Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test started with "start_vm = no" and we are starting the VM
manually. It's not guaranteed, that params were updated in preprocess
so we need to make sure params were updated by calling create with
new params.

One example of incorrect params is when VM exists and is alive from
previous test. In such case preprocess skips initialization (start_vm
= no") and vm.create() destroys and recreates the VM from previous test.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

qemu_devices is becoming way too long. This patch splits it into
logical modules: qdevices, qbuses and qcontainer.

I perserved the qemu_devices_unittest in the same form for now,
but it can be split in the future. It already is sepparated into
three classes.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

There's no need to have the name of the "run_<name>" method (main
test entry point) depend on the test name. So, for simplicity sake,
let's just name the test entry point "run".

I have decided to not fallback gracefully to the previous naming
convention as I believe this can lead to confusion in the future.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Fix the problem with duplicated IDs caused by early device_del execution.
Monitor returns no output but internally waits for OS ACPI respone. This
message, in case kernel is not yet initialized, never occurs.

This version resends the device_del in the first round until it succeeds,
than it just waits until the device disappears.

Additionally the test was rewritten to use qemu_devices to handle the
hotplug/unplug and the test cfg was tweaked to be more intensive.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

In order to make the utils_test module more maintainable,
split the qemu tests specific functions to the module
utils_test.qemu. Of course, eventually modules specific
to other test types, such as libvirt, v2v and the such
can be also carried.

All the API references were fixed, and functions that
were mentioned in a single test module were moved to
said test module.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Before we can cleanup after migrate test the guest worker has to
reconnect to the current VM. The session itself works fine
as the VM was migrated, but the VM object is different and reports
non-living VM, which results in impropper cleanup.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Current kernels report Error 16 Device or resource busy: ... instead
of Error 24 Access is denied. Accept all kind of failures as success
in this particular test.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Currently qemu dies when we try to remove and plug the port couple of
times without delay.

This modification allows to execute x (virtio_console_micro_repeates)
interuptions, where we don't check anything, only do them as fast as
possible. After that we execute normal loop where the transfer is
checked after each interruption.

As the port might not be openable for some time, the guest_worker
supports multiple attemps to open the port. Don't forget to remove
the old /tmp/virtio_console_guest.py* files in order to use the new one.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

No actual change, only put the interrupted_loopback initialization
before the thread initialization (in case of failure it's quicker)
and fix the 80 characters line width.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Some tests were voulnurable to infinite loop due of not finished
background thread. Use funcatexit to register function, which sets
the exit_event and causes the background threads to finish.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Tweak the config to maximize interrupted_loopback test
* short interruption is 0s
* short test_time is 0.1s
* tweak the acceptable_loss
* increase the number of acceptable test_time repetition
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Usually the index of lost characters is reloaded while reconnection.
This patch creates a function, which can be called from outside of this
thread in case we want to reload this index elswhere. Additionally booth
(debug and normal) threads use the same code now.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

err value was overwritten by thread return number which resulted in
incorrectly reported failed tests as pass.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Fix some extra problems with docstrings in the code tree.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Make virt-test to have a docstring style mostly compliant with
the sphinx's [1] format, reStructured text. This means that
we have to trade all '@' tags to be ':', mostly.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

With the help of the PEP8 script, a semi-automated clean
of the code base was made, and effectively changing some
of the coding style rules of our code base:

 * No more 2 lines spacing among functions
 * No more combined module imports

In the end, it was going to be too difficult to take
into consideration E501 (line length > 80) errors and
W601 (.has_key() is deprecated), so these were ignored,
at least for now.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Using the program codespell

https://github.com/lucasdemarchi/codespell

We found a number of spelling mistakes, that could be
found and fixed. The result of the automated script
was double checked, of course.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

The bacground process was likely to be killed by SIGIO while setting
port to async mode. This was due of bad workflow, where first the
port was set to async mode, than the the new SIGIO handler was set.

With this version when we set async mode, we fist add the SIGIO
handler and then set the port to async mode.

When we are setting the port back, the order is the opposite, so
first the port is set to sync mode and then we remove the SIGIO
handler.

Additionally the workaround with test_delete_guest_script() was
removed as it's fixed by this patch.

NOTE: You have to remove the previous /tmp/virtio_console_guest.py*
      in order to let virtio_console test to copy this updated one!
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

virtio_console catches exceptions and raises TestFail when unknown
exception occurs (because any Shell/... exception usually means
test failure.

Anyway it should raise TestError and TestNAError in case they occurs.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Thread might be alive, but hung on data recv/send. With this check
the test is interrupted after 1s of inactivity. It doesn't check for
new data on the sender thread as with some settings (big send buffer
and small recv buffers) the queue might be filled and this thread
has to wait for new data. This is expected behaviour.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch benefits from the better error message when key is missing
in config params. Instead of custom handling of missing param it
uses params[] directly.

I modified those params, where I was 100% certain that they are
required and doesn't have default value:
 * val = params.get(..) followed by if not val: raise...
 * int(params.get(...))
 * params.get(...).split()
 * commands which require string (eg. session.cmd(params.get(...)))

I didn't test all the tests before and after in order to verify
possible typos but I used pylint before and after, watching for
new errors (code rate before and after).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Fail with error.TestError in case suspend or hibernate commands are
missing.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Take care of freezed sender thread everywhere as in 66e657
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Use a experienced reboot timeout 720.
Signed-off-by: NAmos Kong <akong@redhat.com>

Force exit background threads on thest failure.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

When guest script is already running and the test only connects, every
signal destroys the daemon. This is complex problem of python, where
only the main thread accepts the signals. In order to fix this without
much pain I added test_delete_guest_script() at the test beginning thus
fresh instance of guest script is always executed. This solution seems
to work pretty well.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test closes the used ports while transfering data and after the
intr_time opens them again. It randomly closes ports on guest side
and host side, send threads and recv threads thus in the long run
it should cover all cases.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Loopback test might freeze the whole autotest when sender thread
fills the buffer. During cleanup it tries to send additional data
which is blocked due of full buffer. The only solution I found is
to destroy the VM and inform about this bug.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch adds interrupted_transfer suspend variant. This variant
initiates loopback over 2 vio ports and suspends the VM to mem.
After a while it resumes the VM with 'system_wakeup' and verifies
that the data are still transfered in booth directions.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

On heavy loaded VM data might not yet be available, give
the machine second chance in case of Errno 11.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Some tests (polling, sigio, rmmod, ...) raised VirtioPortException
instead of TestFailure. This might be confusing so let's raise
TestFail + original traceback instead.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

A frequent request we've got from users is that the name
'kvm' makes assumptions that are not true for a lot of
the use cases of the test suite. Then let's rename
the subtest 'kvm' to 'qemu', making it more accurate.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Don't wait until the test timeout, but fail the test when thread
or VM fails.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

When VM dies it closes connection to sockets. We might (and often)
stuck in readout for ever.

Additionally get rid of in-test readouts which are handeled by
guest_worker's safe_exit_loopback_threads()
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch removes debug print which somehow made it to upstream, plus
two long line fixes. There are no other (fixable) pylint errors now.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

hibernate test on powerful machines fails very often. It is due to
short timeouts. To fix it I changed the check logic. Acceptable loss
is set to (almost) infinity. On the other hand it's set to 0 when
the initial loss after resume is over (at least 100 chars transferred
without any error).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>