Re: [PATCH v5 1/3] dt-bindings: mtd: partitions: Support label only partition

From: Rob Herring <robh@kernel.org>
To: Ansuel Smith <ansuelsmth@gmail.com>
Cc: "Miquel Raynal" <miquel.raynal@bootlin.com>,
	"Richard Weinberger" <richard@nod.at>,
	"Vignesh Raghavendra" <vigneshr@ti.com>,
	"Krzysztof Kozlowski" <krzysztof.kozlowski+dt@linaro.org>,
	"Greg Kroah-Hartman" <gregkh@linuxfoundation.org>,
	"Jens Axboe" <axboe@kernel.dk>, �ecki <rafal@milecki.pl>,
	"Manivannan Sadhasivam" <mani@kernel.org>,
	linux-mtd@lists.infradead.org, devicetree@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v5 1/3] dt-bindings: mtd: partitions: Support label only partition
Date: Fri, 10 Jun 2022 11:02:05 -0600	[thread overview]
Message-ID: <20220610170205.GD1787330-robh@kernel.org> (raw)
In-Reply-To: <62a250c4.1c69fb81.addd2.0eae@mx.google.com>

On Thu, Jun 09, 2022 at 09:57:52PM +0200, Ansuel Smith wrote:
> On Thu, Jun 09, 2022 at 12:32:52PM -0600, Rob Herring wrote:
> > On Mon, Jun 06, 2022 at 05:14:15PM +0200, Ansuel Smith wrote:
> > > Document new partition nodes that declare only the label instead of the
> > > reg 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.
> > > 
> > > With these special partitions, the reg / offset is not required.
> > > The label binding is used to match the partition allocated by the
> > > parser at runtime and the parser will provide reg and offset of the mtd.
> > > 
> > > NVMEM will use the data from the parser and provide the NVMEM cells
> > > declared in the DTS, "connecting" the dynamic partition with a
> > > static declaration of cells in them.
> > > 
> > > Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> > > ---
> > >  .../bindings/mtd/partitions/partition.yaml       | 16 ++++++++++++++--
> > >  1 file changed, 14 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/Documentation/devicetree/bindings/mtd/partitions/partition.yaml b/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > > index e1ac08064425..bff6fb980e6b 100644
> > > --- a/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > > +++ b/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > > @@ -11,6 +11,13 @@ description: |
> > >    relative offset and size specified. Depending on partition function extra
> > >    properties can be used.
> > >  
> > > +  A partition may be dynamically allocated by a specific parser at runtime.
> > > +  In this specific case, the label is required instead of the reg.
> > > +  This is used to assign an OF node to the dynamiccally allocated partition
> > > +  so that subsystem like NVMEM can provide an OF node and declare NVMEM cells.
> > > +  The OF node will be assigned only if the partition label declared match the
> > > +  one assigned by the parser at runtime.
> > > +
> > >  maintainers:
> > >    - Rafał Miłecki <rafal@milecki.pl>
> > >  
> > > @@ -22,6 +29,8 @@ properties:
> > >    label:
> > >      description: The label / name for this partition. If omitted, the label
> > >        is taken from the node name (excluding the unit address).
> > > +      With dynamically allocated partition the label is required and won't
> > > +      fallback to the node name.
> > 
> > Generally, label is never required being something for humans rather 
> > than the s/w to consume. I don't see any reason why we can't still use 
> > the node name (with 'partition-' stripped off).
> > 
> 
> How to enforce the use of 'partition-'? Should the driver then check the
> node name and reject any wrong node name (and return error)?

The schema can do it either in the parent (of partition nodes) schema or 
with $nodename 'property'.

$nodename:
  oneOf:
    - pattern: '^.*@.*$'
    - pattern: '^partition-.*$'

or:

if:
  not:
    required:
      - reg
then:
  properties:
    $nodename:
      pattern: '^partition-.*$'

The latter is a bit clearer on the intent I think.

> > If the purpose is to define what the partition contains, then 
> > 'compatible' is the right thing for that.
> >
> 
> Introducing a compatible means creating another scheme I think or we can
> add that special compatible in the partition scheme?

It would be another schema. You could make 'compatible' required here 
perhaps, but maybe there's a use for an empty node?

Rob