iwd/doc/test-runner.txt

Notes for test-runner usage
***************************

Tool Description
================

test-runner is an automated test execution tool for IWD. It is capable of
creating the emulated environments representing a variety of network topologies
and run the automated tests of IWD functionality.

Software Prerequisites
======================

The test-runner tool requires the following binaries to be present on the host
OS:

	Name:			Tested ver.:
	 1. qemu		2.4.1
	 2. Linux kernel	4.20+
	 3. dbus-daemon		1.11.18
	 4. iproute2
	 5. iw			3.17
	 6. python		3.9
	 7. haveged		no ver. avail.
	 8. hostapd		recent
	 9. <iwd>/tools/hwsim	0.0
	10. <iwd>/src/iwd	0.0
	11. ofono (optional)
	12. phonesim (optional)
	13. wpa_supplicant	2.9
	14. radvd		2.18
	15. dhcpd

Note: The version for hostapd is listed as 'recent'. Older hostapd versions
will work but we are continually adopting new features from hostapd and using
those to test. Its recommended to use a recent release of hostapd.

Note: The test-runner mounts host's file system in readonly mode and executes
the above binaries inside of an emulated environment directly from it.

Note: Running EAP-SIM/AKA/AKA' tests using oFono will require oFono and
phonesim to be installed on the host. This is explained further in the
"Running with oFono and phonesim" section.

In addition, the following Python packages are required:

	Name:			Tested ver.:
	 1. python-prettytable	0.7.2
	 2. python-termcolor	1.1.0
	 3. python-pyroute2	0.5.14
	 4. python-gobject	3.38.0
	 5. python-dbus		1.2.16
	 6. scapy		2.4.5

Building Kernel
===============

test-runner can work both in a fully virtualized environment (QEMU) or inside
a UML (User Mode Linux) instance. Its recommended using UML as its both faster
and more reliable than QEMU for timing specific scenarios. When building the
kernel for UML its important to always specify ARCH=um for any build command.
The default architectures seem to muck with the kernel config that UML needs
and will likely result in a UML binary that won't run under test-runner.

The test-runner tool requires a kernel that is at least build with these
minimal options for a successful boot and execution:

	<arch>_defconfig 			Default kernel configuration

	kvm_guest.config			Default configuration for
						kvm guests (QEMU only)

	<iwd>/tools/test_runner_kernel_config	The test-runner specific
						configuration

These configurations should be installed as .config in the kernel source
directory. To build a x86_64 UML kernel the sequence of commands may look
as follows:

	$ cd linux-X.X.X

	$ make ARCH=um x86_64_defconfig

	$ ARCH=um sh <iwd>/tools/test_runner_kernel_config

	$ make ARCH=um olddefconfig

After that a default kernel with the required options can be built:

	$ make ARCH=um -j$(nproc)

If you do need to build a QEMU kernel you can remove ARCH=um and make the
kvm_guest.config target as well.

After building (for UML) there should be a 'linux' executable at the root of the
repository, this is the UML binary that should be supplied to test-runner with
the --kernel,-k option.

For QEMU it is instead a kernel image located at arch/boot/<arch>/bzImage.

Note: If your host distribution does not contain a regulatory.db you may get an
error similar to this when building the kernel:

No rule to make target '/lib/firmware/regulatory.db'...

To fix this you must download the regulatory.db manually and place it in
/lib/firmware. This can be found here:

https://git.kernel.org/pub/scm/linux/kernel/git/sforshee/wireless-regdb.git

Note: To catch locking related issues the following set of kernel config
options may be useful:

	CONFIG_LOCKDEP_SUPPORT=y
	CONFIG_DEBUG_SPINLOCK=y
	CONFIG_DEBUG_LOCK_ALLOC=y
	CONFIG_PROVE_LOCKING=y
	CONFIG_LOCKDEP=y
	CONFIG_DEBUG_MUTEXES=y

By default the test-runner will search for the kernel image in these locations:

	<iwd>/tools/bzImage

		or

	<iwd>/tools/arch/x86/boot/bzImage

An arbitrary kernel image location can be specified by using '--kernel <path>'
parameter into test-runner.

Running Automated Tests
=======================
Before running any tests, its expected that the folder /var/lib/iwd exists on
the host machine. If not, you will see a mounting error when starting
test-runner.

	mkdir /var/lib/iwd

By default, the automated test configuration directories reside in
'<iwd>/autotests' and have a mandatory prefix of 'test'.

	<iwd>/autotests/test1
			/test2
			...

The test configurations along with test cases in <iwd>/autotests/test*
directories will be discovered and executed by test-runner in sequential
fashion. The following set of commands is sufficient to run the automated
tests shipped with IWD:

	$ cd <iwd>/tools

	$ sudo ./test-runner -k <kernel binary>

One can specify a particular set of test configurations to be executed by using
'-A <dir1,dir2>' parameter. An absolute path is necessary for the test
configuration directories outside of <iwd>/autotests.

