Re: [PATCH v2 1/3] dt-bindings: mtd: fixed-partitions: Add binman compatible

[Simon Glass]

Hi Michael,

On Wed, 4 Oct 2023 at 11:17, Michael Walle <mwalle@xxxxxxxxxx> wrote:
>
> Hi,
>
> >> >> Add a compatible string for binman, so we can extend fixed-partitions
> >> >> in various ways.
> >> >
> >> > I've been thinking at the proper way to describe the binman partitions.
> >> > I am wondering if we should really extend the fixed-partitions
> >> > schema. This description is really basic and kind of supposed to remain
> >> > like that. Instead, I wonder if we should not just keep the binman
> >> > compatible alone, like many others already. This way it would be very clear
> >> > what is expected and allowed in both cases. I am thinking about
> >> > something like that:
> >> >
> >> >       Documentation/devicetree/bindings/mtd/partitions/brcm,bcm4908-partitions.yaml
> >> >
> >> > this file is also referenced there (but this patch does the same, which
> >> > is what I'd expect):
> >> >
> >> >       Documentation/devicetree/bindings/mtd/partitions/partitions.yaml
> >> >
> >> > I'll let the binding maintainers judge whether they think it's
> >> > relevant, it's not a strong opposition.
> >>
> >> What is the overall goal here? To replace the current binman node
> >> which is
> >> usually contained in the -u-boot.dtsi files? If one is using binman to
> >> create an image, is it expected that one needs to adapt the DT in
> >> linux?
> >> Or will it still be a seperate -u-boot.dtsi? > Because in the latter
> >> case
> >> I see that there will be conflicts because you have to overwrite the
> >> flash node. Or will it be a seperate node with all the information
> >> duplicated?
> >
> > The goal is simply to have a full binding for firmware layout, such
> > that firmware images can be created, examined and updated. The
> > -u-boot.dtsi files are a stopgap while we sort out a real binding.
> > They should eventually go away.
>
> You haven't answered whether this node should be a seperate binman
> node - or if you'll reuse the existing flash (partitions) node(s) and
> add any missing property there. If it's the latter, I don't think
> compatible = "binman", "fixed-partitions"; is correct.

My intent is to make it compatible, so wouldn't it make sense to have
binman as the first compatible, then falling back to fixed-partitions
as the second?

>
> >> Maybe (a more complete) example would be helpful.
> >
> > Can you please be a bit more specific? What is missing from the
> > example?
>
> Like a complete (stripped) DTS. Right now I just see how the individual
> node looks like. But with a complete example DTS, my question from above
> would have been answered.
>
> What if a board uses eMMC to store the firmware binaries? Will that then
> be a subnode to the eMMC device?

I thought there was a way to link the partition nodes and the device
using a property, without having the partition info as a subnode of
the device. But I may have imagined it as I cannot find it now. So
yes, it will be a subnode of the eMMC device.

Regards,
Simon