This commit is contained in:
Mai Gillmann
2019-12-17 14:09:10 +01:00
parent 66e908fc8a
commit 4791d00a43
2122 changed files with 423791 additions and 0 deletions
@@ -0,0 +1,448 @@
from __future__ import absolute_import
import datetime
import logging
import os
import socket
from socket import error as SocketError, timeout as SocketTimeout
import warnings
from .packages import six
from .packages.six.moves.http_client import HTTPConnection as _HTTPConnection
from .packages.six.moves.http_client import HTTPException # noqa: F401
try: # Compiled with SSL?
import ssl
BaseSSLError = ssl.SSLError
except (ImportError, AttributeError): # Platform-specific: No SSL.
ssl = None
class BaseSSLError(BaseException):
pass
try:
# Python 3: not a no-op, we're adding this to the namespace so it can be imported.
ConnectionError = ConnectionError
except NameError:
# Python 2
class ConnectionError(Exception):
pass
from .exceptions import (
NewConnectionError,
ConnectTimeoutError,
SubjectAltNameWarning,
SystemTimeWarning,
)
from .packages.ssl_match_hostname import match_hostname, CertificateError
from .util.ssl_ import (
resolve_cert_reqs,
resolve_ssl_version,
assert_fingerprint,
create_urllib3_context,
ssl_wrap_socket,
)
from .util import connection
from ._collections import HTTPHeaderDict
log = logging.getLogger(__name__)
port_by_scheme = {"http": 80, "https": 443}
# When it comes time to update this value as a part of regular maintenance
# (ie test_recent_date is failing) update it to ~6 months before the current date.
RECENT_DATE = datetime.date(2019, 1, 1)
class DummyConnection(object):
"""Used to detect a failed ConnectionCls import."""
pass
class HTTPConnection(_HTTPConnection, object):
"""
Based on httplib.HTTPConnection but provides an extra constructor
backwards-compatibility layer between older and newer Pythons.
Additional keyword parameters are used to configure attributes of the connection.
Accepted parameters include:
- ``strict``: See the documentation on :class:`urllib3.connectionpool.HTTPConnectionPool`
- ``source_address``: Set the source address for the current connection.
- ``socket_options``: Set specific options on the underlying socket. If not specified, then
defaults are loaded from ``HTTPConnection.default_socket_options`` which includes disabling
Nagle's algorithm (sets TCP_NODELAY to 1) unless the connection is behind a proxy.
For example, if you wish to enable TCP Keep Alive in addition to the defaults,
you might pass::
HTTPConnection.default_socket_options + [
(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1),
]
Or you may want to disable the defaults by passing an empty list (e.g., ``[]``).
"""
default_port = port_by_scheme["http"]
#: Disable Nagle's algorithm by default.
#: ``[(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)]``
default_socket_options = [(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)]
#: Whether this connection verifies the host's certificate.
is_verified = False
def __init__(self, *args, **kw):
if not six.PY2:
kw.pop("strict", None)
# Pre-set source_address.
self.source_address = kw.get("source_address")
#: The socket options provided by the user. If no options are
#: provided, we use the default options.
self.socket_options = kw.pop("socket_options", self.default_socket_options)
_HTTPConnection.__init__(self, *args, **kw)
@property
def host(self):
"""
Getter method to remove any trailing dots that indicate the hostname is an FQDN.
In general, SSL certificates don't include the trailing dot indicating a
fully-qualified domain name, and thus, they don't validate properly when
checked against a domain name that includes the dot. In addition, some
servers may not expect to receive the trailing dot when provided.
However, the hostname with trailing dot is critical to DNS resolution; doing a
lookup with the trailing dot will properly only resolve the appropriate FQDN,
whereas a lookup without a trailing dot will search the system's search domain
list. Thus, it's important to keep the original host around for use only in
those cases where it's appropriate (i.e., when doing DNS lookup to establish the
actual TCP connection across which we're going to send HTTP requests).
"""
return self._dns_host.rstrip(".")
@host.setter
def host(self, value):
"""
Setter for the `host` property.
We assume that only urllib3 uses the _dns_host attribute; httplib itself
only uses `host`, and it seems reasonable that other libraries follow suit.
"""
self._dns_host = value
def _new_conn(self):
""" Establish a socket connection and set nodelay settings on it.
:return: New socket connection.
"""
extra_kw = {}
if self.source_address:
extra_kw["source_address"] = self.source_address
if self.socket_options:
extra_kw["socket_options"] = self.socket_options
try:
conn = connection.create_connection(
(self._dns_host, self.port), self.timeout, **extra_kw
)
except SocketTimeout:
raise ConnectTimeoutError(
self,
"Connection to %s timed out. (connect timeout=%s)"
% (self.host, self.timeout),
)
except SocketError as e:
raise NewConnectionError(
self, "Failed to establish a new connection: %s" % e
)
return conn
def _prepare_conn(self, conn):
self.sock = conn
# Google App Engine's httplib does not define _tunnel_host
if getattr(self, "_tunnel_host", None):
# TODO: Fix tunnel so it doesn't depend on self.sock state.
self._tunnel()
# Mark this connection as not reusable
self.auto_open = 0
def connect(self):
conn = self._new_conn()
self._prepare_conn(conn)
def request_chunked(self, method, url, body=None, headers=None):
"""
Alternative to the common request method, which sends the
body with chunked encoding and not as one block
"""
headers = HTTPHeaderDict(headers if headers is not None else {})
skip_accept_encoding = "accept-encoding" in headers
skip_host = "host" in headers
self.putrequest(
method, url, skip_accept_encoding=skip_accept_encoding, skip_host=skip_host
)
for header, value in headers.items():
self.putheader(header, value)
if "transfer-encoding" not in headers:
self.putheader("Transfer-Encoding", "chunked")
self.endheaders()
if body is not None:
stringish_types = six.string_types + (bytes,)
if isinstance(body, stringish_types):
body = (body,)
for chunk in body:
if not chunk:
continue
if not isinstance(chunk, bytes):
chunk = chunk.encode("utf8")
len_str = hex(len(chunk))[2:]
self.send(len_str.encode("utf-8"))
self.send(b"\r\n")
self.send(chunk)
self.send(b"\r\n")
# After the if clause, to always have a closed body
self.send(b"0\r\n\r\n")
class HTTPSConnection(HTTPConnection):
default_port = port_by_scheme["https"]
ssl_version = None
def __init__(
self,
host,
port=None,
key_file=None,
cert_file=None,
key_password=None,
strict=None,
timeout=socket._GLOBAL_DEFAULT_TIMEOUT,
ssl_context=None,
server_hostname=None,
**kw
):
HTTPConnection.__init__(self, host, port, strict=strict, timeout=timeout, **kw)
self.key_file = key_file
self.cert_file = cert_file
self.key_password = key_password
self.ssl_context = ssl_context
self.server_hostname = server_hostname
# Required property for Google AppEngine 1.9.0 which otherwise causes
# HTTPS requests to go out as HTTP. (See Issue #356)
self._protocol = "https"
def connect(self):
conn = self._new_conn()
self._prepare_conn(conn)
# Wrap socket using verification with the root certs in
# trusted_root_certs
default_ssl_context = False
if self.ssl_context is None:
default_ssl_context = True
self.ssl_context = create_urllib3_context(
ssl_version=resolve_ssl_version(self.ssl_version),
cert_reqs=resolve_cert_reqs(self.cert_reqs),
)
# Try to load OS default certs if none are given.
# Works well on Windows (requires Python3.4+)
context = self.ssl_context
if (
not self.ca_certs
and not self.ca_cert_dir
and default_ssl_context
and hasattr(context, "load_default_certs")
):
context.load_default_certs()
self.sock = ssl_wrap_socket(
sock=conn,
keyfile=self.key_file,
certfile=self.cert_file,
key_password=self.key_password,
ssl_context=self.ssl_context,
server_hostname=self.server_hostname,
)
class VerifiedHTTPSConnection(HTTPSConnection):
"""
Based on httplib.HTTPSConnection but wraps the socket with
SSL certification.
"""
cert_reqs = None
ca_certs = None
ca_cert_dir = None
ssl_version = None
assert_fingerprint = None
def set_cert(
self,
key_file=None,
cert_file=None,
cert_reqs=None,
key_password=None,
ca_certs=None,
assert_hostname=None,
assert_fingerprint=None,
ca_cert_dir=None,
):
"""
This method should only be called once, before the connection is used.
"""
# If cert_reqs is not provided we'll assume CERT_REQUIRED unless we also
# have an SSLContext object in which case we'll use its verify_mode.
if cert_reqs is None:
if self.ssl_context is not None:
cert_reqs = self.ssl_context.verify_mode
else:
cert_reqs = resolve_cert_reqs(None)
self.key_file = key_file
self.cert_file = cert_file
self.cert_reqs = cert_reqs
self.key_password = key_password
self.assert_hostname = assert_hostname
self.assert_fingerprint = assert_fingerprint
self.ca_certs = ca_certs and os.path.expanduser(ca_certs)
self.ca_cert_dir = ca_cert_dir and os.path.expanduser(ca_cert_dir)
def connect(self):
# Add certificate verification
conn = self._new_conn()
hostname = self.host
# Google App Engine's httplib does not define _tunnel_host
if getattr(self, "_tunnel_host", None):
self.sock = conn
# Calls self._set_hostport(), so self.host is
# self._tunnel_host below.
self._tunnel()
# Mark this connection as not reusable
self.auto_open = 0
# Override the host with the one we're requesting data from.
hostname = self._tunnel_host
server_hostname = hostname
if self.server_hostname is not None:
server_hostname = self.server_hostname
is_time_off = datetime.date.today() < RECENT_DATE
if is_time_off:
warnings.warn(
(
"System time is way off (before {0}). This will probably "
"lead to SSL verification errors"
).format(RECENT_DATE),
SystemTimeWarning,
)
# Wrap socket using verification with the root certs in
# trusted_root_certs
default_ssl_context = False
if self.ssl_context is None:
default_ssl_context = True
self.ssl_context = create_urllib3_context(
ssl_version=resolve_ssl_version(self.ssl_version),
cert_reqs=resolve_cert_reqs(self.cert_reqs),
)
context = self.ssl_context
context.verify_mode = resolve_cert_reqs(self.cert_reqs)
# Try to load OS default certs if none are given.
# Works well on Windows (requires Python3.4+)
if (
not self.ca_certs
and not self.ca_cert_dir
and default_ssl_context
and hasattr(context, "load_default_certs")
):
context.load_default_certs()
self.sock = ssl_wrap_socket(
sock=conn,
keyfile=self.key_file,
certfile=self.cert_file,
key_password=self.key_password,
ca_certs=self.ca_certs,
ca_cert_dir=self.ca_cert_dir,
server_hostname=server_hostname,
ssl_context=context,
)
if self.assert_fingerprint:
assert_fingerprint(
self.sock.getpeercert(binary_form=True), self.assert_fingerprint
)
elif (
context.verify_mode != ssl.CERT_NONE
and not getattr(context, "check_hostname", False)
and self.assert_hostname is not False
):
# While urllib3 attempts to always turn off hostname matching from
# the TLS library, this cannot always be done. So we check whether
# the TLS Library still thinks it's matching hostnames.
cert = self.sock.getpeercert()
if not cert.get("subjectAltName", ()):
warnings.warn(
(
"Certificate for {0} has no `subjectAltName`, falling back to check for a "
"`commonName` for now. This feature is being removed by major browsers and "
"deprecated by RFC 2818. (See https://github.com/urllib3/urllib3/issues/497 "
"for details.)".format(hostname)
),
SubjectAltNameWarning,
)
_match_hostname(cert, self.assert_hostname or server_hostname)
self.is_verified = (
context.verify_mode == ssl.CERT_REQUIRED
or self.assert_fingerprint is not None
)
def _match_hostname(cert, asserted_hostname):
try:
match_hostname(cert, asserted_hostname)
except CertificateError as e:
log.warning(
"Certificate did not match expected hostname: %s. Certificate: %s",
asserted_hostname,
cert,
)
# Add cert to exception and reraise so client code can inspect
# the cert when catching the exception, if they want to
e._peer_cert = cert
raise
if ssl:
# Make a copy for testing.
UnverifiedHTTPSConnection = HTTPSConnection
HTTPSConnection = VerifiedHTTPSConnection
else:
HTTPSConnection = DummyConnection
@@ -0,0 +1,493 @@
"""
This module uses ctypes to bind a whole bunch of functions and constants from
SecureTransport. The goal here is to provide the low-level API to
SecureTransport. These are essentially the C-level functions and constants, and
they're pretty gross to work with.
This code is a bastardised version of the code found in Will Bond's oscrypto
library. An enormous debt is owed to him for blazing this trail for us. For
that reason, this code should be considered to be covered both by urllib3's
license and by oscrypto's:
Copyright (c) 2015-2016 Will Bond <will@wbond.net>
Permission is hereby granted, free of charge, to any person obtaining a
copy of this software and associated documentation files (the "Software"),
to deal in the Software without restriction, including without limitation
the rights to use, copy, modify, merge, publish, distribute, sublicense,
and/or sell copies of the Software, and to permit persons to whom the
Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
DEALINGS IN THE SOFTWARE.