The command line may look as follows:

	$ sudo ./test-runner -A test1,test3,/home/test4

Glob matching is also supported when specifying a test list:

	$ sudo ./test-runner -A testWPA*


Creating Test Configurations
============================

A typical test configuration directory may consist of these types of files:

	hw.conf			Defines the network configuration and
				properties of the radios.

	*Test or *Test.py	The set of test cases for IWD functionality
				implemented using Python scripting language.
				These files must have one of the two predefined
				suffixes: 'test' or 'test.py'. The file name
				must also be more descriptive than simply 'test'.
				e.g. 'connection_test' or 'failure_test'. Files
				named 'test' will not be dynamically loaded due
				to python module naming overlap.

	*.conf			A configuration file for an instance of hostapd
	(Defined in hw.conf)	service.

Each configuration directory has exactly one hw.conf, where the number of
Python script files is virtually unlimited. The number of hostapd configuration
files is bounded by the limitation in mac80211_hwsim driver and is set
to 99. (The mac80211_hwsim driver allows to create 100 of simultaneous radios
and one of them is reserved by the test-runner for IWD)

A typical contents of a test configuration directory may look as follows:

	 /test1/hw.conf
		ap1.conf
		ap2.conf
		networkScanTest
		networkConnectTest.py

Note: Any additional files in your test directory will be copied to /tmp inside
the VM. These additional file should be referenced using /tmp/<file> inside
any configuration file for hostapd.

Defining Network
----------------
Network topology along with configuration for the automated test cases is
predetermined in hardware configuration file 'hw.conf'. In addition, it allows
to establish the relationships between the emulated hardware radios and
services that represent various entities of a wireless network.

The following sample hardware configuration file allows to emulate a network
of three nodes. Two of which are access points and the third one represents a
supplicant running IWD:

#~~~~~~~~~~~~~~~~~~~~~~~~~ hw.conf ~~~~~~~~~~~~~~~~~~~~~~~~~
# Lines starting with # are ignored

# 'SETUP' is a manditory configuration group.
[SETUP]
#
# Total number of radios requested per network setup. This includes
# the radios used by APs as well as one for IWD. This field is mandatory and
# has a range of [1, 100].
num_radios=3

# Maximum execution interval per Python script file in seconds. This field is
# optional.
# Default: 20 seconds.
#max_test_exec_interval_sec=5

#
# The following two configuration groups are examples of the radio
# configurations.
#
# This group of settings allows to specify a set of properties for a radio. The
# name of the group represents a radio identifier. These configuration groups
# are optional.
[rad0]

# For each radX group you can specify radio properties:
# Disables the provided interface type(s):
disable_iftype=ap,p2p

# Disables the provided cipher types:
disable_cipher=ccmp,bip

# 'HOSTAPD' configuration group identifies a set of access points (AP) for the
# current network topology. Each key/value pair represents a single AP that is
# emulated by the instance of hostapd service. The key indicates an arbitrary
# radio identifier and value specifies a configuration file for the instance.
# If a radio identifier can not be mapped to a predefined radio configuration
# (identifier is not part of the 'radio_confs' list), then a radio with the
# default configuration is used. This configuration group is optional.
[HOSTAPD]
rad0=ap1.conf
rad1=ap2.conf

# 'radius_server' is a special option used to specify a hostapd instance which
# acts as a radius server only (not an access point). This option does not
# require an additional 'num_radios' to be used.
radius_server=radius.conf
#~~~~~~~~~~~~~~~~~~ end of hw.conf ~~~~~~~~~~~~~~~~~~~~~~~~~

Configuring Access Points
-------------------------
The test-runner tool makes use of the hostapd service to emulate the access
points (AP). Despite the fact that hostapd service comes preinstalled on most
Linux distributions, test-runner uses some of the recently introduced features,
which may only be available from the master tree of the hostapd repository:

	git://w1.fi/srv/git/hostap.git

		OR (its HTTP version)

	http://w1.fi/hostap.git

commit id: 31d3692fe5d56c05753ed4a70c7943979e1d29e7 or above is required.

The sequence of commands to clone, build and install hostapd may look as
follows:

	$ git clone git://w1.fi/srv/git/hostap.git

	$ cd hostap/hostapd

	$ cp <iwd>/doc/hostapd.config .config

	Note: You may need to pre-install: 'gnutls-devel' and 'libgcrypt-devel'
	libraries.

	$ make install

Note: All hostapd build options (CONFIG_*) are stored in doc/hostapd.config.
Any new options which are required for a test should be added there.

Note: If 'make install' fails with the netlink warnings you may need to
install libnl-1.0pre8 (or later).

