iwd/tools/test-runner

1322 lines
35 KiB
Plaintext
Raw Normal View History

auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
#!/usr/bin/python3
import argparse
import os
import shutil
import ctypes
import fcntl
import shlex
import sys
import subprocess
import atexit
import time
import unittest
import importlib
import signal
import pyroute2
import multiprocessing
import re
from configparser import ConfigParser
from prettytable import PrettyTable
from termcolor import colored
from glob import glob
from collections import namedtuple
from time import sleep
import dbus.mainloop.glib
from gi.repository import GLib
libc = ctypes.cdll['libc.so.6']
libc.mount.argtypes = (ctypes.c_char_p, ctypes.c_char_p, ctypes.c_char_p, \
ctypes.c_ulong, ctypes.c_char_p)
# Using ctypes to load the libc library is somewhat low level. Because of this
# we need to define our own flags/options for use with mounting.
MS_NOSUID = 2
MS_NODEV = 4
MS_NOEXEC = 8
MS_STRICTATIME = 1 << 24
STDIN_FILENO = 0
TIOCSTTY = 0x540E
config = None
intf_id = 0
rad_id = 0
TEST_MAX_TIMEOUT = 120
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
dbus.mainloop.glib.DBusGMainLoop(set_as_default=True)
def dbg(*s):
'''
Allows prints if stdout has been re-directed
'''
print(*s, file=sys.__stdout__)
def exit_vm():
if config:
for p in config.ctx.processes:
print("Process %s still running!" % p.name)
if config.ctx and config.ctx.results:
print_results(config.ctx.results)
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
os.sync()
RB_AUTOBOOT = 0x01234567
#
# Calling 'reboot' or 'shutdown' from a shell (e.g. os.system('reboot'))
# is not the same the POSIX reboot() and will cause a kernel panic since
# we are the init process. The libc.reboot() allows the VM to exit
# gracefully.
#
libc.reboot(RB_AUTOBOOT)
def path_exists(path):
'''
Searches PATH as well as absolute paths.
'''
if shutil.which(path):
return True
try:
os.stat(path)
except:
return False
return True
def find_binary(list):
'''
Returns a binary from 'list' if its found in PATH or on a
valid absolute path.
'''
for path in list:
if path_exists(path):
return path
return None
def mount(source, target, fs, flags, options=''):
'''
Python wrapper for libc mount()
'''
ret = libc.mount(source.encode(), target.encode(), fs.encode(), flags,
options.encode())
if ret < 0:
errno = ctypes.get_errno()
raise Exception("Could not mount %s (%d)" % (target, errno))
MountInfo = namedtuple('MountInfo', 'fstype target options flags')
mount_table = [
MountInfo('sysfs', '/sys', '', MS_NOSUID|MS_NOEXEC|MS_NODEV),
MountInfo('proc', '/proc', '', MS_NOSUID|MS_NOEXEC|MS_NODEV),
MountInfo('devpts', '/dev/pts', 'mode=0620', MS_NOSUID|MS_NOEXEC),
MountInfo('tmpfs', '/dev/shm', 'mode=1777', MS_NOSUID|MS_NODEV|MS_STRICTATIME),
MountInfo('tmpfs', '/run', 'mode=0755', MS_NOSUID|MS_NODEV|MS_STRICTATIME),
MountInfo('tmpfs', '/tmp', '', 0),
MountInfo('tmpfs', '/usr/share/dbus-1', 'mode=0755', MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME),
MountInfo('debugfs', '/sys/kernel/debug', '', 0)
]
DevInfo = namedtuple('DevInfo', 'target linkpath')
dev_table = [
DevInfo('/proc/self/fd', '/dev/fd'),
DevInfo('/proc/self/fd/0', '/dev/stdin'),
DevInfo('/proc/self/fd/1', '/dev/stdout'),
DevInfo('/proc/self/fd/2', '/dev/stderr')
]
dbus_config = '''
<!DOCTYPE busconfig PUBLIC \
"-//freedesktop//DTD D-Bus Bus Configuration 1.0//EN" \
"http://www.freedesktop.org/standards/dbus/1.0/\
busconfig.dtd\">
<busconfig>
<type>system</type>
<listen>unix:path=/run/dbus/system_bus_socket</listen>
<limit name=\"reply_timeout\">2147483647</limit>
<auth>ANONYMOUS</auth>
<allow_anonymous/>
<policy context=\"default\">
<allow user=\"*\"/>
<allow own=\"*\"/>
<allow send_type=\"method_call\"/>
<allow send_type=\"signal\"/>
<allow send_type=\"method_return\"/>
<allow send_type=\"error\"/>
<allow receive_type=\"method_call\"/>
<allow receive_type=\"signal\"/>
<allow receive_type=\"method_return\"/>
<allow receive_type=\"error\"/>
<allow send_destination=\"*\" eavesdrop=\"true\"/>
<allow eavesdrop=\"true\"/>
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
</policy>
</busconfig>
'''
class Process:
'''
Start a process. If 'wait' is True the constructor will start
the process and wait for it to exit. No PID is tracked in this
case. If 'multi_test' is True this indicates the process is
run over the entire test run and will not be killed after each
test exits.
