Re: [PATCH 1/4] mailbox: add support for System Control and Power Interface(SCPI) protocol

[Sudeep Holla <sudeep.holla@arm.com>]

On 14/05/15 08:30, Jassi Brar wrote:
> On Thu, May 14, 2015 at 12:32 PM, Jassi Brar <jassisinghbrar@gmail.com> wrote:
>>
>> BTW is scpi_protocol.c meant/tested to work over arm_mhu.c? The spec
>> says so but I don't see how because you pass 'struct scpi_xfer*' as
>> the message whereas arm_mhu.c expects u32*
>>

Yes it's tested using arm_mhu.c, and I have even sent updates to the
binding that's incomplete as of now and *must* be pulled into v4.1.
Please make sure it gets in. Otherwise clocks are optional but the
driver fails to probe without that. I was initially wondering why the
MHU probe is not called.

scpi_xfer has the slot as first element which will have the right
doorbell bit(in this case slot#0) set always.

> It seems your remote doesn't interpret the value in STAT register...
> so it just works.

Not exactly. If you look at Figure 2-1 in the spec, it shows how STAT
(along with SET/CLEAR) is used to identify the protocol. The remote
can implement multiple protocol. E.g. SCPI(on Slot#0 - main topic of
this series), ACPI(PCC/CPPC say on Slot#1), ABC(on Slot#X)...etc.

> However the SCPI spec recommends seeing STAT register as '31 slots'
> ... maybe we should try to support that.
>

Correct but only Slot#0 is used for SCPI.

Regards,
Sudeep