# -*- test-case-name: twisted.web.test.test_http -*- # Copyright (c) Twisted Matrix Laboratories. # See LICENSE for details. """ HyperText Transfer Protocol implementation. This is the basic server-side protocol implementation used by the Twisted Web server. It can parse HTTP 1.0 requests and supports many HTTP 1.1 features as well. Additionally, some functionality implemented here is also useful for HTTP clients (such as the chunked encoding parser). @var CACHED: A marker value to be returned from cache-related request methods to indicate to the caller that a cached response will be usable and no response body should be generated. @var FOUND: An HTTP response code indicating a temporary redirect. @var NOT_MODIFIED: An HTTP response code indicating that a requested pre-condition (for example, the condition represented by an I{If-Modified-Since} header is present in the request) has succeeded. This indicates a response body cached by the client can be used. @var PRECONDITION_FAILED: An HTTP response code indicating that a requested pre-condition (for example, the condition represented by an I{If-None-Match} header is present in the request) has failed. This should typically indicate that the server has not taken the requested action. @var maxChunkSizeLineLength: Maximum allowable length of the CRLF-terminated line that indicates the size of a chunk and the extensions associated with it, as in the HTTP 1.1 chunked I{Transfer-Encoding} (RFC 7230 section 4.1). This limits how much data may be buffered when decoding the line. """ __all__ = [ "SWITCHING", "OK", "CREATED", "ACCEPTED", "NON_AUTHORITATIVE_INFORMATION", "NO_CONTENT", "RESET_CONTENT", "PARTIAL_CONTENT", "MULTI_STATUS", "MULTIPLE_CHOICE", "MOVED_PERMANENTLY", "FOUND", "SEE_OTHER", "NOT_MODIFIED", "USE_PROXY", "TEMPORARY_REDIRECT", "PERMANENT_REDIRECT", "BAD_REQUEST", "UNAUTHORIZED", "PAYMENT_REQUIRED", "FORBIDDEN", "NOT_FOUND", "NOT_ALLOWED", "NOT_ACCEPTABLE", "PROXY_AUTH_REQUIRED", "REQUEST_TIMEOUT", "CONFLICT", "GONE", "LENGTH_REQUIRED", "PRECONDITION_FAILED", "REQUEST_ENTITY_TOO_LARGE", "REQUEST_URI_TOO_LONG", "UNSUPPORTED_MEDIA_TYPE", "REQUESTED_RANGE_NOT_SATISFIABLE", "EXPECTATION_FAILED", "INTERNAL_SERVER_ERROR", "NOT_IMPLEMENTED", "BAD_GATEWAY", "SERVICE_UNAVAILABLE", "GATEWAY_TIMEOUT", "HTTP_VERSION_NOT_SUPPORTED", "INSUFFICIENT_STORAGE_SPACE", "NOT_EXTENDED", "RESPONSES", "CACHED", "urlparse", "parse_qs", "datetimeToString", "datetimeToLogString", "timegm", "stringToDatetime", "toChunk", "fromChunk", "parseContentRange", "StringTransport", "HTTPClient", "NO_BODY_CODES", "Request", "PotentialDataLoss", "HTTPChannel", "HTTPFactory", ] import base64 import binascii import calendar import cgi import math import os import re import tempfile import time import warnings from io import BytesIO from typing import AnyStr, Callable, Optional, Tuple from urllib.parse import ( ParseResultBytes, unquote_to_bytes as unquote, urlparse as _urlparse, ) from zope.interface import Attribute, Interface, implementer, provider from incremental import Version from twisted.internet import address, interfaces, protocol from twisted.internet._producer_helpers import _PullToPush from twisted.internet.defer import Deferred from twisted.internet.interfaces import IProtocol from twisted.logger import Logger from twisted.protocols import basic, policies from twisted.python import log from twisted.python.compat import _PY37PLUS, nativeString, networkString from twisted.python.components import proxyForInterface from twisted.python.deprecate import deprecated from twisted.python.failure import Failure # twisted imports from twisted.web._responses import ( ACCEPTED, BAD_GATEWAY, BAD_REQUEST, CONFLICT, CREATED, EXPECTATION_FAILED, FORBIDDEN, FOUND, GATEWAY_TIMEOUT, GONE, HTTP_VERSION_NOT_SUPPORTED, INSUFFICIENT_STORAGE_SPACE, INTERNAL_SERVER_ERROR, LENGTH_REQUIRED, MOVED_PERMANENTLY, MULTI_STATUS, MULTIPLE_CHOICE, NO_CONTENT, NON_AUTHORITATIVE_INFORMATION, NOT_ACCEPTABLE, NOT_ALLOWED, NOT_EXTENDED, NOT_FOUND, NOT_IMPLEMENTED, NOT_MODIFIED, OK, PARTIAL_CONTENT, PAYMENT_REQUIRED, PERMANENT_REDIRECT, PRECONDITION_FAILED, PROXY_AUTH_REQUIRED, REQUEST_ENTITY_TOO_LARGE, REQUEST_TIMEOUT, REQUEST_URI_TOO_LONG, REQUESTED_RANGE_NOT_SATISFIABLE, RESET_CONTENT, RESPONSES, SEE_OTHER, SERVICE_UNAVAILABLE, SWITCHING, TEMPORARY_REDIRECT, UNAUTHORIZED, UNSUPPORTED_MEDIA_TYPE, USE_PROXY, ) from twisted.web.http_headers import Headers, _sanitizeLinearWhitespace from twisted.web.iweb import IAccessLogFormatter, INonQueuedRequestFactory, IRequest try: from twisted.web._http2 import H2Connection H2_ENABLED = True except ImportError: H2_ENABLED = False # A common request timeout -- 1 minute. This is roughly what nginx uses, and # so it seems to be a good choice for us too. _REQUEST_TIMEOUT = 1 * 60 protocol_version = "HTTP/1.1" CACHED = """Magic constant returned by http.Request methods to set cache validation headers when the request is conditional and the value fails the condition.""" # backwards compatibility responses = RESPONSES # datetime parsing and formatting weekdayname = ["Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"] monthname = [ None, "Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", ] weekdayname_lower = [name.lower() for name in weekdayname] monthname_lower = [name and name.lower() for name in monthname] def _parseHeader(line): # cgi.parse_header requires a str key, pdict = cgi.parse_header(line.decode("charmap")) # We want the key as bytes, and cgi.parse_multipart (which consumes # pdict) expects a dict of str keys but bytes values key = key.encode("charmap") pdict = {x: y.encode("charmap") for x, y in pdict.items()} return (key, pdict) def urlparse(url): """ Parse an URL into six components. This is similar to C{urlparse.urlparse}, but rejects C{str} input and always produces C{bytes} output. @type url: C{bytes} @raise TypeError: The given url was a C{str} string instead of a C{bytes}. @return: The scheme, net location, path, params, query string, and fragment of the URL - all as C{bytes}. @rtype: C{ParseResultBytes} """ if isinstance(url, str): raise TypeError("url must be bytes, not unicode") scheme, netloc, path, params, query, fragment = _urlparse(url) if isinstance(scheme, str): scheme = scheme.encode("ascii") netloc = netloc.encode("ascii") path = path.encode("ascii") query = query.encode("ascii") fragment = fragment.encode("ascii") return ParseResultBytes(scheme, netloc, path, params, query, fragment) def parse_qs(qs, keep_blank_values=0, strict_parsing=0): """ Like C{cgi.parse_qs}, but with support for parsing byte strings on Python 3. @type qs: C{bytes} """ d = {} items = [s2 for s1 in qs.split(b"&") for s2 in s1.split(b";")] for item in items: try: k, v = item.split(b"=", 1) except ValueError: if strict_parsing: raise continue if v or keep_blank_values: k = unquote(k.replace(b"+", b" ")) v = unquote(v.replace(b"+", b" ")) if k in d: d[k].append(v) else: d[k] = [v] return d def datetimeToString(msSinceEpoch=None): """ Convert seconds since epoch to HTTP datetime string. @rtype: C{bytes} """ if msSinceEpoch == None: msSinceEpoch = time.time() year, month, day, hh, mm, ss, wd, y, z = time.gmtime(msSinceEpoch) s = networkString( "%s, %02d %3s %4d %02d:%02d:%02d GMT" % (weekdayname[wd], day, monthname[month], year, hh, mm, ss) ) return s def datetimeToLogString(msSinceEpoch=None): """ Convert seconds since epoch to log datetime string. @rtype: C{str} """ if msSinceEpoch == None: msSinceEpoch = time.time() year, month, day, hh, mm, ss, wd, y, z = time.gmtime(msSinceEpoch) s = "[%02d/%3s/%4d:%02d:%02d:%02d +0000]" % ( day, monthname[month], year, hh, mm, ss, ) return s def timegm(year, month, day, hour, minute, second): """ Convert time tuple in GMT to seconds since epoch, GMT """ EPOCH = 1970 if year < EPOCH: raise ValueError("Years prior to %d not supported" % (EPOCH,)) assert 1 <= month <= 12 days = 365 * (year - EPOCH) + calendar.leapdays(EPOCH, year) for i in range(1, month): days = days + calendar.mdays[i] if month > 2 and calendar.isleap(year): days = days + 1 days = days + day - 1 hours = days * 24 + hour minutes = hours * 60 + minute seconds = minutes * 60 + second return seconds def stringToDatetime(dateString): """ Convert an HTTP date string (one of three formats) to seconds since epoch. @type dateString: C{bytes} """ parts = nativeString(dateString).split() if not parts[0][0:3].lower() in weekdayname_lower: # Weekday is stupid. Might have been omitted. try: return stringToDatetime(b"Sun, " + dateString) except ValueError: # Guess not. pass partlen = len(parts) if (partlen == 5 or partlen == 6) and parts[1].isdigit(): # 1st date format: Sun, 06 Nov 1994 08:49:37 GMT # (Note: "GMT" is literal, not a variable timezone) # (also handles without "GMT") # This is the normal format day = parts[1] month = parts[2] year = parts[3] time = parts[4] elif (partlen == 3 or partlen == 4) and parts[1].find("-") != -1: # 2nd date format: Sunday, 06-Nov-94 08:49:37 GMT # (Note: "GMT" is literal, not a variable timezone) # (also handles without without "GMT") # Two digit year, yucko. day, month, year = parts[1].split("-") time = parts[2] year = int(year) if year < 69: year = year + 2000 elif year < 100: year = year + 1900 elif len(parts) == 5: # 3rd date format: Sun Nov 6 08:49:37 1994 # ANSI C asctime() format. day = parts[2] month = parts[1] year = parts[4] time = parts[3] else: raise ValueError("Unknown datetime format %r" % dateString) day = int(day) month = int(monthname_lower.index(month.lower())) year = int(year) hour, min, sec = map(int, time.split(":")) return int(timegm(year, month, day, hour, min, sec)) def toChunk(data): """ Convert string to a chunk. @type data: C{bytes} @returns: a tuple of C{bytes} representing the chunked encoding of data """ return (networkString(f"{len(data):x}"), b"\r\n", data, b"\r\n") def _ishexdigits(b: bytes) -> bool: """ Is the string case-insensitively hexidecimal? It must be composed of one or more characters in the ranges a-f, A-F and 0-9. """ for c in b: if c not in b"0123456789abcdefABCDEF": return False return b != b"" def _hexint(b: bytes) -> int: """ Decode a hexadecimal integer. Unlike L{int(b, 16)}, this raises L{ValueError} when the integer has a prefix like C{b'0x'}, C{b'+'}, or C{b'-'}, which is desirable when parsing network protocols. """ if not _ishexdigits(b): raise ValueError(b) return int(b, 16) def fromChunk(data: bytes) -> Tuple[bytes, bytes]: """ Convert chunk to string. Note that this function is not specification compliant: it doesn't handle chunk extensions. @type data: C{bytes} @return: tuple of (result, remaining) - both C{bytes}. @raise ValueError: If the given data is not a correctly formatted chunked byte string. """ prefix, rest = data.split(b"\r\n", 1) length = _hexint(prefix) if length < 0: raise ValueError("Chunk length must be >= 0, not %d" % (length,)) if rest[length : length + 2] != b"\r\n": raise ValueError("chunk must end with CRLF") return rest[:length], rest[length + 2 :] def parseContentRange(header): """ Parse a content-range header into (start, end, realLength). realLength