'''
def __init__(self, args, wait=False, multi_test=False, env=None, ctx=None, check=False,
outfile=None):
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
self.args = args
self.wait = wait
self.name = args[0]
self.multi_test = multi_test
self.stdout = subprocess.PIPE
self.stderr = subprocess.PIPE
self.ret = None
self.ctx = ctx
if ctx:
set_stdout = False
if ctx.is_verbose(args[0]):
dbg("Verbose on for %s" % args[0])
set_stdout = True
if os.path.basename(args[0]) == ctx.args.gdb:
self.args = ['gdb', '--args']
self.args.extend(args)
set_stdout = True
if outfile:
set_stdout = True
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
# Anything labeled as multi_test isn't important to
# log. These are processes such as dbus-daemon and
# haveged.
if set_stdout:
if ctx.args.log:
test = os.path.basename(os.getcwd())
test_dir = '%s/%s' % (ctx.args.log, test)
if not path_exists(test_dir):
os.mkdir(test_dir)
os.chown(test_dir, int(ctx.args.log_uid), \
int(ctx.args.log_gid))
self.stdout = open('%s/%s' % (test_dir, args[0]), 'w')
self.stderr = open('%s/%s' % (test_dir, args[0]), 'w')
elif outfile:
self.stdout = open(outfile, 'w')
self.stderr = open(outfile, 'w')
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
else:
self.stdout = sys.__stdout__
self.stderr = sys.__stderr__
if not wait and not check:
self.pid = subprocess.Popen(self.args, stdout=self.stdout, \
stderr=self.stderr, env=env, \
cwd=os.getcwd())
print("Starting process {}".format(self.pid.args))
else:
self.ret = subprocess.call(self.args, stdout=self.stdout, \
stderr=self.stderr)
print("%s returned %d" % (args[0], self.ret))
if check and self.ret != 0:
raise subprocess.CalledProcessError(returncode=self.ret, cmd=self.args)
def __del__(self):
print("Del process %s" % self.args)
if self.ctx and self.ctx.args.log:
self.stdout.close()
self.stderr.close()
def kill(self, force=False):
print("Killing process %s" % self.args)
if force:
os.kill(self.pid.pid, signal.SIGKILL)
else:
self.pid.kill()
self.pid.wait(timeout=15)
def wait_for_socket(self, socket, wait):
waited = 0
while not os.path.exists(socket):
sleep(0.5)
waited += 0.5
if waited > wait:
raise Exception("Timed out waiting for socket")
class Interface:
def __init__(self, name, config):
self.name = name
self.ctrl_interface = '/var/run/hostapd/' + name
self.config = config
def __del__(self):
Process(['iw', 'dev', self.name, 'del'], True)
def set_interface_state(self, state):
Process(['ifconfig', self.name, state], True)
class Radio:
def __init__(self, name):
self.name = name
# hostapd will reset this if this radio is used by it
self.use = 'iwd'
self.interface = None
def __del__(self):
print("Removing radio %s" % self.name)
self.interface = None
def create_interface(self, hapd):
global intf_id
ifname = 'wln%s' % intf_id
intf_id += 1
self.interface = Interface(ifname, hapd.config)
# IWD does not use interfaces in test-runner so any created
# interface is assumed to be used by hostapd.
self.use = 'hostapd'
Process(['iw', 'phy', self.name, 'interface', 'add', ifname,
'type', 'managed'], True)
return self.interface
def __str__(self):
ret = self.name + ':\n'
ret += '\tUsed By: %s ' % self.use
if self.interface:
ret += '(%s)' % self.interface.name
ret += '\n'
return ret
class VirtualRadio(Radio):
'''
A subclass of 'Radio' specific to mac80211_hwsim radios.
TODO: Using D-Bus to create and destroy radios is more desireable
than the command line.
'''
def __init__(self, name, config=None):
global rad_id
super().__init__(name)
self.disable_cipher = None
self.disable_iftype = None
args = ['hwsim', '--create', '--name', self.name, '--nointerface', '--p2p']
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
if config:
self.disable_iftype = config.get('iftype_disable', False)
if self.disable_iftype:
args.append('--iftype-disable')
args.append(self.disable_iftype)
self.disable_cipher = config.get('cipher_disable', False)
if self.disable_cipher:
args.append('--cipher-disable')
args.append(self.disable_cipher)
Process(args, wait=True)
self.id = rad_id
rad_id += 1
def __del__(self):
super().__del__()
Process(['hwsim', '--destroy=%s' % self.id])
def __str__(self):
ret = super().__str__()
if self.disable_iftype:
ret += '\tDisabled interface types: %s\n' % self.disable_iftype
if self.disable_cipher:
ret += '\tDisabled ciphers: %s\n' % self.disable_cipher
ret += '\n'
return ret
class HostapdInstance:
'''
A single instance of hostapd. In reality all hostapd instances
are started as a single process. This class just makes things
convenient for communicating with one of the hostapd APs.
