Functional testing with Python

The tests/functional directory hosts functional tests written in Python. They are usually higher level tests, and may interact with external resources and with various guest operating systems. The functional tests have initially evolved from the Avocado tests, so there is a lot of similarity to those tests here (see Integration testing with Avocado for details about the Avocado tests).

The tests should be written in the style of the Python unittest framework, using stdio for the TAP protocol. The folder tests/functional/qemu_test provides classes (e.g. the QemuBaseTest, QemuUserTest and the QemuSystemTest classes) and utility functions that help to get your test into the right shape, e.g. by replacing the ‘stdout’ python object to redirect the normal output of your test to stderr instead.

Note that if you don’t use one of the QemuBaseTest based classes for your test, or if you spawn subprocesses from your test, you have to make sure that there is no TAP-incompatible output written to stdio, e.g. either by prefixing every line with a “# “ to mark the output as a TAP comment, or e.g. by capturing the stdout output of subprocesses (redirecting it to stderr is OK).

Tests based on qemu_test.QemuSystemTest can easily:

Customize the command line arguments given to the convenience self.vm attribute (a QEMUMachine instance)

Interact with the QEMU monitor, send QMP commands and check their results

Interact with the guest OS, using the convenience console device (which may be useful to assert the effectiveness and correctness of command line arguments or QMP commands)

Download (and cache) remote data files, such as firmware and kernel images

Running tests

You can run the functional tests simply by executing:

make check-functional

It is also possible to run tests for a certain target only, for example the following line will only run the tests for the x86_64 target:

make check-functional-x86_64

To run a single test file without the meson test runner, you can also execute the file directly by specifying two environment variables first, the PYTHONPATH that has to include the python folder and the tests/functional folder of the source tree, and QEMU_TEST_QEMU_BINARY that has to point to the QEMU binary that should be used for the test. The current working directory should be your build folder. For example:

$ export PYTHONPATH=../python:../tests/functional
$ export QEMU_TEST_QEMU_BINARY=$PWD/qemu-system-x86_64
$ pyvenv/bin/python3 ../tests/functional/test_file.py

The test framework will automatically purge any scratch files created during the tests. If needing to debug a failed test, it is possible to keep these files around on disk by setting `QEMU_TEST_KEEP_SCRATCH=1` as an env variable. Any preserved files will be deleted the next time the test is run without this variable set.

Overview

The tests/functional/qemu_test directory provides the qemu_test Python module, containing the qemu_test.QemuSystemTest class. Here is a simple usage example:

#!/usr/bin/env python3

from qemu_test import QemuSystemTest

class Version(QemuSystemTest):

    def test_qmp_human_info_version(self):
        self.vm.launch()
        res = self.vm.cmd('human-monitor-command',
                          command_line='info version')
        self.assertRegex(res, r'^(\d+\.\d+\.\d)')

if __name__ == '__main__':
    QemuSystemTest.main()

By providing the “hash bang” line at the beginning of the script, marking the file as executable and by calling into QemuSystemTest.main(), the test can also be run stand-alone, without a test runner. OTOH when run via a test runner, the QemuSystemTest.main() function takes care of running the test functions in the right fassion (e.g. with TAP output that is required by the meson test runner).

The `qemu_test.QemuSystemTest` base test class

The qemu_test.QemuSystemTest class has a number of characteristics that are worth being mentioned.

First of all, it attempts to give each test a ready to use QEMUMachine instance, available at self.vm. Because many tests will tweak the QEMU command line, launching the QEMUMachine (by using self.vm.launch()) is left to the test writer.

The base test class has also support for tests with more than one QEMUMachine. The way to get machines is through the self.get_vm() method which will return a QEMUMachine instance. The self.get_vm() method accepts arguments that will be passed to the QEMUMachine creation and also an optional name attribute so you can identify a specific machine and get it more than once through the tests methods. A simple and hypothetical example follows:

from qemu_test import QemuSystemTest

class MultipleMachines(QemuSystemTest):
    def test_multiple_machines(self):
        first_machine = self.get_vm()
        second_machine = self.get_vm()
        self.get_vm(name='third_machine').launch()

        first_machine.launch()
        second_machine.launch()

        first_res = first_machine.cmd(
            'human-monitor-command',
            command_line='info version')

        second_res = second_machine.cmd(
            'human-monitor-command',
            command_line='info version')

        third_res = self.get_vm(name='third_machine').cmd(
            'human-monitor-command',
            command_line='info version')

        self.assertEqual(first_res, second_res, third_res)

At test “tear down”, qemu_test.QemuSystemTest handles all the QEMUMachines shutdown.

QEMUMachine

The QEMUMachine API is already widely used in the Python iotests, device-crash-test and other Python scripts. It’s a wrapper around the execution of a QEMU binary, giving its users:

the ability to set command line arguments to be given to the QEMU binary

a ready to use QMP connection and interface, which can be used to send commands and inspect its results, as well as asynchronous events

convenience methods to set commonly used command line arguments in a more succinct and intuitive way

QEMU binary selection

The QEMU binary used for the self.vm QEMUMachine instance will primarily depend on the value of the qemu_bin class attribute. If it is not explicitly set by the test code, its default value will be the result the QEMU_TEST_QEMU_BINARY environment variable.

