bazarr/libs/dns/resolver.py

1576 lines
57 KiB
Python
Raw Normal View History

# Copyright (C) Dnspython Contributors, see LICENSE for text of ISC license
# Copyright (C) 2003-2017 Nominum, Inc.
2018-10-31 16:08:29 +00:00
#
# Permission to use, copy, modify, and distribute this software and its
# documentation for any purpose with or without fee is hereby granted,
# provided that the above copyright notice and this permission notice
# appear in all copies.
#
# THE SOFTWARE IS PROVIDED "AS IS" AND NOMINUM DISCLAIMS ALL WARRANTIES
# WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
# MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL NOMINUM BE LIABLE FOR
# ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
# WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
# ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
# OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
"""DNS stub resolver."""
from urllib.parse import urlparse
import contextlib
2018-10-31 16:08:29 +00:00
import socket
import sys
import time
import random
import warnings
2018-10-31 16:08:29 +00:00
try:
import threading as _threading
except ImportError: # pragma: no cover
import dummy_threading as _threading # type: ignore
2018-10-31 16:08:29 +00:00
import dns.exception
import dns.flags
import dns.inet
2018-10-31 16:08:29 +00:00
import dns.ipv4
import dns.ipv6
import dns.message
import dns.name
import dns.query
import dns.rcode
import dns.rdataclass
import dns.rdatatype
import dns.reversename
import dns.tsig
if sys.platform == 'win32':
import dns.win32util
2018-10-31 16:08:29 +00:00
class NXDOMAIN(dns.exception.DNSException):
"""The DNS query name does not exist."""
supp_kwargs = {'qnames', 'responses'}
2018-10-31 16:08:29 +00:00
fmt = None # we have our own __str__ implementation
# pylint: disable=arguments-differ
def _check_kwargs(self, qnames,
responses=None):
2018-10-31 16:08:29 +00:00
if not isinstance(qnames, (list, tuple, set)):
raise AttributeError("qnames must be a list, tuple or set")
if len(qnames) == 0:
raise AttributeError("qnames must contain at least one element")
if responses is None:
responses = {}
elif not isinstance(responses, dict):
raise AttributeError("responses must be a dict(qname=response)")
kwargs = dict(qnames=qnames, responses=responses)
return kwargs
def __str__(self):
if 'qnames' not in self.kwargs:
return super().__str__()
2018-10-31 16:08:29 +00:00
qnames = self.kwargs['qnames']
if len(qnames) > 1:
msg = 'None of DNS query names exist'
else:
msg = 'The DNS query name does not exist'
2018-10-31 16:08:29 +00:00
qnames = ', '.join(map(str, qnames))
return "{}: {}".format(msg, qnames)
2018-10-31 16:08:29 +00:00
@property
2018-10-31 16:08:29 +00:00
def canonical_name(self):
"""Return the unresolved canonical name."""
if 'qnames' not in self.kwargs:
2018-10-31 16:08:29 +00:00
raise TypeError("parametrized exception required")
for qname in self.kwargs['qnames']:
response = self.kwargs['responses'][qname]
try:
cname = response.canonical_name()
if cname != qname:
return cname
except Exception:
# We can just eat this exception as it means there was
# something wrong with the response.
pass
2018-10-31 16:08:29 +00:00
return self.kwargs['qnames'][0]
def __add__(self, e_nx):
"""Augment by results from another NXDOMAIN exception."""
qnames0 = list(self.kwargs.get('qnames', []))
responses0 = dict(self.kwargs.get('responses', {}))
responses1 = e_nx.kwargs.get('responses', {})
for qname1 in e_nx.kwargs.get('qnames', []):
if qname1 not in qnames0:
qnames0.append(qname1)
if qname1 in responses1:
responses0[qname1] = responses1[qname1]
return NXDOMAIN(qnames=qnames0, responses=responses0)
def qnames(self):
"""All of the names that were tried.
2018-10-31 16:08:29 +00:00
Returns a list of ``dns.name.Name``.
"""
return self.kwargs['qnames']
def responses(self):
"""A map from queried names to their NXDOMAIN responses.
Returns a dict mapping a ``dns.name.Name`` to a
``dns.message.Message``.
"""
return self.kwargs['responses']
def response(self, qname):
"""The response for query *qname*.
Returns a ``dns.message.Message``.
"""
return self.kwargs['responses'][qname]
2018-10-31 16:08:29 +00:00
class YXDOMAIN(dns.exception.DNSException):
2018-10-31 16:08:29 +00:00
"""The DNS query name is too long after DNAME substitution."""
def _errors_to_text(errors):
"""Turn a resolution errors trace into a list of text."""
texts = []
for err in errors:
texts.append('Server {} {} port {} answered {}'.format(err[0],
'TCP' if err[1] else 'UDP', err[2], err[3]))
return texts
2018-10-31 16:08:29 +00:00
class LifetimeTimeout(dns.exception.Timeout):
"""The resolution lifetime expired."""
msg = "The resolution lifetime expired."
fmt = "%s after {timeout} seconds: {errors}" % msg[:-1]
supp_kwargs = {'timeout', 'errors'}
def _fmt_kwargs(self, **kwargs):
srv_msgs = _errors_to_text(kwargs['errors'])
return super()._fmt_kwargs(timeout=kwargs['timeout'],
errors='; '.join(srv_msgs))
# We added more detail to resolution timeouts, but they are still
# subclasses of dns.exception.Timeout for backwards compatibility. We also
# keep dns.resolver.Timeout defined for backwards compatibility.
Timeout = LifetimeTimeout
2018-10-31 16:08:29 +00:00
class NoAnswer(dns.exception.DNSException):
2018-10-31 16:08:29 +00:00
"""The DNS response does not contain an answer to the question."""
fmt = 'The DNS response does not contain an answer ' + \
'to the question: {query}'
supp_kwargs = {'response'}
2018-10-31 16:08:29 +00:00
def _fmt_kwargs(self, **kwargs):
return super()._fmt_kwargs(query=kwargs['response'].question)
2018-10-31 16:08:29 +00:00
def response(self):
return self.kwargs['response']
2018-10-31 16:08:29 +00:00
class NoNameservers(dns.exception.DNSException):
2018-10-31 16:08:29 +00:00
"""All nameservers failed to answer the query.
errors: list of servers and respective errors
The type of errors is
[(server IP address, any object convertible to string)].
