%PDF- %PDF-
Direktori : /usr/lib/python3/dist-packages/twisted/names/ |
Current File : //usr/lib/python3/dist-packages/twisted/names/dns.py |
# -*- test-case-name: twisted.names.test.test_dns -*- # Copyright (c) Twisted Matrix Laboratories. # See LICENSE for details. """ DNS protocol implementation. Future Plans: - Get rid of some toplevels, maybe. """ # System imports import inspect import random import socket import struct from io import BytesIO from itertools import chain from typing import Optional, SupportsInt, Union from zope.interface import Attribute, Interface, implementer # Twisted imports from twisted.internet import defer, protocol from twisted.internet.error import CannotListenError from twisted.python import failure, log, randbytes, util as tputil from twisted.python.compat import cmp, comparable, nativeString __all__ = [ "IEncodable", "IRecord", "IEncodableRecord", "A", "A6", "AAAA", "AFSDB", "CNAME", "DNAME", "HINFO", "MAILA", "MAILB", "MB", "MD", "MF", "MG", "MINFO", "MR", "MX", "NAPTR", "NS", "NULL", "OPT", "PTR", "RP", "SOA", "SPF", "SRV", "TXT", "SSHFP", "TSIG", "WKS", "ANY", "CH", "CS", "HS", "IN", "ALL_RECORDS", "AXFR", "IXFR", "EFORMAT", "ENAME", "ENOTIMP", "EREFUSED", "ESERVER", "EBADVERSION", "EBADSIG", "EBADKEY", "EBADTIME", "Record_A", "Record_A6", "Record_AAAA", "Record_AFSDB", "Record_CNAME", "Record_DNAME", "Record_HINFO", "Record_MB", "Record_MD", "Record_MF", "Record_MG", "Record_MINFO", "Record_MR", "Record_MX", "Record_NAPTR", "Record_NS", "Record_NULL", "Record_PTR", "Record_RP", "Record_SOA", "Record_SPF", "Record_SRV", "Record_SSHFP", "Record_TSIG", "Record_TXT", "Record_WKS", "UnknownRecord", "QUERY_CLASSES", "QUERY_TYPES", "REV_CLASSES", "REV_TYPES", "EXT_QUERIES", "Charstr", "Message", "Name", "Query", "RRHeader", "SimpleRecord", "DNSDatagramProtocol", "DNSMixin", "DNSProtocol", "OK", "OP_INVERSE", "OP_NOTIFY", "OP_QUERY", "OP_STATUS", "OP_UPDATE", "PORT", "AuthoritativeDomainError", "DNSQueryTimeoutError", "DomainError", ] AF_INET6 = socket.AF_INET6 def _ord2bytes(ordinal): """ Construct a bytes object representing a single byte with the given ordinal value. @type ordinal: L{int} @rtype: L{bytes} """ return bytes([ordinal]) def _nicebytes(bytes): """ Represent a mostly textful bytes object in a way suitable for presentation to an end user. @param bytes: The bytes to represent. @rtype: L{str} """ return repr(bytes)[1:] def _nicebyteslist(list): """ Represent a list of mostly textful bytes objects in a way suitable for presentation to an end user. @param list: The list of bytes to represent. @rtype: L{str} """ return "[{}]".format(", ".join([_nicebytes(b) for b in list])) def randomSource(): """ Wrapper around L{twisted.python.randbytes.RandomFactory.secureRandom} to return 2 random bytes. @rtype: L{bytes} """ return struct.unpack("H", randbytes.secureRandom(2, fallback=True))[0] PORT = 53 ( A, NS, MD, MF, CNAME, SOA, MB, MG, MR, NULL, WKS, PTR, HINFO, MINFO, MX, TXT, RP, AFSDB, ) = range(1, 19) AAAA = 28 SRV = 33 NAPTR = 35 A6 = 38 DNAME = 39 OPT = 41 SSHFP = 44 SPF = 99 # These record types do not exist in zones, but are transferred in # messages the same way normal RRs are. TKEY = 249 TSIG = 250 QUERY_TYPES = { A: "A", NS: "NS", MD: "MD", MF: "MF", CNAME: "CNAME", SOA: "SOA", MB: "MB", MG: "MG", MR: "MR", NULL: "NULL", WKS: "WKS", PTR: "PTR", HINFO: "HINFO", MINFO: "MINFO", MX: "MX", TXT: "TXT", RP: "RP", AFSDB: "AFSDB", # 19 through 27? Eh, I'll get to 'em. AAAA: "AAAA", SRV: "SRV", NAPTR: "NAPTR", A6: "A6", DNAME: "DNAME", OPT: "OPT", SSHFP: "SSHFP", SPF: "SPF", TKEY: "TKEY", TSIG: "TSIG", } IXFR, AXFR, MAILB, MAILA, ALL_RECORDS = range(251, 256) # "Extended" queries (Hey, half of these are deprecated, good job) EXT_QUERIES = { IXFR: "IXFR", AXFR: "AXFR", MAILB: "MAILB", MAILA: "MAILA", ALL_RECORDS: "ALL_RECORDS", } REV_TYPES = {v: k for (k, v) in chain(QUERY_TYPES.items(), EXT_QUERIES.items())} IN, CS, CH, HS = range(1, 5) ANY = 255 QUERY_CLASSES = {IN: "IN", CS: "CS", CH: "CH", HS: "HS", ANY: "ANY"} REV_CLASSES = {v: k for (k, v) in QUERY_CLASSES.items()} # Opcodes OP_QUERY, OP_INVERSE, OP_STATUS = range(3) OP_NOTIFY = 4 # RFC 1996 OP_UPDATE = 5 # RFC 2136 # Response Codes OK, EFORMAT, ESERVER, ENAME, ENOTIMP, EREFUSED = range(6) # https://tools.ietf.org/html/rfc6891#section-9 EBADVERSION = 16 # RFC 2845 EBADSIG, EBADKEY, EBADTIME = range(16, 19) class IRecord(Interface): """ A single entry in a zone of authority. """ TYPE = Attribute("An indicator of what kind of record this is.") # Backwards compatibility aliases - these should be deprecated or something I # suppose. -exarkun from twisted.names.error import ( AuthoritativeDomainError, DNSQueryTimeoutError, DomainError, ) def _nameToLabels(name): """ Split a domain name into its constituent labels. @type name: L{bytes} @param name: A fully qualified domain name (with or without a trailing dot). @return: A L{list} of labels ending with an empty label representing the DNS root zone. @rtype: L{list} of L{bytes} """ if name in (b"", b"."): return [b""] labels = name.split(b".") if labels[-1] != b"": labels.append(b"") return labels def domainString(domain): """ Coerce a domain name string to bytes. L{twisted.names} represents domain names as L{bytes}, but many interfaces accept L{bytes} or a text string (L{unicode} on Python 2, L{str} on Python 3). This function coerces text strings using IDNA encoding --- see L{encodings.idna}. Note that DNS is I{case insensitive} but I{case preserving}. This function doesn't normalize case, so you'll still need to do that whenever comparing the strings it returns. @param domain: A domain name. If passed as a text string it will be C{idna} encoded. @type domain: L{bytes} or L{str} @returns: L{bytes} suitable for network transmission. @rtype: L{bytes} @since: Twisted 20.3.0 """ if isinstance(domain, str): domain = domain.encode("idna") if not isinstance(domain, bytes): raise TypeError( "Expected {} or {} but found {!r} of type {}".format( bytes.__name__, str.__name__, domain, type(domain) ) ) return domain def _isSubdomainOf(descendantName, ancestorName): """ Test whether C{descendantName} is equal to or is a I{subdomain} of C{ancestorName}. The names are compared case-insensitively. The names are treated as byte strings containing one or more DNS labels separated by B{.}. C{descendantName} is considered equal if its sequence of labels exactly matches the labels of C{ancestorName}. C{descendantName} is considered a I{subdomain} if its sequence of labels ends with the labels of C{ancestorName}. @type descendantName: L{bytes} @param descendantName: The DNS subdomain name. @type ancestorName: L{bytes} @param ancestorName: The DNS parent or ancestor domain name. @return: C{True} if C{descendantName} is equal to or if it is a subdomain of C{ancestorName}. Otherwise returns C{False}. """ descendantLabels = _nameToLabels(descendantName.lower()) ancestorLabels = _nameToLabels(ancestorName.lower()) return descendantLabels[-len(ancestorLabels) :] == ancestorLabels def str2time(s): """ Parse a string description of an interval into an integer number of seconds. @param s: An interval definition constructed as an interval duration followed by an interval unit. An interval duration is a base ten representation of an integer. An interval unit is one of the following letters: S (seconds), M (minutes), H (hours), D (days), W (weeks), or Y (years). For example: C{"3S"} indicates an interval of three seconds; C{"5D"} indicates an interval of five days. Alternatively, C{s} may be any non-string and it will be returned unmodified. @type s: text string (L{bytes} or L{str}) for parsing; anything else for passthrough. @return: an L{int} giving the interval represented by the string C{s}, or whatever C{s} is if it is not a string. """ suffixes = ( ("S", 1), ("M", 60), ("H", 60 * 60), ("D", 60 * 60 * 24), ("W", 60 * 60 * 24 * 7), ("Y", 60 * 60 * 24 * 365), ) if isinstance(s, bytes): s = s.decode("ascii") if isinstance(s, str): s = s.upper().strip() for (suff, mult) in suffixes: if s.endswith(suff): return int(float(s[:-1]) * mult) try: s = int(s) except ValueError: raise ValueError("Invalid time interval specifier: " + s) return s def readPrecisely(file, l): buff = file.read(l) if len(buff) < l: raise EOFError return buff class IEncodable(Interface): """ Interface for something which can be encoded to and decoded to the DNS wire format. A binary-mode file object (such as L{io.BytesIO}) is used as a buffer when encoding or decoding. """ def encode(strio, compDict=None): """ Write a representation of this object to the given file object. @type strio: File-like object @param strio: The buffer to write to. It must have a C{tell()} method. @type compDict: L{dict} of L{bytes} to L{int} r L{None} @param compDict: A mapping of names to byte offsets that have already been written to the buffer, which may be used for compression (see RFC 1035 section 4.1.4). When L{None}, encode without compression. """ def decode(strio, length=None): """ Reconstruct an object from data read from the given file object. @type strio: File-like object @param strio: A seekable buffer from which bytes may be read. @type length: L{int} or L{None} @param length: The number of bytes in this RDATA field. Most implementations can ignore this value. Only in the case of records similar to TXT where the total length is in no way encoded in the data is it necessary. """ class IEncodableRecord(IEncodable, IRecord): """ Interface for DNS records that can be encoded and decoded. @since: Twisted 21.2.0 """ @implementer(IEncodable) class Charstr: def __init__(self, string=b""): if not isinstance(string, bytes): raise ValueError(f"{string!r} is not a byte string") self.string = string def encode(self, strio, compDict=None): """ Encode this Character string into the appropriate byte format. @type strio: file @param strio: The byte representation of this Charstr will be written to this file. """ string = self.string ind = len(string) strio.write(_ord2bytes(ind)) strio.write(string) def decode(self, strio, length=None): """ Decode a byte string into this Charstr. @type strio: file @param strio: Bytes will be read from this file until the full string is decoded. @raise EOFError: Raised when there are not enough bytes available from C{strio}. """ self.string = b"" l = ord(readPrecisely(strio, 1)) self.string = readPrecisely(strio, l) def __eq__(self, other: object) -> bool: if isinstance(other, Charstr): return self.string == other.string return NotImplemented def __hash__(self): return hash(self.string) def __str__(self) -> str: """ Represent this L{Charstr} instance by its string value. """ return nativeString(self.string) @implementer(IEncodable) class Name: """ A name in the domain name system, made up of multiple labels. For example, I{twistedmatrix.com}. @ivar name: A byte string giving the name. @type name: L{bytes} """ def __init__(self, name=b""): """ @param name: A name. @type name: L{bytes} or L{str} """ self.name = domainString(name) def encode(self, strio, compDict=None): """ Encode this Name into the appropriate byte format. @type strio: file @param strio: The byte representation of this Name will be written to this file. @type compDict: dict @param compDict: dictionary of Names that have already been encoded and whose addresses may be backreferenced by this Name (for the purpose of reducing the message size). """ name = self.name while name: if compDict is not None: if name in compDict: strio.write(struct.pack("!H", 0xC000 | compDict[name])) return else: compDict[name] = strio.tell() + Message.headerSize ind = name.find(b".") if ind > 0: label, name = name[:ind], name[ind + 1 :] else: # This is the last label, end the loop after handling it. label = name name = None ind = len(label) strio.write(_ord2bytes(ind)) strio.write(label) strio.write(b"\x00") def decode(self, strio, length=None): """ Decode a byte string into this Name. @type strio: file @param strio: Bytes will be read from this file until the full Name is decoded. @raise EOFError: Raised when there are not enough bytes available from C{strio}. @raise ValueError: Raised when the name cannot be decoded (for example, because it contains a loop). """ visited = set() self.name = b"" off = 0 while 1: l = ord(readPrecisely(strio, 1)) if l == 0: if off > 0: strio.seek(off) return if (l >> 6) == 3: new_off = (l & 63) << 8 | ord(readPrecisely(strio, 1)) if new_off in visited: raise ValueError("Compression loop in encoded name") visited.add(new_off) if off == 0: off = strio.tell() strio.seek(new_off) continue label = readPrecisely(strio, l) if self.name == b"": self.name = label else: self.name = self.name + b"." + label def __eq__(self, other: object) -> bool: if isinstance(other, Name): return self.name.lower() == other.name.lower() return NotImplemented def __hash__(self): return hash(self.name) def __str__(self) -> str: """ Represent this L{Name} instance by its string name. """ return nativeString(self.name) @comparable @implementer(IEncodable) class Query: """ Represent a single DNS query. @ivar name: The name about which this query is requesting information. @type name: L{Name} @ivar type: The query type. @type type: L{int} @ivar cls: The query class. @type cls: L{int} """ def __init__(self, name: Union[bytes, str] = b"", type: int = A, cls: int = IN): """ @type name: L{bytes} or L{str} @param name: See L{Query.name} @type type: L{int} @param type: The query type. @type cls: L{int} @param cls: The query class. """ self.name = Name(name) self.type = type self.cls = cls def encode(self, strio, compDict=None): self.name.encode(strio, compDict) strio.write(struct.pack("!HH", self.type, self.cls)) def decode(self, strio, length=None): self.name.decode(strio) buff = readPrecisely(strio, 4) self.type, self.cls = struct.unpack("!HH", buff) def __hash__(self): return hash((self.name.name.lower(), self.type, self.cls)) def __cmp__(self, other): if isinstance(other, Query): return cmp( (self.name.name.lower(), self.type, self.cls), (other.name.name.lower(), other.type, other.cls), ) return NotImplemented def __str__(self) -> str: t = QUERY_TYPES.get( self.type, EXT_QUERIES.get(self.type, "UNKNOWN (%d)" % self.type) ) c = QUERY_CLASSES.get(self.cls, "UNKNOWN (%d)" % self.cls) return f"<Query {self.name} {t} {c}>" def __repr__(self) -> str: return f"Query({self.name.name!r}, {self.type!r}, {self.cls!r})" @implementer(IEncodable) class _OPTHeader(tputil.FancyStrMixin, tputil.FancyEqMixin): """ An OPT record header. @ivar name: The DNS name associated with this record. Since this is a pseudo record, the name is always an L{Name} instance with value b'', which represents the DNS root zone. This attribute is a readonly property. @ivar type: The DNS record type. This is a fixed value of 41 C{dns.OPT} for OPT Record. This attribute is a readonly property. @see: L{_OPTHeader.__init__} for documentation of other public instance attributes. @see: U{https://tools.ietf.org/html/rfc6891#section-6.1.2} @since: 13.2 """ showAttributes = ( ("name", lambda n: nativeString(n.name)), "type", "udpPayloadSize", "extendedRCODE", "version", "dnssecOK", "options", ) compareAttributes = ( "name", "type", "udpPayloadSize", "extendedRCODE", "version", "dnssecOK", "options", ) def __init__( self, udpPayloadSize=4096, extendedRCODE=0, version=0, dnssecOK=False, options=None, ): """ @type udpPayloadSize: L{int} @param udpPayloadSize: The number of octets of the largest UDP payload that can be reassembled and delivered in the requestor's network stack. @type extendedRCODE: L{int} @param extendedRCODE: Forms the upper 8 bits of extended 12-bit RCODE (together with the 4 bits defined in [RFC1035]. Note that EXTENDED-RCODE value 0 indicates that an unextended RCODE is in use (values 0 through 15). @type version: L{int} @param version: Indicates the implementation level of the setter. Full conformance with this specification is indicated by version C{0}. @type dnssecOK: L{bool} @param dnssecOK: DNSSEC OK bit as defined by [RFC3225]. @type options: L{list} @param options: A L{list} of 0 or more L{_OPTVariableOption} instances. """ self.udpPayloadSize = udpPayloadSize self.extendedRCODE = extendedRCODE self.version = version self.dnssecOK = dnssecOK if options is None: options = [] self.options = options @property def name(self): """ A readonly property for accessing the C{name} attribute of this record. @return: The DNS name associated with this record. Since this is a pseudo record, the name is always an L{Name} instance with value b'', which represents the DNS root zone. """ return Name(b"") @property def type(self): """ A readonly property for accessing the C{type} attribute of this record. @return: The DNS record type. This is a fixed value of 41 (C{dns.OPT} for OPT Record. """ return OPT def encode(self, strio, compDict=None): """ Encode this L{_OPTHeader} instance to bytes. @type strio: file @param strio: the byte representation of this L{_OPTHeader} will be written to this file. @type compDict: L{dict} or L{None} @param compDict: A dictionary of backreference addresses that have already been written to this stream and that may be used for DNS name compression. """ b = BytesIO() for o in self.options: o.encode(b) optionBytes = b.getvalue() RRHeader( name=self.name.name, type=self.type, cls=self.udpPayloadSize, ttl=(self.extendedRCODE << 24 | self.version << 16 | self.dnssecOK << 15), payload=UnknownRecord(optionBytes), ).encode(strio, compDict) def decode(self, strio, length=None): """ Decode bytes into an L{_OPTHeader} instance. @type strio: file @param strio: Bytes will be read from this file until the full L{_OPTHeader} is decoded. @type length: L{int} or L{None} @param length: Not used. """ h = RRHeader() h.decode(strio, length) h.payload = UnknownRecord(readPrecisely(strio, h.rdlength)) newOptHeader = self.fromRRHeader(h) for attrName in self.compareAttributes: if attrName not in ("name", "type"): setattr(self, attrName, getattr(newOptHeader, attrName)) @classmethod def fromRRHeader(cls, rrHeader): """ A classmethod for constructing a new L{_OPTHeader} from the attributes and payload of an existing L{RRHeader} instance. @type rrHeader: L{RRHeader} @param rrHeader: An L{RRHeader} instance containing an L{UnknownRecord} payload. @return: An instance of L{_OPTHeader}. @rtype: L{_OPTHeader} """ options = None if rrHeader.payload is not None: options = [] optionsBytes = BytesIO(rrHeader.payload.data) optionsBytesLength = len(rrHeader.payload.data) while optionsBytes.tell() < optionsBytesLength: o = _OPTVariableOption() o.decode(optionsBytes) options.append(o) # Decode variable options if present return cls( udpPayloadSize=rrHeader.cls, extendedRCODE=rrHeader.ttl >> 24, version=rrHeader.ttl >> 16 & 0xFF, dnssecOK=(rrHeader.ttl & 0xFFFF) >> 15, options=options, ) @implementer(IEncodable) class _OPTVariableOption(tputil.FancyStrMixin, tputil.FancyEqMixin): """ A class to represent OPT record variable options. @see: L{_OPTVariableOption.