Hi Eric,
[add Sebastian]
> Eric Anholt <eric@xxxxxxxxxx> hat am 30. November 2018 um 21:27 geschrieben:
>
>
> This binding supersedes the bcm2835-pm-wdt binding which only covered
> enough to provide a watchdog, but the HW block is actually mostly
> about power domains.
>
> Signed-off-by: Eric Anholt <eric@xxxxxxxxxx>
> ---
> .../bindings/soc/bcm/brcm,bcm2835-pm.txt | 42 +++++++++++++++++++
> 1 file changed, 42 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/soc/bcm/brcm,bcm2835-pm.txt
>
> diff --git a/Documentation/devicetree/bindings/soc/bcm/brcm,bcm2835-pm.txt b/Documentation/devicetree/bindings/soc/bcm/brcm,bcm2835-pm.txt
> new file mode 100644
> index 000000000000..7818d33a158f
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/soc/bcm/brcm,bcm2835-pm.txt
> @@ -0,0 +1,42 @@
> +BCM2835 PM (Power domains, watchdog)
> +
> +The PM block controls power domains and some reset lines, and includes
> +a watchdog timer. This binding supersedes the brcm,bcm2835-pm-wdt
> +binding which covered some of PM's register range and functionality.
> +
> +Required properties:
> +
> +- compatible: Should be "brcm,bcm2835-pm"
> +- reg: Specifies base physical address and size of the two
> + register ranges ("PM" and "ASYNC_BRIDGE" in that
> + order)
> +- clocks: a) v3d: The V3D clock from CPRMAN
> + b) peri_image: The PERI_IMAGE clock from CPRMAN
> + c) h264: The H264 clock from CPRMAN
> + d) isp: The ISP clock from CPRMAN
> +- #reset-cells: Should be 1. This property follows the reset controller
> + bindings[1].
> +- #power-domain-cells: Should be 1. This property follows the power domain
> + bindings[2].
> +
> +Optional properties:
> +
> +- timeout-sec: Contains the watchdog timeout in seconds
> +
> +[1] Documentation/devicetree/bindings/reset/reset.txt
> +[2] Documentation/devicetree/bindings/power/power_domain.txt
> +
sorry for my late reply. I hope we can take the opportunity of a new binding to fix an old issue of the bcm2835-wdt driver. The watchdog driver sets the pm_power_off callback without checking for "system-power-controller" property [3]. As a result we can't use another poweroff controller like e.g. gpio-poweroff. In downstream this has been workarounded by an additional devicetree property for the gpio-poweroff driver [4]. But this isn't the right way because the issue is in the bcm2835 watchdog driver not in all the other reset controller.
Suggested pseudo code:
if (!pm_power_off) {
/* Preserve the old behavior of watchdog driver in case of old DTB */
if (!of_device_is_compatible(dev->of.node, "brcm,bcm2835-pm") ||
of_device_is_system_power_controller(dev->of.node))
pm_power_off = bcm2835_power_off;
}
Best regards
Stefan
[3] - https://elixir.bootlin.com/linux/latest/source/Documentation/devicetree/bindings/power/power-controller.txt
[4] - https://github.com/raspberrypi/linux/commit/f86dcfac0a4e478fee40b5a3729ff1b159ae91ee