'''
def __init__(self, config, radio):
self.radio = radio
self.config = config
self.intf = radio.create_interface(self)
self.intf.set_interface_state('up')
def __del__(self):
print("Removing HostapdInstance %s" % self.config)
self.intf.set_interface_state('down')
self.radio = None
self.intf = None
def __str__(self):
ret = 'Hostapd (%s)\n' % self.intf.name
ret += '\tConfig: %s\n' % self.config
return ret
class Hostapd:
'''
A set of running hostapd instances. This is really just a single
process since hostapd can be started with multiple config files.
'''
def __init__(self, ctx, radios, configs, radius):
if len(configs) != len(radios):
raise Exception("Config (%d) and radio (%d) list length not equal" % \
(len(configs), len(radios)))
print("Initializing hostapd instances")
ctx.start_process(['ip', 'link', 'set', 'eth0', 'up'], wait=True)
ctx.start_process(['ip', 'link', 'set', 'eth1', 'up'], wait=True)
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
self.global_ctrl_iface = '/var/run/hostapd/ctrl'
self.instances = [HostapdInstance(c, r) for c, r in zip(configs, radios)]
ifaces = [rad.interface.name for rad in radios]
ifaces = ','.join(ifaces)
args = ['hostapd', '-g', self.global_ctrl_iface]
if ifaces:
args.extend(['-i', ifaces])
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
#
# Config files should already be present in /tmp. This appends
# ctrl_interface and does any variable replacement. Currently
# this is just any $ifaceN occurrences.
#
for c in configs:
full_path = '/tmp/%s' % c
args.append(full_path)
self._rewrite_config(full_path)
if radius:
args.append(radius)
if ctx.is_verbose('hostapd'):
args.append('-d')
self.process = ctx.start_process(args)
self.process.wait_for_socket(self.global_ctrl_iface, 30)
def _rewrite_config(self, config):
'''
Replaces any $ifaceN values with the correct interface
names as well as appends the ctrl_interface path to
the config file.
'''
with open(config, 'r+') as f:
data = f.read()
to_replace = []
for match in re.finditer(r'\$iface[0-9]+', data):
tag = data[match.start():match.end()]
idx = tag.split('iface')[1]
to_replace.append((tag, self.instances[int(idx)].intf.name))
for r in to_replace:
data = data.replace(r[0], r[1], 1)
data += '\nctrl_interface=/var/run/hostapd\n'
f.write(data)
def __getitem__(self, config):
if not config:
return self.instances[0]
for hapd in self.instances:
if hapd.config == config:
return hapd
return None
def __del__(self):
print("Removing Hostapd")
try:
os.remove(self.global_ctrl_iface)
except:
dbg("Failed to remove %s" % self.global_ctrl_iface)
self.instances = None
self.process.kill()
class TestContext:
'''
Contains all information for a given set of tests being run
such as processes, radios, interfaces and test results.
'''
def __init__(self, args):
self.processes = []
self.args = args
self.hw_config = None
self.hostapd = None
self.cur_radio_id = 0
self.cur_iface_id = 0
self.radios = []
self.loopback_started = False
self.iwd_extra_options = None
self.results = {}
self.mainloop = GLib.MainLoop()
def start_process(self, args, **kwargs):
p = Process(args, ctx=self, **kwargs)
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
if not kwargs.get('wait', False):
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
self.processes.append(p)
return p
def start_dbus(self):
with open('/usr/share/dbus-1/system.conf', 'w+') as f:
f.write(dbus_config)
os.mkdir('/run/dbus', 755)
self.start_process(['dbus-daemon', '--system', '--nosyslog'], multi_test=True)
def start_dbus_monitor(self):
if not self.is_verbose('dbus-monitor'):
return
self.start_process(['dbus-monitor', '--system'])
def start_haveged(self):
self.start_process(['haveged'], multi_test=True)
def create_radios(self):
setup = self.hw_config['SETUP']
nradios = int(setup['num_radios'])
for i in range(nradios):