"""
from __future__ import absolute_import
import platform
from ctypes.util import find_library
from ctypes import (
c_void_p,
c_int32,
c_char_p,
c_size_t,
c_byte,
c_uint32,
c_ulong,
c_long,
c_bool,
)
from ctypes import CDLL, POINTER, CFUNCTYPE
security_path = find_library("Security")
if not security_path:
raise ImportError("The library Security could not be found")
core_foundation_path = find_library("CoreFoundation")
if not core_foundation_path:
raise ImportError("The library CoreFoundation could not be found")
version = platform.mac_ver()[0]
version_info = tuple(map(int, version.split(".")))
if version_info < (10, 8):
raise OSError(
"Only OS X 10.8 and newer are supported, not %s.%s"
% (version_info[0], version_info[1])
)
Security = CDLL(security_path, use_errno=True)
CoreFoundation = CDLL(core_foundation_path, use_errno=True)
Boolean = c_bool
CFIndex = c_long
CFStringEncoding = c_uint32
CFData = c_void_p
CFString = c_void_p
CFArray = c_void_p
CFMutableArray = c_void_p
CFDictionary = c_void_p
CFError = c_void_p
CFType = c_void_p
CFTypeID = c_ulong
CFTypeRef = POINTER(CFType)
CFAllocatorRef = c_void_p
OSStatus = c_int32
CFDataRef = POINTER(CFData)
CFStringRef = POINTER(CFString)
CFArrayRef = POINTER(CFArray)
CFMutableArrayRef = POINTER(CFMutableArray)
CFDictionaryRef = POINTER(CFDictionary)
CFArrayCallBacks = c_void_p
CFDictionaryKeyCallBacks = c_void_p
CFDictionaryValueCallBacks = c_void_p
SecCertificateRef = POINTER(c_void_p)
SecExternalFormat = c_uint32
SecExternalItemType = c_uint32
SecIdentityRef = POINTER(c_void_p)
SecItemImportExportFlags = c_uint32
SecItemImportExportKeyParameters = c_void_p
SecKeychainRef = POINTER(c_void_p)
SSLProtocol = c_uint32
SSLCipherSuite = c_uint32
SSLContextRef = POINTER(c_void_p)
SecTrustRef = POINTER(c_void_p)
SSLConnectionRef = c_uint32
SecTrustResultType = c_uint32
SecTrustOptionFlags = c_uint32
SSLProtocolSide = c_uint32
SSLConnectionType = c_uint32
SSLSessionOption = c_uint32
try:
Security.SecItemImport.argtypes = [
CFDataRef,
CFStringRef,
POINTER(SecExternalFormat),
POINTER(SecExternalItemType),
SecItemImportExportFlags,
POINTER(SecItemImportExportKeyParameters),
SecKeychainRef,
POINTER(CFArrayRef),
]
Security.SecItemImport.restype = OSStatus
Security.SecCertificateGetTypeID.argtypes = []
Security.SecCertificateGetTypeID.restype = CFTypeID
Security.SecIdentityGetTypeID.argtypes = []
Security.SecIdentityGetTypeID.restype = CFTypeID
Security.SecKeyGetTypeID.argtypes = []
Security.SecKeyGetTypeID.restype = CFTypeID
Security.SecCertificateCreateWithData.argtypes = [CFAllocatorRef, CFDataRef]
Security.SecCertificateCreateWithData.restype = SecCertificateRef
Security.SecCertificateCopyData.argtypes = [SecCertificateRef]
Security.SecCertificateCopyData.restype = CFDataRef
Security.SecCopyErrorMessageString.argtypes = [OSStatus, c_void_p]
Security.SecCopyErrorMessageString.restype = CFStringRef
Security.SecIdentityCreateWithCertificate.argtypes = [
CFTypeRef,
SecCertificateRef,
POINTER(SecIdentityRef),
]
Security.SecIdentityCreateWithCertificate.restype = OSStatus
Security.SecKeychainCreate.argtypes = [
c_char_p,
c_uint32,
c_void_p,
Boolean,
c_void_p,
POINTER(SecKeychainRef),
]
Security.SecKeychainCreate.restype = OSStatus
Security.SecKeychainDelete.argtypes = [SecKeychainRef]
Security.SecKeychainDelete.restype = OSStatus
Security.SecPKCS12Import.argtypes = [
CFDataRef,
CFDictionaryRef,
POINTER(CFArrayRef),
]
Security.SecPKCS12Import.restype = OSStatus
SSLReadFunc = CFUNCTYPE(OSStatus, SSLConnectionRef, c_void_p, POINTER(c_size_t))
SSLWriteFunc = CFUNCTYPE(
OSStatus, SSLConnectionRef, POINTER(c_byte), POINTER(c_size_t)
)
Security.SSLSetIOFuncs.argtypes = [SSLContextRef, SSLReadFunc, SSLWriteFunc]
Security.SSLSetIOFuncs.restype = OSStatus
Security.SSLSetPeerID.argtypes = [SSLContextRef, c_char_p, c_size_t]
Security.SSLSetPeerID.restype = OSStatus
Security.SSLSetCertificate.argtypes = [SSLContextRef, CFArrayRef]
Security.SSLSetCertificate.restype = OSStatus
Security.SSLSetCertificateAuthorities.argtypes = [SSLContextRef, CFTypeRef, Boolean]
Security.SSLSetCertificateAuthorities.restype = OSStatus
Security.SSLSetConnection.argtypes = [SSLContextRef, SSLConnectionRef]
Security.SSLSetConnection.restype = OSStatus
Security.SSLSetPeerDomainName.argtypes = [SSLContextRef, c_char_p, c_size_t]
Security.SSLSetPeerDomainName.restype = OSStatus
Security.SSLHandshake.argtypes = [SSLContextRef]
Security.SSLHandshake.restype = OSStatus
Security.SSLRead.argtypes = [SSLContextRef, c_char_p, c_size_t, POINTER(c_size_t)]
Security.SSLRead.restype = OSStatus
Security.SSLWrite.argtypes = [SSLContextRef, c_char_p, c_size_t, POINTER(c_size_t)]
Security.SSLWrite.restype = OSStatus
Security.SSLClose.argtypes = [SSLContextRef]
Security.SSLClose.restype = OSStatus
Security.SSLGetNumberSupportedCiphers.argtypes = [SSLContextRef, POINTER(c_size_t)]
Security.SSLGetNumberSupportedCiphers.restype = OSStatus
Security.SSLGetSupportedCiphers.argtypes = [
SSLContextRef,
POINTER(SSLCipherSuite),
POINTER(c_size_t),
]
Security.SSLGetSupportedCiphers.restype = OSStatus
Security.SSLSetEnabledCiphers.argtypes = [
SSLContextRef,
POINTER(SSLCipherSuite),
c_size_t,
]
Security.SSLSetEnabledCiphers.restype = OSStatus
Security.SSLGetNumberEnabledCiphers.argtype = [SSLContextRef, POINTER(c_size_t)]
Security.SSLGetNumberEnabledCiphers.restype = OSStatus
Security.SSLGetEnabledCiphers.argtypes = [
SSLContextRef,
POINTER(SSLCipherSuite),
POINTER(c_size_t),
]
Security.SSLGetEnabledCiphers.restype = OSStatus
Security.SSLGetNegotiatedCipher.argtypes = [SSLContextRef, POINTER(SSLCipherSuite)]
Security.SSLGetNegotiatedCipher.restype = OSStatus
Security.SSLGetNegotiatedProtocolVersion.argtypes = [
SSLContextRef,
POINTER(SSLProtocol),
]
Security.SSLGetNegotiatedProtocolVersion.restype = OSStatus
Security.SSLCopyPeerTrust.argtypes = [SSLContextRef, POINTER(SecTrustRef)]
Security.SSLCopyPeerTrust.restype = OSStatus
Security.SecTrustSetAnchorCertificates.argtypes = [SecTrustRef, CFArrayRef]
Security.SecTrustSetAnchorCertificates.restype = OSStatus
Security.SecTrustSetAnchorCertificatesOnly.argstypes = [SecTrustRef, Boolean]
Security.SecTrustSetAnchorCertificatesOnly.restype = OSStatus
Security.SecTrustEvaluate.argtypes = [SecTrustRef, POINTER(SecTrustResultType)]
Security.SecTrustEvaluate.restype = OSStatus
Security.SecTrustGetCertificateCount.argtypes = [SecTrustRef]
Security.SecTrustGetCertificateCount.restype = CFIndex
Security.SecTrustGetCertificateAtIndex.argtypes = [SecTrustRef, CFIndex]
Security.SecTrustGetCertificateAtIndex.restype = SecCertificateRef
Security.SSLCreateContext.argtypes = [
CFAllocatorRef,
SSLProtocolSide,
SSLConnectionType,
]
Security.SSLCreateContext.restype = SSLContextRef
Security.SSLSetSessionOption.argtypes = [SSLContextRef, SSLSessionOption, Boolean]
Security.SSLSetSessionOption.restype = OSStatus
Security.SSLSetProtocolVersionMin.argtypes = [SSLContextRef, SSLProtocol]
Security.SSLSetProtocolVersionMin.restype = OSStatus
Security.SSLSetProtocolVersionMax.argtypes = [SSLContextRef, SSLProtocol]
Security.SSLSetProtocolVersionMax.restype = OSStatus
Security.SecCopyErrorMessageString.argtypes = [OSStatus, c_void_p]
Security.SecCopyErrorMessageString.restype = CFStringRef
Security.SSLReadFunc = SSLReadFunc
Security.SSLWriteFunc = SSLWriteFunc
Security.SSLContextRef = SSLContextRef
Security.SSLProtocol = SSLProtocol
Security.SSLCipherSuite = SSLCipherSuite
Security.SecIdentityRef = SecIdentityRef
Security.SecKeychainRef = SecKeychainRef
Security.SecTrustRef = SecTrustRef
Security.SecTrustResultType = SecTrustResultType
Security.SecExternalFormat = SecExternalFormat
Security.OSStatus = OSStatus
Security.kSecImportExportPassphrase = CFStringRef.in_dll(
Security, "kSecImportExportPassphrase"
)
Security.kSecImportItemIdentity = CFStringRef.in_dll(
Security, "kSecImportItemIdentity"
)
# CoreFoundation time!
CoreFoundation.CFRetain.argtypes = [CFTypeRef]
CoreFoundation.CFRetain.restype = CFTypeRef
CoreFoundation.CFRelease.argtypes = [CFTypeRef]
CoreFoundation.CFRelease.restype = None
CoreFoundation.CFGetTypeID.argtypes = [CFTypeRef]
CoreFoundation.CFGetTypeID.restype = CFTypeID
CoreFoundation.CFStringCreateWithCString.argtypes = [
CFAllocatorRef,
c_char_p,
CFStringEncoding,
]
CoreFoundation.CFStringCreateWithCString.restype = CFStringRef
CoreFoundation.CFStringGetCStringPtr.argtypes = [CFStringRef, CFStringEncoding]
CoreFoundation.CFStringGetCStringPtr.restype = c_char_p
CoreFoundation.CFStringGetCString.argtypes = [
CFStringRef,
c_char_p,
CFIndex,
CFStringEncoding,
]
CoreFoundation.CFStringGetCString.restype = c_bool
CoreFoundation.CFDataCreate.argtypes = [CFAllocatorRef, c_char_p, CFIndex]
CoreFoundation.CFDataCreate.restype = CFDataRef
CoreFoundation.CFDataGetLength.argtypes = [CFDataRef]
CoreFoundation.CFDataGetLength.restype = CFIndex
CoreFoundation.CFDataGetBytePtr.argtypes = [CFDataRef]
CoreFoundation.CFDataGetBytePtr.restype = c_void_p
CoreFoundation.CFDictionaryCreate.argtypes = [
CFAllocatorRef,
POINTER(CFTypeRef),
POINTER(CFTypeRef),
CFIndex,
CFDictionaryKeyCallBacks,
CFDictionaryValueCallBacks,
]
CoreFoundation.CFDictionaryCreate.restype = CFDictionaryRef
CoreFoundation.CFDictionaryGetValue.argtypes = [CFDictionaryRef, CFTypeRef]
CoreFoundation.CFDictionaryGetValue.restype = CFTypeRef
CoreFoundation.CFArrayCreate.argtypes = [
CFAllocatorRef,
POINTER(CFTypeRef),
CFIndex,
CFArrayCallBacks,
]
CoreFoundation.CFArrayCreate.restype = CFArrayRef
CoreFoundation.CFArrayCreateMutable.argtypes = [
CFAllocatorRef,
CFIndex,
CFArrayCallBacks,
]
CoreFoundation.CFArrayCreateMutable.restype = CFMutableArrayRef
CoreFoundation.CFArrayAppendValue.argtypes = [CFMutableArrayRef, c_void_p]
CoreFoundation.CFArrayAppendValue.restype = None
CoreFoundation.CFArrayGetCount.argtypes = [CFArrayRef]
CoreFoundation.CFArrayGetCount.restype = CFIndex
CoreFoundation.CFArrayGetValueAtIndex.argtypes = [CFArrayRef, CFIndex]
CoreFoundation.CFArrayGetValueAtIndex.restype = c_void_p
CoreFoundation.kCFAllocatorDefault = CFAllocatorRef.in_dll(
CoreFoundation, "kCFAllocatorDefault"
)
CoreFoundation.kCFTypeArrayCallBacks = c_void_p.in_dll(
CoreFoundation, "kCFTypeArrayCallBacks"
)
CoreFoundation.kCFTypeDictionaryKeyCallBacks = c_void_p.in_dll(
CoreFoundation, "kCFTypeDictionaryKeyCallBacks"
)
CoreFoundation.kCFTypeDictionaryValueCallBacks = c_void_p.in_dll(
CoreFoundation, "kCFTypeDictionaryValueCallBacks"
)
CoreFoundation.CFTypeRef = CFTypeRef
CoreFoundation.CFArrayRef = CFArrayRef
CoreFoundation.CFStringRef = CFStringRef
CoreFoundation.CFDictionaryRef = CFDictionaryRef
except (AttributeError):
raise ImportError("Error initializing ctypes")
class CFConst(object):
"""
A class object that acts as essentially a namespace for CoreFoundation
constants.
"""
kCFStringEncodingUTF8 = CFStringEncoding(0x08000100)
class SecurityConst(object):
"""
A class object that acts as essentially a namespace for Security constants.
"""
kSSLSessionOptionBreakOnServerAuth = 0
kSSLProtocol2 = 1
kSSLProtocol3 = 2
kTLSProtocol1 = 4
kTLSProtocol11 = 7
kTLSProtocol12 = 8
# SecureTransport does not support TLS 1.3 even if there's a constant for it
kTLSProtocol13 = 10
kTLSProtocolMaxSupported = 999
kSSLClientSide = 1
kSSLStreamType = 0
kSecFormatPEMSequence = 10
kSecTrustResultInvalid = 0
kSecTrustResultProceed = 1
# This gap is present on purpose: this was kSecTrustResultConfirm, which
# is deprecated.
kSecTrustResultDeny = 3
kSecTrustResultUnspecified = 4
kSecTrustResultRecoverableTrustFailure = 5
kSecTrustResultFatalTrustFailure = 6
kSecTrustResultOtherError = 7
errSSLProtocol = -9800
errSSLWouldBlock = -9803
errSSLClosedGraceful = -9805
errSSLClosedNoNotify = -9816
errSSLClosedAbort = -9806
errSSLXCertChainInvalid = -9807
errSSLCrypto = -9809
errSSLInternal = -9810
errSSLCertExpired = -9814
errSSLCertNotYetValid = -9815
errSSLUnknownRootCert = -9812
errSSLNoRootCert = -9813
errSSLHostNameMismatch = -9843
errSSLPeerHandshakeFail = -9824
errSSLPeerUserCancelled = -9839
errSSLWeakPeerEphemeralDHKey = -9850
errSSLServerAuthCompleted = -9841
errSSLRecordOverflow = -9847
errSecVerifyFailed = -67808
errSecNoTrustSettings = -25263
errSecItemNotFound = -25300
errSecInvalidTrustSettings = -25262
# Cipher suites. We only pick the ones our default cipher string allows.
# Source: https://developer.apple.com/documentation/security/1550981-ssl_cipher_suite_values
TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384 = 0xC02C
TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 = 0xC030
TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256 = 0xC02B
TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256 = 0xC02F
TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256 = 0xCCA9
TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256 = 0xCCA8
TLS_DHE_RSA_WITH_AES_256_GCM_SHA384 = 0x009F
TLS_DHE_RSA_WITH_AES_128_GCM_SHA256 = 0x009E
TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA384 = 0xC024
TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA384 = 0xC028
TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA = 0xC00A
TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA = 0xC014
TLS_DHE_RSA_WITH_AES_256_CBC_SHA256 = 0x006B
TLS_DHE_RSA_WITH_AES_256_CBC_SHA = 0x0039
TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256 = 0xC023
TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256 = 0xC027
TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA = 0xC009
TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA = 0xC013
TLS_DHE_RSA_WITH_AES_128_CBC_SHA256 = 0x0067
TLS_DHE_RSA_WITH_AES_128_CBC_SHA = 0x0033
TLS_RSA_WITH_AES_256_GCM_SHA384 = 0x009D
TLS_RSA_WITH_AES_128_GCM_SHA256 = 0x009C
TLS_RSA_WITH_AES_256_CBC_SHA256 = 0x003D
TLS_RSA_WITH_AES_128_CBC_SHA256 = 0x003C
TLS_RSA_WITH_AES_256_CBC_SHA = 0x0035
TLS_RSA_WITH_AES_128_CBC_SHA = 0x002F
TLS_AES_128_GCM_SHA256 = 0x1301
TLS_AES_256_GCM_SHA384 = 0x1302
TLS_AES_128_CCM_8_SHA256 = 0x1305
TLS_AES_128_CCM_SHA256 = 0x1304
@@ -0,0 +1,121 @@
"""
NTLM authenticating pool, contributed by erikcederstran
Issue #10, see: http://code.google.com/p/urllib3/issues/detail?id=10
"""
from __future__ import absolute_import
from logging import getLogger
from ntlm import ntlm
from .. import HTTPSConnectionPool
from ..packages.six.moves.http_client import HTTPSConnection
log = getLogger(__name__)
class NTLMConnectionPool(HTTPSConnectionPool):
"""
Implements an NTLM authentication version of an urllib3 connection pool
"""
scheme = "https"
def __init__(self, user, pw, authurl, *args, **kwargs):
"""
authurl is a random URL on the server that is protected by NTLM.
user is the Windows user, probably in the DOMAIN\\username format.
pw is the password for the user.
