
On 08/10/2012 05:32 PM, Scott Wood wrote:
On 07/30/2012 12:37 PM, Stephen Warren wrote:
diff --git a/board/nvidia/dts/tegra20-harmony.dts b/board/nvidia/dts/tegra20-harmony.dts
- nand-controller@70008000 {
...
nand@0 {
reg = <0>;
compatible = "hynix,hy27uf4g2b", "nand-flash";
Where is the binding for a generic "nand-flash" compatible?
I don't know if there is one. This patch was cribbed from one by Simon (CC'd) for Seaboard. Simon do you know the answer?
That said, it seems like it's quite reasonable to create or assume such a binding, since NAND flash appears self-describing.
That said, looking for precedence in the Linux kernel indicates some inconsistencies. There exist:
* Many (ARM and PowerPC) bindings that have no node for the NAND chip itself, but rather the NAND controller has children that describe the partitioning of the NAND chip.
* Many (I think only PowerPC) bindings that have a node for the NAND chip, but containing little but #address/#size-cells, and this node then contains children that describe the partitioning.
* I guess there is at least one PowerPC example just like this patch, that contains a child node containing a compatible value describing the flash model number.