# Copyright (c) 2005-2009, eventlet contributors # Copyright (c) 2009-2018, gevent contributors """ A pure-Python, gevent-friendly WSGI server implementing HTTP/1.1. The server is provided in :class:`WSGIServer`, but most of the actual WSGI work is handled by :class:`WSGIHandler` --- a new instance is created for each request. The server can be customized to use different subclasses of :class:`WSGIHandler`. .. important:: This server is intended primarily for development and testing, and secondarily for other "safe" scenarios where it will not be exposed to potentially malicious input. The code has not been security audited, and is not intended for direct exposure to the public Internet. For production usage on the Internet, either choose a production-strength server such as gunicorn, or put a reverse proxy between gevent and the Internet. .. versionchanged:: 23.9.0 Complies more closely with the HTTP specification for chunked transfer encoding. In particular, we are much stricter about trailers, and trailers that are invalid (too long or featuring disallowed characters) forcibly close the connection to the client *after* the results have been sent. Trailers otherwise continue to be ignored and are not available to the WSGI application. """ from __future__ import absolute_import # FIXME: Can we refactor to make smallor? # pylint:disable=too-many-lines import errno from io import BytesIO import string import sys import time import traceback from datetime import datetime from urllib.parse import unquote from gevent import socket import gevent from gevent.server import StreamServer from gevent.hub import GreenletExit from gevent._compat import reraise from functools import partial unquote_latin1 = partial(unquote, encoding='latin-1') _no_undoc_members = True # Don't put undocumented things into sphinx __all__ = [ 'WSGIServer', 'WSGIHandler', 'LoggingLogAdapter', 'Environ', 'SecureEnviron', 'WSGISecureEnviron', ] MAX_REQUEST_LINE = 8192 # Weekday and month names for HTTP date/time formatting; always English! _WEEKDAYNAME = ("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun") _MONTHNAME = (None, # Dummy so we can use 1-based month numbers "Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec") # The contents of the "HEX" grammar rule for HTTP, upper and lowercase A-F plus digits, # in byte form for comparing to the network. _HEX = string.hexdigits.encode('ascii') # The characters allowed in "token" rules. # token = 1*tchar # tchar = "!" / "#" / "$" / "%" / "&" / "'" / "*" # / "+" / "-" / "." / "^" / "_" / "`" / "|" / "~" # / DIGIT / ALPHA # ; any VCHAR, except delimiters # ALPHA = %x41-5A / %x61-7A ; A-Z / a-z _ALLOWED_TOKEN_CHARS = frozenset( # Remember we have to be careful because bytestrings # inexplicably iterate as integers, which are not equal to bytes. # explicit chars then DIGIT (c.encode('ascii') for c in "!#$%&'*+-.^_`|~0123456789") # Then we add ALPHA ) | {c.encode('ascii') for c in string.ascii_letters} assert b'A' in _ALLOWED_TOKEN_CHARS # Errors _ERRORS = {} _INTERNAL_ERROR_STATUS = '500 Internal Server Error' _INTERNAL_ERROR_BODY = b'Internal Server Error' _INTERNAL_ERROR_HEADERS = ( ('Content-Type', 'text/plain'), ('Connection', 'close'), ('Content-Length', str(len(_INTERNAL_ERROR_BODY))) ) _ERRORS[500] = (_INTERNAL_ERROR_STATUS, _INTERNAL_ERROR_HEADERS, _INTERNAL_ERROR_BODY) _BAD_REQUEST_STATUS = '400 Bad Request' _BAD_REQUEST_BODY = '' _BAD_REQUEST_HEADERS = ( ('Content-Type', 'text/plain'), ('Connection', 'close'), ('Content-Length', str(len(_BAD_REQUEST_BODY))) ) _ERRORS[400] = (_BAD_REQUEST_STATUS, _BAD_REQUEST_HEADERS, _BAD_REQUEST_BODY) _REQUEST_TOO_LONG_RESPONSE = b"HTTP/1.1 414 Request URI Too Long\r\nConnection: close\r\nContent-length: 0\r\n\r\n" _BAD_REQUEST_RESPONSE = b"HTTP/1.1 400 Bad Request\r\nConnection: close\r\nContent-length: 0\r\n\r\n" _CONTINUE_RESPONSE = b"HTTP/1.1 100 Continue\r\n\r\n" def format_date_time(timestamp): # Return a byte-string of the date and time in HTTP format # .. versionchanged:: 1.1b5 # Return a byte string, not a native string year, month, day, hh, mm, ss, wd, _y, _z = time.gmtime(timestamp) value = "%s, %02d %3s %4d %02d:%02d:%02d GMT" % (_WEEKDAYNAME[wd], day, _MONTHNAME[month], year, hh, mm, ss) value = value.encode("latin-1") return value class _InvalidClientInput(IOError): # Internal exception raised by Input indicating that the client # sent invalid data at the lowest level of the stream. The result # *should* be a HTTP 400 error. pass class _InvalidClientRequest(ValueError): # Internal exception raised by WSGIHandler.read_request indicating # that the client sent an HTTP request that cannot be parsed # (e.g., invalid grammar). The result *should* be an HTTP 400 # error. It must have exactly one argument, the fully formatted # error string. def __init__(self, message): ValueError.__init__(self, message) self.formatted_message = message class Input(object): __slots__ = ('rfile', 'content_length', 'socket', 'position', 'chunked_input', 'chunk_length', '_chunked_input_error') def __init__(self, rfile, content_length, socket=None, chunked_input=False): # pylint:disable=redefined-outer-name self.rfile = rfile self.content_length = content_length self.socket = socket self.position = 0 self.chunked_input = chunked_input self.chunk_length = -1 self._chunked_input_error = False def _discard(self): if self._chunked_input_error: # We are in an unknown state, so we can't necessarily discard # the body (e.g., if the client keeps the socket open, we could hang # here forever). # In this case, we've raised an exception and the user of this object # is going to close the socket, so we don't have to discard return if self.socket is None and (self.position < (self.content_length or 0) or self.chunked_input): # ## Read and discard body while 1: d = self.read(16384) if not d: break def _send_100_continue(self): if self.socket is not None: self.socket.sendall(_CONTINUE_RESPONSE) self.socket = None def _do_read(self, length=None, use_readline=False): if use_readline: reader = self.rfile.readline else: reader = self.rfile.read content_length = self.content_length if content_length is None: # Either Content-Length or "Transfer-Encoding: chunked" must be present in a request with a body # if it was chunked, then this function would have not been called return b'' self._send_100_continue() left = content_length - self.position if length is None: length = left elif length > left: length = left if not length: return b'' # On Python 2, self.rfile is usually socket.makefile(), which # uses cStringIO.StringIO. If *length* is greater than the C # sizeof(int) (typically 32 bits signed), parsing the argument to # readline raises OverflowError. StringIO.read(), OTOH, uses # PySize_t, typically a long (64 bits). In a bare readline() # case, because the header lines we're trying to read with # readline are typically expected to be small, we can