;
parser->flags = 0;
parser->content_length = -1;
CALLBACK2(message_begin);
if (ch == 'H')
state = s_res_or_resp_H;
else {
parser->type = HTTP_REQUEST;
goto start_req_method_assign;
}
break;
}
case s_res_or_resp_H:
if (ch == 'T') {
parser->type = HTTP_RESPONSE;
state = s_res_HT;
} else {
if (ch != 'E') goto error;
parser->type = HTTP_REQUEST;
parser->method = HTTP_HEAD;
index = 2;
state = s_req_method;
}
break;
case s_start_res:
{
parser->flags = 0;
parser->content_length = -1;
CALLBACK2(message_begin);
switch (ch) {
case 'H':
state = s_res_H;
break;
case CR:
case LF:
break;
default:
goto error;
}
break;
}
case s_res_H:
STRICT_CHECK(ch != 'T');
state = s_res_HT;
break;
case s_res_HT:
STRICT_CHECK(ch != 'T');
state = s_res_HTT;
break;
case s_res_HTT:
STRICT_CHECK(ch != 'P');
state = s_res_HTTP;
break;
case s_res_HTTP:
STRICT_CHECK(ch != '/');
state = s_res_first_http_major;
break;
case s_res_first_http_major:
if (ch < '1' || ch > '9') goto error;
parser->http_major = ch - '0';
state = s_res_http_major;
break;
/* major HTTP version or dot */
case s_res_http_major:
{
if (ch == '.') {
state = s_res_first_http_minor;
break;
}
if (ch < '0' || ch > '9') goto error;
parser->http_major *= 10;
parser->http_major += ch - '0';
if (parser->http_major > 999) goto error;
break;
}
/* first digit of minor HTTP version */
case s_res_first_http_minor:
if (ch < '0' || ch > '9') goto error;
parser->http_minor = ch - '0';
state = s_res_http_minor;
break;
/* minor HTTP version or end of request line */
case s_res_http_minor:
{
if (ch == ' ') {
state = s_res_first_status_code;
break;
}
if (ch < '0' || ch > '9') goto error;
parser->http_minor *= 10;
parser->http_minor += ch - '0';
if (parser->http_minor > 999) goto error;
break;
}
case s_res_first_status_code:
{
if (ch < '0' || ch > '9') {
if (ch == ' ') {
break;
}
goto error;
}
parser->status_code = ch - '0';
state = s_res_status_code;
break;
}
case s_res_status_code:
{
if (ch < '0' || ch > '9') {
switch (ch) {
case ' ':
state = s_res_status;
break;
case CR:
state = s_res_line_almost_done;
break;
case LF:
state = s_header_field_start;
break;
default:
goto error;
}
break;
}
parser->status_code *= 10;
parser->status_code += ch - '0';
if (parser->status_code > 999) goto error;
break;
}
case s_res_status:
/* the human readable status. e.g. "NOT FOUND"
* we are not humans so just ignore this */
if (ch == CR) {
state = s_res_line_almost_done;
break;
}
if (ch == LF) {
state = s_header_field_start;
break;
}
break;
case s_res_line_almost_done:
STRICT_CHECK(ch != LF);
state = s_header_field_start;
break;
case s_start_req:
{
if (ch == CR || ch == LF)
break;
parser->flags = 0;
parser->content_length = -1;
CALLBACK2(message_begin);
if (ch < 'A' || 'Z' < ch) goto error;
start_req_method_assign:
parser->method = (enum http_method) 0;
index = 1;
switch (ch) {
case 'C': parser->method = HTTP_CONNECT; /* or COPY, CHECKOUT */ break;
case 'D': parser->method = HTTP_DELETE; break;
case 'G': parser->method = HTTP_GET; break;
case 'H': parser->method = HTTP_HEAD; break;
case 'L': parser->method = HTTP_LOCK; break;
case 'M': parser->method = HTTP_MKCOL; /* or MOVE, MKACTIVITY, MERGE */ break;
case 'O': parser->method = HTTP_OPTIONS; break;
case 'P': parser->method = HTTP_POST; /* or PROPFIND or PROPPATCH or PUT */ break;
case 'R': parser->method = HTTP_REPORT; break;
case 'T': parser->method = HTTP_TRACE; break;
case 'U': parser->method = HTTP_UNLOCK; break;
default: goto error;
}
state = s_req_method;
break;
}
case s_req_method:
{
if (ch == '\0')
goto error;
const char *matcher = method_strings[parser->method];
if (ch == ' ' && matcher[index] == '\0') {
state = s_req_spaces_before_url;
} else if (ch == matcher[index]) {
; /* nada */
} else if (parser->method == HTTP_CONNECT) {
if (index == 1 && ch == 'H') {
parser->method = HTTP_CHECKOUT;
} else if (index == 2 && ch == 'P') {
parser->method = HTTP_COPY;
}
} else if (parser->method == HTTP_MKCOL) {
if (index == 1 && ch == 'O') {
parser->method = HTTP_MOVE;
} else if (index == 1 && ch == 'E') {
parser->method = HTTP_MERGE;
} else if (index == 2 && ch == 'A') {
parser->method = HTTP_MKACTIVITY;
}
} else if (index == 1 && parser->method == HTTP_POST && ch == 'R') {
parser->method = HTTP_PROPFIND; /* or HTTP_PROPPATCH */
} else if (index == 1 && parser->method == HTTP_POST && ch == 'U') {
parser->method = HTTP_PUT;
} else if (index == 4 && parser->method == HTTP_PROPFIND && ch == 'P') {
parser->method = HTTP_PROPPATCH;
} else {
goto error;
}
++index;
break;
}
case s_req_spaces_before_url:
{
if (ch == ' ') break;
if (ch == '/') {
MARK(url);
MARK(path);
state = s_req_path;
break;
}
c = LOWER(ch);
if (c >= 'a' && c <= 'z') {
MARK(url);
state = s_req_schema;
break;
}
goto error;
}
case s_req_schema:
{
c = LOWER(ch);
if (c >= 'a' && c <= 'z') break;
if (ch == ':') {
state = s_req_schema_slash;
break;
} else if (ch == '.') {
state = s_req_host;
break;
}
goto error;
}
case s_req_schema_slash:
STRICT_CHECK(ch != '/');
state = s_req_schema_slash_slash;
break;
case s_req_schema_slash_slash:
STRICT_CHECK(ch != '/');
state = s_req_host;
break;
case s_req_host:
{
c = LOWER(ch);
if (c >= 'a' && c <= 'z') break;
if ((ch >= '0' && ch <= '9') || ch == '.' || ch == '-') break;
switch (ch) {
case ':':
state = s_req_port;
break;
case '/':
MARK(path);
state = s_req_path;
break;
case ' ':
/* The request line looks like:
* "GET http://foo.bar.com HTTP/1.1"
* That is, there is no path.
