[PATCH v4 2/2] dt: power: st: Provide bindings for ST's OPPs

[Rob Herring]

On Thu, Jul 30, 2015 at 3:46 AM, Lee Jones <lee.jones at linaro.org> wrote:
> On Wed, 29 Jul 2015, Stephen Boyd wrote:
>
>> On 07/29, Lee Jones wrote:
>> > On Tue, 28 Jul 2015, Stephen Boyd wrote:
>> >
>> > > On 07/28, Viresh Kumar wrote:
>> > > > Cc'ing few people (whom I cc'd last time as well :)).
>> > > >
>> > > > On 27-07-15, 16:20, Lee Jones wrote:
>> > > > > + - opp-hz            : CPU frequency [Hz] for this OPP [See: ./opp.txt]
>> > > > > + - st,avs            : List of available voltages [uV] indexed by process code
>> > > > > + - st,cuts           : Cut version this OPP is suitable for [0xFF means ALL]
>> > > > > + - st,substrate              : Substrate version this OPP is suitable for [0xFF means ALL]
>> [...]
>> > > > > +cpu0-opp-list {
>> > > > > +     compatible      = "operating-points-v2-sti";
>> > > > > +     st,syscfg       = <&syscfg [major_offset]>;
>> > > > > +     st,syscfg-eng   = <&syscfg_eng [pcode_offset] [minor_offset]>;
>> > > > > +
>> > > > > +     opp0 {
>> > > > > +             opp-hz          = <1200000000>;
>> > > > > +             st,avs          = <1110 1150 1100 1080 1040 1020 980 930>;
>> > > > > +             st,substrate    = <0xff>;
>> > > > > +             st,cuts         = <0xff>;
>> > > > > +     };
>> > > > > +     opp1 {
>> > > > > +             opp-hz          = <1500000000>;
>> > > > > +             st,avs          = <1200 1200 1200 1200 1170 1140 1100 1070>;
>> > > > > +             st,substrate    = <0xff>;
>> > > > > +             st,cuts         = <0x2>;
>> > > > > +     };
>> > > > > +};
>> > > >
>> > > > I don't see more problems here, unless we can move some of this to the
>> > > > generic bindings.
>> > > >
>> > > > @Rob/Stephen: Please respond before it is late :)
>> > >
>> > > It's interesting to have vendor specific properties like avs,
>> > > cuts, and substrate. That could replace our planned usage of the
>> > > opp-names property where we encode similar information (speed
>> > > bin, revision, etc.) into a string that we look for.
>> > >
>> > > So I wonder why the avs/cut/substrate information can't be
>> > > encoded into the opp name? That would make these properties
>> > > obsolete, given that all they're used for is to pick out the
>> > > correct OPP?
>> >
>> > You could hack the substrate and cut version into a string, but that's
>> > exactly what it would be, a hack.  I'm struggling how you would do the
>> > same for 'st,avs', which is an array of u32s.
>> >
>>
>>
>> (I don't understand the st,avs property to begin with, so feel
>> free to ignore the rest of this mail.)
>>
>> For qcom platforms we have the pvs bin and speed bin, and
>> sometimes a revision number. Each one of these properties
>> corresponds to a different set of OPPs (opp table). So we might
>> have speed1-pvs2-v0 for speed bin 1, pvs bin 2 and version 0. We
>> fill out an opp table for this and then point the opps property
>> at the table and have a corresponding opp-name "speed1-pvs2-v0"
>> in the "consumer" node.
>>
>>       operating-points-v2 = <&speed1_pvs2_v0>;
>>       operating-points-names = "speed1-pvs2-v0";
>>
>> We have quite a few of these tables because the values are
>> always different. If the values were the same then we could use
>> the same table with different names I suppose, but we're not
>> doing that.
>>
>> From a quick read of the st properties (that I admit I don't
>> understand), it looks like we're trying to compress the OPP
>> tables by listing all the voltages that could be used for a
>> particular frequency depending on which avs is present on the
>> device? And then limiting the frequency voltage pairs depending
>> on which cut and substrate is present?
>>
>> So we'd probably have to expand out the tables to be unique per
>> avs/cut/substrate parameter. Something like:
>>
>> avs0-cut2 {
>>       compatible = "operating-points-v2";
>>
>>       opp0 {
>>               opp-hz = <1200000000>;
>>               opp-microvolt = <1100>;
>>       };
>>
>>       opp1 {
>>               opp-hz = <1500000000>;
>>               opp-microvolt = <1200>;
>>       };
>> };
>>
>> avs1-cut2 {
>>       compatible = "operating-points-v2";
>>
>>       opp0 {
>>               opp-hz = <1200000000>;
>>               opp-microvolt = <1150>;
>>       };
>>
>>       opp1 {
>>               opp-hz = <1500000000>;
>>               opp-microvolt = <1200>;
>>       };
>> };
>>
>> And then another copy of these for the devices without cuts != 2
>> where the top frequency is gone?
>
> There is nothing stopping us from representing the data in this way.
> On the plus side, it would mean that we wouldn't need any vendor
> specific properties.  However, far outweighing the positives are the
> fact that, even in our very simple example provided, where we only
> have 2 frequencies, differ between only 1 cut and support all
> substrates, we would still need 16 OPP tables.  When any one of those
> variables increase (and they will), we would then have a large number
> of permutations and subsequently and unruly amount of OPP tables.
>
> (... and we haven't even provided the body-biasing information yet.)
>
> The way we've chosen to represent our circumstance is the least
> intrusive and the most succinct way we can think of.  Which IMHO
> outweighs the fact that we have to introduce a couple of vendor
> properties by a country mile.

Regardless of which is better or worse, you are both doing essentially
the same thing. You are selecting OPPs based on some Si parameters. We
should not really be doing this 2 different ways. I'd be fine with 2
ways if it was 2 for all SOCs, but right now it is 2 for 2 SOCs.
Really, I'd like to see most if not all the selection or fixup of OPPs
be done in the bootloader and the kernel just deals with the correct
OPP table.

Where are you storing the data that gets selected to fill into these
properties? Is it just look-up tables or is there some kind of
algorithm to generate the values? If the former, then DT is as good a
data struct as any to store the tables. There is a lot of duplication
though with only a single property varying in each set. If you both
have that problem, then perhaps we can come up with a generic way to
list all possible values more concisely.

Rob