correct # that failure by simply doing a smaller call to readline and # appending; failures in read we let propagate. try: read = reader(length) except OverflowError: if not use_readline: # Expecting to read more than 64 bits of data. Ouch! raise # We could loop on calls to smaller readline(), appending them # until we actually get a newline. For uses in this module, # we expect the actual length to be small, but WSGI applications # are allowed to pass in an arbitrary length. (This loop isn't optimal, # but even client applications *probably* have short lines.) read = b'' while len(read) < length and not read.endswith(b'\n'): read += reader(MAX_REQUEST_LINE) self.position += len(read) if len(read) < length: if (use_readline and not read.endswith(b"\n")) or not use_readline: raise IOError("unexpected end of file while reading request at position %s" % (self.position,)) return read def __read_chunk_length(self, rfile): # Read and return the next integer chunk length. If no # chunk length can be read, raises _InvalidClientInput. # Here's the production for a chunk (actually the whole body): # (https://www.rfc-editor.org/rfc/rfc7230#section-4.1) # chunked-body = *chunk # last-chunk # trailer-part # CRLF # # chunk = chunk-size [ chunk-ext ] CRLF # chunk-data CRLF # chunk-size = 1*HEXDIG # last-chunk = 1*("0") [ chunk-ext ] CRLF # trailer-part = *( header-field CRLF ) # chunk-data = 1*OCTET ; a sequence of chunk-size octets # # chunk-ext = *( ";" chunk-ext-name [ "=" chunk-ext-val ] ) # # chunk-ext-name = token # chunk-ext-val = token / quoted-string # To cope with malicious or broken clients that fail to send # valid chunk lines, the strategy is to read character by # character until we either reach a ; or newline. If at any # time we read a non-HEX digit, we bail. If we hit a ;, # indicating an chunk-extension, we'll read up to the next # MAX_REQUEST_LINE characters ("A server ought to limit the # total length of chunk extensions received") looking for the # CRLF, and if we don't find it, we bail. If we read more than # 16 hex characters, (the number needed to represent a 64-bit # chunk size), we bail (this protects us from a client that # sends an infinite stream of `F`, for example). buf = BytesIO() while 1: char = rfile.read(1) if not char: self._chunked_input_error = True raise _InvalidClientInput("EOF before chunk end reached") if char in ( b'\r', # Beginning EOL b';', # Beginning extension ): break if char not in _HEX: # Invalid data. self._chunked_input_error = True raise _InvalidClientInput("Non-hex data", char) buf.write(char) if buf.tell() > 16: # Too many hex bytes self._chunked_input_error = True raise _InvalidClientInput("Chunk-size too large.") if char == b';': i = 0 while i < MAX_REQUEST_LINE: char = rfile.read(1) if char == b'\r': break i += 1 else: # we read more than MAX_REQUEST_LINE without # hitting CR self._chunked_input_error = True raise _InvalidClientInput("Too large chunk extension") if char == b'\r': # We either got here from the main loop or from the # end of an extension self.__read_chunk_size_crlf(rfile, newline_only=True) result = int(buf.getvalue(), 16) if result == 0: # The only time a chunk size of zero is allowed is the final # chunk. It is either followed by another \r\n, or some trailers # which are then followed by \r\n. while self.__read_chunk_trailer(rfile): pass return result # Trailers have the following production (they are a header-field followed by CRLF) # See above for the definition of "token". # # header-field = field-name ":" OWS field-value OWS # field-name = token # field-value = *( field-content / obs-fold ) # field-content = field-vchar [ 1*( SP / HTAB ) field-vchar ] # field-vchar = VCHAR / obs-text # obs-fold = CRLF 1*( SP / HTAB ) # ; obsolete line folding # ; see Section 3.2.4 def __read_chunk_trailer(self, rfile, ): # With rfile positioned just after a \r\n, read a trailer line. # Return a true value if a non-empty trailer was read, and # return false if an empty trailer was read (meaning the trailers are # done). # If a single line exceeds the MAX_REQUEST_LINE, raise an exception. # If the field-name portion contains invalid characters, raise an exception. i = 0 empty = True seen_field_name = False while i < MAX_REQUEST_LINE: char = rfile.read(1) if char == b'\r': # Either read the next \n or raise an error. self.__read_chunk_size_crlf(rfile, newline_only=True) break # Not a \r, so we are NOT an empty chunk. empty = False if char == b':' and i > 0: # We're ending the field-name part; stop validating characters. # Unless : was the first character... seen_field_name = True if not seen_field_name and char not in _ALLOWED_TOKEN_CHARS: raise _InvalidClientInput('Invalid token character: %r' % (char,)) i += 1 else: # We read too much self._chunked_input_error = True raise _InvalidClientInput("Too large chunk trailer") return not empty def __read_chunk_size_crlf(self, rfile, newline_only=False): # Also for safety, correctly verify that we get \r\n when expected. if not newline_only: char = rfile.read(1) if char != b'\r': self._chunked_input_error = True raise _InvalidClientInput("Line didn't end in CRLF: %r" % (char,)) char = rfile.read(1) if char != b'\n': self._chunked_input_error = True raise _InvalidClientInput("Line didn't end in LF: %r" % (char,)) def _chunked_read(self, length=None, use_readline=False): # pylint:disable=too-many-branches rfile = self.rfile self._send_100_continue() if length == 0: return b"" if use_readline: reader = self.rfile.readline else: reader = self.rfile.read response = [] while self.chunk_length != 0: maxreadlen = self.chunk_length - self.position if length is not None and length < maxreadlen: maxreadlen = length if maxreadlen > 0: data = reader(maxreadlen) if not data: self.chunk_length = 0 self._chunked_input_error = True raise IOError("unexpected end of file while parsing chunked data") datalen = len(data) response.append(data) self.position += datalen if self.chunk_length == self.position: self.__read_chunk_size_crlf(rfile) if length is not None: length -= datalen if length == 0: break if use_readline and data[-1] == b"\n"[0]: break else: # We're at the beginning of a chunk, so we need to # determine the next size to read self.chunk_length = self.__read_chunk_length(rfile) self.position = 0 # If chunk_length was 0, we already read any trailers and # validated that we have ended with \r\n\r\n. return b''.join(response) def read(self, length=None): if length is not None and length < 0: length = None if self.chunked_input: return self._chunked_read(length) return self._do_read(length) def readline(self, size=None): if size is not None and size < 0: size = None if self.chunked_input: return self._chunked_read(size, True) return self._do_read(size, use_readline=True) def readlines(self, hint=None): # pylint:disable=unused-argument return list(self) def __iter__(self): return self def next(self): line = self.readline() if not line: raise StopIteration return line __next__ = next try: import mimetools headers_factory = mimetools.Message except ImportError: # adapt Python 3 HTTP headers to old API from http import client # pylint:disable=import-error class OldMessage(client.HTTPMessage): def __init__(self, **kwargs): super(client.HTTPMessage, self).