*/
CALLBACK(url);
state = s_req_http_start;
break;
default:
goto error;
}
break;
}
case s_req_port:
{
if (ch >= '0' && ch <= '9') break;
switch (ch) {
case '/':
MARK(path);
state = s_req_path;
break;
case ' ':
/* The request line looks like:
* "GET http://foo.bar.com:1234 HTTP/1.1"
* That is, there is no path.
*/
CALLBACK(url);
state = s_req_http_start;
break;
default:
goto error;
}
break;
}
case s_req_path:
{
if (normal_url_char[(unsigned char)ch]) break;
switch (ch) {
case ' ':
CALLBACK(url);
CALLBACK(path);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
CALLBACK(path);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
CALLBACK(path);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '?':
CALLBACK(path);
state = s_req_query_string_start;
break;
case '#':
CALLBACK(path);
state = s_req_fragment_start;
break;
default:
goto error;
}
break;
}
case s_req_query_string_start:
{
if (normal_url_char[(unsigned char)ch]) {
MARK(query_string);
state = s_req_query_string;
break;
}
switch (ch) {
case '?':
break; /* XXX ignore extra '?' ... is this right? */
case ' ':
CALLBACK(url);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '#':
state = s_req_fragment_start;
break;
default:
goto error;
}
break;
}
case s_req_query_string:
{
if (normal_url_char[(unsigned char)ch]) break;
switch (ch) {
case '?':
/* allow extra '?' in query string */
break;
case ' ':
CALLBACK(url);
CALLBACK(query_string);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
CALLBACK(query_string);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
CALLBACK(query_string);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '#':
CALLBACK(query_string);
state = s_req_fragment_start;
break;
default:
goto error;
}
break;
}
case s_req_fragment_start:
{
if (normal_url_char[(unsigned char)ch]) {
MARK(fragment);
state = s_req_fragment;
break;
}
switch (ch) {
case ' ':
CALLBACK(url);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '?':
MARK(fragment);
state = s_req_fragment;
break;
case '#':
break;
default:
goto error;
}
break;
}
case s_req_fragment:
{
if (normal_url_char[(unsigned char)ch]) break;
switch (ch) {
case ' ':
CALLBACK(url);
CALLBACK(fragment);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
CALLBACK(fragment);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
CALLBACK(fragment);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '?':
case '#':
break;
default:
goto error;
}
break;
}
case s_req_http_start:
switch (ch) {
case 'H':
state = s_req_http_H;
break;
case ' ':
break;
default:
goto error;
}
break;
case s_req_http_H:
STRICT_CHECK(ch != 'T');
state = s_req_http_HT;
break;
case s_req_http_HT:
STRICT_CHECK(ch != 'T');
state = s_req_http_HTT;
break;
case s_req_http_HTT:
STRICT_CHECK(ch != 'P');
state = s_req_http_HTTP;
break;
case s_req_http_HTTP:
STRICT_CHECK(ch != '/');
state = s_req_first_http_major;
break;
/* first digit of major HTTP version */
case s_req_first_http_major:
if (ch < '1' || ch > '9') goto error;
parser->http_major = ch - '0';
state = s_req_http_major;
break;
/* major HTTP version or dot */
case s_req_http_major:
{
if (ch == '.') {
state = s_req_first_http_minor;
break;
}
if (ch < '0' || ch > '9') goto error;
parser->http_major *= 10;
parser->http_major += ch - '0';
if (parser->http_major > 999) goto error;
break;
}
/* first digit of minor HTTP version */
case s_req_first_http_minor:
if (ch < '0' || ch > '9') goto error;
parser->http_minor = ch - '0';
state = s_req_http_minor;
break;
/* minor HTTP version or end of request line */
case s_req_http_minor:
{
if (ch == CR) {
state = s_req_line_almost_done;
break;
}
if (ch == LF) {
state = s_header_field_start;
break;
}
/* XXX allow spaces after digit? */
if (ch < '0' || ch > '9') goto error;
parser->http_minor *= 10;
parser->http_minor += ch - '0';
if (parser->http_minor > 999) goto error;
break;
}
/* end of request line */
case s_req_line_almost_done:
{
if (ch != LF) goto error;
state = s_header_field_start;
break;
}
case s_header_field_start:
{
if (ch == CR) {
state = s_headers_almost_done;
break;
}
if (ch == LF) {
/* they might be just sending \n instead of \r\n so this would be
* the second \n to denote the end of headers*/
state = s_headers_almost_done;
goto headers_almost_done;
}
c = LOWER(ch);
if (c < 'a' || 'z' < c) goto error;
MARK(header_field);
index = 0;
state = s_header_field;
switch (c) {
case 'c':
header_state = h_C;
break;
case 'p':
header_state = h_matching_proxy_connection;
break;
case 't':
header_state = h_matching_transfer_encoding;
break;
case 'u':
header_state = h_matching_upgrade;
break;
default:
header_state = h_general;
break;
}
break;
}
case s_header_field:
{
c = acceptable_header[(unsigned char)ch];
if (c) {
switch (header_state) {
case h_general:
break;
case h_C:
index++;
header_state = (c == 'o' ? h_CO : h_general);
break;
case h_CO:
index++;
header_state = (c == 'n' ? h_CON : h_general);
break;
case h_CON:
index++;
switch (c) {
case 'n':
header_state = h_matching_connection;
break;
case 't':
header_state = h_matching_content_length;
break;
default:
header_state = h_general;
break;
}
break;
/* connection */
case h_matching_connection:
index++;