"""
super(NTLMConnectionPool, self).__init__(*args, **kwargs)
self.authurl = authurl
self.rawuser = user
user_parts = user.split("\\", 1)
self.domain = user_parts[0].upper()
self.user = user_parts[1]
self.pw = pw
def _new_conn(self):
# Performs the NTLM handshake that secures the connection. The socket
# must be kept open while requests are performed.
self.num_connections += 1
log.debug(
"Starting NTLM HTTPS connection no. %d: https://%s%s",
self.num_connections,
self.host,
self.authurl,
)
headers = {"Connection": "Keep-Alive"}
req_header = "Authorization"
resp_header = "www-authenticate"
conn = HTTPSConnection(host=self.host, port=self.port)
# Send negotiation message
headers[req_header] = "NTLM %s" % ntlm.create_NTLM_NEGOTIATE_MESSAGE(
self.rawuser
)
log.debug("Request headers: %s", headers)
conn.request("GET", self.authurl, None, headers)
res = conn.getresponse()
reshdr = dict(res.getheaders())
log.debug("Response status: %s %s", res.status, res.reason)
log.debug("Response headers: %s", reshdr)
log.debug("Response data: %s [...]", res.read(100))
# Remove the reference to the socket, so that it can not be closed by
# the response object (we want to keep the socket open)
res.fp = None
# Server should respond with a challenge message
auth_header_values = reshdr[resp_header].split(", ")
auth_header_value = None
for s in auth_header_values:
if s[:5] == "NTLM ":
auth_header_value = s[5:]
if auth_header_value is None:
raise Exception(
"Unexpected %s response header: %s" % (resp_header, reshdr[resp_header])
)
# Send authentication message
ServerChallenge, NegotiateFlags = ntlm.parse_NTLM_CHALLENGE_MESSAGE(
auth_header_value
)
auth_msg = ntlm.create_NTLM_AUTHENTICATE_MESSAGE(
ServerChallenge, self.user, self.domain, self.pw, NegotiateFlags
)
headers[req_header] = "NTLM %s" % auth_msg
log.debug("Request headers: %s", headers)
conn.request("GET", self.authurl, None, headers)
res = conn.getresponse()
log.debug("Response status: %s %s", res.status, res.reason)
log.debug("Response headers: %s", dict(res.getheaders()))
log.debug("Response data: %s [...]", res.read()[:100])
if res.status != 200:
if res.status == 401:
raise Exception("Server rejected request: wrong username or password")
raise Exception("Wrong server response: %s %s" % (res.status, res.reason))
res.fp = None
log.debug("Connection established")
return conn
def urlopen(
self,
method,
url,
body=None,
headers=None,
retries=3,
redirect=True,
assert_same_host=True,
):
if headers is None:
headers = {}
headers["Connection"] = "Keep-Alive"
return super(NTLMConnectionPool, self).urlopen(
method, url, body, headers, retries, redirect, assert_same_host
)
@@ -0,0 +1,859 @@
"""
SecureTranport support for urllib3 via ctypes.
This makes platform-native TLS available to urllib3 users on macOS without the
use of a compiler. This is an important feature because the Python Package
Index is moving to become a TLSv1.2-or-higher server, and the default OpenSSL
that ships with macOS is not capable of doing TLSv1.2. The only way to resolve
this is to give macOS users an alternative solution to the problem, and that
solution is to use SecureTransport.
We use ctypes here because this solution must not require a compiler. That's
because pip is not allowed to require a compiler either.
This is not intended to be a seriously long-term solution to this problem.
The hope is that PEP 543 will eventually solve this issue for us, at which
point we can retire this contrib module. But in the short term, we need to
solve the impending tire fire that is Python on Mac without this kind of
contrib module. So...here we are.
To use this module, simply import and inject it::
import urllib3.contrib.securetransport
urllib3.contrib.securetransport.inject_into_urllib3()
Happy TLSing!
This code is a bastardised version of the code found in Will Bond's oscrypto
library. An enormous debt is owed to him for blazing this trail for us. For
that reason, this code should be considered to be covered both by urllib3's
license and by oscrypto's:
Copyright (c) 2015-2016 Will Bond <will@wbond.net>
Permission is hereby granted, free of charge, to any person obtaining a
copy of this software and associated documentation files (the "Software"),
to deal in the Software without restriction, including without limitation
the rights to use, copy, modify, merge, publish, distribute, sublicense,
and/or sell copies of the Software, and to permit persons to whom the
Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
DEALINGS IN THE SOFTWARE.
"""
from __future__ import absolute_import
import contextlib
import ctypes
import errno
import os.path
import shutil
import socket
import ssl
import threading
import weakref
from .. import util
from ._securetransport.bindings import Security, SecurityConst, CoreFoundation
from ._securetransport.low_level import (
_assert_no_error,
_cert_array_from_pem,
_temporary_keychain,
_load_client_cert_chain,
)
try: # Platform-specific: Python 2
from socket import _fileobject
except ImportError: # Platform-specific: Python 3
_fileobject = None
from ..packages.backports.makefile import backport_makefile
__all__ = ["inject_into_urllib3", "extract_from_urllib3"]
# SNI always works
HAS_SNI = True
orig_util_HAS_SNI = util.HAS_SNI
orig_util_SSLContext = util.ssl_.SSLContext
# This dictionary is used by the read callback to obtain a handle to the
# calling wrapped socket. This is a pretty silly approach, but for now it'll
# do. I feel like I should be able to smuggle a handle to the wrapped socket
# directly in the SSLConnectionRef, but for now this approach will work I
# guess.
#
# We need to lock around this structure for inserts, but we don't do it for
# reads/writes in the callbacks. The reasoning here goes as follows:
#
# 1. It is not possible to call into the callbacks before the dictionary is
# populated, so once in the callback the id must be in the dictionary.
# 2. The callbacks don't mutate the dictionary, they only read from it, and
# so cannot conflict with any of the insertions.
#
# This is good: if we had to lock in the callbacks we'd drastically slow down
# the performance of this code.
_connection_refs = weakref.WeakValueDictionary()
_connection_ref_lock = threading.Lock()
# Limit writes to 16kB. This is OpenSSL's limit, but we'll cargo-cult it over
# for no better reason than we need *a* limit, and this one is right there.
SSL_WRITE_BLOCKSIZE = 16384
# This is our equivalent of util.ssl_.DEFAULT_CIPHERS, but expanded out to
# individual cipher suites. We need to do this because this is how
# SecureTransport wants them.
CIPHER_SUITES = [
SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384,
SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256,
SecurityConst.TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384,
SecurityConst.TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256,
SecurityConst.TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256,
SecurityConst.TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256,
SecurityConst.TLS_DHE_RSA_WITH_AES_256_GCM_SHA384,
SecurityConst.TLS_DHE_RSA_WITH_AES_128_GCM_SHA256,
SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA384,
SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA,
SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256,
SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA,
SecurityConst.TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA384,
SecurityConst.TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA,
SecurityConst.TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256,
SecurityConst.TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA,
SecurityConst.TLS_DHE_RSA_WITH_AES_256_CBC_SHA256,
SecurityConst.TLS_DHE_RSA_WITH_AES_256_CBC_SHA,
SecurityConst.TLS_DHE_RSA_WITH_AES_128_CBC_SHA256,
SecurityConst.TLS_DHE_RSA_WITH_AES_128_CBC_SHA,
SecurityConst.TLS_AES_256_GCM_SHA384,
SecurityConst.TLS_AES_128_GCM_SHA256,
SecurityConst.TLS_RSA_WITH_AES_256_GCM_SHA384,
SecurityConst.TLS_RSA_WITH_AES_128_GCM_SHA256,
SecurityConst.TLS_AES_128_CCM_8_SHA256,
SecurityConst.TLS_AES_128_CCM_SHA256,
SecurityConst.TLS_RSA_WITH_AES_256_CBC_SHA256,
SecurityConst.TLS_RSA_WITH_AES_128_CBC_SHA256,
SecurityConst.TLS_RSA_WITH_AES_256_CBC_SHA,
SecurityConst.TLS_RSA_WITH_AES_128_CBC_SHA,
]
# Basically this is simple: for PROTOCOL_SSLv23 we turn it into a low of
# TLSv1 and a high of TLSv1.2. For everything else, we pin to that version.
# TLSv1 to 1.2 are supported on macOS 10.8+
_protocol_to_min_max = {
util.PROTOCOL_TLS: (SecurityConst.kTLSProtocol1, SecurityConst.kTLSProtocol12)
}
if hasattr(ssl, "PROTOCOL_SSLv2"):
_protocol_to_min_max[ssl.PROTOCOL_SSLv2] = (
SecurityConst.kSSLProtocol2,
SecurityConst.kSSLProtocol2,
)
if hasattr(ssl, "PROTOCOL_SSLv3"):
_protocol_to_min_max[ssl.PROTOCOL_SSLv3] = (
SecurityConst.kSSLProtocol3,
SecurityConst.kSSLProtocol3,
)
if hasattr(ssl, "PROTOCOL_TLSv1"):
_protocol_to_min_max[ssl.PROTOCOL_TLSv1] = (
SecurityConst.kTLSProtocol1,
SecurityConst.kTLSProtocol1,
)
if hasattr(ssl, "PROTOCOL_TLSv1_1"):
_protocol_to_min_max[ssl.PROTOCOL_TLSv1_1] = (
SecurityConst.kTLSProtocol11,
SecurityConst.kTLSProtocol11,
)
if hasattr(ssl, "PROTOCOL_TLSv1_2"):
_protocol_to_min_max[ssl.PROTOCOL_TLSv1_2] = (
SecurityConst.kTLSProtocol12,
SecurityConst.kTLSProtocol12,
)
def inject_into_urllib3():
"""
Monkey-patch urllib3 with SecureTransport-backed SSL-support.
"""
util.SSLContext = SecureTransportContext
util.ssl_.SSLContext = SecureTransportContext
util.HAS_SNI = HAS_SNI
util.ssl_.HAS_SNI = HAS_SNI
util.IS_SECURETRANSPORT = True
util.ssl_.IS_SECURETRANSPORT = True
def extract_from_urllib3():
"""
Undo monkey-patching by :func:`inject_into_urllib3`.
"""
util.SSLContext = orig_util_SSLContext
util.ssl_.SSLContext = orig_util_SSLContext
util.HAS_SNI = orig_util_HAS_SNI
util.ssl_.HAS_SNI = orig_util_HAS_SNI
util.IS_SECURETRANSPORT = False
util.ssl_.IS_SECURETRANSPORT = False
def _read_callback(connection_id, data_buffer, data_length_pointer):
"""
SecureTransport read callback. This is called by ST to request that data
be returned from the socket.
"""
wrapped_socket = None
try:
wrapped_socket = _connection_refs.get(connection_id)
if wrapped_socket is None:
return SecurityConst.errSSLInternal
base_socket = wrapped_socket.socket
requested_length = data_length_pointer[0]
timeout = wrapped_socket.gettimeout()
error = None
read_count = 0
try:
while read_count < requested_length:
if timeout is None or timeout >= 0:
if not util.wait_for_read(base_socket, timeout):
raise socket.error(errno.EAGAIN, "timed out")
remaining = requested_length - read_count
buffer = (ctypes.c_char * remaining).from_address(
data_buffer + read_count
)
chunk_size = base_socket.recv_into(buffer, remaining)
read_count += chunk_size
if not chunk_size:
if not read_count:
return SecurityConst.errSSLClosedGraceful
break
except (socket.error) as e:
error = e.errno
if error is not None and error != errno.EAGAIN:
data_length_pointer[0] = read_count
if error == errno.ECONNRESET or error == errno.EPIPE:
return SecurityConst.errSSLClosedAbort
raise
data_length_pointer[0] = read_count
if read_count != requested_length:
return SecurityConst.errSSLWouldBlock
return 0
except Exception as e:
if wrapped_socket is not None:
wrapped_socket._exception = e
return SecurityConst.errSSLInternal
def _write_callback(connection_id, data_buffer, data_length_pointer):
"""
SecureTransport write callback. This is called by ST to request that data
actually be sent on the network.
"""
wrapped_socket = None
try:
wrapped_socket = _connection_refs.get(connection_id)
if wrapped_socket is None:
return SecurityConst.errSSLInternal
base_socket = wrapped_socket.socket
bytes_to_write = data_length_pointer[0]
data = ctypes.string_at(data_buffer, bytes_to_write)
timeout = wrapped_socket.gettimeout()
error = None
sent = 0
try:
while sent < bytes_to_write:
if timeout is None or timeout >= 0:
if not util.wait_for_write(base_socket, timeout):
raise socket.error(errno.EAGAIN, "timed out")
chunk_sent = base_socket.send(data)
sent += chunk_sent
# This has some needless copying here, but I'm not sure there's
# much value in optimising this data path.
data = data[chunk_sent:]
except (socket.error) as e:
error = e.errno
if error is not None and error != errno.EAGAIN:
data_length_pointer[0] = sent
if error == errno.ECONNRESET or error == errno.EPIPE:
return SecurityConst.errSSLClosedAbort
raise
data_length_pointer[0] = sent
if sent != bytes_to_write:
return SecurityConst.errSSLWouldBlock
return 0
except Exception as e:
if wrapped_socket is not None:
wrapped_socket._exception = e
return SecurityConst.errSSLInternal
# We need to keep these two objects references alive: if they get GC'd while
# in use then SecureTransport could attempt to call a function that is in freed
# memory. That would be...uh...bad. Yeah, that's the word. Bad.
_read_callback_pointer = Security.SSLReadFunc(_read_callback)
_write_callback_pointer = Security.SSLWriteFunc(_write_callback)
class WrappedSocket(object):
"""
API-compatibility wrapper for Python's OpenSSL wrapped socket object.
Note: _makefile_refs, _drop(), and _reuse() are needed for the garbage
collector of PyPy.
"""
def __init__(self, socket):
self.socket = socket
self.context = None
self._makefile_refs = 0
self._closed = False
self._exception = None
self._keychain = None
self._keychain_dir = None
self._client_cert_chain = None
# We save off the previously-configured timeout and then set it to
# zero. This is done because we use select and friends to handle the
# timeouts, but if we leave the timeout set on the lower socket then
# Python will "kindly" call select on that socket again for us. Avoid
# that by forcing the timeout to zero.
self._timeout = self.socket.gettimeout()
self.socket.settimeout(0)
@contextlib.contextmanager
def _raise_on_error(self):
"""
A context manager that can be used to wrap calls that do I/O from
SecureTransport. If any of the I/O callbacks hit an exception, this
context manager will correctly propagate the exception after the fact.