__init__(**kwargs) # pylint:disable=bad-super-call self.status = '' def getheader(self, name, default=None): return self.get(name, default) @property def headers(self): for key, value in self._headers: yield '%s: %s\r\n' % (key, value) @property def typeheader(self): return self.get('content-type') def headers_factory(fp, *args): # pylint:disable=unused-argument try: ret = client.parse_headers(fp, _class=OldMessage) except client.LineTooLong: ret = OldMessage() ret.status = 'Line too long' return ret class WSGIHandler(object): """ Handles HTTP requests from a socket, creates the WSGI environment, and interacts with the WSGI application. This is the default value of :attr:`WSGIServer.handler_class`. This class may be subclassed carefully, and that class set on a :class:`WSGIServer` instance through a keyword argument at construction time. Instances are constructed with the same arguments as passed to the server's :meth:`WSGIServer.handle` method followed by the server itself. The application and environment are obtained from the server. """ # pylint:disable=too-many-instance-attributes protocol_version = 'HTTP/1.1' def MessageClass(self, *args): return headers_factory(*args) # Attributes reset at various times for each request; not public # documented. Class attributes to keep the constructor fast # (but not make lint tools complain) status = None # byte string: b'200 OK' _orig_status = None # native string: '200 OK' response_headers = None # list of tuples (b'name', b'value') code = None # Integer parsed from status provided_date = None provided_content_length = None close_connection = False time_start = 0 # time.time() when begin handling request time_finish = 0 # time.time() when done handling request headers_sent = False # Have we already sent headers? response_use_chunked = False # Write with transfer-encoding chunked # Was the connection upgraded? We shouldn't try to chunk writes in that # case. connection_upgraded = False environ = None # Dict from self.get_environ application = None # application callable from self.server.application requestline = None # native str 'GET / HTTP/1.1' response_length = 0 # How much data we sent result = None # The return value of the WSGI application wsgi_input = None # Instance of Input() content_length = 0 # From application-provided headers Incoming # request headers, instance of MessageClass (gunicorn uses hasattr # on this so the default value needs to be compatible with the # API) headers = headers_factory(BytesIO()) request_version = None # str: 'HTTP 1.1' command = None # str: 'GET' path = None # str: '/' def __init__(self, sock, address, server, rfile=None): # Deprecation: The rfile kwarg was introduced in 1.0a1 as part # of a refactoring. It was never documented or used. It is # considered DEPRECATED and may be removed in the future. Its # use is not supported. self.socket = sock self.client_address = address self.server = server if rfile is None: self.rfile = sock.makefile('rb', -1) else: self.rfile = rfile def handle(self): """ The main request handling method, called by the server. This method runs a request handling loop, calling :meth:`handle_one_request` until all requests on the connection have been handled (that is, it implements keep-alive). """ try: while self.socket is not None: self.time_start = time.time() self.time_finish = 0 result = self.handle_one_request() if result is None: break if result is True: continue self.status, response_body = result # pylint:disable=unpacking-non-sequence self.socket.sendall(response_body) if self.time_finish == 0: self.time_finish = time.time() self.log_request() break finally: if self.socket is not None: _sock = getattr(self.socket, '_sock', None) # Python 3 try: # read out request data to prevent error: [Errno 104] Connection reset by peer if _sock: try: # socket.recv would hang _sock.recv(16384) finally: _sock.close() self.socket.close() except socket.error: pass self.__dict__.pop('socket', None) self.__dict__.pop('rfile', None) self.__dict__.pop('wsgi_input', None) def _check_http_version(self): version_str = self.request_version if not version_str.startswith("HTTP/"): return False version = tuple(int(x) for x in version_str[5:].split(".")) # "HTTP/" if version[1] < 0 or version < (0, 9) or version >= (2, 0): return False return True def read_request(self, raw_requestline): """ Parse the incoming request. Parses various headers into ``self.headers`` using :attr:`MessageClass`. Other attributes that are set upon a successful return of this method include ``self.content_length`` and ``self.close_connection``. :param str raw_requestline: A native :class:`str` representing the request line. A processed version of this will be stored into ``self.requestline``. :raises ValueError: If the request is invalid. This error will not be logged as a traceback (because it's a client issue, not a server problem). :return: A boolean value indicating whether the request was successfully parsed. This method should either return a true value or have raised a ValueError with details about the parsing error. .. versionchanged:: 1.1b6 Raise the previously documented :exc:`ValueError` in more cases instead of returning a false value; this allows subclasses more opportunity to customize behaviour. """ # pylint:disable=too-many-branches self.requestline = raw_requestline.rstrip() words = self.requestline.split() if len(words) == 3: self.command, self.path, self.request_version = words if not self._check_http_version(): raise _InvalidClientRequest('Invalid http version: %r' % (raw_requestline,)) elif len(words) == 2: self.command, self.path = words if self.command != "GET": raise _InvalidClientRequest('Expected GET method; Got command=%r; path=%r; raw=%r' % ( self.command, self.path, raw_requestline,)) self.request_version = "HTTP/0.9" # QQQ I'm pretty sure we can drop support for HTTP/0.9 else: raise _InvalidClientRequest('Invalid HTTP method: %r' % (raw_requestline,)) self.headers = self.MessageClass(self.rfile, 0) if self.headers.status: raise _InvalidClientRequest('Invalid headers status: %r' % (self.headers.status,)) if self.headers.get("transfer-encoding", "").lower() == "chunked": try: del self.headers["content-length"] except KeyError: pass content_length = self.headers.get("content-length") if content_length is not None: content_length = int(content_length) if content_length < 0: raise _InvalidClientRequest('Invalid Content-Length: %r' % (content_length,)) if content_length and self.command in ('HEAD', ): raise _InvalidClientRequest('Unexpected