if (index > sizeof(CONNECTION)-1
|| c != CONNECTION[index]) {
header_state = h_general;
} else if (index == sizeof(CONNECTION)-2) {
header_state = h_connection;
}
break;
/* proxy-connection */
case h_matching_proxy_connection:
index++;
if (index > sizeof(PROXY_CONNECTION)-1
|| c != PROXY_CONNECTION[index]) {
header_state = h_general;
} else if (index == sizeof(PROXY_CONNECTION)-2) {
header_state = h_connection;
}
break;
/* content-length */
case h_matching_content_length:
index++;
if (index > sizeof(CONTENT_LENGTH)-1
|| c != CONTENT_LENGTH[index]) {
header_state = h_general;
} else if (index == sizeof(CONTENT_LENGTH)-2) {
header_state = h_content_length;
}
break;
/* transfer-encoding */
case h_matching_transfer_encoding:
index++;
if (index > sizeof(TRANSFER_ENCODING)-1
|| c != TRANSFER_ENCODING[index]) {
header_state = h_general;
} else if (index == sizeof(TRANSFER_ENCODING)-2) {
header_state = h_transfer_encoding;
}
break;
/* upgrade */
case h_matching_upgrade:
index++;
if (index > sizeof(UPGRADE)-1
|| c != UPGRADE[index]) {
header_state = h_general;
} else if (index == sizeof(UPGRADE)-2) {
header_state = h_upgrade;
}
break;
case h_connection:
case h_content_length:
case h_transfer_encoding:
case h_upgrade:
if (ch != ' ') header_state = h_general;
break;
default:
assert(0 && "Unknown header_state");
break;
}
break;
}
if (ch == ':') {
CALLBACK(header_field);
state = s_header_value_start;
break;
}
if (ch == CR) {
state = s_header_almost_done;
CALLBACK(header_field);
break;
}
if (ch == LF) {
CALLBACK(header_field);
state = s_header_field_start;
break;
}
goto error;
}
case s_header_value_start:
{
if (ch == ' ') break;
MARK(header_value);
state = s_header_value;
index = 0;
c = acceptable_header[(unsigned char)ch];
if (!c) {
if (ch == CR) {
CALLBACK(header_value);
header_state = h_general;
state = s_header_almost_done;
break;
}
if (ch == LF) {
CALLBACK(header_value);
state = s_header_field_start;
break;
}
header_state = h_general;
break;
}
switch (header_state) {
case h_upgrade:
parser->flags |= F_UPGRADE;
header_state = h_general;
break;
case h_transfer_encoding:
/* looking for 'Transfer-Encoding: chunked' */
if ('c' == c) {
header_state = h_matching_transfer_encoding_chunked;
} else {
header_state = h_general;
}
break;
case h_content_length:
if (ch < '0' || ch > '9') goto error;
parser->content_length = ch - '0';
break;
case h_connection:
/* looking for 'Connection: keep-alive' */
if (c == 'k') {
header_state = h_matching_connection_keep_alive;
/* looking for 'Connection: close' */
} else if (c == 'c') {
header_state = h_matching_connection_close;
} else {
header_state = h_general;
}
break;
default:
header_state = h_general;
break;
}
break;
}
case s_header_value:
{
c = acceptable_header[(unsigned char)ch];
if (!c) {
if (ch == CR) {
CALLBACK(header_value);
state = s_header_almost_done;
break;
}
if (ch == LF) {
CALLBACK(header_value);
goto header_almost_done;
}
break;
}
switch (header_state) {
case h_general:
break;
case h_connection:
case h_transfer_encoding:
assert(0 && "Shouldn't get here.");
break;
case h_content_length:
if (ch < '0' || ch > '9') goto error;
parser->content_length *= 10;
parser->content_length += ch - '0';
break;
/* Transfer-Encoding: chunked */
case h_matching_transfer_encoding_chunked:
index++;
if (index > sizeof(CHUNKED)-1
|| c != CHUNKED[index]) {
header_state = h_general;
} else if (index == sizeof(CHUNKED)-2) {
header_state = h_transfer_encoding_chunked;
}
break;
/* looking for 'Connection: keep-alive' */
case h_matching_connection_keep_alive:
index++;
if (index > sizeof(KEEP_ALIVE)-1
|| c != KEEP_ALIVE[index]) {
header_state = h_general;
} else if (index == sizeof(KEEP_ALIVE)-2) {
header_state = h_connection_keep_alive;
}
break;
/* looking for 'Connection: close' */
case h_matching_connection_close:
index++;
if (index > sizeof(CLOSE)-1 || c != CLOSE[index]) {
header_state = h_general;
} else if (index == sizeof(CLOSE)-2) {
header_state = h_connection_close;
}
break;
case h_transfer_encoding_chunked:
case h_connection_keep_alive:
case h_connection_close:
if (ch != ' ') header_state = h_general;
break;
default:
state = s_header_value;
header_state = h_general;
break;
}
break;
}
case s_header_almost_done:
header_almost_done:
{
STRICT_CHECK(ch != LF);
state = s_header_field_start;
switch (header_state) {
case h_connection_keep_alive:
parser->flags |= F_CONNECTION_KEEP_ALIVE;
break;
case h_connection_close:
parser->flags |= F_CONNECTION_CLOSE;
break;
case h_transfer_encoding_chunked:
parser->flags |= F_CHUNKED;
break;
default:
break;
}
break;
}
case s_headers_almost_done:
headers_almost_done:
{
STRICT_CHECK(ch != LF);
if (parser->flags & F_TRAILING) {
/* End of a chunked request */
CALLBACK2(message_complete);
state = NEW_MESSAGE();
break;
}
nread = 0;
if (parser->flags & F_UPGRADE || parser->method == HTTP_CONNECT) {
parser->upgrade = 1;
}
/* Here we call the headers_complete callback. This is somewhat
* different than other callbacks because if the user returns 1, we
* will interpret that as saying that this message has no body. This
* is needed for the annoying case of recieving a response to a HEAD
* request.
