Normally, these configs are important to a test job, thus show
them all in test's name.
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

Since funcatexit are cleanup functions, make the test fail with
TestWarn and proceed with postprocess_on_error.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Local functions are not pickable. Raise a sane exception when the test
tries to use them.

Global functions (defined on the module scope) are fine.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Because all arguments are stored in env we have to ensure the arguments
are pickable before storing them. Otherwise users would get weird errors
pointing at env:
    TypeError: can't pickle ... objects
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

* Get rid of old virt-test reference to 'kvm' driver_type
* virtnet style doesn't care about kvm vs qemu difference
* unattended_install uses it for both qemu and libvirt virt-test
* No other usage found for vm.driver_type sensitive to 'kvm' value
Signed-off-by: NChris Evich <cevich@redhat.com>

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

When the guest_name changes (arch, machine, ...) the TAG_INDEX length
might differs from the previous one. Store indexes per name to avoid
shortname corruption.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

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

Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

The save_restore.py script contains some linux-specific
operation, thus add a filter to limit it on Linux guest.
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

This exception might make user confused, so add some more info
to its message.
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

To make sure boot_nr tests are working correctly it's necessarily to boot
(and open) the port with nr=1.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NYiqiao Pu <ypu@redhat.com>

Signed-off-by: NLi Yang <liyang.fnst@cn.fujitsu.com>

Signed-off-by: NLi Yang <liyang.fnst@cn.fujitsu.com>

Signed-off-by: NLi Yang <liyang.fnst@cn.fujitsu.com>

The name of the clock on Power guest is "timebase" which is
different from it counterpart in x86 "kvm-clock"

To makes it can work in both Power and x86 guest, add the
platform check.
Signed-off-by: NMike Qiu <qiudayu@linux.vnet.ibm.com>

Nowadays, netdev_id has been a part of nic, and not
belongs to the vm object anymore.

At the sametime nic already has the key 'device_id', so
use nic.device_id is OK.
Signed-off-by: NMike Qiu <qiudayu@linux.vnet.ibm.com>
Acked-by: NFeng Yang <fyang@redhat.com>

Currently, when running the tests while the clean
up didn't run on the previous, it will raise an
error like:

Test failed: CmdError:
Command <fuser -k /dev/vg_kvm_test_enospc/lv_kvm_test_enospc>
failed, rc=1, Command returned non-zero exit status
[context: performing enospc setup --> performing enospc cleanup]
* Command:
    fuser -k /dev/vg_kvm_test_enospc/lv_kvm_test_enospc
Exit status: 1
Duration: 0.0501132011414

That is the link /dev/vg_kvm_test_enospc/lv_kvm_test_enospc
exist, but no user. thus the command line:
    fuser -k /dev/vg_kvm_test_enospc/lv_kvm_test_enospc
will return 1(fuser will always return none-zero for nobody use).

So just ignore the return status of the fuser.
Signed-off-by: NMike Qiu <qiudayu@linux.vnet.ibm.com>

Currently, the VM names has changed to like virt-tests-vm1 style.
But the VM names in test case time_manage are still the old version

Modify the VM names to the new style so the test case pass.
Signed-off-by: NMike Qiu <qiudayu@linux.vnet.ibm.com>
Acked-by: NFeng Yang <fyang@redhat.com>

Modify code to catch the word 'processor' rather than 'model name'
in /proc/cpuinfo to count cpu number. Becasue the Power cpu haven't
the 'model name' field, and both x86 and Power cpu has the
'processor' field.
Signed-off-by: NMike Qiu <qiudayu@linux.vnet.ibm.com>
Signed-off-by: NLiu Sheng <liusheng@linux.vnet.ibm.com>

This would prevent case error when test with QMP
as it uses 'inject-nmi' instead of 'nmi'
Signed-off-by: NXiaoqing Wei <xwei@redhat.com>
Acked-by: NLukas Doktor <ldoktor@redhat.com>

both HMP and QMP
Changlog:
use self.cmd() on QMP monitor
Signed-off-by: NXiaoqing Wei <xwei@redhat.com>
Acked-by: NLukas Doktor <ldoktor@redhat.com>

Because some old versions of qemu-kvm take more than we'd like to
initialize, and this is not a bug or one that will be fixed.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

This patch introduces using macvtap nic in test, when using
macvtap nic better have a switch support macvtap, if not,
need have a host with two nic devices when using this patch.

if you want to use macvtap in your test, just configure
"nettype = macvtap" in your .cfg file, or provide
--nettype macvtap as a flag for the virt test runner.
Signed-off-by: NYunping Zheng <yunzheng@redhat.com>

Signed-off-by: NJiří Župka <jzupka@redhat.com>

Signed-off-by: NJiří Župka <jzupka@redhat.com>

If the parent directory of the image file is missing,
the qemu-img command will throw error messages that
are not very helpful. This condition usually indicates
a serious setup problem.

However, It is possible that the missing image dir is a
localized problem and the job can continue. So, instead of
erroring the test, like previously considered, just try
to create the missing parent image dir, and proceed with
the image creation, after logging the condition and giving
some idea to the tester what could be wrong there (if
the setup problem is so serious the test will fail anyway,
but now in a more clear and understandable way).
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

By default, the qemu_storage code is ignoring failures
when qemu-img fails to create a disk image. This happens
because the failure is actually considered only if the
(terribly undocumented) param 'check_output' is set to
yes.

Since there's no 'check_output' on any .cfg file
in the virt-test repo, the failures will never be
considered and virt-tests will happily proceed, even
though a world of pain and suffering will follow.

Since there are future tests that might require the
output of the creation command, we'll fix the problem
by:

1) Introducing a ignore_errors boolean parameter to
the .create() method, when set to true it'll make
the method to not raise an error on failure.

2) Making the method return a fixed tuple
(path to the image created, object with the result
of the command). With the result object, it is easy
to get outputs as they are attributes of the said
object.

This also has the nice side effect of not requiring
the check_output parameter anymore.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

We depend on '/usr/bin/fakeroot' to install from iso's, but
it was not required or suggested by get_started.py.
Signed-off-by: NChris Evich <cevich@redhat.com>

Setting the environment variable MALLOC_PERTURB_ to
a non zero value [1] helps to find out some subtle
memory allocation problems, and it is being used in
our test grid. However, by default, the virt test
runner doesn't use it.

Add the new, qemu test specific --use-malloc-perturb flag
and set it to 'yes' (use the env variable) by default.
This way we're being more strict, which makes sense
in the context of a test suite. If people want to double
check the behavior of the test without the flag, just
use '--use-malloc-perturb no'.

[1] http://udrepper.livejournal.com/11429.htmlSigned-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

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

We do all possible to run static checking of modules
on older python, given that it is easier to spot
python 2.4 non compliances. However, we have a bunch
of quirks with files on older python. Therefore, let's
skip certain files from the checks if running under
py < 2.5.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

pylint (or some of the dependent libraries) also has
bugs, so when they happen we don't want to spoil our
entire run. Let's catch the exceptions and report them.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Add a file with the maintainer's names for the virt
test project.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Windows doesn't have killall command and the worker is killed by the
missing connection anyway.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

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