This avoids silently swallowing those exceptions.
It also correctly forces the socket closed.
"""
self._exception = None
# We explicitly don't catch around this yield because in the unlikely
# event that an exception was hit in the block we don't want to swallow
# it.
yield
if self._exception is not None:
exception, self._exception = self._exception, None
self.close()
raise exception
def _set_ciphers(self):
"""
Sets up the allowed ciphers. By default this matches the set in
util.ssl_.DEFAULT_CIPHERS, at least as supported by macOS. This is done
custom and doesn't allow changing at this time, mostly because parsing
OpenSSL cipher strings is going to be a freaking nightmare.
"""
ciphers = (Security.SSLCipherSuite * len(CIPHER_SUITES))(*CIPHER_SUITES)
result = Security.SSLSetEnabledCiphers(
self.context, ciphers, len(CIPHER_SUITES)
)
_assert_no_error(result)
def _custom_validate(self, verify, trust_bundle):
"""
Called when we have set custom validation. We do this in two cases:
first, when cert validation is entirely disabled; and second, when
using a custom trust DB.
"""
# If we disabled cert validation, just say: cool.
if not verify:
return
# We want data in memory, so load it up.
if os.path.isfile(trust_bundle):
with open(trust_bundle, "rb") as f:
trust_bundle = f.read()
cert_array = None
trust = Security.SecTrustRef()
try:
# Get a CFArray that contains the certs we want.
cert_array = _cert_array_from_pem(trust_bundle)
# Ok, now the hard part. We want to get the SecTrustRef that ST has
# created for this connection, shove our CAs into it, tell ST to
# ignore everything else it knows, and then ask if it can build a
# chain. This is a buuuunch of code.
result = Security.SSLCopyPeerTrust(self.context, ctypes.byref(trust))
_assert_no_error(result)
if not trust:
raise ssl.SSLError("Failed to copy trust reference")
result = Security.SecTrustSetAnchorCertificates(trust, cert_array)
_assert_no_error(result)
result = Security.SecTrustSetAnchorCertificatesOnly(trust, True)
_assert_no_error(result)
trust_result = Security.SecTrustResultType()
result = Security.SecTrustEvaluate(trust, ctypes.byref(trust_result))
_assert_no_error(result)
finally:
if trust:
CoreFoundation.CFRelease(trust)
if cert_array is not None:
CoreFoundation.CFRelease(cert_array)
# Ok, now we can look at what the result was.
successes = (
SecurityConst.kSecTrustResultUnspecified,
SecurityConst.kSecTrustResultProceed,
)
if trust_result.value not in successes:
raise ssl.SSLError(
"certificate verify failed, error code: %d" % trust_result.value
)
def handshake(
self,
server_hostname,
verify,
trust_bundle,
min_version,
max_version,
client_cert,
client_key,
client_key_passphrase,
):
"""
Actually performs the TLS handshake. This is run automatically by
wrapped socket, and shouldn't be needed in user code.
"""
# First, we do the initial bits of connection setup. We need to create
# a context, set its I/O funcs, and set the connection reference.
self.context = Security.SSLCreateContext(
None, SecurityConst.kSSLClientSide, SecurityConst.kSSLStreamType
)
result = Security.SSLSetIOFuncs(
self.context, _read_callback_pointer, _write_callback_pointer
)
_assert_no_error(result)
# Here we need to compute the handle to use. We do this by taking the
# id of self modulo 2**31 - 1. If this is already in the dictionary, we
# just keep incrementing by one until we find a free space.
with _connection_ref_lock:
handle = id(self) % 2147483647
while handle in _connection_refs:
handle = (handle + 1) % 2147483647
_connection_refs[handle] = self
result = Security.SSLSetConnection(self.context, handle)
_assert_no_error(result)
# If we have a server hostname, we should set that too.
if server_hostname:
if not isinstance(server_hostname, bytes):
server_hostname = server_hostname.encode("utf-8")
result = Security.SSLSetPeerDomainName(
self.context, server_hostname, len(server_hostname)
)
_assert_no_error(result)
# Setup the ciphers.
self._set_ciphers()
# Set the minimum and maximum TLS versions.
result = Security.SSLSetProtocolVersionMin(self.context, min_version)
_assert_no_error(result)
result = Security.SSLSetProtocolVersionMax(self.context, max_version)
_assert_no_error(result)
# If there's a trust DB, we need to use it. We do that by telling
# SecureTransport to break on server auth. We also do that if we don't
# want to validate the certs at all: we just won't actually do any
# authing in that case.
if not verify or trust_bundle is not None:
result = Security.SSLSetSessionOption(
self.context, SecurityConst.kSSLSessionOptionBreakOnServerAuth, True
)
_assert_no_error(result)
# If there's a client cert, we need to use it.
if client_cert:
self._keychain, self._keychain_dir = _temporary_keychain()
self._client_cert_chain = _load_client_cert_chain(
self._keychain, client_cert, client_key
)
result = Security.SSLSetCertificate(self.context, self._client_cert_chain)
_assert_no_error(result)
while True:
with self._raise_on_error():
result = Security.SSLHandshake(self.context)
if result == SecurityConst.errSSLWouldBlock:
raise socket.timeout("handshake timed out")
elif result == SecurityConst.errSSLServerAuthCompleted:
self._custom_validate(verify, trust_bundle)
continue
else:
_assert_no_error(result)
break
def fileno(self):
return self.socket.fileno()
# Copy-pasted from Python 3.5 source code
def _decref_socketios(self):
if self._makefile_refs > 0:
self._makefile_refs -= 1
if self._closed:
self.close()
def recv(self, bufsiz):
buffer = ctypes.create_string_buffer(bufsiz)
bytes_read = self.recv_into(buffer, bufsiz)
data = buffer[:bytes_read]
return data
def recv_into(self, buffer, nbytes=None):
# Read short on EOF.
if self._closed:
return 0
if nbytes is None:
nbytes = len(buffer)
buffer = (ctypes.c_char * nbytes).from_buffer(buffer)
processed_bytes = ctypes.c_size_t(0)
with self._raise_on_error():
result = Security.SSLRead(
self.context, buffer, nbytes, ctypes.byref(processed_bytes)
)
# There are some result codes that we want to treat as "not always
# errors". Specifically, those are errSSLWouldBlock,
# errSSLClosedGraceful, and errSSLClosedNoNotify.
if result == SecurityConst.errSSLWouldBlock:
# If we didn't process any bytes, then this was just a time out.
# However, we can get errSSLWouldBlock in situations when we *did*
# read some data, and in those cases we should just read "short"
# and return.
if processed_bytes.value == 0:
# Timed out, no data read.
raise socket.timeout("recv timed out")
elif result in (
SecurityConst.errSSLClosedGraceful,
SecurityConst.errSSLClosedNoNotify,
):
# The remote peer has closed this connection. We should do so as
# well. Note that we don't actually return here because in
# principle this could actually be fired along with return data.
# It's unlikely though.
self.close()
else:
_assert_no_error(result)
# Ok, we read and probably succeeded. We should return whatever data
# was actually read.
return processed_bytes.value
def settimeout(self, timeout):
self._timeout = timeout
def gettimeout(self):
return self._timeout
def send(self, data):
processed_bytes = ctypes.c_size_t(0)
with self._raise_on_error():
result = Security.SSLWrite(
self.context, data, len(data), ctypes.byref(processed_bytes)
)
if result == SecurityConst.errSSLWouldBlock and processed_bytes.value == 0:
# Timed out
raise socket.timeout("send timed out")
else:
_assert_no_error(result)
# We sent, and probably succeeded. Tell them how much we sent.
return processed_bytes.value
def sendall(self, data):
total_sent = 0
while total_sent < len(data):
sent = self.send(data[total_sent : total_sent + SSL_WRITE_BLOCKSIZE])
total_sent += sent
def shutdown(self):
with self._raise_on_error():
Security.SSLClose(self.context)
def close(self):
# TODO: should I do clean shutdown here? Do I have to?
if self._makefile_refs < 1:
self._closed = True
if self.context:
CoreFoundation.CFRelease(self.context)
self.context = None
if self._client_cert_chain:
CoreFoundation.CFRelease(self._client_cert_chain)
self._client_cert_chain = None
if self._keychain:
Security.SecKeychainDelete(self._keychain)
CoreFoundation.CFRelease(self._keychain)
shutil.rmtree(self._keychain_dir)
self._keychain = self._keychain_dir = None
return self.socket.close()
else:
self._makefile_refs -= 1
def getpeercert(self, binary_form=False):
# Urgh, annoying.
#
# Here's how we do this:
#
# 1. Call SSLCopyPeerTrust to get hold of the trust object for this
# connection.
# 2. Call SecTrustGetCertificateAtIndex for index 0 to get the leaf.
# 3. To get the CN, call SecCertificateCopyCommonName and process that
# string so that it's of the appropriate type.
# 4. To get the SAN, we need to do something a bit more complex:
# a. Call SecCertificateCopyValues to get the data, requesting
# kSecOIDSubjectAltName.
# b. Mess about with this dictionary to try to get the SANs out.
#
# This is gross. Really gross. It's going to be a few hundred LoC extra
# just to repeat something that SecureTransport can *already do*. So my
# operating assumption at this time is that what we want to do is
# instead to just flag to urllib3 that it shouldn't do its own hostname
# validation when using SecureTransport.
if not binary_form:
raise ValueError("SecureTransport only supports dumping binary certs")
trust = Security.SecTrustRef()
certdata = None
der_bytes = None
try:
# Grab the trust store.
result = Security.SSLCopyPeerTrust(self.context, ctypes.byref(trust))
_assert_no_error(result)
if not trust:
# Probably we haven't done the handshake yet. No biggie.
return None
cert_count = Security.SecTrustGetCertificateCount(trust)
if not cert_count:
# Also a case that might happen if we haven't handshaked.
# Handshook? Handshaken?
return None
leaf = Security.SecTrustGetCertificateAtIndex(trust, 0)
assert leaf
# Ok, now we want the DER bytes.
certdata = Security.SecCertificateCopyData(leaf)
assert certdata
data_length = CoreFoundation.CFDataGetLength(certdata)
data_buffer = CoreFoundation.CFDataGetBytePtr(certdata)
der_bytes = ctypes.string_at(data_buffer, data_length)
finally:
if certdata:
CoreFoundation.CFRelease(certdata)
if trust:
CoreFoundation.CFRelease(trust)
return der_bytes
def version(self):
protocol = Security.SSLProtocol()
result = Security.SSLGetNegotiatedProtocolVersion(
self.context, ctypes.byref(protocol)
)
_assert_no_error(result)
if protocol.value == SecurityConst.kTLSProtocol13:
raise ssl.SSLError("SecureTransport does not support TLS 1.3")
elif protocol.value == SecurityConst.kTLSProtocol12:
return "TLSv1.2"
elif protocol.value == SecurityConst.kTLSProtocol11:
return "TLSv1.1"
elif protocol.value == SecurityConst.kTLSProtocol1:
return "TLSv1"
elif protocol.value == SecurityConst.kSSLProtocol3:
return "SSLv3"
elif protocol.value == SecurityConst.kSSLProtocol2:
return "SSLv2"
else:
raise ssl.SSLError("Unknown TLS version: %r" % protocol)
def _reuse(self):
self._makefile_refs += 1
def _drop(self):
if self._makefile_refs < 1:
self.close()
else:
self._makefile_refs -= 1
if _fileobject: # Platform-specific: Python 2
def makefile(self, mode, bufsize=-1):
self._makefile_refs += 1
return _fileobject(self, mode, bufsize, close=True)
else: # Platform-specific: Python 3
def makefile(self, mode="r", buffering=None, *args, **kwargs):
# We disable buffering with SecureTransport because it conflicts with
# the buffering that ST does internally (see issue #1153 for more).
buffering = 0
return backport_makefile(self, mode, buffering, *args, **kwargs)
WrappedSocket.makefile = makefile
class SecureTransportContext(object):
"""
I am a wrapper class for the SecureTransport library, to translate the
interface of the standard library ``SSLContext`` object to calls into
SecureTransport.
"""
def __init__(self, protocol):
self._min_version, self._max_version = _protocol_to_min_max[protocol]
self._options = 0
self._verify = False
self._trust_bundle = None
self._client_cert = None
self._client_key = None
self._client_key_passphrase = None
@property
def check_hostname(self):
"""
SecureTransport cannot have its hostname checking disabled. For more,
see the comment on getpeercert() in this file.
"""
return True
@check_hostname.setter
def check_hostname(self, value):
"""
SecureTransport cannot have its hostname checking disabled. For more,
see the comment on getpeercert() in this file.
"""
pass
@property
def options(self):
# TODO: Well, crap.
#
# So this is the bit of the code that is the most likely to cause us
# trouble. Essentially we need to enumerate all of the SSL options that
# users might want to use and try to see if we can sensibly translate
# them, or whether we should just ignore them.
return self._options
@options.setter
def options(self, value):
# TODO: Update in line with above.
self._options = value
@property
def verify_mode(self):
return ssl.CERT_REQUIRED if self._verify else ssl.CERT_NONE
@verify_mode.setter
def verify_mode(self, value):
self._verify = True if value == ssl.CERT_REQUIRED else False
def set_default_verify_paths(self):
# So, this has to do something a bit weird. Specifically, what it does
# is nothing.
#
# This means that, if we had previously had load_verify_locations
# called, this does not undo that. We need to do that because it turns
# out that the rest of the urllib3 code will attempt to load the
# default verify paths if it hasn't been told about any paths, even if
# the context itself was sometime earlier. We resolve that by just
# ignoring it.
pass
def load_default_certs(self):
return self.set_default_verify_paths()
def set_ciphers(self, ciphers):
# For now, we just require the default cipher string.
if ciphers != util.ssl_.DEFAULT_CIPHERS:
raise ValueError("SecureTransport doesn't support custom cipher strings")
def load_verify_locations(self, cafile=None, capath=None, cadata=None):
# OK, we only really support cadata and cafile.
if capath is not None:
raise ValueError("SecureTransport does not support cert directories")
self._trust_bundle = cafile or cadata
def load_cert_chain(self, certfile, keyfile=None, password=None):
self._client_cert = certfile
self._client_key = keyfile
self._client_cert_passphrase = password
def wrap_socket(
self,
sock,
server_side=False,
do_handshake_on_connect=True,
suppress_ragged_eofs=True,
server_hostname=None,
):
# So, what do we do here? Firstly, we assert some properties. This is a
# stripped down shim, so there is some functionality we don't support.
# See PEP 543 for the real deal.
assert not server_side
assert do_handshake_on_connect
assert suppress_ragged_eofs
# Ok, we're good to go. Now we want to create the wrapped socket object
# and store it in the appropriate place.
wrapped_socket = WrappedSocket(sock)
# Now we can handshake
wrapped_socket.handshake(
server_hostname,
self._verify,
self._trust_bundle,
self._min_version,
self._max_version,
self._client_cert,
self._client_key,
self._client_key_passphrase,
)
return wrapped_socket
@@ -0,0 +1,98 @@
from __future__ import absolute_import
import binascii
import codecs
import os
from io import BytesIO
from .packages import six
from .packages.six import b
from .fields import RequestField
writer = codecs.lookup("utf-8")[3]
def choose_boundary():
"""
Our embarrassingly-simple replacement for mimetools.choose_boundary.