__init__} for documentation of public instance attributes. @see: U{https://tools.ietf.org/html/rfc6891#section-6.1.2} @since: 13.2 """ showAttributes = ("code", ("data", nativeString)) compareAttributes = ("code", "data") _fmt = "!HH" def __init__(self, code=0, data=b""): """ @type code: L{int} @param code: The option code @type data: L{bytes} @param data: The option data """ self.code = code self.data = data def encode(self, strio, compDict=None): """ Encode this L{_OPTVariableOption} to bytes. @type strio: file @param strio: the byte representation of this L{_OPTVariableOption} will be written to this file. @type compDict: L{dict} or L{None} @param compDict: A dictionary of backreference addresses that have already been written to this stream and that may be used for DNS name compression. """ strio.write(struct.pack(self._fmt, self.code, len(self.data)) + self.data) def decode(self, strio, length=None): """ Decode bytes into an L{_OPTVariableOption} instance. @type strio: file @param strio: Bytes will be read from this file until the full L{_OPTVariableOption} is decoded. @type length: L{int} or L{None} @param length: Not used. """ l = struct.calcsize(self._fmt) buff = readPrecisely(strio, l) self.code, length = struct.unpack(self._fmt, buff) self.data = readPrecisely(strio, length) @implementer(IEncodable) class RRHeader(tputil.FancyEqMixin): """ A resource record header. @cvar fmt: L{str} specifying the byte format of an RR. @ivar name: The name about which this reply contains information. @type name: L{Name} @ivar type: The query type of the original request. @type type: L{int} @ivar cls: The query class of the original request. @ivar ttl: The time-to-live for this record. @type ttl: L{int} @ivar payload: The record described by this header. @type payload: L{IEncodableRecord} or L{None} @ivar auth: A L{bool} indicating whether this C{RRHeader} was parsed from an authoritative message. """ compareAttributes = ("name", "type", "cls", "ttl", "payload", "auth") fmt = "!HHIH" rdlength = None cachedResponse = None def __init__( self, name: Union[bytes, str] = b"", type: int = A, cls: int = IN, ttl: SupportsInt = 0, payload: Optional[IEncodableRecord] = None, auth: bool = False, ): """ @type name: L{bytes} or L{str} @param name: See L{RRHeader.name} @type type: L{int} @param type: The query type. @type cls: L{int} @param cls: The query class. @type ttl: L{int} @param ttl: Time to live for this record. This will be converted to an L{int}. @type payload: L{IEncodableRecord} or L{None} @param payload: An optional Query Type specific data object. @raises TypeError: if the ttl cannot be converted to an L{int}. @raises ValueError: if the ttl is negative. @raises ValueError: if the payload type is not equal to the C{type} argument. """ payloadType = None if payload is None else payload.TYPE if payloadType is not None and payloadType != type: raise ValueError( "Payload type (%s) does not match given type (%s)" % ( QUERY_TYPES.get(payloadType, payloadType), QUERY_TYPES.get(type, type), ) ) integralTTL = int(ttl) if integralTTL < 0: raise ValueError("TTL cannot be negative") self.name = Name(name) self.type = type self.cls = cls self.ttl = integralTTL self.payload = payload self.auth = auth def encode(self, strio, compDict=None): self.name.encode(strio, compDict) strio.write(struct.pack(self.fmt, self.type, self.cls, self.ttl, 0)) if self.payload: prefix = strio.tell() self.payload.encode(strio, compDict) aft = strio.tell() strio.seek(prefix - 2, 0) strio.write(struct.pack("!H", aft - prefix)) strio.seek(aft, 0) def decode(self, strio, length=None): self.name.decode(strio) l = struct.calcsize(self.fmt) buff = readPrecisely(strio, l) r = struct.unpack(self.fmt, buff) self.type, self.cls, self.ttl, self.rdlength = r def isAuthoritative(self): return self.auth def __str__(self) -> str: t = QUERY_TYPES.get( self.type, EXT_QUERIES.get(self.type, "UNKNOWN (%d)" % self.type) ) c = QUERY_CLASSES.get(self.cls, "UNKNOWN (%d)" % self.cls) return "<RR name=%s type=%s class=%s ttl=%ds auth=%s>" % ( self.name, t, c, self.ttl, self.auth and "True" or "False", ) __repr__ = __str__ @implementer(IEncodableRecord) class SimpleRecord(tputil.FancyStrMixin, tputil.FancyEqMixin): """ A Resource Record which consists of a single RFC 1035 domain-name. @type name: L{Name} @ivar name: The name associated with this record. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ showAttributes = (("name", "name", "%s"), "ttl") compareAttributes = ("name", "ttl") TYPE: Optional[int] = None name = None def __init__(self, name=b"", ttl=None): """ @param name: See L{SimpleRecord.name} @type name: L{bytes} or L{str} """ self.name = Name(name) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): self.name.encode(strio, compDict) def decode(self, strio, length=None): self.name = Name() self.name.decode(strio) def __hash__(self): return hash(self.name) # Kinds of RRs - oh my! class Record_NS(SimpleRecord): """ An authoritative nameserver. """ TYPE = NS fancybasename = "NS" class Record_MD(SimpleRecord): """ A mail destination. This record type is obsolete. @see: L{Record_MX} """ TYPE = MD fancybasename = "MD" class Record_MF(SimpleRecord): """ A mail forwarder. This record type is obsolete. @see: L{Record_MX} """ TYPE = MF fancybasename = "MF" class Record_CNAME(SimpleRecord): """ The canonical name for an alias. """ TYPE = CNAME fancybasename = "CNAME" class Record_MB(SimpleRecord): """ A mailbox domain name. This is an experimental record type. """ TYPE = MB fancybasename = "MB" class Record_MG(SimpleRecord): """ A mail group member. This is an experimental record type. """ TYPE = MG fancybasename = "MG" class Record_MR(SimpleRecord): """ A mail rename domain name. This is an experimental record type. """ TYPE = MR fancybasename = "MR" class Record_PTR(SimpleRecord): """ A domain name pointer. """ TYPE = PTR fancybasename = "PTR" class Record_DNAME(SimpleRecord): """ A non-terminal DNS name redirection. This record type provides the capability to map an entire subtree of the DNS name space to another domain. It differs from the CNAME record which maps a single node of the name space. @see: U{http://www.faqs.org/rfcs/rfc2672.html} @see: U{http://www.faqs.org/rfcs/rfc3363.html} """ TYPE = DNAME fancybasename = "DNAME" @implementer(IEncodableRecord) class Record_A(tputil.FancyEqMixin): """ An IPv4 host address. @type address: L{bytes} @ivar address: The packed network-order representation of the IPv4 address associated with this record. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ compareAttributes = ("address", "ttl") TYPE = A address = None def __init__(self, address="0.0.0.0", ttl=None): """ @type address: L{bytes} or L{str} @param address: The IPv4 address associated with this record, in quad-dotted notation. """ if isinstance(address, bytes): address = address.decode("ascii") address = socket.inet_aton(address) self.address = address self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(self.address) def decode(self, strio, length=None): self.address = readPrecisely(strio, 4) def __hash__(self): return hash(self.address) def __str__(self) -> str: return f"<A address={self.dottedQuad()} ttl={self.ttl}>" __repr__ = __str__ def dottedQuad(self): return socket.inet_ntoa(self.address) @implementer(IEncodableRecord) class Record_SOA(tputil.FancyEqMixin, tputil.FancyStrMixin): """ Marks the start of a zone of authority. This record describes parameters which are shared by all records within a particular zone. @type mname: L{Name} @ivar mname: The domain-name of the name server that was the original or primary source of data for this zone. @type rname: L{Name} @ivar rname: A domain-name which specifies the mailbox of the person responsible for this zone. @type serial: L{int} @ivar serial: The unsigned 32 bit version number of the original copy of the zone. Zone transfers preserve this value. This value wraps and should be compared using sequence space arithmetic. @type refresh: L{int} @ivar refresh: A 32 bit time interval before the zone should be refreshed. @type minimum: L{int} @ivar minimum: The unsigned 32 bit minimum TTL field that should be exported with any RR from this zone. @type expire: L{int} @ivar expire: A 32 bit time value that specifies the upper limit on the time interval that can elapse before the zone is no longer authoritative. @type retry: L{int} @ivar retry: A 32 bit time interval that should elapse before a failed refresh should be retried. @type ttl: L{int} @ivar ttl: The default TTL to use for records served from this zone. """ fancybasename = "SOA" compareAttributes = ( "serial", "mname", "rname", "refresh", "expire", "retry", "minimum", "ttl", ) showAttributes = ( ("mname", "mname", "%s"), ("rname", "rname", "%s"), "serial", "refresh", "retry", "expire", "minimum", "ttl", ) TYPE = SOA def __init__( self, mname=b"", rname=b"", serial=0, refresh=0, retry=0, expire=0, minimum=0, ttl=None, ): """ @param mname: See L{Record_SOA.mname} @type mname: L{bytes} or L{str} @param rname: See L{Record_SOA.rname} @type rname: L{bytes} or L{str} """ self.mname, self.rname = Name(mname), Name(rname) self.serial, self.refresh = str2time(serial), str2time(refresh) self.minimum, self.expire = str2time(minimum), str2time(expire) self.retry = str2time(retry) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): self.mname.encode(strio, compDict) self.rname.encode(strio, compDict) strio.write( struct.pack( "!LlllL", self.serial, self.refresh, self.retry, self.expire, self.minimum, ) ) def decode(self, strio, length=None): self.mname, self.rname = Name(), Name() self.mname.decode(strio) self.rname.decode(strio) r = struct.unpack("!LlllL", readPrecisely(strio, 20)) self.serial, self.refresh, self.retry, self.expire, self.minimum = r def __hash__(self): return hash( (self.serial, self.mname, self.rname, self.refresh, self.expire, self.retry) ) @implementer(IEncodableRecord) class Record_NULL(tputil.FancyStrMixin, tputil.FancyEqMixin): """ A null record. This is an experimental record type. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ fancybasename = "NULL" showAttributes = (("payload", _nicebytes), "ttl") compareAttributes = ("payload", "ttl") TYPE = NULL def __init__(self, payload=None, ttl=None): self.payload = payload self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(self.payload) def decode(self, strio, length=None): self.payload = readPrecisely(strio, length) def __hash__(self): return hash(self.payload) @implementer(IEncodableRecord) class Record_WKS(tputil.FancyEqMixin, tputil.FancyStrMixin): """ A well known service description. This record type is obsolete. See L{Record_SRV}. @type address: L{bytes} @ivar address: The packed network-order representation of the IPv4 address associated with this record. @type protocol: L{int} @ivar protocol: The 8 bit IP protocol number for which this service map is relevant. @type map: L{bytes} @ivar map: A bitvector indicating the services available at the specified address. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ fancybasename = "WKS" compareAttributes = ("address", "protocol", "map", "ttl") showAttributes = [("_address", "address", "%s"), "protocol", "ttl"] TYPE = WKS @property def _address(self): return socket.inet_ntoa(self.address) def __init__(self, address="0.0.0.0", protocol=0, map=b"", ttl=None): """ @type address: L{bytes} or L{str} @param address: The IPv4 address associated with this record, in quad-dotted notation. """ if isinstance(address, bytes): address = address.decode("idna") self.address = socket.inet_aton(address) self.protocol, self.map = protocol, map self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(self.address) strio.write(struct.pack("!B", self.protocol)) strio.write(self.map) def decode(self, strio, length=None): self.address = readPrecisely(strio, 4) self.protocol = struct.unpack("!B", readPrecisely(strio, 1))[0] self.map = readPrecisely(strio, length - 5) def __hash__(self): return hash((self.address, self.protocol, self.map)) @implementer(IEncodableRecord) class Record_AAAA(tputil.FancyEqMixin, tputil.FancyStrMixin): """ An IPv6 host address. @type address: L{bytes} @ivar address: The packed network-order representation of the IPv6 address associated with this record. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. @see: U{http://www.faqs.org/rfcs/rfc1886.html} """ TYPE = AAAA fancybasename = "AAAA" showAttributes = (("_address", "address", "%s"), "ttl") compareAttributes = ("address", "ttl") @property def _address(self): return socket.inet_ntop(AF_INET6, self.address) def __init__(self, address="::", ttl=None): """ @type address: L{bytes} or L{str} @param address: The IPv6 address for this host, in RFC 2373 format. """ if isinstance(address, bytes): address = address.decode("idna") self.address = socket.inet_pton(AF_INET6, address) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(self.address) def decode(self, strio, length=None): self.address = readPrecisely(strio, 16) def __hash__(self): return hash(self.address) @implementer(IEncodableRecord) class Record_A6(tputil.FancyStrMixin, tputil.FancyEqMixin): """ An IPv6 address. This is an experimental record type. @type prefixLen: L{int} @ivar prefixLen: The length of the suffix. @type suffix: L{bytes} @ivar suffix: An IPv6 address suffix in network order. @type prefix: L{Name} @ivar prefix: If specified, a name which will be used as a prefix for other A6 records. @type bytes: L{int} @ivar bytes: The length of the prefix. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. @see: U{http://www.faqs.org/rfcs/rfc2874.html} @see: U{http://www.faqs.org/rfcs/rfc3363.html} @see: U{http://www.faqs.org/rfcs/rfc3364.html} """ TYPE = A6 fancybasename = "A6" showAttributes = (("_suffix", "suffix", "%s"), ("prefix", "prefix", "%s"), "ttl") compareAttributes = ("prefixLen", "prefix", "suffix", "ttl") @property def _suffix(self): return socket.inet_ntop(AF_INET6, self.suffix) def __init__(self, prefixLen=0, suffix="::", prefix=b"", ttl=None): """ @param suffix: An IPv6 address suffix in in RFC 2373 format. @type suffix: L{bytes} or L{str} @param prefix: An IPv6 address prefix for other A6 records. @type prefix: L{bytes} or L{str} """ if isinstance(suffix, bytes): suffix = suffix.decode("idna") self.prefixLen = prefixLen self.suffix = socket.inet_pton(AF_INET6, suffix) self.prefix = Name(prefix) self.bytes = int((128 - self.prefixLen) / 8.0) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(struct.pack("!B", self.prefixLen)) if self.bytes: strio.write(self.suffix[-self.bytes :]) if self.prefixLen: # This may not be compressed self.prefix.encode(strio, None) def decode(self, strio, length=None): self.prefixLen = struct.unpack("!B", readPrecisely(strio, 1))[0] self.bytes = int((128 - self.prefixLen) / 8.0) if self.bytes: self.suffix = b"\x00" * (16 - self.bytes) + readPrecisely(strio, self.bytes) if self.prefixLen: self.prefix.decode(strio) def __eq__(self, other: object) -> bool: if isinstance(other, Record_A6): return ( self.prefixLen == other.prefixLen and self.suffix[-self.bytes :] == other.suffix[-self.bytes :] and self.prefix == other.prefix and self.ttl == other.ttl ) return NotImplemented def __hash__(self): return hash((self.prefixLen, self.suffix[-self.bytes :], self.prefix)) def __str__(self) -> str: return "<A6 %s %s (%d) ttl=%s>" % ( self.prefix, socket.inet_ntop(AF_INET6, self.suffix), self.prefixLen, self.ttl, ) @implementer(IEncodableRecord) class Record_SRV(tputil.FancyEqMixin, tputil.FancyStrMixin): """ The location of the server(s) for a specific protocol and domain. This is an experimental record type. @type priority: L{int} @ivar priority: The priority of this target host. A client MUST attempt to contact the target host with the lowest-numbered priority it can reach; target hosts with the same priority SHOULD be tried in an order defined by the weight field. @type weight: L{int} @ivar weight: Specifies a relative weight for entries with the same priority. Larger weights SHOULD be given a proportionately higher probability of being selected. @type port: L{int} @ivar port: The port on this target host of this service. @type target: L{Name} @ivar target: The domain name of the target host. There MUST be one or more address records for this name, the name MUST NOT be an alias (in the sense of RFC 1034 or RFC 2181). Implementors are urged, but not required, to return the address record(s) in the Additional Data section. Unless and until permitted by future standards action, name compression is not to be used for this field. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. @see: U{http://www.faqs.org/rfcs/rfc2782.html} """ TYPE = SRV fancybasename = "SRV" compareAttributes = ("priority", "weight", "target", "port", "ttl") showAttributes = ("priority", "weight", ("target", "target", "%s"), "port", "ttl") def __init__(self, priority=0, weight=0, port=0, target=b"", ttl=None): """ @param target: See L{Record_SRV.target} @type target: L{bytes} or L{str} """ self.priority = int(priority) self.weight = int(weight) self.port = int(port) self.target = Name(target) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(struct.pack("!HHH", self.priority, self.weight, self.port)) # This can't be compressed self.target.encode(strio, None) def decode(self, strio, length=None): r = struct.unpack("!HHH", readPrecisely(strio, struct.calcsize("!HHH"))) self.priority, self.weight, self.port = r self.target = Name() self.target.decode(strio) def __hash__(self): return hash((self.priority, self.weight, self.port, self.target)) @implementer(IEncodableRecord) class Record_NAPTR(tputil.FancyEqMixin, tputil.FancyStrMixin): """ The location of the server(s) for a specific protocol and domain. @type order: L{int} @ivar order: An integer specifying the order in which the NAPTR records MUST be processed to ensure the correct ordering of rules. Low numbers are processed before high numbers. @type preference: L{int} @ivar preference: An integer that specifies the order in which NAPTR records with equal "order" values SHOULD be processed, low numbers being processed before high numbers. @type flag: L{Charstr} @ivar flag: A <character-string> containing flags to control aspects of the rewriting and interpretation of the fields in the record. Flags are single characters from the set [A-Z0-9]. The case of the alphabetic characters is not significant. At this time only four flags, "S", "A", "U", and "P", are defined. @type service: L{Charstr} @ivar service: Specifies the service(s) available down this rewrite path. It may also specify the particular protocol that is used to talk with a service. A protocol MUST be specified if the flags field states that the NAPTR is terminal. @type regexp: L{Charstr} @ivar regexp: A STRING containing a substitution expression that is applied to the original string held by the client in order to construct the next domain name to lookup. @type replacement: L{Name} @ivar replacement: The next NAME to query for NAPTR, SRV, or address records depending on the value of the flags field. This MUST be a fully qualified domain-name. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. @see: U{http://www.faqs.org/rfcs/rfc2915.html} """ TYPE = NAPTR compareAttributes = ( "order", "preference", "flags", "service", "regexp", "replacement", ) fancybasename = "NAPTR" showAttributes = ( "order", "preference", ("flags", "flags", "%s"), ("service", "service", "%s"), ("regexp", "regexp", "%s"), ("replacement", "replacement", "%s"), "ttl", ) def __init__( self, order=0, preference=0, flags=b"", service=b"", regexp=b"", replacement=b"", ttl=None, ): """ @param replacement: See L{Record_NAPTR.replacement} @type replacement: L{bytes} or L{str} """ self.order = int(order) self.preference = int(preference) self.flags = Charstr(flags) self.service = Charstr(service) self.regexp = Charstr(regexp) self.replacement = Name(replacement) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(struct.pack("!HH", self.order, self.preference)) # This can't be compressed self.flags.encode(strio, None) self.service.encode(strio, None) self.regexp.encode(strio, None) self.replacement.encode(strio, None) def decode(self, strio, length=None): r = struct.unpack("!HH", readPrecisely(strio, struct.calcsize("!HH"))) self.order, self.preference = r self.flags = Charstr() self.service = Charstr() self.regexp = Charstr() self.replacement = Name() self.flags.decode(strio) self.service.decode(strio) self.regexp.decode(strio) self.replacement.decode(strio) def __hash__(self): return hash( ( self.order, self.preference, self.flags, self.service, self.regexp, self.replacement, ) ) @implementer(IEncodableRecord) class Record_AFSDB(tputil.FancyStrMixin, tputil.FancyEqMixin): """ Map from a domain name to the name of an AFS cell database server. @type subtype: L{int} @ivar subtype: In the case of subtype 1, the host has an AFS version 3.0 Volume Location Server for the named AFS cell. In the case of subtype 2, the host has an authenticated name server holding the cell-root directory node for the named DCE/NCA cell. @type hostname: L{Name} @ivar hostname: The domain name of a host that has a server for the cell named by this record. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. @see: U{http://www.faqs.org/rfcs/rfc1183.html} """ TYPE = AFSDB fancybasename = "AFSDB" compareAttributes = ("subtype", "hostname", "ttl") showAttributes = ("subtype", ("hostname", "hostname", "%s"), "ttl") def __init__(self, subtype=0, hostname=b"", ttl=None): """ @param hostname: See L{Record_AFSDB.hostname} @type hostname: L{bytes} or L{str} """ self.subtype = int(subtype) self.hostname = Name(hostname) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(struct.pack("!H", self.subtype)) self.hostname.encode(strio, compDict) def decode(self, strio, length=None): r = struct.unpack("!H", readPrecisely(strio, struct.calcsize("!H"))) (self.subtype,) = r self.hostname.decode(strio) def __hash__(self): return hash((self.subtype, self.hostname)) @implementer(IEncodableRecord) class Record_RP(tputil.FancyEqMixin, tputil.FancyStrMixin): """ The responsible person for a domain. @type mbox: L{Name} @ivar mbox: A domain name that specifies the mailbox for the responsible person. @type txt: L{Name} @ivar txt: A domain name for which TXT RR's exist (indirection through which allows information sharing about the contents of this RP record). @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. @see: U{http://www.faqs.org/rfcs/rfc1183.html} """ TYPE = RP fancybasename = "RP" compareAttributes = ("mbox", "txt", "ttl") showAttributes = (("mbox", "mbox", "%s"), ("txt", "txt", "%s"), "ttl") def __init__(self, mbox=b"", txt=b"", ttl=None): """ @param mbox: See L{Record_RP.mbox}. @type mbox: L{bytes} or L{str} @param txt: See L{Record_RP.txt} @type txt: L{bytes} or L{str} """ self.mbox = Name(mbox) self.txt = Name(txt) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): self.mbox.encode(strio, compDict) self.txt.encode(strio, compDict) def decode(self, strio, length=None): self.mbox = Name() self.txt = Name() self.mbox.decode(strio) self.txt.decode(strio) def __hash__(self): return hash((self.mbox, self.txt)) @implementer(IEncodableRecord) class Record_HINFO(tputil.FancyStrMixin, tputil.FancyEqMixin): """ Host information. @type cpu: L{bytes} @ivar cpu: Specifies the CPU type. @type os: L{bytes} @ivar os: Specifies the OS. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ TYPE = HINFO fancybasename = "HINFO" showAttributes = (("cpu", _nicebytes), ("os", _nicebytes), "ttl") compareAttributes = ("cpu", "os", "ttl") def __init__(self, cpu=b"", os=b"", ttl=None): self.cpu, self.os = cpu, os self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(struct.pack("!B", len(self.cpu)) + self.cpu) strio.write(struct.pack("!B", len(self.os)) + self.os) def decode(self, strio, length=None): cpu = struct.unpack("!B", readPrecisely(strio, 1))[0] self.cpu = readPrecisely(strio, cpu) os = struct.unpack("!B", readPrecisely(strio, 1))[0] self.os = readPrecisely(strio, os) def __eq__(self, other: object) -> bool: if isinstance(other, Record_HINFO): return ( self.os.lower() == other.os.lower() and self.cpu.lower() == other.cpu.lower() and self.ttl == other.ttl ) return NotImplemented def __hash__(self): return hash((self.os.lower(), self.cpu.lower())) @implementer(IEncodableRecord) class Record_MINFO(tputil.FancyEqMixin, tputil.FancyStrMixin): """ Mailbox or mail list information. This is an experimental record type. @type rmailbx: L{Name} @ivar rmailbx: A domain-name which specifies a mailbox which is responsible for the mailing list or mailbox. If this domain name names the root, the owner of the MINFO RR is responsible for itself. @type emailbx: L{Name} @ivar emailbx: A domain-name which specifies a mailbox which is to receive error messages related to the mailing list or mailbox specified by the owner of the MINFO record. If this domain name names the root, errors should be returned to the sender of the message. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ TYPE = MINFO rmailbx = None emailbx = None fancybasename = "MINFO" compareAttributes = ("rmailbx", "emailbx", "ttl") showAttributes = ( ("rmailbx", "responsibility", "%s"), ("emailbx", "errors", "%s"), "ttl", ) def __init__(self, rmailbx=b"", emailbx=b"", ttl=None): """ @param rmailbx: See L{Record_MINFO.rmailbx}. @type rmailbx: L{bytes} or L{str} @param emailbx: See L{Record_MINFO.rmailbx}. @type emailbx: L{bytes} or L{str} """ self.rmailbx, self.emailbx = Name(rmailbx), Name(emailbx) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): self.rmailbx.encode(strio, compDict) self.emailbx.encode(strio, compDict) def decode(self, strio, length=None): self.rmailbx, self.emailbx = Name(), Name() self.rmailbx.decode(strio) self.emailbx.decode(strio) def __hash__(self): return hash((self.rmailbx, self.emailbx)) @implementer(IEncodableRecord) class Record_MX(tputil.FancyStrMixin, tputil.FancyEqMixin): """ Mail exchange. @type preference: L{int} @ivar preference: Specifies the preference given to this RR among others at the same owner. Lower values are preferred. @type name: L{Name} @ivar name: A domain-name which specifies a host willing to act as a mail exchange. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ TYPE = MX fancybasename = "MX" compareAttributes = ("preference", "name", "ttl") showAttributes = ("preference", ("name", "name", "%s"), "ttl") def __init__(self, preference=0, name=b"", ttl=None, **kwargs): """ @param name: See L{Record_MX.name}. @type name: L{bytes} or L{str} """ self.preference = int(preference) self.name = Name(kwargs.get("exchange", name)) self.ttl = str2time(ttl) def encode(self, strio, compDict=None): strio.write(struct.pack("!H", self.preference)) self.name.encode(strio, compDict) def decode(self, strio, length=None): self.preference = struct.unpack("!H", readPrecisely(strio, 2))[0] self.name = Name() self.name.decode(strio) def __hash__(self): return hash((self.preference, self.name)) @implementer(IEncodableRecord) class Record_SSHFP(tputil.FancyEqMixin, tputil.FancyStrMixin): """ A record containing the fingerprint of an SSH key. @type algorithm: L{int} @ivar algorithm: The SSH key's algorithm, such as L{ALGORITHM_RSA}. Note that the numbering used for SSH key algorithms is specific to the SSHFP record, and is not the same as the numbering used for KEY or SIG records. @type fingerprintType: L{int} @ivar fingerprintType: The fingerprint type, such as L{FINGERPRINT_TYPE_SHA256}. @type fingerprint: L{bytes} @ivar fingerprint: The key's fingerprint, e.g. a 32-byte SHA-256 digest. @cvar ALGORITHM_RSA: The algorithm value for C{ssh-rsa} keys. @cvar ALGORITHM_DSS: The algorithm value for C{ssh-dss} keys. @cvar ALGORITHM_ECDSA: The algorithm value for C{ecdsa-sha2-*} keys. @cvar ALGORITHM_Ed25519: The algorithm value for C{ed25519} keys. @cvar FINGERPRINT_TYPE_SHA1: The type for SHA-1 fingerprints. @cvar FINGERPRINT_TYPE_SHA256: The type for SHA-256 fingerprints. @see: U{RFC 4255 <https://tools.ietf.org/html/rfc4255>} and U{RFC 6594 <https://tools.ietf.org/html/rfc6594>} """ fancybasename = "SSHFP" compareAttributes = ("algorithm", "fingerprintType", "fingerprint", "ttl") showAttributes = ("algorithm", "fingerprintType", "fingerprint") TYPE = SSHFP ALGORITHM_RSA = 1 ALGORITHM_DSS = 2 ALGORITHM_ECDSA = 3 ALGORITHM_Ed25519 = 4 FINGERPRINT_TYPE_SHA1 = 1 FINGERPRINT_TYPE_SHA256 = 2 def __init__(self, algorithm=0, fingerprintType=0, fingerprint=b"", ttl=0): self.algorithm = algorithm self.fingerprintType = fingerprintType self.fingerprint = fingerprint self.ttl = ttl def encode(self, strio, compDict=None): strio.write(struct.pack("!BB", self.algorithm, self.fingerprintType)) strio.write(self.fingerprint) def decode(self, strio, length=None): r = struct.unpack("!BB", readPrecisely(strio, 2)) (self.algorithm, self.fingerprintType) = r self.fingerprint = readPrecisely(strio, length - 2) def __hash__(self): return hash((self.algorithm, self.fingerprintType, self.fingerprint)) @implementer(IEncodableRecord) class Record_TXT(tputil.FancyEqMixin, tputil.FancyStrMixin): """ Freeform text. @type data: L{list} of L{bytes} @ivar data: Freeform text which makes up this record. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ TYPE = TXT fancybasename = "TXT" showAttributes = (("data", _nicebyteslist), "ttl") compareAttributes = ("data", "ttl") def __init__(self, *data, **kw): self.data = list(data) # arg man python sucks so bad self.ttl = str2time(kw.get("ttl", None)) def encode(self, strio, compDict=None): for d in self.data: strio.write(struct.pack("!B", len(d)) + d) def decode(self, strio, length=None): soFar = 0 self.data = [] while soFar < length: L = struct.unpack("!B", readPrecisely(strio, 1))[0] self.data.append(readPrecisely(strio, L)) soFar += L + 1 if soFar != length: log.msg( "Decoded %d bytes in %s record, but rdlength is %d" % (soFar, self.fancybasename, length) ) def __hash__(self): return hash(tuple(self.data)) @implementer(IEncodableRecord) class UnknownRecord(tputil.FancyEqMixin, tputil.FancyStrMixin): """ Encapsulate the wire data for unknown record types so that they can pass through the system unchanged. @type data: L{bytes} @ivar data: Wire data which makes up this record. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. @since: 11.1 """ TYPE = None fancybasename = "UNKNOWN" compareAttributes = ("data", "ttl") showAttributes = (("data", _nicebytes), "ttl") def __init__(self, data=b"", ttl=None): self.data = data self.ttl = str2time(ttl) def encode(self, strio, compDict=None): """ Write the raw bytes corresponding to this record's payload to the stream. """ strio.write(self.data) def decode(self, strio, length=None): """ Load the bytes which are part of this record from the stream and store them unparsed and unmodified. """ if length is None: raise Exception("must know length for unknown record types") self.data = readPrecisely(strio, length) def __hash__(self): return hash((self.data, self.ttl)) class Record_SPF(Record_TXT): """ Structurally, freeform text. Semantically, a policy definition, formatted as defined in U{rfc 4408<http://www.faqs.org/rfcs/rfc4408.html>}. @type data: L{list} of L{bytes} @ivar data: Freeform text which makes up this record. @type ttl: L{int} @ivar ttl: The maximum number of seconds which this record should be cached. """ TYPE = SPF fancybasename = "SPF" @implementer(IEncodableRecord) class Record_TSIG(tputil.FancyEqMixin, tputil.FancyStrMixin): """ A transaction signature, encapsulated in a RR, as described in U{RFC 2845 <https://tools.ietf.org/html/rfc2845>}. @type algorithm: L{Name} @ivar algorithm: The name of the signature or MAC algorithm. @type timeSigned: L{int} @ivar timeSigned: Signing time, as seconds from the POSIX epoch. @type fudge: L{int} @ivar fudge: Allowable time skew, in seconds. @type MAC: L{bytes} @ivar MAC: The message digest or signature. @type originalID: L{int} @ivar originalID: A message ID. @type error: L{int} @ivar error: An error code (extended C{RCODE}) carried in exceptional cases. @type otherData: L{bytes} @ivar otherData: Other data carried in exceptional cases. """ fancybasename = "TSIG" compareAttributes = ( "algorithm", "timeSigned", "fudge", "MAC", "originalID", "error", "otherData", "ttl", ) showAttributes = ["algorithm", "timeSigned", "MAC", "error", "otherData"] TYPE = TSIG def __init__( self, algorithm=None, timeSigned=None, fudge=5, MAC=None, originalID=0, error=OK, otherData=b"", ttl=0, ): # All of our init arguments have to have defaults, because of # the way IEncodable and Message.parseRecords() work, but for # some of our arguments there is no reasonable default; we use # invalid values here to prevent a user of this class from # relying on what's really an internal implementation detail. self.algorithm = None if algorithm is None else Name(algorithm) self.timeSigned = timeSigned self.fudge = str2time(fudge) self.MAC = MAC self.originalID = originalID self.error = error self.otherData = otherData self.ttl = ttl def encode(self, strio, compDict=None): self.algorithm.encode(strio, compDict) strio.write(struct.pack("!Q", self.timeSigned)[2:]) # 48-bit number strio.write(struct.pack("!HH", self.fudge, len(self.MAC))) strio.write(self.MAC) strio.write( struct.pack("!HHH", self.originalID, self.error, len(self.otherData)) ) strio.write(self.otherData) def decode(self, strio, length=None): algorithm = Name() algorithm.decode(strio) self.algorithm = algorithm fields = struct.unpack("!QHH", b"\x00\x00" + readPrecisely(strio, 10)) self.timeSigned, self.fudge, macLength = fields self.MAC = readPrecisely(strio, macLength) fields = struct.unpack("!HHH", readPrecisely(strio, 6)) self.originalID, self.error, otherLength = fields self.otherData = readPrecisely(strio, otherLength) def __hash__(self): return hash((self.algorithm, self.timeSigned, self.MAC, self.originalID)) def _responseFromMessage(responseConstructor, message, **kwargs): """ Generate a L{Message} like instance suitable for use as the response to C{message}. The C{queries}, C{id} attributes will be copied from C{message} and the C{answer} flag will be set to L{True}. @param responseConstructor: A response message constructor with an initializer signature matching L{dns.Message.__init__}. @type responseConstructor: C{callable} @param message: A request message. @type message: L{Message} @param kwargs: Keyword arguments which will be passed to the initialiser of the response message. @type kwargs: L{dict} @return: A L{Message} like response instance. @rtype: C{responseConstructor} """ response = responseConstructor(id=message.id, answer=True, **kwargs) response.queries = message.queries[:] return response def _getDisplayableArguments(obj, alwaysShow, fieldNames): """ Inspect the function signature of C{obj}'s constructor, and get a list of which arguments should be displayed. This is a helper function for C{_compactRepr}. @param obj: The instance whose repr is being generated. @param alwaysShow: A L{list} of field names which should always be shown. @param fieldNames: A L{list} of field attribute names which should be shown if they have non-default values. @return: A L{list} of displayable arguments. """ displayableArgs = [] # Get the argument names and values from the constructor. signature = inspect.signature(obj.__class__.__init__) for name in fieldNames: defaultValue = signature.parameters[name].default fieldValue = getattr(obj, name, defaultValue) if (name in alwaysShow) or (fieldValue != defaultValue): displayableArgs.append(f" {name}={fieldValue!r}") return displayableArgs def _compactRepr( obj, alwaysShow=None, flagNames=None, fieldNames=None, sectionNames=None ): """ Return a L{str} representation of C{obj} which only shows fields with non-default values, flags which are True and sections which have been explicitly set. @param obj: The instance whose repr is being generated. @param alwaysShow: A L{list} of field names which should always be shown. @param flagNames: A L{list} of flag attribute names which should be shown if they are L{True}. @param fieldNames: A L{list} of field attribute names which should be shown if they have non-default values. @param sectionNames: A L{list} of section attribute names which should be shown if they have been assigned a value. @return: A L{str} representation of C{obj}. """ if alwaysShow is None: alwaysShow = [] if flagNames is None: flagNames = [] if fieldNames is None: fieldNames = [] if sectionNames is None: sectionNames = [] setFlags = [] for name in flagNames: if name in alwaysShow or getattr(obj, name, False) == True: setFlags.append(name) displayableArgs = _getDisplayableArguments(obj, alwaysShow, fieldNames) out = ["<", obj.__class__.__name__] + displayableArgs if setFlags: out.append(" flags={}".format(",".join(setFlags))) for name in sectionNames: section = getattr(obj, name, []) if section: out.append(f" {name}={section!r}") out.append(">") return "".join(out) class Message(tputil.FancyEqMixin): """ L{Message} contains all the information represented by a single DNS request or response. @ivar id: See L{__init__} @ivar answer: See L{__init__} @ivar opCode: See L{__init__} @ivar recDes: See L{__init__} @ivar recAv: See L{__init__} @ivar auth: See L{__init__} @ivar rCode: See L{__init__} @ivar trunc: See L{__init__} @ivar maxSize: See L{__init__} @ivar authenticData: See L{__init__} @ivar checkingDisabled: See L{__init__} @ivar queries: The queries which are being asked of or answered by DNS server. @type queries: L{list} of L{Query} @ivar answers: Records containing the answers to C{queries} if this is a response message. @type answers: L{list} of L{RRHeader} @ivar authority: Records containing information about the authoritative DNS servers for the names in C{queries}. @type authority: L{list} of L{RRHeader} @ivar additional: Records containing IP addresses of host names in C{answers} and C{authority}. @type additional: L{list} of L{RRHeader} @ivar _flagNames: The names of attributes representing the flag header fields. @ivar _fieldNames: The names of attributes representing non-flag fixed header fields. @ivar _sectionNames: The names of attributes representing the record sections of this message. """ compareAttributes = ( "id", "answer", "opCode", "recDes", "recAv", "auth", "rCode", "trunc", "maxSize", "authenticData", "checkingDisabled", "queries", "answers", "authority", "additional", ) headerFmt = "!H2B4H" headerSize = struct.calcsize(headerFmt) # Question, answer, additional, and nameserver lists queries = answers = add = ns = None def __init__( self, id=0, answer=0, opCode=0, recDes=0, recAv=0, auth=0, rCode=OK, trunc=0, maxSize=512, authenticData=0, checkingDisabled=0, ): """ @param id: A 16 bit identifier assigned by the program that generates any kind of query. This identifier is copied to the corresponding reply and can be used by the requester to match up replies to outstanding queries. @type id: L{int} @param answer: A one bit field that specifies whether this message is a query (0), or a response (1). @type answer: L{int} @param opCode: A four bit field that specifies kind of query in this message. This value is set by the originator of a query and copied into the response. @type opCode: L{int} @param recDes: Recursion Desired - this bit may be set in a query and is copied into the response. If RD is set, it directs the name server to pursue the query recursively. Recursive query support is optional. @type recDes: L{int} @param recAv: Recursion Available - this bit is set or cleared in a response and denotes whether recursive query support is available in the name server. @type recAv: L{int} @param auth: Authoritative Answer - this bit is valid in responses and specifies that the responding name server is an authority for the domain name in question section. @type auth: L{int} @ivar rCode: A response code, used to indicate success or failure in a message which is a response from a server to a client request. @type rCode: C{0 <= int < 16} @param trunc: A flag indicating that this message was truncated due to