On 21.11.2022 11:21, Roger Pau Monne wrote:
> --- a/drivers/acpi/processor_pdc.c
> +++ b/drivers/acpi/processor_pdc.c
> @@ -137,6 +137,14 @@ acpi_processor_eval_pdc(acpi_handle handle, struct acpi_object_list *pdc_in)
> buffer[2] &= ~(ACPI_PDC_C_C2C3_FFH | ACPI_PDC_C_C1_FFH);
>
> }
> + if (xen_initial_domain())
> + /*
> + * When Linux is running as Xen dom0 it's the hypervisor the
> + * entity in charge of the processor power management, and so
> + * Xen needs to check the OS capabilities reported in the _PDC
> + * buffer matches what the hypervisor driver supports.
> + */
> + xen_sanitize_pdc((uint32_t *)pdc_in->pointer->buffer.pointer);
> status = acpi_evaluate_object(handle, "_PDC", pdc_in, NULL);
Again looking at our old XenoLinux forward port we had this inside the
earlier if(), as an _alternative_ to the &= (I don't think it's valid
to apply both the kernel's and Xen's adjustments). That would also let
you use "buffer" rather than re-calculating it via yet another (risky
from an abstract pov) cast.
It was the very nature of requiring Xen-specific conditionals which I
understand was the reason why so far no attempt was made to get this
(incl the corresponding logic for patch 1) into any upstream kernel.
Jan
On 21.11.2022 15:10, Jan Beulich wrote:
> On 21.11.2022 11:21, Roger Pau Monne wrote:
>> --- a/drivers/acpi/processor_pdc.c
>> +++ b/drivers/acpi/processor_pdc.c
>> @@ -137,6 +137,14 @@ acpi_processor_eval_pdc(acpi_handle handle, struct acpi_object_list *pdc_in)
>> buffer[2] &= ~(ACPI_PDC_C_C2C3_FFH | ACPI_PDC_C_C1_FFH);
>>
>> }
>> + if (xen_initial_domain())
>> + /*
>> + * When Linux is running as Xen dom0 it's the hypervisor the
>> + * entity in charge of the processor power management, and so
>> + * Xen needs to check the OS capabilities reported in the _PDC
>> + * buffer matches what the hypervisor driver supports.
>> + */
>> + xen_sanitize_pdc((uint32_t *)pdc_in->pointer->buffer.pointer);
>> status = acpi_evaluate_object(handle, "_PDC", pdc_in, NULL);
>
> Again looking at our old XenoLinux forward port we had this inside the
> earlier if(), as an _alternative_ to the &= (I don't think it's valid
> to apply both the kernel's and Xen's adjustments). That would also let
> you use "buffer" rather than re-calculating it via yet another (risky
> from an abstract pov) cast.
Oh, I notice this can end up being misleading: Besides having it in the
earlier if() we had also #ifdef-ed out that if() itself (keeping just
the body). The equivalent of this here might then be
if (boot_option_idle_override == IDLE_NOMWAIT || xen_initial_domain()) {
Jan