Re: [PATCH V2 03/10] irqchip,GICv3,ACPI: Add redistributor support via GICC structures.

From: Tomasz Nowicki
Date: Tue Jan 12 2016 - 12:16:34 EST


On 12.01.2016 17:16, Marc Zyngier wrote:
>>Also, the spec says:
>>
>>"On systems supporting GICv3 and above, this field holds the 64-bit
>>physical address of the associated Redistributor. If all of the GIC
>>Redistributors are in the always-on power domain, GICR structures should
>>be used to describe the Redistributors instead, and this field must be
>>set to 0."
>>
>>which triggers two questions:
>>- Can you access always the GICR_TYPER register without waking the
>>redistributor up?
>
>I missed this part, can you suggest how can we do that? accessing some
>register before access to redistributor?
This redistributor may be in a power-domain that is off. Are you
guaranteed that you can access GICR_TYPER even when it is off?

>
>>- How do you cope with situations where some redistributors are in the
>>always-on domain, and some are not?
>
>I'm not sure if there is such hardware, if yes, do we need to fix
>the spec first?
It is something that should definitely be clarified. Can we end-up in a
situation where some redistributors are described via the GICR
structure, and some via the GICC structure? The spec is a bit ambiguous.


To recap:
1. GICR and GICC subtables should be mutually exclusive, but ACPI spec should be clear about this.
2. We need to know if we can access GICR_TYPER without waking the redistributor up.

Tirumalesh, can you please advice?

Tomasz