NetworkManager/src/tests/client/test-client.py

2790 lines
94 KiB
Python
Raw Normal View History

#!/usr/bin/env python
# coding: utf-8
from __future__ import print_function
###############################################################################
#
# This test starts NetworkManager stub service in a user D-Bus session,
# and runs nmcli against it. The output is recorded and compared to a pre-generated
# expected output (src/tests/client/test-client.check-on-disk/*.expected) which
# is also committed to git.
#
###############################################################################
#
# HOWTO: Regenerate output
#
# When adjusting the tests, or when making changes to nmcli that intentionally
# change the output, the expected output must be regenerated.
#
# For that, you'd setup your system correctly (see SETUP below) and then simply:
#
# $ meson -Ddocs=true --prefix=/tmp/nm1 build
# $ ninja -C build
# $ ninja -C build install
# $ NM_TEST_REGENERATE=1 ninja -C build test
# $ git diff ... ; git add ...
# # The previous step regenerated the expected output. Review the changes
# # and consider whether they are correct. Then commit the changes to git.
#
# Beware that you need to install the sources, and beware to choose a prefix that doesn't
# mess up your system (see SETUP below).
#
# SETUP: For regenerating the output, the translations must work. First
# test whether the following works:
#
# 1) LANG=pl_PL.UTF-8 /usr/bin/nmcli --version
# # Ensure that Polish output works for the system-installed nmcli.
# # If not, you should ensure that `locale -a` reports the Polish
# # locale. If that is not the case, how to enable the locale depends on
# # your distro.
# #
# # On Debian, you might do:
# # sed -i 's/^# \(pl_PL.UTF-8 .*\)$/\1/p' /etc/locale.gen
# # locale-gen pl_PL.UTF-8
# # On Fedora, you might install `glibc-langpack-pl` package.
#
# 2) LANG=pl_PL.UTF-8 ./src/nmcli/nmcli --version
# # Ensure that the built nmcli has Polish locale working. If not,
# # you probably need to first `make install` the application at the
# # correct prefix. Take care to configure the build with the desired
# # prefix, like `meson setup build --prefix=/opt/tmp`. Usually, you want to avoid
# # using /usr as prefix, because that might overwrite files from your
# # package management system.
#
###############################################################################
#
# Environment variables to configure test:
# (optional) The build dir. Optional, mainly used to find the nmcli binary (in case
# ENV_NM_TEST_CLIENT_NMCLI_PATH is not set.
ENV_NM_TEST_CLIENT_BUILDDIR = "NM_TEST_CLIENT_BUILDDIR"
# (optional) Path to nmcli. By default, it looks for nmcli in build dir.
# In particular, you can test also a nmcli binary installed somewhere else.
ENV_NM_TEST_CLIENT_NMCLI_PATH = "NM_TEST_CLIENT_NMCLI_PATH"
# (optional) Path to nm-cloud-setup. By default, it looks for nm-cloud-setup
# in build dir.
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH = "NM_TEST_CLIENT_CLOUD_SETUP_PATH"
# (optional) The test also compares tranlsated output (l10n). This requires,
# that you first install the translation in the right place. So, by default,
# if a test for a translation fails, it will mark the test as skipped, and not
# fail the tests. Under the assumption, that the test cannot succeed currently.
# By setting NM_TEST_CLIENT_CHECK_L10N=1, you can force a failure of the test.
ENV_NM_TEST_CLIENT_CHECK_L10N = "NM_TEST_CLIENT_CHECK_L10N"
# Regenerate the .expected files. Instead of asserting, rewrite the files
# on disk with the expected output.
ENV_NM_TEST_REGENERATE = "NM_TEST_REGENERATE"
# whether the file location should include the line number. That is useful
# only for debugging, to correlate the expected output with the test.
# Obviously, since the expected output is commited to git without line numbers,
# you'd have to first NM_TEST_REGENERATE the test expected data, with line
# numbers enabled.
ENV_NM_TEST_WITH_LINENO = "NM_TEST_WITH_LINENO"
ENV_NM_TEST_ASAN_OPTIONS = "NM_TEST_ASAN_OPTIONS"
ENV_NM_TEST_LSAN_OPTIONS = "NM_TEST_LSAN_OPTIONS"
ENV_NM_TEST_UBSAN_OPTIONS = "NM_TEST_UBSAN_OPTIONS"
# Run nmcli under valgrind. If unset, we honor NMTST_USE_VALGRIND instead.
# Valgrind is always disabled, if NM_TEST_REGENERATE is enabled.
ENV_NM_TEST_VALGRIND = "NM_TEST_VALGRIND"
ENV_LIBTOOL = "LIBTOOL"
###############################################################################
import collections
import dbus
import dbus.mainloop.glib
import dbus.service
import errno
import fcntl
import io
import itertools
import os
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
import random
import re
import shlex
import signal
import socket
import subprocess
import sys
import tempfile
import time
import unittest
import gi
try:
from gi.repository import GLib
except ImportError:
GLib = None
try:
gi.require_version("NM", "1.0")
except ValueError:
NM = None
else:
try:
from gi.repository import NM
except ImportError:
NM = None
try:
import pexpect
except ImportError:
pexpect = None
try:
from http.server import HTTPServer
from http.server import BaseHTTPRequestHandler
from http.client import HTTPConnection, HTTPResponse
except ImportError:
HTTPServer = None
###############################################################################
class PathConfiguration:
@staticmethod
def srcdir():
# this is the directory where the test script itself lies.
# Based on this directory, we find other parts that we expect
# in the source repository.
return os.path.dirname(os.path.abspath(__file__))
@staticmethod
def top_srcdir():
return os.path.abspath(PathConfiguration.srcdir() + "/../../..")
@staticmethod
def test_networkmanager_service_path():
v = os.path.abspath(
PathConfiguration.top_srcdir() + "/tools/test-networkmanager-service.py"
)
assert os.path.exists(v), 'Cannot find test server at "%s"' % (v)
return v
@staticmethod
def test_cloud_meta_mock_path():
v = os.path.abspath(
PathConfiguration.top_srcdir() + "/tools/test-cloud-meta-mock.py"
)
assert os.path.exists(v), 'Cannot find cloud metadata mock server at "%s"' % (v)
return v
@staticmethod
def canonical_script_filename():
p = "src/tests/client/test-client.py"
assert (PathConfiguration.top_srcdir() + "/" + p) == os.path.abspath(__file__)
return p
###############################################################################
dbus_session_inited = False
_DEFAULT_ARG = object()
_UNSTABLE_OUTPUT = object()
###############################################################################
class Util:
_signal_no_lookup = {
1: "SIGHUP",
2: "SIGINT",
3: "SIGQUIT",
4: "SIGILL",
5: "SIGTRAP",
6: "SIGABRT",
8: "SIGFPE",
9: "SIGKILL",
11: "SIGSEGV",
12: "SIGSYS",
13: "SIGPIPE",
14: "SIGALRM",
15: "SIGTERM",
16: "SIGURG",
17: "SIGSTOP",
18: "SIGTSTP",
19: "SIGCONT",
20: "SIGCHLD",
21: "SIGTTIN",
22: "SIGTTOU",
23: "SIGPOLL",
24: "SIGXCPU",
25: "SIGXFSZ",
26: "SIGVTALRM",
27: "SIGPROF",
30: "SIGUSR1",
31: "SIGUSR2",
}
@classmethod
def signal_no_to_str(cls, sig):
s = cls._signal_no_lookup.get(sig, None)
if s is None:
return "<unknown %d>" % (sig)
return s
@staticmethod
def python_has_version(major, minor=0):
return sys.version_info[0] > major or (
sys.version_info[0] == major and sys.version_info[1] >= minor
)
@staticmethod
def is_string(s):
if Util.python_has_version(3):
t = str
else:
t = basestring
return isinstance(s, t)
@staticmethod
def is_bool(s, defval=False):
if s is None:
return defval
if isinstance(s, int):
return s != 0
if isinstance(s, str):
if s.lower() in ["1", "y", "yes", "true", "on"]:
return True
if s.lower() in ["0", "n", "no", "false", "off"]:
return False
raise ValueError('Argument "%s" is not a boolean' % (s,))
@staticmethod
def as_bytes(s):
if Util.is_string(s):
return s.encode("utf-8")
assert isinstance(s, bytes)
return s
@staticmethod
def memoize_nullary(nullary_func):
result = []
def closure():
if not result:
result.append(nullary_func())
return result[0]
return closure
_find_unsafe = re.compile(
r"[^\w@%+=:,./-]", re.ASCII if sys.version_info[0] >= 3 else 0
).search
@staticmethod
def shlex_quote(s):
# Reimplement shlex.quote().
if Util.python_has_version(3, 3):
return shlex.quote(s)
if not s:
return "''"
if Util._find_unsafe(s) is None:
return s
return "'" + s.replace("'", "'\"'\"'") + "'"
@staticmethod
def shlex_join(args):
# Reimplement shlex.join()
return " ".join(Util.shlex_quote(s) for s in args)
@staticmethod
def popen_wait(p, timeout=0):
(res, b_stdout, b_stderr) = Util.popen_wait_read(
p, timeout=timeout, read_std_pipes=False
)
return res
@staticmethod
def popen_wait_read(p, timeout=0, read_std_pipes=True):
start = NM.utils_get_timestamp_msec()
delay = 0.0005
b_stdout = b""
b_stderr = b""
res = None
while True:
if read_std_pipes:
b_stdout += Util.buffer_read(p.stdout)
b_stderr += Util.buffer_read(p.stderr)
if p.poll() is not None:
res = p.returncode
break
if timeout == 0:
break
assert timeout > 0
remaining = timeout - ((NM.utils_get_timestamp_msec() - start) / 1000.0)
if remaining <= 0:
break
delay = min(delay * 2, remaining, 0.05)
time.sleep(delay)
return (res, b_stdout, b_stderr)
@staticmethod
def buffer_read(buf):
b = b""
while True:
try:
b1 = buf.read()
except io.BlockingIOError:
b1 = b""
except IOError:
b1 = b""
if not b1:
return b
b += b1
@staticmethod
def buffer_set_nonblock(buf):
fd = buf.fileno()
fl = fcntl.fcntl(fd, fcntl.F_GETFL)
fcntl.fcntl(fd, fcntl.F_SETFL, fl | os.O_NONBLOCK)
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
@staticmethod
def random_job(jobs):
jobs = list(jobs)
l = len(jobs)
t = l * (l + 1) / 2
while True:
