This patch adds windows support for sending/receiving data
from/to guest.

It also adds send_loop support for guest sender thread.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch enables virtio_console_guest.py to initiate loopback
under windows.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch enables virtio_console_guest.py script to run on Windows.
It's a first very limited version.

VirtioGuest class now contains all functions, which might be used in
virtio_console tests and should be implemented.

The Posix version is the original Linux version with minor bugfixes.

The Nt version is simplified version which currently doesn't support
port reconnection (migration) and implements only "open" and "close"
features.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

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

Signed-off-by: Yu Mingfei<yumingfei@cn.fujitsu.com>

Signed-off-by: Yu Mingfei<yumingfei@cn.fujitsu.com>

Signed-off-by: Yu Mingfei<yumingfei@cn.fujitsu.com>

Signed-off-by: Yu Mingfei<yumingfei@cn.fujitsu.com>

It took us a while to notice that the source for good
images was incorrect in the backup code:

10/19 08:07:59 DEBUG|   storage:0185| Copying
/tmp/kvm_autotest_root/images/win7-64-sp1.qcow2.backup ->
/tmp/kvm_autotest_root/images_archive/win7-64-sp1.qcow2.good.XWUf
10/19 08:07:59 INFO |   storage:0189| No source file
/tmp/kvm_autotest_root/images/win7-64-sp1.qcow2.backup, skipping copy...

Let's fix that.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Test the command virsh nodecpustats

(1) Call the virsh nodecpustats command for all cpu host cpus
    separately
(2) Get the output
(3) Check the against /proc/stat output(o) for respective cpu
    user: o[0] + o[1]
    system: o[2] + o[5] + o[6]
    idle: o[3]
    iowait: o[4]
(4) Call the virsh nodecpustats command with an unexpected option
(5) Call the virsh nodecpustats command with libvirtd service stop
Signed-off-by: NSatheesh Rajendran <sathnaga@linux.vnet.ibm.com>

Signed-off-by: NSatheesh Rajendran <sathnaga@linux.vnet.ibm.com>

Signed-off-by: NSatheesh Rajendran <sathnaga@linux.vnet.ibm.com>

connect_uri is used for canonical_uri(),but it will not
be changed even remote test. While the connect_uri for
remote test should be target_uri.

Signed-off-by: Yu Mingfei<yumingfei@cn.fujitsu.com>

The tools subdirectory will include useful scripts for
users of the virt tests. The first app is the koji/brew
pkg spec tool, that allows people to list information
about a koji or brew build.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Pylint correction makes migration_multi_host_fd test dysfunctional.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

The test iterates through pkt_sizes even we're not on
windows, so move its definition up.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

I found later that the problem I was referring in the
below issue relates only to the virsh vcpupin testcase
since it uses qemu_monitor_command function.

After the introduction of virsh class, there is a function
named 'command' got introduced, which was getting conflict
with the command variable and below patch solves the issue.

https://github.com/autotest/autotest/issues/555

Excerpt of error log:
----
17:52:44 DEBUG| Destroying VM
17:52:44 WARNI| Requested MAC address release from persistent vm vm1.
Ignoring.
17:52:44 ERROR| FAIL -> TypeError: 'str' object is not callable
FAIL (3.42 s)
----

Sorry for the confusion if any caused.
Signed-off-by: NSatheesh Rajendran <sathnaga@linux.vnet.ibm.com>

Adding a test to verify a remote-viewer connection
with the full screen option the guest takes the
resolution of the client, and it also supports the
negative test if connected w/o the full screen
option the resolution of the guest will not take
on the resolution of the client.
Signed-off-by: NVimal Patel <vipatel@redhat.com>

The purpose of the test is to setup the VMs for
the full screen tests, change the resolution of
the guest VM so the VMs have different resolutions,
and start spice-vdagentd prior to remote-viewer
connection.
Signed-off-by: NVimal Patel <vipatel@redhat.com>

Signed-off-by: NVimal Patel <vipatel@redhat.com>

Defining fullscreen_setup & fullscreen tests, and
add a full_screen parameter to the rv_connect test.
Signed-off-by: NVimal Patel <vipatel@redhat.com>

Signed-off-by: NVimal Patel <vipatel@redhat.com>

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

This patch groups multihost test into one group to which is
added machine types variant of test.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

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

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

Before testing multihost.srv prints list of test which will be tested.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

From time to time[1], we see cases where, after a successful
guest install, a lot of tests will fail because we can't
get a connection to the unix socket tied to the monitor.

So, let's at least print what the exception was, so we
have a better idea of what's going on.

[1] Rarely, actually.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

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

During pylint checking under 2.4, we've found:

************* Module ksm_overcommit_guest
E1102:222:die: exit is not callable

Let's fix it by making the script call sys.exit(0) instead.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Found with automated pylint checks on py 2.4:

************* Module client.tests.virt.virttest.aexpect
E1102:183: exit is not callable

Turns out that indeed in that scope we needed sys.exit(0)
rather than exit(0).
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

During the automated pylint checks, I've noticed that we
were getting an error report from pylint:

************* Module client.tests.virt.kvm.tests.time_manage
E0602:104:run_time_manage: Undefined variable 'format'

As a matter of fact, this builtin was introduced with
python 2.6, therefore it's illegal in py 2.4. Let's fix
it to use an equivalent, py 2.4 legal statement.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

With commit e0251772, that makes serial ports configurable,
isa_serials is a mandatory configuration key for a vm, and
that was making our unittest vms to fail, since they use
those tests use their own unittests.cfg. Fix this and
get the ability to run the unittests back.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

There was problem with qemu parameter -S.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

As part of the work to resolve the issue:

https://github.com/autotest/autotest/issues/555

I've started to write unittests for the Env class,
and those unittests started to show some small
problems on that class.

So, create a new set of unittests for Env, and
modify it in the following ways:

1) Env._filename will always get set, to avoid
an Env object instantiated with no filename passed
to the constructor can be saved if no filename
was passed to the save() method;
2) All set/get operations will now be performed
in Env.data, for safety
3) The get_all_vms method was depending on a not
quite correct is_vm() function, that was checking
the name of the class of the object, while the
proper way to do it would be calling isinstance,
instead since we guarantee all vm keys will
start with vm__, use that instead, creating a
more correct and more testable method.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

The test names for VirtNet were mixing capitalization
and underscores, let's go with simple CamelCase, as it's
more in line with our coding style.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Instead of making such a mess, use a relative import
and let's call it a day.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Follow the same schema used in the kvm, libvirt and v2v
subfolders, so the code can actually find an autotest
checkout to import modules from.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

This is a solution for the issue #554.

Since we have to do a bunch of hacks to supress stdout,
it's not possible to use conveniently, for example, the
python debugger to debug your test code. By adding a
configuration option to the StreamProxy class constructor,
allow full stdout output in case the --verbose option
was passed to the run program.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

This patch add a function called has_device, so that we can
check if qemu-kvm support given device.

Changes from v1:
 * Was using help, code was refactored upstream, it created
   a bug that had to be fixed by changing help with hlp.
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>