"""
boundary = binascii.hexlify(os.urandom(16))
if not six.PY2:
boundary = boundary.decode("ascii")
return boundary
def iter_field_objects(fields):
"""
Iterate over fields.
Supports list of (k, v) tuples and dicts, and lists of
:class:`~urllib3.fields.RequestField`.
"""
if isinstance(fields, dict):
i = six.iteritems(fields)
else:
i = iter(fields)
for field in i:
if isinstance(field, RequestField):
yield field
else:
yield RequestField.from_tuples(*field)
def iter_fields(fields):
"""
.. deprecated:: 1.6
Iterate over fields.
The addition of :class:`~urllib3.fields.RequestField` makes this function
obsolete. Instead, use :func:`iter_field_objects`, which returns
:class:`~urllib3.fields.RequestField` objects.
Supports list of (k, v) tuples and dicts.
"""
if isinstance(fields, dict):
return ((k, v) for k, v in six.iteritems(fields))
return ((k, v) for k, v in fields)
def encode_multipart_formdata(fields, boundary=None):
"""
Encode a dictionary of ``fields`` using the multipart/form-data MIME format.
:param fields:
Dictionary of fields or list of (key, :class:`~urllib3.fields.RequestField`).
:param boundary:
If not specified, then a random boundary will be generated using
:func:`urllib3.filepost.choose_boundary`.
"""
body = BytesIO()
if boundary is None:
boundary = choose_boundary()
for field in iter_field_objects(fields):
body.write(b("--%s\r\n" % (boundary)))
writer(body).write(field.render_headers())
data = field.data
if isinstance(data, int):
data = str(data) # Backwards compatibility
if isinstance(data, six.text_type):
writer(body).write(data)
else:
body.write(data)
body.write(b"\r\n")
body.write(b("--%s--\r\n" % (boundary)))
content_type = str("multipart/form-data; boundary=%s" % boundary)
return body.getvalue(), content_type
@@ -0,0 +1,52 @@
# -*- coding: utf-8 -*-
"""
backports.makefile
~~~~~~~~~~~~~~~~~~
Backports the Python 3 ``socket.makefile`` method for use with anything that
wants to create a "fake" socket object.
"""
import io
from socket import SocketIO
def backport_makefile(
self, mode="r", buffering=None, encoding=None, errors=None, newline=None
):
"""
Backport of ``socket.makefile`` from Python 3.5.
"""
if not set(mode) <= {"r", "w", "b"}:
raise ValueError("invalid mode %r (only r, w, b allowed)" % (mode,))
writing = "w" in mode
reading = "r" in mode or not writing
assert reading or writing
binary = "b" in mode
rawmode = ""
if reading:
rawmode += "r"
if writing:
rawmode += "w"
raw = SocketIO(self, rawmode)
self._makefile_refs += 1
if buffering is None:
buffering = -1
if buffering < 0:
buffering = io.DEFAULT_BUFFER_SIZE
if buffering == 0:
if not binary:
raise ValueError("unbuffered streams must be binary")
return raw
if reading and writing:
buffer = io.BufferedRWPair(raw, raw, buffering)
elif reading:
buffer = io.BufferedReader(raw, buffering)
else:
assert writing
buffer = io.BufferedWriter(raw, buffering)
if binary:
return buffer
text = io.TextIOWrapper(buffer, encoding, errors, newline)
text.mode = mode
return text
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,171 @@
from __future__ import absolute_import
from .filepost import encode_multipart_formdata
from .packages.six.moves.urllib.parse import urlencode
__all__ = ["RequestMethods"]
class RequestMethods(object):
"""
Convenience mixin for classes who implement a :meth:`urlopen` method, such
as :class:`~urllib3.connectionpool.HTTPConnectionPool` and
:class:`~urllib3.poolmanager.PoolManager`.
Provides behavior for making common types of HTTP request methods and
decides which type of request field encoding to use.
Specifically,
:meth:`.request_encode_url` is for sending requests whose fields are
encoded in the URL (such as GET, HEAD, DELETE).
:meth:`.request_encode_body` is for sending requests whose fields are
encoded in the *body* of the request using multipart or www-form-urlencoded
(such as for POST, PUT, PATCH).
:meth:`.request` is for making any kind of request, it will look up the
appropriate encoding format and use one of the above two methods to make
the request.
Initializer parameters:
:param headers:
Headers to include with all requests, unless other headers are given
explicitly.
"""
_encode_url_methods = {"DELETE", "GET", "HEAD", "OPTIONS"}
def __init__(self, headers=None):
self.headers = headers or {}
def urlopen(
self,
method,
url,
body=None,
headers=None,
encode_multipart=True,
multipart_boundary=None,
**kw
): # Abstract
raise NotImplementedError(
"Classes extending RequestMethods must implement "
"their own ``urlopen`` method."
)
def request(self, method, url, fields=None, headers=None, **urlopen_kw):
"""
Make a request using :meth:`urlopen` with the appropriate encoding of
``fields`` based on the ``method`` used.
This is a convenience method that requires the least amount of manual
effort. It can be used in most situations, while still having the
option to drop down to more specific methods when necessary, such as
:meth:`request_encode_url`, :meth:`request_encode_body`,
or even the lowest level :meth:`urlopen`.
"""
method = method.upper()
urlopen_kw["request_url"] = url
if method in self._encode_url_methods:
return self.request_encode_url(
method, url, fields=fields, headers=headers, **urlopen_kw
)
else:
return self.request_encode_body(
method, url, fields=fields, headers=headers, **urlopen_kw
)
def request_encode_url(self, method, url, fields=None, headers=None, **urlopen_kw):
"""
Make a request using :meth:`urlopen` with the ``fields`` encoded in
the url. This is useful for request methods like GET, HEAD, DELETE, etc.
"""
if headers is None:
headers = self.headers
extra_kw = {"headers": headers}
extra_kw.update(urlopen_kw)
if fields:
url += "?" + urlencode(fields)
return self.urlopen(method, url, **extra_kw)
def request_encode_body(
self,
method,
url,
fields=None,
headers=None,
encode_multipart=True,
multipart_boundary=None,
**urlopen_kw
):
"""
Make a request using :meth:`urlopen` with the ``fields`` encoded in
the body. This is useful for request methods like POST, PUT, PATCH, etc.
When ``encode_multipart=True`` (default), then
:meth:`urllib3.filepost.encode_multipart_formdata` is used to encode
the payload with the appropriate content type. Otherwise
:meth:`urllib.urlencode` is used with the
'application/x-www-form-urlencoded' content type.
Multipart encoding must be used when posting files, and it's reasonably
safe to use it in other times too. However, it may break request
signing, such as with OAuth.
Supports an optional ``fields`` parameter of key/value strings AND
key/filetuple. A filetuple is a (filename, data, MIME type) tuple where
the MIME type is optional. For example::
fields = {
'foo': 'bar',
'fakefile': ('foofile.txt', 'contents of foofile'),
'realfile': ('barfile.txt', open('realfile').read()),
'typedfile': ('bazfile.bin', open('bazfile').read(),
'image/jpeg'),
'nonamefile': 'contents of nonamefile field',
}
When uploading a file, providing a filename (the first parameter of the
tuple) is optional but recommended to best mimic behavior of browsers.
Note that if ``headers`` are supplied, the 'Content-Type' header will
be overwritten because it depends on the dynamic random boundary string
which is used to compose the body of the request. The random boundary
string can be explicitly set with the ``multipart_boundary`` parameter.
"""
if headers is None:
headers = self.headers
extra_kw = {"headers": {}}
if fields:
if "body" in urlopen_kw:
raise TypeError(
"request got values for both 'fields' and 'body', can only specify one."
)
if encode_multipart:
body, content_type = encode_multipart_formdata(
fields, boundary=multipart_boundary
)
else:
body, content_type = (
urlencode(fields),
"application/x-www-form-urlencoded",
)
extra_kw["body"] = body
extra_kw["headers"] = {"Content-Type": content_type}
extra_kw["headers"].update(headers)
extra_kw.update(urlopen_kw)
return self.urlopen(method, url, **extra_kw)
@@ -0,0 +1,809 @@
from __future__ import absolute_import
from contextlib import contextmanager
import zlib
import io
import logging
from socket import timeout as SocketTimeout
from socket import error as SocketError
try:
import brotli
except ImportError:
brotli = None
from ._collections import HTTPHeaderDict
from .exceptions import (
BodyNotHttplibCompatible,
ProtocolError,
DecodeError,
ReadTimeoutError,
ResponseNotChunked,
IncompleteRead,
InvalidHeader,
)
from .packages.six import string_types as basestring, PY3
from .packages.six.moves import http_client as httplib
from .connection import HTTPException, BaseSSLError
from .util.response import is_fp_closed, is_response_to_head
log = logging.getLogger(__name__)
class DeflateDecoder(object):
def __init__(self):
self._first_try = True
self._data = b""
self._obj = zlib.decompressobj()
def __getattr__(self, name):
return getattr(self._obj, name)
def decompress(self, data):
if not data:
return data
if not self._first_try:
return self._obj.decompress(data)
self._data += data
try:
decompressed = self._obj.decompress(data)
if decompressed:
self._first_try = False
self._data = None
return decompressed
except zlib.error:
self._first_try = False
self._obj = zlib.decompressobj(-zlib.MAX_WBITS)
try:
return self.decompress(self._data)
finally:
self._data = None
class GzipDecoderState(object):
FIRST_MEMBER = 0
OTHER_MEMBERS = 1
SWALLOW_DATA = 2
class GzipDecoder(object):
def __init__(self):
self._obj = zlib.decompressobj(16 + zlib.MAX_WBITS)
self._state = GzipDecoderState.FIRST_MEMBER
def __getattr__(self, name):
return getattr(self._obj, name)
def decompress(self, data):
ret = bytearray()
if self._state == GzipDecoderState.SWALLOW_DATA or not data:
return bytes(ret)
while True:
try:
ret += self._obj.decompress(data)
except zlib.error:
previous_state = self._state
# Ignore data after the first error
self._state = GzipDecoderState.SWALLOW_DATA
if previous_state == GzipDecoderState.OTHER_MEMBERS:
# Allow trailing garbage acceptable in other gzip clients
return bytes(ret)
raise
data = self._obj.unused_data
if not data:
return bytes(ret)
self._state = GzipDecoderState.OTHER_MEMBERS
self._obj = zlib.decompressobj(16 + zlib.MAX_WBITS)
if brotli is not None:
class BrotliDecoder(object):
# Supports both 'brotlipy' and 'Brotli' packages
# since they share an import name. The top branches
# are for 'brotlipy' and bottom branches for 'Brotli'
def __init__(self):
self._obj = brotli.Decompressor()
def decompress(self, data):
if hasattr(self._obj, "decompress"):
return self._obj.decompress(data)
return self._obj.process(data)
def flush(self):
if hasattr(self._obj, "flush"):
return self._obj.flush()
return b""
class MultiDecoder(object):
"""
From RFC7231:
If one or more encodings have been applied to a representation, the
sender that applied the encodings MUST generate a Content-Encoding
header field that lists the content codings in the order in which
they were applied.
"""
def __init__(self, modes):
self._decoders = [_get_decoder(m.strip()) for m in modes.split(",")]
def flush(self):
return self._decoders[0].flush()
def decompress(self, data):
for d in reversed(self._decoders):
data = d.decompress(data)
return data
def _get_decoder(mode):
if "," in mode:
return MultiDecoder(mode)
if mode == "gzip":
return GzipDecoder()
if brotli is not None and mode == "br":
return BrotliDecoder()
return DeflateDecoder()
class HTTPResponse(io.IOBase):
"""
HTTP Response container.
Backwards-compatible to httplib's HTTPResponse but the response ``body`` is
loaded and decoded on-demand when the ``data`` property is accessed. This
class is also compatible with the Python standard library's :mod:`io`
module, and can hence be treated as a readable object in the context of that
framework.
Extra parameters for behaviour not present in httplib.HTTPResponse:
:param preload_content:
If True, the response's body will be preloaded during construction.
:param decode_content:
If True, will attempt to decode the body based on the
'content-encoding' header.
:param original_response:
When this HTTPResponse wrapper is generated from an httplib.HTTPResponse
object, it's convenient to include the original for debug purposes. It's
otherwise unused.
:param retries:
The retries contains the last :class:`~urllib3.util.retry.Retry` that
was used during the request.
:param enforce_content_length:
Enforce content length checking. Body returned by server must match
value of Content-Length header, if present. Otherwise, raise error.
"""
CONTENT_DECODERS = ["gzip", "deflate"]
if brotli is not None:
CONTENT_DECODERS += ["br"]
REDIRECT_STATUSES = [301, 302, 303, 307, 308]
def __init__(
self,
body="",
headers=None,
status=0,
version=0,
reason=None,
strict=0,
preload_content=True,
decode_content=True,
original_response=None,
pool=None,
connection=None,
msg=None,
retries=None,
enforce_content_length=False,
request_method=None,
request_url=None,
auto_close=True,
):
if isinstance(headers, HTTPHeaderDict):
self.headers = headers
else:
self.headers = HTTPHeaderDict(headers)
self.status = status
self.version = version
self.reason = reason
self.strict = strict
self.decode_content = decode_content
self.retries = retries
self.enforce_content_length = enforce_content_length
self.auto_close = auto_close
self._decoder = None
self._body = None
self._fp = None
self._original_response = original_response
self._fp_bytes_read = 0
self.msg = msg
self._request_url = request_url
if body and isinstance(body, (basestring, bytes)):
self._body = body
self._pool = pool
self._connection = connection
if hasattr(body, "read"):
self._fp = body
# Are we using the chunked-style of transfer encoding?
self.chunked = False
self.chunk_left = None
tr_enc = self.headers.get("transfer-encoding", "").lower()
# Don't incur the penalty of creating a list and then discarding it
encodings = (enc.strip() for enc in tr_enc.split(","))
if "chunked" in encodings:
self.chunked = True
# Determine length of response
self.length_remaining = self._init_length(request_method)
# If requested, preload the body.
if preload_content and not self._body:
self._body = self.read(decode_content=decode_content)
def get_redirect_location(self):
"""
Should we redirect and where to?
:returns: Truthy redirect location string if we got a redirect status
code and valid location. ``None`` if redirect status and no
location. ``False`` if not a redirect status code.
"""
if self.status in self.REDIRECT_STATUSES:
return self.headers.get("location")
return False
def release_conn(self):
if not self._pool or not self._connection:
return
self._pool._put_conn(self._connection)
self._connection = None
@property
def data(self):
# For backwords-compat with earlier urllib3 0.4 and earlier.
if self._body:
return self._body
if self._fp:
return self.read(cache_content=True)
@property
def connection(self):
return self._connection
def isclosed(self):
return is_fp_closed(self._fp)
def tell(self):
"""
Obtain the number of bytes pulled over the wire so far. May differ from
the amount of content returned by :meth:``HTTPResponse.read`` if bytes
are encoded on the wire (e.g, compressed).
"""
return self._fp_bytes_read
def _init_length(self, request_method):
"""
Set initial length value for Response content if available.
"""
length = self.headers.get("content-length")
if length is not None:
if self.chunked:
# This Response will fail with an IncompleteRead if it can't be
# received as chunked. This method falls back to attempt reading
# the response before raising an exception.
log.warning(
"Received response with both Content-Length and "
"Transfer-Encoding set. This is expressly forbidden "
"by RFC 7230 sec 3.3.2. Ignoring Content-Length and "
"attempting to process response as Transfer-Encoding: "
"chunked."
