Hi,
On Mon, Dec 16, 2019 at 10:06 PM Douglas Anderson <dianders@xxxxxxxxxxxx> wrote:
>
> On many arm64 qcom device trees, running `make dtbs_check` yells:
>
> timer@17c20000: #size-cells:0:0: 1 was expected
>
> It appears that someone was trying to assert the fact that sub-nodes
> describing frames would never have a size that's more than 32-bits
> big. That does indeed appear to be true for all cases I could find.
>
> Currently many arm64 qcom device tree files have a #address-cells and
> about in commit bede7d2dc8f3 ("arm64: dts: qcom: sdm845: Increase
> address and size cells for soc"). That means the only way we can
> shrink them down is to use a non-empty ranges.
>
> Since forever it has said in "writing-bindings.txt" to "DO use
> non-empty 'ranges' to limit the size of child buses/devices". I guess
> we should start listening to it.
>
> I believe (but am not certain) that this also means that we should use
> "ranges" to simplify the "reg" of our sub devices by specifying an
> offset. Let's update the example in the bindings to make this
> obvious.
>
> Signed-off-by: Douglas Anderson <dianders@xxxxxxxxxxxx>
> ---
> See:
> https://lore.kernel.org/r/20191212113540.7.Ia9bd3fca24ad34a5faaf1c3e58095c74b38abca1@changeid
>
> ...for the patch that sparked this change.
>
> .../devicetree/bindings/timer/arm,arch_timer_mmio.yaml | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/Documentation/devicetree/bindings/timer/arm,arch_timer_mmio.yaml b/Documentation/devicetree/bindings/timer/arm,arch_timer_mmio.yaml
> index b3f0fe96ff0d..d927b42ddeb8 100644
> --- a/Documentation/devicetree/bindings/timer/arm,arch_timer_mmio.yaml
> +++ b/Documentation/devicetree/bindings/timer/arm,arch_timer_mmio.yaml
> @@ -99,22 +99,22 @@ examples:
> compatible = "arm,armv7-timer-mem";
> #address-cells = <1>;
> #size-cells = <1>;
> - ranges;
> + ranges = <0 0xf0000000 0x1000>;
I checked this over a few times and yet I still screwed it up. :(
This should be:
ranges = <0 0xf0001000 0x1000>;
...which makes the first "frame" below actually start at 0. I'll wait
before sending out a v2, though, in case this patch is totally wrong
or something.
> reg = <0xf0000000 0x1000>;
> clock-frequency = <50000000>;
>
> - frame@f0001000 {
> + frame@0 {
> frame-number = <0>;
> interrupts = <0 13 0x8>,
> <0 14 0x8>;
> - reg = <0xf0001000 0x1000>,
> - <0xf0002000 0x1000>;
> + reg = <0x0000 0x1000>,
> + <0x1000 0x1000>;
> };
>
> frame@f0003000 {
> frame-number = <1>;
> interrupts = <0 15 0x8>;
> - reg = <0xf0003000 0x1000>;
> + reg = <0x2000 0x1000>;
> };
> };