Debugging hung QEMU

When test cases go wrong it may be helpful to debug a stalled QEMU process. While the QEMUMachine class owns the primary QMP monitor socket, it is possible to request a second QMP monitor be created by setting the QEMU_TEST_QMP_BACKDOOR env variable to refer to a UNIX socket name. The qmp-shell command can then be attached to the stalled QEMU to examine its live state.

Attribute reference

QemuBaseTest

The following attributes are available on any qemu_test.QemuBaseTest instance.

arch

The target architecture of the QEMU binary.

Tests are also free to use this attribute value, for their own needs. A test may, for instance, use this value when selecting the architecture of a kernel or disk image to boot a VM with.

qemu_bin

The preserved value of the QEMU_TEST_QEMU_BINARY environment variable.

QemuUserTest

The QemuUserTest class can be used for running an executable via the usermode emulation binaries.

QemuSystemTest

The QemuSystemTest class can be used for running tests via one of the qemu-system-* binaries.

vm

A QEMUMachine instance, initially configured according to the given qemu_bin parameter.

cpu

The cpu model that will be set to all QEMUMachine instances created by the test.

machine

The machine type that will be set to all QEMUMachine instances created by the test. By using the set_machine() function of the QemuSystemTest class to set this attribute, you can automatically check whether the machine is available to skip the test in case it is not built into the QEMU binary.

Asset handling

Many functional tests download assets (e.g. Linux kernels, initrds, firmware images, etc.) from the internet to be able to run tests with them. This imposes additional challenges to the test framework.

First there is the the problem that some people might not have an unconstrained internet connection, so such tests should not be run by default when running make check. To accomplish this situation, the tests that download files should only be added to the “thorough” speed mode in the meson.build file, while the “quick” speed mode is fine for functional tests that can be run without downloading files. make check then only runs the quick functional tests along with the other quick tests from the other test suites. If you choose to run only run make check-functional, the “thorough” tests will be executed, too. And to run all functional tests along with the others, you can use something like:

make -j$(nproc) check SPEED=thorough

The second problem with downloading files from the internet are time constraints. The time for downloading files should not be taken into account when the test is running and the timeout of the test is ticking (since downloading can be very slow, depending on the network bandwidth). This problem is solved by downloading the assets ahead of time, before the tests are run. This pre-caching is done with the qemu_test.Asset class. To use it in your test, declare an asset in your test class with its URL and SHA256 checksum like this:

ASSET_somename = (
    ('https://www.qemu.org/assets/images/qemu_head_200.png'),
    '34b74cad46ea28a2966c1d04e102510daf1fd73e6582b6b74523940d5da029dd')

In your test function, you can then get the file name of the cached asset like this:

def test_function(self):
    file_path = self.ASSET_somename.fetch()

The pre-caching will be done automatically when running make check-functional (but not when running e.g. make check-functional-<target>). In case you just want to download the assets without running the tests, you can do so by running:

make precache-functional

The cache is populated in the ~/.cache/qemu/download directory by default, but the location can be changed by setting the QEMU_TEST_CACHE_DIR environment variable.

Skipping tests

Since the test framework is based on the common Python unittest framework, you can use the usual Python decorators which allow for easily skipping tests running under certain conditions, for example, on the lack of a binary on the test system or when the running environment is a CI system. For further information about those decorators, please refer to:

https://docs.python.org/3/library/unittest.html#skipping-tests-and-expected-failures

While the conditions for skipping tests are often specifics of each one, there are recurring scenarios identified by the QEMU developers and the use of environment variables became a kind of standard way to enable/disable tests.

Here is a list of the most used variables:

QEMU_TEST_ALLOW_LARGE_STORAGE

Tests which are going to fetch or produce assets considered large are not going to run unless that QEMU_TEST_ALLOW_LARGE_STORAGE=1 is exported on the environment.

The definition of large is a bit arbitrary here, but it usually means an asset which occupies at least 1GB of size on disk when uncompressed.

QEMU_TEST_ALLOW_UNTRUSTED_CODE

There are tests which will boot a kernel image or firmware that can be considered not safe to run on the developer’s workstation, thus they are skipped by default. The definition of not safe is also arbitrary but usually it means a blob which either its source or build process aren’t public available.

You should export QEMU_TEST_ALLOW_UNTRUSTED_CODE=1 on the environment in order to allow tests which make use of those kind of assets.

QEMU_TEST_FLAKY_TESTS

Some tests are not working reliably and thus are disabled by default. This includes tests that don’t run reliably on GitLab’s CI which usually expose real issues that are rarely seen on developer machines due to the constraints of the CI environment. If you encounter a similar situation then raise a bug and then mark the test as shown on the code snippet below:

# See https://gitlab.com/qemu-project/qemu/-/issues/nnnn
@skipUnless(os.getenv('QEMU_TEST_FLAKY_TESTS'), 'Test is unstable on GitLab')
def test(self):
    do_something()

Tests should not live in this state forever and should either be fixed or eventually removed.