name = 'rad%u' % i
# Get any [radX] sections. These are for configuring
# any special radios. This no longer requires a
# radio_conf list, we just assume radios start rad0
# and increment.
rad_config = None
if self.hw_config.has_section(name):
rad_config = self.hw_config[name]
self.radios.append(VirtualRadio(name, rad_config))
self.cur_radio_id += 1
if self.hw_config['SETUP'].get('hwsim_medium', 'yes') in ['yes', '1', 'true']:
# register hwsim as medium
self.start_process(['hwsim'])
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
def discover_radios(self):
phys = []
iw = pyroute2.iwutil.IW()
attrs = [phy['attrs'] for phy in iw.list_wiphy()]
for attr in attrs:
for key, value in attr:
if key == 'NL80211_ATTR_WIPHY_NAME':
if value not in phys:
phys.append(value)
break
print('Discovered radios: %s' % str(phys))
self.radios = [Radio(name) for name in phys]
def start_radios(self):
reg_domain = self.hw_config['SETUP'].get('reg_domain', None)
if reg_domain:
Process(['iw', 'reg', 'set', reg_domain], True)
if self.args.hw:
self.discover_radios()
else:
self.create_radios()
def start_iwd(self, config_dir = '/tmp'):
args = []
iwd_radios = ','.join([r.name for r in self.radios if r.use == 'iwd'])
if self.args.valgrind:
args.extend(['valgrind', '--leak-check=full', '--log-file=%s' % \
'/tmp/valgrind.log'])
args.extend(['iwd', '-p', iwd_radios])
if self.is_verbose(args[0]):
args.append('-d')
if self.iwd_extra_options:
args.append(self.iwd_extra_options)
env = os.environ.copy()
env['CONFIGURATION_DIRECTORY'] = config_dir
env['STATE_DIRECTORY'] = '/tmp/iwd'
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
if self.is_verbose('iwd-dhcp'):
env['IWD_DHCP_DEBUG'] = '1'
if self.is_verbose('iwd-tls'):
env['IWD_TLS_DEBUG'] = '1'
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
pid = self.start_process(args, env=env)
return pid
def start_hostapd(self):
if not 'HOSTAPD' in self.hw_config:
return
settings = self.hw_config['HOSTAPD']
if self.args.hw:
# Just grab the first N radios. It gets rather
# complicated trying to map radX radios specified in
# hw.conf so any passed through physical adapters are
# just given to hostapd/IWD as they appear during
# discovery.
#
# TODO: It may be desireable to map PCI/USB adapters to
# specific radX radios specified in the config but
# there are really 2 separate use cases here.
# 1. You want to test a *specific* radio with IWD
# or hostapd. For this you would want radX
# to map to a specific radio
# 2. You have many adapters in use to run multiple
# tests. In this case you would not care what
# was using each radio, just that there was
# enough to run all tests.
nradios = 0
for k, _ in settings.items():
if k == 'radius_server':
continue
nradios += 1
hapd_radios = self.radios[:nradios]
else:
hapd_radios = [rad for rad in self.radios if rad.name in settings]
hapd_configs = [conf for rad, conf in settings.items() if rad != 'radius_server']
radius_config = settings.get('radius_server', None)
self.hostapd = Hostapd(self, hapd_radios, hapd_configs, radius_config)
def start_ofono(self):
sim_keys = self.hw_config['SETUP'].get('sim_keys', None)
if not sim_keys:
print("Ofono not requred")
return
elif sim_keys != 'ofono':
os.environ['IWD_SIM_KEYS'] = sim_keys
return
if not find_binary(['ofonod']) or not find_binary(['phonesim']):
print("Ofono or Phonesim not found, skipping test")
return
Process(['ifconfig', 'lo', 'up'], wait=True)
os.environ['OFONO_PHONESIM_CONFIG'] = '/tmp/phonesim.conf'
phonesim_args = ['phonesim', '-p', '12345', '/usr/share/phonesim/default.xml']
self.start_process(phonesim_args)
#
# TODO:
# Is there something to wait for? Without this phonesim rejects
# connections on all but the fist test.
#
time.sleep(3)
ofono_args = ['ofonod', '-n', '--plugin=atmodem,phonesim']
if self.is_verbose('ofonod'):
ofono_args.append('-d')
self.start_process(ofono_args)
print("Ofono started")
def is_verbose(self, process):
process = os.path.basename(process)
if self.args is None:
return False
# every process is verbose when logging is enabled
if self.args.log:
return True
if process in self.args.verbose:
return True
# Special case here to enable verbose output with valgrind running
if process == 'valgrind' and 'iwd' in self.args.verbose:
return True
# Handle any glob matches
for item in self.args.verbose:
if process in glob(item):
return True
return False
def stop_process(self, p, force=False):
p.kill(force)
self.processes.remove(p)
def stop_test_processes(self):
self.radios = []
self.hostapd = None
self.iwd_extra_options = None
for p in [p for p in self.processes if p.multi_test is False]:
print("Killing process %s" % p.name)
self.stop_process(p)
def is_process_running(self, process):
for p in self.processes:
if p.name == process:
return True
return False
def __str__(self):
ret = 'Arguments:\n'
for arg in vars(self.args):
ret += '\t --%s %s\n' % (arg, str(getattr(self.args, arg)))
ret += 'Processes:\n'
for p in self.processes:
ret += '\t%s\n' % str(p.args)
ret += 'Radios:\n'
if len(self.radios) > 0:
for r in self.radios:
ret += '\t%s\n' % str(r)
else:
ret += '\tNo Radios\n'
ret += 'Hostapd:\n'
if self.hostapd:
for h in self.hostapd.instances:
ret += '\t%s\n' % str(h)
else:
ret += '\tNo Hostapd instances\n'
return ret
def prepare_sandbox():
print('Preparing sandbox')
for entry in mount_table:
try:
os.lstat(entry.target)
except:
os.mkdir(entry.target, 755)
mount(entry.fstype, entry.target, entry.fstype, entry.flags,
entry.options)
for entry in dev_table:
os.symlink(entry.target, entry.linkpath)
os.mkdir('/tmp/iwd')
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
os.setsid()
fcntl.ioctl(STDIN_FILENO, TIOCSTTY, 1)
def build_unit_list(args):
'''
Build list of unit tests based on passed arguments. This first
checks for literal names provided in the arguments, then if
no matches were found, checks for a glob match.