*/
if (settings->on_headers_complete) {
switch (settings->on_headers_complete(parser)) {
case 0:
break;
case 1:
parser->flags |= F_SKIPBODY;
break;
default:
return p - data; /* Error */
}
}
/* Exit, the rest of the connect is in a different protocol. */
if (parser->upgrade) {
CALLBACK2(message_complete);
return (p - data);
}
if (parser->flags & F_SKIPBODY) {
CALLBACK2(message_complete);
state = NEW_MESSAGE();
} else if (parser->flags & F_CHUNKED) {
/* chunked encoding - ignore Content-Length header */
state = s_chunk_size_start;
} else {
if (parser->content_length == 0) {
/* Content-Length header given but zero: Content-Length: 0\r\n */
CALLBACK2(message_complete);
state = NEW_MESSAGE();
} else if (parser->content_length > 0) {
/* Content-Length header given and non-zero */
state = s_body_identity;
} else {
if (parser->type == HTTP_REQUEST || http_should_keep_alive(parser)) {
/* Assume content-length 0 - read the next */
CALLBACK2(message_complete);
state = NEW_MESSAGE();
} else {
/* Read body until EOF */
state = s_body_identity_eof;
}
}
}
break;
}
case s_body_identity:
to_read = MIN(pe - p, (int64_t)parser->content_length);
if (to_read > 0) {
if (settings->on_body) settings->on_body(parser, p, to_read);
p += to_read - 1;
parser->content_length -= to_read;
if (parser->content_length == 0) {
CALLBACK2(message_complete);
state = NEW_MESSAGE();
}
}
break;
/* read until EOF */
case s_body_identity_eof:
to_read = pe - p;
if (to_read > 0) {
if (settings->on_body) settings->on_body(parser, p, to_read);
p += to_read - 1;
}
break;
case s_chunk_size_start:
{
assert(parser->flags & F_CHUNKED);
c = unhex[(unsigned char)ch];
if (c == -1) goto error;
parser->content_length = c;
state = s_chunk_size;
break;
}
case s_chunk_size:
{
assert(parser->flags & F_CHUNKED);
if (ch == CR) {
state = s_chunk_size_almost_done;
break;
}
c = unhex[(unsigned char)ch];
if (c == -1) {
if (ch == ';' || ch == ' ') {
state = s_chunk_parameters;
break;
}
goto error;
}
parser->content_length *= 16;
parser->content_length += c;
break;
}
case s_chunk_parameters:
{
assert(parser->flags & F_CHUNKED);
/* just ignore this shit. TODO check for overflow */
if (ch == CR) {
state = s_chunk_size_almost_done;
break;
}
break;
}
case s_chunk_size_almost_done:
{
assert(parser->flags & F_CHUNKED);
STRICT_CHECK(ch != LF);
if (parser->content_length == 0) {
parser->flags |= F_TRAILING;
state = s_header_field_start;
} else {
state = s_chunk_data;
}
break;
}
case s_chunk_data:
{
assert(parser->flags & F_CHUNKED);
to_read = MIN(pe - p, (int64_t)(parser->content_length));
if (to_read > 0) {
if (settings->on_body) settings->on_body(parser, p, to_read);
p += to_read - 1;
}
if (to_read == parser->content_length) {
state = s_chunk_data_almost_done;
}
parser->content_length -= to_read;
break;
}
case s_chunk_data_almost_done:
assert(parser->flags & F_CHUNKED);
STRICT_CHECK(ch != CR);
state = s_chunk_data_done;
break;
case s_chunk_data_done:
assert(parser->flags & F_CHUNKED);
STRICT_CHECK(ch != LF);
state = s_chunk_size_start;
break;
default:
assert(0 && "unhandled state");
goto error;
}
}
CALLBACK_NOCLEAR(header_field);
CALLBACK_NOCLEAR(header_value);
CALLBACK_NOCLEAR(fragment);
CALLBACK_NOCLEAR(query_string);
CALLBACK_NOCLEAR(path);
CALLBACK_NOCLEAR(url);
parser->state = state;
parser->header_state = header_state;
parser->index = index;
parser->nread = nread;
return len;
error:
return (p - data);
}
size_t http_parser_execute (http_parser *parser,
const http_parser_settings *settings,
const char *data,
size_t len)
{
char c, ch;
const char *p = data, *pe;
ssize_t to_read;
enum state state = (enum state) parser->state;
enum header_states header_state = (enum header_states) parser->header_state;
size_t index = parser->index;
size_t nread = parser->nread;
if (len == 0) {
if (state == s_body_identity_eof) {
CALLBACK2(message_complete);
}
return 0;
}
if (parser->header_field_mark) parser->header_field_mark = data;
if (parser->header_value_mark) parser->header_value_mark = data;
if (parser->fragment_mark) parser->fragment_mark = data;
if (parser->query_string_mark) parser->query_string_mark = data;
if (parser->path_mark) parser->path_mark = data;
if (parser->url_mark) parser->url_mark = data;
for (p=data, pe=data+len; p != pe; p++) {
ch = *p;
if (++nread > HTTP_MAX_HEADER_SIZE && PARSING_HEADER(state)) {
/* Buffer overflow attack */
goto error;
}
switch (state) {
case s_dead:
/* this state is used after a 'Connection: close' message
* the parser will error out if it reads another message
*/
goto error;
