Hi Morimoto-san,
On Thu, 17 Apr 2025 at 01:52, Kuninori Morimoto
<kuninori.morimoto.gx@xxxxxxxxxxx> wrote:
> > > > + # "MSIOF-SPI" specific
> > > > + - if:
> > > > + properties:
> > > > + $nodename:
> > > > + pattern: '^spi@'
> > >
> > > This condition does not match what you wrote in the cover letter:
> > > the controller is used in I2S mode when a port(s) subnode is present,
> > > and in SPI mode when no port(s) subnode is present.
> > >
> > > > + then:
> > > > + allOf:
> > > > + - $ref: spi-controller.yaml#
> > >
> > > Documentation/devicetree/bindings/spi/spi-controller.yaml indeed
> > > requires that the node-name matches "^spi(@.*|-([0-9]|[1-9][0-9]+))?$".
> > > The controller's node is located in the SoC-specific .dtsi, where its
> > > intended use case is not yet known, and its node name cannot easily be
> > > overridden in the board .dts that specifies the use case. Hence the
> > > node name must always be "spi" (and cannot be e.g. "serial-engine").
> > > Let's hope there is no other use case for MSIOF that requires using
> > > a different node name...
>
> Hmm...
>
> Now, MSIOF node has "spi@xxxx".
> SoC file indicates MSIOF-SPI as default, so it has below lines
>
> --- SoC file ----
> msiof1: spi@xxxx {
> ...
> #address-cells = <1>;
> #size-cells = <0>;
> ...
> };
>
> These are not needed for MSIOF-I2S, so removes these
>
> --- Board file ----
> &msiof1 {
> ...
> /delete-property/#address-cells;
> /delete-property/#size-cells;
> ...
> };
>
> Now, my dt-bindings doesn't load spi-controller.yaml (as sample), but I got
>
> [SoC file]: Warning (spi_bus_bridge): /soc/spi@xxxx: incorrect #address-cells for SPI bus
> also defined at [Board file]
> [SoC file]: Warning (spi_bus_bridge): /soc/spi@xxxx: incorrect #size-cells for SPI bus
> also defined at [Board file]
>
> MSIOF dt-bindings doesn't load spi-controller.yaml, but why I got "spi_bus_bridge"
> warning ?? I wonder dt compiler (?) automatically check "spi" node ?
> I have tryed some code, my expectation seems correct (In case of node name was "spi@xxx",
> I got many SPI related warnings even though I didn't load spi-controller).
These come from dtc, which makes its own assumptions:
$ git grep spi_bus_bridge
scripts/dtc/checks.c:static void check_spi_bus_bridge(struct check
*c, struct dt_info *dti, struct node *node)
scripts/dtc/checks.c:WARNING(spi_bus_bridge, check_spi_bus_bridge,
NULL, &addr_size_cells);
scripts/dtc/checks.c:WARNING(spi_bus_reg, check_spi_bus_reg, NULL,
®_format, &spi_bus_bridge);
scripts/dtc/checks.c: &spi_bus_bridge,
Perhaps we do need to extend the use of role-specifying properties
like "interrupt-controller" (in Device Tree Specification v0.4 and in
dt-schema) and the few others in Documentation/devicetree/bindings:
gpio-controller
mctp-controller
msi-controller
system-power-controller
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