might be None if real length is not known ('*'). """ kind, other = header.strip().split() if kind.lower() != "bytes": raise ValueError("a range of type %r is not supported") startend, realLength = other.split("/") start, end = map(int, startend.split("-")) if realLength == "*": realLength = None else: realLength = int(realLength) return (start, end, realLength) class _IDeprecatedHTTPChannelToRequestInterface(Interface): """ The interface L{HTTPChannel} expects of L{Request}. """ requestHeaders = Attribute( "A L{http_headers.Headers} instance giving all received HTTP request " "headers." ) responseHeaders = Attribute( "A L{http_headers.Headers} instance holding all HTTP response " "headers to be sent." ) def connectionLost(reason): """ The underlying connection has been lost. @param reason: A failure instance indicating the reason why the connection was lost. @type reason: L{twisted.python.failure.Failure} """ def gotLength(length): """ Called when L{HTTPChannel} has determined the length, if any, of the incoming request's body. @param length: The length of the request's body. @type length: L{int} if the request declares its body's length and L{None} if it does not. """ def handleContentChunk(data): """ Deliver a received chunk of body data to the request. Note this does not imply chunked transfer encoding. @param data: The received chunk. @type data: L{bytes} """ def parseCookies(): """ Parse the request's cookies out of received headers. """ def requestReceived(command, path, version): """ Called when the entire request, including its body, has been received. @param command: The request's HTTP command. @type command: L{bytes} @param path: The request's path. Note: this is actually what RFC7320 calls the URI. @type path: L{bytes} @param version: The request's HTTP version. @type version: L{bytes} """ def __eq__(other: object) -> bool: """ Determines if two requests are the same object. @param other: Another object whose identity will be compared to this instance's. @return: L{True} when the two are the same object and L{False} when not. """ def __ne__(other: object) -> bool: """ Determines if two requests are not the same object. @param other: Another object whose identity will be compared to this instance's. @return: L{True} when the two are not the same object and L{False} when they are. """ def __hash__(): """ Generate a hash value for the request. @return: The request's hash value. @rtype: L{int} """ class StringTransport: """ I am a BytesIO wrapper that conforms for the transport API. I support the `writeSequence' method. """ def __init__(self): self.s = BytesIO() def writeSequence(self, seq): self.s.write(b"".join(seq)) def __getattr__(self, attr): return getattr(self.__dict__["s"], attr) class HTTPClient(basic.LineReceiver): """ A client for HTTP 1.0. Notes: You probably want to send a 'Host' header with the name of the site you're connecting to, in order to not break name based virtual hosting. @ivar length: The length of the request body in bytes. @type length: C{int} @ivar firstLine: Are we waiting for the first header line? @type firstLine: C{bool} @ivar __buffer: The buffer that stores the response to the HTTP request. @type __buffer: A C{BytesIO} object. @ivar _header: Part or all of an HTTP request header. @type _header: C{bytes} """ length = None firstLine = True __buffer = None _header = b"" def sendCommand(self, command, path): self.transport.writeSequence([command, b" ", path, b" HTTP/1.0\r\n"]) def sendHeader(self, name, value): if not isinstance(value, bytes): # XXX Deprecate this case value = networkString(str(value)) santizedName = _sanitizeLinearWhitespace(name) santizedValue = _sanitizeLinearWhitespace(value) self.transport.writeSequence([santizedName, b": ", santizedValue, b"\r\n"]) def endHeaders(self): self.transport.write(b"\r\n") def extractHeader(self, header): """ Given a complete HTTP header, extract the field name and value and process the header. @param header: a complete HTTP request header of the form 'field-name: value'. @type header: C{bytes} """ key, val = header.split(b":", 1) val = val.lstrip() self.handleHeader(key, val) if key.lower() == b"content-length": self.length = int(val) def lineReceived(self, line): """ Parse the status line and headers for an HTTP request. @param line: Part of an HTTP request header. Request bodies are parsed in L{HTTPClient.rawDataReceived}. @type line: C{bytes} """ if self.firstLine: self.firstLine = False l = line.split(None, 2) version = l[0] status = l[1] try: message = l[2] except IndexError: # sometimes there is no message message = b"" self.handleStatus(version, status, message) return if not line: if self._header != b"": # Only extract headers if there are any self.extractHeader(self._header) self.__buffer = BytesIO() self.handleEndHeaders() self.setRawMode() return if line.startswith(b"\t") or line.startswith(b" "): # This line is part of a multiline header. According to RFC 822, in # "unfolding" multiline headers you do not strip the leading # whitespace on the continuing line. self._header = self._header + line elif self._header: # This line starts a new header, so process the previous one. self.extractHeader(self._header) self._header = line else: # First header self._header = line def connectionLost(self, reason): self.handleResponseEnd() def handleResponseEnd(self): """ The response has been completely received. This callback may be invoked more than once per request. """ if self.__buffer is not None: b = self.__buffer.getvalue() self.__buffer = None self.handleResponse(b) def handleResponsePart(self, data): self.__buffer.write(data) def connectionMade(self): pass def handleStatus(self, version, status, message): """ Called when the status-line is received. @param version: e.g. 'HTTP/1.0' @param status: e.g. '200' @type status: C{bytes} @param message: e.g. 'OK' """ def handleHeader(self, key, val): """ Called every time a header is received. """ def handleEndHeaders(self): """ Called when all headers have been received. """ def rawDataReceived(self, data): if self.length is not None: data, rest = data[: self.length], data[self.length :] self.length -= len(data) else: rest = b"" self.handleResponsePart(data) if self.length == 0: self.handleResponseEnd() self.setLineMode(rest) # response codes that must have empty bodies NO_BODY_CODES = (204, 304) # Sentinel object that detects people explicitly passing `queued` to Request. _QUEUED_SENTINEL = object() def _getContentFile(length): """ Get a writeable file-like object to which request content can be written. """ if length is not None and length < 100000: return BytesIO() return tempfile.TemporaryFile() _hostHeaderExpression = re.compile(br"^\[?(?P.*?)\]?(:\d+)?$") @implementer(interfaces.IConsumer, _IDeprecatedHTTPChannelToRequestInterface) class Request: """ A HTTP request. Subclasses should override the process() method to determine how the request will be processed. @ivar method: The HTTP method that was used, e.g. C{b'GET'}. @type method: L{bytes} @ivar uri: The full encoded URI which was requested (including query arguments), e.g. C{b'/a/b%20/c?q=v'}. @type uri: L{bytes} @ivar path: The encoded path of the request URI (not including query arguments), e.g. C{b'/a/b%20/c'}. @type path: L{bytes} @ivar args: A mapping of decoded query argument names as L{bytes} to corresponding query argument values as L{list}s of L{bytes}. For example, for a URI with C{foo=bar&foo=baz&quux=spam} as its query part C{args} will be C{{b'foo': [b'bar', b'baz'], b'quux': [b'spam']}}. @type args: L{dict} of L{bytes} to L{list} of L{bytes} @ivar content: A file-like object giving the request body. This may be a file on disk, an L{io.BytesIO}, or some other type. The implementation is free to decide on a per-request basis. @type content: L{typing.BinaryIO} @ivar cookies: The cookies that will be sent in the response. @type cookies: L{list} of L{bytes} @type requestHeaders: L{http_headers.Headers} @ivar requestHeaders: All received HTTP request headers. @type responseHeaders: L{http_headers.Headers} @ivar responseHeaders: All HTTP response headers to be sent. @ivar notifications: A L{list} of L{Deferred}s which are waiting for notification that the response to this request has been finished (successfully or with an error). Don't use this attribute directly, instead use the L{Request.notifyFinish} method. @ivar _disconnected: A flag which is C{False} until the connection over which this request was received is closed and which is C{True} after that. @type _disconnected: L{bool} @ivar _log: A logger instance for request related messages. @type _log: L{twisted.logger.Logger} """ producer = None finished = 0 code = OK code_message = RESPONSES[OK] method = b"(no method yet)" clientproto = b"(no clientproto yet)" uri = b"(no uri yet)" startedWriting = 0 chunked = 0 sentLength = 0 # content-length of response, or total bytes sent via chunking etag = None lastModified = None args = None path = None content = None _forceSSL = 0 _disconnected = False _log = Logger() def __init__(self, channel, queued=_QUEUED_SENTINEL): """ @param channel: the channel we're connected to. @param queued: (deprecated) are we in the request queue, or can we start writing to the transport? """ self.notifications = [] self.channel = channel # Cache the client and server information, we'll need this # later to be serialized and sent with the request so CGIs # will work remotely self.client = self.channel.getPeer() self.host = self.channel.getHost() self.requestHeaders: Headers = Headers() self.received_cookies = {} self.responseHeaders = Headers() self.cookies = [] # outgoing cookies self.transport = self.channel.transport if queued is _QUEUED_SENTINEL: queued = False self.queued = queued def _cleanup(self): """ Called when have finished responding and are no longer queued. """ if self.producer: self._log.failure( "", Failure(RuntimeError(f"Producer was not unregistered for {self.uri}")), ) self.unregisterProducer() self.channel.requestDone(self) del self.channel if self.content is not None: try: self.content.close() except OSError: # win32 suckiness, no idea why it does this pass del self.content for d in self.notifications: d.callback(None) self.notifications = [] # methods for channel - end users should not use these @deprecated(Version("Twisted", 16, 3, 0)) def noLongerQueued(self): """ Notify the object that it is no longer queued. We start writing whatever data we have to the transport, etc. This method is not intended for users. In 16.3 this method was changed to become a no-op, as L{Request} objects are now never queued. """ pass def gotLength(self, length): """ Called when HTTP channel got length of content in this request. This method is not intended for users. @param length: The length of the request body, as indicated by the request headers. L{None} if the request headers do not indicate a length. """ self.content = _getContentFile(length) def parseCookies(self): """ Parse cookie headers. This method is not intended for users. """ cookieheaders = self.requestHeaders.getRawHeaders(b"cookie") if