This introduces the flag --qemu-bin, that allows people
to specify an arbitrary qemu built on your box, such as:

./run -t kvm --qemu-bin /home/lmr/Code/virt-test.git/kvm/install_root/bin/qemu-system-x86_64

And this will automatically run the tests with that qemu.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

There's no reason why all tests can't use a single roof
for their activities, moreover, /var/tmp is a safer place
to store isos and disk images, considering that we might
have /tmp on tmpfs on Fedora 18 on.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Now setup log goes to main debug log, setup phase is
displayed as a test. Also, there was a bug on the
uncompress stage in case the image was already
uncompressed, fixed by passing the right flags to 7za.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

In Fedora, it requires less dependencies.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Modify the virt assistant function to:

 1) Download the JeOS image
 2) Setup the environment non interactively

So it all will be done without user intervention.
The first time the JeOS will have to be built,
a couple of minutes to download the 120MB compressed
image + 1 minute to uncompress the image will be
required, after that it'll keep using the uncompressed
image, unless you want to replace it by the pristine
file you have on your hard drive, then you have to call
run with --restore-image.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

After exploring some possibilities, we came up with a
minimal Fedora install, which presents an interesting
compromise between features and disk space. The idea
is to host the compressed image on my FedoraPeople
area and have code to automatically download it if it's
missing, and subsequently cache it.

The unattended install variant is in fact only for
people that, for some weird reason want to rebuild
the minimal guest on a convenient way, nothing more
than that.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

+ libvirt_xml:
    + Improved docstring
    + Removed uuid module usage
    + Removed duplicate method definitions
    + Fixed comment spelling mistakes
    + Added os_arch_machine_map property to LibvirtXML

+ utils_misc:
    + Remove unused Closure Class
    + Fixed indentation / spacing
    + Added comments

+ utils_misc unittest:
    + Cleaned up nested class/method definitions
    + Unfolded some complex assignments
    + Fixed indentation / spacing
    + Add some dict-method unittests

+ virsh:
    + Updated/Clarified module docstring
    + Removed unnecessary DaArgMangler class
    + Fixed specialized VirshClosure Class
    + Cleaned up VIRSH_EXEC usage & warning
    + Replaced old use of PROPERTIES with __slots__ (resolved TODO)
    + Enforced setting of 'uri' and 'virsh_exec' properties
    + Made VirshSession class record the uri for later comparison
    + Major VirshPersistant cleanup/bugfix
        + Added session acounting (helps unittests)
        + Added close_session method to unify behavior
        + Fixed new_session to properly close alive existing sessions
        + Fixed set_uri to open new session only on uri change WRT old session
        + Cleaned up virsh.command to work with VirshPersistant properly

+ virsh_unittest:
    + Added back unittest module (some tests ignored if no libvirt)
    + Added 8 tests for VirshClosure
    + Added VirshPersistent session count check to ConstructorsTest
    + Added help command cache tests
    + Added VirshPersistant test teardown double-check session accounting

+ libvirt_vm: Cleaned up hvm/pvm arch and machine-type detection/check

+ guest-os.cfg.sample:
    + Added vm_arch_name param for every OS defaulting uname -m
    + Removed a few guest os types that aren't used anymore
Signed-off-by: NChris Evich <cevich@redhat.com>

Thanks to Yu Mingfei <yumingfei@cn.fujitsu.com> for the original
virsh_rename_domain test which this is based upon. Exercises
virsh define and virsh_undefine plus auto-generated uuids.
Signed-off-by: NChris Evich <cevich@redhat.com>

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

Follow a properties and accessor-method pattern so allowing higher
level code to ignore all details of underlying XML format.  Also,
provide intermediate methods for working with XML related virsh
commands along-side the abstracted XML-manupulator classes.
Signed-off-by: NChris Evich <cevich@redhat.com>

The virt networking code in utils_misc and base classes from the virsh
and libvirt_xml modules were duplicating a lot of functionality.  Moved
the comonalities under a new utils_misc PropCanBase base class. Added
new unittests for the new base class and fixed up all usage mismatches.

Also moved Closure class from virsh into utils_misc so it's more
centrally located for common use.
Signed-off-by: NChris Evich <cevich@redhat.com>

    Test uses virttest for virtual machine management.

    /tests/openvswitch/
    preparing test environment:
      ./get_started.py

    tests:
    1) Performance test.
    2) Vlan tests.
    3) Ping test.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

Allow creating of machine with tap devices which are not
connected to bridge, and add function to fill virtnet
object with address.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

When autotest tries add tap to bridge then test recognize if
test is bridge is standard linux or OpenVSwitch.

