On Wed, Sep 03, 2014 at 01:23:21PM +0200, Arnd Bergmann wrote:
On Tuesday 02 September 2014 16:15:05 Ashwin Chaugule wrote:
Right. At least not in a way DT does. PCC clients know if something needs to be written/read via PCC mailbox and can identify a PCC subspace. (i.e. Mailbox channel). The PCC mailbox is uniquely identified/defined in the spec.
#define ACPI_ADR_SPACE_PLATFORM_COMM (acpi_adr_space_type) 10
So we could use this ID instead of a string and use that to look up the PCC controller for a PCC client.
I didn't realize this was the case. Does that mean we can treat pcc as a linearly accessible address space the way we do for system memory, pci-config etc?
If that works, we should probably just have a regmap for it rather than expose the mailbox API to client drivers.
A regmap doesn't seem to map very well here - as far as I can tell the addresses referred to are mailboxes rather than registers or memory addresses. I could be misunderstanding though.