'''
tests = []
test_root = args.testhome + '/unit'
for unit in args.unit_tests.split(','):
path = '%s/%s' % (test_root, unit)
if os.access(unit, os.X_OK):
tests.append(unit)
elif os.access(path, os.X_OK):
tests.append(path)
else:
# Full list or glob, first build up valid list of tests
matches = glob(path)
if matches == []:
raise Exception("Could not find test %s" % unit)
matches = [exe for exe in matches if os.access(exe, os.X_OK)]
tests.extend(matches)
return sorted(tests)
def build_test_list(args):
'''
Build list of auto test directories based on passed arguments.
First check for absolute paths, then look in <iwd>/autotests,
then glob match.
'''
tests = []
test_root = args.testhome + '/autotests'
# Run all tests
if not args.auto_tests:
# --shell with no tests implies 'shell' test
if args.shell:
return [test_root + '/shell']
tests = os.listdir(test_root)
# Pair down any non-tests and append full path
tests = [test_root + '/' + t for t in tests if t.startswith('test')]
else:
print("Generating partial test list")
for t in args.auto_tests.split(','):
path = '%s/%s' % (test_root, t)
# Full test path specified
if os.path.exists(t):
tests.append(t)
elif os.path.exists(path):
tests.append(path)
else:
matches = glob(path)
if matches == []:
raise Exception("Could not find test %s" % t)
tests.extend(matches)
return sorted(tests)
SimpleResult = namedtuple('SimpleResult', 'run failures errors skipped time')
def start_test(ctx, subtests, rqueue):
'''
Run an individual test. 'subtests' are parsed prior to calling
but these effectively make up a single test. 'rqueue' is the
results queue which is required since this is using
multiprocessing.
'''
suite = unittest.TestSuite()
#
# Iterate through each individual python test.
#
for s in subtests:
loader = unittest.TestLoader()
subtest = importlib.import_module(os.path.splitext(s)[0])
suite.addTests(loader.loadTestsFromModule(subtest))
# Prevents future test modules with the same name (e.g.
# connection_test.py) from being loaded from the cache
sys.modules.pop(subtest.__name__)
start = time.time()
runner = unittest.TextTestRunner()
result = runner.run(suite)
#
# The multiprocessing queue is picky with what objects it will serialize
# and send between processes. Because of this we put the important bits
# of the result into our own 'SimpleResult' tuple.
#
sresult = SimpleResult(run=result.testsRun, failures=len(result.failures),
errors=len(result.errors), skipped=len(result.skipped),
time=time.time() - start)
rqueue.put(sresult)
# This may not be required since we are manually popping sys.modules
importlib.invalidate_caches()
def pre_test(ctx, test):
'''
Copy test files, start processes, and any other pre test work.
'''
os.chdir(test)
dbg("Starting test %s" % test)
if not os.path.exists(test + '/hw.conf'):
print("No hw.conf found for %s" % test)
exit()
ctx.hw_config = ConfigParser()
ctx.hw_config.read(test + '/hw.conf')
#
# We have two types of test files: tests and everything else. Rather
# than require each test to specify the files needing to be copied to
# /tmp (previously 'tmpfs_extra_stuff'), we just copy everything which
# isn't a test. There is really no reason not to do this as any file
# present in a test directory should be needed by the test.
#
# All files
files = os.listdir(test)
# Tests (starts or ends with 'test')
subtests = [f for f in files if f.startswith('test') or \
os.path.splitext(f)[0].endswith('test')]
# Everything else (except .py files)
to_copy = [f for f in list(set(files) - set(subtests)) if not f.endswith('.py')]
for f in to_copy:
if os.path.isdir(f):
shutil.copytree(f, '/tmp/' + f)
else:
shutil.copy(f, '/tmp')
ctx.start_dbus_monitor()
ctx.start_radios()
ctx.start_hostapd()
ctx.start_ofono()
if ctx.args.log:
ctx.start_process(['iwmon'])
elif ctx.args.monitor:
ctx.start_process(['iwmon'], outfile=ctx.args.monitor)
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
if ctx.hw_config.has_option('SETUP', 'start_iwd'):
start = ctx.hw_config.getboolean('SETUP', 'start_iwd')
else:
start = True
if start:
ctx.start_iwd()
else:
print("Not starting IWD from test-runner")
print(ctx)
sys.path.insert(1, test)
return (to_copy, sorted(subtests))
def post_test(ctx, to_copy):
'''
Remove copied files, and stop test processes.