)
return None
try:
# RFC 7230 section 3.3.2 specifies multiple content lengths can
# be sent in a single Content-Length header
# (e.g. Content-Length: 42, 42). This line ensures the values
# are all valid ints and that as long as the `set` length is 1,
# all values are the same. Otherwise, the header is invalid.
lengths = set([int(val) for val in length.split(",")])
if len(lengths) > 1:
raise InvalidHeader(
"Content-Length contained multiple "
"unmatching values (%s)" % length
)
length = lengths.pop()
except ValueError:
length = None
else:
if length < 0:
length = None
# Convert status to int for comparison
# In some cases, httplib returns a status of "_UNKNOWN"
try:
status = int(self.status)
except ValueError:
status = 0
# Check for responses that shouldn't include a body
if status in (204, 304) or 100 <= status < 200 or request_method == "HEAD":
length = 0
return length
def _init_decoder(self):
"""
Set-up the _decoder attribute if necessary.
"""
# Note: content-encoding value should be case-insensitive, per RFC 7230
# Section 3.2
content_encoding = self.headers.get("content-encoding", "").lower()
if self._decoder is None:
if content_encoding in self.CONTENT_DECODERS:
self._decoder = _get_decoder(content_encoding)
elif "," in content_encoding:
encodings = [
e.strip()
for e in content_encoding.split(",")
if e.strip() in self.CONTENT_DECODERS
]
if len(encodings):
self._decoder = _get_decoder(content_encoding)
DECODER_ERROR_CLASSES = (IOError, zlib.error)
if brotli is not None:
DECODER_ERROR_CLASSES += (brotli.error,)
def _decode(self, data, decode_content, flush_decoder):
"""
Decode the data passed in and potentially flush the decoder.
"""
if not decode_content:
return data
try:
if self._decoder:
data = self._decoder.decompress(data)
except self.DECODER_ERROR_CLASSES as e:
content_encoding = self.headers.get("content-encoding", "").lower()
raise DecodeError(
"Received response with content-encoding: %s, but "
"failed to decode it." % content_encoding,
e,
)
if flush_decoder:
data += self._flush_decoder()
return data
def _flush_decoder(self):
"""
Flushes the decoder. Should only be called if the decoder is actually
being used.
"""
if self._decoder:
buf = self._decoder.decompress(b"")
return buf + self._decoder.flush()
return b""
@contextmanager
def _error_catcher(self):
"""
Catch low-level python exceptions, instead re-raising urllib3
variants, so that low-level exceptions are not leaked in the
high-level api.
On exit, release the connection back to the pool.
"""
clean_exit = False
try:
try:
yield
except SocketTimeout:
# FIXME: Ideally we'd like to include the url in the ReadTimeoutError but
# there is yet no clean way to get at it from this context.
raise ReadTimeoutError(self._pool, None, "Read timed out.")
except BaseSSLError as e:
# FIXME: Is there a better way to differentiate between SSLErrors?
if "read operation timed out" not in str(e): # Defensive:
# This shouldn't happen but just in case we're missing an edge
# case, let's avoid swallowing SSL errors.
raise
raise ReadTimeoutError(self._pool, None, "Read timed out.")
except (HTTPException, SocketError) as e:
# This includes IncompleteRead.
raise ProtocolError("Connection broken: %r" % e, e)
# If no exception is thrown, we should avoid cleaning up
# unnecessarily.
clean_exit = True
finally:
# If we didn't terminate cleanly, we need to throw away our
# connection.
if not clean_exit:
# The response may not be closed but we're not going to use it
# anymore so close it now to ensure that the connection is
# released back to the pool.
if self._original_response:
self._original_response.close()
# Closing the response may not actually be sufficient to close
# everything, so if we have a hold of the connection close that
# too.
if self._connection:
self._connection.close()
# If we hold the original response but it's closed now, we should
# return the connection back to the pool.
if self._original_response and self._original_response.isclosed():
self.release_conn()
def read(self, amt=None, decode_content=None, cache_content=False):
"""
Similar to :meth:`httplib.HTTPResponse.read`, but with two additional
parameters: ``decode_content`` and ``cache_content``.
:param amt:
How much of the content to read. If specified, caching is skipped
because it doesn't make sense to cache partial content as the full
response.
:param decode_content:
If True, will attempt to decode the body based on the
'content-encoding' header.
:param cache_content:
If True, will save the returned data such that the same result is
returned despite of the state of the underlying file object. This
is useful if you want the ``.data`` property to continue working
after having ``.read()`` the file object. (Overridden if ``amt`` is
set.)
"""
self._init_decoder()
if decode_content is None:
decode_content = self.decode_content
if self._fp is None:
return
flush_decoder = False
fp_closed = getattr(self._fp, "closed", False)
with self._error_catcher():
if amt is None:
# cStringIO doesn't like amt=None
data = self._fp.read() if not fp_closed else b""
flush_decoder = True
else:
cache_content = False
data = self._fp.read(amt) if not fp_closed else b""
if (
amt != 0 and not data
): # Platform-specific: Buggy versions of Python.
# Close the connection when no data is returned
#
# This is redundant to what httplib/http.client _should_
# already do. However, versions of python released before
# December 15, 2012 (http://bugs.python.org/issue16298) do
# not properly close the connection in all cases. There is
# no harm in redundantly calling close.
self._fp.close()
flush_decoder = True
if self.enforce_content_length and self.length_remaining not in (
0,
None,
):
# This is an edge case that httplib failed to cover due
# to concerns of backward compatibility. We're
# addressing it here to make sure IncompleteRead is
# raised during streaming, so all calls with incorrect
# Content-Length are caught.
raise IncompleteRead(self._fp_bytes_read, self.length_remaining)
if data:
self._fp_bytes_read += len(data)
if self.length_remaining is not None:
self.length_remaining -= len(data)
data = self._decode(data, decode_content, flush_decoder)
if cache_content:
self._body = data
return data
def stream(self, amt=2 ** 16, decode_content=None):
"""
A generator wrapper for the read() method. A call will block until
``amt`` bytes have been read from the connection or until the
connection is closed.
:param amt:
How much of the content to read. The generator will return up to
much data per iteration, but may return less. This is particularly
likely when using compressed data. However, the empty string will
never be returned.
:param decode_content:
If True, will attempt to decode the body based on the
'content-encoding' header.
"""
if self.chunked and self.supports_chunked_reads():
for line in self.read_chunked(amt, decode_content=decode_content):
yield line
else:
while not is_fp_closed(self._fp):
data = self.read(amt=amt, decode_content=decode_content)
if data:
yield data
@classmethod
def from_httplib(ResponseCls, r, **response_kw):
"""
Given an :class:`httplib.HTTPResponse` instance ``r``, return a
corresponding :class:`urllib3.response.HTTPResponse` object.
Remaining parameters are passed to the HTTPResponse constructor, along
with ``original_response=r``.
"""
headers = r.msg
if not isinstance(headers, HTTPHeaderDict):
if PY3:
headers = HTTPHeaderDict(headers.items())
else:
# Python 2.7
headers = HTTPHeaderDict.from_httplib(headers)
# HTTPResponse objects in Python 3 don't have a .strict attribute
strict = getattr(r, "strict", 0)
resp = ResponseCls(
body=r,
headers=headers,
status=r.status,
version=r.version,
reason=r.reason,
strict=strict,
original_response=r,
**response_kw
)
return resp
# Backwards-compatibility methods for httplib.HTTPResponse
def getheaders(self):
return self.headers
def getheader(self, name, default=None):
return self.headers.get(name, default)
# Backwards compatibility for http.cookiejar
def info(self):
return self.headers
# Overrides from io.IOBase
def close(self):
if not self.closed:
self._fp.close()
if self._connection:
self._connection.close()
if not self.auto_close:
io.IOBase.close(self)
@property
def closed(self):
if not self.auto_close:
return io.IOBase.closed.__get__(self)
elif self._fp is None:
return True
elif hasattr(self._fp, "isclosed"):
return self._fp.isclosed()
elif hasattr(self._fp, "closed"):
return self._fp.closed
else:
return True
def fileno(self):
if self._fp is None:
raise IOError("HTTPResponse has no file to get a fileno from")
elif hasattr(self._fp, "fileno"):
return self._fp.fileno()
else:
raise IOError(
"The file-like object this HTTPResponse is wrapped "
"around has no file descriptor"
)
def flush(self):
if (
self._fp is not None
and hasattr(self._fp, "flush")
and not getattr(self._fp, "closed", False)
):
return self._fp.flush()
def readable(self):
# This method is required for `io` module compatibility.
return True
def readinto(self, b):
# This method is required for `io` module compatibility.
temp = self.read(len(b))
if len(temp) == 0:
return 0
else:
b[: len(temp)] = temp
return len(temp)
def supports_chunked_reads(self):
"""
Checks if the underlying file-like object looks like a
httplib.HTTPResponse object. We do this by testing for the fp
attribute. If it is present we assume it returns raw chunks as
processed by read_chunked().
"""
return hasattr(self._fp, "fp")
def _update_chunk_length(self):
# First, we'll figure out length of a chunk and then
# we'll try to read it from socket.
if self.chunk_left is not None:
return
line = self._fp.fp.readline()
line = line.split(b";", 1)[0]
try:
self.chunk_left = int(line, 16)
except ValueError:
# Invalid chunked protocol response, abort.
self.close()
raise httplib.IncompleteRead(line)
def _handle_chunk(self, amt):
returned_chunk = None
if amt is None:
chunk = self._fp._safe_read(self.chunk_left)
returned_chunk = chunk
self._fp._safe_read(2) # Toss the CRLF at the end of the chunk.
self.chunk_left = None
elif amt < self.chunk_left:
value = self._fp._safe_read(amt)
self.chunk_left = self.chunk_left - amt
returned_chunk = value
elif amt == self.chunk_left:
value = self._fp._safe_read(amt)
self._fp._safe_read(2) # Toss the CRLF at the end of the chunk.
self.chunk_left = None
returned_chunk = value
else: # amt > self.chunk_left
returned_chunk = self._fp._safe_read(self.chunk_left)
self._fp._safe_read(2) # Toss the CRLF at the end of the chunk.
self.chunk_left = None
return returned_chunk
def read_chunked(self, amt=None, decode_content=None):
"""
Similar to :meth:`HTTPResponse.read`, but with an additional
parameter: ``decode_content``.
:param amt:
How much of the content to read. If specified, caching is skipped
because it doesn't make sense to cache partial content as the full
response.
:param decode_content:
If True, will attempt to decode the body based on the
'content-encoding' header.
"""
self._init_decoder()
# FIXME: Rewrite this method and make it a class with a better structured logic.
if not self.chunked:
raise ResponseNotChunked(
"Response is not chunked. "
"Header 'transfer-encoding: chunked' is missing."
)
if not self.supports_chunked_reads():
raise BodyNotHttplibCompatible(
"Body should be httplib.HTTPResponse like. "
"It should have have an fp attribute which returns raw chunks."
)
with self._error_catcher():
# Don't bother reading the body of a HEAD request.
if self._original_response and is_response_to_head(self._original_response):
self._original_response.close()
return
# If a response is already read and closed
# then return immediately.
if self._fp.fp is None:
return
while True:
self._update_chunk_length()
if self.chunk_left == 0:
break
chunk = self._handle_chunk(amt)
decoded = self._decode(
chunk, decode_content=decode_content, flush_decoder=False
)
if decoded:
yield decoded
if decode_content:
# On CPython and PyPy, we should never need to flush the
# decoder. However, on Jython we *might* need to, so
# lets defensively do it anyway.
decoded = self._flush_decoder()
if decoded: # Platform-specific: Jython.
yield decoded
# Chunk content ends with \r\n: discard it.
while True:
line = self._fp.fp.readline()
if not line:
# Some sites may not end with '\r\n'.
break
if line == b"\r\n":
break
# We read everything; close the "file".
if self._original_response:
self._original_response.close()
def geturl(self):
"""
Returns the URL that was the source of this response.
If the request that generated this response redirected, this method
will return the final redirect location.
"""
if self.retries is not None and len(self.retries.history):
return self.retries.history[-1].redirect_location
else:
return self._request_url
def __iter__(self):
buffer = [b""]
for chunk in self.stream(decode_content=True):
if b"\n" in chunk:
chunk = chunk.split(b"\n")
yield b"".join(buffer) + chunk[0] + b"\n"
for x in chunk[1:-1]:
yield x + b"\n"
if chunk[-1]:
buffer = [chunk[-1]]
else:
buffer = []
else:
buffer.append(chunk)
if buffer:
yield b"".join(buffer)
@@ -0,0 +1,46 @@
from __future__ import absolute_import
# For backwards compatibility, provide imports that used to be here.
from .connection import is_connection_dropped
from .request import make_headers
from .response import is_fp_closed
from .ssl_ import (
SSLContext,
HAS_SNI,
IS_PYOPENSSL,
IS_SECURETRANSPORT,
assert_fingerprint,
resolve_cert_reqs,
resolve_ssl_version,
ssl_wrap_socket,
PROTOCOL_TLS,
)
from .timeout import current_time, Timeout
from .retry import Retry
from .url import get_host, parse_url, split_first, Url
from .wait import wait_for_read, wait_for_write
__all__ = (
"HAS_SNI",
"IS_PYOPENSSL",
"IS_SECURETRANSPORT",
"SSLContext",
"PROTOCOL_TLS",
"Retry",
"Timeout",
"Url",
"assert_fingerprint",
"current_time",
"is_connection_dropped",
"is_fp_closed",
"get_host",
"parse_url",
"make_headers",
"resolve_cert_reqs",
"resolve_ssl_version",
"split_first",
"ssl_wrap_socket",
"wait_for_read",
"wait_for_write",
)
@@ -0,0 +1,138 @@
from __future__ import absolute_import
import socket
from .wait import NoWayToWaitForSocketError, wait_for_read
from ..contrib import _appengine_environ
def is_connection_dropped(conn): # Platform-specific
"""
Returns True if the connection is dropped and should be closed.
:param conn:
:class:`httplib.HTTPConnection` object.
Note: For platforms like AppEngine, this will always return ``False`` to
let the platform handle connection recycling transparently for us.
"""
sock = getattr(conn, "sock", False)
if sock is False: # Platform-specific: AppEngine
return False
if sock is None: # Connection already closed (such as by httplib).
return True
try:
# Returns True if readable, which here means it's been dropped
return wait_for_read(sock, timeout=0.0)
except NoWayToWaitForSocketError: # Platform-specific: AppEngine
return False
# This function is copied from socket.py in the Python 2.7 standard
# library test suite. Added to its signature is only `socket_options`.
# One additional modification is that we avoid binding to IPv6 servers
# discovered in DNS if the system doesn't have IPv6 functionality.
def create_connection(
address,
timeout=socket._GLOBAL_DEFAULT_TIMEOUT,
source_address=None,
socket_options=None,
):
"""Connect to *address* and return the socket object.
Convenience function. Connect to *address* (a 2-tuple ``(host,
port)``) and return the socket object. Passing the optional
*timeout* parameter will set the timeout on the socket instance
before attempting to connect. If no *timeout* is supplied, the
global default timeout setting returned by :func:`getdefaulttimeout`
is used. If *source_address* is set it must be a tuple of (host, port)
for the socket to bind as a source address before making the connection.
An host of '' or port 0 tells the OS to use the default.