Content-Length') self.content_length = content_length if self.request_version == "HTTP/1.1": conntype = self.headers.get("Connection", "").lower() self.close_connection = (conntype == 'close') # pylint:disable=superfluous-parens elif self.request_version == 'HTTP/1.0': conntype = self.headers.get("Connection", "close").lower() self.close_connection = (conntype != 'keep-alive') # pylint:disable=superfluous-parens else: # XXX: HTTP 0.9. We should drop support self.close_connection = True return True _print_unexpected_exc = staticmethod(traceback.print_exc) def log_error(self, msg, *args): if not args: # Already fully formatted, no need to do it again; msg # might contain % chars that would lead to a formatting # error. message = msg else: try: message = msg % args except Exception: # pylint:disable=broad-except self._print_unexpected_exc() message = '%r %r' % (msg, args) try: message = '%s: %s' % (self.socket, message) except Exception: # pylint:disable=broad-except pass try: self.server.error_log.write(message + '\n') except Exception: # pylint:disable=broad-except self._print_unexpected_exc() def read_requestline(self): """ Read and return the HTTP request line. Under both Python 2 and 3, this should return the native ``str`` type; under Python 3, this probably means the bytes read from the network need to be decoded (using the ISO-8859-1 charset, aka latin-1). """ line = self.rfile.readline(MAX_REQUEST_LINE) line = line.decode('latin-1') return line def handle_one_request(self): """ Handles one HTTP request using ``self.socket`` and ``self.rfile``. Each invocation of this method will do several things, including (but not limited to): - Read the request line using :meth:`read_requestline`; - Read the rest of the request, including headers, with :meth:`read_request`; - Construct a new WSGI environment in ``self.environ`` using :meth:`get_environ`; - Store the application in ``self.application``, retrieving it from the server; - Handle the remainder of the request, including invoking the application, with :meth:`handle_one_response` There are several possible return values to indicate the state of the client connection: - ``None`` The client connection is already closed or should be closed because the WSGI application or client set the ``Connection: close`` header. The request handling loop should terminate and perform cleanup steps. - (status, body) An HTTP status and body tuple. The request was in error, as detailed by the status and body. The request handling loop should terminate, close the connection, and perform cleanup steps. Note that the ``body`` is the complete contents to send to the client, including all headers and the initial status line. - ``True`` The literal ``True`` value. The request was successfully handled and the response sent to the client by :meth:`handle_one_response`. The connection remains open to process more requests and the connection handling loop should call this method again. This is the typical return value. .. seealso:: :meth:`handle` .. versionchanged:: 1.1b6 Funnel exceptions having to do with invalid HTTP requests through :meth:`_handle_client_error` to allow subclasses to customize. Note that this is experimental and may change in the future. """ # pylint:disable=too-many-return-statements if self.rfile.closed: return try: self.requestline = self.read_requestline() # Account for old subclasses that haven't done this if isinstance(self.requestline, bytes): self.requestline = self.requestline.decode('latin-1') except socket.error: # "Connection reset by peer" or other socket errors aren't interesting here return if not self.requestline: return self.response_length = 0 if len(self.requestline) >= MAX_REQUEST_LINE: return ('414', _REQUEST_TOO_LONG_RESPONSE) try: # for compatibility with older versions of pywsgi, we pass self.requestline as an argument there # NOTE: read_request is supposed to raise ValueError on invalid input; allow old # subclasses that return a False value instead. # NOTE: This can mutate the value of self.headers, so self.get_environ() must not be # called until AFTER this call is done. if not self.read_request(self.requestline): return ('400', _BAD_REQUEST_RESPONSE) except Exception as ex: # pylint:disable=broad-except # Notice we don't use self.handle_error because it reports # a 500 error to the client, and this is almost certainly # a client error. # Provide a hook for subclasses. return self._handle_client_error(ex) self.environ = self.get_environ() self.application = self.server.application self.handle_one_response() if self.close_connection: return if self.rfile.closed: return return True # read more requests def _connection_upgrade_requested(self): if self.headers.get('Connection', '').lower() == 'upgrade': return True if self.headers.get('Upgrade', '').lower() == 'websocket': return True return False def finalize_headers(self): if self.provided_date is None: self.response_headers.append((b'Date', format_date_time(time.time()))) self.connection_upgraded = self.code == 101 if self.code not in (304, 204): # the reply will include message-body; make sure we have either Content-Length or chunked if self.provided_content_length is None: if hasattr(self.result, '__len__'): total_len = sum(len(chunk) for chunk in self.result) total_len_str = str(total_len) total_len_str = total_len_str.encode("latin-1") self.response_headers.append((b'Content-Length', total_len_str)) else: self.response_use_chunked = ( not self.connection_upgraded and self.request_version != 'HTTP/1.0' ) if self.response_use_chunked: self.response_headers.append((b'Transfer-Encoding', b'chunked')) def _sendall(self, data): try: self.socket.sendall(data) except socket.error as ex: self.status = 'socket error: %s' % ex if self.code > 0: self.code = -self.code raise self.response_length += len(data) def _write(self, data, _bytearray=bytearray): if not data: # The application/middleware are allowed to yield # empty bytestrings. return if self.response_use_chunked: # Write the chunked encoding header header_str = b'%x\r\n' % len(data) towrite = _bytearray(header_str) # data towrite += data # trailer towrite += b'\r\n' self._sendall(towrite) else: self._sendall(data) ApplicationError = AssertionError def write(self, data): # The write() callable we return from start_response. # https://www.python.org/dev/peps/pep-3333/#the-write-callable # Supposed to do pretty much the same thing as yielding values # from the application's return. if self.code in (304, 204) and data: raise self.ApplicationError('The %s response must have no body' % self.code) if self.headers_sent: self._write(data) else: if not self.status: raise self.ApplicationError("The application did not call start_response()") self._write_with_headers(data) def _write_with_headers(self, data): self.headers_sent = True self.finalize_headers() # self.response_headers and self.status are already in latin-1, as encoded by