On 03/10/2023 00:03, Stefano Stabellini wrote:
On Mon, 2 Oct 2023, Julien Grall wrote:
On 29/09/2023 21:29, Stefano Stabellini wrote:
Now regarding whether the PPI is used. AFAICT, the secure timer PPI is still
present in the firmware tables (ACPI and DT) passed to dom0. So strictly
speaking we want to ensure the PPI value is reserved.
That said, the ACPI spec suggests that the value will be ignored by the guest.
The Device-Tree binding doesn't have such statement, but I suspect this may be
the same. So it should be ok to skip reserving the PPI and therefore allow the
event channel interrupt to use if it is not reserved by someone else.
You looked into this more deeply then I did. Your suggestion makes sense
to me.
I have sent a patch [1]. This is not quite the same as I discussed
before. See the commit message for more details.
Cheers,
[1] https://lore.kernel.org/[email protected]
--
Julien Grall