# we return a random jobs from the list, but the indexes at the front of
# the list are more likely. The idea is, that those jobs were started first,
# and are expected to complete first. As we poll, we want to check more frequently
# on the elements at the beginning of the list...
#
# Let's assign probabilities with an arithmetic series.
# That is, if there are 16 jobs, then the first gets weighted
# with 16, the second with 15, then 14, and so on, until the
# last has weight 1. That means, the first element is 16 times
# more probable than the last.
# Element at idx (starting with 0) is picked with probability
# 1 / (l*(l+1)/2) * (l - idx)
r = random.random() * t
idx = 0
rx = 0
while True:
rx += l - idx
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
if rx >= r or idx == l - 1:
yield jobs[idx]
break
idx += 1
@staticmethod
def iter_single(itr, min_num=1, max_num=1):
itr = list(itr)
n = 0
v = None
for c in itr:
n += 1
if n > 1:
break
v = c
if n < min_num:
raise AssertionError(
"Expected at least %s elements, but %s found" % (min_num, n)
)
if n > max_num:
raise AssertionError(
"Expected at most %s elements, but %s found" % (max_num, n)
)
return v
@staticmethod
def file_read(filename):
try:
with open(filename, "rb") as f:
return f.read()
except:
return None
@staticmethod
def file_read_expected(filename):
results_expect = []
content_expect = Util.file_read(filename)
try:
base_idx = 0
size_prefix = "size: ".encode("utf8")
while True:
if not content_expect[base_idx : base_idx + 10].startswith(size_prefix):
raise Exception("Unexpected token")
j = base_idx + len(size_prefix)
i = j
if Util.python_has_version(3, 0):
eol = ord("\n")
else:
eol = "\n"
while content_expect[i] != eol:
i += 1
i = i + 1 + int(content_expect[j:i])
results_expect.append(content_expect[base_idx:i])
if len(content_expect) == i:
break
base_idx = i
except Exception as e:
results_expect = None
return content_expect, results_expect
@staticmethod
def _replace_text_match_join(split_arr, replacement):
yield split_arr[0]
for t in split_arr[1:]:
yield (replacement,)
yield t
@staticmethod
def ReplaceTextSimple(search, replacement):
# This gives a function that can be used by Util.replace_text().
# The function replaces an input bytes string @t. It must either return
# a bytes string, a list containing bytes strings and/or 1-tuples (the
# latter containing one bytes string).
# The 1-tuple acts as a placeholder for atomic text, that cannot be replaced
# a second time.
#
# Search for replace_text_fcn in Util.replace_text() where this is called.
replacement = Util.as_bytes(replacement)
if callable(search):
search_fcn = search
else:
search_fcn = lambda: search
def replace_fcn(t):
assert isinstance(t, bytes)
search_txt = search_fcn()
if search_txt is None:
return t
search_txt = Util.as_bytes(search_txt)
return Util._replace_text_match_join(t.split(search_txt), replacement)
return replace_fcn
@staticmethod
def ReplaceTextRegex(pattern, replacement):
# See ReplaceTextSimple.
pattern = Util.as_bytes(pattern)
replacement = Util.as_bytes(replacement)
p = re.compile(pattern)
return lambda t: Util._replace_text_match_join(p.split(t), replacement)
clients/tests: fix regular expression match in Util.replace_text() Seems the previous code did not work properly: With python36-3.6.8-38.module+el8.5.0+12207+5c5719bc.x86_6 on rhel-8.6: Traceback (most recent call last): File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 1157, in f func(self) File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 1724, in test_offline replace_stdout=replace_uuids, File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 797, in call_nmcli frame, File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 997, in _call_nmcli self.async_start(wait_all=sync_barrier) File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 1032, in async_start async_job.wait_and_complete() File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 670, in wait_and_complete self._complete_cb(self, return_code, stdout, stderr) File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 919, in complete_cb stdout = Util.replace_text(stdout, replace_stdout) File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 362, in replace_text if Util.is_regex_pattern(v_search): File "/root/nm-build/NetworkManager/src/tests/client/test-client.py", line 208, in is_regex_pattern t = re.Pattern AttributeError: module 're' has no attribute 'Pattern' On this python version, re.compile() give an object of type _sre.SRE_Pattern. # python -c 'import re; print(type(re.compile("a")))' <class '_sre.SRE_Pattern'> Fixes: beebde9e56c3 ('client/test: allow matching and replacing regex-es in nmcli output')
2022-04-20 08:37:09 +02:00
@staticmethod
def replace_text(text, replace_arr):
if not replace_arr:
return text
needs_encode = Util.python_has_version(3) and Util.is_string(text)
if needs_encode:
text = text.encode("utf-8")
text = [text]
for replace_text_fcn in replace_arr:
text2 = []
for t in text:
# tuples are markers for atomic strings. They won't be replaced a second
# time.
if not isinstance(t, tuple):
t = replace_text_fcn(t)
if isinstance(t, bytes) or isinstance(t, tuple):
text2.append(t)
else:
text2.extend(t)
text = text2
bb = b"".join([(t[0] if isinstance(t, tuple) else t) for t in text])
if needs_encode:
bb = bb.decode("utf-8")
return bb
@staticmethod
def replace_text_sort_list(lst, replace_arr):
lst = [(Util.replace_text(elem, replace_arr), elem) for elem in lst]
lst = sorted(lst)
lst = [tup[1] for tup in lst]
return list(lst)
@staticmethod
def debug_dbus_interface():
# this is for printf debugging, not used in actual code.
os.system(
"busctl --user --verbose call org.freedesktop.NetworkManager /org/freedesktop org.freedesktop.DBus.ObjectManager GetManagedObjects | cat"
)
@staticmethod
def iter_nmcli_output_modes():
for mode in [[], ["--mode", "tabular"], ["--mode", "multiline"]]:
for fmt in [[], ["--pretty"], ["--terse"]]:
for color in [[], ["--color", "yes"]]:
2018-06-21 16:51:43 +02:00
yield mode + fmt + color
@staticmethod
def valgrind_check_log(valgrind_log, logname):
if valgrind_log is None:
return
fd, name = valgrind_log
os.close(fd)
if not os.path.isfile(name):
raise Exception("valgrind log %s unexpectedly does not exist" % (name,))
if os.path.getsize(name) != 0:
out = subprocess.run(
[
"sed",
"-e",
r"/^--[0-9]\+-- WARNING: unhandled .* syscall: /,/^--[0-9]\+-- it at http.*\.$/d",
name,
],
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
)
if out.returncode != 0:
raise Exception('Calling "sed" to search valgrind log failed')
if out.stdout:
print("valgrind log %s for %s is not empty:" % (name, logname))
print("\n%s\n" % (out.stdout.decode("utf-8", errors="replace"),))
raise Exception("valgrind log %s unexpectedly is not empty" % (name,))
os.remove(name)
test-client: fix race in test_monitor() test During srv_shutdown() we do p.stdin.close() p.kill() Usually, the kill wins and the service just drops off the bus: libnm-dbus[3201919]: <debug> [438617.45324] nmclient[40f7938626f3f5f0]: name owner changed: ":1.1" -> (null) libnm-dbus[3201919]: <debug> [438617.45332] nmclient[40f7938626f3f5f0]: release all at which point all objects in NMClient get destroyed and the signals get emitted in the order: libnm-dbus[3201919]: <trace> [438617.45574] nmclient[40f7938626f3f5f0]: [nmclient] emit "device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 nmcli[out]: eth0: device removed libnm-dbus[3201919]: <trace> [438617.45590] nmclient[40f7938626f3f5f0]: [nmclient] emit "any-device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 libnm-dbus[3201919]: <trace> [438617.45593] nmclient[40f7938626f3f5f0]: [nmclient] emit "connection-removed" signal for /org/freedesktop/NetworkManager/Settings/Connectio> nmcli[out]: con-1: connection profile removed However, sometimes the stub service notices that stdin was closed and it sends signals about shutting down: libnm-dbus[3201061]: <trace> [438226.44965] nmclient[401639659459c316]: interfaces-removed: [/org/freedesktop/NetworkManager/Settings] receive interface remove event for > libnm-dbus[3201061]: <trace> [438226.44966] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: changed-type 0x01 linked libnm-dbus[3201061]: <trace> [438226.44967] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: changed-type 0x01 consumed libnm-dbus[3201061]: <trace> [438226.44968] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: changed-type 0x02 linked libnm-dbus[3201061]: <trace> [438226.44969] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: unregister NMClient from D-Bus object libnm-dbus[3201061]: <trace> [438226.44971] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: drop D-Bus instance libnm-dbus[3201061]: <trace> [438226.44971] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: set D-Bus object state unlinked libnm-dbus[3201061]: <trace> [438226.44972] nmclient[401639659459c316]: [nmclient] emit "connection-removed" signal for /org/freedesktop/NetworkManager/Settings/Connectio> nmcli[out]: con-1: connection profile removed libnm-dbus[3201061]: <trace> [438226.44992] nmclient[401639659459c316]: interfaces-removed: [/org/freedesktop/NetworkManager] receive interface remove event for interface> libnm-dbus[3201061]: <trace> [438226.44994] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: changed-type 0x01 linked libnm-dbus[3201061]: <trace> [438226.44995] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: changed-type 0x01 consumed libnm-dbus[3201061]: <trace> [438226.44996] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: changed-type 0x02 linked libnm-dbus[3201061]: <trace> [438226.44998] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: unregister NMClient from D-Bus object libnm-dbus[3201061]: <trace> [438226.44999] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: drop D-Bus instance libnm-dbus[3201061]: <trace> [438226.45000] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: set D-Bus object state unlinked libnm-dbus[3201061]: <trace> [438226.45001] nmclient[401639659459c316]: [nmclient] emit "device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 nmcli[out]: eth0: device removed libnm-dbus[3201061]: <trace> [438226.45005] nmclient[401639659459c316]: [nmclient] emit "any-device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 nmcli[out]: NetworkManager is stopped libnm-dbus[3201061]: <debug> [438226.45545] nmclient[401639659459c316]: name owner changed: ":1.1" -> (null) libnm-dbus[3201061]: <debug> [438226.45550] nmclient[401639659459c316]: release all The fix is to accept the events in any order. (cherry picked from commit 8548ab29ee1eaad43bbfe099181dab905e229446)
2023-02-07 11:26:58 +01:00
@staticmethod
def pexpect_expect_all(pexp, *pattern_list):
# This will call "pexpect.expect()" on pattern_list,
# expecting all entries to match exactly once, in any
# order.
pattern_list = list(pattern_list)
while pattern_list:
idx = pexp.expect(pattern_list)
del pattern_list[idx]
@staticmethod
def skip_without_pexpect(_func=None):
if _func is None:
if pexpect is None:
raise unittest.SkipTest("pexpect not available")
return
def f(*a, **kw):
Util.skip_without_pexpect()
_func(*a, **kw)
return f
@staticmethod
def skip_without_dbus_session(_func=None):
if _func is None:
if not dbus_session_inited:
raise unittest.SkipTest(
"Own D-Bus session for testing is not initialized. Do you have dbus-run-session available?"
)
return
def f(*a, **kw):
Util.skip_without_dbus_session()
_func(*a, **kw)
return f
@staticmethod
def skip_without_NM(_func=None):
if _func is None:
if NM is None:
raise unittest.SkipTest(
"gi.NM is not available. Did you build with introspection?"