2018-10-31 16:08:29 +00:00
Non-empty errors list will add explanatory message ()
"""
msg = "All nameservers failed to answer the query."
fmt = "%s {query}: {errors}" % msg[:-1]
supp_kwargs = {'request', 'errors'}
2018-10-31 16:08:29 +00:00
def _fmt_kwargs(self, **kwargs):
srv_msgs = _errors_to_text(kwargs['errors'])
return super()._fmt_kwargs(query=kwargs['request'].question,
errors='; '.join(srv_msgs))
2018-10-31 16:08:29 +00:00
class NotAbsolute(dns.exception.DNSException):
"""An absolute domain name is required but a relative name was provided."""
class NoRootSOA(dns.exception.DNSException):
"""There is no SOA RR at the DNS root name. This should never happen!"""
class NoMetaqueries(dns.exception.DNSException):
"""DNS metaqueries are not allowed."""
class NoResolverConfiguration(dns.exception.DNSException):
"""Resolver configuration could not be read or specified no nameservers."""
2018-10-31 16:08:29 +00:00
class Answer:
"""DNS stub resolver answer.
2018-10-31 16:08:29 +00:00
Instances of this class bundle up the result of a successful DNS
resolution.
For convenience, the answer object implements much of the sequence
protocol, forwarding to its ``rrset`` attribute. E.g.
``for a in answer`` is equivalent to ``for a in answer.rrset``.
``answer[i]`` is equivalent to ``answer.rrset[i]``, and
``answer[i:j]`` is equivalent to ``answer.rrset[i:j]``.
2018-10-31 16:08:29 +00:00
Note that CNAMEs or DNAMEs in the response may mean that answer
RRset's name might not be the query name.
2018-10-31 16:08:29 +00:00
"""
def __init__(self, qname, rdtype, rdclass, response, nameserver=None,
port=None):
2018-10-31 16:08:29 +00:00
self.qname = qname
self.rdtype = rdtype
self.rdclass = rdclass
self.response = response
self.nameserver = nameserver
self.port = port
self.chaining_result = response.resolve_chaining()
# Copy some attributes out of chaining_result for backwards
# compatibility and convenience.
self.canonical_name = self.chaining_result.canonical_name
self.rrset = self.chaining_result.answer
self.expiration = time.time() + self.chaining_result.minimum_ttl
def __getattr__(self, attr): # pragma: no cover
2018-10-31 16:08:29 +00:00
if attr == 'name':
return self.rrset.name
elif attr == 'ttl':
return self.rrset.ttl
elif attr == 'covers':
return self.rrset.covers
elif attr == 'rdclass':
return self.rrset.rdclass
elif attr == 'rdtype':
return self.rrset.rdtype
else:
raise AttributeError(attr)
def __len__(self):
return self.rrset and len(self.rrset) or 0
def __iter__(self):
return self.rrset and iter(self.rrset) or iter(tuple())
def __getitem__(self, i):
if self.rrset is None:
raise IndexError
2018-10-31 16:08:29 +00:00
return self.rrset[i]
def __delitem__(self, i):
if self.rrset is None:
raise IndexError
2018-10-31 16:08:29 +00:00
del self.rrset[i]
class CacheStatistics:
"""Cache Statistics
"""
2018-10-31 16:08:29 +00:00
def __init__(self, hits=0, misses=0):
self.hits = hits
self.misses = misses
2018-10-31 16:08:29 +00:00
def reset(self):
self.hits = 0
self.misses = 0
2018-10-31 16:08:29 +00:00
def clone(self):
return CacheStatistics(self.hits, self.misses)
class CacheBase:
def __init__(self):
self.lock = _threading.Lock()
self.statistics = CacheStatistics()
def reset_statistics(self):
"""Reset all statistics to zero."""
with self.lock:
self.statistics.reset()
def hits(self):
"""How many hits has the cache had?"""
with self.lock:
return self.statistics.hits
2018-10-31 16:08:29 +00:00
def misses(self):
"""How many misses has the cache had?"""
with self.lock:
return self.statistics.misses
def get_statistics_snapshot(self):
"""Return a consistent snapshot of all the statistics.
If running with multiple threads, it's better to take a
snapshot than to call statistics methods such as hits() and
misses() individually.
"""
with self.lock:
return self.statistics.clone()
class Cache(CacheBase):
"""Simple thread-safe DNS answer cache."""
def __init__(self, cleaning_interval=300.0):
"""*cleaning_interval*, a ``float`` is the number of seconds between
periodic cleanings.
2018-10-31 16:08:29 +00:00
"""
super().__init__()
2018-10-31 16:08:29 +00:00
self.data = {}
self.cleaning_interval = cleaning_interval
self.next_cleaning = time.time() + self.cleaning_interval
def _maybe_clean(self):
"""Clean the cache if it's time to do so."""
now = time.time()
if self.next_cleaning <= now:
keys_to_delete = []
for (k, v) in self.data.items():
if v.expiration <= now:
keys_to_delete.append(k)
for k in keys_to_delete:
del self.data[k]
now = time.time()
self.next_cleaning = now + self.cleaning_interval
def get(self, key):
"""Get the answer associated with *key*.
Returns None if no answer is cached for the key.
*key*, a ``(dns.name.Name, int, int)`` tuple whose values are the
query name, rdtype, and rdclass respectively.
Returns a ``dns.resolver.Answer`` or ``None``.
2018-10-31 16:08:29 +00:00
"""
with self.lock:
2018-10-31 16:08:29 +00:00
self._maybe_clean()
v = self.data.get(key)
if v is None or v.expiration <= time.time():
self.statistics.misses += 1
2018-10-31 16:08:29 +00:00
return None
self.statistics.hits += 1
2018-10-31 16:08:29 +00:00
return v
def put(self, key, value):
"""Associate key and value in the cache.
*key*, a ``(dns.name.Name, int, int)`` tuple whose values are the
query name, rdtype, and rdclass respectively.
*value*, a ``dns.resolver.Answer``, the answer.
2018-10-31 16:08:29 +00:00
"""
with self.lock:
2018-10-31 16:08:29 +00:00
self._maybe_clean()
self.data[key] = value
def flush(self, key=None):
"""Flush the cache.
If *key* is not ``None``, only that item is flushed. Otherwise
2018-10-31 16:08:29 +00:00
the entire cache is flushed.
*key*, a ``(dns.name.Name, int, int)`` tuple whose values are the
query name, rdtype, and rdclass respectively.
2018-10-31 16:08:29 +00:00
"""
with self.lock:
2018-10-31 16:08:29 +00:00
if key is not None:
if key in self.data:
del self.data[key]
else:
self.data = {}
self.next_cleaning = time.time() + self.cleaning_interval
class LRUCacheNode:
"""LRUCache node."""