self.start_response towrite = bytearray(b'HTTP/1.1 ') towrite += self.status towrite += b'\r\n' for header, value in self.response_headers: towrite += header towrite += b': ' towrite += value towrite += b"\r\n" towrite += b'\r\n' self._sendall(towrite) # No need to copy the data into towrite; we may make an extra syscall # but the copy time could be substantial too, and it reduces the chances # of sendall being able to send everything in one go self._write(data) def start_response(self, status, headers, exc_info=None): """ .. versionchanged:: 1.2a1 Avoid HTTP header injection by raising a :exc:`ValueError` if *status* or any *header* name or value contains a carriage return or newline. .. versionchanged:: 1.1b5 Pro-actively handle checking the encoding of the status line and headers during this method. On Python 2, avoid some extra encodings. """ # pylint:disable=too-many-branches,too-many-statements if exc_info: try: if self.headers_sent: # Re-raise original exception if headers sent reraise(*exc_info) finally: # Avoid dangling circular ref exc_info = None # Pep 3333, "The start_response callable": # https://www.python.org/dev/peps/pep-3333/#the-start-response-callable # "Servers should check for errors in the headers at the time # start_response is called, so that an error can be raised # while the application is still running." Here, we check the encoding. # This aids debugging: headers especially are generated programmatically # and an encoding error in a loop or list comprehension yields an opaque # UnicodeError without any clue which header was wrong. # Note that this results in copying the header list at this point, not modifying it, # although we are allowed to do so if needed. This slightly increases memory usage. # We also check for HTTP Response Splitting vulnerabilities response_headers = [] header = None value = None try: for header, value in headers: if not isinstance(header, str): raise UnicodeError("The header must be a native string", header, value) if not isinstance(value, str): raise UnicodeError("The value must be a native string", header, value) if '\r' in header or '\n' in header: raise ValueError('carriage return or newline in header name', header) if '\r' in value or '\n' in value: raise ValueError('carriage return or newline in header value', value) # Either we're on Python 2, in which case bytes is correct, or # we're on Python 3 and the user screwed up (because it should be a native # string). In either case, make sure that this is latin-1 compatible. Under # Python 2, bytes.encode() will take a round-trip through the system encoding, # which may be ascii, which is not really what we want. However, the latin-1 encoding # can encode everything except control characters and the block from 0x7F to 0x9F, so # explicitly round-tripping bytes through the encoding is unlikely to be of much # benefit, so we go for speed (the WSGI spec specifically calls out allowing the range # from 0x00 to 0xFF, although the HTTP spec forbids the control characters). # Note: Some Python 2 implementations, like Jython, may allow non-octet (above 255) values # in their str implementation; this is mentioned in the WSGI spec, but we don't # run on any platform like that so we can assume that a str value is pure bytes. response_headers.append((header.encode("latin-1"), value.encode("latin-1"))) except UnicodeEncodeError: # If we get here, we're guaranteed to have a header and value raise UnicodeError("Non-latin1 header", repr(header), repr(value)) # Same as above if not isinstance(status, str): raise UnicodeError("The status string must be a native string") if '\r' in status or '\n' in status: raise ValueError("carriage return or newline in status", status) # don't assign to anything until the validation is complete, including parsing the # code code = int(status.split(' ', 1)[0]) self.status = status.encode("latin-1") self._orig_status = status # Preserve the native string for logging self.response_headers = response_headers self.code = code provided_connection = None # Did the wsgi app give us a Connection header? self.provided_date = None self.provided_content_length = None for header, value in headers: header = header.lower() if header == 'connection': provided_connection = value elif header == 'date': self.provided_date = value elif header == 'content-length': self.provided_content_length = value if self.request_version == 'HTTP/1.0' and provided_connection is None: conntype = b'close' if self.close_connection else b'keep-alive' response_headers.append((b'Connection', conntype)) elif provided_connection == 'close': self.close_connection = True if self.code in (304, 204): if self.provided_content_length is not None and self.provided_content_length != '0': msg = 'Invalid Content-Length for %s response: %r (must be absent or zero)' % (self.code, self.provided_content_length) msg = msg.encode('latin-1') raise self.ApplicationError(msg) return self.write def log_request(self): self.server.log.write(self.format_request() + '\n') def format_request(self): now = datetime.now().replace(microsecond=0) length = self.response_length or '-' if self.time_finish: delta = '%.6f' % (self.time_finish - self.time_start) else: delta = '-' client_address = self.client_address[0] if isinstance(self.client_address, tuple) else self.client_address return '%s - - [%s] "%s" %s %s %s' % ( client_address or '-', now, self.requestline or '', # Use the native string version of the status, saved so we don't have to # decode. But fallback to the encoded 'status' in case of subclasses # (Is that really necessary? At least there's no overhead.) (self._orig_status or self.status or '000').split()[0], length, delta) def process_result(self): for data in self.result: if data: self.write(data) if self.status and not self.headers_sent: # In other words, the application returned an empty # result iterable (and did not use the write callable) # Trigger the flush of the headers. self.write(b'') if self.response_use_chunked: self._sendall(b'0\r\n\r\n') def run_application(self): assert self.result is None try: self.result = self.application(self.environ, self.start_response) self.process_result() finally: close = getattr(self.result, 'close', None) try: if close is not None: close() finally: # Discard the result. If it's a generator this can # free a lot of hidden resources (if we failed to iterate # all the way through it---the frames are automatically # cleaned up when StopIteration is raised); but other cases # could still free up resources sooner than otherwise. close = None self.result = None #: These errors are silently ignored by :meth:`handle_one_response` to avoid producing #: excess log entries on normal operating conditions. They indicate #: a remote client has disconnected and there is little or nothing #: this process can be expected to do about it. You may change this #: value in a subclass. #: #: The default value