And add some utils for bridge manipulation.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

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

Turns out this caused for a long time a namespace
collision that confused pylint. So, in order to get
a fully clean run_pylint.py -q in the entire tree,
let's rename the test to autotest_control.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

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

Add a script that will run all unittests available on
virt tests source dir. With this it'll be easier to
run the unittests and have a quick glance of the code
state.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Fedora 18 introduces 'xterm-256color', which was not
accounted for on initial versions of the bcolors class.
Let's add it so we get colored output for this new
version of Fedora.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

This test script runs live snapshot after guest
fs is frozen, then check if the file created on
guest disk is correct.
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

This patch contains several test cases:
 * Execute "guest-sync" command to guest agent
 * Halt guest with command "guest-shutdown"
 * Reboot guest with command "guest-shutdown"
 * Shutdown guest with command "guest-shutdown"
 * Check the guest status with command "guest-fsfreeze-status"
 * Freeze guest with command "guest-fsfreeze-freeze"
 * Thaw guest with command "guest-fsfreeze-thaw"

Changes from V2:
 * Add some hook method which can inject test steps in
   guest fsfreeze testing.
   The methods are:
    - _action_before_fsfreeze
    - _action_after_fsfreeze
    - _action_before_fsthaw
    - _action_after_fsthaw
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

Create a new class QemuGuestAgentTest which contains
qemu guest agent installation/startup action.
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

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

The reboot variant is also test name. This leads to automatical runs
of this test when reboot test is specified. Changing this variant name
to reset should fix this issue.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Make it easier to apply and verify patches going on the
virt-test repo.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

The standalone_test script sets log directory for every
separated virt test now. But the log_line function in
utils_misc module can't work with this new directory
structure when there are multiple tests, it only checks
log file name, ignores file path info, and writes all
guest logs into the first virt test's log file.

This patch makes this log_line function check opened
file list with the full path of a file, to make log
msg are written into separated file for every test.
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

Signed-off-by: NShuping Cui <scui@redhat.com>

Currently, the commit test just use qemu_img, which
install in the system as default, to make the backup
file, and use the qemu_img_binary which get the path
from env file. Thus may bring some bugs.

For example, if just use the qemu source code and
don't make install to the system(that means the
system haven't installed a qemu to /usr/local/bin
or other system bin path, and autotest just test
the qemu using the git tree), the test will not
find the qemu_img_binary to create the backing_file.

and another problem is the binary used to backup and
commit may be different version, this may cause a bug.

To solve the problem, make the same binary to backup
and commit.
Signed-off-by: NMike Qiu <qiudayu@linux.vnet.ibm.com>

Those tools are more useful in the virt_test than they
are on autotest:

 * cd_hash.py: Tool to get disk md5/sha1 sums from files
 * regression.py: Tool to analyze regression results
 * virt_disk.py: Tool to create virtual cds and floppies
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

We might use 'iterations' parameters to execute one case
repeatedly, this patch makes the filename unique.
Signed-off-by: NAmos Kong <akong@redhat.com>

Expected result when opening virtconsole on windows is fail.
(actually it's dependent on the way we open it and currently
we use exclusive access so this test is not exactly related
to virtconsole, but to windows CreateFile function)
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Don't pass *args to Exceptions, use '%' instead.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Console version of test_rw_notconnect_guest used the second vio port,
but should use the first one instead. Test might fail in case VM
was recreated by the test.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The bus handling in virtio_ports is broken. This patch modifies
it to realy work as described:
1) manual port definition
2) virtio_port_spread
3) without definition (but create the first bus)
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This setting suits the current version of virtio_console loopback
test.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This test starts loopback between first 2 ports and do the interruption.
Than it verifies, whether new data still arrives and the loss/dup is
in the allowed range. Multiple interruption loops are supported.

Variants:
* stop and cont
* port replug (sender, receiver, random)
* hibernate (S4)
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch adds _reconnect_none_mode version of the loopback thread.
It works as ordinary none_mode, but additionally disconnects in case
the port is unreadable and reconnects when possible.

It was necessarily to make virt global variable to be able to access
it in this thread.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch adds new features to GuestWorker:
* read_nonblocking - read-out the output of guest script (used when
                     the test fails and we want to see the output)
* ThSendCheck - adds reduced_set option (use only A-Z, simplifies debuging
* ThRecvCheck - adds debug mode. This mode executes completely different
                loop. It's better for debuging purposes. Not all
                functionalities are kept so you should know when you want
                to use it.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>