)
return
def f(*a, **kw):
Util.skip_without_NM()
_func(*a, **kw)
return f
@staticmethod
def cmd_create_env(
lang="C",
calling_num=None,
fatal_warnings=_DEFAULT_ARG,
extra_env=None,
):
if lang == "C":
language = ""
elif lang == "de_DE.utf8":
language = "de"
elif lang == "pl_PL.UTF-8":
language = "pl"
else:
raise AssertionError("invalid language %s" % (lang))
env = {}
for k in [
"LD_LIBRARY_PATH",
"DBUS_SESSION_BUS_ADDRESS",
"LIBNM_CLIENT_DEBUG",
"LIBNM_CLIENT_DEBUG_FILE",
]:
val = os.environ.get(k, None)
if val is not None:
env[k] = val
env["LANG"] = lang
env["LANGUAGE"] = language
env["LIBNM_USE_SESSION_BUS"] = "1"
env["LIBNM_USE_NO_UDEV"] = "1"
env["TERM"] = "linux"
env["ASAN_OPTIONS"] = conf.get(ENV_NM_TEST_ASAN_OPTIONS)
env["LSAN_OPTIONS"] = conf.get(ENV_NM_TEST_LSAN_OPTIONS)
env["LBSAN_OPTIONS"] = conf.get(ENV_NM_TEST_UBSAN_OPTIONS)
env["XDG_CONFIG_HOME"] = PathConfiguration.srcdir()
if calling_num is not None:
env["NM_TEST_CALLING_NUM"] = str(calling_num)
if fatal_warnings is _DEFAULT_ARG or fatal_warnings:
env["G_DEBUG"] = "fatal-warnings"
if extra_env is not None:
for k, v in extra_env.items():
env[k] = v
return env
@staticmethod
def cmd_create_argv(cmd_path, args, with_valgrind=None):
if with_valgrind is None:
with_valgrind = conf.get(ENV_NM_TEST_VALGRIND)
valgrind_log = None
cmd = conf.get(cmd_path)
if with_valgrind:
valgrind_log = tempfile.mkstemp(prefix="nm-test-client-valgrind.")
argv = [
"valgrind",
"--quiet",
"--error-exitcode=37",
"--leak-check=full",
"--gen-suppressions=all",
(
"--suppressions="
+ PathConfiguration.top_srcdir()
+ "/valgrind.suppressions"
),
"--num-callers=100",
"--log-file=" + valgrind_log[1],
cmd,
]
libtool = conf.get(ENV_LIBTOOL)
if libtool:
argv = list(libtool) + ["--mode=execute"] + argv
else:
argv = [cmd]
argv.extend(args)
return argv, valgrind_log
@staticmethod
def cmd_call_pexpect(cmd_path, args, extra_env):
argv, valgrind_log = Util.cmd_create_argv(cmd_path, args)
env = Util.cmd_create_env(extra_env=extra_env)
pexp = pexpect.spawn(argv[0], argv[1:], timeout=10, env=env)
pexp.str_last_chars = 100000
typ = collections.namedtuple("CallPexpect", ["pexp", "valgrind_log"])
return typ(pexp, valgrind_log)
@staticmethod
def cmd_call_pexpect_nmcli(args, extra_env={}):
extra_env = extra_env.copy()
extra_env.update({"NO_COLOR": "1"})
return Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_NMCLI_PATH,
args,
extra_env,
)
@staticmethod
def get_nmcli_version():
ver = NM.utils_version()
micro = ver & 0xFF
minor = (ver >> 8) & 0xFF
major = ver >> 16
return "%s.%s.%s" % (major, minor, micro)
###############################################################################
class Configuration:
def __init__(self):
self._values = {}
def get(self, name):
v = self._values.get(name, None)
if name in self._values:
return v
if name == ENV_NM_TEST_CLIENT_BUILDDIR:
v = os.environ.get(
ENV_NM_TEST_CLIENT_BUILDDIR, PathConfiguration.top_srcdir()
)
if not os.path.isdir(v):
raise Exception("Missing builddir. Set NM_TEST_CLIENT_BUILDDIR?")
elif name == ENV_NM_TEST_CLIENT_NMCLI_PATH:
v = os.environ.get(ENV_NM_TEST_CLIENT_NMCLI_PATH, None)
if v is None:
try:
v = os.path.abspath(
self.get(ENV_NM_TEST_CLIENT_BUILDDIR) + "/src/nmcli/nmcli"
)
except:
pass
if not os.path.exists(v):
raise Exception("Missing nmcli binary. Set NM_TEST_CLIENT_NMCLI_PATH?")
elif name == ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH:
v = os.environ.get(ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH, None)
if v is None:
try:
v = os.path.abspath(
self.get(ENV_NM_TEST_CLIENT_BUILDDIR)
+ "/src/nm-cloud-setup/nm-cloud-setup"
)
except:
pass
if not os.path.exists(v):
raise Exception(
"Missing nm-cloud-setup binary. Set NM_TEST_CLIENT_CLOUD_SETUP_PATH?"
)
elif name == ENV_NM_TEST_CLIENT_CHECK_L10N:
# if we test locales other than 'C', the output of nmcli depends on whether
# nmcli can load the translations. Unfortunately, I cannot find a way to
# make gettext use the po/*.gmo files from the build-dir.
#
# hence, such tests only work, if you also issue `make-install`
#
# Only by setting NM_TEST_CLIENT_CHECK_L10N=1, these tests are included
# as well.
v = Util.is_bool(os.environ.get(ENV_NM_TEST_CLIENT_CHECK_L10N, None))
elif name == ENV_NM_TEST_REGENERATE:
# in the "regenerate" mode, the tests will rewrite the files on disk against
# which we assert. That is useful, if there are intentional changes and
# we want to regenerate the expected output.
v = Util.is_bool(os.environ.get(ENV_NM_TEST_REGENERATE, None))
elif name == ENV_NM_TEST_WITH_LINENO:
v = Util.is_bool(os.environ.get(ENV_NM_TEST_WITH_LINENO, None))
elif name == ENV_NM_TEST_VALGRIND:
if self.get(ENV_NM_TEST_REGENERATE):
v = False
else:
v = os.environ.get(ENV_NM_TEST_VALGRIND, None)
if v:
v = Util.is_bool(v)
else:
v = Util.is_bool(os.environ.get("NMTST_USE_VALGRIND", None))
elif name in [
ENV_NM_TEST_ASAN_OPTIONS,
ENV_NM_TEST_LSAN_OPTIONS,
ENV_NM_TEST_UBSAN_OPTIONS,
]:
v = os.environ.get(name, None)
if v is None:
if name == ENV_NM_TEST_ASAN_OPTIONS:
v = "detect_leaks=1"
# v += ' fast_unwind_on_malloc=false'
elif name == ENV_NM_TEST_LSAN_OPTIONS:
v = ""
elif name == ENV_NM_TEST_UBSAN_OPTIONS:
v = "print_stacktrace=1:halt_on_error=1"
else:
assert False
elif name == ENV_LIBTOOL:
v = os.environ.get(name, None)
if v is None:
v = os.path.abspath(
os.path.dirname(self.get(ENV_NM_TEST_CLIENT_NMCLI_PATH))
+ "/../../libtool"
)
if not os.path.isfile(v):
v = None
else:
v = [v]
elif not v:
v = None
else:
v = shlex.split(v)
else:
raise Exception()
self._values[name] = v
return v
conf = Configuration()
###############################################################################
class NMStubServer:
@staticmethod
def _conn_get_main_object(conn):
try:
return conn.get_object(
"org.freedesktop.NetworkManager", "/org/freedesktop/NetworkManager"
)
except:
return None
def __init__(self, seed, version=None):
service_path = PathConfiguration.test_networkmanager_service_path()
self._conn = dbus.SessionBus()
clients/tests: seed generated numbers for test-networkmanager-service.py At several places, "test-networkmanager-service.py" uses generated numbers with a defined seed. For example, generated connection's UUID is generated in a predictable, but randomized way (if you forgive the inprecise use of the word "random" in context of using a deterministic seed). Aside the connection's UUID, this becomes more interesting in the next commit where the stub server generates a list of IP and DHCP settings in a predictable randomized way. For "clients/tests" we spawn the test service multiple times, but also create similar environments by calling init_001(). This is done for convenience, where out of lazyness all the tests share one setup. But it's still a good idea that these tests generate slightly different setups, wherever applicable. this increases the possible setups which get tested. For example, the number of static IPv4 addresses (the following commit) is interested to explicitly test for zero or a non-zero number of addresses. If all tests happen to use the same seed, the tests are expected to also generate the same number of addresses, and we miss an opportunity to hit interesting test cases. There is still no guarantee that all interesting cases are hit, the chances are just better. The approach of generating the setup randomly, does not preclude that the stub-server allows to explicitly configure the setup. However, due to the sheer number of combinations that might be interesting to test, it's much simpler to rely on some randomization and have the justifid hope we catch interesting cases. Also in terms of runtime of the test, the cli unit tests should complete within few seconds. Testing every combination would result in huge tests and long runtimes. Also, the patch refactors generating random numbers in "test-networkmanager-service.py". For example, it introduces Util.RandomSeed(), which can be used to generate a sequence of different random numbers. It works by having an internal state and a counter which is combined to chain the seed and generate different numbers on each call.
2018-06-07 17:42:31 +02:00
env = os.environ.copy()
env["NM_TEST_NETWORKMANAGER_SERVICE_SEED"] = seed
if version is not None:
env["NM_TEST_NETWORKMANAGER_SERVICE_VERSION"] = version
else:
env["NM_TEST_NETWORKMANAGER_SERVICE_VERSION"] = Util.get_nmcli_version()
p = subprocess.Popen(
[sys.executable, service_path], stdin=subprocess.PIPE, env=env
)
start = NM.utils_get_timestamp_msec()
while True:
if p.poll() is not None:
p.stdin.close()
if p.returncode == 77:
raise unittest.SkipTest(
"the stub service %s exited with status 77" % (service_path)
)
raise Exception(
"the stub service %s exited unexpectedly" % (service_path)
)
nmobj = self._conn_get_main_object(self._conn)
if nmobj is not None:
break
if (NM.utils_get_timestamp_msec() - start) >= 4000:
p.stdin.close()
p.kill()
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
Util.popen_wait(p, 1)
raise Exception(
"after starting stub service the D-Bus name was not claimed in time"
)
self._nmobj = nmobj
self._nmiface = dbus.Interface(
nmobj, "org.freedesktop.NetworkManager.LibnmGlibTest"
)
self._p = p
def shutdown(self, kill_mode="random"):
