Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

There is no reason to hard-code tests.cfg for --list-tests.
With a custom config we better don't strip the shortnames though
as we don't know whenever the first six elements of the shortname
are really identical.  So don't do that.

[ Note: We could try something more clever like figuring whenever
        all tests have a common prefix and if that happens to be
        the case strip that.  I'll leave that as excercise for
        someone else though ]
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

I can't find a reason why combining --config and --tests is disallowed.
Picking a subset of the tests on the command line, without having to
edit my custom cartesian config file, is a useful feature.
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

The test runner applies a default test selection in case --tests is not
specified.  That is a pretty pointless in case a custom cartesian config
is specified, so don't do that.
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

It's not in the repos (or at least not on the DVD image), thus
trying to run a unattended Fedora 16 install fails because the
packet is missing.
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

Currently, in hmp monitor, try 'info qtree' it will display
some infomation, but as    dev: spapr-nvram, id "nvram@1001"
                                reg = 4097
                                drive = <null>
                                irq = 4098
This drive=<null> will cause a lot errors in qtree related test.
and if drive=<null> , it will not map to a disk in 'info block',
and it is not a bug. So add that as a exception.
Signed-off-by: NMike Qiu <qiudayu@linux.vnet.ibm.com>

xml_utils were moved from autotest to virt-test,
given that it was the only user of the code.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

If a vm is already running when autotest starts and there is no
env or address_pool state, vm.get_address() is unable to obtain
a mac address (it isn't stored anywhere).  However, this information
is retrievable, from the command-line or libvirt vm's XML.

Checking the xml, a secondary XML parsing bug was reveiled in the
libvirt vm.get_virsh_mac_address method.  Switching to using xml_utils
resolved this. As long as a DHCP request comes in (to update
address_cache) a libvirt test is able to continue.

In order for this to work with KVM, some mechanism is needed
to retrieve the mac address from the already running qemu-kvm
process.  This is beyond the scope of this fix.
Signed-off-by: NChris Evich <cevich@redhat.com>

+ Reorganized code/config with result-based logic
+ Added code/config to allow older libvirt failure testing
+ Added / Fixed typos on a few comments
Signed-off-by: NChris Evich <cevich@redhat.com>

+ Fixed up imports
+ Fixed up a few comments
+ Added has_command_help_match
+ Added unittest for has_command_help_match
+ Updated setmem method to accept all combinations of parameters
Signed-off-by: NChris Evich <cevich@redhat.com>

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

Multiple migration of VM between master and
slave with different types of stress (no stress,
disk stress, memory stress, both).
Signed-off-by: NJiří Župka <jzupka@redhat.com>

1) adds support for continuous memory fill in time in cpuflags-test stressmem
2) limits usage of memory for stressmem.
    --memstress memory_usage[,memory_fill]
      it fills memory_fill amount of memory by memory_usage parts.
      rounds = memory_fill/memory_usage
      roundtime = 1/rounds
      for rounds:
         fill(memory_usage)
         sleep(roundtime-time) - rest of time to next fill roundvirt: Improve cpuflags-test stressmem util.

1) adds support for spread memory fill in time in cpuflags-test stressmem
2) limit usage memory for stressmem.
    --memstress memory_usage[,memory_fill]
      it fills memory_fill amount of memory by memory_usage parts.
      rounds = memory_fill/memory_usage
      for rounds:
         fill(memory_usage)
         sleep(rest of time to next fill round for spread filling over time)
Signed-off-by: NJiří Župka <jzupka@redhat.com>

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

virtconsole uses virtserialport and creates console on top of it.
It creates an additional layer between the port and the user
and as the result it handles certain things differently, like
Non-blocking mode and not-connected-host are situations, which are
not handled and there is currently no plan of when and how it will
be handled as console is always connected to virtserialport and
console always uses blocking mode to talk to virtserialport.
That's why it doesn't make sense to run tests and pretend it
should act as virtserialport.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Background threads of Th*Check might be blocked. Even use of
InterruptableThread() from base_utils didn't help.
Anyway in this scenarios threads are usually unblocked by cleaning
the ports so it's usually safe to wait till the test end and
recheck, whether the thread finished or not.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

In fact, the parent dir of the data dir can be safely
used even when virt-test is running inside autotest.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Patch 65080193 introduce os_arch_machine_map for checking supported
guest type. However, libvirt uses 'xen' as name for XEN PV guests so
test fails as we use 'pv' for this type of guests.