case s_start_res:
{
parser->flags = 0;
parser->content_length = -1;
CALLBACK2(message_begin);
switch (ch) {
case 'H':
state = s_res_H;
break;
case CR:
case LF:
break;
default:
goto error;
}
break;
}
case s_res_H:
STRICT_CHECK(ch != 'T');
state = s_res_HT;
break;
case s_res_HT:
STRICT_CHECK(ch != 'T');
state = s_res_HTT;
break;
case s_res_HTT:
STRICT_CHECK(ch != 'P');
state = s_res_HTTP;
break;
case s_res_HTTP:
STRICT_CHECK(ch != '/');
state = s_res_first_http_major;
break;
case s_res_first_http_major:
if (ch < '1' || ch > '9') goto error;
parser->http_major = ch - '0';
state = s_res_http_major;
break;
/* major HTTP version or dot */
case s_res_http_major:
{
if (ch == '.') {
state = s_res_first_http_minor;
break;
}
if (ch < '0' || ch > '9') goto error;
parser->http_major *= 10;
parser->http_major += ch - '0';
if (parser->http_major > 999) goto error;
break;
}
/* first digit of minor HTTP version */
case s_res_first_http_minor:
if (ch < '0' || ch > '9') goto error;
parser->http_minor = ch - '0';
state = s_res_http_minor;
break;
/* minor HTTP version or end of request line */
case s_res_http_minor:
{
if (ch == ' ') {
state = s_res_first_status_code;
break;
}
if (ch < '0' || ch > '9') goto error;
parser->http_minor *= 10;
parser->http_minor += ch - '0';
if (parser->http_minor > 999) goto error;
break;
}
case s_res_first_status_code:
{
if (ch < '0' || ch > '9') {
if (ch == ' ') {
break;
}
goto error;
}
parser->status_code = ch - '0';
state = s_res_status_code;
break;
}
case s_res_status_code:
{
if (ch < '0' || ch > '9') {
switch (ch) {
case ' ':
state = s_res_status;
break;
case CR:
state = s_res_line_almost_done;
break;
case LF:
state = s_header_field_start;
break;
default:
goto error;
}
break;
}
parser->status_code *= 10;
parser->status_code += ch - '0';
if (parser->status_code > 999) goto error;
break;
}
case s_res_status:
/* the human readable status. e.g. "NOT FOUND"
* we are not humans so just ignore this */
if (ch == CR) {
state = s_res_line_almost_done;
break;
}
if (ch == LF) {
state = s_header_field_start;
break;
}
break;
case s_res_line_almost_done:
STRICT_CHECK(ch != LF);
state = s_header_field_start;
break;
case s_start_req:
{
if (ch == CR || ch == LF)
break;
parser->flags = 0;
parser->content_length = -1;
CALLBACK2(message_begin);
if (ch < 'A' || 'Z' < ch) goto error;
parser->method = (enum http_method) 0;
index = 0;
parser->buffer[0] = ch;
state = s_req_method;
break;
}
case s_req_method:
if (ch == ' ') {
assert(index+1 < HTTP_PARSER_MAX_METHOD_LEN);
parser->buffer[index+1] = '\0';
switch (index+1) {
case 3:
if (ngx_str3_cmp(parser->buffer, 'G', 'E', 'T')) {
parser->method = HTTP_GET;
break;
}
if (ngx_str3_cmp(parser->buffer, 'P', 'U', 'T')) {
parser->method = HTTP_PUT;
break;
}
break;
case 4:
if (ngx_str4cmp(parser->buffer, 'P', 'O', 'S', 'T')) {
parser->method = HTTP_POST;
break;
}
if (ngx_str4cmp(parser->buffer, 'H', 'E', 'A', 'D')) {
parser->method = HTTP_HEAD;
break;
}
if (ngx_str4cmp(parser->buffer, 'C', 'O', 'P', 'Y')) {
parser->method = HTTP_COPY;
break;
}
if (ngx_str4cmp(parser->buffer, 'M', 'O', 'V', 'E')) {
parser->method = HTTP_MOVE;
break;
}
break;
case 5:
if (ngx_str5cmp(parser->buffer, 'M', 'K', 'C', 'O', 'L')) {
parser->method = HTTP_MKCOL;
break;
}
if (ngx_str5cmp(parser->buffer, 'T', 'R', 'A', 'C', 'E')) {
parser->method = HTTP_TRACE;
break;
}
break;
case 6:
if (ngx_str6cmp(parser->buffer, 'D', 'E', 'L', 'E', 'T', 'E')) {
parser->method = HTTP_DELETE;
break;
}
if (ngx_str6cmp(parser->buffer, 'U', 'N', 'L', 'O', 'C', 'K')) {
parser->method = HTTP_UNLOCK;
break;
}
break;
case 7:
if (ngx_str7_cmp(parser->buffer,
'O', 'P', 'T', 'I', 'O', 'N', 'S', '\0')) {
parser->method = HTTP_OPTIONS;
break;
}
if (ngx_str7_cmp(parser->buffer,
'C', 'O', 'N', 'N', 'E', 'C', 'T', '\0')) {
parser->method = HTTP_CONNECT;
break;
}
break;
case 8:
if (ngx_str8cmp(parser->buffer,
'P', 'R', 'O', 'P', 'F', 'I', 'N', 'D')) {
parser->method = HTTP_PROPFIND;
break;
}
break;
case 9:
if (ngx_str9cmp(parser->buffer,
'P', 'R', 'O', 'P', 'P', 'A', 'T', 'C', 'H')) {
parser->method = HTTP_PROPPATCH;
break;
}
break;
}
state = s_req_spaces_before_url;
break;
}
if (ch < 'A' || 'Z' < ch) goto error;
if (++index >= HTTP_PARSER_MAX_METHOD_LEN - 1) {
goto error;
}
parser->buffer[index] = ch;
break;
case s_req_spaces_before_url:
{
if (ch == ' ') break;
if (ch == '/') {
MARK(url);
MARK(path);
state = s_req_path;
break;
}
c = LOWER(ch);
if (c >= 'a' && c <= 'z') {
MARK(url);
state = s_req_schema;
break;
}
goto error;
}
case s_req_schema:
{
c = LOWER(ch);