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
conn = self._conn
p = self._p
self._nmobj = None
self._nmiface = None
self._conn = None
self._p = None
# The test stub service watches stdin and will do a proper
# shutdown when it closes. That means, to send signals about
# going away.
# On the other hand, just killing it will cause the process
# from dropping off the bus.
if kill_mode == "kill":
p.kill()
elif kill_mode == "stdin-close":
p.stdin.close()
else:
assert kill_mode == "random"
ops = [p.stdin.close, p.kill]
random.shuffle(ops)
ops[0]()
r = random.random()
if r < 0.75:
if r < 0.5:
time.sleep(r * 0.2)
ops[1]()
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
if Util.popen_wait(p, 1) is None:
raise Exception("Stub service did not exit in time")
p.stdin.close()
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
if self._conn_get_main_object(conn) is not None:
raise Exception(
"Stub service is not still here although it should shut down"
)
class _MethodProxy:
def __init__(self, parent, method_name):
self._parent = parent
self._method_name = method_name
def __call__(self, *args, **kwargs):
dbus_iface = kwargs.pop("dbus_iface", None)
if dbus_iface is None:
dbus_iface = self._parent._nmiface
method = dbus_iface.get_dbus_method(self._method_name)
if kwargs:
# for convenience, we allow the caller to specify arguments
# as kwargs. In this case, we construct a a{sv} array as last argument.
args = list(args)
args.append(kwargs)
return method(*args)
def __getattr__(self, member):
if not member.startswith("op_"):
raise AttributeError(member)
return self._MethodProxy(self, member[3:])
def addConnection(self, connection, do_verify_strict=True):
return self.op_AddConnection(connection, do_verify_strict)
def findConnections(self, **kwargs):
if kwargs:
lst = self.op_FindConnections(**kwargs)
else:
lst = self.op_FindConnections({})
return list([(str(elem[0]), str(elem[1]), str(elem[2])) for elem in lst])
def findConnectionUuid(self, con_id, required=True):
try:
u = Util.iter_single(self.findConnections(con_id=con_id))[1]
assert u, "Invalid uuid %s" % (u)
except Exception as e:
if not required:
return None
raise AssertionError(
"Unexpectedly not found connection %s: %s" % (con_id, str(e))
)
return u
def ReplaceTextConUuid(self, con_name, replacement):
return Util.ReplaceTextSimple(
Util.memoize_nullary(lambda: self.findConnectionUuid(con_name)),
replacement,
)
def setProperty(self, path, propname, value, iface_name=None):
if iface_name is None:
iface_name = ""
self.op_SetProperties([(path, [(iface_name, [(propname, value)])])])
def addAndActivateConnection(
self, connection, device, specific_object="", delay=None
):
if delay is not None:
self.op_SetActiveConnectionStateChangedDelay(device, delay)
nm_iface = self._conn_get_main_object(self._conn)
self.op_AddAndActivateConnection(
connection, device, specific_object, dbus_iface=nm_iface
)
###############################################################################
class AsyncProcess:
def __init__(self, args, env, complete_cb, max_waittime_msec=20000):
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
self._args = list(args)
self._env = env
self._complete_cb = complete_cb
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
self._max_waittime_msec = max_waittime_msec
def start(self):
if not hasattr(self, "_p"):
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
self._p_start_timestamp = NM.utils_get_timestamp_msec()
self._p_stdout_buf = b""
self._p_stderr_buf = b""
self._p = subprocess.Popen(
self._args,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
env=self._env,
)
Util.buffer_set_nonblock(self._p.stdout)
Util.buffer_set_nonblock(self._p.stderr)
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
def _timeout_remaining_time(self):
# note that we call this during poll() and wait_and_complete().
# we don't know the exact time when the process terminated,
# so this is only approximately correct, if we call poll/wait
# frequently.
# Worst case, we will think that the process did not time out,
# when in fact it was running longer than max-waittime.
return self._max_waittime_msec - (
NM.utils_get_timestamp_msec() - self._p_start_timestamp
)
def poll(self, timeout=0):
self.start()
(return_code, b_stdout, b_stderr) = Util.popen_wait_read(self._p, timeout)
self._p_stdout_buf += b_stdout
self._p_stderr_buf += b_stderr
if return_code is None and self._timeout_remaining_time() <= 0:
raise Exception(
"process is still running after timeout: %s" % (" ".join(self._args))
)
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
return return_code
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
def wait_and_complete(self):
self.start()
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
p = self._p
self._p = None
(return_code, b_stdout, b_stderr) = Util.popen_wait_read(
p, max(0, self._timeout_remaining_time()) / 1000
)
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
(stdout, stderr) = (p.stdout.read(), p.stderr.read())
p.stdout.close()
p.stderr.close()
stdout = self._p_stdout_buf + b_stdout + stdout
stderr = self._p_stderr_buf + b_stderr + stderr
del self._p_stdout_buf
del self._p_stderr_buf
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
if return_code is None:
print(stdout)
print(stderr)
raise Exception(
"process did not complete in time: %s" % (" ".join(self._args))
)
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
self._complete_cb(self, return_code, stdout, stderr)
###############################################################################
class NMTestContext:
MAX_JOBS = 15
def __init__(self, testMethodName):
self.testMethodName = testMethodName
self._calling_num = {}
self._skip_test_for_l10n_diff = []
self._async_jobs = []
self.ctx_results = []
self.srv = None
def calling_num(self, calling_fcn):
calling_num = self._calling_num.get(calling_fcn, 0) + 1
self._calling_num[calling_fcn] = calling_num
return calling_num
def srv_start(self, srv_version=None):
self.srv_shutdown()
self.srv = NMStubServer(self.testMethodName, srv_version)
def srv_shutdown(self):
if self.srv is not None:
srv = self.srv
self.srv = None
srv.shutdown()
def async_start(self, wait_all=False):
while True:
while True:
for async_job in list(self._async_jobs[0 : self.MAX_JOBS]):
async_job.start()
# start up to MAX_JOBS jobs, but poll() and complete those
# that are already exited. Retry, until there are no more
# jobs to start, or until MAX_JOBS are running.
jobs_running = []
for async_job in list(self._async_jobs[0 : self.MAX_JOBS]):
if async_job.poll() is not None:
self._async_jobs.remove(async_job)
async_job.wait_and_complete()
continue
jobs_running.append(async_job)
if len(jobs_running) >= len(self._async_jobs):
break
if len(jobs_running) >= self.MAX_JOBS:
break
if not jobs_running:
return
if not wait_all:
return
# in a loop, indefinitely poll the running jobs until we find one that
# completes. Note that poll() itself will raise an exception if a
# jobs times out.
for async_job in Util.random_job(jobs_running):
if async_job.poll(timeout=0.03) is not None:
self._async_jobs.remove(async_job)
async_job.wait_and_complete()
break
def async_wait(self):
return self.async_start(wait_all=True)
def async_append_job(self, async_job):
self._async_jobs.append(async_job)
###############################################################################
class TestNmcli(unittest.TestCase):
def setUp(self):
Util.skip_without_dbus_session()
Util.skip_without_NM()
self.ctx = NMTestContext(self._testMethodName)
self._skip_test_for_l10n_diff = []
def call_nmcli_l(
self,
args,
check_on_disk=_DEFAULT_ARG,
fatal_warnings=_DEFAULT_ARG,
expected_returncode=_DEFAULT_ARG,
expected_stdout=_DEFAULT_ARG,
expected_stderr=_DEFAULT_ARG,
replace_stdout=None,
replace_stderr=None,
replace_cmd=None,
sort_lines_stdout=False,
extra_env=None,
sync_barrier=False,
):
frame = sys._getframe(1)
for lang in ["C", "pl"]:
self._call_nmcli(
args,
lang,
check_on_disk,
fatal_warnings,
expected_returncode,
expected_stdout,
expected_stderr,
replace_stdout,
replace_stderr,
replace_cmd,
sort_lines_stdout,
extra_env,
sync_barrier,
frame,
)
def call_nmcli(
self,
args,
langs=None,
lang=None,
check_on_disk=_DEFAULT_ARG,
fatal_warnings=_DEFAULT_ARG,
expected_returncode=_DEFAULT_ARG,
expected_stdout=_DEFAULT_ARG,
expected_stderr=_DEFAULT_ARG,
replace_stdout=None,
replace_stderr=None,
replace_cmd=None,
sort_lines_stdout=False,
extra_env=None,
sync_barrier=None,
):
frame = sys._getframe(1)
if langs is not None:
assert lang is None
else:
if lang is None:
lang = "C"
langs = [lang]
if sync_barrier is None:
sync_barrier = len(langs) == 1
for lang in langs:
self._call_nmcli(
args,
lang,
check_on_disk,
fatal_warnings,
expected_returncode,
expected_stdout,
expected_stderr,
replace_stdout,
replace_stderr,
replace_cmd,
sort_lines_stdout,
extra_env,
sync_barrier,
frame,
)
def _call_nmcli(
self,
args,
lang,
check_on_disk,
fatal_warnings,
expected_returncode,
expected_stdout,
expected_stderr,
replace_stdout,
replace_stderr,
replace_cmd,
sort_lines_stdout,
extra_env,
sync_barrier,
frame,
):
if sync_barrier:
self.ctx.async_wait()
calling_fcn = frame.f_code.co_name
calling_num = self.ctx.calling_num(calling_fcn)
test_name = "%s-%03d" % (calling_fcn, calling_num)
# we cannot use frame.f_code.co_filename directly, because it might be different depending
# on where the file lies and which is CWD. We still want to give the location of
# the file, so that the user can easier find the source (when looking at the .expected files)
self.assertTrue(
os.path.abspath(frame.f_code.co_filename).endswith(
"/" + PathConfiguration.canonical_script_filename()
)
)
if conf.get(ENV_NM_TEST_WITH_LINENO):
calling_location = "%s:%d:%s()/%d" % (
PathConfiguration.canonical_script_filename(),
frame.f_lineno,
frame.f_code.co_name,
calling_num,
)
else:
calling_location = "%s:%s()/%d" % (
PathConfiguration.canonical_script_filename(),
frame.f_code.co_name,
calling_num,
)
if lang is None or lang == "C":
lang = "C"
elif lang == "de":
lang = "de_DE.utf8"
elif lang == "pl":
lang = "pl_PL.UTF-8"
else:
self.fail("invalid language %s" % (lang))
