On Thu, Apr 27, 2017 at 09:42:24AM +0200, Andrea Bolognani wrote:
On Wed, 2017-04-26 at 18:59 +0200, Pavel Hrdina wrote:
> > If qemu-xhci is not available but usb-nec-xhci is, it would
> > probably be nicer to pick the latter rather than leaving it
> > unspecified (which will result in an error later on).
>
> This is not true, if the qemu-xhci is not available libvirt will set
> piix3-uhci if available as a default for aarch64. Only in case that
> piix3-uhci is not available it will fail.
You're technically correct[1]. However, piix3-uhci is
another piece of Intel-derived hardware so in practice
qemu-system-aarch64 is very unlikely to have it compiled
in and most users will end up getting the error instead.
Isn't the nec-xhci also Intel hardware, so the same would apply to that
controller as well. Moreover, this probably happens only for downstream
builds of QEMU (most likely only RHEL/CentOS) as there is no configure
option for that. QEMU has some default configs for different architectures
but in upstream QEMU the set of UHCI usb controllers is enabled by default
for aarch64.
> > So I would basically use the same code as ppc64 here,
minus
> > the pci-ohci part of course.
>
> That should be addressed by a separate patch.
Agreed. I suggest adding a patch at the beginning of the
series that makes nec-xhci the default for aarch64 if
available as in ppc64, and then make qemu-xhci the new
default for both in this patch. Would that work for you?
I don't mind if we add the same logic for ppc64 and aarch64, so I'll send
a v2 with this change included.
Thanks for review.
Pavel
[1] Which is the best kind of correct! ;)
--
Andrea Bolognani / Red Hat / Virtualization
--
libvir-list mailing list
libvir-list(a)redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list