"""
host, port = address
if host.startswith("["):
host = host.strip("[]")
err = None
# Using the value from allowed_gai_family() in the context of getaddrinfo lets
# us select whether to work with IPv4 DNS records, IPv6 records, or both.
# The original create_connection function always returns all records.
family = allowed_gai_family()
for res in socket.getaddrinfo(host, port, family, socket.SOCK_STREAM):
af, socktype, proto, canonname, sa = res
sock = None
try:
sock = socket.socket(af, socktype, proto)
# If provided, set socket level options before connecting.
_set_socket_options(sock, socket_options)
if timeout is not socket._GLOBAL_DEFAULT_TIMEOUT:
sock.settimeout(timeout)
if source_address:
sock.bind(source_address)
sock.connect(sa)
return sock
except socket.error as e:
err = e
if sock is not None:
sock.close()
sock = None
if err is not None:
raise err
raise socket.error("getaddrinfo returns an empty list")
def _set_socket_options(sock, options):
if options is None:
return
for opt in options:
sock.setsockopt(*opt)
def allowed_gai_family():
"""This function is designed to work in the context of
getaddrinfo, where family=socket.AF_UNSPEC is the default and
will perform a DNS search for both IPv6 and IPv4 records."""
family = socket.AF_INET
if HAS_IPV6:
family = socket.AF_UNSPEC
return family
def _has_ipv6(host):
""" Returns True if the system can bind an IPv6 address. """
sock = None
has_ipv6 = False
# App Engine doesn't support IPV6 sockets and actually has a quota on the
# number of sockets that can be used, so just early out here instead of
# creating a socket needlessly.
# See https://github.com/urllib3/urllib3/issues/1446
if _appengine_environ.is_appengine_sandbox():
return False
if socket.has_ipv6:
# has_ipv6 returns true if cPython was compiled with IPv6 support.
# It does not tell us if the system has IPv6 support enabled. To
# determine that we must bind to an IPv6 address.
# https://github.com/urllib3/urllib3/pull/611
# https://bugs.python.org/issue658327
try:
sock = socket.socket(socket.AF_INET6)
sock.bind((host, 0))
has_ipv6 = True
except Exception:
pass
if sock:
sock.close()
return has_ipv6
HAS_IPV6 = _has_ipv6("::1")
@@ -0,0 +1,21 @@
import collections
from ..packages import six
from ..packages.six.moves import queue
if six.PY2:
# Queue is imported for side effects on MS Windows. See issue #229.
import Queue as _unused_module_Queue # noqa: F401
class LifoQueue(queue.Queue):
def _init(self, _):
self.queue = collections.deque()
def _qsize(self, len=len):
return len(self.queue)
def _put(self, item):
self.queue.append(item)
def _get(self):
return self.queue.pop()
@@ -0,0 +1,135 @@
from __future__ import absolute_import
from base64 import b64encode
from ..packages.six import b, integer_types
from ..exceptions import UnrewindableBodyError
ACCEPT_ENCODING = "gzip,deflate"
try:
import brotli as _unused_module_brotli # noqa: F401
except ImportError:
pass
else:
ACCEPT_ENCODING += ",br"
_FAILEDTELL = object()
def make_headers(
keep_alive=None,
accept_encoding=None,
user_agent=None,
basic_auth=None,
proxy_basic_auth=None,
disable_cache=None,
):
"""
Shortcuts for generating request headers.
:param keep_alive:
If ``True``, adds 'connection: keep-alive' header.
:param accept_encoding:
Can be a boolean, list, or string.
``True`` translates to 'gzip,deflate'.
List will get joined by comma.
String will be used as provided.
:param user_agent:
String representing the user-agent you want, such as
"python-urllib3/0.6"
:param basic_auth:
Colon-separated username:password string for 'authorization: basic ...'
auth header.
:param proxy_basic_auth:
Colon-separated username:password string for 'proxy-authorization: basic ...'
auth header.
:param disable_cache:
If ``True``, adds 'cache-control: no-cache' header.
Example::
>>> make_headers(keep_alive=True, user_agent="Batman/1.0")
{'connection': 'keep-alive', 'user-agent': 'Batman/1.0'}
>>> make_headers(accept_encoding=True)
{'accept-encoding': 'gzip,deflate'}
"""
headers = {}
if accept_encoding:
if isinstance(accept_encoding, str):
pass
elif isinstance(accept_encoding, list):
accept_encoding = ",".join(accept_encoding)
else:
accept_encoding = ACCEPT_ENCODING
headers["accept-encoding"] = accept_encoding
if user_agent:
headers["user-agent"] = user_agent
if keep_alive:
headers["connection"] = "keep-alive"
if basic_auth:
headers["authorization"] = "Basic " + b64encode(b(basic_auth)).decode("utf-8")
if proxy_basic_auth:
headers["proxy-authorization"] = "Basic " + b64encode(
b(proxy_basic_auth)
).decode("utf-8")
if disable_cache:
headers["cache-control"] = "no-cache"
return headers
def set_file_position(body, pos):
"""
If a position is provided, move file to that point.
Otherwise, we'll attempt to record a position for future use.
"""
if pos is not None:
rewind_body(body, pos)
elif getattr(body, "tell", None) is not None:
try:
pos = body.tell()
except (IOError, OSError):
# This differentiates from None, allowing us to catch
# a failed `tell()` later when trying to rewind the body.
pos = _FAILEDTELL
return pos
def rewind_body(body, body_pos):
"""
Attempt to rewind body to a certain position.
Primarily used for request redirects and retries.
:param body:
File-like object that supports seek.
:param int pos:
Position to seek to in file.
"""
body_seek = getattr(body, "seek", None)
if body_seek is not None and isinstance(body_pos, integer_types):
try:
body_seek(body_pos)
except (IOError, OSError):
raise UnrewindableBodyError(
"An error occurred when rewinding request body for redirect/retry."
)
elif body_pos is _FAILEDTELL:
raise UnrewindableBodyError(
"Unable to record file position for rewinding "
"request body during a redirect/retry."
)
else:
raise ValueError(
"body_pos must be of type integer, instead it was %s." % type(body_pos)
)
@@ -0,0 +1,450 @@
from __future__ import absolute_import
import time
import logging
from collections import namedtuple
from itertools import takewhile
import email
import re
from ..exceptions import (
ConnectTimeoutError,
MaxRetryError,
ProtocolError,
ReadTimeoutError,
ResponseError,
InvalidHeader,
)
from ..packages import six
log = logging.getLogger(__name__)
# Data structure for representing the metadata of requests that result in a retry.
RequestHistory = namedtuple(
"RequestHistory", ["method", "url", "error", "status", "redirect_location"]
)
class Retry(object):
""" Retry configuration.
Each retry attempt will create a new Retry object with updated values, so
they can be safely reused.
Retries can be defined as a default for a pool::
retries = Retry(connect=5, read=2, redirect=5)
http = PoolManager(retries=retries)
response = http.request('GET', 'http://example.com/')
Or per-request (which overrides the default for the pool)::
response = http.request('GET', 'http://example.com/', retries=Retry(10))
Retries can be disabled by passing ``False``::
response = http.request('GET', 'http://example.com/', retries=False)
Errors will be wrapped in :class:`~urllib3.exceptions.MaxRetryError` unless
retries are disabled, in which case the causing exception will be raised.
:param int total:
Total number of retries to allow. Takes precedence over other counts.
Set to ``None`` to remove this constraint and fall back on other
counts. It's a good idea to set this to some sensibly-high value to
account for unexpected edge cases and avoid infinite retry loops.
Set to ``0`` to fail on the first retry.
Set to ``False`` to disable and imply ``raise_on_redirect=False``.
:param int connect:
How many connection-related errors to retry on.
These are errors raised before the request is sent to the remote server,
which we assume has not triggered the server to process the request.
Set to ``0`` to fail on the first retry of this type.
:param int read:
How many times to retry on read errors.
These errors are raised after the request was sent to the server, so the
request may have side-effects.
Set to ``0`` to fail on the first retry of this type.
:param int redirect:
How many redirects to perform. Limit this to avoid infinite redirect
loops.
A redirect is a HTTP response with a status code 301, 302, 303, 307 or
308.
Set to ``0`` to fail on the first retry of this type.
Set to ``False`` to disable and imply ``raise_on_redirect=False``.
:param int status:
How many times to retry on bad status codes.
These are retries made on responses, where status code matches
``status_forcelist``.
Set to ``0`` to fail on the first retry of this type.
:param iterable method_whitelist:
Set of uppercased HTTP method verbs that we should retry on.
By default, we only retry on methods which are considered to be
idempotent (multiple requests with the same parameters end with the
same state). See :attr:`Retry.DEFAULT_METHOD_WHITELIST`.
Set to a ``False`` value to retry on any verb.
:param iterable status_forcelist:
A set of integer HTTP status codes that we should force a retry on.
A retry is initiated if the request method is in ``method_whitelist``
and the response status code is in ``status_forcelist``.
By default, this is disabled with ``None``.
:param float backoff_factor:
A backoff factor to apply between attempts after the second try
(most errors are resolved immediately by a second try without a
delay). urllib3 will sleep for::
{backoff factor} * (2 ** ({number of total retries} - 1))
seconds. If the backoff_factor is 0.1, then :func:`.sleep` will sleep
for [0.0s, 0.2s, 0.4s, ...] between retries. It will never be longer
than :attr:`Retry.BACKOFF_MAX`.
By default, backoff is disabled (set to 0).
:param bool raise_on_redirect: Whether, if the number of redirects is
exhausted, to raise a MaxRetryError, or to return a response with a
response code in the 3xx range.
:param bool raise_on_status: Similar meaning to ``raise_on_redirect``:
whether we should raise an exception, or return a response,
if status falls in ``status_forcelist`` range and retries have
been exhausted.
:param tuple history: The history of the request encountered during
each call to :meth:`~Retry.increment`. The list is in the order
the requests occurred. Each list item is of class :class:`RequestHistory`.
:param bool respect_retry_after_header:
Whether to respect Retry-After header on status codes defined as
:attr:`Retry.RETRY_AFTER_STATUS_CODES` or not.
:param iterable remove_headers_on_redirect:
Sequence of headers to remove from the request when a response
indicating a redirect is returned before firing off the redirected
request.
"""
DEFAULT_METHOD_WHITELIST = frozenset(
["HEAD", "GET", "PUT", "DELETE", "OPTIONS", "TRACE"]
)
RETRY_AFTER_STATUS_CODES = frozenset([413, 429, 503])
DEFAULT_REDIRECT_HEADERS_BLACKLIST = frozenset(["Authorization"])
#: Maximum backoff time.
BACKOFF_MAX = 120
def __init__(
self,
total=10,
connect=None,
read=None,
redirect=None,
status=None,
method_whitelist=DEFAULT_METHOD_WHITELIST,
status_forcelist=None,
backoff_factor=0,
raise_on_redirect=True,
raise_on_status=True,
history=None,
respect_retry_after_header=True,
remove_headers_on_redirect=DEFAULT_REDIRECT_HEADERS_BLACKLIST,
):
self.total = total
self.connect = connect
self.read = read
self.status = status
if redirect is False or total is False:
redirect = 0
raise_on_redirect = False
self.redirect = redirect
self.status_forcelist = status_forcelist or set()
self.method_whitelist = method_whitelist
self.backoff_factor = backoff_factor
self.raise_on_redirect = raise_on_redirect
self.raise_on_status = raise_on_status
self.history = history or tuple()
self.respect_retry_after_header = respect_retry_after_header
self.remove_headers_on_redirect = frozenset(
[h.lower() for h in remove_headers_on_redirect]
)
def new(self, **kw):
params = dict(
total=self.total,
connect=self.connect,
read=self.read,
redirect=self.redirect,
status=self.status,
method_whitelist=self.method_whitelist,
status_forcelist=self.status_forcelist,
backoff_factor=self.backoff_factor,
raise_on_redirect=self.raise_on_redirect,
raise_on_status=self.raise_on_status,
history=self.history,
remove_headers_on_redirect=self.remove_headers_on_redirect,
respect_retry_after_header=self.respect_retry_after_header,
)
params.update(kw)
return type(self)(**params)
@classmethod
def from_int(cls, retries, redirect=True, default=None):
""" Backwards-compatibility for the old retries format."""
if retries is None:
retries = default if default is not None else cls.DEFAULT
if isinstance(retries, Retry):
return retries
redirect = bool(redirect) and None
new_retries = cls(retries, redirect=redirect)
log.debug("Converted retries value: %r -> %r", retries, new_retries)
return new_retries
def get_backoff_time(self):
""" Formula for computing the current backoff
:rtype: float
"""
# We want to consider only the last consecutive errors sequence (Ignore redirects).
consecutive_errors_len = len(
list(
takewhile(lambda x: x.redirect_location is None, reversed(self.history))
)
)
if consecutive_errors_len <= 1:
return 0
backoff_value = self.backoff_factor * (2 ** (consecutive_errors_len - 1))
return min(self.BACKOFF_MAX, backoff_value)
def parse_retry_after(self, retry_after):
# Whitespace: https://tools.ietf.org/html/rfc7230#section-3.2.4
if re.match(r"^\s*[0-9]+\s*$", retry_after):
seconds = int(retry_after)
else:
retry_date_tuple = email.utils.parsedate(retry_after)
if retry_date_tuple is None:
raise InvalidHeader("Invalid Retry-After header: %s" % retry_after)
retry_date = time.mktime(retry_date_tuple)
seconds = retry_date - time.time()
if seconds < 0:
seconds = 0
return seconds
def get_retry_after(self, response):
""" Get the value of Retry-After in seconds. """
retry_after = response.getheader("Retry-After")
if retry_after is None:
return None
return self.parse_retry_after(retry_after)
def sleep_for_retry(self, response=None):
retry_after = self.get_retry_after(response)
if retry_after:
time.sleep(retry_after)
return True
return False
def _sleep_backoff(self):
backoff = self.get_backoff_time()
if backoff <= 0:
return
time.sleep(backoff)
def sleep(self, response=None):
""" Sleep between retry attempts.
This method will respect a server's ``Retry-After`` response header
and sleep the duration of the time requested. If that is not present, it
will use an exponential backoff. By default, the backoff factor is 0 and
this method will return immediately.
"""
if self.respect_retry_after_header and response:
slept = self.sleep_for_retry(response)
if slept:
return
self._sleep_backoff()
def _is_connection_error(self, err):
""" Errors when we're fairly sure that the server did not receive the
request, so it should be safe to retry.
"""
return isinstance(err, ConnectTimeoutError)
def _is_read_error(self, err):
""" Errors that occur after the request has been started, so we should
assume that the server began processing it.
"""
return isinstance(err, (ReadTimeoutError, ProtocolError))
def _is_method_retryable(self, method):
""" Checks if a given HTTP method should be retried upon, depending if
it is included on the method whitelist.
"""
if self.method_whitelist and method.upper() not in self.method_whitelist:
return False
return True
def is_retry(self, method, status_code, has_retry_after=False):
""" Is this method/status code retryable? (Based on whitelists and control
variables such as the number of total retries to allow, whether to
respect the Retry-After header, whether this header is present, and
whether the returned status code is on the list of status codes to
be retried upon on the presence of the aforementioned header)
"""
if not self._is_method_retryable(method):
return False
if self.status_forcelist and status_code in self.status_forcelist:
return True
return (
self.total
and self.respect_retry_after_header
and has_retry_after
and (status_code in self.RETRY_AFTER_STATUS_CODES)
)
def is_exhausted(self):
""" Are we out of retries? """
retry_counts = (self.total, self.connect, self.read, self.redirect, self.status)
retry_counts = list(filter(None, retry_counts))
if not retry_counts:
return False
return min(retry_counts) < 0
def increment(
self,
method=None,
url=None,
response=None,
error=None,
_pool=None,
_stacktrace=None,
):
""" Return a new Retry object with incremented retry counters.