# Running under valgrind is not yet supported for those tests.
args, valgrind_log = Util.cmd_create_argv(
ENV_NM_TEST_CLIENT_NMCLI_PATH, args, with_valgrind=False
)
assert valgrind_log is None
if replace_stdout is not None:
replace_stdout = list(replace_stdout)
if replace_stderr is not None:
replace_stderr = list(replace_stderr)
if replace_cmd is not None:
replace_cmd = list(replace_cmd)
if check_on_disk is _DEFAULT_ARG:
check_on_disk = (
expected_returncode is _DEFAULT_ARG
and (
expected_stdout is _DEFAULT_ARG
or expected_stdout is _UNSTABLE_OUTPUT
)
and (
expected_stderr is _DEFAULT_ARG
or expected_stderr is _UNSTABLE_OUTPUT
)
)
if expected_returncode is _DEFAULT_ARG:
expected_returncode = None
if expected_stdout is _DEFAULT_ARG:
expected_stdout = None
if expected_stderr is _DEFAULT_ARG:
expected_stderr = None
results_idx = len(self.ctx.ctx_results)
self.ctx.ctx_results.append(None)
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
def complete_cb(async_job, returncode, stdout, stderr):
if expected_stdout is _UNSTABLE_OUTPUT:
stdout = "<UNSTABLE OUTPUT>".encode("utf-8")
else:
stdout = Util.replace_text(stdout, replace_stdout)
if expected_stderr is _UNSTABLE_OUTPUT:
stderr = "<UNSTABLE OUTPUT>".encode("utf-8")
else:
stderr = Util.replace_text(stderr, replace_stderr)
if sort_lines_stdout:
stdout = b"\n".join(sorted(stdout.split(b"\n")))
ignore_l10n_diff = lang != "C" and not conf.get(
ENV_NM_TEST_CLIENT_CHECK_L10N
)
if expected_stderr is not None and expected_stderr is not _UNSTABLE_OUTPUT:
if expected_stderr != stderr:
if ignore_l10n_diff:
self._skip_test_for_l10n_diff.append(test_name)
else:
self.assertEqual(expected_stderr, stderr)
if expected_stdout is not None and expected_stdout is not _UNSTABLE_OUTPUT:
if expected_stdout != stdout:
if ignore_l10n_diff:
self._skip_test_for_l10n_diff.append(test_name)
else:
self.assertEqual(expected_stdout, stdout)
if expected_returncode is not None:
self.assertEqual(expected_returncode, returncode)
if fatal_warnings is _DEFAULT_ARG:
if expected_returncode != -5:
self.assertNotEqual(returncode, -5)
elif fatal_warnings:
if expected_returncode is None:
self.assertEqual(returncode, -5)
if check_on_disk:
cmd = "$NMCLI %s" % (Util.shlex_join(args[1:]),)
cmd = Util.replace_text(cmd, replace_cmd)
if returncode < 0:
returncode_str = "%d (SIGNAL %s)" % (
returncode,
Util.signal_no_to_str(-returncode),
)
else:
returncode_str = "%d" % (returncode)
content = (
("location: %s\n" % (calling_location)).encode("utf8")
+ ("cmd: %s\n" % (cmd)).encode("utf8")
+ ("lang: %s\n" % (lang)).encode("utf8")
+ ("returncode: %s\n" % (returncode_str)).encode("utf8")
)
if len(stdout) > 0:
content += (
("stdout: %d bytes\n>>>\n" % (len(stdout))).encode("utf8")
+ stdout
+ "\n<<<\n".encode("utf8")
)
if len(stderr) > 0:
content += (
("stderr: %d bytes\n>>>\n" % (len(stderr))).encode("utf8")
+ stderr
+ "\n<<<\n".encode("utf8")
)
content = ("size: %s\n" % (len(content))).encode("utf8") + content
self.ctx.ctx_results[results_idx] = {
"test_name": test_name,
"ignore_l10n_diff": ignore_l10n_diff,
"content": content,
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
}
env = Util.cmd_create_env(lang, calling_num, fatal_warnings, extra_env)
async_job = AsyncProcess(args=args, env=env, complete_cb=complete_cb)
self.ctx.async_append_job(async_job)
self.ctx.async_start(wait_all=sync_barrier)
clients/tests: don't wait for first job before scheduling parallel jobs Previously, the test would kick off 15 processes in parallel, but the first job in the queue would block more processes from being started. That is, async_start() would only start 15 processes, but since none of them were reaped before async_wait() was called, no more than 15 jobs were running during the start phase. That is not a real issue, because the start phase is non-blocking and queues all the jobs quickly. It's not really expected that during that time many processes already completed. Anyway, this was a bit ugly. The bigger problem is that async_wait() would always block for the first job to complete, before starting more processes. That means, if the first job in the queue takes unusually long, then this blocks other processes from getting reaped and new processes from being started. Instead, don't block only one one jobs, but poll them in turn for a short amount of time. Whichever process exits first will be completed and more jobs will be started. In fact, in the current setup it's hard to notice any difference, because all nmcli invocations take about the same time and are relatively fast. That this approach parallelizes better can be seen when the runtime of jobs varies stronger (and some invocations take a notably longer time). As we later want to run nmcli under valgrind, this probably will make a difference. An alternative would be not to poll()/wait() for child processes, but somehow get notified. For example, we could use a GMainContext and watch child processes. But that's probably more complicated to do, so let's keep the naive approach with polling.
2019-10-12 11:02:21 +02:00
def run_post(self):
self.ctx.async_wait()
self.ctx.srv_shutdown()
self.ctx._calling_num = None
results = self.ctx.ctx_results
self.ctx.ctx_results = None
if len(results) == 0:
return
skip_test_for_l10n_diff = self._skip_test_for_l10n_diff
self._skip_test_for_l10n_diff = None
filename = os.path.abspath(
PathConfiguration.srcdir()
+ "/test-client.check-on-disk/"
+ self._testMethodName
+ ".expected"
)
regenerate = conf.get(ENV_NM_TEST_REGENERATE)
content_expect, results_expect = Util.file_read_expected(filename)
if results_expect is None:
if not regenerate:
self.fail(
"Failed to parse expected file '%s'. Let the test write the file by rerunning with NM_TEST_REGENERATE=1"
% (filename)
)
else:
for i in range(0, min(len(results_expect), len(results))):
n = results[i]
if results_expect[i] == n["content"]:
continue
if regenerate:
continue
if n["ignore_l10n_diff"]:
skip_test_for_l10n_diff.append(n["test_name"])
continue
print(
"\n\n\nThe file '%s' does not have the expected content:"
% (filename)
)
print("ACTUAL OUTPUT:\n[[%s]]\n" % (n["content"]))
print("EXPECT OUTPUT:\n[[%s]]\n" % (results_expect[i]))
print(
"Let the test write the file by rerunning with NM_TEST_REGENERATE=1"
)
print(
"See howto in %s for details.\n"
% (PathConfiguration.canonical_script_filename())
)
sys.stdout.flush()
self.fail(
"Unexpected output of command, expected %s. Rerun test with NM_TEST_REGENERATE=1 to regenerate files"
% (filename)
)
if len(results_expect) != len(results):
if not regenerate:
print(
"\n\n\nThe number of tests in %s does not match the expected content (%s vs %s):"
% (filename, len(results_expect), len(results))
)
if len(results_expect) < len(results):
print(
"ACTUAL OUTPUT:\n[[%s]]\n"
% (results[len(results_expect)]["content"])
)
else:
print(
"EXPECT OUTPUT:\n[[%s]]\n" % (results_expect[len(results)])
)
print(
"Let the test write the file by rerunning with NM_TEST_REGENERATE=1"
)
print(
"See howto in %s for details.\n"
% (PathConfiguration.canonical_script_filename())
)
sys.stdout.flush()
self.fail(
"Unexpected output of command, expected %s. Rerun test with NM_TEST_REGENERATE=1 to regenerate files"
% (filename)
)
if regenerate:
content_new = b"".join([r["content"] for r in results])
if content_new != content_expect:
try:
with open(filename, "wb") as content_file:
content_file.write(content_new)
except Exception as e:
self.fail("Failure to write '%s': %s" % (filename, e))
if skip_test_for_l10n_diff:
# nmcli loads translations from the installation path. This failure commonly
# happens because you did not install the binary in the --prefix, before
# running the test. Hence, translations are not available or differ.
raise unittest.SkipTest(
"Skipped asserting for localized tests %s. Set NM_TEST_CLIENT_CHECK_L10N=1 to force fail."
% (",".join(skip_test_for_l10n_diff))
)
def nm_test(func):
def f(self):
self.ctx.srv_start()
func(self)
self.run_post()
return f
def nm_test_no_dbus(func):
def f(self):
func(self)
self.run_post()
return f
def init_001(self):
self.ctx.srv.op_AddObj("WiredDevice", iface="eth0")
self.ctx.srv.op_AddObj("WiredDevice", iface="eth1")
self.ctx.srv.op_AddObj("WifiDevice", iface="wlan0")
self.ctx.srv.op_AddObj("WifiDevice", iface="wlan1")
# add another device with an identical ifname. The D-Bus API itself
# does not enforce the ifnames are unique.
self.ctx.srv.op_AddObj("WifiDevice", ident="wlan1/x", iface="wlan1")
self.ctx.srv.op_AddObj("WifiAp", device="wlan0", rsnf=0x0)
self.ctx.srv.op_AddObj("WifiAp", device="wlan0")
NM_AP_FLAGS = getattr(NM, "80211ApSecurityFlags")
rsnf = 0x0
rsnf = rsnf | NM_AP_FLAGS.PAIR_TKIP
rsnf = rsnf | NM_AP_FLAGS.PAIR_CCMP
rsnf = rsnf | NM_AP_FLAGS.GROUP_TKIP
rsnf = rsnf | NM_AP_FLAGS.GROUP_CCMP
rsnf = rsnf | NM_AP_FLAGS.KEY_MGMT_SAE
self.ctx.srv.op_AddObj("WifiAp", device="wlan0", wpaf=0x0, rsnf=rsnf)
self.ctx.srv.op_AddObj("WifiAp", device="wlan1")
self.ctx.srv.addConnection(
{"connection": {"type": "802-3-ethernet", "id": "con-1"}}
)
@nm_test
def test_001(self):
self.call_nmcli_l([])
self.call_nmcli_l(
["-f", "AP", "-mode", "multiline", "-p", "d", "show", "wlan0"]
)
self.call_nmcli_l(["c", "s"])
self.call_nmcli_l(["bogus", "s"])
for mode in Util.iter_nmcli_output_modes():
self.call_nmcli_l(mode + ["general", "permissions"])
@nm_test
def test_002(self):
self.init_001()
self.call_nmcli_l(["d"])
self.call_nmcli_l(["-f", "all", "d"])
self.call_nmcli_l([])
self.call_nmcli_l(["-f", "AP", "-mode", "multiline", "d", "show", "wlan0"])
self.call_nmcli_l(
["-f", "AP", "-mode", "multiline", "-p", "d", "show", "wlan0"]
)
self.call_nmcli_l(
["-f", "AP", "-mode", "multiline", "-t", "d", "show", "wlan0"]
)
self.call_nmcli_l(["-f", "AP", "-mode", "tabular", "d", "show", "wlan0"])
self.call_nmcli_l(["-f", "AP", "-mode", "tabular", "-p", "d", "show", "wlan0"])
self.call_nmcli_l(["-f", "AP", "-mode", "tabular", "-t", "d", "show", "wlan0"])
self.call_nmcli_l(["-f", "ALL", "d", "wifi"])
self.call_nmcli_l(["c"])
self.call_nmcli_l(["c", "s", "con-1"])
@nm_test
def test_003(self):
con_gsm_list = [
("con-gsm1", "xyz.con-gsm1"),
("con-gsm2", ""),
("con-gsm3", " "),
]
self.init_001()
replace_uuids = []
replace_uuids.append(
self.ctx.srv.ReplaceTextConUuid(
"con-xx1", "UUID-con-xx1-REPLACED-REPLACED-REPLA"
)
)
self.call_nmcli(
["c", "add", "type", "ethernet", "ifname", "*", "con-name", "con-xx1"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(["c", "s"], replace_stdout=replace_uuids)
for con_name, apn in con_gsm_list:
replace_uuids.append(
self.ctx.srv.ReplaceTextConUuid(
con_name, "UUID-" + con_name + "-REPLACED-REPLACED-REPL"
)
)
self.call_nmcli(
[
"connection",
"add",
"type",
"gsm",
"autoconnect",
"no",
"con-name",
con_name,
"ifname",
"*",
"apn",
apn,
"serial.baud",
"5",
"serial.send-delay",
"100",
"serial.pari",
"1",
"ipv4.dns-options",
" ",
],
replace_stdout=replace_uuids,
)
replace_uuids.append(
self.ctx.srv.ReplaceTextConUuid(
"ethernet", "UUID-ethernet-REPLACED-REPLACED-REPL"
)
)
self.call_nmcli(
["c", "add", "type", "ethernet", "ifname", "*"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(["c", "s"], replace_stdout=replace_uuids)
self.call_nmcli_l(["-f", "ALL", "c", "s"], replace_stdout=replace_uuids)
self.call_nmcli_l(
["--complete-args", "-f", "ALL", "c", "s", ""],
replace_stdout=replace_uuids,
sort_lines_stdout=True,
)
for con_name, apn in con_gsm_list:
self.call_nmcli_l(["con", "s", con_name], replace_stdout=replace_uuids)
self.call_nmcli_l(
["-g", "all", "con", "s", con_name], replace_stdout=replace_uuids
)