length greater than that permitted on the transmission channel. @type trunc: L{int} @param maxSize: The requestor's UDP payload size is the number of octets of the largest UDP payload that can be reassembled and delivered in the requestor's network stack. @type maxSize: L{int} @param authenticData: A flag indicating in a response that all the data included in the answer and authority portion of the response has been authenticated by the server according to the policies of that server. See U{RFC2535 section-6.1<https://tools.ietf.org/html/rfc2535#section-6.1>}. @type authenticData: L{int} @param checkingDisabled: A flag indicating in a query that pending (non-authenticated) data is acceptable to the resolver sending the query. See U{RFC2535 section-6.1<https://tools.ietf.org/html/rfc2535#section-6.1>}. @type authenticData: L{int} """ self.maxSize = maxSize self.id = id self.answer = answer self.opCode = opCode self.auth = auth self.trunc = trunc self.recDes = recDes self.recAv = recAv self.rCode = rCode self.authenticData = authenticData self.checkingDisabled = checkingDisabled self.queries = [] self.answers = [] self.authority = [] self.additional = [] def __repr__(self) -> str: """ Generate a repr of this L{Message}. Only includes the non-default fields and sections and only includes flags which are set. The C{id} is always shown. @return: The native string repr. """ return _compactRepr( self, flagNames=( "answer", "auth", "trunc", "recDes", "recAv", "authenticData", "checkingDisabled", ), fieldNames=("id", "opCode", "rCode", "maxSize"), sectionNames=("queries", "answers", "authority", "additional"), alwaysShow=("id",), ) def addQuery(self, name, type=ALL_RECORDS, cls=IN): """ Add another query to this Message. @type name: L{bytes} @param name: The name to query. @type type: L{int} @param type: Query type @type cls: L{int} @param cls: Query class """ self.queries.append(Query(name, type, cls)) def encode(self, strio): compDict = {} body_tmp = BytesIO() for q in self.queries: q.encode(body_tmp, compDict) for q in self.answers: q.encode(body_tmp, compDict) for q in self.authority: q.encode(body_tmp, compDict) for q in self.additional: q.encode(body_tmp, compDict) body = body_tmp.getvalue() size = len(body) + self.headerSize if self.maxSize and size > self.maxSize: self.trunc = 1 body = body[: self.maxSize - self.headerSize] byte3 = ( ((self.answer & 1) << 7) | ((self.opCode & 0xF) << 3) | ((self.auth & 1) << 2) | ((self.trunc & 1) << 1) | (self.recDes & 1) ) byte4 = ( ((self.recAv & 1) << 7) | ((self.authenticData & 1) << 5) | ((self.checkingDisabled & 1) << 4) | (self.rCode & 0xF) ) strio.write( struct.pack( self.headerFmt, self.id, byte3, byte4, len(self.queries), len(self.answers), len(self.authority), len(self.additional), ) ) strio.write(body) def decode(self, strio, length=None): self.maxSize = 0 header = readPrecisely(strio, self.headerSize) r = struct.unpack(self.headerFmt, header) self.id, byte3, byte4, nqueries, nans, nns, nadd = r self.answer = (byte3 >> 7) & 1 self.opCode = (byte3 >> 3) & 0xF self.auth = (byte3 >> 2) & 1 self.trunc = (byte3 >> 1) & 1 self.recDes = byte3 & 1 self.recAv = (byte4 >> 7) & 1 self.authenticData = (byte4 >> 5) & 1 self.checkingDisabled = (byte4 >> 4) & 1 self.rCode = byte4 & 0xF self.queries = [] for i in range(nqueries): q = Query() try: q.decode(strio) except EOFError: return self.queries.append(q) items = ((self.answers, nans), (self.authority, nns), (self.additional, nadd)) for (l, n) in items: self.parseRecords(l, n, strio) def parseRecords(self, list, num, strio): for i in range(num): header = RRHeader(auth=self.auth) try: header.decode(strio) except EOFError: return t = self.lookupRecordType(header.type) if not t: continue header.payload = t(ttl=header.ttl) try: header.payload.decode(strio, header.rdlength) except EOFError: return list.append(header) # Create a mapping from record types to their corresponding Record_* # classes. This relies on the global state which has been created so # far in initializing this module (so don't define Record classes after # this). _recordTypes = {} for name in globals(): if name.startswith("Record_"): _recordTypes[globals()[name].TYPE] = globals()[name] # Clear the iteration variable out of the class namespace so it # doesn't become an attribute. del name def lookupRecordType(self, type): """ Retrieve the L{IRecord} implementation for the given record type. @param type: A record type, such as C{A} or L{NS}. @type type: L{int} @return: An object which implements L{IRecord} or L{None} if none can be found for the given type. @rtype: C{Type[IRecord]} """ return self._recordTypes.get(type, UnknownRecord) def toStr(self): """ Encode this L{Message} into a byte string in the format described by RFC 1035. @rtype: L{bytes} """ strio = BytesIO() self.encode(strio) return strio.getvalue() def fromStr(self, str): """ Decode a byte string in the format described by RFC 1035 into this L{Message}. @param str: L{bytes} """ strio = BytesIO(str) self.decode(strio) class _EDNSMessage(tputil.FancyEqMixin): """ An I{EDNS} message. Designed for compatibility with L{Message} but with a narrower public interface. Most importantly, L{_EDNSMessage.fromStr} will interpret and remove I{OPT} records that are present in the additional records section. The I{OPT} records are used to populate certain I{EDNS} specific attributes. L{_EDNSMessage.toStr} will add suitable I{OPT} records to the additional section to represent the extended EDNS information. @see: U{https://tools.ietf.org/html/rfc6891} @ivar id: See L{__init__} @ivar answer: See L{__init__} @ivar opCode: See L{__init__} @ivar auth: See L{__init__} @ivar trunc: See L{__init__} @ivar recDes: See L{__init__} @ivar recAv: See L{__init__} @ivar rCode: See L{__init__} @ivar ednsVersion: See L{__init__} @ivar dnssecOK: See L{__init__} @ivar authenticData: See L{__init__} @ivar checkingDisabled: See L{__init__} @ivar maxSize: See L{__init__} @ivar queries: See L{__init__} @ivar answers: See L{__init__} @ivar authority: See L{__init__} @ivar additional: See L{__init__} @ivar _messageFactory: A constructor of L{Message} instances. Called by C{_toMessage} and C{_fromMessage}. """ compareAttributes = ( "id", "answer", "opCode", "auth", "trunc", "recDes", "recAv", "rCode", "ednsVersion", "dnssecOK", "authenticData", "checkingDisabled", "maxSize", "queries", "answers", "authority", "additional", ) _messageFactory = Message def __init__( self, id=0, answer=False, opCode=OP_QUERY, auth=False, trunc=False, recDes=False, recAv=False, rCode=0, ednsVersion=0, dnssecOK=False, authenticData=False, checkingDisabled=False, maxSize=512, queries=None, answers=None, authority=None, additional=None, ): """ Construct a new L{_EDNSMessage} @see: U{RFC1035 section-4.1.1<https://tools.ietf.org/html/rfc1035#section-4.1.1>} @see: U{RFC2535 section-6.1<https://tools.ietf.org/html/rfc2535#section-6.1>} @see: U{RFC3225 section-3<https://tools.ietf.org/html/rfc3225#section-3>} @see: U{RFC6891 section-6.1.3<https://tools.ietf.org/html/rfc6891#section-6.1.3>} @param id: A 16 bit identifier assigned by the program that generates any kind of query. This identifier is copied the corresponding reply and can be used by the requester to match up replies to outstanding queries. @type id: L{int} @param answer: A one bit field that specifies whether this message is a query (0), or a response (1). @type answer: L{bool} @param opCode: A four bit field that specifies kind of query in this message. This value is set by the originator of a query and copied into the response. @type opCode: L{int} @param auth: Authoritative Answer - this bit is valid in responses, and specifies that the responding name server is an authority for the domain name in question section. @type auth: L{bool} @param trunc: Truncation - specifies that this message was truncated due to length greater than that permitted on the transmission channel. @type trunc: L{bool} @param recDes: Recursion Desired - this bit may be set in a query and is copied into the response. If set, it directs the name server to pursue the query recursively. Recursive query support is optional. @type recDes: L{bool} @param recAv: Recursion Available - this bit is set or cleared in a response, and denotes whether recursive query support is available in the name server. @type recAv: L{bool} @param rCode: Extended 12-bit RCODE. Derived from the 4 bits defined in U{RFC1035 4.1.1<https://tools.ietf.org/html/rfc1035#section-4.1.1>} and the upper 8bits defined in U{RFC6891 6.1.3<https://tools.ietf.org/html/rfc6891#section-6.1.3>}. @type rCode: L{int} @param ednsVersion: Indicates the EDNS implementation level. Set to L{None} to prevent any EDNS attributes and options being added to the encoded byte string. @type ednsVersion: L{int} or L{None} @param dnssecOK: DNSSEC OK bit as defined by U{RFC3225 3<https://tools.ietf.org/html/rfc3225#section-3>}. @type dnssecOK: L{bool} @param authenticData: A flag indicating in a response that all the data included in the answer and authority portion of the response has been authenticated by the server according to the policies of that server. See U{RFC2535 section-6.1<https://tools.ietf.org/html/rfc2535#section-6.1>}. @type authenticData: L{bool} @param checkingDisabled: A flag indicating in a query that pending (non-authenticated) data is acceptable to the resolver sending the query. See U{RFC2535 section-6.1<https://tools.ietf.org/html/rfc2535#section-6.1>}. @type authenticData: L{bool} @param maxSize: The requestor's UDP payload size is the number of octets of the largest UDP payload that can be reassembled and delivered in the requestor's network stack. @type maxSize: L{int} @param queries: The L{list} of L{Query} associated with this message. @type queries: L{list} of L{Query} @param answers: The L{list} of answers associated