cookieheaders is None: return for cookietxt in cookieheaders: if cookietxt: for cook in cookietxt.split(b";"): cook = cook.lstrip() try: k, v = cook.split(b"=", 1) self.received_cookies[k] = v except ValueError: pass def handleContentChunk(self, data): """ Write a chunk of data. This method is not intended for users. """ self.content.write(data) def requestReceived(self, command, path, version): """ Called by channel when all data has been received. This method is not intended for users. @type command: C{bytes} @param command: The HTTP verb of this request. This has the case supplied by the client (eg, it maybe "get" rather than "GET"). @type path: C{bytes} @param path: The URI of this request. @type version: C{bytes} @param version: The HTTP version of this request. """ clength = self.content.tell() self.content.seek(0, 0) self.args = {} self.method, self.uri = command, path self.clientproto = version x = self.uri.split(b"?", 1) if len(x) == 1: self.path = self.uri else: self.path, argstring = x self.args = parse_qs(argstring, 1) # Argument processing args = self.args ctype = self.requestHeaders.getRawHeaders(b"content-type") if ctype is not None: ctype = ctype[0] if self.method == b"POST" and ctype and clength: mfd = b"multipart/form-data" key, pdict = _parseHeader(ctype) # This weird CONTENT-LENGTH param is required by # cgi.parse_multipart() in some versions of Python 3.7+, see # bpo-29979. It looks like this will be relaxed and backported, see # https://github.com/python/cpython/pull/8530. pdict["CONTENT-LENGTH"] = clength if key == b"application/x-www-form-urlencoded": args.update(parse_qs(self.content.read(), 1)) elif key == mfd: try: if _PY37PLUS: cgiArgs = cgi.parse_multipart( self.content, pdict, encoding="utf8", errors="surrogateescape", ) else: cgiArgs = cgi.parse_multipart(self.content, pdict) if _PY37PLUS: # The parse_multipart function on Python 3.7+ # decodes the header bytes as iso-8859-1 and # decodes the body bytes as utf8 with # surrogateescape -- we want bytes self.args.update( { x.encode("iso-8859-1"): [ z.encode("utf8", "surrogateescape") if isinstance(z, str) else z for z in y ] for x, y in cgiArgs.items() if isinstance(x, str) } ) else: # The parse_multipart function on Python 3 # decodes the header bytes as iso-8859-1 and # returns a str key -- we want bytes so encode # it back self.args.update( {x.encode("iso-8859-1"): y for x, y in cgiArgs.items()} ) except Exception as e: # It was a bad request, or we got a signal. self.channel._respondToBadRequestAndDisconnect() if isinstance(e, (TypeError, ValueError, KeyError)): return else: # If it's not a userspace error from CGI, reraise raise self.content.seek(0, 0) self.process() def __repr__(self) -> str: """ Return a string description of the request including such information as the request method and request URI. @return: A string loosely describing this L{Request} object. @rtype: L{str} """ return "<{} at 0x{:x} method={} uri={} clientproto={}>".format( self.__class__.__name__, id(self), nativeString(self.method), nativeString(self.uri), nativeString(self.clientproto), ) def process(self): """ Override in subclasses. This method is not intended for users. """ pass # consumer interface def registerProducer(self, producer, streaming): """ Register a producer. """ if self.producer: raise ValueError( "registering producer %s before previous one (%s) was " "unregistered" % (producer, self.producer) ) self.streamingProducer = streaming self.producer = producer self.channel.registerProducer(producer, streaming) def unregisterProducer(self): """ Unregister the producer. """ self.channel.unregisterProducer() self.producer = None # The following is the public interface that people should be # writing to. def getHeader(self, key: AnyStr) -> Optional[AnyStr]: """ Get an HTTP request header. @type key: C{bytes} or C{str} @param key: The name of the header to get the value of. @rtype: C{bytes} or C{str} or L{None} @return: The value of the specified header, or L{None} if that header was not present in the request. The string type of the result matches the type of C{key}. """ value = self.requestHeaders.getRawHeaders(key) if value is not None: return value[-1] return None def getCookie(self, key): """ Get a cookie that was sent from the network. @type key: C{bytes} @param key: The name of the cookie to get. @rtype: C{bytes} or C{None} @returns: The value of the specified cookie, or L{None} if that cookie was not present in the request. """ return self.received_cookies.get(key) def notifyFinish(self): """ Notify when the response to this request has finished. @note: There are some caveats around the reliability of the delivery of this notification. 1. If this L{Request}'s channel is paused, the notification will not be delivered. This can happen in one of two ways; either you can call C{request.transport.pauseProducing} yourself, or, 2. In order to deliver this notification promptly when a client disconnects, the reactor must continue reading from the transport, so that it can tell when the underlying network connection has gone away. Twisted Web will only keep reading up until a finite (small) maximum buffer size before it gives up and pauses the transport itself. If this occurs, you will not discover that the connection has gone away until a timeout fires or until the application attempts to send some data via L{Request.write}. 3. It is theoretically impossible to distinguish between successfully I{sending} a response and the peer successfully I{receiving} it. There are several networking edge cases where the L{Deferred}s returned by C{notifyFinish} will indicate success, but the data will never be received. There are also edge cases where the connection will appear to fail, but in reality the response was delivered. As a result, the information provided by the result of the L{Deferred}s returned by this method should be treated as a guess; do not make critical decisions in your applications based upon it. @rtype: L{Deferred} @return: A L{Deferred} which will be triggered when the request is finished -- with a L{None} value if the request finishes successfully or with an error if the request is interrupted by an error (for example, the client closing the connection prematurely). """ self.notifications.append(Deferred()) return self.notifications[-1] def finish(self): """ Indicate that all response data has been written to this L{Request}. """ if self._disconnected: raise RuntimeError( "Request.finish called on a request after its connection was lost; " "use Request.notifyFinish to keep track of this." ) if self.finished: warnings.warn("Warning! request.finish called twice.", stacklevel=2) return if not self.startedWriting: # write headers self.write(b"") if self.chunked: # write last chunk and closing CRLF self.channel.write(b"0\r\n\r\n") # log request if hasattr(self.channel, "factory") and self.channel.factory is not None: self.channel.factory.log(self) self.finished = 1 if not self.queued: self._cleanup() def write(self, data): """ Write some data as a result of an HTTP request. The first time this is called, it writes out response data. @type data: C{bytes} @param data: Some bytes to be sent as part of the response body. """ if self.finished: raise RuntimeError( "Request.write called on a request after " "Request.finish was called." ) if self._disconnected: # Don't attempt to write any data to a disconnected client. # The RuntimeError exception will be thrown as usual when # request.finish is called return if not self.startedWriting: self.startedWriting = 1 version = self.clientproto code = b"%d" % (self.code,) reason = self.code_message headers = [] # if we don't have a content length, we send data in # chunked mode, so that we can support pipelining in # persistent connections. if ( (version == b"HTTP/1.1") and (self.responseHeaders.getRawHeaders(b"content-length") is None) and self.method != b"HEAD" and self.code not in NO_BODY_CODES ): headers.append((b"Transfer-Encoding", b"chunked")) self.chunked = 1 if self.lastModified is not None: if self.responseHeaders.hasHeader(b"last-modified"): self._log.info( "Warning: last-modified specified both in" " header list and lastModified attribute." ) else: self.responseHeaders.setRawHeaders( b"last-modified", [datetimeToString(self.lastModified)] ) if self.etag is not None: self.responseHeaders.setRawHeaders(b"ETag", [self.etag]) for name, values in self.responseHeaders.getAllRawHeaders(): for value in values: headers.append((name, value)) for cookie in self.cookies: headers.append((b"Set-Cookie", cookie)) self.channel.writeHeaders(version, code, reason, headers) # if this is a "HEAD" request, we shouldn't return any data if self.method == b"HEAD": self.write = lambda data: None return # for certain result codes, we should never return any data if self.code in NO_BODY_CODES: self.write = lambda data: None return self.sentLength = self.sentLength + len(data) if data: if self.chunked: self.channel.writeSequence(toChunk(data)) else: self.channel.write(data) def addCookie( self, k, v, expires=None, domain=None, path=None, max_age=None, comment=None, secure=None, httpOnly=False, sameSite=None, ): """ Set an outgoing HTTP cookie. In general, you should consider using sessions instead of cookies, see L{twisted.web.server.Request.getSession} and the L{twisted.web.server.Session} class for details. @param k: cookie name @type k: L{bytes} or L{str} @param v: cookie value @type v: L{bytes} or L{str} @param expires: cookie expire attribute value in "Wdy, DD Mon YYYY HH:MM:SS GMT" format @type expires: L{bytes} or L{str} @param domain: cookie domain @type domain: L{bytes} or L{str} @param path: cookie path @type path: L{bytes} or L{str} @param max_age: cookie expiration in seconds from reception @type max_age: L{bytes} or L{str} @param comment: cookie comment @type comment: L{bytes} or L{str} @param secure: direct browser to send the cookie on encrypted connections only @type secure: L{bool} @param httpOnly: direct browser not to expose cookies through channels other than HTTP (and HTTPS) requests @type httpOnly: L{bool} @param sameSite: One of L{None} (default), C{'lax'} or C{'strict'}. Direct browsers not to send this cookie on cross-origin requests. Please see: U{https://tools.ietf.org/html/draft-west-first-party-cookies-07} @type sameSite: L{None}, L{bytes} or L{str} @raise ValueError: If the value for C{sameSite} is not supported. """ def _ensureBytes(val): """ Ensure that C{val} is bytes, encoding using UTF-8 if needed. @param val: L{bytes} or L{str} @return: L{bytes} """ if val is None: # It's None, so we don't want to touch it return val if isinstance(val, bytes): return val else: return val.encode("utf8") def _sanitize(val): r""" Replace linear whitespace (C{\r}, C{\n}, C{\r\n}) and semicolons C{;} in C{val} with a single space. @param val: L{bytes} @return: L{bytes} """ return _sanitizeLinearWhitespace(val).replace(b";", b" ") cookie = _sanitize(_ensureBytes(k)) + b"=" + _sanitize(_ensureBytes(v)) if expires is not None: cookie = cookie + b"; Expires=" + _sanitize(_ensureBytes(expires)) if