qemu.cfg: Typo fix in cdrom_test.cfg

Signed-off-by: NFeng Yang <fyang@redhat.com>

Autotest regression test

libvirt.tests: Support multiple devices for multifunction test.

tests.kdump: Add new sub test for multiple guests

qemu.test: Update the check method in smbios_table

In the machine_type check for smbios_table, we only care the main version
information but not the details. And the command line output from guest
will include more than the version info:
System.Version mismatch, Set 'RHEL 7.0.0 PC'
but guest is : 'RHEL 7.0.0 PC (i440FX + PIIX, 1996)'

So update the check method for it.
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

Disable balloon_fix_value case in ppc64 platform

libvirt_vm: remove setup_serial_ports() method.

libvirt: use handle_prompts() to wait for end of 'ZZ' in session.

qemu.tests: Update get_property useage in vhost_with_cgroup

The parameter in get_property should be filename without \.
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

Currently, we use time.sleep(1) to wait for the end of 'ZZ' in session,
but it does not work well in any conditions.

This patch replace time.sleep(1) with remote.handle_prompts().
Signed-off-by: NDongsheng Yang <yangds.fnst@cn.fujitsu.com>

Issue 1303 fix

Add test for snapshot current and parent test

shared.downloads: Add Fedora 20 asset file

It is handy for people wanting to download the recently
released OS's iso.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Indentation fixes

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

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

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

virt-test: add Fedora 20

qemu.tests.multi_disk_random_hotplug: Couple of little improvements and bugfixes

So that we can run the shorter set of unittests if desired.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

I've found that it's useful to check just the autotest installation steps
with a single VM with increased resources, mainly memory.

First 1G of memory is too little to do things like compiling the GWT client,
2G is OK, but 4G is preferrable. On most development machines (mine included),
starting two VMs with 2G or 4G is usually not a good idea.

To set this single VM (server only) mode, simply set the "vms" parameter to
a single (server) VM name.

Also, be a lot more generous about the installation timeout, as based on my
experience, things like downloading GWT and the git trees can take a long
time (YMMV).
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NCleber Rosa <crosa@redhat.com>

Somewhere the new-ness of the class heirarchy is getting lost leading
to these errors:

    # tools/run_pylint.py -q virttest/utils_net.py
    ************* Module autotest.client.tests.virt.virttest.utils_net
    E1001:1221,0: VirtIface: Use of __slots__ on an old style class
    E1001:1334,0: LibvirtIface: Use of __slots__ on an old style class
    E1001:1342,0: QemuIface: Use of __slots__ on an old style class

Adding a multi-inheritance of ``object`` to the base class makes pylint
happy and all unittests run clean.  I'm assuming that means there are
no negative consequences.

Signed-off-by: NChris Evich <cevich@redhat.com>

Running these unittests in a specific order resulted in many
failures.  The problem is, the unittests all modify contents
of the virsh module, then they proceede to make shared-access
of the module.  To fix this, isolation of the virsh module for
each unittest would be needed.  However this is not practical
since python likes to save memory by sharing loaded modules.

Since all unittests that use the virsh module normally create
an instance of the Virsh() class anyway, it's much easier to
not isolate the module, but work with a copy of a specially
prepared Virsh() class instance.  This instance can be generated
with the new FakeVirshFactory() function from the virsh_unittest
module.
Signed-off-by: NChris Evich <cevich@redhat.com>

Many unittests require testing codepaths containing
the virsh module and/or Virsh() class instances.  However
it is necessary to isolate them from eachother AND provide
a check to make sure no "real" system-virsh calls are made.

To make this work easier for other unittests, added a
factory function that returns a mocked Virsh() class instance.
Every closed method on this instance will raise a
bogusVirshFailureException (subclass of
unittest.TestCase.failureException).

Now, unittest tests only need to call this function and
override the resulting instance's methods as needed for
testing.  I also added a unittest to do a basic check of
this new unittest helper function.
Signed-off-by: NChris Evich <cevich@redhat.com>

This virsh module function becomes a Virsh class method, but
internally it makes other calls to the virsh module directly.
Sometimes this is not what is needed, rather a specific Virsh
class instance should be used (i.e. to run against a remote
system or with special args/dargs).  Added a parameter
with None default to allow this.  Fixed up all references to
this function so they also pass through proper virsh_instance
state.
Signed-off-by: NChris Evich <cevich@redhat.com>

SR-IOV: Support get VF from specified sr-iov PF.

Signed-off-by: NRoss Brattain <ross.b.brattain@intel.com>

Commit c37a86ca picked out the create_serial_console() from setup_serial_ports(),
and call it in env_process(). But it is not robust enough.

This patch rename the setup_serial_ports() to create_serial_console() to be called
in env_process() and remove the older create_serial_console().
Reported-by: NYanbing Du <ydu@redhat.com>
Signed-off-by: NDongsheng Yang <yangds.fnst@cn.fujitsu.com>
Reviewed-by: NYanbing Du <ydu@redhat.com>
Tested-by: NYanbing Du <ydu@redhat.com>

With incorrect configuration it might happen that no disks can be added.
Throw exception instead of simple warning and clean test run.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The qdev.images_define_by_variables needs to be in the try/except
block in order to work as expected (remove format and try another round)
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Add context for easier debugging.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch changes the handling of unplug. The simplu_unplug has sometimes
problem detecting correctly the unplug status, as it might take seconds
before the disk is removed (monitor returns output, but qemu still
waits for OS repsonse before it removes the device from qtree).

This custom handler waits up to 1s for device to be removed and even
when it's not proved to be removed it removes it from representation.
We can afford this as all devices are again checked later in the test
and the VM state is for the whole period set to dirty.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Libvirt bench

Init serial_ports to [] instead of None