:param response: A response object, or None, if the server did not
return a response.
:type response: :class:`~urllib3.response.HTTPResponse`
:param Exception error: An error encountered during the request, or
None if the response was received successfully.
:return: A new ``Retry`` object.
"""
if self.total is False and error:
# Disabled, indicate to re-raise the error.
raise six.reraise(type(error), error, _stacktrace)
total = self.total
if total is not None:
total -= 1
connect = self.connect
read = self.read
redirect = self.redirect
status_count = self.status
cause = "unknown"
status = None
redirect_location = None
if error and self._is_connection_error(error):
# Connect retry?
if connect is False:
raise six.reraise(type(error), error, _stacktrace)
elif connect is not None:
connect -= 1
elif error and self._is_read_error(error):
# Read retry?
if read is False or not self._is_method_retryable(method):
raise six.reraise(type(error), error, _stacktrace)
elif read is not None:
read -= 1
elif response and response.get_redirect_location():
# Redirect retry?
if redirect is not None:
redirect -= 1
cause = "too many redirects"
redirect_location = response.get_redirect_location()
status = response.status
else:
# Incrementing because of a server error like a 500 in
# status_forcelist and a the given method is in the whitelist
cause = ResponseError.GENERIC_ERROR
if response and response.status:
if status_count is not None:
status_count -= 1
cause = ResponseError.SPECIFIC_ERROR.format(status_code=response.status)
status = response.status
history = self.history + (
RequestHistory(method, url, error, status, redirect_location),
)
new_retry = self.new(
total=total,
connect=connect,
read=read,
redirect=redirect,
status=status_count,
history=history,
)
if new_retry.is_exhausted():
raise MaxRetryError(_pool, url, error or ResponseError(cause))
log.debug("Incremented Retry for (url='%s'): %r", url, new_retry)
return new_retry
def __repr__(self):
return (
"{cls.__name__}(total={self.total}, connect={self.connect}, "
"read={self.read}, redirect={self.redirect}, status={self.status})"
).format(cls=type(self), self=self)
# For backwards compatibility (equivalent to pre-v1.9):
Retry.DEFAULT = Retry(3)
@@ -0,0 +1,407 @@
from __future__ import absolute_import
import errno
import warnings
import hmac
import sys
from binascii import hexlify, unhexlify
from hashlib import md5, sha1, sha256
from .url import IPV4_RE, BRACELESS_IPV6_ADDRZ_RE
from ..exceptions import SSLError, InsecurePlatformWarning, SNIMissingWarning
from ..packages import six
SSLContext = None
HAS_SNI = False
IS_PYOPENSSL = False
IS_SECURETRANSPORT = False
# Maps the length of a digest to a possible hash function producing this digest
HASHFUNC_MAP = {32: md5, 40: sha1, 64: sha256}
def _const_compare_digest_backport(a, b):
"""
Compare two digests of equal length in constant time.
The digests must be of type str/bytes.
Returns True if the digests match, and False otherwise.
"""
result = abs(len(a) - len(b))
for l, r in zip(bytearray(a), bytearray(b)):
result |= l ^ r
return result == 0
_const_compare_digest = getattr(hmac, "compare_digest", _const_compare_digest_backport)
try: # Test for SSL features
import ssl
from ssl import wrap_socket, CERT_REQUIRED
from ssl import HAS_SNI # Has SNI?
except ImportError:
pass
try: # Platform-specific: Python 3.6
from ssl import PROTOCOL_TLS
PROTOCOL_SSLv23 = PROTOCOL_TLS
except ImportError:
try:
from ssl import PROTOCOL_SSLv23 as PROTOCOL_TLS
PROTOCOL_SSLv23 = PROTOCOL_TLS
except ImportError:
PROTOCOL_SSLv23 = PROTOCOL_TLS = 2
try:
from ssl import OP_NO_SSLv2, OP_NO_SSLv3, OP_NO_COMPRESSION
except ImportError:
OP_NO_SSLv2, OP_NO_SSLv3 = 0x1000000, 0x2000000
OP_NO_COMPRESSION = 0x20000
# A secure default.
# Sources for more information on TLS ciphers:
#
# - https://wiki.mozilla.org/Security/Server_Side_TLS
# - https://www.ssllabs.com/projects/best-practices/index.html
# - https://hynek.me/articles/hardening-your-web-servers-ssl-ciphers/
#
# The general intent is:
# - prefer cipher suites that offer perfect forward secrecy (DHE/ECDHE),
# - prefer ECDHE over DHE for better performance,
# - prefer any AES-GCM and ChaCha20 over any AES-CBC for better performance and
# security,
# - prefer AES-GCM over ChaCha20 because hardware-accelerated AES is common,
# - disable NULL authentication, MD5 MACs, DSS, and other
# insecure ciphers for security reasons.
# - NOTE: TLS 1.3 cipher suites are managed through a different interface
# not exposed by CPython (yet!) and are enabled by default if they're available.
DEFAULT_CIPHERS = ":".join(
[
"ECDHE+AESGCM",
"ECDHE+CHACHA20",
"DHE+AESGCM",
"DHE+CHACHA20",
"ECDH+AESGCM",
"DH+AESGCM",
"ECDH+AES",
"DH+AES",
"RSA+AESGCM",
"RSA+AES",
"!aNULL",
"!eNULL",
"!MD5",
"!DSS",
]
)
try:
from ssl import SSLContext # Modern SSL?
except ImportError:
class SSLContext(object): # Platform-specific: Python 2
def __init__(self, protocol_version):
self.protocol = protocol_version
# Use default values from a real SSLContext
self.check_hostname = False
self.verify_mode = ssl.CERT_NONE
self.ca_certs = None
self.options = 0
self.certfile = None
self.keyfile = None
self.ciphers = None
def load_cert_chain(self, certfile, keyfile):
self.certfile = certfile
self.keyfile = keyfile
def load_verify_locations(self, cafile=None, capath=None):
self.ca_certs = cafile
if capath is not None:
raise SSLError("CA directories not supported in older Pythons")
def set_ciphers(self, cipher_suite):
self.ciphers = cipher_suite
def wrap_socket(self, socket, server_hostname=None, server_side=False):
warnings.warn(
"A true SSLContext object is not available. This prevents "
"urllib3 from configuring SSL appropriately and may cause "
"certain SSL connections to fail. You can upgrade to a newer "
"version of Python to solve this. For more information, see "
"https://urllib3.readthedocs.io/en/latest/advanced-usage.html"
"#ssl-warnings",
InsecurePlatformWarning,
)
kwargs = {
"keyfile": self.keyfile,
"certfile": self.certfile,
"ca_certs": self.ca_certs,
"cert_reqs": self.verify_mode,
"ssl_version": self.protocol,
"server_side": server_side,
}
return wrap_socket(socket, ciphers=self.ciphers, **kwargs)
def assert_fingerprint(cert, fingerprint):
"""
Checks if given fingerprint matches the supplied certificate.
:param cert:
Certificate as bytes object.
:param fingerprint:
Fingerprint as string of hexdigits, can be interspersed by colons.
"""
fingerprint = fingerprint.replace(":", "").lower()
digest_length = len(fingerprint)
hashfunc = HASHFUNC_MAP.get(digest_length)
if not hashfunc:
raise SSLError("Fingerprint of invalid length: {0}".format(fingerprint))
# We need encode() here for py32; works on py2 and p33.
fingerprint_bytes = unhexlify(fingerprint.encode())
cert_digest = hashfunc(cert).digest()
if not _const_compare_digest(cert_digest, fingerprint_bytes):
raise SSLError(
'Fingerprints did not match. Expected "{0}", got "{1}".'.format(
fingerprint, hexlify(cert_digest)
)
)
def resolve_cert_reqs(candidate):
"""
Resolves the argument to a numeric constant, which can be passed to
the wrap_socket function/method from the ssl module.
Defaults to :data:`ssl.CERT_NONE`.
If given a string it is assumed to be the name of the constant in the
:mod:`ssl` module or its abbreviation.
(So you can specify `REQUIRED` instead of `CERT_REQUIRED`.
If it's neither `None` nor a string we assume it is already the numeric
constant which can directly be passed to wrap_socket.
"""
if candidate is None:
return CERT_REQUIRED
if isinstance(candidate, str):
res = getattr(ssl, candidate, None)
if res is None:
res = getattr(ssl, "CERT_" + candidate)
return res
return candidate
def resolve_ssl_version(candidate):
"""
like resolve_cert_reqs
"""
if candidate is None:
return PROTOCOL_TLS
if isinstance(candidate, str):
res = getattr(ssl, candidate, None)
if res is None:
res = getattr(ssl, "PROTOCOL_" + candidate)
return res
return candidate
def create_urllib3_context(
ssl_version=None, cert_reqs=None, options=None, ciphers=None
):
"""All arguments have the same meaning as ``ssl_wrap_socket``.
By default, this function does a lot of the same work that
``ssl.create_default_context`` does on Python 3.4+. It:
- Disables SSLv2, SSLv3, and compression
- Sets a restricted set of server ciphers
If you wish to enable SSLv3, you can do::
from urllib3.util import ssl_
context = ssl_.create_urllib3_context()
context.options &= ~ssl_.OP_NO_SSLv3
You can do the same to enable compression (substituting ``COMPRESSION``
for ``SSLv3`` in the last line above).
:param ssl_version:
The desired protocol version to use. This will default to
PROTOCOL_SSLv23 which will negotiate the highest protocol that both
the server and your installation of OpenSSL support.
:param cert_reqs:
Whether to require the certificate verification. This defaults to
``ssl.CERT_REQUIRED``.
:param options:
Specific OpenSSL options. These default to ``ssl.OP_NO_SSLv2``,
``ssl.OP_NO_SSLv3``, ``ssl.OP_NO_COMPRESSION``.
:param ciphers:
Which cipher suites to allow the server to select.
:returns:
Constructed SSLContext object with specified options
:rtype: SSLContext
"""
context = SSLContext(ssl_version or PROTOCOL_TLS)
context.set_ciphers(ciphers or DEFAULT_CIPHERS)
# Setting the default here, as we may have no ssl module on import
cert_reqs = ssl.CERT_REQUIRED if cert_reqs is None else cert_reqs
if options is None:
options = 0
# SSLv2 is easily broken and is considered harmful and dangerous
options |= OP_NO_SSLv2
# SSLv3 has several problems and is now dangerous
options |= OP_NO_SSLv3
# Disable compression to prevent CRIME attacks for OpenSSL 1.0+
# (issue #309)
options |= OP_NO_COMPRESSION
context.options |= options
# Enable post-handshake authentication for TLS 1.3, see GH #1634. PHA is
# necessary for conditional client cert authentication with TLS 1.3.
# The attribute is None for OpenSSL <= 1.1.0 or does not exist in older
# versions of Python. We only enable on Python 3.7.4+ or if certificate
# verification is enabled to work around Python issue #37428
# See: https://bugs.python.org/issue37428
if (cert_reqs == ssl.CERT_REQUIRED or sys.version_info >= (3, 7, 4)) and getattr(
context, "post_handshake_auth", None
) is not None:
context.post_handshake_auth = True
context.verify_mode = cert_reqs
if (
getattr(context, "check_hostname", None) is not None
): # Platform-specific: Python 3.2
# We do our own verification, including fingerprints and alternative
# hostnames. So disable it here
context.check_hostname = False
return context
def ssl_wrap_socket(
sock,
keyfile=None,
certfile=None,
cert_reqs=None,
ca_certs=None,
server_hostname=None,
ssl_version=None,
ciphers=None,
ssl_context=None,
ca_cert_dir=None,
key_password=None,
):
"""
All arguments except for server_hostname, ssl_context, and ca_cert_dir have
the same meaning as they do when using :func:`ssl.wrap_socket`.
:param server_hostname:
When SNI is supported, the expected hostname of the certificate
:param ssl_context:
A pre-made :class:`SSLContext` object. If none is provided, one will
be created using :func:`create_urllib3_context`.
:param ciphers:
A string of ciphers we wish the client to support.
:param ca_cert_dir:
A directory containing CA certificates in multiple separate files, as
supported by OpenSSL's -CApath flag or the capath argument to
SSLContext.load_verify_locations().
:param key_password:
Optional password if the keyfile is encrypted.
"""
context = ssl_context
if context is None:
# Note: This branch of code and all the variables in it are no longer
# used by urllib3 itself. We should consider deprecating and removing
# this code.
context = create_urllib3_context(ssl_version, cert_reqs, ciphers=ciphers)
if ca_certs or ca_cert_dir:
try:
context.load_verify_locations(ca_certs, ca_cert_dir)
except IOError as e: # Platform-specific: Python 2.7
raise SSLError(e)
# Py33 raises FileNotFoundError which subclasses OSError
# These are not equivalent unless we check the errno attribute
except OSError as e: # Platform-specific: Python 3.3 and beyond
if e.errno == errno.ENOENT:
raise SSLError(e)
raise
elif ssl_context is None and hasattr(context, "load_default_certs"):
# try to load OS default certs; works well on Windows (require Python3.4+)
context.load_default_certs()
# Attempt to detect if we get the goofy behavior of the
# keyfile being encrypted and OpenSSL asking for the
# passphrase via the terminal and instead error out.
if keyfile and key_password is None and _is_key_file_encrypted(keyfile):
raise SSLError("Client private key is encrypted, password is required")
if certfile:
if key_password is None:
context.load_cert_chain(certfile, keyfile)
else:
context.load_cert_chain(certfile, keyfile, key_password)
# If we detect server_hostname is an IP address then the SNI
# extension should not be used according to RFC3546 Section 3.1
# We shouldn't warn the user if SNI isn't available but we would
# not be using SNI anyways due to IP address for server_hostname.
if (
server_hostname is not None and not is_ipaddress(server_hostname)
) or IS_SECURETRANSPORT:
if HAS_SNI and server_hostname is not None:
return context.wrap_socket(sock, server_hostname=server_hostname)
warnings.warn(
"An HTTPS request has been made, but the SNI (Server Name "
"Indication) extension to TLS is not available on this platform. "
"This may cause the server to present an incorrect TLS "
"certificate, which can cause validation failures. You can upgrade to "
"a newer version of Python to solve this. For more information, see "
"https://urllib3.readthedocs.io/en/latest/advanced-usage.html"
"#ssl-warnings",
SNIMissingWarning,
)
return context.wrap_socket(sock)
def is_ipaddress(hostname):
"""Detects whether the hostname given is an IPv4 or IPv6 address.
Also detects IPv6 addresses with Zone IDs.
:param str hostname: Hostname to examine.
:return: True if the hostname is an IP address, False otherwise.
"""
if not six.PY2 and isinstance(hostname, bytes):
# IDN A-label bytes are ASCII compatible.
hostname = hostname.decode("ascii")
return bool(IPV4_RE.match(hostname) or BRACELESS_IPV6_ADDRZ_RE.match(hostname))
def _is_key_file_encrypted(key_file):
"""Detects if a key file is encrypted or not."""
with open(key_file, "r") as f:
for line in f:
# Look for Proc-Type: 4,ENCRYPTED
if "ENCRYPTED" in line:
return True
return False