# activate the same profile on multiple devices. Our stub-implmentation
# is fine with that... although NetworkManager service would reject
# such a configuration by deactivating the profile first. But note that
# that is only an internal behavior of NetworkManager service. The D-Bus
# API perfectly allows for one profile to be active multiple times. Also
# note, that there is always a short time where one profile goes down,
# while another is activating. Hence, while real NetworkManager commonly
# does not allow that multiple profiles *stay* connected at the same
# time, there is always the possibility that a profile is activating/active
# on a device, while also activating/deactivating in parallel.
for dev in ["eth0", "eth1"]:
self.call_nmcli(["con", "up", "ethernet", "ifname", dev])
self.call_nmcli_l(["con"], replace_stdout=replace_uuids)
self.call_nmcli_l(["-f", "ALL", "con"], replace_stdout=replace_uuids)
self.call_nmcli_l(
["-f", "ALL", "con", "s", "-a"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
["-f", "ACTIVE-PATH,DEVICE,UUID", "con", "s", "-act"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
["-f", "UUID,NAME", "con", "s", "--active"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
["-f", "ALL", "con", "s", "ethernet"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
["-f", "GENERAL.STATE", "con", "s", "ethernet"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(["con", "s", "ethernet"], replace_stdout=replace_uuids)
self.call_nmcli_l(
["-f", "ALL", "dev", "status"], replace_stdout=replace_uuids
)
# test invalid call ('s' abbrevates 'status' and not 'show'
self.call_nmcli_l(
["-f", "ALL", "dev", "s", "eth0"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
["-f", "ALL", "dev", "show", "eth0"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
["-f", "ALL", "-t", "dev", "show", "eth0"], replace_stdout=replace_uuids
)
self.ctx.async_wait()
self.ctx.srv.setProperty(
"/org/freedesktop/NetworkManager/ActiveConnection/1",
"State",
dbus.UInt32(NM.ActiveConnectionState.DEACTIVATING),
)
self.call_nmcli_l(["-f", "all", "d"], replace_stdout=replace_uuids)
self.call_nmcli_l([], replace_stdout=replace_uuids)
for i in [0, 1]:
if i == 1:
self.ctx.async_wait()
self.ctx.srv.op_ConnectionSetVisible(False, con_id="ethernet")
for mode in Util.iter_nmcli_output_modes():
self.call_nmcli_l(
mode + ["-f", "ALL", "con"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
mode + ["-f", "UUID,TYPE", "con"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
mode + ["con", "s", "ethernet"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
mode
+ ["c", "s", "/org/freedesktop/NetworkManager/ActiveConnection/1"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["-f", "all", "dev", "show", "eth0"],
replace_stdout=replace_uuids,
)
@nm_test
def test_004(self):
self.init_001()
replace_uuids = []
replace_uuids.append(
self.ctx.srv.ReplaceTextConUuid(
"con-xx1", "UUID-con-xx1-REPLACED-REPLACED-REPLA"
)
)
self.call_nmcli(
[
"c",
"add",
"type",
"wifi",
"ifname",
"*",
"ssid",
"foobar",
"con-name",
"con-xx1",
],
replace_stdout=replace_uuids,
)
self.call_nmcli(["connection", "mod", "con-xx1", "ip.gateway", ""])
self.call_nmcli(
["connection", "mod", "con-xx1", "ipv4.gateway", "172.16.0.1"], lang="pl"
)
self.call_nmcli(["connection", "mod", "con-xx1", "ipv6.gateway", "::99"])
self.call_nmcli(["connection", "mod", "con-xx1", "802.abc", ""])
self.call_nmcli(
[
"connection",
"mod",
"con-xx1",
"802-11-wireless.band",
"a",
"802-11-wireless.mac-address-denylist",
"aA:Bb:cC:dd:EE:f",
]
)
self.call_nmcli(
[
"connection",
"mod",
"con-xx1",
"ipv4.addresses",
"192.168.77.5/24",
"ipv4.routes",
"2.3.4.5/32 192.168.77.1",
"ipv6.addresses",
"1:2:3:4::6/64",
"ipv6.routes",
"1:2:3:4:5:6::5/128",
]
)
self.call_nmcli_l(["con", "s", "con-xx1"], replace_stdout=replace_uuids)
self.ctx.async_wait()
replace_uuids.append(
self.ctx.srv.ReplaceTextConUuid(
"con-vpn-1", "UUID-con-vpn-1-REPLACED-REPLACED-REP"
)
)
self.call_nmcli(
[
"connection",
"add",
"type",
"vpn",
"con-name",
"con-vpn-1",
"ifname",
"*",
"vpn-type",
"openvpn",
"vpn.data",
"key1 = val1, key2 = val2, key3=val3",
],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(["con", "s"], replace_stdout=replace_uuids)
self.call_nmcli_l(["con", "s", "con-vpn-1"], replace_stdout=replace_uuids)
self.call_nmcli(["con", "up", "con-xx1"])
self.call_nmcli_l(["con", "s"], replace_stdout=replace_uuids)
self.call_nmcli(["con", "up", "con-vpn-1"])
self.call_nmcli_l(["con", "s"], replace_stdout=replace_uuids)
self.call_nmcli_l(["con", "s", "con-vpn-1"], replace_stdout=replace_uuids)
self.ctx.async_wait()
self.ctx.srv.setProperty(
"/org/freedesktop/NetworkManager/ActiveConnection/2",
"VpnState",
dbus.UInt32(NM.VpnConnectionState.ACTIVATED),
)
uuids = Util.replace_text_sort_list(
[c[1] for c in self.ctx.srv.findConnections()], replace_uuids
)
self.call_nmcli_l([], replace_stdout=replace_uuids)
self.call_nmcli(
["-f", "all", "connection", "show", "--order", "na:-active"],
replace_stdout=replace_uuids,
)
self.call_nmcli(
["-f", "all", "connection", "show", "--order", "active:-na"],
replace_stdout=replace_uuids,
)
for mode in Util.iter_nmcli_output_modes():
self.call_nmcli_l(
mode + ["con", "s", "con-vpn-1"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
mode + ["con", "s", "con-vpn-1"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
mode + ["-f", "ALL", "con", "s", "con-vpn-1"],
replace_stdout=replace_uuids,
)
# This only filters 'vpn' settings from the connection profile.
# Contrary to '-f GENERAL' below, it does not show the properties of
# the activated VPN connection. This is a nmcli bug.
self.call_nmcli_l(
mode + ["-f", "VPN", "con", "s", "con-vpn-1"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["-f", "GENERAL", "con", "s", "con-vpn-1"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(mode + ["dev", "s"], replace_stdout=replace_uuids)
self.call_nmcli_l(
mode + ["-f", "all", "dev", "status"], replace_stdout=replace_uuids
)
self.call_nmcli_l(mode + ["dev", "show"], replace_stdout=replace_uuids)
self.call_nmcli_l(
mode + ["-f", "all", "dev", "show"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
mode + ["dev", "show", "wlan0"], replace_stdout=replace_uuids
)
self.call_nmcli_l(
mode + ["-f", "all", "dev", "show", "wlan0"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ [
"-f",
"GENERAL,GENERAL.HWADDR,WIFI-PROPERTIES",
"dev",
"show",
"wlan0",
],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ [
"-f",
"GENERAL,GENERAL.HWADDR,WIFI-PROPERTIES",
"dev",
"show",
"wlan0",
],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["-f", "DEVICE,TYPE,DBUS-PATH", "dev"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["-f", "ALL", "device", "wifi", "list"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["-f", "COMMON", "device", "wifi", "list"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ [
"-f",
"NAME,SSID,SSID-HEX,BSSID,MODE,CHAN,FREQ,RATE,SIGNAL,BARS,SECURITY,WPA-FLAGS,RSN-FLAGS,DEVICE,ACTIVE,IN-USE,DBUS-PATH",
"device",
"wifi",
"list",
],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ ["-f", "ALL", "device", "wifi", "list", "bssid", "C0:E2:BE:E8:EF:B6"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ [
"-f",
"COMMON",
"device",
"wifi",
"list",
"bssid",
"C0:E2:BE:E8:EF:B6",
],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ [
"-f",
"NAME,SSID,SSID-HEX,BSSID,MODE,CHAN,FREQ,RATE,SIGNAL,BARS,SECURITY,WPA-FLAGS,RSN-FLAGS,DEVICE,ACTIVE,IN-USE,DBUS-PATH",
"device",
"wifi",
"list",
"bssid",
"C0:E2:BE:E8:EF:B6",
],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["-f", "ALL", "device", "show", "wlan0"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["-f", "COMMON", "device", "show", "wlan0"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ [
"-f",
"GENERAL,CAPABILITIES,WIFI-PROPERTIES,AP,WIRED-PROPERTIES,WIMAX-PROPERTIES,NSP,IP4,DHCP4,IP6,DHCP6,BOND,TEAM,BRIDGE,VLAN,BLUETOOTH,CONNECTIONS",
"device",
"show",
"wlan0",
],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode + ["dev", "lldp", "list", "ifname", "eth0"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
mode
+ [
"-f",
"connection.id,connection.uuid,connection.type,connection.interface-name,802-3-ethernet.mac-address,vpn.user-name",
"connection",
"show",
]
+ uuids,
replace_stdout=replace_uuids,
replace_cmd=replace_uuids,
)
replace_uuids.append(
self.ctx.srv.ReplaceTextConUuid(
"con-xx2", "UUID-con-xx2-REPLACED-REPLACED-REPLA"
)
)
self.call_nmcli(
["c", "add", "type", "ethernet", "con-name", "con-xx2", "ifname", "eth1"],
replace_stdout=replace_uuids,
)
self.ctx.srv.op_SetActiveConnectionStateChangedDelay(
"/org/freedesktop/NetworkManager/Devices/2", 50000
)
self.call_nmcli(["-wait", "0", "con", "up", "con-xx2"])
self.call_nmcli(["con", "up", "con-xx1"])
self.call_nmcli_l(
["-f", "all", "device", "status"],
replace_stdout=replace_uuids,
)
self.call_nmcli_l(
["-f", "all", "connection", "show"],
replace_stdout=replace_uuids,
)
@nm_test_no_dbus
def test_offline(self):
