test_uuid: test_getnode and test_windll_getnode fail if connected to the Internet via an Android phone

[native-api]

BPO	31643
Nosy	@warsaw, @xdegaye, @serhiy-storchaka, @native-api
PRs	bpo-31643: test_uuid: allow zero MAC address #4448 bpo-31643: test_uuid: allow zero MAC address #4804

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2017-09-29.21:37:18.944>
labels = ['type-bug', 'tests', '3.9', '3.10', '3.11']
title = 'test_uuid: test_getnode and test_windll_getnode fail if connected to the Internet via an Android phone'
updated_at = <Date 2021-06-14.21:55:20.385>
user = 'https://github.com/native-api'

bugs.python.org fields:

activity = <Date 2021-06-14.21:55:20.385>
actor = 'iritkatriel'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Tests']
creation = <Date 2017-09-29.21:37:18.944>
creator = 'Ivan.Pozdeev'
dependencies = []
files = []
hgrepos = []
issue_num = 31643
keywords = []
message_count = 7.0
messages = ['303360', '303373', '308097', '308100', '308104', '308152', '308194']
nosy_count = 5.0
nosy_names = ['barry', '__Vano', 'xdegaye', 'serhiy.storchaka', 'Ivan.Pozdeev']
pr_nums = ['4448', '4804']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue31643'
versions = ['Python 3.9', 'Python 3.10', 'Python 3.11']

[native-api]

Ethernet emulation for some devices like Android phones' tethering use all-zero MAC addresses (which is okay since they don't actually pass Ethernet frames to other NICs). This results in a node ID of 0 if I'm currently connected to the Net via such a device. Which fails range checks in the corresponding tests.

RFC 4122 doesn't actually have any prohibitions of using a node ID of 0. Neither does IEEE 802.3 (or rather, whatever info I gathered on it since the standard's text is not freely available) assign any special meaning to an all-zero MAC address.

The patch also corrects the check call in test_windll_getnode since the tested function always generates UUID from a MAC address.

[native-api]

Example failure:

======================================================================
FAIL: test_windll_getnode (test.test_uuid.TestInternals)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "C:\Ivan\cpython\lib\test\test_uuid.py", line 501, in test_windll_getnode

    self.check_node(node)
  File "C:\Ivan\cpython\lib\test\test_uuid.py", line 449, in check_node
    "%s is not an RFC 4122 node ID" % hex)
AssertionError: False is not true : 000000000000 is not an RFC 4122 node ID

======================================================================
FAIL: test_getnode (test.test_uuid.TestUUID)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "C:\Ivan\cpython\lib\test\test_uuid.py", line 296, in test_getnode
    self.assertTrue(0 < node1 < (1 << 48), '%012x' % node1)
AssertionError: False is not true : 000000000000

[xdegaye]

On a Samsung device with Android 5.1 and tethering enabled, only one network device has a MAC address and it is not an all-zero MAC address.

Ivan can you please provide more information about this problem: which device, what Android version, what is the output of 'ip link' ?

[serhiy-storchaka]

I'm not sure that 000000000000 is better than a random address. Even if an all-zero MAC address is not explicitly prohibited it is used in a multiple independent devices. A random address looks more appropriate fore the purpose of generating a UUID.

[xdegaye]

I'm not sure that 000000000000 is better than a random address.

I agree, getnode() when obtained from a MAC address should not return 0.

RFC 4122 states (emphasis on 'uniqueness' is mine):

This document specifies three algorithms to generate UUIDs: the
first leverages the unique values of 802 MAC addresses to
guarantee *uniqueness*, ...

[Vano]

Xavier de Gaye <[email protected]> added the comment:

On a Samsung device with Android 5.1 and tethering enabled, only one network device has a MAC address and it is not an all-zero MAC address.

Ivan can you please provide more information about this problem: which device, what Android version, what is the output of 'ip link' ?

C:\> ipconfig -all
<...>
Samsung Galaxy S7 - Ethernet адаптер:

        DNS-суффикс этого подключения . . :
        Описание  . . . . . . . . . . . . : Android USB Ethernet/RNDIS
        Физический адрес. . . . . . . . . : 00-00-00-00-00-00
        Dhcp включен. . . . . . . . . . . : да
        Автонастройка включена  . . . . . : да
        IP-адрес  . . . . . . . . . . . . : 192.168.42.2
        Маска подсети . . . . . . . . . . : 255.255.255.0
        Основной шлюз . . . . . . . . . . : 192.168.42.129
        DHCP-сервер . . . . . . . . . . . : 192.168.42.129
        DNS-серверы . . . . . . . . . . . : 192.168.42.129

Device name: Samsung Galaxy S7
Model number: SM-G930F
Android version: 6.0.1
Firmware version: G930FXXU1APGJ
Baseband version: G930VVRS4APH1
Build number: MMB29M.G930VVRU2APG5

Serhiy Storchaka<[email protected]> added the comment:

I'm not sure that 000000000000 is better than a random address. Even if an all-zero MAC address is not explicitly prohibited it is used in a multiple independent devices. A random address looks more appropriate fore the purpose of generating a UUID.
The question is not whether it's good UUID source material. The
question, for the purpose of the test, is whether it's valid getnode()
output -- which it is for the current logic.

[xdegaye]

C:\> ipconfig -all

So the all-zero MAC address occurs on a Windows box and that explains the test_windll_getnode failure (this test is skipped on Android). I missed that point in your initial post and knowing what Android device you are using is irrelevant.

[mhsmith]

According to my tests, on Android API level 21, getnode uses the ip command to return the actual MAC address of the device. On API level 34, the ip command no longer returns anything for an unprivileged user, so getnode falls back on a random address, as documented. Either way, the resulting address is non-zero.

@encukou: So I think this issue can be closed.