includes :data:`errno.EPIPE` and :data:`errno.ECONNRESET`. #: On Windows this also includes :data:`errno.WSAECONNABORTED`. #: #: This is a provisional API, subject to change. See :pr:`377`, :pr:`999` #: and :issue:`136`. #: #: .. versionadded:: 1.3 ignored_socket_errors = (errno.EPIPE, errno.ECONNRESET) try: ignored_socket_errors += (errno.WSAECONNABORTED,) except AttributeError: pass # Not windows def handle_one_response(self): """ Invoke the application to produce one response. This is called by :meth:`handle_one_request` after all the state for the request has been established. It is responsible for error handling. """ self.time_start = time.time() self.status = None self.headers_sent = False self.result = None self.response_use_chunked = False self.connection_upgraded = False self.response_length = 0 try: try: self.run_application() finally: try: self.wsgi_input._discard() except _InvalidClientInput: # This one is deliberately raised to the outer # scope, because, with the incoming stream in some bad state, # we can't be sure we can synchronize and properly parse the next # request. raise except socket.error: # Don't let socket exceptions during discarding # input override any exception that may have been # raised by the application, such as our own _InvalidClientInput. # In the general case, these aren't even worth logging (see the comment # just below) pass except _InvalidClientInput as ex: # DO log this one because: # - Some of the data may have been read and acted on by the # application; # - The response may or may not have been sent; # - It's likely that the client is bad, or malicious, and # users might wish to take steps to block the client. self._handle_client_error(ex) self.close_connection = True self._send_error_response_if_possible(400) except socket.error as ex: if ex.args[0] in self.ignored_socket_errors: # See description of self.ignored_socket_errors. self.close_connection = True else: self.handle_error(*sys.exc_info()) except: # pylint:disable=bare-except self.handle_error(*sys.exc_info()) finally: self.time_finish = time.time() self.log_request() def _send_error_response_if_possible(self, error_code): if self.response_length: self.close_connection = True else: status, headers, body = _ERRORS[error_code] try: self.start_response(status, headers[:]) self.write(body) except socket.error: self.close_connection = True def _log_error(self, t, v, tb): # TODO: Shouldn't we dump this to wsgi.errors? If we did that now, it would # wind up getting logged twice if not issubclass(t, GreenletExit): context = self.environ if not isinstance(context, self.server.secure_environ_class): context = self.server.secure_environ_class(context) self.server.loop.handle_error(context, t, v, tb) def handle_error(self, t, v, tb): # Called for internal, unexpected errors, NOT invalid client input self._log_error(t, v, tb) t = v = tb = None self._send_error_response_if_possible(500) def _handle_client_error(self, ex): # Called for invalid client input # Returns the appropriate error response. if not isinstance(ex, (ValueError, _InvalidClientInput)): # XXX: Why not self._log_error to send it through the loop's # handle_error method? # _InvalidClientRequest is a ValueError; _InvalidClientInput is an IOError. traceback.print_exc() if isinstance(ex, _InvalidClientRequest): # No formatting needed, that's already been handled. In fact, because the # formatted message contains user input, it might have a % in it, and attempting # to format that with no arguments would be an error. # However, the error messages do not include the requesting IP # necessarily, so we do add that. self.log_error('(from %s) %s', self.client_address, ex.formatted_message) else: self.log_error('Invalid request (from %s): %s', self.client_address, str(ex) or ex.__class__.__name__) return ('400', _BAD_REQUEST_RESPONSE) def _headers(self): key = None value = None IGNORED_KEYS = (None, 'CONTENT_TYPE', 'CONTENT_LENGTH') for header in self.headers.headers: if key is not None and header[:1] in " \t": value += header continue if key not in IGNORED_KEYS: yield 'HTTP_' + key, value.strip() key, value = header.split(':', 1) if '_' in key: # strip incoming bad veaders key = None else: key = key.replace('-', '_').upper() if key not in IGNORED_KEYS: yield 'HTTP_' + key, value.strip() def get_environ(self): """ Construct and return a new WSGI environment dictionary for a specific request. This should begin with asking the server for the base environment using :meth:`WSGIServer.get_environ`, and then proceed to add the request specific values. By the time this method is invoked the request line and request shall have been parsed and ``self.headers`` shall be populated. """ env = self.server.get_environ() env['REQUEST_METHOD'] = self.command # SCRIPT_NAME is explicitly implementation defined. Using an # empty value for SCRIPT_NAME is both explicitly allowed by # both the CGI standard and WSGI PEPs, and also the thing that # makes the most sense from a generic server perspective (we # have no hierarchy or understanding of URLs or files, just a # single application to call. The empty string represents the # application root, which is what we have). Different WSGI # implementations handle this very differently, so portable # applications that rely on SCRIPT_NAME will have to use a # WSGI middleware to set it to a defined value, or otherwise # rely on server-specific mechanisms (e.g, on waitress, use # ``--url-prefix``, in gunicorn set the ``SCRIPT_NAME`` header # or process environment variable, in gevent subclass # WSGIHandler.) # # See https://github.com/gevent/gevent/issues/1667 for discussion. env['SCRIPT_NAME'] = '' path, query = self.path.split('?', 1) if '?' in self.path else (self.path, '') # Note that self.path contains the original str object; if it contains # encoded escapes, it will NOT match PATH_INFO. env['PATH_INFO'] = unquote_latin1(path) env['QUERY_STRING'] = query if self.headers.typeheader is not None: env['CONTENT_TYPE'] = self.headers.typeheader length = self.headers.getheader('content-length') if length: env['CONTENT_LENGTH'] = length env['SERVER_PROTOCOL'] = self.request_version client_address = self.client_address if isinstance(client_address, tuple): env['REMOTE_ADDR'] = str(client_address[0]) env['REMOTE_PORT'] = str(client_address[1]) for key, value in self._headers(): if key in env: if 'COOKIE' in key: env[key] += '; ' + value else: env[key] += ',' + value else: env[key] = value sock = self.socket if env.get('HTTP_EXPECT') == '100-continue' else None chunked = env.get('HTTP_TRANSFER_ENCODING', '').lower() == 'chunked' # Input refuses to read if the data isn't chunked, and there is no content_length # provided. For 'Upgrade: Websocket' requests, neither of those things is true. handling_reads = not self._connection_upgrade_requested() self.wsgi_input = Input(self.rfile, self.content_length, socket=sock, chunked_input=chunked) env['wsgi.input'] = self.wsgi_input