domain is not None: cookie = cookie + b"; Domain=" + _sanitize(_ensureBytes(domain)) if path is not None: cookie = cookie + b"; Path=" + _sanitize(_ensureBytes(path)) if max_age is not None: cookie = cookie + b"; Max-Age=" + _sanitize(_ensureBytes(max_age)) if comment is not None: cookie = cookie + b"; Comment=" + _sanitize(_ensureBytes(comment)) if secure: cookie = cookie + b"; Secure" if httpOnly: cookie = cookie + b"; HttpOnly" if sameSite: sameSite = _ensureBytes(sameSite).lower() if sameSite not in [b"lax", b"strict"]: raise ValueError("Invalid value for sameSite: " + repr(sameSite)) cookie += b"; SameSite=" + sameSite self.cookies.append(cookie) def setResponseCode(self, code, message=None): """ Set the HTTP response code. @type code: L{int} @type message: L{bytes} """ if not isinstance(code, int): raise TypeError("HTTP response code must be int or long") self.code = code if message: if not isinstance(message, bytes): raise TypeError("HTTP response status message must be bytes") self.code_message = message else: self.code_message = RESPONSES.get(code, b"Unknown Status") def setHeader(self, name, value): """ Set an HTTP response header. Overrides any previously set values for this header. @type name: L{bytes} or L{str} @param name: The name of the header for which to set the value. @type value: L{bytes} or L{str} @param value: The value to set for the named header. A L{str} will be UTF-8 encoded, which may not interoperable with other implementations. Avoid passing non-ASCII characters if possible. """ self.responseHeaders.setRawHeaders(name, [value]) def redirect(self, url): """ Utility function that does a redirect. Set the response code to L{FOUND} and the I{Location} header to the given URL. The request should have C{finish()} called after this. @param url: I{Location} header value. @type url: L{bytes} or L{str} """ self.setResponseCode(FOUND) self.setHeader(b"location", url) def setLastModified(self, when): """ Set the C{Last-Modified} time for the response to this request. If I am called more than once, I ignore attempts to set Last-Modified earlier, only replacing the Last-Modified time if it is to a later value. If I am a conditional request, I may modify my response code to L{NOT_MODIFIED} if appropriate for the time given. @param when: The last time the resource being returned was modified, in seconds since the epoch. @type when: number @return: If I am a I{If-Modified-Since} conditional request and the time given is not newer than the condition, I return L{http.CACHED} to indicate that you should write no body. Otherwise, I return a false value. """ # time.time() may be a float, but the HTTP-date strings are # only good for whole seconds. when = int(math.ceil(when)) if (not self.lastModified) or (self.lastModified < when): self.lastModified = when modifiedSince = self.getHeader(b"if-modified-since") if modifiedSince: firstPart = modifiedSince.split(b";", 1)[0] try: modifiedSince = stringToDatetime(firstPart) except ValueError: return None if modifiedSince >= self.lastModified: self.setResponseCode(NOT_MODIFIED) return CACHED return None def setETag(self, etag): """ Set an C{entity tag} for the outgoing response. That's \"entity tag\" as in the HTTP/1.1 C{ETag} header, \"used for comparing two or more entities from the same requested resource.\" If I am a conditional request, I may modify my response code to L{NOT_MODIFIED} or L{PRECONDITION_FAILED}, if appropriate for the tag given. @param etag: The entity tag for the resource being returned. @type etag: string @return: If I am a C{If-None-Match} conditional request and the tag matches one in the request, I return L{http.CACHED} to indicate that you should write no body. Otherwise, I return a false value. """ if etag: self.etag = etag tags = self.getHeader(b"if-none-match") if tags: tags = tags.split() if (etag in tags) or (b"*" in tags): self.setResponseCode( ((self.method in (b"HEAD", b"GET")) and NOT_MODIFIED) or PRECONDITION_FAILED ) return CACHED return None def getAllHeaders(self): """ Return dictionary mapping the names of all received headers to the last value received for each. Since this method does not return all header information, C{self.requestHeaders.getAllRawHeaders()} may be preferred. """ headers = {} for k, v in self.requestHeaders.getAllRawHeaders(): headers[k.lower()] = v[-1] return headers def getRequestHostname(self): """ Get the hostname that the HTTP client passed in to the request. @see: L{IRequest.getRequestHostname} @returns: the requested hostname @rtype: C{bytes} """ host = self.getHeader(b"host") if host is not None: match = _hostHeaderExpression.match(host) if match is not None: return match.group("host") return networkString(self.getHost().host) def getHost(self): """ Get my originally requesting transport's host. Don't rely on the 'transport' attribute, since Request objects may be copied remotely. For information on this method's return value, see L{twisted.internet.tcp.Port}. """ return self.host def setHost(self, host, port, ssl=0): """ Change the host and port the request thinks it's using. This method is useful for working with reverse HTTP proxies (e.g. both Squid and Apache's mod_proxy can do this), when the address the HTTP client is using is different than the one we're listening on. For example, Apache may be listening on https://www.example.com/, and then forwarding requests to http://localhost:8080/, but we don't want HTML produced by Twisted to say b'http://localhost:8080/', they should say b'https://www.example.com/', so we do:: request.setHost(b'www.example.com', 443, ssl=1) @type host: C{bytes} @param host: The value to which to change the host header. @type ssl: C{bool} @param ssl: A flag which, if C{True}, indicates that the request is considered secure (if C{True}, L{isSecure} will return C{True}). """ self._forceSSL = ssl # set first so isSecure will work if self.isSecure(): default = 443 else: default = 80 if port == default: hostHeader = host else: hostHeader = b"%b:%d" % (host, port) self.requestHeaders.setRawHeaders(b"host", [hostHeader]) self.host = address.IPv4Address("TCP", host, port) @deprecated(Version("Twisted", 18, 4, 0), replacement="getClientAddress") def getClientIP(self): """ Return the IP address of the client who submitted this request. This method is B{deprecated}. Use L{getClientAddress} instead. @returns: the client IP address @rtype: C{str} """ if isinstance(self.client, (address.IPv4Address, address.IPv6Address)): return self.client.host else: return None def getClientAddress(self): """ Return the address of the client who submitted this request. This may not be a network address (e.g., a server listening on a UNIX domain socket will cause this to return L{UNIXAddress}). Callers must check the type of the returned address. @since: 18.4 @return: the client's address. @rtype: L{IAddress} """ return self.client def isSecure(self): """ Return L{True} if this request is using a secure transport. Normally this method returns L{True} if this request's L{HTTPChannel} instance is using a transport that implements L{interfaces.ISSLTransport}. This will also return L{True} if L{Request.setHost} has been called with C{ssl=True}. @returns: L{True} if this request is secure @rtype: C{bool} """ if self._forceSSL: return True channel = getattr(self, "channel", None) if channel is None: return False return channel.isSecure() def _authorize(self): # Authorization, (mostly) per the RFC try: authh = self.getHeader(b"Authorization") if not authh: self.user = self.password = b"" return bas, upw = authh.split() if bas.lower() != b"basic": raise ValueError() upw = base64.b64decode(upw) self.user, self.password = upw.split(b":", 1) except (binascii.Error, ValueError): self.user = self.password = b"" except BaseException: self._log.failure("") self.user = self.password = b"" def getUser(self): """ Return the HTTP user sent with this request, if any. If no user was supplied, return the empty string. @returns: the HTTP user, if any @rtype: C{bytes} """ try: return self.user except BaseException: pass self._authorize() return self.user def getPassword(self): """ Return the HTTP password sent with this request, if any. If no password was supplied, return the empty string. @returns: the HTTP password, if any @rtype: C{bytes} """ try: return self.password except BaseException: pass self._authorize() return self.password def connectionLost(self, reason): """ There is no longer a connection for this request to respond over. Clean up anything which can't be useful anymore. """ self._disconnected = True self.channel = None if self.content is not None: self.content.close() for d in self.notifications: d.errback(reason) self.notifications = [] def loseConnection(self): """ Pass the loseConnection through to the underlying channel. """ if self.channel is not None: self.channel.loseConnection() def __eq__(self, other: object) -> bool: """ Determines if two requests are the same object. @param other: Another object whose identity will be compared to this instance's. @return: L{True} when the two are the same object and L{False} when not. @rtype: L{bool} """ # When other is not an instance of request, return # NotImplemented so that Python uses other.__eq__ to perform # the comparison. This ensures that a Request proxy generated # by proxyForInterface compares equal to an actual Request # instanceby turning request != proxy into proxy != request. if isinstance(other, Request): return self is other return NotImplemented def __hash__(self): """ A C{Request} is hashable so that it can be used as a mapping key. @return: A C{int} based on the instance's identity. """ return id(self) class _DataLoss(Exception): """ L{_DataLoss} indicates that not all of a message body was received. This is only one of several possible exceptions which may indicate that data was lost. Because of this, it should not be checked for by specifically; any unexpected exception should be treated as having caused data loss. """ class PotentialDataLoss(Exception): """ L{PotentialDataLoss} may be raised by a transfer encoding decoder's C{noMoreData} method to indicate that it cannot be determined if the entire response body has been delivered. This only occurs when making requests to HTTP servers which do not set I{Content-Length} or a I{Transfer-Encoding} in the response because in this case the end of the response is indicated by the connection being closed, an event which may also be due to a transient network problem or other error. """ class _MalformedChunkedDataError(Exception): """ C{_ChunkedTransferDecoder} raises L{_MalformedChunkedDataError} from its C{dataReceived} method when it encounters malformed data. This exception indicates a client-side error. If this exception is raised, the connection should be dropped with a 400 error. """ class _IdentityTransferDecoder: """ Protocol for accumulating bytes up to a specified length. This handles the case where no I{Transfer-Encoding} is specified. @ivar contentLength: Counter keeping track of how many more bytes there are to receive. @ivar