On 10/15/14 15:55, Michal Privoznik wrote:
On 14.10.2014 09:42, Cole Robinson wrote:
> It's supported on aarch64 and armv7l as well, so just drop the
> restriction
> entirely since it doesn't add much.
> ---
> src/qemu/qemu_command.c | 8 --------
> 1 file changed, 8 deletions(-)
>
> diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
> index 8cb0865..2872e47 100644
> --- a/src/qemu/qemu_command.c
> +++ b/src/qemu/qemu_command.c
> @@ -7672,14 +7672,6 @@ qemuBuildDomainLoaderCommandLine(virCommandPtr
> cmd,
> break;
>
> case VIR_DOMAIN_LOADER_TYPE_PFLASH:
> - /* UEFI is supported only for x86_64 currently */
> - if (def->os.arch != VIR_ARCH_X86_64) {
I think it would be better to deliberately allow arches we want. As far
as I understand the cmd line generated to use UEFI can vary depending on
the guest architecture, is that right Laszlo? If it is so, ACK to the
opposite patch :)
> - virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
> - _("pflash is not supported for %s guest
> architecture"),
> - virArchToString(def->os.arch));
> - goto cleanup;
> - }
> -
> if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_DRIVE)) {
> virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> _("this QEMU binary doesn't support
> -drive"));
>
Michal
Sorry about the late response.
I wasn't CC'd on either this patch or the "opposite" one. But, I think
I
agree with Michal -- enumerating architectures that are allowed to use
UEFI seems safer at this point than a whole-sale enablement (and also
safer than enumerating the architectures that are *not* allowed to use
UEFI).
So, default to "off", list the arches one by one that are allowed to use
UEFI, and (if necessary) customize the command line / pflash switches
for the given arch.
Thanks
Laszlo