QEMU "info qtree" output depends on information generated by guest
firmware. One example is the bus number in PCI device address. QEMU
retrieves it from the PCI configuration space of the bridge which
creates the bus that a PCI device is attached to. So if test runs
"info qtree" before firmware finishes initializing pci devices, its
output can be incorrect.

The current code works around this timing issue by sleeping 5 seconds.
That is not reliable and doesn't work on my aarch64 system. So I'd
suggest to change the test to not start until it verifies the guest
has booted up. This will make the test execution time slightly longer
but is reliable.
Signed-off-by: NHuan Xiong <huan.xiong@hxt-semitech.com>

With the py3k changes some returns are not strings, but decoded
unicodes. Currently this mainly breaks monitor returns, where human
monitor gets recognized as qmp one. Let's use "six.string_types" that
provides the right string types for given python version.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NXu Han <xuhan@redhat.com>

Signed-off-by: NXu Han <xuhan@redhat.com>

Signed-off-by: NXu Han <xuhan@redhat.com>

The format of pci addr was changed to hex numbers with leading "0x" a
while ago, but this test was not adjusted resulting in reserved address
having unusable addresses. Let's adjust the format.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The timeout 10s was chosed to not slow-down the test execution much, but
usually less than 1s is enough. These days the boot takes < 20s,
sometimes even 5s so let's decrease the timeout to suitable but still
safe 5s.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "pci_bridge1" became widely used and it clashes with the default
name used by this test. Let's use "test_pci_%s" to emphasize it's
test-defined pci controller.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The make_create_command used to return "devices", but now it returns
"tuple(make_create_devices, spice_options)". Let's allow both by
checking the type.
Signed-off-by: NLukáš Doktor <ldoktor@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>

Clean up imports across the board in tp-qemu. The criteria
followed was:

1) Remove unnecessary imports
2) Turn multi imports from a single entry point to one
   import per line. This increases both patch and
   source code readability.
3) Separate blocks of imports with a line:
   * Python standard library imports
   * Autotest imports
   * virttest imports
   * Internal tp-qemu imports
4) Remove try/except blocks refering autotest libs and
   stick with the virttest.staging copy

This paves the way for the future removal of autotest
references, replacing those references with avocado ones.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

There's no need to have the name of the "run_<name>" method (main
test entry point) depend on the test name. So, for simplicity sake,
let's just name the test entry point "run".

I have decided to not fallback gracefully to the previous naming
convention as I believe this can lead to confusion in the future.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Make it possible to use ich9-usb-uhci controller for pci_devices testing
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch adds the possibility to use virtio_disks for testing.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Before "info qtree" returns the correct values, firmware initialization
has to finish. 10s should be enough (in case we won't define 1000
devices)
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Catching up with some problems found w.r.t. coding style
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Add some docstrings and other pylint fixes.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Additional check for (not only) correct pci addresses in qtree and lspci.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Split functions into subfunctions and prepare the test for multiple
parts with exception in the end.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

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

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

add_devices_all occupied all ports even when the bus was used for another
buses. This caused problems with occupied ports. This patch increases
the first_port number of the upper bus with each inserted bus thus the
bus could be plugged there.

The drawback is that buses which has some buses inside will have those
buses on the first addresses and the testing devices would not start with
addr=0x0. On the other hand missing bus would also test that the bus is
incapable of handling the first slot correctly.

Additionally this patch adds wait_for_login() in order to verify that
the VM booted correctly.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

1) change dest_fname to "autoinst.xml", as most cfg used.
2) change autoyast_param from floppy to 'autoyast=device://fd0/autoinst.xml'
    for old SLES OS compatibility.
3) add kernel_params for all existing SLES
Signed-off-by: NXiaoqing Wei <xwei@redhat.com>

This test boots VM with a define set of PCI devices and compares
the qtree and internal representation.

In the future this test should also check the operation system view
of the devices (lspci or the windows equivalent)

The testing setup is done via graphical representation of the setup.
Note that the number of spaces is irelevant. The location is defined
by the number of '->' characters. You can verify the setup in the test.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>