Note: It is recommended to override the pre-installed version of hostapd with
the newly built one to avoid any confusion. The simplest way to make sure
that the correct version of hostapd is used is to execute the following
command:

	$ hostapd -h

Make sure that '-i' option is available in the list of option.
For more information on hostapd refer to this page:

	https://wireless.wiki.kernel.org/en/users/documentation/hostapd

A full set of the hostapd configurations along with explanation can be
found at:

	https://w1.fi/cgit/hostap/plain/hostapd/hostapd.conf

Note: for EAP-SIM/AKA/AKA' hostapd needs an authenticator running separately.
IWD has a python version of hostapd's "hlrauc.c". This may work out of the box
on your system, but the pycrypto library is required. This can be installed
with python pip3:

	sudo pip3 install pycrypto

Running with oFono and phonesim
-------------------------------
EAP-SIM/AKA/AKA' require SIM card access to perform the authentication
algorithms. This is achieved in test runner using oFono and phonesim. If
either oFono or phonesim are not found when test runner starts, any test
involving oFono will be skipped. Using the option "sim_keys=ofono" in the
hardware config file will tell test runner that the test should use oFono.
There is some setup that needs to be done before test runner will work with
ofono/phonesim

setup ofono:

	$ git clone git://git.kernel.org/pub/scm/network/ofono/ofono.git
	$ cd ofono
	$ ./bootstrap-configure
	$ make install

setup phonesim:

	$ git clone git://git.kernel.org/pub/scm/network/ofono/phonesim.git
	$ cd phonesim
	$ ./bootstrap-configure
	$ make install

Now test runner should pick up both installed binaries.

Note: EAP-SIM/AKA/AKA' can also be tested using the hardcoded SIM plugin. This
just reads hardcoded SIM values from a local file. Tests using this plugin
should not need any additional setup. This plugin is enabled by setting
"sim_keys=<file>" in the hardware config file.

Writing Python Test Scripts
---------------------------
The test-runner tool relies on test cases written in Python script language
to exercise the functionality of IWD. The outcomes of the tests are determined
by the exit status of a process running test and reported on per Python file
bases. The test creators are highly encouraged to use the Python unit test
framework.

For more information on Python unit test framework refer to the following page:

	http://pyunit.sourceforge.net/pyunit.html

Using hardware passthrough
---------------------------

The --hw, -w flag lets you pass in a config file containing USB/PCI adapter
addresses, which can then be used as radios inside the test/VM just as the
virtual mac80211_hwsim radios are used. Note: physical radios cannot be used at
the same time as mac80211_hwsim radios.

Using this option, in some cases, does require some pre-configuration that won't
be described in this document. Specifically, PCI adapters are very involved to
get setup, and require special kernel boot options (on the host), BIOS changes,
and most likely a lot of time to get the system working reliably. Because of
this only USB adapters will be discussed in this document.

If PCI passthrough is something you need, it would be best to follow this guide:

https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF

First, whatever kernel you are using must contain the adapters driver and, if
required, firmware built in. The driver can be built in using 'make menuconfig'
and finding the correct driver for your adapter:

Device Driver -> Network Device Support -> Wireless LAN

Enable [*] the driver(s) you need, save, and exit.

The firmware also needs to be built in, and this will require you finding the
right firmware file (/lib/firmware/) required for your adapter and adding it to
CONFIG_EXTRA_FIRMWARE in your .config file. It is sometimes not very obvious
what firmware you need. I have found that during the kernel boot some adapters
will print out if the firmware was not found, and the name of the firmware file
they expect. If you are having trouble finding the firmware file try continuing
on and see what happens when test-runner starts. Google is also your friend.

Once you have the kernel built you can write your hardware config file for
test-runner. Find the USB bus and device for the adapter:

$ lsusb

You should see your device listed with a 'Bus' and 'Device' number:

$ Bus 001 Device 002: ........

Put these into your passthrough config file under a 'USBAdapters' group:

[USBAdapters]
rad0=1,2

Note: The 'rad#' does not matter at this time. These named keys will not
correspond to rad0, rad1, etc in your test config file. This may change in the
future.

You can then run test-runner using this config file:

./test-runner -k <kernel> --hw passthrough.conf ...

If running specific tests you need to ensure you have enough adapters defined
in the config file, and that the adapters support the features you expect. For
example, some adapters cannot go into AP mode, or use certain channels. If your
test expects these features and the adapters do not support them, the test will
fail in sometimes unexpected ways.

Using the 'shell' feature
---------------------------

The --shell,-s flag allows you to boot into a shell inside the test-runner VM.
If this flag is used the python test will not actually run, only the environment
will be setup. Tis is useful for diagnosing issues with a particular test
quickly without having to modify the python test and restart the VM. The shell
flag is meant to be used in conjunction with --autotest,-A. If no specific test
is specified test-runner will default to the 'shell' test, which is just an
empty test with one adapter.

Using the shell with real hardware (--hw flag) is even more powerful. If your
system is setup for USB/PCI passthrough you can expose physical network cards
in the VM and use them in the shell sandbox. This allows you to try out
different kernels in the VM very quickly (no reboots/swapping out kernels on
the host system).

Here are some examples of --shell usage:

Setup environment for 'testWPA' and boot into shell:
./test-runner -k <kernel> -A testWPA --shell

Boot directly into 'shell' test (sandbox):
./test-runner -k <kernel> --shell

Use hardware passthrough:
./test-runner -k <kernel> --hw <hw.conf> --shell