2018-10-31 16:08:29 +00:00
def __init__(self, key, value):
self.key = key
self.value = value
self.hits = 0
2018-10-31 16:08:29 +00:00
self.prev = self
self.next = self
def link_after(self, node):
self.prev = node
self.next = node.next
node.next.prev = self
node.next = self
def unlink(self):
self.next.prev = self.prev
self.prev.next = self.next
class LRUCache(CacheBase):
"""Thread-safe, bounded, least-recently-used DNS answer cache.
2018-10-31 16:08:29 +00:00
This cache is better than the simple cache (above) if you're
running a web crawler or other process that does a lot of
resolutions. The LRUCache has a maximum number of nodes, and when
it is full, the least-recently used node is removed to make space
for a new one.
"""
def __init__(self, max_size=100000):
"""*max_size*, an ``int``, is the maximum number of nodes to cache;
it must be greater than 0.
2018-10-31 16:08:29 +00:00
"""
super().__init__()
2018-10-31 16:08:29 +00:00
self.data = {}
self.set_max_size(max_size)
self.sentinel = LRUCacheNode(None, None)
self.sentinel.prev = self.sentinel
self.sentinel.next = self.sentinel
2018-10-31 16:08:29 +00:00
def set_max_size(self, max_size):
if max_size < 1:
max_size = 1
self.max_size = max_size
def get(self, key):
"""Get the answer associated with *key*.
Returns None if no answer is cached for the key.
*key*, a ``(dns.name.Name, int, int)`` tuple whose values are the
query name, rdtype, and rdclass respectively.
Returns a ``dns.resolver.Answer`` or ``None``.
2018-10-31 16:08:29 +00:00
"""
with self.lock:
2018-10-31 16:08:29 +00:00
node = self.data.get(key)
if node is None:
self.statistics.misses += 1
2018-10-31 16:08:29 +00:00
return None
# Unlink because we're either going to move the node to the front
# of the LRU list or we're going to free it.
node.unlink()
if node.value.expiration <= time.time():
del self.data[node.key]
self.statistics.misses += 1
2018-10-31 16:08:29 +00:00
return None
node.link_after(self.sentinel)
self.statistics.hits += 1
node.hits += 1
2018-10-31 16:08:29 +00:00
return node.value
def get_hits_for_key(self, key):
"""Return the number of cache hits associated with the specified key."""
with self.lock:
node = self.data.get(key)
if node is None or node.value.expiration <= time.time():
return 0
else:
return node.hits
2018-10-31 16:08:29 +00:00
def put(self, key, value):
"""Associate key and value in the cache.
*key*, a ``(dns.name.Name, int, int)`` tuple whose values are the
query name, rdtype, and rdclass respectively.
*value*, a ``dns.resolver.Answer``, the answer.
2018-10-31 16:08:29 +00:00
"""
with self.lock:
2018-10-31 16:08:29 +00:00
node = self.data.get(key)
if node is not None:
node.unlink()
del self.data[node.key]
while len(self.data) >= self.max_size:
node = self.sentinel.prev
node.unlink()
del self.data[node.key]
node = LRUCacheNode(key, value)
node.link_after(self.sentinel)
self.data[key] = node
def flush(self, key=None):
"""Flush the cache.
If *key* is not ``None``, only that item is flushed. Otherwise
2018-10-31 16:08:29 +00:00
the entire cache is flushed.
*key*, a ``(dns.name.Name, int, int)`` tuple whose values are the
query name, rdtype, and rdclass respectively.
2018-10-31 16:08:29 +00:00
"""
with self.lock:
2018-10-31 16:08:29 +00:00
if key is not None:
node = self.data.get(key)
if node is not None:
node.unlink()
del self.data[node.key]
else:
node = self.sentinel.next
while node != self.sentinel:
next = node.next
node.unlink()
2018-10-31 16:08:29 +00:00
node = next
self.data = {}
class _Resolution:
"""Helper class for dns.resolver.Resolver.resolve().
All of the "business logic" of resolution is encapsulated in this
class, allowing us to have multiple resolve() implementations
using different I/O schemes without copying all of the
complicated logic.
This class is a "friend" to dns.resolver.Resolver and manipulates
resolver data structures directly.
2018-10-31 16:08:29 +00:00
"""
def __init__(self, resolver, qname, rdtype, rdclass, tcp,
raise_on_no_answer, search):
if isinstance(qname, str):
qname = dns.name.from_text(qname, None)
rdtype = dns.rdatatype.RdataType.make(rdtype)
if dns.rdatatype.is_metatype(rdtype):
raise NoMetaqueries
rdclass = dns.rdataclass.RdataClass.make(rdclass)
if dns.rdataclass.is_metaclass(rdclass):
raise NoMetaqueries
self.resolver = resolver
self.qnames_to_try = resolver._get_qnames_to_try(qname, search)
self.qnames = self.qnames_to_try[:]
self.rdtype = rdtype
self.rdclass = rdclass
self.tcp = tcp
self.raise_on_no_answer = raise_on_no_answer
self.nxdomain_responses = {}
#
# Initialize other things to help analysis tools
self.qname = dns.name.empty
self.nameservers = []
self.current_nameservers = []
self.errors = []
self.nameserver = None
self.port = 0
self.tcp_attempt = False
self.retry_with_tcp = False
self.request = None
self.backoff = 0
def next_request(self):
"""Get the next request to send, and check the cache.
Returns a (request, answer) tuple. At most one of request or
answer will not be None.
"""
# We return a tuple instead of Union[Message,Answer] as it lets
# the caller avoid isinstance().
while len(self.qnames) > 0:
self.qname = self.qnames.pop(0)
# Do we know the answer?
if self.resolver.cache:
answer = self.resolver.cache.get((self.qname, self.rdtype,
self.rdclass))
if answer is not None:
if answer.rrset is None and self.raise_on_no_answer:
raise NoAnswer(response=answer.response)
else:
return (None, answer)
answer = self.resolver.cache.get((self.qname,
dns.rdatatype.ANY,
self.rdclass))
if answer is not None and \
answer.response.rcode() == dns.rcode.NXDOMAIN:
# cached NXDOMAIN; record it and continue to next
# name.
self.nxdomain_responses[self.qname] = answer.response
continue
# Build the request
request = dns.message.make_query(self.qname, self.rdtype,
self.rdclass)
if self.resolver.keyname is not None:
request.use_tsig(self.resolver.keyring, self.resolver.keyname,
algorithm=self.resolver.keyalgorithm)
request.use_edns(self.resolver.edns, self.resolver.ednsflags,
self.resolver.payload)
if self.resolver.flags is not None:
request.flags = self.resolver.flags
self.nameservers = self.resolver.nameservers[:]
if self.resolver.rotate:
random.shuffle(self.nameservers)
self.current_nameservers = self.nameservers[:]
self.errors = []
self.nameserver = None
self.tcp_attempt = False
self.retry_with_tcp = False
self.request = request
self.backoff = 0.10
return (request, None)