if (c >= 'a' && c <= 'z') break;
if (ch == ':') {
state = s_req_schema_slash;
break;
}
goto error;
}
case s_req_schema_slash:
STRICT_CHECK(ch != '/');
state = s_req_schema_slash_slash;
break;
case s_req_schema_slash_slash:
STRICT_CHECK(ch != '/');
state = s_req_host;
break;
case s_req_host:
{
c = LOWER(ch);
if (c >= 'a' && c <= 'z') break;
if ((ch >= '0' && ch <= '9') || ch == '.' || ch == '-') break;
switch (ch) {
case ':':
state = s_req_port;
break;
case '/':
MARK(path);
state = s_req_path;
break;
case ' ':
/* The request line looks like:
* "GET http://foo.bar.com HTTP/1.1"
* That is, there is no path.
*/
CALLBACK(url);
state = s_req_http_start;
break;
default:
goto error;
}
break;
}
case s_req_port:
{
if (ch >= '0' && ch <= '9') break;
switch (ch) {
case '/':
MARK(path);
state = s_req_path;
break;
case ' ':
/* The request line looks like:
* "GET http://foo.bar.com:1234 HTTP/1.1"
* That is, there is no path.
*/
CALLBACK(url);
state = s_req_http_start;
break;
default:
goto error;
}
break;
}
case s_req_path:
{
if (USUAL(ch)) break;
switch (ch) {
case ' ':
CALLBACK(url);
CALLBACK(path);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
CALLBACK(path);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
CALLBACK(path);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '?':
CALLBACK(path);
state = s_req_query_string_start;
break;
case '#':
CALLBACK(path);
state = s_req_fragment_start;
break;
default:
goto error;
}
break;
}
case s_req_query_string_start:
{
if (USUAL(ch)) {
MARK(query_string);
state = s_req_query_string;
break;
}
switch (ch) {
case '?':
break; // XXX ignore extra '?' ... is this right?
case ' ':
CALLBACK(url);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '#':
state = s_req_fragment_start;
break;
default:
goto error;
}
break;
}
case s_req_query_string:
{
if (USUAL(ch)) break;
switch (ch) {
case '?':
// allow extra '?' in query string
break;
case ' ':
CALLBACK(url);
CALLBACK(query_string);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
CALLBACK(query_string);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
CALLBACK(query_string);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '#':
CALLBACK(query_string);
state = s_req_fragment_start;
break;
default:
goto error;
}
break;
}
case s_req_fragment_start:
{
if (USUAL(ch)) {
MARK(fragment);
state = s_req_fragment;
break;
}
switch (ch) {
case ' ':
CALLBACK(url);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '?':
MARK(fragment);
state = s_req_fragment;
break;
case '#':
break;
default:
goto error;
}
break;
}
case s_req_fragment:
{
if (USUAL(ch)) break;
switch (ch) {
case ' ':
CALLBACK(url);
CALLBACK(fragment);
state = s_req_http_start;
break;
case CR:
CALLBACK(url);
CALLBACK(fragment);
parser->http_minor = 9;
state = s_req_line_almost_done;
break;
case LF:
CALLBACK(url);
CALLBACK(fragment);
parser->http_minor = 9;
state = s_header_field_start;
break;
case '?':
case '#':
break;
default:
goto error;
}
break;
}
case s_req_http_start:
switch (ch) {
case 'H':
state = s_req_http_H;
break;
case ' ':
break;
default:
goto error;
}
break;
case s_req_http_H:
STRICT_CHECK(ch != 'T');
state = s_req_http_HT;
break;
case s_req_http_HT:
STRICT_CHECK(ch != 'T');
state = s_req_http_HTT;
break;
case s_req_http_HTT:
STRICT_CHECK(ch != 'P');
state = s_req_http_HTTP;
break;
case s_req_http_HTTP:
STRICT_CHECK(ch != '/');
state = s_req_first_http_major;
break;
/* first digit of major HTTP version */
case s_req_first_http_major:
if (ch < '1' || ch > '9') goto error;
parser->http_major = ch - '0';
state = s_req_http_major;
break;
/* major HTTP version or dot */
case s_req_http_major:
{
if (ch == '.') {
state = s_req_first_http_minor;
break;
}
if (ch < '0' || ch > '9') goto error;
parser->http_major *= 10;
parser->http_major += ch - '0';
if (parser->http_major > 999) goto error;
break;
}
/* first digit of minor HTTP version */
case s_req_first_http_minor:
if (ch < '0' || ch > '9') goto error;
parser->http_minor = ch - '0';
state = s_req_http_minor;
break;
/* minor HTTP version or end of request line */
case s_req_http_minor:
{
if (ch == CR) {
state = s_req_line_almost_done;
break;
}
if (ch == LF) {
state = s_header_field_start;
break;
}
/* XXX allow spaces after digit? */
if (ch < '0' || ch > '9') goto error;
parser->http_minor *= 10;
parser->http_minor += ch - '0';
if (parser->http_minor > 999) goto error;
break;
}
/* end of request line */
case s_req_line_almost_done:
{
if (ch != LF) goto error;
state = s_header_field_start;
break;
}
case s_header_field_start:
{
if (ch == CR) {