'''
for f in to_copy:
if os.path.isdir('/tmp/' + f):
shutil.rmtree('/tmp/' + f)
else:
os.remove('/tmp/' + f)
Process(['ifconfig', 'lo', 'down'], wait=True)
ctx.stop_test_processes()
if ctx.args.valgrind:
with open('/tmp/valgrind.log', 'r') as f:
dbg(f.read())
dbg("\n")
def print_results(results):
table = PrettyTable(['Test', colored('Passed', 'green'), colored('Failed', 'red'), \
colored('Skipped', 'cyan'), colored('Time', 'yellow')])
total_pass = 0
total_fail = 0
total_skip = 0
total_time = 0
for test, result in results.items():
if result.time != TEST_MAX_TIMEOUT:
failed = result.failures + result.errors
passed = result.run - failed
total_pass += passed
total_fail += failed
total_skip += result.skipped
else:
failed = "Timed out"
passed = "Timed out"
total_time += result.time
time = '%.2f' % result.time
table.add_row([test, colored(passed, 'green'), colored(failed, 'red'), \
colored(result.skipped, 'cyan'), colored(time, 'yellow')])
total_time = '%.2f' % total_time
table.add_row(['Total', colored(total_pass, 'green'), colored(total_fail, 'red'), \
colored(total_skip, 'cyan'), colored(total_time, 'yellow')])
dbg(table)
def run_auto_tests(ctx, args):
tests = build_test_list(args)
ctx.start_dbus()
ctx.start_haveged()
# Copy autotests/misc/{certs,secrets,phonesim} so any test can refer to them
shutil.copytree(args.testhome + '/autotests/misc/certs', '/tmp/certs')
shutil.copytree(args.testhome + '/autotests/misc/secrets', '/tmp/secrets')
shutil.copy(args.testhome + '/autotests/misc/phonesim/phonesim.conf', '/tmp')
if args.shell:
#
# Shell really isn't meant to be used with multiple tests. If
# a set of tests was passed in just start out in the first.
#
os.chdir(tests[0])
os.system('/bin/bash')
exit()
for test in tests:
try:
copied, subtests = pre_test(ctx, test)
rqueue = multiprocessing.Queue()
p = multiprocessing.Process(target=start_test, args=(ctx, subtests, rqueue))
p.start()
# Rather than time each subtest we just time the total but
# mutiply the default time by the number of tests being run.
p.join(TEST_MAX_TIMEOUT * len(subtests))
if p.is_alive():
# Timeout
p.terminate()
ctx.results[os.path.basename(test)] = SimpleResult(run=0,
failures=0, errors=0,
skipped=0, time=TEST_MAX_TIMEOUT)
else:
ctx.results[os.path.basename(test)] = rqueue.get()
except Exception as ex:
print(ex)
print("Uncaught exception thrown for %s" % test)
ctx.results[os.path.basename(test)] = SimpleResult(run=-1, failures=-1,
errors=-1, skipped=-1, time=-1)
finally:
post_test(ctx, copied)
# Write out kernel log
if ctx.args.log:
Process(["dmesg"], ctx=ctx, wait=True)
def run_unit_tests(ctx, args):
os.chdir(args.testhome + '/unit')
units = build_unit_list(args)
for u in units:
if ctx.start_process([u], wait=True).ret != 0:
dbg("Unit test %s failed" % os.path.basename(u))
else:
dbg("Unit test %s passed" % os.path.basename(u))
def run_tests():
global config
with open('/proc/cmdline', 'r') as f:
cmdline = f.read()
start = cmdline.find('--testhome')
options = shlex.split(cmdline[start:])
parser = argparse.ArgumentParser()
parser.add_argument('--testhome')
parser.add_argument('--auto_tests')
parser.add_argument('--unit_tests')
parser.add_argument('--verbose', default=[])
parser.add_argument('--debug')
parser.add_argument('--path')
parser.add_argument('--valgrind')
parser.add_argument('--gdb')
parser.add_argument('--shell')
parser.add_argument('--log')
parser.add_argument('--log-gid')
parser.add_argument('--log-uid')
parser.add_argument('--hw')
parser.add_argument('--monitor')
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
args = parser.parse_args(options)
#
# This prevents any print() calls in this script from printing unless
# --debug is passed. For an 'always print' option use dbg()
#
if not args.debug:
sys.stdout = open(os.devnull, 'w')
os.environ['PATH'] = '%s/src' % args.testhome
os.environ['PATH'] += ':%s/tools' % args.testhome
os.environ['PATH'] += ':%s/client' % args.testhome
os.environ['PATH'] += ':%s/wired' % args.testhome
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
os.environ['PATH'] += ':' + args.path
sys.path.append(args.testhome + '/autotests/util')