with this message. @type answers: L{list} of L{RRHeader} @param authority: The L{list} of authority records associated with this message. @type authority: L{list} of L{RRHeader} @param additional: The L{list} of additional records associated with this message. @type additional: L{list} of L{RRHeader} """ self.id = id self.answer = answer self.opCode = opCode self.auth = auth self.trunc = trunc self.recDes = recDes self.recAv = recAv self.rCode = rCode self.ednsVersion = ednsVersion self.dnssecOK = dnssecOK self.authenticData = authenticData self.checkingDisabled = checkingDisabled self.maxSize = maxSize if queries is None: queries = [] self.queries = queries if answers is None: answers = [] self.answers = answers if authority is None: authority = [] self.authority = authority if additional is None: additional = [] self.additional = additional def __repr__(self) -> str: return _compactRepr( self, flagNames=( "answer", "auth", "trunc", "recDes", "recAv", "authenticData", "checkingDisabled", "dnssecOK", ), fieldNames=("id", "opCode", "rCode", "maxSize", "ednsVersion"), sectionNames=("queries", "answers", "authority", "additional"), alwaysShow=("id",), ) def _toMessage(self): """ Convert to a standard L{dns.Message}. If C{ednsVersion} is not None, an L{_OPTHeader} instance containing all the I{EDNS} specific attributes and options will be appended to the list of C{additional} records. @return: A L{dns.Message} @rtype: L{dns.Message} """ m = self._messageFactory( id=self.id, answer=self.answer, opCode=self.opCode, auth=self.auth, trunc=self.trunc, recDes=self.recDes, recAv=self.recAv, # Assign the lower 4 bits to the message rCode=self.rCode & 0xF, authenticData=self.authenticData, checkingDisabled=self.checkingDisabled, ) m.queries = self.queries[:] m.answers = self.answers[:] m.authority = self.authority[:] m.additional = self.additional[:] if self.ednsVersion is not None: o = _OPTHeader( version=self.ednsVersion, dnssecOK=self.dnssecOK, udpPayloadSize=self.maxSize, # Assign the upper 8 bits to the OPT record extendedRCODE=self.rCode >> 4, ) m.additional.append(o) return m def toStr(self): """ Encode to wire format by first converting to a standard L{dns.Message}. @return: A L{bytes} string. """ return self._toMessage().toStr() @classmethod def _fromMessage(cls, message): """ Construct and return a new L{_EDNSMessage} whose attributes and records are derived from the attributes and records of C{message} (a L{Message} instance). If present, an C{OPT} record will be extracted from the C{additional} section and its attributes and options will be used to set the EDNS specific attributes C{extendedRCODE}, C{ednsVersion}, C{dnssecOK}, C{ednsOptions}. The C{extendedRCODE} will be combined with C{message.rCode} and assigned to C{self.rCode}. @param message: The source L{Message}. @type message: L{Message} @return: A new L{_EDNSMessage} @rtype: L{_EDNSMessage} """ additional = [] optRecords = [] for r in message.additional: if r.type == OPT: optRecords.append(_OPTHeader.fromRRHeader(r)) else: additional.append(r) newMessage = cls( id=message.id, answer=message.answer, opCode=message.opCode, auth=message.auth, trunc=message.trunc, recDes=message.recDes, recAv=message.recAv, rCode=message.rCode, authenticData=message.authenticData, checkingDisabled=message.checkingDisabled, # Default to None, it will be updated later when the OPT records are # parsed. ednsVersion=None, dnssecOK=False, queries=message.queries[:], answers=message.answers[:], authority=message.authority[:], additional=additional, ) if len(optRecords) == 1: # XXX: If multiple OPT records are received, an EDNS server should # respond with FORMERR. See ticket:5669#comment:1. opt = optRecords[0] newMessage.ednsVersion = opt.version newMessage.dnssecOK = opt.dnssecOK newMessage.maxSize = opt.udpPayloadSize newMessage.rCode = opt.extendedRCODE << 4 | message.rCode return newMessage def fromStr(self, bytes): """ Decode from wire format, saving flags, values and records to this L{_EDNSMessage} instance in place. @param bytes: The full byte string to be decoded. @type bytes: L{bytes} """ m = self._messageFactory() m.fromStr(bytes) ednsMessage = self._fromMessage(m) for attrName in self.compareAttributes: setattr(self, attrName, getattr(ednsMessage, attrName)) class DNSMixin: """ DNS protocol mixin shared by UDP and TCP implementations. @ivar _reactor: A L{IReactorTime} and L{IReactorUDP} provider which will be used to issue DNS queries and manage request timeouts. """ id = None liveMessages = None def __init__(self, controller, reactor=None): self.controller = controller self.id = random.randrange(2 ** 10, 2 ** 15) if reactor is None: from twisted.internet import reactor self._reactor = reactor def pickID(self): """ Return a unique ID for queries. """ while True: id = randomSource() if id not in self.liveMessages: return id def callLater(self, period, func, *args): """ Wrapper around reactor.callLater, mainly for test purpose. """ return self._reactor.callLater(period, func, *args) def _query(self, queries, timeout, id, writeMessage): """ Send out a message with the given queries. @type queries: L{list} of C{Query} instances @param queries: The queries to transmit @type timeout: L{int} or C{float} @param timeout: How long to wait before giving up @type id: L{int} @param id: Unique key for this request @type writeMessage: C{callable} @param writeMessage: One-parameter callback which writes the message @rtype: C{Deferred} @return: a C{Deferred} which will be fired with the result of the query, or errbacked with any errors that could happen (exceptions during writing of the query, timeout errors, ...). """ m = Message(id, recDes=1) m.queries = queries try: writeMessage(m) except BaseException: return defer.fail() resultDeferred = defer.Deferred() cancelCall = self.callLater(timeout, self._clearFailed, resultDeferred, id) self.liveMessages[id] = (resultDeferred, cancelCall) return resultDeferred def _clearFailed(self, deferred, id): """ Clean the Deferred after a timeout. """ try: del self.liveMessages[id] except KeyError: pass deferred.errback(failure.Failure(DNSQueryTimeoutError(id))) class DNSDatagramProtocol(DNSMixin, protocol.DatagramProtocol): """ DNS protocol over UDP. """ resends = None def stopProtocol(self): """ Stop protocol: reset state variables. """ self.liveMessages = {} self.resends = {} self.transport = None def startProtocol(self): """ Upon start, reset internal state. """ self.liveMessages = {} self.resends = {} def writeMessage(self, message, address): """ Send a message holding DNS queries. @type message: L{Message} """ self.transport.write(message.toStr(), address) def startListening(self): self._reactor.listenUDP(0, self, maxPacketSize=512) def datagramReceived(self, data, addr): """ Read a datagram, extract the message in it and trigger the associated Deferred. """ m = Message() try: m.fromStr(data) except EOFError: log.msg("Truncated packet (%d bytes) from %s" % (len(data), addr)) return except BaseException: # Nothing should trigger this, but since we're potentially # invoking a lot of different decoding methods, we might as well # be extra cautious. Anything that triggers this is itself # buggy. log.err(failure.Failure(), "Unexpected decoding error") return if m.id in self.liveMessages: d, canceller = self.liveMessages[m.id] del self.liveMessages[m.id] canceller.cancel() # XXX we shouldn't need this hack of catching exception on callback() try: d.callback(m) except BaseException: log.err() else: if m.id not in self.resends: self.controller.messageReceived(m, self, addr) def removeResend(self, id): """ Mark message ID as no longer having duplication suppression. """ try: del self.resends[id] except KeyError: pass def query(self, address, queries, timeout=10, id=None): """ Send out a message with the given queries. @type address: L{tuple} of L{str} and L{int} @param address: The address to which to send the query @type queries: L{list} of C{Query} instances @param queries: The queries to transmit @rtype: C{Deferred} """ if not self.transport: # XXX transport might not get created automatically, use callLater? try: self.startListening() except CannotListenError: return defer.fail() if id is None: id = self.pickID() else: self.resends[id] = 1 def writeMessage(m): self.writeMessage(m, address) return self._query(queries, timeout, id, writeMessage) class DNSProtocol(DNSMixin, protocol.Protocol): """ DNS protocol over TCP. """ length = None buffer = b"" def writeMessage(self, message): """ Send a message holding DNS queries. @type message: L{Message} """ s = message.toStr() self.transport.write(struct.pack("!H", len(s)) + s) def connectionMade(self): """ Connection is made: reset internal state, and notify the controller. """ self.liveMessages = {} self.controller.connectionMade(self) def connectionLost(self, reason): """ Notify the controller that this protocol is no longer connected. """ self.controller.connectionLost(self) def dataReceived(self, data): self.buffer += data while self.buffer: if self.length is None and len(self.buffer) >= 2: self.length = struct.unpack("!H", self.buffer[:2])[0] self.buffer = self.buffer[2:] if len(self.buffer) >= self.length: myChunk = self.buffer[: self.length] m = Message() m.fromStr(myChunk) try: d, canceller = self.liveMessages[m.id] except KeyError: self.controller.messageReceived(m, self) else: del self.liveMessages[m.id] canceller.cancel() # XXX we shouldn't need this hack try: d.callback(m) except BaseException: log.err() self.buffer = self.buffer[self.length :] self.length = None else: break def query(self, queries, timeout=60): """ Send out a message with the given queries. @type queries: L{list} of C{Query} instances @param queries: The queries to transmit @rtype: C{Deferred} """ id = self.pickID() return self._query(queries, timeout, id, self.writeMessage)