Don't rely on libvirt's default. This makes any XML changes more
explicit, and can help other parts of the code that may depend on
a machine type being encoded
We shouldn't concern ourselves with whether the OS supports qemu_ga
out of the box, just set up the channel regardless, as long as
virtio-serial is supported.
This adds the channel for some aarch64 cases which didn't have OS
specified, but that's desired IMO
This case will still work, but be a bit slower, which is fine. Nowadays
-M virt is much better for virt usage and nearly everyone is using that,
so save us the complication. This was really only useful when
bootstrapping arm virt support
Let libvirt/libxl fill it in. This was required for rhel5 xen at
least, but we don't support that anymore and modern libvirt seems
to do the right thing and fill it in for us.
The validation is already handled by libvirt, and setting q35 for
smm=on is overkill and just hardcoding some libvirt logic here.
I think the 'secboot' detection in Guest.py is the preferred
magic here, otherwise let users specify all the correct values.
An emulated backend doesn't require any path, since libvirt will take
care of finding the emulator and managing the storage. However, the
version to emulate can be specified.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Add codec support to virt-install so that it can accommodate
multiple instances of codec configuration.
The commandline argument:
--sound codec0.type=micro,codec1.type=duplex,codec2.type=output
maps to the sound XML below:
<sound model="es1370">
<codec type="micro"/>
<codec type="duplex"/>
<codec type="output"/>
</sound>
Signed-off-by: Anya Harter <aharter@redhat.com>
Lookup the domain capabilities CPU model and compare with
the host capabilities CPU model and if they are not equal
set the guest's CPU model to None.
(crobinso: compare against 'custom' list not 'host-model', move
to separate function)
non-treeinfo redhat only applies to pre RHEL5.4 and very old
Fedora. It's not worth it anymore to slow down all URL lookups
and maintain code complexity to handle such long out of date
distros.
GenericDistro doesn't actually apply to any public trees that I
can find, except for some with TreeInfo. So turn it into
GenericTreeinfoDistro. If random URL trees want to work with
virt-install, add a treeinfo file
Right now, most test cases will create a libvirt test driver using
tests/testdriver.xml. This is problematic for 2 reasons:
1) testdriver.xml is 3500 lines of XML, and it's parsed hundreds of
times. Opening it is responsible for about 25% of test suite time
2) Older libvirt chokes on testdriver.xml, meaning the test suite will
be useless there. This is a recurring problem as new features are
added to testdriver.xml
Most test cases don't actually need all the test state in testdriver.xml
though. So this creates a smaller testsuite.xml which has a lower
libvirt requirement and is much quicker to parse. New XML bits should
continue to go into testdriver.xml so we can keep testsuite.xml as
the more stripped down option.
Moves the libxml2 bits to a separate xmlapi file and class, a bunch
of cleanups to xmlbuilder internals dealing with XML stuff.
The main point is to experiment with different XML library impls,
since libxml2 is unfun to deal with and we are having python3
issues like
https://bugzilla.gnome.org/show_bug.cgi?id=776815
Add vcpupin support to virt-install so that it can create guest
domains with statically allocated vcpu pinning towards a given cpuset.
Syntax: to pin vcpu=0 to cpuset="1,3" and vcpu=1 to cpuset=2
--cputune vcpupin0.vcpu=0,vcpupin0.cpuset=1,3,vcpupin1.vcpu=1,vcpupin1.cpuset=2
generates below XML description for the guest domain.
<cputune>
<vcpupin vcpu="0" cpuset="1,3"/>
<vcpupin vcpu="1" cpuset="2"/>
</cputune>
Signed-off-by: Wim ten Have <wim.ten.have@oracle.com>
Signed-off-by: Menno Lageman <menno.lageman@oracle.com>
Now that libvirt has support for administration of distances between NUMA cells
it would be nice to be able to set those with virt-install directly instead of
having to 'virsh edit' the domain XML manually after installation.
For example
--cpu cell0.memory=1234,cell0.cpus=0-3,cell1.memory=5678,cell1.cpus=4-7,\
cell0.distances.sibling0.id=0,cell0.distances.sibling0.value=10,\
cell0.distances.sibling1.id=1,cell0.distances.sibling1.value=21,\
cell1.distances.sibling0.id=0,cell1.distances.sibling0.value=21,\
cell1.distances.sibling1.id=1,cell1.distances.sibling1.value=10
would generate the following XML:
<cpu>
<numa>
<cell cpus="0-3" memory="1234">
<distances>
<sibling id="0" value="10"/>
<sibling id="1" value="21"/>
</distances>
</cell>
<cell cpus="4-7" memory="5678">
<distances>
<sibling id="0" value="21"/>
<sibling id="1" value="10"/>
</distances>
</cell>
</numa>
</cpu>
Signed-off-by: Menno Lageman <menno.lageman@oracle.com>
(crobinso: rework cli format, drop some validation, drop man changes)
libvirt supports guest CPU cache by commit df13c0b, So add this feature
to virt-install to configure cpu L3 cache mode.
Currently, The valid values are 'passthrough', 'emulate' or 'disable'.
say:
--cpu host-passthrough,cache.mode=passthrough
or
--cpu $CPU,cache.mode=emulate,cache.level=3
or
--cpu $CPU,cache.mode=disable
Signed-off-by: Lin Ma <lma@suse.com>
Using "listens.*" allows better configuration of listen elements for
graphics devices. Currently the only way how to configure a listen
type is to abuse "listen" parameter and there is no way how to configure
exact "network".
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
The default port and tlsPort should be configured only if no listen
type was specified or the listen type is "address" or "network".
For other listen types the port and tlsPort doesn't make sense.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Cole Robinson
Marc-André Lureau
Anya Harter
Charles Arnold
Wim ten Have
Rauno Väli
Menno Lageman
Scott Garfinkle
Lin Ma
Pavel Hrdina