Message-ID: <61f05dca.1c69fb81.82753.b5a4@mx.google.com>
Date:   Tue, 25 Jan 2022 21:30:01 +0100
From:   Ansuel Smith <ansuelsmth@...il.com>
To:     Rafał Miłecki <zajec5@...il.com>
Cc:     Miquel Raynal <miquel.raynal@...tlin.com>,
        Richard Weinberger <richard@....at>,
        Vignesh Raghavendra <vigneshr@...com>,
        Rob Herring <robh+dt@...nel.org>,
        linux-mtd@...ts.infradead.org, devicetree@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH 1/2] dt-bindings: mtd: partitions: Document new
 dynamic-partitions node

On Tue, Jan 25, 2022 at 09:21:04PM +0100, Rafał Miłecki wrote:
> On 24.01.2022 23:12, Ansuel Smith wrote:
> > On Mon, Jan 24, 2022 at 11:02:24PM +0100, Rafał Miłecki wrote:
> > > On 20.01.2022 21:26, Ansuel Smith wrote:
> > > > Document new dynamic-partitions node used to provide an of node for
> > > > partition registred at runtime by parsers. This is required for nvmem
> > > > system to declare and detect nvmem-cells.
> > > > 
> > > > Signed-off-by: Ansuel Smith <ansuelsmth@...il.com>
> > > > ---
> > > >    .../mtd/partitions/dynamic-partitions.yaml    | 59 +++++++++++++++++++
> > > >    1 file changed, 59 insertions(+)
> > > >    create mode 100644 Documentation/devicetree/bindings/mtd/partitions/dynamic-partitions.yaml
> > > > 
> > > > diff --git a/Documentation/devicetree/bindings/mtd/partitions/dynamic-partitions.yaml b/Documentation/devicetree/bindings/mtd/partitions/dynamic-partitions.yaml
> > > > new file mode 100644
> > > > index 000000000000..7528e49f2d7e
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/mtd/partitions/dynamic-partitions.yaml
> > > > @@ -0,0 +1,59 @@
> > > > +# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
> > > > +%YAML 1.2
> > > > +---
> > > > +$id: http://devicetree.org/schemas/mtd/partitions/dynamic-partitions.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: Dynamic partitions
> > > > +
> > > > +description: |
> > > > +  This binding can be used on platforms which have partitions registered at
> > > > +  runtime by parsers or partition table present on the flash. Example are
> > > > +  partitions declared from smem parser or cmdlinepart. This will create an
> > > > +  of node for these dynamic partition where systems like Nvmem can get a
> > > > +  reference to register nvmem-cells.
> > > > +
> > > > +  The partition table should be a node named "dynamic-partitions".
> > > > +  Partitions are then defined as subnodes. Only the label is required
> > > > +  as any other data will be taken from the parser.
> > > > +
> > > > +maintainers:
> > > > +  - Ansuel Smith <ansuelsmth@...il.com>
> > > > +
> > > > +properties:
> > > > +  compatible:
> > > > +    const: dynamic-partitions
> > > > +
> > > > +patternProperties:
> > > > +  "@[0-9a-f]+$":
> > > > +    $ref: "partition.yaml#"
> > > > +
> > > > +additionalProperties: true
> > > > +
> > > > +examples:
> > > > +  - |
> > > > +    partitions {
> > > > +        compatible = "qcom,smem";
> > > > +        #address-cells = <1>;
> > > > +        #size-cells = <1>;
> > > > +    };
> > > > +
> > > > +    dynamic-partitions {
> > > > +      compatible = "dynamic-partitions";
> > > > +
> > > > +      art: art {
> > > > +        label = "0:art";
> > > > +        read-only;
> > > > +        compatible = "nvmem-cells";
> > > > +        #address-cells = <1>;
> > > > +        #size-cells = <1>;
> > > > +
> > > > +        macaddr_art_0: macaddr@0 {
> > > > +          reg = <0x0 0x6>;
> > > > +        };
> > > > +
> > > > +        macaddr_art_6: macaddr@6 {
> > > > +          reg = <0x6 0x6>;
> > > > +        };
> > > > +      };
> > > > +    };
> > > 
> > > First of all: I fully support such a feature. I need it for Broadom
> > > platforms that use "brcm,bcm947xx-cfe-partitions" dynamic partitions.
> > > In my case bootloader partition is created dynamically (it doesn't have
> > > const offset and size). It contains NVMEM data however that needs to be
> > > described in DT.
> > > 
> > > This binding however looks loose and confusing to me.
> > > 
> > 
> > I agree.
> > 
> > > First of all did you really mean to use "qcom,smem"? My first guess is
> > > you meant "qcom,smem-part".
> > > 
> > 
> > Yes sorry, I was referring to the smem parser qcom,smem-part
> > 
> > > Secondly can't we have partitions defined just as subnodes of the
> > > partitions { ... }; node?
> > > 
> > 
> > I would love to use it. My only concern is that due to the fact
> > that we have to support legacy partition declaring, wonder if this could
> > create some problem. I'm referring to declaring fixed partition without
> > using any compatible/standard binding name.
> 
> Legacy partitioning won't kick in if you have "partitions" with
> "compatible" string. We're safe here. Just checked to be sure.
>

Oh ok then the dynamic partition compatible stuff is not needed.
To make sure I will change the "connect" function part and skip the
of_node assign if a compatible is not present. (The of_node assign
should be done only with the nvmem-cell compatible currently.)

> 
> > I remember we improved that with the introduction of the nvmem binding
> > by making the fixed-partition compatible mandatory. But I would like to
> > have extra check. Wonder if to be on the safe part we can consider
> > appending to the "dynamic parser" a compatible like "dynamic-partitions"
> > and use your way to declare them (aka keeping the dynamic-partition and
> > removing the extra parallel partitions list)
> > 
> > Feel free to tell me it's just a stupid and unnecessary idea. I just
> > have fear to introduce regression in the partition parsing logic.
> 
> I'm confused. I think all dynamic partitioners already have a
> "compatible" set.

Now that I think about it you are right. If a dynamic partition is
present in the system, a compatible must be present to use the correct
parser. And as I said up, all the nvmem cells should have the
correct compatible.

> 
> Can you post an example of DT binging you described above, please?

Was thinking something like this. But not needed.

partitions {
     compatible = "brcm,bcm947xx-cfe-partitions", "dynamic-partitions";

     partition-0 {
         compatible = "nvmem-cells";
         label = "boot";

         #address-cells = <1>;
         #size-cells = <1>;

         mac: macaddr@0 {
             reg = <0x100 0x6>;
         };
     }
};

So in short, a scheme like this should NOT be handled/should not have
of_node assigned. (and is actually very wrong)

partitions {
     compatible = "brcm,bcm947xx-cfe-partitions";

     partition-0 {
         label = "boot";

         #address-cells = <1>;
         #size-cells = <1>;

         mac: macaddr@0 {
             reg = <0x100 0x6>;
         };
     }
};

-- 
	Ansuel

Powered by blists - more mailing lists