#
# This allows all autotest utils (iwd/hostapd/etc) to access the
# TestContext. Any other module or script (in the same interpreter) can
# simply import config.ctx and access all live test information,
# start/stop processes, see active radios etc.
#
config = importlib.import_module('config')
config.ctx = TestContext(args)
if args.log:
mount('logdir', args.log, '9p', 0, 'trans=virtio,version=9p2000.L')
elif args.monitor:
parent = os.path.abspath(os.path.join(args.monitor, os.pardir))
mount('mondir', parent, '9p', 0, 'trans=virtio,version=9p2000.L')
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
if config.ctx.args.unit_tests is None:
run_auto_tests(config.ctx, args)
else:
run_unit_tests(config.ctx, args)
class Main:
def __init__(self):
self.parser = argparse.ArgumentParser(
description='IWD Test Runner')
self.parser.add_argument('--qemu', '-q',
metavar='<QEMU binary>', type=str,
help='QEMU binary to use',
dest='qemu')
self.parser.add_argument('--kernel', '-k', metavar='<kernel>',
type=str,
help='Path to kernel image',
dest='kernel')
self.parser.add_argument('--verbose', '-v', metavar='<list>',
type=str,
help='Comma separated list of applications',
dest='verbose',
default=[])
self.parser.add_argument('--debug', '-d',
action='store_true',
help='Enable test-runner debugging',
dest='debug')
self.parser.add_argument('--shell', '-s', action='store_true',
help='Boot into shell', dest='shell')
self.parser.add_argument('--log', '-l', type=str,
help='Directory for log files')
self.parser.add_argument('--hw', '-w', type=str, nargs=1,
help='Use physical adapters for tests (passthrough)')
self.parser.add_argument('--monitor', '-m', type=str,
help='Enables iwmon output to file')
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
# Prevent --autotest/--unittest from being used together
auto_unit_group = self.parser.add_mutually_exclusive_group()
auto_unit_group.add_argument('--auto-tests', '-A',
metavar='<tests>', type=str, nargs=1,
help='List of tests to run',
default=None,
dest='auto_tests')
auto_unit_group.add_argument('--unit-tests', '-U',
metavar='<tests>', type=str, nargs='?',
const='*',
help='List of unit tests to run',
dest='unit_tests')
# Prevent --valgrind/--gdb from being used together
valgrind_gdb_group = self.parser.add_mutually_exclusive_group()
valgrind_gdb_group.add_argument('--gdb', '-g', metavar='<exec>',
type=str, nargs=1,
help='Run gdb on specified executable',
dest='gdb')
valgrind_gdb_group.add_argument('--valgrind', '-V', action='store_true',
help='Run valgrind on IWD', dest='valgrind')
self.args = self.parser.parse_args()
if self.args.log and self.args.unit_tests:
dbg("Cannot use --log with --unit-tests")
quit()
def start(self):
usb_adapters = None
qemu_table = [
'qemu-system-x86_64',
'/usr/bin/qemu-system-x86_64'
]
kernel_table = [
'bzImage',
'arch/x86/boot/bzImage',
'vmlinux',
'arch/x86/boot/vmlinux'
]
if self.args.qemu is None:
qemu_binary = find_binary(qemu_table)
else:
if path_exists(self.args.qemu):
qemu_binary = self.args.qemu
else:
print("QEMU binary %s does not exist" % \
self.args.qemu)
quit()
if self.args.kernel is None:
kernel_binary = find_binary(kernel_table)
else:
if path_exists(self.args.kernel):
kernel_binary = self.args.kernel
else:
print("Kernel image %s does not exist" % \
self.args.kernel)
quit()
if self.args.hw:
hw_conf = ConfigParser()
hw_conf.read(self.args.hw)
# TODO: Parse PCI adapters
if hw_conf.has_section('USBAdapters'):
# The actual key name of the adapter
# doesn't matter since all we need is the
# bus/address. This gets named by the kernel
# anyways once in the VM.
usb_adapters = [v for v in hw_conf['USBAdapters'].values()]
#
# Additional arguments not provided to test-runner which are
# needed once booted into the kernel.
#
options = 'init=%s' % os.path.realpath(sys.argv[0])
# Support running from top level as well as tools
if os.getcwd().endswith('tools'):
options += ' --testhome %s/../' % os.getcwd()
else:
options += ' --testhome %s' % os.getcwd()
options += ' --path "%s"' % os.environ['PATH']
if self.args.auto_tests:
options += ' --auto_tests %s' % ','.join(self.args.auto_tests)
if self.args.log:
if os.environ.get('SUDO_GID', None) is None:
print("--log can only be used as root user")
quit()
self.args.log = os.path.abspath(self.args.log)
uid = int(os.environ['SUDO_UID'])
gid = int(os.environ['SUDO_GID'])
if not path_exists(self.args.log):
os.mkdir(self.args.log)
os.chown(self.args.log, uid, gid)
options += ' --log-gid %u' % gid
options += ' --log-uid %u' % uid
if self.args.monitor:
self.args.monitor = os.path.abspath(self.args.monitor)
mon_parent_dir = os.path.abspath(os.path.join(self.args.monitor, os.pardir))
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
denylist = [
'auto_tests',
'qemu',
'kernel'
]
