Re: [RFC PATCH 1/5] mtd: nand: ecc-qcom: Add support for ECC Engine Driver

[Miquel Raynal <miquel.raynal@bootlin.com>]

Hello,

> Based on below feedback [1] and NAK on the device node patch
> got idea of having separate device node for ECC is not acceptable.
> Could you please help to clarify that.

If I may try to compare with the Macronix situation, the ECC engine
was an independent hardware block, with its own mapping and its own
registers, so it was described as an independent node in the DT. The
type of ECC controller (pipelined or external) is described by the
nand-ecc-engine property which either points at the parent node
(pipelined) or an external node (external). The SPI host would itself
point at the external ECC engine node with its own nand-ecc-engine
property (see mtd/mxicy,nand-ecc-engine.yaml in the bindings).

> Since ECC block is inlined with QPIC controller so is the below
> device node acceptable ?
> 
>     bch: qpic_ecc {
>                            compatible = "qcom,ipq9574-ecc";
>                            status = "ok";
>                    };

If it does not has its own mapping and if you access the ECC engine
through the host registers then the controller should be part of the
host node, but I am not sure it really needs to be described
explicitly, most of them are not for historical reasons. Conceptually
there is a problem with subnodes of each of these controllers having
a signification already: SPI devices or NAND chips.

Thanks,
Miquèl