[PATCH v1] mtd: parsers: ofpart: Fix parsing when size-cells is 0

[Miquel Raynal]

Hi Francesco,

francesco at dolcini.it wrote on Fri, 2 Dec 2022 17:45:37 +0100:

> On Fri, Dec 02, 2022 at 05:17:59PM +0100, Marek Vasut wrote:
> > On 12/2/22 16:49, Miquel Raynal wrote:  
> > > , not the NAND controller node. I hope this
> > > is correctly supported in U-Boot though. So if there is a NAND chip
> > > subnode, I suppose U-Boot would try to create the partitions that are
> > > inside, or even in the sub "partitions" container.  
> > 
> > My understanding is that U-Boot checks the nand-controller node size-cells,
> > not the nand-chip{} or partitions{} subnode size-cells .  
> Not 100% correct.
> 
>  - U-Boot before v2022.04 updates the nand-controller{} node, no matter what.
>  - U-Boot starting from v2022.04 looks for `partitions{}` into the
>    nand-controller{} node, and creates the partition into it if found.
>    If not found it behaves the same way as the previous versions.
>    See commit 36fee2f7621e ("common: fdt_support: add support for "partitions" subnode to fdt_fixup_mtdparts()")
> 
> I'd like to stress once more the fact that we cannot expect old U-Boot
> to be updated in the field, and they will keep generating the partitions
> as child of the nand-controller node whatever we do with the dts file.
> 
> I think that this should be treated the same way as any other fixup we
> might have for broken firmware, especially considering that this used to
> "work" (yes, I can agree that it horrible, but I cannot change the past)
> without even a warning since the imx7 support was first introduced in
> the linux kernel years ago.
> 
> > Francesco, can you please share the DT, including the U-Boot generated
> > partitions, which is passed to Linux on Colibri MX7 ? I think that should
> > make all confusion go away.  
> 
> The device tree part is easy, just
> arch/arm/boot/dts/imx7d-colibri-eval-v3.dts.
> 
> and the nand-controller node is coming from
> 
> #include "imx7d.dtsi"
> 
> plus
> 
> &gpmi {
> 	fsl,use-minimum-ecc;
> 	nand-ecc-mode = "hw";
> 	nand-on-flash-bbt;
> 	pinctrl-names = "default";
> 	pinctrl-0 = <&pinctrl_gpmi_nand>;
> };
> 
> The partitions nodes are generated 100% by U-Boot, nothing is present in
> the dts source files.

I hope if you provide a NAND chip child node, the partitions are created
at the right location, otherwise this is so, so wrong...

> 
> With this DTS file as input, whatever U-Boot version is used I have the
> following generated:
> 
> root at colibri-imx7-02844233:/# ls /proc/device-tree/soc/nand-controller at 33002000/
> #address-cells          dma-names               nand-on-flash-bbt       pinctrl-0
> #size-cells             dmas                    partition at 0             pinctrl-names
> assigned-clock-parents  fsl,use-minimum-ecc     partition at 200000        reg
> assigned-clocks         interrupt-names         partition at 380000        reg-names
> clock-names             interrupts              partition at 400000        status
> clocks                  name                    partition at 80000
> compatible              nand-ecc-mode           phandle
> 
> root at colibri-imx7-02844233:/# ls /proc/device-tree/soc/nand-controller at 33002000/partition@*
> /proc/device-tree/soc/nand-controller at 33002000/partition at 0:
> label  name   reg
> 
> /proc/device-tree/soc/nand-controller at 33002000/partition at 200000:
> label      name       read_only  reg
> 
> /proc/device-tree/soc/nand-controller at 33002000/partition at 380000:
> label  name   reg
> 
> /proc/device-tree/soc/nand-controller at 33002000/partition at 400000:
> label  name   reg
> 
> /proc/device-tree/soc/nand-controller at 33002000/partition at 80000:
> label      name       read_only  reg
> 
> 


Thanks,
Miquèl