if handling_reads else self.rfile # This is a non-standard flag indicating that our input stream is # self-terminated (returns EOF when consumed). # See https://github.com/gevent/gevent/issues/1308 env['wsgi.input_terminated'] = handling_reads return env class _NoopLog(object): # Does nothing; implements just enough file-like methods # to pass the WSGI validator def write(self, *args, **kwargs): # pylint:disable=unused-argument return def flush(self): pass def writelines(self, *args, **kwargs): pass class LoggingLogAdapter(object): """ An adapter for :class:`logging.Logger` instances to let them be used with :class:`WSGIServer`. .. warning:: Unless the entire process is monkey-patched at a very early part of the lifecycle (before logging is configured), loggers are likely to not be gevent-cooperative. For example, the socket and syslog handlers use the socket module in a way that can block, and most handlers acquire threading locks. .. warning:: It *may* be possible for the logging functions to be called in the :class:`gevent.Hub` greenlet. Code running in the hub greenlet cannot use any gevent blocking functions without triggering a ``LoopExit``. .. versionadded:: 1.1a3 .. versionchanged:: 1.1b6 Attributes not present on this object are proxied to the underlying logger instance. This permits using custom :class:`~logging.Logger` subclasses (or indeed, even duck-typed objects). .. versionchanged:: 1.1 Strip trailing newline characters on the message passed to :meth:`write` because log handlers will usually add one themselves. """ # gevent avoids importing and using logging because importing it and # creating loggers creates native locks unless monkey-patched. __slots__ = ('_logger', '_level') def __init__(self, logger, level=20): """ Write information to the *logger* at the given *level* (default to INFO). """ self._logger = logger self._level = level def write(self, msg): if msg and msg.endswith('\n'): msg = msg[:-1] self._logger.log(self._level, msg) def flush(self): "No-op; required to be a file-like object" def writelines(self, lines): for line in lines: self.write(line) def __getattr__(self, name): return getattr(self._logger, name) def __setattr__(self, name, value): if name not in LoggingLogAdapter.__slots__: setattr(self._logger, name, value) else: object.__setattr__(self, name, value) def __delattr__(self, name): delattr(self._logger, name) #### ## Environ classes. # These subclass dict. They could subclass collections.UserDict on # 3.3+ and proxy to the underlying real dict to avoid a copy if we # have to print them (on 2.7 it's slightly more complicated to be an # instance of collections.MutableMapping; UserDict.UserDict isn't.) # Then we could have either the WSGIHandler.get_environ or the # WSGIServer.get_environ return one of these proxies, and # WSGIHandler.run_application would know to access the `environ.data` # attribute to be able to pass the *real* dict to the application # (because PEP3333 requires no subclasses, only actual dict objects; # wsgiref.validator and webob.Request both enforce this). This has the # advantage of not being fragile if anybody else tries to print/log # self.environ (and not requiring a copy). However, if there are any # subclasses of Handler or Server, this could break if they don't know # to return this type. #### class Environ(dict): """ A base class that can be used for WSGI environment objects. Provisional API. .. versionadded:: 1.2a1 """ __slots__ = () # add no ivars or weakref ability def copy(self): return self.__class__(self) if not hasattr(dict, 'iteritems'): # Python 3 def iteritems(self): return self.items() def __reduce_ex__(self, proto): return (dict, (), None, None, iter(self.iteritems())) class SecureEnviron(Environ): """ An environment that does not print its keys and values by default. Provisional API. This is intended to keep potentially sensitive information like HTTP authorization and cookies from being inadvertently printed or logged. For debugging, each instance can have its *secure_repr* attribute set to ``False``, which will cause it to print like a normal dict. When *secure_repr* is ``True`` (the default), then the value of the *whitelist_keys* attribute is consulted; if this value is true-ish, it should be a container (something that responds to ``in``) of key names (typically a list or set). Keys and values in this dictionary that are in *whitelist_keys* will then be printed, while all other values will be masked. These values may be customized on the class by setting the *default_secure_repr* and *default_whitelist_keys*, respectively:: >>> environ = SecureEnviron(key='value') >>> environ # doctest: +ELLIPSIS >> environ.whitelist_keys = {'key'} >>> environ {'key': 'value'} A non-whitelisted key (*only*, to avoid doctest issues) is masked:: >>> environ['secure'] = 'secret'; del environ['key'] >>> environ {'secure': ''} We can turn it off entirely for the instance:: >>> environ.secure_repr = False >>> environ {'secure': 'secret'} We can also customize it at the class level (here we use a new class to be explicit and to avoid polluting the true default values; we would set this class to be the ``environ_class`` of the server):: >>> class MyEnviron(SecureEnviron): ... default_whitelist_keys = ('key',) ... >>> environ = MyEnviron({'key': 'value'}) >>> environ {'key': 'value'} .. versionadded:: 1.2a1 """ default_secure_repr = True default_whitelist_keys = () default_print_masked_keys = True # Allow instances to override the class values, # but inherit from the class if not present. Keeps instances # small since we can't combine __slots__ with class attributes # of the same name. __slots__ = ('secure_repr', 'whitelist_keys', 'print_masked_keys') def __getattr__(self, name): if name in SecureEnviron.__slots__: return getattr(type(self), 'default_' + name) raise AttributeError(name) def __repr__(self): if self.secure_repr: whitelist = self.whitelist_keys print_masked = self.print_masked_keys if whitelist: safe = {k: self[k] if k in whitelist else "" for k in self if k in whitelist or print_masked} safe_repr = repr(safe) if not print_masked and len(safe) != len(self): safe_repr = safe_repr[:-1] + ", (hidden keys: %d)}" % (len(self) - len(safe)) return safe_repr return "" % (len(self), id(self)) return Environ.