#
# We've tried everything and only gotten NXDOMAINs. (We know
# it's only NXDOMAINs as anything else would have returned
# before now.)
#
raise NXDOMAIN(qnames=self.qnames_to_try,
responses=self.nxdomain_responses)
def next_nameserver(self):
if self.retry_with_tcp:
assert self.nameserver is not None
self.tcp_attempt = True
self.retry_with_tcp = False
return (self.nameserver, self.port, True, 0)
backoff = 0
if not self.current_nameservers:
if len(self.nameservers) == 0:
# Out of things to try!
raise NoNameservers(request=self.request, errors=self.errors)
self.current_nameservers = self.nameservers[:]
backoff = self.backoff
self.backoff = min(self.backoff * 2, 2)
self.nameserver = self.current_nameservers.pop(0)
self.port = self.resolver.nameserver_ports.get(self.nameserver,
self.resolver.port)
self.tcp_attempt = self.tcp
return (self.nameserver, self.port, self.tcp_attempt, backoff)
def query_result(self, response, ex):
#
# returns an (answer: Answer, end_loop: bool) tuple.
#
if ex:
# Exception during I/O or from_wire()
assert response is None
self.errors.append((self.nameserver, self.tcp_attempt, self.port,
ex, response))
if isinstance(ex, dns.exception.FormError) or \
isinstance(ex, EOFError) or \
isinstance(ex, OSError) or \
isinstance(ex, NotImplementedError):
# This nameserver is no good, take it out of the mix.
self.nameservers.remove(self.nameserver)
elif isinstance(ex, dns.message.Truncated):
if self.tcp_attempt:
# Truncation with TCP is no good!
self.nameservers.remove(self.nameserver)
else:
self.retry_with_tcp = True
return (None, False)
# We got an answer!
assert response is not None
rcode = response.rcode()
if rcode == dns.rcode.NOERROR:
try:
answer = Answer(self.qname, self.rdtype, self.rdclass, response,
self.nameserver, self.port)
except Exception as e:
self.errors.append((self.nameserver, self.tcp_attempt,
self.port, e, response))
# The nameserver is no good, take it out of the mix.
self.nameservers.remove(self.nameserver)
return (None, False)
if self.resolver.cache:
self.resolver.cache.put((self.qname, self.rdtype,
self.rdclass), answer)
if answer.rrset is None and self.raise_on_no_answer:
raise NoAnswer(response=answer.response)
return (answer, True)
elif rcode == dns.rcode.NXDOMAIN:
# Further validate the response by making an Answer, even
# if we aren't going to cache it.
try:
answer = Answer(self.qname, dns.rdatatype.ANY,
dns.rdataclass.IN, response)
except Exception as e:
self.errors.append((self.nameserver, self.tcp_attempt,
self.port, e, response))
# The nameserver is no good, take it out of the mix.
self.nameservers.remove(self.nameserver)
return (None, False)
self.nxdomain_responses[self.qname] = response
if self.resolver.cache:
self.resolver.cache.put((self.qname,
dns.rdatatype.ANY,
self.rdclass), answer)
# Make next_nameserver() return None, so caller breaks its
# inner loop and calls next_request().
return (None, True)
elif rcode == dns.rcode.YXDOMAIN:
yex = YXDOMAIN()
self.errors.append((self.nameserver, self.tcp_attempt,
self.port, yex, response))
raise yex
else:
#
# We got a response, but we're not happy with the
# rcode in it.
#
if rcode != dns.rcode.SERVFAIL or not self.resolver.retry_servfail:
self.nameservers.remove(self.nameserver)
self.errors.append((self.nameserver, self.tcp_attempt, self.port,
dns.rcode.to_text(rcode), response))
return (None, False)
class BaseResolver:
"""DNS stub resolver."""
# We initialize in reset()
#
# pylint: disable=attribute-defined-outside-init
2018-10-31 16:08:29 +00:00
def __init__(self, filename='/etc/resolv.conf', configure=True):
"""*filename*, a ``str`` or file object, specifying a file
in standard /etc/resolv.conf format. This parameter is meaningful
only when *configure* is true and the platform is POSIX.
*configure*, a ``bool``. If True (the default), the resolver
instance is configured in the normal fashion for the operating
system the resolver is running on. (I.e. by reading a
/etc/resolv.conf file on POSIX systems and from the registry
on Windows systems.)
"""
2018-10-31 16:08:29 +00:00
self.reset()
if configure:
if sys.platform == 'win32':
self.read_registry()
elif filename:
self.read_resolv_conf(filename)
def reset(self):
"""Reset all resolver configuration to the defaults."""
2018-10-31 16:08:29 +00:00
self.domain = \
dns.name.Name(dns.name.from_text(socket.gethostname())[1:])
if len(self.domain) == 0:
self.domain = dns.name.root
self.nameservers = []
self.nameserver_ports = {}
self.port = 53
self.search = []
self.use_search_by_default = False
2018-10-31 16:08:29 +00:00
self.timeout = 2.0
self.lifetime = 5.0
2018-10-31 16:08:29 +00:00
self.keyring = None
self.keyname = None
self.keyalgorithm = dns.tsig.default_algorithm
self.edns = -1
self.ednsflags = 0
self.payload = 0
self.cache = None
self.flags = None
self.retry_servfail = False
self.rotate = False
self.ndots = None
2018-10-31 16:08:29 +00:00
def read_resolv_conf(self, f):
"""Process *f* as a file in the /etc/resolv.conf format. If f is
a ``str``, it is used as the name of the file to open; otherwise it
is treated as the file itself.