state = s_headers_almost_done;
break;
}
if (ch == LF) {
/* they might be just sending \n instead of \r\n so this would be
* the second \n to denote the end of headers*/
state = s_headers_almost_done;
goto headers_almost_done;
}
c = LOWER(ch);
if (c < 'a' || 'z' < c) goto error;
MARK(header_field);
index = 0;
state = s_header_field;
switch (c) {
case 'c':
header_state = h_C;
break;
case 'p':
header_state = h_matching_proxy_connection;
break;
case 't':
header_state = h_matching_transfer_encoding;
break;
case 'u':
header_state = h_matching_upgrade;
break;
default:
header_state = h_general;
break;
}
break;
}
case s_header_field:
{
c = lowcase[(int)ch];
if (c) {
switch (header_state) {
case h_general:
break;
case h_C:
index++;
header_state = (c == 'o' ? h_CO : h_general);
break;
case h_CO:
index++;
header_state = (c == 'n' ? h_CON : h_general);
break;
case h_CON:
index++;
switch (c) {
case 'n':
header_state = h_matching_connection;
break;
case 't':
header_state = h_matching_content_length;
break;
default:
header_state = h_general;
break;
}
break;
/* connection */
case h_matching_connection:
index++;
if (index > sizeof(CONNECTION)-1
|| c != CONNECTION[index]) {
header_state = h_general;
} else if (index == sizeof(CONNECTION)-2) {
header_state = h_connection;
}
break;
/* proxy-connection */
case h_matching_proxy_connection:
index++;
if (index > sizeof(PROXY_CONNECTION)-1
|| c != PROXY_CONNECTION[index]) {
header_state = h_general;
} else if (index == sizeof(PROXY_CONNECTION)-2) {
header_state = h_connection;
}
break;
/* content-length */
case h_matching_content_length:
index++;
if (index > sizeof(CONTENT_LENGTH)-1
|| c != CONTENT_LENGTH[index]) {
header_state = h_general;
} else if (index == sizeof(CONTENT_LENGTH)-2) {
header_state = h_content_length;
}
break;
/* transfer-encoding */
case h_matching_transfer_encoding:
index++;
if (index > sizeof(TRANSFER_ENCODING)-1
|| c != TRANSFER_ENCODING[index]) {
header_state = h_general;
} else if (index == sizeof(TRANSFER_ENCODING)-2) {
header_state = h_transfer_encoding;
}
break;
/* upgrade */
case h_matching_upgrade:
index++;
if (index > sizeof(UPGRADE)-1
|| c != UPGRADE[index]) {
header_state = h_general;
} else if (index == sizeof(UPGRADE)-2) {
header_state = h_upgrade;
}
break;
case h_connection:
case h_content_length:
case h_transfer_encoding:
case h_upgrade:
if (ch != ' ') header_state = h_general;
break;
default:
assert(0 && "Unknown header_state");
break;
}
break;
}
if (ch == ':') {
CALLBACK(header_field);
state = s_header_value_start;
break;
}
if (ch == CR) {
state = s_header_almost_done;
CALLBACK(header_field);
break;
}
if (ch == LF) {
CALLBACK(header_field);
state = s_header_field_start;
break;
}
goto error;
}
case s_header_value_start:
{
if (ch == ' ') break;
MARK(header_value);
state = s_header_value;
index = 0;
c = lowcase[(int)ch];
if (!c) {
if (ch == CR) {
header_state = h_general;
state = s_header_almost_done;
break;
}
if (ch == LF) {
state = s_header_field_start;
break;
}
header_state = h_general;
break;
}
switch (header_state) {
case h_upgrade:
parser->flags |= F_UPGRADE;
header_state = h_general;
break;
case h_transfer_encoding:
/* looking for 'Transfer-Encoding: chunked' */
if ('c' == c) {
header_state = h_matching_transfer_encoding_chunked;
} else {
header_state = h_general;
}
break;
case h_content_length:
if (ch < '0' || ch > '9') goto error;
parser->content_length = ch - '0';
break;
case h_connection:
/* looking for 'Connection: keep-alive' */
if (c == 'k') {
header_state = h_matching_connection_keep_alive;
/* looking for 'Connection: close' */
} else if (c == 'c') {
header_state = h_matching_connection_close;
} else {
header_state = h_general;
}
break;
default:
header_state = h_general;
break;
}
break;
}
case s_header_value:
{
c = lowcase[(int)ch];
if (!c) {
if (ch == CR) {
CALLBACK(header_value);
state = s_header_almost_done;
break;
}
if (ch == LF) {
CALLBACK(header_value);
goto header_almost_done;
}
break;
}
switch (header_state) {
case h_general:
break;
case h_connection:
case h_transfer_encoding:
assert(0 && "Shouldn't get here.");
break;
case h_content_length:
if (ch < '0' || ch > '9') goto error;
parser->content_length *= 10;
parser->content_length += ch - '0';
break;
/* Transfer-Encoding: chunked */
case h_matching_transfer_encoding_chunked:
index++;
if (index > sizeof(CHUNKED)-1
|| c != CHUNKED[index]) {
header_state = h_general;
} else if (index == sizeof(CHUNKED)-2) {
header_state = h_transfer_encoding_chunked;
}
break;
/* looking for 'Connection: keep-alive' */