dataCallback: A one-argument callable which will be invoked each time application data is received. @ivar finishCallback: A one-argument callable which will be invoked when the terminal chunk is received. It will be invoked with all bytes which were delivered to this protocol which came after the terminal chunk. """ def __init__(self, contentLength, dataCallback, finishCallback): self.contentLength = contentLength self.dataCallback = dataCallback self.finishCallback = finishCallback def dataReceived(self, data): """ Interpret the next chunk of bytes received. Either deliver them to the data callback or invoke the finish callback if enough bytes have been received. @raise RuntimeError: If the finish callback has already been invoked during a previous call to this methood. """ if self.dataCallback is None: raise RuntimeError( "_IdentityTransferDecoder cannot decode data after finishing" ) if self.contentLength is None: self.dataCallback(data) elif len(data) < self.contentLength: self.contentLength -= len(data) self.dataCallback(data) else: # Make the state consistent before invoking any code belonging to # anyone else in case noMoreData ends up being called beneath this # stack frame. contentLength = self.contentLength dataCallback = self.dataCallback finishCallback = self.finishCallback self.dataCallback = self.finishCallback = None self.contentLength = 0 dataCallback(data[:contentLength]) finishCallback(data[contentLength:]) def noMoreData(self): """ All data which will be delivered to this decoder has been. Check to make sure as much data as was expected has been received. @raise PotentialDataLoss: If the content length is unknown. @raise _DataLoss: If the content length is known and fewer than that many bytes have been delivered. @return: L{None} """ finishCallback = self.finishCallback self.dataCallback = self.finishCallback = None if self.contentLength is None: finishCallback(b"") raise PotentialDataLoss() elif self.contentLength != 0: raise _DataLoss() maxChunkSizeLineLength = 1024 _chunkExtChars = ( b"\t !\"#$%&'()*+,-./0123456789:;<=>?@" b"ABCDEFGHIJKLMNOPQRSTUVWXYZ[]^_`" b"abcdefghijklmnopqrstuvwxyz{|}~" b"\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" b"\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" b"\xa0\xa1\xa2\xa3\xa4\xa5\xa6\xa7\xa8\xa9\xaa\xab\xac\xad\xae\xaf" b"\xb0\xb1\xb2\xb3\xb4\xb5\xb6\xb7\xb8\xb9\xba\xbb\xbc\xbd\xbe\xbf" b"\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" b"\xd0\xd1\xd2\xd3\xd4\xd5\xd6\xd7\xd8\xd9\xda\xdb\xdc\xdd\xde\xdf" b"\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" b"\xf0\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\xfb\xfc\xfd\xfe\xff" ) """ Characters that are valid in a chunk extension. See RFC 7230 section 4.1.1:: chunk-ext = *( ";" chunk-ext-name [ "=" chunk-ext-val ] ) chunk-ext-name = token chunk-ext-val = token / quoted-string And section 3.2.6:: token = 1*tchar tchar = "!" / "#" / "$" / "%" / "&" / "'" / "*" / "+" / "-" / "." / "^" / "_" / "`" / "|" / "~" / DIGIT / ALPHA ; any VCHAR, except delimiters quoted-string = DQUOTE *( qdtext / quoted-pair ) DQUOTE qdtext = HTAB / SP /%x21 / %x23-5B / %x5D-7E / obs-text obs-text = %x80-FF We don't check if chunk extensions are well-formed beyond validating that they don't contain characters outside this range. """ class _ChunkedTransferDecoder: """ Protocol for decoding I{chunked} Transfer-Encoding, as defined by RFC 7230, section 4.1. This protocol can interpret the contents of a request or response body which uses the I{chunked} Transfer-Encoding. It cannot interpret any of the rest of the HTTP protocol. It may make sense for _ChunkedTransferDecoder to be an actual IProtocol implementation. Currently, the only user of this class will only ever call dataReceived on it. However, it might be an improvement if the user could connect this to a transport and deliver connection lost notification. This way, `dataCallback` becomes `self.transport.write` and perhaps `finishCallback` becomes `self.transport.loseConnection()` (although I'm not sure where the extra data goes in that case). This could also allow this object to indicate to the receiver of data that the stream was not completely received, an error case which should be noticed. -exarkun @ivar dataCallback: A one-argument callable which will be invoked each time application data is received. This callback is not reentrant. @ivar finishCallback: A one-argument callable which will be invoked when the terminal chunk is received. It will be invoked with all bytes which were delivered to this protocol which came after the terminal chunk. @ivar length: Counter keeping track of how many more bytes in a chunk there are to receive. @ivar state: One of C{'CHUNK_LENGTH'}, C{'CRLF'}, C{'TRAILER'}, C{'BODY'}, or C{'FINISHED'}. For C{'CHUNK_LENGTH'}, data for the chunk length line is currently being read. For C{'CRLF'}, the CR LF pair which follows each chunk is being read. For C{'TRAILER'}, the CR LF pair which follows the terminal 0-length chunk is currently being read. For C{'BODY'}, the contents of a chunk are being read. For C{'FINISHED'}, the last chunk has been completely read and no more input is valid. @ivar _buffer: Accumulated received data for the current state. At each state transition this is truncated at the front so that index 0 is where the next state shall begin. @ivar _start: While in the C{'CHUNK_LENGTH'} state, tracks the index into the buffer at which search for CRLF should resume. Resuming the search at this position avoids doing quadratic work if the chunk length line arrives over many calls to C{dataReceived}. Not used in any other state. """ state = "CHUNK_LENGTH" def __init__( self, dataCallback: Callable[[bytes], None], finishCallback: Callable[[bytes], None], ) -> None: self.dataCallback = dataCallback self.finishCallback = finishCallback self._buffer = bytearray() self._start = 0 def _dataReceived_CHUNK_LENGTH(self) -> bool: """ Read the chunk size line, ignoring any extensions. @returns: C{True} once the line has been read and removed from C{self._buffer}. C{False} when more data is required. @raises _MalformedChunkedDataError: when the chunk size cannot be decoded or the length of the line exceeds L{maxChunkSizeLineLength}. """ eolIndex = self._buffer.find(b"\r\n", self._start) if eolIndex >= maxChunkSizeLineLength or ( eolIndex == -1 and len(self._buffer) > maxChunkSizeLineLength ): raise _MalformedChunkedDataError( "Chunk size line exceeds maximum of {} bytes.".format( maxChunkSizeLineLength ) ) if eolIndex == -1: # Restart the search upon receipt of more data at the start of the # new data, minus one in case the last character of the buffer is # CR. self._start = len(self._buffer) - 1 return False endOfLengthIndex = self._buffer.find(b";", 0, eolIndex) if endOfLengthIndex == -1: endOfLengthIndex = eolIndex rawLength = self._buffer[0:endOfLengthIndex] try: length = _hexint(rawLength) except ValueError: raise _MalformedChunkedDataError("Chunk-size must be an integer.") ext = self._buffer[endOfLengthIndex + 1 : eolIndex] if ext and ext.translate(None, _chunkExtChars) != b"": raise _MalformedChunkedDataError( f"Invalid characters in chunk extensions: {ext!r}." ) if length == 0: self.state = "TRAILER" else: self.state = "BODY" self.length = length del self._buffer[0 : eolIndex + 2] self._start = 0 return True def _dataReceived_CRLF(self) -> bool: """ Await the carriage return and line feed characters that are the end of chunk marker that follow the chunk data. @returns: C{True} when the CRLF have been read, otherwise C{False}. @raises _MalformedChunkedDataError: when anything other than CRLF are received. """ if len(self._buffer) < 2: return False if not self._buffer.startswith(b"\r\n"): raise _MalformedChunkedDataError("Chunk did not end with CRLF") self.state = "CHUNK_LENGTH" del self._buffer[0:2] return True def _dataReceived_TRAILER(self) -> bool: """ Await the carriage return and line feed characters that follow the terminal zero-length chunk. Then invoke C{finishCallback} and switch to state C{'FINISHED'}. @returns: C{False}, as there is either insufficient data to continue, or no data remains. @raises _MalformedChunkedDataError: when anything other than CRLF is received. """ if len(self._buffer) < 2: return False if not self._buffer.startswith(b"\r\n"): raise _MalformedChunkedDataError("Chunk did not end with CRLF") data = memoryview(self._buffer)[2:].tobytes() del self._buffer[:] self.state = "FINISHED" self.finishCallback(data) return False def _dataReceived_BODY(self) -> bool: """ Deliver any available chunk data to the C{dataCallback}. When all the remaining data for the chunk arrives, switch to state C{'CRLF'}. @returns: C{True} to continue processing of any buffered data. """ if len(self._buffer) >= self.length: chunk = memoryview(self._buffer)[: self.length].tobytes() del self._buffer[: self.length] self.state = "CRLF" self.dataCallback(chunk) else: chunk = bytes(self._buffer) self.length -= len(chunk) del self._buffer[:] self.dataCallback(chunk) return True def _dataReceived_FINISHED(self) -> bool: """ Once C{finishCallback} has been invoked receipt of additional data raises L{RuntimeError} because it represents a programming error in the caller. """ raise RuntimeError( "_ChunkedTransferDecoder.dataReceived called after last " "chunk was processed" ) def dataReceived(self, data: bytes) -> None: """ Interpret data from a request or response body which uses the I{chunked} Transfer-Encoding. """ self._buffer += data goOn = True while goOn and self._buffer: goOn = getattr(self, "_dataReceived_" + self.state)() def noMoreData(self) -> None: """ Verify that all data has been received. If it has not been, raise L{_DataLoss}. """ if self.state != "FINISHED": raise _DataLoss( "Chunked decoder in %r state, still expecting more data to " "get to 'FINISHED' state." % (self.state,) ) @implementer(interfaces.IPushProducer) class _NoPushProducer: """ A no-op version of L{interfaces.IPushProducer}, used to abstract over the possibility that a L{HTTPChannel} transport does not provide L{IPushProducer}. """ def pauseProducing(self): """ Pause producing data. Tells a producer that it has produced too much data to process for the time being, and to stop until resumeProducing() is called. """ def resumeProducing(self): """ Resume producing data. This tells a producer to re-add itself to the main loop and produce more data for its consumer. """ def registerProducer(self, producer, streaming): """ Register to receive data from a producer. @param producer: The producer to register. @param streaming: Whether this is a streaming producer or not. """ def unregisterProducer(self): """ Stop consuming data from a producer, without disconnecting. """ def stopProducing(self): """ IProducer.stopProducing """ @implementer(interfaces.ITransport, interfaces.IPushProducer, interfaces.IConsumer) class HTTPChannel(basic.LineReceiver, policies.TimeoutMixin): """ A receiver for HTTP requests. The L{HTTPChannel} provides L{interfaces.ITransport} and L{interfaces.IConsumer} to the L{Request} objects it