Il 15/03/2013 20:35, Matthew Garrett ha scritto:
> On Thu, 2013-03-14 at 16:51 +0800, Hu Tao wrote:
>
>> + { "MSFT0001", 0},
>
> This seems wrong, and it looks like qemu agrees. Can you resubmit when
> there's agreement on the name?
>
>> + acpi_evaluate_object(handle, "WRPT", &arg_list, NULL);
>
> Is there a spec for this?
Not yet. Hu, in the next version of your QEMU patches you probably will
not need anymore usage information in docs/pvevent.txt. Instead, please
add a spec for both the ISA and ACPI interfaces in docs/spec/pvevent.txt.
> Is the only reason for this to allow guests to notify the host that
> they've panicked? It seems like making use of pstore to push the crash
> dump to the host. as well would be a useful thing to do.
Since we try to limit the amount of VM-specific interfaces we have
(yeah, I know this is one), that would mean implementing APEI in QEMU, I
guess. It is definitely a useful thing to have, but a bit wider in
scope than a simple device to distinguish idle and crashed VMs.
Paolo
