[Skiboot] [PATCH V6 3/3] occ: Add support for Version 0x90 OCC_OPAL shared memory region

[Vaidyanathan Srinivasan]

* Shilpa Bhat <shilpa.bhat at linux.vnet.ibm.com> [2017-02-02 12:45:57]:

> Hi,
> 
> On 02/02/2017 10:33 AM, Benjamin Herrenschmidt wrote:
> > On Thu, 2017-02-02 at 16:01 +1100, Stewart Smith wrote:
> >>> Shilpasri G Bhat <shilpa.bhat at linux.vnet.ibm.com> writes:
> >>> This patch enables pstate table parsing support for P9. It
> >>> introduces below device tree changes.
> >>> - Add a new node per chip in /ibm,opal/power-mgt called occ.
> >>>>>        occ at 3ffd9f8000 {
> >>>                reg = <0x3f 0xfd9f8000>;
> >>>                ibm,chip-id = <0x0>;
> >>>                ibm,pstate-vdds = <0x4e4e4f4f 0x4f505152 0x53545556
> >>>                0x5758595a 0x5b5c5d5e 0x5f606162 0x63646565 0x66676868
> >>>                0x696a6a6b 0x6c6d6d6e 0x6f6f7071 0x72727374 0x74757677
> >>>                0x7778797a 0x7a7b7c7c>;
> >>>                ibm,pstate-vcss = <0x4446484a 0x4c4e4f50 0x50515253
> >>>                0x54555556 0x5758595a 0x5a5b5c5d 0x5d5e5e5e 0x5e5f5f5f
> >>>                0x5f606060 0x61616161 0x62626263 0x63636364 0x64646465
> >>>                0x65656666 0x66666767>;
> >>>                ibm,pstate-core-max = <0x0 0x0 0x0 0x0 0x0 0x0 0x0 0x0>;
> >>>                phandle = <0x10000326>;
> >>>        };
> >>
> >> Ben - any thoughts on device tree layout for ^ ?
> > 
> > Yup, not too sure. Looks ok but what is the core-max 0 means ?
> 
> 'core-max' property holds good for WOF enabled systems where the max pstate of
> the chip varies with the number of active cores.
> Here core-max[0] = 0 points out the max pstate that can be attained by the chip
> with 1 active core. Similarly core-max[1] is with 2 active cores and so on.
> 
> The above example is from WOF disabled system where the Pmax for N active cores
> is constant which is max turbo pstate.

Max pstate ID that can be achieved for given number of active core is
provided by OCC like this.  We actually need to be able to extend the
device tree property since in future different levels of active vs
idle state can also be captures in independent array elements.

Pointing to pre defined array is a good idea since the information
will more or less be same.

> Thanks and Regards,
> Shilpa
> 
> > 
> >> I'm wondering if we need it per chip or if the information in the
> >> ibm,pstate-* is mostly common, and we should just point to it rather
> >> than have a copy per chip?
> > 
> > Maybe. Can they be different ? Vaidy ?

The vcs and vdd voltage IDs are unique per-chip.  The
ibm,pstate-core-max is practically same for all chips in the system,
but the information as per spec is provided on a per-chip basis ;)

Lets move this one out to common area and have only vcs/vdd within the
per-occ device-tree node.

--Vaidy