Re: [PATCH v4 3/3] arm64: dts: qcom: sc8280xp-x13s: model the PMU of the on-board wcn6855

[Stephan Gerhold <stephan.gerhold@linaro.org>]

On Thu, Oct 03, 2024 at 05:16:59AM -0700, Bartosz Golaszewski wrote:
> On Thu, 3 Oct 2024 13:38:35 +0200, Bartosz Golaszewski <brgl@bgdev.pl> said:
> > On Thu, Oct 3, 2024 at 1:07 PM Johan Hovold <johan@kernel.org> wrote:
> >>
> >> Without this patch I'm seeing an indefinite probe deferral with
> >> 6.12-rc1:
> >>
> >>         platform 1c00000.pcie:pcie@0:wifi@0: deferred probe pending: pci-pwrctl-pwrseq: Failed to get the power sequencer
> >>
> >> Can you please look into that and make sure that the existing DT
> >> continues to work without such warnings.
> >>
> >
> > Ah, dammit, I missed the fact that X13s already has this node defined
> > so PCI pwrctl will consume it and try to get the power sequencer
> > handle. I'm wondering how to tackle it though... It will most likely
> > require some kind of a driver quirk where we check if we have the PMU
> > node and if not, then don't try to set up power sequencing. Any other
> > ideas?
> >
> 
> This is untested but would it make sense?
> 
> diff --git a/drivers/pci/pwrctl/pci-pwrctl-pwrseq.c
> b/drivers/pci/pwrctl/pci-pwrctl-pwrseq.c
> index a23a4312574b..071ee77c763d 100644
> --- a/drivers/pci/pwrctl/pci-pwrctl-pwrseq.c
> +++ b/drivers/pci/pwrctl/pci-pwrctl-pwrseq.c
> @@ -3,6 +3,7 @@
>   * Copyright (C) 2024 Linaro Ltd.
>   */
> 
> +#include <linux/cleanup.h>
>  #include <linux/device.h>
>  #include <linux/mod_devicetable.h>
>  #include <linux/module.h>
> @@ -87,7 +88,31 @@ static struct platform_driver pci_pwrctl_pwrseq_driver = {
>  	},
>  	.probe = pci_pwrctl_pwrseq_probe,
>  };
> -module_platform_driver(pci_pwrctl_pwrseq_driver);
> +
> +static int __init pci_pwrctl_pwrseq_init(void)
> +{
> +	/*
> +	 * Old device trees for the Lenovo X13s have the "pci17cb,1103" node
> +	 * defined but don't use power sequencing yet. If we register this
> +	 * driver, it will match against this node and lead to emitting of
> +	 * a warning in the kernel log when we cannot get the power sequencing
> +	 * handle. Let's skip registering the platform driver if we're on X13s
> +	 * but don't have the PMU node.
> +	 */
> +	if (of_machine_is_compatible("lenovo,thinkpad-x13s")) {
> +		struct device_node *dn __free(device_node) =
> +			of_find_compatible_node(NULL, NULL, "qcom,wcn6855-pmu");
> +		if (!dn)
> +			return 0;
> +	}
> +
> +	return platform_driver_register(&pci_pwrctl_pwrseq_driver);
> +}
> +
> +static void __exit pci_pwrctl_pwrseq_exit(void)
> +{
> +	platform_driver_unregister(&pci_pwrctl_pwrseq_driver);
> +}
> 
>  MODULE_AUTHOR("Bartosz Golaszewski <bartosz.golaszewski@linaro.org>");
>  MODULE_DESCRIPTION("Generic PCI Power Control module for power
> sequenced devices");
> 
> X13s is the only platform that would use one of the compatibles supported by
> this driver before power sequencing so it should be a one-off quirk.
> 

I'm guessing the pci17cb,1107 node in x1e80100-lenovo-yoga-slim7x is
also affected?

Maybe you can check if the node has one of the -supply properties and
return -ENODEV from pci_pwrctl_pwrseq_probe() otherwise?

Thanks,
Stephan