This patch map xen type to pv as we use this value on various place
(mainly in unattended install) for XEN PV related quirks.
Signed-off-by: NMiroslav Rezanina <mrezanin@redhat.com>

Some of the params can be changed when we setup unattended install.
However, updating params is not done properly in every case - code
mix update only internal variables and params dictionary update.

This patch ensure that all changes done during setup are properly
reflected in params dictionary.
Signed-off-by: NMiroslav Rezanina <mrezanin@redhat.com>

.git can be also a file, in case virt test is a submodule.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

We want to make the test runner run the default test
set to root, so set nettype for the base test cfg
file to 'user', which will make qemu to use slirp
instead of bridge.

I made a mistake on a previous patch that went to master,
setting netttype = it in the base cfg file, so this patch
reverts that change too.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

When running virt test outside the autotest context,
don't try to find the guest console helper script
in autotest.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

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

The command output sometimes include the command line itself which will
cause the results get wrong. So just get the last 4 value which should
be the real data we wanted.
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

Changes from v1:
Delete client_path = "D:\\netperf.exe"
Changes from v2:
Make windows could be client or host.
Add host sub parameters.
Signed-off-by: NFeng Yang <fyang@redhat.com>

variants:
    - netperf_exe:
        variants:
           - default_setting:
           - best_registry_setting:
    - netperf_cygwin:

Changes from v1:

Use telnet for win2008, win2008r2 and win2003 for
best_registry_setting case. netsh int tcp set global
autotuninglevel=disabled command always fails
with nc in these guests.
Signed-off-by: NFeng Yang <fyang@redhat.com>

We have netserver.exe now, so make netserver.exe works in netperf_win.
Also correct remote.wait_for_login.

changes from v1:
  - commits and log message clean up
Signed-off-by: NFeng Yang <fyang@redhat.com>

In some guest the ssh_cmd will return the commmand line in output
too which may cause the problem for find cpu number. So only take
the numbers in the output.
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

Run netserver in windows guest and the TCP scenarios in localhost.

changes from v1:
  - Add some log
  - typo error fix
  - configure parameters fix
changes from v2:
  - modify the parameters to the default test env
changes from v3:
  - format and commit clean up
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

Add the command line for system that don't support numactl.

changes from v1:
  - modify to fix the change from pre patches
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

As some ssh version in different OS may not support the ssh for
command execute.

changes from v1:
  - Add session for netperf_thread
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

In different version mpstat the parameters show in results are
different. Use the name of the parameter to get the index of
CPU use status related parameter from the command output head.
This will avoid the wrong cpu values calculated in hosts with
different mpstat.
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

Update the record function in netperf to let it can records
different kind of results according the result dict.

changes from v1:
  - params and log fix
Signed-off-by: NYiqiao Pu <ypu@redhat.com>

CC: Amos Kong <akong@redhat.com>
Signed-off-by: NQingtang Zhou <qzhou@redhat.com>

The code was missing the proper domstate namespace and
was passing a non existing **dargs dict.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

virsh.restore only takes the savefile path as an argument,
let's fix it.

Also, let's fix a redundant import and some inconsistent
comments.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

New tests:
  1) change downtime of migration after start of migration
  2) change migration speed after start of migration.
  3) cancel migration after start.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

It allows use migration over fd in all multihost_migration tests.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

If force_image_clone == no then only check vm image consistency.
If vm image consistency is good use old image else copy again from original.
Remove image if consistency is not correct after finish of test.
Signed-off-by: NJiří Župka <jzupka@redhat.com>

1) Start migration with stressed VM.
2) Wait cancel_delay and then cancel migration.
3) Check if machine is alive on source host.
4) Stop stress of VM.
5) Migrate machine again.
6) Check if machine is alive on destination host.
Signed-off-by: NJiří Župka <jzupka@redhat.com>