On 14.08.2025 12:35, Roger Pau Monné wrote:
> On Thu, Aug 14, 2025 at 09:21:53AM +0200, Jan Beulich wrote:
>> On 13.08.2025 11:32, Roger Pau Monne wrote:
>>> For x86 PVH dom0 makes uses of the host ACPI tables, and hence the status
>>> of Host ACPI entry should have been updated to reflect that PVH dom0 is now
>>> supported, and Host ACPI support is not one of the caveats.
>>
>> Well, not explicitly, but seeing changes like 6378909b41c4 ("x86/PVH: expose
>> OEMx ACPI tables to Dom0") I'm not convinced we're there yet.
>
> How would we know when we are there?
That's the downside of the whitelisting approach: We may make it there
temporarily, but then a new table type may break us again.
> I think listing it here means it
> had some degree of testing (which it does in Gitlab CI), plus that we
> will address issues and backport fixes (which to my knowledge we
> already do).
We do, yes. We did already before it was supported, to encourage its use.
> AFAICT the only way x86 PVH dom0 can get it's hardware information is
> from ACPI, so not listing it as supported is IMO weird, it's not like
> PCI passthrough or some additional feature, without ACPI support you
> cannot really boot a PVH dom0 I think.
I agree. And from that angle I also agree it ought to be marked as
supported. I really don't have a good suggestion here. My concern, in
any event, shouldn't stand in the way of this going in with someone
else's ack (e.g. Stefano's, that iirc you already got). Yet I'm
hesitant to provide one myself.
Jan