Interprets the following items:
- nameserver - name server IP address
- domain - local domain name
- search - search list for host-name lookup
- options - supported options are rotate, timeout, edns0, and ndots
"""
with contextlib.ExitStack() as stack:
if isinstance(f, str):
try:
f = stack.enter_context(open(f))
except OSError:
# /etc/resolv.conf doesn't exist, can't be read, etc.
raise NoResolverConfiguration(f'cannot open {f}')
2018-10-31 16:08:29 +00:00
for l in f:
if len(l) == 0 or l[0] == '#' or l[0] == ';':
continue
tokens = l.split()
# Any line containing less than 2 tokens is malformed
if len(tokens) < 2:
continue
if tokens[0] == 'nameserver':
self.nameservers.append(tokens[1])
elif tokens[0] == 'domain':
self.domain = dns.name.from_text(tokens[1])
# domain and search are exclusive
self.search = []
2018-10-31 16:08:29 +00:00
elif tokens[0] == 'search':
# the last search wins
self.search = []
2018-10-31 16:08:29 +00:00
for suffix in tokens[1:]:
self.search.append(dns.name.from_text(suffix))
# We don't set domain as it is not used if
# len(self.search) > 0
2018-10-31 16:08:29 +00:00
elif tokens[0] == 'options':
for opt in tokens[1:]:
if opt == 'rotate':
self.rotate = True
elif opt == 'edns0':
self.use_edns()
elif 'timeout' in opt:
try:
self.timeout = int(opt.split(':')[1])
except (ValueError, IndexError):
pass
elif 'ndots' in opt:
try:
self.ndots = int(opt.split(':')[1])
except (ValueError, IndexError):
pass
2018-10-31 16:08:29 +00:00
if len(self.nameservers) == 0:
raise NoResolverConfiguration('no nameservers')
2018-10-31 16:08:29 +00:00
def read_registry(self):
"""Extract resolver configuration from the Windows registry."""
try:
info = dns.win32util.get_dns_info()
if info.domain is not None:
self.domain = info.domain
self.nameservers = info.nameservers
self.search = info.search
except AttributeError:
raise NotImplementedError
def _compute_timeout(self, start, lifetime=None, errors=None):
lifetime = self.lifetime if lifetime is None else lifetime
2018-10-31 16:08:29 +00:00
now = time.time()
duration = now - start
if errors is None:
errors = []
2018-10-31 16:08:29 +00:00
if duration < 0:
if duration < -1:
# Time going backwards is bad. Just give up.
raise LifetimeTimeout(timeout=duration, errors=errors)
2018-10-31 16:08:29 +00:00
else:
# Time went backwards, but only a little. This can
# happen, e.g. under vmware with older linux kernels.
# Pretend it didn't happen.
now = start
if duration >= lifetime:
raise LifetimeTimeout(timeout=duration, errors=errors)
return min(lifetime - duration, self.timeout)
def _get_qnames_to_try(self, qname, search):
# This is a separate method so we can unit test the search
# rules without requiring the Internet.
if search is None:
search = self.use_search_by_default
2018-10-31 16:08:29 +00:00
qnames_to_try = []
if qname.is_absolute():
qnames_to_try.append(qname)
else:
abs_qname = qname.concatenate(dns.name.root)
if search:
if len(self.search) > 0:
# There is a search list, so use it exclusively
search_list = self.search[:]
elif self.domain != dns.name.root and self.domain is not None:
# We have some notion of a domain that isn't the root, so
# use it as the search list.
search_list = [self.domain]
else:
search_list = []
# Figure out the effective ndots (default is 1)
if self.ndots is None:
ndots = 1
else:
ndots = self.ndots
for suffix in search_list:
qnames_to_try.append(qname + suffix)
if len(qname) > ndots:
# The name has at least ndots dots, so we should try an
# absolute query first.
qnames_to_try.insert(0, abs_qname)
else:
# The name has less than ndots dots, so we should search
# first, then try the absolute name.
qnames_to_try.append(abs_qname)
2018-10-31 16:08:29 +00:00
else:
qnames_to_try.append(abs_qname)
return qnames_to_try
2018-10-31 16:08:29 +00:00
def use_tsig(self, keyring, keyname=None,
algorithm=dns.tsig.default_algorithm):
"""Add a TSIG signature to each query.
The parameters are passed to ``dns.message.Message.use_tsig()``;
see its documentation for details.
"""
2018-10-31 16:08:29 +00:00
self.keyring = keyring
self.keyname = keyname
2018-10-31 16:08:29 +00:00
self.keyalgorithm = algorithm
def use_edns(self, edns=0, ednsflags=0,
payload=dns.message.DEFAULT_EDNS_PAYLOAD):
"""Configure EDNS behavior.
*edns*, an ``int``, is the EDNS level to use. Specifying
``None``, ``False``, or ``-1`` means "do not use EDNS", and in this case
the other parameters are ignored. Specifying ``True`` is
equivalent to specifying 0, i.e. "use EDNS0".
2018-10-31 16:08:29 +00:00
*ednsflags*, an ``int``, the EDNS flag values.
*payload*, an ``int``, is the EDNS sender's payload field, which is the
maximum size of UDP datagram the sender can handle. I.e. how big
a response to this message can be.
"""
2018-10-31 16:08:29 +00:00
if edns is None or edns is False:
2018-10-31 16:08:29 +00:00
edns = -1
elif edns is True:
edns = 0
2018-10-31 16:08:29 +00:00
self.edns = edns
self.ednsflags = ednsflags
self.payload = payload
def set_flags(self, flags):
"""Overrides the default flags with your own.
*flags*, an ``int``, the message flags to use.
"""
2018-10-31 16:08:29 +00:00
self.flags = flags
@property
def nameservers(self):
return self._nameservers
@nameservers.setter
def nameservers(self, nameservers):
"""
*nameservers*, a ``list`` of nameservers.
Raises ``ValueError`` if *nameservers* is anything other than a
``list``.
"""
if isinstance(nameservers, list):
for nameserver in nameservers:
if not dns.inet.is_address(nameserver):
try:
if urlparse(nameserver).scheme != 'https':
raise NotImplementedError
except Exception:
raise ValueError(f'nameserver {nameserver} is not an '
'IP address or valid https URL')
self._nameservers = nameservers
else:
raise ValueError('nameservers must be a list'
' (not a {})'.format(type(nameservers)))
class Resolver(BaseResolver):
"""DNS stub resolver."""
def resolve(self, qname, rdtype=dns.rdatatype.A, rdclass=dns.rdataclass.IN,
tcp=False, source=None, raise_on_no_answer=True, source_port=0,
lifetime=None, search=None): # pylint: disable=arguments-differ
"""Query nameservers to find the answer to the question.
The *qname*, *rdtype*, and *rdclass* parameters may be objects
of the appropriate type, or strings that can be converted into objects
of the appropriate type.
*qname*, a ``dns.name.Name`` or ``str``, the query name.
*rdtype*, an ``int`` or ``str``, the query type.
*rdclass*, an ``int`` or ``str``, the query class.
*tcp*, a ``bool``. If ``True``, use TCP to make the query.