creates. It also implements L{interfaces.IPushProducer} to C{self.transport}, allowing the transport to pause it. @ivar MAX_LENGTH: Maximum length for initial request line and each line from the header. @ivar _transferDecoder: L{None} or a decoder instance if the request body uses the I{chunked} Transfer-Encoding. @type _transferDecoder: L{_ChunkedTransferDecoder} @ivar maxHeaders: Maximum number of headers allowed per request. @type maxHeaders: C{int} @ivar totalHeadersSize: Maximum bytes for request line plus all headers from the request. @type totalHeadersSize: C{int} @ivar _receivedHeaderSize: Bytes received so far for the header. @type _receivedHeaderSize: C{int} @ivar _handlingRequest: Whether a request is currently being processed. @type _handlingRequest: L{bool} @ivar _dataBuffer: Any data that has been received from the connection while processing an outstanding request. @type _dataBuffer: L{list} of L{bytes} @ivar _networkProducer: Either the transport, if it provides L{interfaces.IPushProducer}, or a null implementation of L{interfaces.IPushProducer}. Used to attempt to prevent the transport from producing excess data when we're responding to a request. @type _networkProducer: L{interfaces.IPushProducer} @ivar _requestProducer: If the L{Request} object or anything it calls registers itself as an L{interfaces.IProducer}, it will be stored here. This is used to create a producing pipeline: pause/resume producing methods will be propagated from the C{transport}, through the L{HTTPChannel} instance, to the c{_requestProducer}. The reason we proxy through the producing methods rather than the old behaviour (where we literally just set the L{Request} object as the producer on the transport) is because we want to be able to exert backpressure on the client to prevent it from sending in arbitrarily many requests without ever reading responses. Essentially, if the client never reads our responses we will eventually stop reading its requests. @type _requestProducer: L{interfaces.IPushProducer} @ivar _requestProducerStreaming: A boolean that tracks whether the producer on the L{Request} side of this channel has registered itself as a L{interfaces.IPushProducer} or an L{interfaces.IPullProducer}. @type _requestProducerStreaming: L{bool} or L{None} @ivar _waitingForTransport: A boolean that tracks whether the transport has asked us to stop producing. This is used to keep track of what we're waiting for: if the transport has asked us to stop producing then we don't want to unpause the transport until it asks us to produce again. @type _waitingForTransport: L{bool} @ivar abortTimeout: The number of seconds to wait after we attempt to shut the transport down cleanly to give up and forcibly terminate it. This is only used when we time a connection out, to prevent errors causing the FD to get leaked. If this is L{None}, we will wait forever. @type abortTimeout: L{int} @ivar _abortingCall: The L{twisted.internet.base.DelayedCall} that will be used to forcibly close the transport if it doesn't close cleanly. @type _abortingCall: L{twisted.internet.base.DelayedCall} @ivar _optimisticEagerReadSize: When a resource takes a long time to answer a request (via L{twisted.web.server.NOT_DONE_YET}, hopefully one day by a L{Deferred}), we would like to be able to let that resource know about the underlying transport disappearing as promptly as possible, via L{Request.notifyFinish}, and therefore via C{self.requests[...].connectionLost()} on this L{HTTPChannel}. However, in order to simplify application logic, we implement head-of-line blocking, and do not relay pipelined requests to the application until the previous request has been answered. This means that said application cannot dispose of any entity-body that comes in from those subsequent requests, which may be arbitrarily large, and it may need to be buffered in memory. To implement this tradeoff between prompt notification when possible (in the most frequent case of non-pipelined requests) and correct behavior when not (say, if a client sends a very long-running GET request followed by a PUT request with a very large body) we will continue reading pipelined requests into C{self._dataBuffer} up to a given limit. C{_optimisticEagerReadSize} is the number of bytes we will accept from the client and buffer before pausing the transport. This behavior has been in place since Twisted 17.9.0 . @type _optimisticEagerReadSize: L{int} """ maxHeaders = 500 totalHeadersSize = 16384 abortTimeout = 15 length = 0 persistent = 1 __header = b"" __first_line = 1 __content = None # set in instances or subclasses requestFactory = Request _savedTimeOut = None _receivedHeaderCount = 0 _receivedHeaderSize = 0 _requestProducer = None _requestProducerStreaming = None _waitingForTransport = False _abortingCall = None _optimisticEagerReadSize = 0x4000 _log = Logger() def __init__(self): # the request queue self.requests = [] self._handlingRequest = False self._dataBuffer = [] self._transferDecoder = None def connectionMade(self): self.setTimeout(self.timeOut) self._networkProducer = interfaces.IPushProducer( self.transport, _NoPushProducer() ) self._networkProducer.registerProducer(self, True) def lineReceived(self, line): """ Called for each line from request until the end of headers when it enters binary mode. """ self.resetTimeout() self._receivedHeaderSize += len(line) if self._receivedHeaderSize > self.totalHeadersSize: self._respondToBadRequestAndDisconnect() return if self.__first_line: # if this connection is not persistent, drop any data which # the client (illegally) sent after the last request. if not self.persistent: self.dataReceived = self.lineReceived = lambda *args: None return # IE sends an extraneous empty line (\r\n) after a POST request; # eat up such a line, but only ONCE if not line and self.__first_line == 1: self.__first_line = 2 return # create a new Request object if INonQueuedRequestFactory.providedBy(self.requestFactory): request = self.requestFactory(self) else: request = self.requestFactory(self, len(self.requests)) self.requests.append(request) self.__first_line = 0 parts = line.split() if len(parts) != 3: self._respondToBadRequestAndDisconnect() return command, request, version = parts try: command.decode("ascii") except UnicodeDecodeError: self._respondToBadRequestAndDisconnect() return self._command = command self._path = request self._version = version elif line == b"": # End of headers. if self.__header: ok = self.headerReceived(self.__header) # If the last header we got is invalid, we MUST NOT proceed # with processing. We'll have sent a 400 anyway, so just stop. if not ok: return self.__header = b"" self.allHeadersReceived() if self.length == 0: self.allContentReceived() else: self.setRawMode() elif line[0] in b" \t": # Continuation of a multi line header. self.__header += b" " + line.lstrip(b" \t") # Regular header line. # Processing of header line is delayed to allow accumulating multi # line headers. else: if self.__header: self.headerReceived(self.__header) self.__header = line def _finishRequestBody(self, data): self.allContentReceived() self._dataBuffer.append(data) def _maybeChooseTransferDecoder(self, header, data): """ If the provided header is C{content-length} or C{transfer-encoding}, choose the appropriate decoder if any. Returns L{True} if the request can proceed and L{False} if not. """ def fail(): self._respondToBadRequestAndDisconnect() self.length = None return False # Can this header determine the length? if header == b"content-length": if not data.isdigit(): return fail() try: length = int(data) except ValueError: return fail() newTransferDecoder = _IdentityTransferDecoder( length, self.requests[-1].handleContentChunk, self._finishRequestBody ) elif header == b"transfer-encoding": # XXX Rather poorly tested code block, apparently only exercised by # test_chunkedEncoding if data.lower() == b"chunked": length = None newTransferDecoder = _ChunkedTransferDecoder( self.requests[-1].handleContentChunk, self._finishRequestBody ) elif data.lower() == b"identity": return True else: return fail() else: # It's not a length related header, so exit return True if self._transferDecoder is not None: return fail() else: self.length = length self._transferDecoder = newTransferDecoder return True def headerReceived(self, line): """ Do pre-processing (for content-length) and store this header away. Enforce the per-request header limit. @type line: C{bytes} @param line: A line from the header section of a request, excluding the line delimiter. @return: A flag indicating whether the header was valid. @rtype: L{bool} """ try: header, data = line.split(b":", 1) except ValueError: self._respondToBadRequestAndDisconnect() return False if not header or header[-1:].isspace(): self._respondToBadRequestAndDisconnect() return False header = header.lower() data = data.strip(b" \t") if not self._maybeChooseTransferDecoder(header, data): return False reqHeaders = self.requests[-1].requestHeaders values = reqHeaders.getRawHeaders(header) if values is not None: values.append(data) else: reqHeaders.setRawHeaders(header, [data]) self._receivedHeaderCount += 1 if self._receivedHeaderCount > self.maxHeaders: self._respondToBadRequestAndDisconnect() return False return True def allContentReceived(self): command = self._command path = self._path version = self._version # reset ALL state variables, so we don't interfere with next request self.length = 0 self._receivedHeaderCount = 0 self._receivedHeaderSize = 0 self.