__repr__(self) __str__ = __repr__ class WSGISecureEnviron(SecureEnviron): """ Specializes the default list of whitelisted keys to a few common WSGI variables. Example:: >>> environ = WSGISecureEnviron(REMOTE_ADDR='::1', HTTP_AUTHORIZATION='secret') >>> environ {'REMOTE_ADDR': '::1', (hidden keys: 1)} >>> import pprint >>> pprint.pprint(environ) {'REMOTE_ADDR': '::1', (hidden keys: 1)} >>> print(pprint.pformat(environ)) {'REMOTE_ADDR': '::1', (hidden keys: 1)} """ default_whitelist_keys = ('REMOTE_ADDR', 'REMOTE_PORT', 'HTTP_HOST') default_print_masked_keys = False class WSGIServer(StreamServer): """ A WSGI server based on :class:`StreamServer` that supports HTTPS. :keyword log: If given, an object with a ``write`` method to which request (access) logs will be written. If not given, defaults to :obj:`sys.stderr`. You may pass ``None`` to disable request logging. You may use a wrapper, around e.g., :mod:`logging`, to support objects that don't implement a ``write`` method. (If you pass a :class:`~logging.Logger` instance, or in general something that provides a ``log`` method but not a ``write`` method, such a wrapper will automatically be created and it will be logged to at the :data:`~logging.INFO` level.) :keyword error_log: If given, a file-like object with ``write``, ``writelines`` and ``flush`` methods to which error logs will be written. If not given, defaults to :obj:`sys.stderr`. You may pass ``None`` to disable error logging (not recommended). You may use a wrapper, around e.g., :mod:`logging`, to support objects that don't implement the proper methods. This parameter will become the value for ``wsgi.errors`` in the WSGI environment (if not already set). (As with *log*, wrappers for :class:`~logging.Logger` instances and the like will be created automatically and logged to at the :data:`~logging.ERROR` level.) .. seealso:: :class:`LoggingLogAdapter` See important warnings before attempting to use :mod:`logging`. .. versionchanged:: 1.1a3 Added the ``error_log`` parameter, and set ``wsgi.errors`` in the WSGI environment to this value. .. versionchanged:: 1.1a3 Add support for passing :class:`logging.Logger` objects to the ``log`` and ``error_log`` arguments. .. versionchanged:: 20.6.0 Passing a ``handle`` kwarg to the constructor is now officially deprecated. """ #: A callable taking three arguments: (socket, address, server) and returning #: an object with a ``handle()`` method. The callable is called once for #: each incoming socket request, as is its handle method. The handle method should not #: return until all use of the socket is complete. #: #: This class uses the :class:`WSGIHandler` object as the default value. You may #: subclass this class and set a different default value, or you may pass #: a value to use in the ``handler_class`` keyword constructor argument. handler_class = WSGIHandler #: The object to which request logs will be written. #: It must never be None. Initialized from the ``log`` constructor #: parameter. log = None #: The object to which error logs will be written. #: It must never be None. Initialized from the ``error_log`` constructor #: parameter. error_log = None #: The class of environ objects passed to the handlers. #: Must be a dict subclass. For compliance with :pep:`3333` #: and libraries like WebOb, this is simply :class:`dict` #: but this can be customized in a subclass or per-instance #: (probably to :class:`WSGISecureEnviron`). #: #: .. versionadded:: 1.2a1 environ_class = dict # Undocumented internal detail: the class that WSGIHandler._log_error # will cast to before passing to the loop. secure_environ_class = WSGISecureEnviron base_env = {'GATEWAY_INTERFACE': 'CGI/1.1', 'SERVER_SOFTWARE': 'gevent/%d.%d Python/%d.%d' % (gevent.version_info[:2] + sys.version_info[:2]), 'SCRIPT_NAME': '', 'wsgi.version': (1, 0), 'wsgi.multithread': False, # XXX: Aren't we really, though? 'wsgi.multiprocess': False, 'wsgi.run_once': False} def __init__(self, listener, application=None, backlog=None, spawn='default', log='default', error_log='default', handler_class=None, environ=None, **ssl_args): if 'handle' in ssl_args: # The ultimate base class (BaseServer) uses 'handle' for # the thing we call 'application'. We never deliberately # bass a `handle` argument to the base class, but one # could sneak in through ``**ssl_args``, even though that # is not the intent, while application is None. That # causes our own ``def handle`` method to be replaced, # probably leading to bad results. Passing a 'handle' # instead of an 'application' can really confuse things. import warnings warnings.warn("Passing 'handle' kwarg to WSGIServer is deprecated. " "Did you mean application?", DeprecationWarning, stacklevel=2) StreamServer.__init__(self, listener, backlog=backlog, spawn=spawn, **ssl_args) if application is not None: self.application = application if handler_class is not None: self.handler_class = handler_class # Note that we can't initialize these as class variables: # sys.stderr might get monkey patched at runtime. def _make_log(l, level=20): if l == 'default': return sys.stderr if l is None: return _NoopLog() if not hasattr(l, 'write') and hasattr(l, 'log'): return LoggingLogAdapter(l, level) return l self.log = _make_log(log) self.error_log = _make_log(error_log, 40) # logging.ERROR self.set_environ(environ) self.set_max_accept() def set_environ(self, environ=None): if environ is not None: self.environ = environ environ_update = getattr(self, 'environ', None) self.environ = self.environ_class(self.base_env) if self.ssl_enabled: self.environ['wsgi.url_scheme'] = 'https' else: self.environ['wsgi.url_scheme'] = 'http' if environ_update is not None: self.environ.update(environ_update) if self.environ.get('wsgi.errors') is None: self.environ['wsgi.errors'] = self.error_log def set_max_accept(self): if self.environ.get('wsgi.multiprocess'): self.max_accept = 1 def get_environ(self): return self.environ_class(self.environ) def init_socket(self): StreamServer.init_socket(self) self.update_environ() def update_environ(self): """ Called before the first request is handled to fill in WSGI environment values. This includes getting the correct server name and port. """ address = self.address if isinstance(address, tuple): if 'SERVER_NAME' not in self.environ: try: name = socket.getfqdn(address[0]) except socket.error: name = str(address[0]) if not isinstance(name, str): name = name.decode('ascii') self.environ['SERVER_NAME'] = name self.environ.setdefault('SERVER_PORT', str(address[1])) else: self.environ.setdefault('SERVER_NAME', '') self.environ.setdefault('SERVER_PORT', '') def handle(self, sock, address): """ Create an instance of :attr:`handler_class` to handle the request. This method blocks until the handler returns. """ # pylint:disable=method-hidden handler = self.handler_class(sock, address, self) handler.handle() def _main(): # Provisional main handler, for quick tests, not production # usage. from gevent import monkey; monkey.patch_all() import argparse import importlib parser = argparse.ArgumentParser() parser.add_argument("app", help="dotted name of WSGI app callable [module:callable]") parser.add_argument("-b", "--bind", help="The socket to bind", default=":8080") args = parser.parse_args() module_name, app_name = args.app.split(':') module = importlib.import_module(module_name) app = getattr(module, app_name) bind = args.bind server = WSGIServer(bind, app) server.serve_forever() if __name__ == '__main__': _main()