*source*, a ``str`` or ``None``. If not ``None``, bind to this IP
address when making queries.
*raise_on_no_answer*, a ``bool``. If ``True``, raise
``dns.resolver.NoAnswer`` if there's no answer to the question.
*source_port*, an ``int``, the port from which to send the message.
*lifetime*, a ``float``, how many seconds a query should run
before timing out.
*search*, a ``bool`` or ``None``, determines whether the
search list configured in the system's resolver configuration
are used for relative names, and whether the resolver's domain
may be added to relative names. The default is ``None``,
which causes the value of the resolver's
``use_search_by_default`` attribute to be used.
Raises ``dns.resolver.LifetimeTimeout`` if no answers could be found
in the specified lifetime.
Raises ``dns.resolver.NXDOMAIN`` if the query name does not exist.
Raises ``dns.resolver.YXDOMAIN`` if the query name is too long after
DNAME substitution.
Raises ``dns.resolver.NoAnswer`` if *raise_on_no_answer* is
``True`` and the query name exists but has no RRset of the
desired type and class.
Raises ``dns.resolver.NoNameservers`` if no non-broken
nameservers are available to answer the question.
Returns a ``dns.resolver.Answer`` instance.
"""
resolution = _Resolution(self, qname, rdtype, rdclass, tcp,
raise_on_no_answer, search)
start = time.time()
while True:
(request, answer) = resolution.next_request()
# Note we need to say "if answer is not None" and not just
# "if answer" because answer implements __len__, and python
# will call that. We want to return if we have an answer
# object, including in cases where its length is 0.
if answer is not None:
# cache hit!
return answer
done = False
while not done:
(nameserver, port, tcp, backoff) = resolution.next_nameserver()
if backoff:
time.sleep(backoff)
timeout = self._compute_timeout(start, lifetime,
resolution.errors)
try:
if dns.inet.is_address(nameserver):
if tcp:
response = dns.query.tcp(request, nameserver,
timeout=timeout,
port=port,
source=source,
source_port=source_port)
else:
response = dns.query.udp(request,
nameserver,
timeout=timeout,
port=port,
source=source,
source_port=source_port,
raise_on_truncation=True)
else:
response = dns.query.https(request, nameserver,
timeout=timeout)
except Exception as ex:
(_, done) = resolution.query_result(None, ex)
continue
(answer, done) = resolution.query_result(response, None)
# Note we need to say "if answer is not None" and not just
# "if answer" because answer implements __len__, and python
# will call that. We want to return if we have an answer
# object, including in cases where its length is 0.
if answer is not None:
return answer
def query(self, qname, rdtype=dns.rdatatype.A, rdclass=dns.rdataclass.IN,
tcp=False, source=None, raise_on_no_answer=True, source_port=0,
lifetime=None): # pragma: no cover
"""Query nameservers to find the answer to the question.
This method calls resolve() with ``search=True``, and is
provided for backwards compatbility with prior versions of
dnspython. See the documentation for the resolve() method for
further details.
"""
warnings.warn('please use dns.resolver.Resolver.resolve() instead',
DeprecationWarning, stacklevel=2)
return self.resolve(qname, rdtype, rdclass, tcp, source,
raise_on_no_answer, source_port, lifetime,
True)
def resolve_address(self, ipaddr, *args, **kwargs):
"""Use a resolver to run a reverse query for PTR records.
This utilizes the resolve() method to perform a PTR lookup on the
specified IP address.
*ipaddr*, a ``str``, the IPv4 or IPv6 address you want to get
the PTR record for.
All other arguments that can be passed to the resolve() function
except for rdtype and rdclass are also supported by this
function.
"""
return self.resolve(dns.reversename.from_address(ipaddr),
rdtype=dns.rdatatype.PTR,
rdclass=dns.rdataclass.IN,
*args, **kwargs)
# pylint: disable=redefined-outer-name
def canonical_name(self, name):
"""Determine the canonical name of *name*.
The canonical name is the name the resolver uses for queries
after all CNAME and DNAME renamings have been applied.
*name*, a ``dns.name.Name`` or ``str``, the query name.
This method can raise any exception that ``resolve()`` can
raise, other than ``dns.resolver.NoAnswer`` and
``dns.resolver.NXDOMAIN``.
Returns a ``dns.name.Name``.
"""
try:
answer = self.resolve(name, raise_on_no_answer=False)
canonical_name = answer.canonical_name
except dns.resolver.NXDOMAIN as e:
canonical_name = e.canonical_name
return canonical_name
# pylint: enable=redefined-outer-name
#: The default resolver.
2018-10-31 16:08:29 +00:00
default_resolver = None
def get_default_resolver():
"""Get the default resolver, initializing it if necessary."""
if default_resolver is None:
reset_default_resolver()
return default_resolver
def reset_default_resolver():
"""Re-initialize default resolver.
Note that the resolver configuration (i.e. /etc/resolv.conf on UNIX
systems) will be re-read immediately.
2018-10-31 16:08:29 +00:00
"""
2018-10-31 16:08:29 +00:00
global default_resolver
default_resolver = Resolver()
def resolve(qname, rdtype=dns.rdatatype.A, rdclass=dns.rdataclass.IN,
tcp=False, source=None, raise_on_no_answer=True,
source_port=0, lifetime=None, search=None):
2018-10-31 16:08:29 +00:00
"""Query nameservers to find the answer to the question.
This is a convenience function that uses the default resolver
object to make the query.
See ``dns.resolver.Resolver.resolve`` for more information on the
parameters.
"""
return get_default_resolver().resolve(qname, rdtype, rdclass, tcp, source,
raise_on_no_answer, source_port,
lifetime, search)
def query(qname, rdtype=dns.rdatatype.A, rdclass=dns.rdataclass.IN,
tcp=False, source=None, raise_on_no_answer=True,
source_port=0, lifetime=None): # pragma: no cover
"""Query nameservers to find the answer to the question.
This method calls resolve() with ``search=True``, and is
provided for backwards compatbility with prior versions of
dnspython. See the documentation for the resolve() method for
further details.
"""
warnings.warn('please use dns.resolver.resolve() instead',
DeprecationWarning, stacklevel=2)
return resolve(qname, rdtype, rdclass, tcp, source,
raise_on_no_answer, source_port, lifetime,
True)
def resolve_address(ipaddr, *args, **kwargs):
"""Use a resolver to run a reverse query for PTR records.
See ``dns.resolver.Resolver.resolve_address`` for more information on the
parameters.