__first_line = 1 self._transferDecoder = None del self._command, self._path, self._version # Disable the idle timeout, in case this request takes a long # time to finish generating output. if self.timeOut: self._savedTimeOut = self.setTimeout(None) self._handlingRequest = True req = self.requests[-1] req.requestReceived(command, path, version) def dataReceived(self, data): """ Data was received from the network. Process it. """ # If we're currently handling a request, buffer this data. if self._handlingRequest: self._dataBuffer.append(data) if ( sum(map(len, self._dataBuffer)) > self._optimisticEagerReadSize ) and not self._waitingForTransport: # If we received more data than a small limit while processing # the head-of-line request, apply TCP backpressure to our peer # to get them to stop sending more request data until we're # ready. See docstring for _optimisticEagerReadSize above. self._networkProducer.pauseProducing() return return basic.LineReceiver.dataReceived(self, data) def rawDataReceived(self, data): self.resetTimeout() try: self._transferDecoder.dataReceived(data) except _MalformedChunkedDataError: self._respondToBadRequestAndDisconnect() def allHeadersReceived(self): req = self.requests[-1] req.parseCookies() self.persistent = self.checkPersistence(req, self._version) req.gotLength(self.length) # Handle 'Expect: 100-continue' with automated 100 response code, # a simplistic implementation of RFC 2686 8.2.3: expectContinue = req.requestHeaders.getRawHeaders(b"expect") if ( expectContinue and expectContinue[0].lower() == b"100-continue" and self._version == b"HTTP/1.1" ): self._send100Continue() def checkPersistence(self, request, version): """ Check if the channel should close or not. @param request: The request most recently received over this channel against which checks will be made to determine if this connection can remain open after a matching response is returned. @type version: C{bytes} @param version: The version of the request. @rtype: C{bool} @return: A flag which, if C{True}, indicates that this connection may remain open to receive another request; if C{False}, the connection must be closed in order to indicate the completion of the response to C{request}. """ connection = request.requestHeaders.getRawHeaders(b"connection") if connection: tokens = [t.lower() for t in connection[0].split(b" ")] else: tokens = [] # Once any HTTP 0.9 or HTTP 1.0 request is received, the connection is # no longer allowed to be persistent. At this point in processing the # request, we don't yet know if it will be possible to set a # Content-Length in the response. If it is not, then the connection # will have to be closed to end an HTTP 0.9 or HTTP 1.0 response. # If the checkPersistence call happened later, after the Content-Length # has been determined (or determined not to be set), it would probably # be possible to have persistent connections with HTTP 0.9 and HTTP 1.0. # This may not be worth the effort, though. Just use HTTP 1.1, okay? if version == b"HTTP/1.1": if b"close" in tokens: request.responseHeaders.setRawHeaders(b"connection", [b"close"]) return False else: return True else: return False def requestDone(self, request): """ Called by first request in queue when it is done. """ if request != self.requests[0]: raise TypeError del self.requests[0] # We should only resume the producer if we're not waiting for the # transport. if not self._waitingForTransport: self._networkProducer.resumeProducing() if self.persistent: self._handlingRequest = False if self._savedTimeOut: self.setTimeout(self._savedTimeOut) # Receive our buffered data, if any. data = b"".join(self._dataBuffer) self._dataBuffer = [] self.setLineMode(data) else: self.loseConnection() def timeoutConnection(self): self._log.info("Timing out client: {peer}", peer=str(self.transport.getPeer())) if self.abortTimeout is not None: # We use self.callLater because that's what TimeoutMixin does. self._abortingCall = self.callLater( self.abortTimeout, self.forceAbortClient ) self.loseConnection() def forceAbortClient(self): """ Called if C{abortTimeout} seconds have passed since the timeout fired, and the connection still hasn't gone away. This can really only happen on extremely bad connections or when clients are maliciously attempting to keep connections open. """ self._log.info( "Forcibly timing out client: {peer}", peer=str(self.transport.getPeer()) ) # We want to lose track of the _abortingCall so that no-one tries to # cancel it. self._abortingCall = None self.transport.abortConnection() def connectionLost(self, reason): self.setTimeout(None) for request in self.requests: request.connectionLost(reason) # If we were going to force-close the transport, we don't have to now. if self._abortingCall is not None: self._abortingCall.cancel() self._abortingCall = None def isSecure(self): """ Return L{True} if this channel is using a secure transport. Normally this method returns L{True} if this instance is using a transport that implements L{interfaces.ISSLTransport}. @returns: L{True} if this request is secure @rtype: C{bool} """ if interfaces.ISSLTransport(self.transport, None) is not None: return True return False def writeHeaders(self, version, code, reason, headers): """ Called by L{Request} objects to write a complete set of HTTP headers to a transport. @param version: The HTTP version in use. @type version: L{bytes} @param code: The HTTP status code to write. @type code: L{bytes} @param reason: The HTTP reason phrase to write. @type reason: L{bytes} @param headers: The headers to write to the transport. @type headers: L{twisted.web.http_headers.Headers} """ sanitizedHeaders = Headers() for name, value in headers: sanitizedHeaders.addRawHeader(name, value) responseLine = version + b" " + code + b" " + reason + b"\r\n" headerSequence = [responseLine] headerSequence.extend( name + b": " + value + b"\r\n" for name, values in sanitizedHeaders.getAllRawHeaders() for value in values ) headerSequence.append(b"\r\n") self.transport.writeSequence(headerSequence) def write(self, data): """ Called by L{Request} objects to write response data. @param data: The data chunk to write to the stream. @type data: L{bytes} @return: L{None} """ self.transport.write(data) def writeSequence(self, iovec): """ Write a list of strings to the HTTP response. @param iovec: A list of byte strings to write to the stream. @type iovec: L{list} of L{bytes} @return: L{None} """ self.transport.writeSequence(iovec) def getPeer(self): """ Get the remote address of this connection. @return: An L{IAddress} provider. """ return self.transport.getPeer() def getHost(self): """ Get the local address of this connection. @return: An L{IAddress} provider. """ return self.transport.getHost() def loseConnection(self): """ Closes the connection. Will write any data that is pending to be sent on the network, but if this response has not yet been written to the network will not write anything. @return: L{None} """ self._networkProducer.unregisterProducer() return self.transport.loseConnection() def registerProducer(self, producer, streaming): """ Register to receive data from a producer. This sets self to be a consumer for a producer. When this object runs out of data (as when a send(2) call on a socket succeeds in moving the last data from a userspace buffer into a kernelspace buffer), it will ask the producer to resumeProducing(). For L{IPullProducer} providers, C{resumeProducing} will be called once each time data is required. For L{IPushProducer} providers, C{pauseProducing} will be called whenever the write buffer fills up and C{resumeProducing} will only be called when it empties. @type producer: L{IProducer} provider @param producer: The L{IProducer} that will be producing data. @type streaming: L{bool} @param streaming: C{True} if C{producer} provides L{IPushProducer}, C{False} if C{producer} provides L{IPullProducer}. @raise RuntimeError: If a producer is already registered. @return: L{None} """ if self._requestProducer is not None: raise RuntimeError( "Cannot register producer %s, because producer %s was never " "unregistered." % (producer, self._requestProducer) ) if not streaming: producer = _PullToPush(producer, self) self._requestProducer = producer self._requestProducerStreaming = streaming if not streaming: producer.startStreaming() def unregisterProducer(self): """ Stop consuming data from a producer, without disconnecting. @return: L{None} """ if self._requestProducer is None: return if not self._requestProducerStreaming: self._requestProducer.stopStreaming() self._requestProducer = None self._requestProducerStreaming = None def stopProducing(self): """ Stop producing data. The HTTPChannel doesn't *actually* implement this, beacuse the assumption is that it will only be called just before C{loseConnection} is called. There's nothing sensible we can do other than call C{loseConnection} anyway. """ if self._requestProducer is not None: self._requestProducer.stopProducing() def pauseProducing(self): """ Pause producing data. This will be called by the transport when the send buffers have been filled up. We want to simultaneously pause the producing L{Request} object and also pause our transport. The logic behind pausing the transport is specifically to avoid issues like https://twistedmatrix.com/trac/ticket/8868. In this case, our inability to send does not prevent us handling more requests, which means we increasingly queue up more responses in our send buffer without end. The easiest way to handle this is to ensure that if we are unable to send our responses, we will not read further data from the connection until the client pulls some data out. This is a bit of a blunt instrument, but it's ok. Note that this potentially interacts with timeout handling in a positive way. Once the transport is paused the client may run into a timeout which will cause us to tear the connection down. That's a good thing! """ self._waitingForTransport = True # The first step is to tell any producer we might currently have # registered to stop producing. If we can slow our applications down # we should. if self._requestProducer is not None: self._requestProducer.pauseProducing() # The next step here is to pause our own transport, as discussed in the # docstring. if not self._handlingRequest: self._networkProducer.pauseProducing() def resumeProducing(self): """ Resume producing data. This will be called by the transport when the send buffer has dropped enough to actually send more data. When this happens we can unpause any outstanding L{Request} producers we have, and also unpause our transport. """ self._waitingForTransport = False if self._requestProducer is not None: self._requestProducer.resumeProducing() # We only want to resume the network producer if we're not currently # waiting for a response to show up. if not self._handlingRequest: self._networkProducer.resumeProducing() def _send100Continue(self): """ Sends a 100 Continue response, used to signal to clients that further processing will be performed. """ self.transport.write(b"HTTP/1.1 100 Continue\r\n\r\n") def _respondToBadRequestAndDisconnect(self): """ This is a quick and dirty way of responding to bad requests. As described by HTTP standard we should be patient and accept the whole request from the client before sending a polite bad request response, even in the case when clients send tons of data. """ self.transport.write(b"HTTP/1.1 400 Bad Request\r\n\r\n") self.loseConnection() def _escape(s): """ Return a string like python repr, but always escaped as if surrounding quotes were double quotes. @param s: The string to escape. @type s: L{bytes} or L{str} @return: An escaped string. @rtype: L{str} """ if not isinstance(s, bytes): s = s.encode("ascii") r = repr(s) if not isinstance(r, str): r = r.decode("ascii") if r.startswith("b"): r = r[1:] if r.startswith("'"): return r[1:-1].replace('"', '\\"').replace("\\'", "'") return r[1:-1] @provider(IAccessLogFormatter) def combinedLogFormatter(timestamp, request): """ @return: A combined log formatted log line for the given request. @see: L{IAccessLogFormatter} """ clientAddr = request.getClientAddress() if isinstance( clientAddr, (address.IPv4Address, address.IPv6Address, _XForwardedForAddress) ): ip = clientAddr.host else: ip = b"-" referrer = _escape(request.getHeader(b"referer") or b"-") agent = _escape(request.getHeader(b"user-agent") or b"-") line = ( '"%(ip)s" - - %(timestamp)s "%(method)s %(uri)s %(protocol)s" ' '%(code)d %(length)s "%(referrer)s" "%(agent)s"' % dict( ip=_escape(ip), timestamp=timestamp, method=_escape(request.method), uri=_escape(request.uri), protocol=_escape(request.clientproto), code=request.code, length=request.sentLength