case h_matching_connection_keep_alive:
index++;
if (index > sizeof(KEEP_ALIVE)-1
|| c != KEEP_ALIVE[index]) {
header_state = h_general;
} else if (index == sizeof(KEEP_ALIVE)-2) {
header_state = h_connection_keep_alive;
}
break;
/* looking for 'Connection: close' */
case h_matching_connection_close:
index++;
if (index > sizeof(CLOSE)-1 || c != CLOSE[index]) {
header_state = h_general;
} else if (index == sizeof(CLOSE)-2) {
header_state = h_connection_close;
}
break;
case h_transfer_encoding_chunked:
case h_connection_keep_alive:
case h_connection_close:
if (ch != ' ') header_state = h_general;
break;
default:
state = s_header_value;
header_state = h_general;
break;
}
break;
}
case s_header_almost_done:
header_almost_done:
{
STRICT_CHECK(ch != LF);
state = s_header_field_start;
switch (header_state) {
case h_connection_keep_alive:
parser->flags |= F_CONNECTION_KEEP_ALIVE;
break;
case h_connection_close:
parser->flags |= F_CONNECTION_CLOSE;
break;
case h_transfer_encoding_chunked:
parser->flags |= F_CHUNKED;
break;
default:
break;
}
break;
}
case s_headers_almost_done:
headers_almost_done:
{
STRICT_CHECK(ch != LF);
if (parser->flags & F_TRAILING) {
/* End of a chunked request */
CALLBACK2(message_complete);
state = NEW_MESSAGE();
break;
}
parser->body_read = 0;
nread = 0;
if (parser->flags & F_UPGRADE) parser->upgrade = 1;
CALLBACK2(headers_complete);
// Exit, the rest of the connect is in a different protocol.
if (parser->flags & F_UPGRADE) {
CALLBACK2(message_complete);
return (p - data);
}
if (parser->flags & F_CHUNKED) {
/* chunked encoding - ignore Content-Length header */
state = s_chunk_size_start;
} else {
if (parser->content_length == 0) {
/* Content-Length header given but zero: Content-Length: 0\r\n */
CALLBACK2(message_complete);
state = NEW_MESSAGE();
} else if (parser->content_length > 0) {
/* Content-Length header given and non-zero */
state = s_body_identity;
} else {
if (parser->type == HTTP_REQUEST || http_should_keep_alive(parser)) {
/* Assume content-length 0 - read the next */
CALLBACK2(message_complete);
state = NEW_MESSAGE();
} else {
/* Read body until EOF */
state = s_body_identity_eof;
}
}
}
break;
}
case s_body_identity:
to_read = MIN(pe - p, (ssize_t)(parser->content_length - parser->body_read));
if (to_read > 0) {
if (settings->on_body) settings->on_body(parser, p, to_read);
p += to_read - 1;
parser->body_read += to_read;
if (parser->body_read == parser->content_length) {
CALLBACK2(message_complete);
state = NEW_MESSAGE();
}
}
break;
/* read until EOF */
case s_body_identity_eof:
to_read = pe - p;
if (to_read > 0) {
if (settings->on_body) settings->on_body(parser, p, to_read);
p += to_read - 1;
parser->body_read += to_read;
}
break;
case s_chunk_size_start:
{
assert(parser->flags & F_CHUNKED);
c = unhex[(int)ch];
if (c == -1) goto error;
parser->content_length = c;
state = s_chunk_size;
break;
}
case s_chunk_size:
{
assert(parser->flags & F_CHUNKED);
if (ch == CR) {
state = s_chunk_size_almost_done;
break;
}
c = unhex[(int)ch];
if (c == -1) {
if (ch == ';' || ch == ' ') {
state = s_chunk_parameters;
break;
}
goto error;
}
parser->content_length *= 16;
parser->content_length += c;
break;
}
case s_chunk_parameters:
{
assert(parser->flags & F_CHUNKED);
/* just ignore this shit. TODO check for overflow */
if (ch == CR) {
state = s_chunk_size_almost_done;
break;
}
break;
}
case s_chunk_size_almost_done:
{
assert(parser->flags & F_CHUNKED);
STRICT_CHECK(ch != LF);
if (parser->content_length == 0) {
parser->flags |= F_TRAILING;
state = s_header_field_start;
} else {
state = s_chunk_data;
}
break;
}
case s_chunk_data:
{
assert(parser->flags & F_CHUNKED);
to_read = MIN(pe - p, (ssize_t)(parser->content_length));
if (to_read > 0) {
if (settings->on_body) settings->on_body(parser, p, to_read);
p += to_read - 1;
}
if (to_read == parser->content_length) {
state = s_chunk_data_almost_done;
}
parser->content_length -= to_read;
break;
}
case s_chunk_data_almost_done:
assert(parser->flags & F_CHUNKED);
STRICT_CHECK(ch != CR);
state = s_chunk_data_done;
break;
case s_chunk_data_done:
assert(parser->flags & F_CHUNKED);
STRICT_CHECK(ch != LF);
state = s_chunk_size_start;
break;
default:
assert(0 && "unhandled state");
goto error;
}
}
CALLBACK_NOCLEAR(header_field);
CALLBACK_NOCLEAR(header_value);
CALLBACK_NOCLEAR(fragment);
CALLBACK_NOCLEAR(query_string);
CALLBACK_NOCLEAR(path);
CALLBACK_NOCLEAR(url);
parser->state = state;
parser->header_state = header_state;
parser->index = index;
parser->nread = nread;
return len;
error:
return (p - data);
}