"""
return get_default_resolver().resolve_address(ipaddr, *args, **kwargs)
def canonical_name(name):
"""Determine the canonical name of *name*.
See ``dns.resolver.Resolver.canonical_name`` for more information on the
parameters and possible exceptions.
"""
2018-10-31 16:08:29 +00:00
return get_default_resolver().canonical_name(name)
def zone_for_name(name, rdclass=dns.rdataclass.IN, tcp=False, resolver=None,
lifetime=None):
2018-10-31 16:08:29 +00:00
"""Find the name of the zone which contains the specified name.
*name*, an absolute ``dns.name.Name`` or ``str``, the query name.
*rdclass*, an ``int``, the query class.
*tcp*, a ``bool``. If ``True``, use TCP to make the query.
*resolver*, a ``dns.resolver.Resolver`` or ``None``, the resolver to use.
If ``None``, the default, then the default resolver is used.
*lifetime*, a ``float``, the total time to allow for the queries needed
to determine the zone. If ``None``, the default, then only the individual
query limits of the resolver apply.
Raises ``dns.resolver.NoRootSOA`` if there is no SOA RR at the DNS
root. (This is only likely to happen if you're using non-default
root servers in your network and they are misconfigured.)
Raises ``dns.resolver.LifetimeTimeout`` if the answer could not be
found in the alotted lifetime.
Returns a ``dns.name.Name``.
"""
if isinstance(name, str):
2018-10-31 16:08:29 +00:00
name = dns.name.from_text(name, dns.name.root)
if resolver is None:
resolver = get_default_resolver()
if not name.is_absolute():
raise NotAbsolute(name)
start = time.time()
if lifetime is not None:
expiration = start + lifetime
else:
expiration = None
2018-10-31 16:08:29 +00:00
while 1:
try:
if expiration:
rlifetime = expiration - time.time()
if rlifetime <= 0:
rlifetime = 0
else:
rlifetime = None
answer = resolver.resolve(name, dns.rdatatype.SOA, rdclass, tcp,
lifetime=rlifetime)
2018-10-31 16:08:29 +00:00
if answer.rrset.name == name:
return name
# otherwise we were CNAMEd or DNAMEd and need to look higher
except (dns.resolver.NXDOMAIN, dns.resolver.NoAnswer) as e:
if isinstance(e, dns.resolver.NXDOMAIN):
response = e.responses().get(name)
else:
response = e.response() # pylint: disable=no-value-for-parameter
if response:
for rrs in response.authority:
if rrs.rdtype == dns.rdatatype.SOA and \
rrs.rdclass == rdclass:
(nr, _, _) = rrs.name.fullcompare(name)
if nr == dns.name.NAMERELN_SUPERDOMAIN:
# We're doing a proper superdomain check as
# if the name were equal we ought to have gotten
# it in the answer section! We are ignoring the
# possibility that the authority is insane and
# is including multiple SOA RRs for different
# authorities.
return rrs.name
# we couldn't extract anything useful from the response (e.g. it's
# a type 3 NXDOMAIN)
2018-10-31 16:08:29 +00:00
try:
name = name.parent()
except dns.name.NoParent:
raise NoRootSOA
#
# Support for overriding the system resolver for all python code in the
# running process.
#
_protocols_for_socktype = {
socket.SOCK_DGRAM: [socket.SOL_UDP],
socket.SOCK_STREAM: [socket.SOL_TCP],
}
_resolver = None
_original_getaddrinfo = socket.getaddrinfo
_original_getnameinfo = socket.getnameinfo
_original_getfqdn = socket.getfqdn
_original_gethostbyname = socket.gethostbyname
_original_gethostbyname_ex = socket.gethostbyname_ex
_original_gethostbyaddr = socket.gethostbyaddr
def _getaddrinfo(host=None, service=None, family=socket.AF_UNSPEC, socktype=0,
proto=0, flags=0):
if flags & socket.AI_NUMERICHOST != 0:
# Short circuit directly into the system's getaddrinfo(). We're
# not adding any value in this case, and this avoids infinite loops
# because dns.query.* needs to call getaddrinfo() for IPv6 scoping
# reasons. We will also do this short circuit below if we
# discover that the host is an address literal.
return _original_getaddrinfo(host, service, family, socktype, proto,
flags)
2018-10-31 16:08:29 +00:00
if flags & (socket.AI_ADDRCONFIG | socket.AI_V4MAPPED) != 0:
# Not implemented. We raise a gaierror as opposed to a
# NotImplementedError as it helps callers handle errors more
# appropriately. [Issue #316]
#
# We raise EAI_FAIL as opposed to EAI_SYSTEM because there is
# no EAI_SYSTEM on Windows [Issue #416]. We didn't go for
# EAI_BADFLAGS as the flags aren't bad, we just don't
# implement them.
raise socket.gaierror(socket.EAI_FAIL,
'Non-recoverable failure in name resolution')
2018-10-31 16:08:29 +00:00
if host is None and service is None:
raise socket.gaierror(socket.EAI_NONAME, 'Name or service not known')
2018-10-31 16:08:29 +00:00
v6addrs = []
v4addrs = []
canonical_name = None # pylint: disable=redefined-outer-name
# Is host None or an address literal? If so, use the system's
# getaddrinfo().
if host is None:
return _original_getaddrinfo(host, service, family, socktype,
proto, flags)
2018-10-31 16:08:29 +00:00
try:
# We don't care about the result of af_for_address(), we're just
# calling it so it raises an exception if host is not an IPv4 or
# IPv6 address.
dns.inet.af_for_address(host)
return _original_getaddrinfo(host, service, family, socktype,
proto, flags)
2018-10-31 16:08:29 +00:00
except Exception:
pass
# Something needs resolution!
try:
if family == socket.AF_INET6 or family == socket.AF_UNSPEC:
v6 = _resolver.resolve(host, dns.rdatatype.AAAA,
raise_on_no_answer=False)
# Note that setting host ensures we query the same name
# for A as we did for AAAA.
host = v6.qname
canonical_name = v6.canonical_name.to_text(True)
if v6.rrset is not None:
for rdata in v6.rrset:
v6addrs.append(rdata.address)
if family == socket.AF_INET or family == socket.AF_UNSPEC:
v4 = _resolver.resolve(host, dns.rdatatype.A,
raise_on_no_answer=False)
host = v4.qname
canonical_name = v4.canonical_name.to_text(True)
if v4.rrset is not None:
for rdata in v4.rrset:
v4addrs.append(rdata.address)
except dns.resolver.NXDOMAIN:
raise socket.gaierror(socket.EAI_NONAME, 'Name or service not known')
except Exception:
# We raise EAI_AGAIN here as the failure may be temporary
# (e.g. a timeout) and EAI_SYSTEM isn't defined on Windows.
# [Issue #416]
raise socket.gaierror(socket.EAI_AGAIN,
'Temporary failure in name resolution')
2018-10-31 16:08:29 +00:00
port = None
try:
# Is it a port literal?
if service is None:
port = 0
else:
port = int(service)
except Exception:
if flags & socket.AI_NUMERICSERV == 0:
try:
port = socket.getservbyname(service)
except Exception:
pass
if port is None:
raise socket.gaierror(socket.EAI_NONAME, 'Name or service not known')
2018-10-31 16:08:29 +00:00
tuples = []
if socktype == 0:
socktypes = [socket.SOCK_DGRAM, socket.SOCK_STREAM]
else:
socktypes = [socktype]
if flags & socket.AI_CANONNAME != 0:
cname = canonical_name
else:
cname = ''
if family == socket.AF_INET6 or family == socket.AF_UNSPEC:
for addr in v6addrs:
for socktype in socktypes:
for proto in _protocols_for_socktype[socktype]:
tuples.append((socket.AF_INET6, socktype, proto,
cname, (addr, port, 0, 0)))
if family == socket.AF_INET or family == socket.AF_UNSPEC:
for addr in v4addrs:
for socktype in socktypes:
for proto in _protocols_for_socktype[socktype]:
tuples.append((socket.AF_INET, socktype, proto,
cname, (addr, port)))
if len(tuples) == 0:
raise socket.gaierror(socket.EAI_NONAME, 'Name or service not known')
2018-10-31 16:08:29 +00:00
return tuples
def _getnameinfo(sockaddr, flags=0):
host = sockaddr[0]
port = sockaddr[1]
if len(sockaddr) == 4:
scope = sockaddr[3]
family = socket.AF_INET6
else:
scope = None
family = socket.AF_INET
tuples = _getaddrinfo(host, port, family, socket.SOCK_STREAM,
socket.SOL_TCP, 0)
if len(tuples) > 1:
raise socket.error('sockaddr resolved to multiple addresses')
addr = tuples[0][4][0]
if flags & socket.NI_DGRAM:
pname = 'udp'
else:
pname = 'tcp'
qname = dns.reversename.from_address(addr)
if flags & socket.NI_NUMERICHOST == 0:
try:
answer = _resolver.resolve(qname, 'PTR')
2018-10-31 16:08:29 +00:00
hostname = answer.rrset[0].target.to_text(True)
except (dns.resolver.NXDOMAIN, dns.resolver.NoAnswer):
if flags & socket.NI_NAMEREQD:
raise socket.gaierror(socket.EAI_NONAME,
'Name or service not known')
2018-10-31 16:08:29 +00:00
hostname = addr
if scope is not None:
hostname += '%' + str(scope)
else:
hostname = addr
if scope is not None:
hostname += '%' + str(scope)
if flags & socket.NI_NUMERICSERV:
service = str(port)
else:
service = socket.getservbyport(port, pname)
return (hostname, service)
def _getfqdn(name=None):
if name is None:
name = socket.gethostname()
try:
(name, _, _) = _gethostbyaddr(name)
# Python's version checks aliases too, but our gethostbyname
# ignores them, so we do so here as well.
2018-10-31 16:08:29 +00:00
except Exception:
pass
return name
2018-10-31 16:08:29 +00:00
def _gethostbyname(name):
return _gethostbyname_ex(name)[2][0]
def _gethostbyname_ex(name):
aliases = []
addresses = []
tuples = _getaddrinfo(name, 0, socket.AF_INET, socket.SOCK_STREAM,
socket.SOL_TCP, socket.AI_CANONNAME)
canonical = tuples[0][3]
for item in tuples:
addresses.append(item[4][0])
# XXX we just ignore aliases
return (canonical, aliases, addresses)
def _gethostbyaddr(ip):
try:
dns.ipv6.inet_aton(ip)
sockaddr = (ip, 80, 0, 0)
family = socket.AF_INET6
except Exception:
try:
dns.ipv4.inet_aton(ip)
except Exception:
raise socket.gaierror(socket.EAI_NONAME,
'Name or service not known')
2018-10-31 16:08:29 +00:00
sockaddr = (ip, 80)
family = socket.AF_INET
(name, _) = _getnameinfo(sockaddr, socket.NI_NAMEREQD)
2018-10-31 16:08:29 +00:00
aliases = []
addresses = []
tuples = _getaddrinfo(name, 0, family, socket.SOCK_STREAM, socket.SOL_TCP,
socket.AI_CANONNAME)
canonical = tuples[0][3]
# We only want to include an address from the tuples if it's the
# same as the one we asked about. We do this comparison in binary
# to avoid any differences in text representations.
bin_ip = dns.inet.inet_pton(family, ip)
2018-10-31 16:08:29 +00:00
for item in tuples:
addr = item[4][0]
bin_addr = dns.inet.inet_pton(family, addr)
if bin_ip == bin_addr:
addresses.append(addr)
2018-10-31 16:08:29 +00:00
# XXX we just ignore aliases
return (canonical, aliases, addresses)
def override_system_resolver(resolver=None):
"""Override the system resolver routines in the socket module with
versions which use dnspython's resolver.
This can be useful in testing situations where you want to control
the resolution behavior of python code without having to change
the system's resolver settings (e.g. /etc/resolv.conf).
The resolver to use may be specified; if it's not, the default
resolver will be used.
resolver, a ``dns.resolver.Resolver`` or ``None``, the resolver to use.
2018-10-31 16:08:29 +00:00
"""
2018-10-31 16:08:29 +00:00
if resolver is None:
resolver = get_default_resolver()
global _resolver
_resolver = resolver
socket.getaddrinfo = _getaddrinfo
socket.getnameinfo = _getnameinfo
socket.getfqdn = _getfqdn
socket.gethostbyname = _gethostbyname
socket.gethostbyname_ex = _gethostbyname_ex
socket.gethostbyaddr = _gethostbyaddr
def restore_system_resolver():
"""Undo the effects of prior override_system_resolver()."""
2018-10-31 16:08:29 +00:00
global _resolver
_resolver = None
socket.getaddrinfo = _original_getaddrinfo
socket.getnameinfo = _original_getnameinfo
socket.getfqdn = _original_getfqdn
socket.gethostbyname = _original_gethostbyname
socket.gethostbyname_ex = _original_gethostbyname_ex
socket.gethostbyaddr = _original_gethostbyaddr