or "-", referrer=referrer, agent=agent, ) ) return line @implementer(interfaces.IAddress) class _XForwardedForAddress: """ L{IAddress} which represents the client IP to log for a request, as gleaned from an X-Forwarded-For header. @ivar host: An IP address or C{b"-"}. @type host: L{bytes} @see: L{proxiedLogFormatter} """ def __init__(self, host): self.host = host class _XForwardedForRequest(proxyForInterface(IRequest, "_request")): # type: ignore[misc] """ Add a layer on top of another request that only uses the value of an X-Forwarded-For header as the result of C{getClientAddress}. """ def getClientAddress(self): """ The client address (the first address) in the value of the I{X-Forwarded-For header}. If the header is not present, the IP is considered to be C{b"-"}. @return: L{_XForwardedForAddress} which wraps the client address as expected by L{combinedLogFormatter}. """ host = ( self._request.requestHeaders.getRawHeaders(b"x-forwarded-for", [b"-"])[0] .split(b",")[0] .strip() ) return _XForwardedForAddress(host) # These are missing from the interface. Forward them manually. @property def clientproto(self): """ @return: The protocol version in the request. @rtype: L{bytes} """ return self._request.clientproto @property def code(self): """ @return: The response code for the request. @rtype: L{int} """ return self._request.code @property def sentLength(self): """ @return: The number of bytes sent in the response body. @rtype: L{int} """ return self._request.sentLength @provider(IAccessLogFormatter) def proxiedLogFormatter(timestamp, request): """ @return: A combined log formatted log line for the given request but use the value of the I{X-Forwarded-For} header as the value for the client IP address. @see: L{IAccessLogFormatter} """ return combinedLogFormatter(timestamp, _XForwardedForRequest(request)) class _GenericHTTPChannelProtocol(proxyForInterface(IProtocol, "_channel")): # type: ignore[misc] """ A proxy object that wraps one of the HTTP protocol objects, and switches between them depending on TLS negotiated protocol. @ivar _negotiatedProtocol: The protocol negotiated with ALPN or NPN, if any. @type _negotiatedProtocol: Either a bytestring containing the ALPN token for the negotiated protocol, or L{None} if no protocol has yet been negotiated. @ivar _channel: The object capable of behaving like a L{HTTPChannel} that is backing this object. By default this is a L{HTTPChannel}, but if a HTTP protocol upgrade takes place this may be a different channel object. Must implement L{IProtocol}. @type _channel: L{HTTPChannel} @ivar _requestFactory: A callable to use to build L{IRequest} objects. @type _requestFactory: L{IRequest} @ivar _site: A reference to the creating L{twisted.web.server.Site} object. @type _site: L{twisted.web.server.Site} @ivar _factory: A reference to the creating L{HTTPFactory} object. @type _factory: L{HTTPFactory} @ivar _timeOut: A timeout value to pass to the backing channel. @type _timeOut: L{int} or L{None} @ivar _callLater: A value for the C{callLater} callback. @type _callLater: L{callable} """ _negotiatedProtocol = None _requestFactory = Request _factory = None _site = None _timeOut = None _callLater = None @property def factory(self): """ @see: L{_genericHTTPChannelProtocolFactory} """ return self._channel.factory @factory.setter def factory(self, value): self._factory = value self._channel.factory = value @property def requestFactory(self): """ A callable to use to build L{IRequest} objects. Retries the object from the current backing channel. """ return self._channel.requestFactory @requestFactory.setter def requestFactory(self, value): """ A callable to use to build L{IRequest} objects. Sets the object on the backing channel and also stores the value for propagation to any new channel. @param value: The new callable to use. @type value: A L{callable} returning L{IRequest} """ self._requestFactory = value self._channel.requestFactory = value @property def site(self): """ A reference to the creating L{twisted.web.server.Site} object. Returns the site object from the backing channel. """ return self._channel.site @site.setter def site(self, value): """ A reference to the creating L{twisted.web.server.Site} object. Sets the object on the backing channel and also stores the value for propagation to any new channel. @param value: The L{twisted.web.server.Site} object to set. @type value: L{twisted.web.server.Site} """ self._site = value self._channel.site = value @property def timeOut(self): """ The idle timeout for the backing channel. """ return self._channel.timeOut @timeOut.setter def timeOut(self, value): """ The idle timeout for the backing channel. Sets the idle timeout on both the backing channel and stores it for propagation to any new backing channel. @param value: The timeout to set. @type value: L{int} or L{float} """ self._timeOut = value self._channel.timeOut = value @property def callLater(self): """ A value for the C{callLater} callback. This callback is used by the L{twisted.protocols.policies.TimeoutMixin} to handle timeouts. """ return self._channel.callLater @callLater.setter def callLater(self, value): """ Sets the value for the C{callLater} callback. This callback is used by the L{twisted.protocols.policies.TimeoutMixin} to handle timeouts. @param value: The new callback to use. @type value: L{callable} """ self._callLater = value self._channel.callLater = value def dataReceived(self, data): """ An override of L{IProtocol.dataReceived} that checks what protocol we're using. """ if self._negotiatedProtocol is None: try: negotiatedProtocol = self._channel.transport.negotiatedProtocol except AttributeError: # Plaintext HTTP, always HTTP/1.1 negotiatedProtocol = b"http/1.1" if negotiatedProtocol is None: negotiatedProtocol = b"http/1.1" if negotiatedProtocol == b"h2": if not H2_ENABLED: raise ValueError("Negotiated HTTP/2 without support.") # We need to make sure that the HTTPChannel is unregistered # from the transport so that the H2Connection can register # itself if possible. networkProducer = self._channel._networkProducer networkProducer.unregisterProducer() # Cancel the old channel's timeout. self._channel.setTimeout(None) transport = self._channel.transport self._channel = H2Connection() self._channel.requestFactory = self._requestFactory self._channel.site = self._site self._channel.factory = self._factory self._channel.timeOut = self._timeOut self._channel.callLater = self._callLater self._channel.makeConnection(transport) # Register the H2Connection as the transport's # producer, so that the transport can apply back # pressure. networkProducer.registerProducer(self._channel, True) else: # Only HTTP/2 and HTTP/1.1 are supported right now. assert ( negotiatedProtocol == b"http/1.1" ), "Unsupported protocol negotiated" self._negotiatedProtocol = negotiatedProtocol return self._channel.dataReceived(data) def _genericHTTPChannelProtocolFactory(self): """ Returns an appropriately initialized _GenericHTTPChannelProtocol. """ return _GenericHTTPChannelProtocol(HTTPChannel()) class HTTPFactory(protocol.ServerFactory): """ Factory for HTTP server. @ivar _logDateTime: A cached datetime string for log messages, updated by C{_logDateTimeCall}. @type _logDateTime: C{str} @ivar _logDateTimeCall: A delayed call for the next update to the cached log datetime string. @type _logDateTimeCall: L{IDelayedCall} provided @ivar _logFormatter: See the C{logFormatter} parameter to L{__init__} @ivar _nativeize: A flag that indicates whether the log file being written to wants native strings (C{True}) or bytes (C{False}). This is only to support writing to L{twisted.python.log} which, unfortunately, works with native strings. @ivar reactor: An L{IReactorTime} provider used to manage connection timeouts and compute logging timestamps. """ # We need to ignore the mypy error here, because # _genericHTTPChannelProtocolFactory is a callable which returns a proxy # to a Protocol, instead of a concrete Protocol object, as expected in # the protocol.Factory interface protocol = _genericHTTPChannelProtocolFactory # type: ignore[assignment] logPath = None timeOut = _REQUEST_TIMEOUT def __init__( self, logPath=None, timeout=_REQUEST_TIMEOUT, logFormatter=None, reactor=None ): """ @param logPath: File path to which access log messages will be written or C{None} to disable logging. @type logPath: L{str} or L{bytes} @param timeout: The initial value of L{timeOut}, which defines the idle connection timeout in seconds, or C{None} to disable the idle timeout. @type timeout: L{float} @param logFormatter: An object to format requests into log lines for the access log. L{combinedLogFormatter} when C{None} is passed. @type logFormatter: L{IAccessLogFormatter} provider @param reactor: An L{IReactorTime} provider used to manage connection timeouts and compute logging timestamps. Defaults to the global reactor. """ if not reactor: from twisted.internet import reactor self.reactor = reactor if logPath is not None: logPath = os.path.abspath(logPath) self.logPath = logPath self.timeOut = timeout if logFormatter is None: logFormatter = combinedLogFormatter self._logFormatter = logFormatter # For storing the cached log datetime and the callback to update it self._logDateTime = None self._logDateTimeCall = None def _updateLogDateTime(self): """ Update log datetime periodically, so we aren't always recalculating it. """ self._logDateTime = datetimeToLogString(self.reactor.seconds()) self._logDateTimeCall = self.reactor.callLater(1, self._updateLogDateTime) def buildProtocol(self, addr): p = protocol.ServerFactory.buildProtocol(self, addr) # This is a bit of a hack to ensure that the HTTPChannel timeouts # occur on the same reactor as the one we're using here. This could # ideally be resolved by passing the reactor more generally to the # HTTPChannel, but that won't work for the TimeoutMixin until we fix # https://twistedmatrix.com/trac/ticket/8488 p.callLater = self.reactor.callLater # timeOut needs to be on the Protocol instance cause # TimeoutMixin expects it there p.timeOut = self.timeOut return p def startFactory(self): """ Set up request logging if necessary. """ if self._logDateTimeCall is None: self._updateLogDateTime() if self.logPath: self.logFile = self._openLogFile(self.logPath) else: self.logFile = log.logfile def stopFactory(self): if hasattr(self, "logFile"): if self.logFile != log.logfile: self.logFile.close() del self.logFile if self._logDateTimeCall is not None and self._logDateTimeCall.active(): self._logDateTimeCall.cancel() self._logDateTimeCall = None def _openLogFile(self, path): """ Override in subclasses, e.g. to use L{twisted.python.logfile}. """ f = open(path, "ab", 1) return f def log(self, request): """ Write a line representing C{request} to the access log file. @param request: The request object about which to log. @type request: L{Request} """ try: logFile = self.logFile except AttributeError: pass else: line = self._logFormatter(self._logDateTime, request) + "\n" logFile.write(line.encode("utf8"))