# Make sure we're not using D-Bus
no_dbus_env = {
"DBUS_SYSTEM_BUS_ADDRESS": "very:invalid",
"DBUS_SESSION_BUS_ADDRESS": "very:invalid",
}
# This check just makes sure the above works and the
# "nmcli g" command indeed fails talking to D-Bus
self.call_nmcli(
["g"],
extra_env=no_dbus_env,
replace_stderr=[
Util.ReplaceTextRegex(
# depending on glib version, it prints `%s', '%s', or “%s”.
# depending on libc version, it converts unicode to ? or *.
r"Key/Value pair 0, [`*?']invalid[*?'], in address element [`*?']very:invalid[*?'] does not contain an equal sign",
"Key/Value pair 0, 'invalid', in address element 'very:invalid' does not contain an equal sign",
)
],
)
replace_uuids = [
Util.ReplaceTextRegex(
r"\buuid=[-a-f0-9]+\b", "uuid=UUID-WAS-HERE-BUT-IS-NO-MORE-SADLY"
)
]
self.call_nmcli(
["--offline", "c", "add", "type", "ethernet"],
extra_env=no_dbus_env,
replace_stdout=replace_uuids,
)
self.call_nmcli(
["--offline", "c", "show"],
extra_env=no_dbus_env,
)
self.call_nmcli(
["--offline", "g"],
extra_env=no_dbus_env,
)
self.call_nmcli(
["--offline"],
extra_env=no_dbus_env,
)
self.call_nmcli(
[
"--offline",
"c",
"add",
"type",
"wifi",
"ssid",
"lala",
"802-1x.eap",
"pwd",
"802-1x.identity",
"foo",
"802-1x.password",
"bar",
],
extra_env=no_dbus_env,
replace_stdout=replace_uuids,
)
self.call_nmcli(
[
"--offline",
"c",
"add",
"type",
"wifi",
"ssid",
"lala",
"802-1x.eap",
"pwd",
"802-1x.identity",
"foo",
"802-1x.password",
"bar",
"802-1x.password-flags",
"agent-owned",
],
extra_env=no_dbus_env,
replace_stdout=replace_uuids,
)
self.call_nmcli(
["--complete-args", "--offline", "conn", "modify", "ipv6.ad"],
extra_env=no_dbus_env,
)
@Util.skip_without_pexpect
2022-06-28 14:29:24 +02:00
@nm_test
def test_ask_mode(self):
nmc = Util.cmd_call_pexpect_nmcli(["--ask", "c", "add"])
nmc.pexp.expect("Connection type:")
nmc.pexp.sendline("ethernet")
nmc.pexp.expect("Interface name:")
nmc.pexp.sendline("eth0")
nmc.pexp.expect("There are 3 optional settings for Wired Ethernet.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect("There are 2 optional settings for IPv4 protocol.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect("There are 2 optional settings for IPv6 protocol.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect("There are 4 optional settings for Proxy.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect(r"Connection 'ethernet' \(.*\) successfully added.")
nmc.pexp.expect(pexpect.EOF)
Util.valgrind_check_log(nmc.valgrind_log, "test_ask_mode")
2022-06-28 14:29:24 +02:00
@Util.skip_without_pexpect
@nm_test
def test_ask_offline(self):
# Make sure we're not using D-Bus
no_dbus_env = {
"DBUS_SYSTEM_BUS_ADDRESS": "very:invalid",
"DBUS_SESSION_BUS_ADDRESS": "very:invalid",
}
nmc = Util.cmd_call_pexpect_nmcli(
["--offline", "--ask", "c", "add"], extra_env=no_dbus_env
)
nmc.pexp.expect("Connection type:")
nmc.pexp.sendline("ethernet")
nmc.pexp.expect("Interface name:")
nmc.pexp.sendline("eth0")
nmc.pexp.expect("There are 3 optional settings for Wired Ethernet.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect("There are 2 optional settings for IPv4 protocol.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect("There are 2 optional settings for IPv6 protocol.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect("There are 4 optional settings for Proxy.")
nmc.pexp.expect(r"Do you want to provide them\? \(yes/no\) \[yes]")
nmc.pexp.sendline("no")
nmc.pexp.expect(
r"\[connection\]\r\n"
r"id=ethernet\r\n"
r"uuid=.*\r\n"
r"type=ethernet\r\n"
r"interface-name=eth0\r\n"
r"\r\n"
r"\[ethernet\]\r\n"
r"\r\n"
r"\[ipv4\]\r\n"
r"method=auto\r\n"
r"\r\n"
r"\[ipv6\]\r\n"
r"addr-gen-mode=default\r\n"
r"method=auto\r\n"
r"\r\n"
r"\[proxy\]\r\n"
)
nmc.pexp.expect(pexpect.EOF)
Util.valgrind_check_log(nmc.valgrind_log, "test_ask_offline")
@Util.skip_without_pexpect
@nm_test
def test_monitor(self):
def start_mon(self):
nmc = Util.cmd_call_pexpect_nmcli(["monitor"])
nmc.pexp.expect("NetworkManager is running")
return nmc
def end_mon(self, nmc):
nmc.pexp.kill(signal.SIGINT)
nmc.pexp.expect(pexpect.EOF)
Util.valgrind_check_log(nmc.valgrind_log, "test_monitor")
nmc = start_mon(self)
self.ctx.srv.op_AddObj("WiredDevice", iface="eth0")
nmc.pexp.expect("eth0: device created\r\n")
self.ctx.srv.addConnection(
{"connection": {"type": "802-3-ethernet", "id": "con-1"}}
)
nmc.pexp.expect("con-1: connection profile created\r\n")
end_mon(self, nmc)
nmc = start_mon(self)
self.ctx.srv_shutdown()
test-client: fix race in test_monitor() test During srv_shutdown() we do p.stdin.close() p.kill() Usually, the kill wins and the service just drops off the bus: libnm-dbus[3201919]: <debug> [438617.45324] nmclient[40f7938626f3f5f0]: name owner changed: ":1.1" -> (null) libnm-dbus[3201919]: <debug> [438617.45332] nmclient[40f7938626f3f5f0]: release all at which point all objects in NMClient get destroyed and the signals get emitted in the order: libnm-dbus[3201919]: <trace> [438617.45574] nmclient[40f7938626f3f5f0]: [nmclient] emit "device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 nmcli[out]: eth0: device removed libnm-dbus[3201919]: <trace> [438617.45590] nmclient[40f7938626f3f5f0]: [nmclient] emit "any-device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 libnm-dbus[3201919]: <trace> [438617.45593] nmclient[40f7938626f3f5f0]: [nmclient] emit "connection-removed" signal for /org/freedesktop/NetworkManager/Settings/Connectio> nmcli[out]: con-1: connection profile removed However, sometimes the stub service notices that stdin was closed and it sends signals about shutting down: libnm-dbus[3201061]: <trace> [438226.44965] nmclient[401639659459c316]: interfaces-removed: [/org/freedesktop/NetworkManager/Settings] receive interface remove event for > libnm-dbus[3201061]: <trace> [438226.44966] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: changed-type 0x01 linked libnm-dbus[3201061]: <trace> [438226.44967] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: changed-type 0x01 consumed libnm-dbus[3201061]: <trace> [438226.44968] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: changed-type 0x02 linked libnm-dbus[3201061]: <trace> [438226.44969] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: unregister NMClient from D-Bus object libnm-dbus[3201061]: <trace> [438226.44971] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: drop D-Bus instance libnm-dbus[3201061]: <trace> [438226.44971] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager/Settings]: set D-Bus object state unlinked libnm-dbus[3201061]: <trace> [438226.44972] nmclient[401639659459c316]: [nmclient] emit "connection-removed" signal for /org/freedesktop/NetworkManager/Settings/Connectio> nmcli[out]: con-1: connection profile removed libnm-dbus[3201061]: <trace> [438226.44992] nmclient[401639659459c316]: interfaces-removed: [/org/freedesktop/NetworkManager] receive interface remove event for interface> libnm-dbus[3201061]: <trace> [438226.44994] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: changed-type 0x01 linked libnm-dbus[3201061]: <trace> [438226.44995] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: changed-type 0x01 consumed libnm-dbus[3201061]: <trace> [438226.44996] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: changed-type 0x02 linked libnm-dbus[3201061]: <trace> [438226.44998] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: unregister NMClient from D-Bus object libnm-dbus[3201061]: <trace> [438226.44999] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: drop D-Bus instance libnm-dbus[3201061]: <trace> [438226.45000] nmclient[401639659459c316]: [/org/freedesktop/NetworkManager]: set D-Bus object state unlinked libnm-dbus[3201061]: <trace> [438226.45001] nmclient[401639659459c316]: [nmclient] emit "device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 nmcli[out]: eth0: device removed libnm-dbus[3201061]: <trace> [438226.45005] nmclient[401639659459c316]: [nmclient] emit "any-device-removed" signal for /org/freedesktop/NetworkManager/Devices/1 nmcli[out]: NetworkManager is stopped libnm-dbus[3201061]: <debug> [438226.45545] nmclient[401639659459c316]: name owner changed: ":1.1" -> (null) libnm-dbus[3201061]: <debug> [438226.45550] nmclient[401639659459c316]: release all The fix is to accept the events in any order. (cherry picked from commit 8548ab29ee1eaad43bbfe099181dab905e229446)
2023-02-07 11:26:58 +01:00
Util.pexpect_expect_all(
nmc.pexp,
"con-1: connection profile removed",
"eth0: device removed",
)
nmc.pexp.expect("NetworkManager is stopped")
end_mon(self, nmc)
@nm_test_no_dbus # we need dbus, but we need to pass arguments to srv_start
def test_version_warn(self):
self.ctx.srv_start(srv_version="A.B.C")
self.call_nmcli_l(
["c"],
replace_stderr=[
Util.ReplaceTextRegex(
r"\(" + Util.get_nmcli_version() + r"\)", "(X.Y.Z)"
)
],
)
@nm_test_no_dbus
def test_daemon_not_running(self):
self.call_nmcli(["c"])
###############################################################################
class TestNmCloudSetup(unittest.TestCase):
def setUp(self):
Util.skip_without_dbus_session()
Util.skip_without_NM()
self.ctx = NMTestContext(self._testMethodName)
_mac1 = "cc:00:00:00:00:01"
_mac2 = "cc:00:00:00:00:02"
_ip1 = "172.31.26.249"
_ip2 = "172.31.176.249"
def cloud_setup_test(func):
"""
Runs the mock NetworkManager along with a mock cloud metadata service.