#
# This passes through most of the command line options to
# the kernel command line. Some are not relevant (e.g. qemu)
# so similar options are added in the denylist above. This excludes
# any unset options which are assumed to be None or False. This
# is done so default arguments can be filled once in the VM. If
# we pass and basic types (None, False etc.) they are turned into
# a string representation ('None', 'False', etc.) which is not
# desirable.
#
for arg in vars(self.args):
if arg in denylist or getattr(self.args, arg) in [None, False, []]:
continue
options += ' --%s %s' % (arg, str(getattr(self.args, arg)))
kern_log = "ignore_loglevel" if "kernel" in self.args.verbose else "quiet"
qemu_cmdline = [
qemu_binary,
'-machine', 'type=q35,accel=kvm:tcg',
'-nodefaults', '-no-user-config', '-monitor', 'none',
'-display', 'none', '-m', '192M', '-nographic', '-vga',
'none', '-no-acpi', '-no-hpet',
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
'-no-reboot', '-fsdev',
'local,id=fsdev-root,path=/,readonly,security_model=none,multidevs=remap',
'-device',
'virtio-9p-pci,fsdev=fsdev-root,mount_tag=/dev/root',
'-chardev', 'stdio,id=chardev-serial0,signal=off',
'-device', 'pci-serial,chardev=chardev-serial0',
'-device', 'virtio-rng-pci',
'-kernel',
kernel_binary,
'-append',
'console=ttyS0,115200n8 earlyprintk=serial \
rootfstype=9p root=/dev/root \
rootflags=trans=virtio,version=9p2000.u \
acpi=off pci=noacpi %s ro \
mac80211_hwsim.radios=0 %s' % (kern_log, options),
'-cpu', 'host'
]
# Add two ethernet devices for testing EAD
qemu_cmdline.extend([
'-net', 'nic,model=virtio',
'-net', 'nic,model=virtio',
'-net', 'user'
])
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
if usb_adapters:
for bus, addr in [s.split(',') for s in usb_adapters]:
qemu_cmdline.extend(['-usb',
'-device',
'usb-host,hostbus=%s,hostaddr=%s' % \
(bus, addr)])
if self.args.log:
#
# Creates a virtfs device that can be mounted. This mount
# will point back to the provided log directory and is
# writable unlike the rest of the mounted file system.
#
qemu_cmdline.extend([
'-virtfs',
'local,path=%s,mount_tag=logdir,security_model=passthrough,id=logdir' \
% self.args.log
])
if self.args.monitor:
qemu_cmdline.extend([
'-virtfs',
'local,path=%s,mount_tag=mondir,security_model=passthrough,id=mondir' \
% mon_parent_dir
])
auto-t: introduce pure python test-runner re-write This patch completely re-writes test-runner in Python. This was done because the existing C test-runner had some clunky work arounds and maintaining or adding new features was starting to become a huge pain. There were a few aspects of test-runner which continually had to be dealt with when adding any new functionality: * Argument parsing: Adding new arguments to test-runner wasn't so bad, but if you wanted those arguments passed into the VM it became a huge pain. Arguments needed to be parsed, then re-formatted into the qemu command line, then re-parsed in a special order (backwards) once in the VM. The burden for adding new arguments was quite high so it was avoided (at least by me) at all costs. * The separation between C and Python: The tests are all written in python, but the executables, radios, and interfaces were all created from C. The way we solved this was by encoding the require info as environment variables, then parsing those from Python. It worked, but it was, again, a huge pain. * Process management: It started with all processes being launched from C, but eventually tests required the ability to start IWD, or kill hostapd ungracefully in order to test certain functionality. Since the processes were tracked in C, Python had no way of signalling that it killed a process and when it started one C had no idea. This was mitigated (basically by killall), but it was no where close to an elegant solution. Re-writing test-runner in python solves all these problems and will be much easier to maintain. * Argument parsing: Now all arguments are forwarded automatically to the VM. The ArgParse library takes care of parsing and each argument is stored in a dictionary. * Separation between C and Python: No more C, so no more separation. * Process management: Python will now manage all processes. This allows a test to kill, restart, or start a new process and not have to remember the PID or to kill it after the test. There are a few more important aspects of the python implementation that should now be considered when writing new tests: * The IWD constructor now has different default arugments. IWD will always be started unless specified and the configuration directory will always be /tmp * Any non *.py file in the test directory will be copied to /tmp. This avoids the need for 'tmpfs_extra_stuff' completely. * ctrl_interface will automatically be appended to every hostapd config. There is no need to include this in a config file from now on. * Test cleanup is extremely important. All tests get run in the same interpreter now and the tests themselves are actually loaded as python modules. This means e.g. if you somehow kept a reference to IWD() any subsequent tests would not start since IWD is still running. * For debugging, the test context can be printed which shows running processes, radios, and interfaces. Three non-native python modules were used: PrettyTable, colored, and pyroute2 $ pip3 install prettytable $ pip3 install termcolor $ pip3 install pyroute2
2020-09-11 01:12:23 +02:00
os.execlp(qemu_cmdline[0], *qemu_cmdline)
if __name__ == '__main__':
if os.getpid() == 1 and os.getppid() == 0:
atexit.register(exit_vm)
prepare_sandbox()
run_tests()
exit()
main = Main()
main.start()