The iscsi target should be configured by users and not hardcoded. Let's
use some dummy values which usually result in:

    AttributeError: 'NoneType' object has no attribute 'logged_in'

and from traceback it's obvious it's checking iscsidevice.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The newer `dd` reports human readable records separate for in and out
records:

    1142986240 bytes (1.1 GB) copied, 0.479684 s, 2.4 GB/s

old:

    1142986240 bytes (1.1 GB, 1.1 GiB) copied, 0.479684 s, 2.4 GB/s

Let's improve the regexp to work well with both.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "process.system" does not uses "shell" by default, but it's expected
in most command in cgroup test. Let's set the "shell=True" explicitly
whenever needed.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

In case of failure this test used to leave all the images. Let's use
try/finall to avoid uncleaned files.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The current variant name RHEL.7.devel.x86_64 clashes with existing
filters enabling this test on all architectures and actually corrupting
the cartesian product. Let's use `-` to separate the parts.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

All recent kernels provide MemAvailable in /proc/meminfo, let's try to
use that instead of MemFree. For backward compatibility also support
MemFree approach but drop caches before evaluation.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

On s390x the balloon device is virtio-balloon-ccw and not -pci.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The s390x supported drive formats is rather limited so it does not make
sense to even pretend to support this test. Let's just filter it out.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test uses suspend and hibernate and currently is disabled on ppc64
and aarch64. As it does not work on ppc64le and s390 let's disable it
everywhere but x86.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The isa-serial is not available outside x86 world, let's change the
filter-out to filter-only.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test only makes sense in x86 world. Let's use only instead of out.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test uses "/home/test" by default which is usually present as a
directory. Let's use "/home/test.img" to decrease the probability of
existing directory.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test only makes sense on x86 archs and it is currently disabled on
q35. Let's enable it only on i440fx then.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

On s390x rtl8139 and e1000 are not supported (similarly to ppc64) and
the virtio-net is called virtio-net-ccw.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "dev_type" used to be used to parse qtree, but there is a generic
implementation which does not rely on dev_type which is not used in the
test.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "virtio_blk" is generally usable nowadays while it should not clash
with the "virtio_scsi" used in the test. Let's use it to boot the
machine instead of ide/ahci, which is not supported on all archs.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test relies on outdated first vm name, while there is "main_vm"
parameter pointing to the correct name. Also it stores all created
vmnames, but instead of relying on this variable it tries to create the
name in later loops.

Ideally one would store the actual vm objects, but let's use this as a
hotfix.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Ideally this test should not exists in Avocado-vt as kvm-unit-tests are
dynamic and more and more are added each day. Anyway as it exists let's
limit it to x86 only as it defines x86 variants.

Those who want to run kvm-unit-tests please use pure Avocado instead.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The process library uses it's own "process.CmdError" exception.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test allows to check all snapshots, but currently only checks the
main image and last snapshot. Let's check one of the middle snapshots as
well to see it behaves correctly.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Newer qemu version modified the human protocol to report block devices,
removing the "file=" headers and using the filename as the first thing.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "live_snapshot_chain" test might pass even when the snapshot is not
actually created. Let's verify the file exists.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "drop_caches" should only be performed when we need to measure the
disk performance, which is not the case of live_snapshot_chain. Linux
kernel should make sure it uses the files it's suppose to be using and
caching is useful feature used out there in production.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "/home/test" destination is likely to be present on some systems.
Let's use "/home/test.img" to decrease the probability of existing
directory with this name.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test uses hardcoded ide/ahci drive, which is only available on
i386/x86_64. Let's blacklist it for other platforms.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test uses hardcoded i386 and x86_64 machine types, let's restrict
it to those archs only.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

utils_misc.find_substring will logged every mismatch info. When
we call it in "get_match_cdrom", the mismatched info will be logged
util the matched cdrom found. This will cause misleading. So rewrite
it.
Signed-off-by: NAihua Liang <aliang@redhat.com>

Mostly spelling wording changes to make it update with the
Avocado-VT/VirtTest terms.  And the shebang which can safely
go.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

1. Should report error when leaked clusters are founded
2. As image locking is enabled, must use force shared option to
check a stopped image.
3. Update the function to print the debug info
Signed-off-by: NPing Li <pingl@redhat.com>

Signed-off-by: NQianqian Zhu <qizhu@redhat.com>

When the winutils disk letter happen to change in guest,
the "cmd = D:\hddsn.exe C:" cannot be executed successfully.
Here changed it to "cmd = WIN_UTILS:\hddsn.exe C:",
and used the set_winutils_letter() to set the correct disk letter.
Signed-off-by: Npeixiu <phou@redhat.com>

1. The return value of utils_misc.get_support_machine_type has changed,
   update function call to match it.
2. Add "remove_alias = True" for smbios_table.
Signed-off-by: NYanan Fu <yfu@redhat.com>

In case no cpuflags were detected to be tested (after blacklisting) this
test silently passes not testing anything at all. Let's use
"test.cancel" to notify the user that nothing was actually executed.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Those 3 variants currently only work for the old "qemu -cpu ?" handling,
but it only informs the user about not-running when new (1350) type is
detected. In other cases (not old, nor 1350) it silently passes without
notifying the user that nothing was actually executed. Let's use "else"
instead of "elif" to avoid silent skips.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

When the "get_all_qemu_flags_1350" fail to find flags, it crashes due to
uninitialized "flags" variable. Let's initialize it to list.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

To enable "-U" option for "qemu-img compare", avoid access error.
Signed-off-by: NQianqian Zhu <qizhu@redhat.com>

Signed-off-by: NZhengtong Liu <zhengtli@redhat.com>

1.add gagent-stop function.
2.add case to stop/restart qemu-ga service repeat 100 times.
3.since gagent_start() function has changed, modify qemu-ga suspend file.
Signed-off-by: NXiaoling Gao <xiagao@redhat.com>

add optional steps to support reboot/shutdown guest after hotplug
or hot-unplug rng devices
Signed-off-by: Nlijin <lijin@redhat.com>

device

add optional steps to support reboot/shutdown guest after hotplug
or hot-unplug balloon devices
Signed-off-by: NLi Jin <lijin@redhat.com>