"""
def f(self):
Util.skip_without_pexpect()
if tuple(sys.version_info[0:2]) < (3, 2):
# subprocess.Popen()'s "pass_fd" argument requires at least Python 3.2.
raise unittest.SkipTest("This test requires at least Python 3.2")
s = socket.socket()
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEPORT, 1)
s.bind(("localhost", 0))
# The same value as Python's TCPServer uses.
# Chosen by summoning the sprit of TCP under influence of
# hallucinogenic substances.
s.listen(5)
def pass_socket():
os.dup2(s.fileno(), 3)
service_path = PathConfiguration.test_cloud_meta_mock_path()
env = os.environ.copy()
env["LISTEN_FDS"] = "1"
p = subprocess.Popen(
[sys.executable, service_path, "--empty"],
stdin=subprocess.PIPE,
env=env,
pass_fds=(3,),
preexec_fn=pass_socket,
)
(hostaddr, port) = s.getsockname()
self.md_conn = HTTPConnection(hostaddr, port=port)
self.md_url = "http://%s:%d" % (hostaddr, port)
s.close()
error = None
self.ctx.srv_start()
try:
func(self)
except Exception as e:
error = e
self.ctx.async_wait()
self.ctx.srv_shutdown()
self.md_conn.close()
p.stdin.close()
p.terminate()
p.wait()
if error:
raise error
return f
def _mock_devices(self):
# Add a device with an active connection that has IPv4 configured
self.ctx.srv.op_AddObj("WiredDevice", iface="eth0", mac="cc:00:00:00:00:01")
self.ctx.srv.addAndActivateConnection(
{
"connection": {"type": "802-3-ethernet", "id": "con-eth0"},
"ipv4": {"method": "auto"},
},
"/org/freedesktop/NetworkManager/Devices/1",
delay=0,
)
# The second connection has no IPv4
self.ctx.srv.op_AddObj("WiredDevice", iface="eth1", mac="cc:00:00:00:00:02")
self.ctx.srv.addAndActivateConnection(
{"connection": {"type": "802-3-ethernet", "id": "con-eth1"}},
"/org/freedesktop/NetworkManager/Devices/2",
"",
delay=0,
)
def _mock_path(self, path, body):
self.md_conn.request("PUT", path, body=body)
self.md_conn.getresponse().read()
@cloud_setup_test
def test_aliyun(self):
self._mock_devices()
_aliyun_meta = "/2016-01-01/meta-data/"
_aliyun_macs = _aliyun_meta + "network/interfaces/macs/"
self._mock_path(_aliyun_meta, "ami-id\n")
self._mock_path(
_aliyun_macs, TestNmCloudSetup._mac2 + "\n" + TestNmCloudSetup._mac1
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac2 + "/vpc-cidr-block", "172.31.16.0/20"
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac2 + "/private-ipv4s",
TestNmCloudSetup._ip1,
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac2 + "/primary-ip-address",
TestNmCloudSetup._ip1,
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac2 + "/netmask", "255.255.255.0"
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac2 + "/gateway", "172.31.26.2"
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac1 + "/vpc-cidr-block", "172.31.166.0/20"
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac1 + "/private-ipv4s",
TestNmCloudSetup._ip2,
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac1 + "/primary-ip-address",
TestNmCloudSetup._ip2,
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac1 + "/netmask", "255.255.255.0"
)
self._mock_path(
_aliyun_macs + TestNmCloudSetup._mac1 + "/gateway", "172.31.176.2"
)
# Run nm-cloud-setup for the first time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_ALIYUN_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_ALIYUN": "yes",
},
)
nmc.pexp.expect("provider aliyun detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("get-config: start fetching meta data")
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# One of the devices has no IPv4 configuration to be modified
nmc.pexp.expect("device has no suitable applied connection. Skip")
# The other one was lacking an address set it up.
nmc.pexp.expect("some changes were applied for provider aliyun")
nmc.pexp.expect(pexpect.EOF)
# Run nm-cloud-setup for the second time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_ALIYUN_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_ALIYUN": "yes",
},
)
nmc.pexp.expect("provider aliyun detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("get-config: starting")
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# No changes this time
nmc.pexp.expect('device needs no update to applied connection "con-eth0"')
nmc.pexp.expect("no changes were applied for provider aliyun")
nmc.pexp.expect(pexpect.EOF)
Util.valgrind_check_log(nmc.valgrind_log, "test_aliyun")
@cloud_setup_test
def test_azure(self):
self._mock_devices()
_azure_meta = "/metadata/instance"
_azure_iface = _azure_meta + "/network/interface/"
_azure_query = "?format=text&api-version=2017-04-02"
self._mock_path(_azure_meta + _azure_query, "")
self._mock_path(_azure_iface + _azure_query, "0\n1\n")
self._mock_path(
_azure_iface + "0/macAddress" + _azure_query, TestNmCloudSetup._mac1
)
self._mock_path(
_azure_iface + "1/macAddress" + _azure_query, TestNmCloudSetup._mac2
)
self._mock_path(_azure_iface + "0/ipv4/ipAddress/" + _azure_query, "0\n")
self._mock_path(_azure_iface + "1/ipv4/ipAddress/" + _azure_query, "0\n")
self._mock_path(
_azure_iface + "0/ipv4/ipAddress/0/privateIpAddress" + _azure_query,
TestNmCloudSetup._ip1,
)
self._mock_path(
_azure_iface + "1/ipv4/ipAddress/0/privateIpAddress" + _azure_query,
TestNmCloudSetup._ip2,
)
self._mock_path(
_azure_iface + "0/ipv4/subnet/0/address/" + _azure_query, "172.31.16.0"
)
self._mock_path(
_azure_iface + "1/ipv4/subnet/0/address/" + _azure_query, "172.31.166.0"
)
self._mock_path(_azure_iface + "0/ipv4/subnet/0/prefix/" + _azure_query, "20")
self._mock_path(_azure_iface + "1/ipv4/subnet/0/prefix/" + _azure_query, "20")
# Run nm-cloud-setup for the first time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_AZURE_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_AZURE": "yes",
},
)
nmc.pexp.expect("provider azure detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("found azure interfaces: 2")
nmc.pexp.expect(r"interface\[0]: found a matching device with hwaddr")
nmc.pexp.expect(
r"interface\[0]: (received subnet address|received subnet prefix 20)"
)
nmc.pexp.expect(
r"interface\[0]: (received subnet address|received subnet prefix 20)"
)
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# One of the devices has no IPv4 configuration to be modified
nmc.pexp.expect("device has no suitable applied connection. Skip")
# The other one was lacking an address set it up.
nmc.pexp.expect("some changes were applied for provider azure")
nmc.pexp.expect(pexpect.EOF)
# Run nm-cloud-setup for the second time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_AZURE_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_AZURE": "yes",
},
)
nmc.pexp.expect("provider azure detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("get-config: starting")
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# No changes this time
nmc.pexp.expect('device needs no update to applied connection "con-eth0"')
nmc.pexp.expect("no changes were applied for provider azure")
nmc.pexp.expect(pexpect.EOF)
Util.valgrind_check_log(nmc.valgrind_log, "test_azure")
@cloud_setup_test
def test_ec2(self):
self._mock_devices()
_ec2_macs = "/2018-09-24/meta-data/network/interfaces/macs/"
self._mock_path("/latest/meta-data/", "ami-id\n")
self._mock_path(
_ec2_macs, TestNmCloudSetup._mac2 + "\n" + TestNmCloudSetup._mac1
)
self._mock_path(
_ec2_macs + TestNmCloudSetup._mac2 + "/subnet-ipv4-cidr-block",
"172.31.16.0/20",
)
self._mock_path(
_ec2_macs + TestNmCloudSetup._mac2 + "/local-ipv4s", TestNmCloudSetup._ip1
)
self._mock_path(
_ec2_macs + TestNmCloudSetup._mac1 + "/subnet-ipv4-cidr-block",
"172.31.166.0/20",
)
self._mock_path(
_ec2_macs + TestNmCloudSetup._mac1 + "/local-ipv4s", TestNmCloudSetup._ip2
)
# Run nm-cloud-setup for the first time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_EC2_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_EC2": "yes",
},
)
nmc.pexp.expect("provider ec2 detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("get-config: starting")
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# One of the devices has no IPv4 configuration to be modified
nmc.pexp.expect("device has no suitable applied connection. Skip")
# The other one was lacking an address set it up.
nmc.pexp.expect("some changes were applied for provider ec2")
nmc.pexp.expect(pexpect.EOF)
# Run nm-cloud-setup for the second time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_EC2_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_EC2": "yes",
},
)
nmc.pexp.expect("provider ec2 detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("get-config: starting")
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# No changes this time
nmc.pexp.expect('device needs no update to applied connection "con-eth0"')
nmc.pexp.expect("no changes were applied for provider ec2")
nmc.pexp.expect(pexpect.EOF)
Util.valgrind_check_log(nmc.valgrind_log, "test_ec2")
@cloud_setup_test
def test_gcp(self):
self._mock_devices()
gcp_meta = "/computeMetadata/v1/instance/"
gcp_iface = gcp_meta + "network-interfaces/"
self._mock_path(gcp_meta + "id", "")
self._mock_path(gcp_iface, "0\n1\n")
self._mock_path(gcp_iface + "0/mac", TestNmCloudSetup._mac1)
self._mock_path(gcp_iface + "1/mac", TestNmCloudSetup._mac2)
self._mock_path(gcp_iface + "0/forwarded-ips/", "0\n")
self._mock_path(gcp_iface + "0/forwarded-ips/0", TestNmCloudSetup._ip1)
self._mock_path(gcp_iface + "1/forwarded-ips/", "0\n")
self._mock_path(gcp_iface + "1/forwarded-ips/0", TestNmCloudSetup._ip2)
# Run nm-cloud-setup for the first time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_GCP_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_GCP": "yes",
},
)
nmc.pexp.expect("provider GCP detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("found GCP interfaces: 2")
nmc.pexp.expect(r"GCP interface\[0]: found a requested device with hwaddr")
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# One of the devices has no IPv4 configuration to be modified
nmc.pexp.expect("device has no suitable applied connection. Skip")
# The other one was lacking an address set it up.
nmc.pexp.expect("some changes were applied for provider GCP")
nmc.pexp.expect(pexpect.EOF)
# Run nm-cloud-setup for the second time
nmc = Util.cmd_call_pexpect(
ENV_NM_TEST_CLIENT_CLOUD_SETUP_PATH,
[],
{
"NM_CLOUD_SETUP_GCP_HOST": self.md_url,
"NM_CLOUD_SETUP_LOG": "trace",
"NM_CLOUD_SETUP_GCP": "yes",
},
)
nmc.pexp.expect("provider GCP detected")
nmc.pexp.expect("found interfaces: CC:00:00:00:00:01, CC:00:00:00:00:02")
nmc.pexp.expect("get-config: starting")
nmc.pexp.expect("get-config: success")
nmc.pexp.expect("meta data received")
# No changes this time
nmc.pexp.expect('device needs no update to applied connection "con-eth0"')
nmc.pexp.expect("no changes were applied for provider GCP")
nmc.pexp.expect(pexpect.EOF)
Util.valgrind_check_log(nmc.valgrind_log, "test_gcp")
###############################################################################
def main():
global dbus_session_inited
if len(sys.argv) >= 2 and sys.argv[1] == "--started-with-dbus-session":
dbus_session_inited = True
del sys.argv[1]
if not dbus_session_inited:
# we don't have yet our own dbus-session. Reexec ourself with
# a new dbus-session.
try:
try:
os.execlp(
"dbus-run-session",
"dbus-run-session",
"--",
sys.executable,
__file__,
"--started-with-dbus-session",
*sys.argv[1:]
)
except OSError as e:
if e.errno != errno.ENOENT:
raise
# we have no dbus-run-session in path? Fall-through
# to skip tests gracefully
else:
raise Exception("unknown error during exec")
except Exception as e:
assert False, "Failure to re-exec dbus-run-session: %s" % (str(e))
if not dbus_session_inited:
# we still don't have a D-Bus session. Probably dbus-run-session is not available.
# retry with dbus-launch
if os.system("type dbus-launch 1>/dev/null") == 0:
try:
os.execlp(
"bash",
"bash",
"-e",
"-c",
"eval `dbus-launch --sh-syntax`;\n"
+ 'trap "kill $DBUS_SESSION_BUS_PID" EXIT;\n'
+ "\n"
+ Util.shlex_join(
[
sys.executable,
__file__,
"--started-with-dbus-session",
]
+ sys.argv[1:]
)
+ " \n"
+ "",
)
except Exception as e:
m = str(e)
else:
m = "unknown error"
assert False, "Failure to re-exec to start script with dbus-launch: %s" % (
m
)
r = unittest.main(exit=False)
sys.exit(not r.result.wasSuccessful())
if __name__ == "__main__":
main()