On Sun, 2016-02-07 at 09:42 -0500, John Ferlan wrote:
> On 02/03/2016 03:26 PM, Andrea Bolognani wrote:
> > GIC is always available to ARM virt machines, and the domain XML should
> > reflect this fact.
> > ---
> > src/qemu/qemu_domain.c | 14 ++++++++++++++
> > .../qemuxml2argv-aarch64-aavmf-virtio-mmio.xml | 1 +
> > 2 files changed, 15 insertions(+)
> >
> > diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
> > index d120e15..5017cbb 100644
> > --- a/src/qemu/qemu_domain.c
> > +++ b/src/qemu/qemu_domain.c
> > @@ -1241,6 +1241,20 @@ qemuDomainDefAddDefaultDevices(virDomainDefPtr def,
> > static int
> > qemuDomainDefAddDefaultFeatures(virDomainDefPtr def)
> > {
> > + switch (def->os.arch) {
> > + case VIR_ARCH_ARMV7L:
> > + case VIR_ARCH_AARCH64:
> > + if (STREQ(def->os.machine, "virt") ||
> > + STRPREFIX(def->os.machine, "virt-")) {
> > + /* GIC is always available to ARM virt machines */
> > + def->features[VIR_DOMAIN_FEATURE_GIC] = VIR_TRISTATE_SWITCH_ON;
>
> See once on - we then have a version='host' and we're good to go.
>
> Of course, as I'm typing I realize that we wouldn't print out
> version='host' if it were the default... But that may not be a bad
> thing - although we could.
That was pretty much the issue :)
For existing guests, <gic/> means <gic version='2'/>, but we pick the
default every single time the XML is loaded instead of explicitly
writing it out in the XML. Which was probably a mistake, because now
we can't really change the default without affecting existing guests :(
On the other hand, QEMU defaults to version 2 as well, so it kinda
fits nicely I guess?
Point is, I think keeping the default to version 2 in libvirt and
have upper layers (eg. virt-install via libosinfo) pick a better
value when creating new guests is in line with how we do a lot of
other stuff in libvirt.
Cheers.
--
Andrea Bolognani
Software Engineer - Virtualization Team
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list