[PATCH] arch: arm: kirkwood: support nvmem mac address

[Andrew Lunn]

On Mon, Sep 30, 2024 at 02:59:34PM -0700, Rosen Penev wrote:
> of_get_ethdev_address gets called too early for nvmem. If EPROBE_DEFER
> gets called, skip so that the ethernet driver can adjust the MAC address
> through nvmem.

Is this from code analysis or do you have a board with real issues? Do
we want to add a Fixed: so it gets back ported in stable?

> 
> Signed-off-by: Rosen Penev <rosenp at gmail.com>
> ---
>  arch/arm/mach-mvebu/kirkwood.c | 11 ++++++++---
>  1 file changed, 8 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/arm/mach-mvebu/kirkwood.c b/arch/arm/mach-mvebu/kirkwood.c
> index 73b2a86d6489..da347f66900b 100644
> --- a/arch/arm/mach-mvebu/kirkwood.c
> +++ b/arch/arm/mach-mvebu/kirkwood.c
> @@ -86,13 +86,18 @@ static void __init kirkwood_dt_eth_fixup(void)
>  		void __iomem *io;
>  		u8 *macaddr;
>  		u32 reg;
> +		int err;
>  
>  		if (!pnp)
>  			continue;
>  
> -		/* skip disabled nodes or nodes with valid MAC address*/
> -		if (!of_device_is_available(pnp) ||
> -		    !of_get_mac_address(np, tmpmac))
> +		/* skip disabled nodes */
> +		if (!of_device_is_available(pnp))
> +			goto eth_fixup_skip;
> +
> +		/* skip nodes with valid MAC address*/
> +		err = of_get_mac_address(np, tmpmac);
> +		if (err == -EPROBE_DEFER || !err)
>  			goto eth_fixup_skip;

I'm wondering about ordering here. What exactly does EPROBE_DEFER
mean? Does it mean we know there is a MAC address in nvmem, but the
nvmem has not probed yet? Or can it mean, the nvmem has not probed
yet, and maybe there is a MAC address in it, maybe not?

In the maybe not case, we should still be trying to read the MAC from
the hardware and storing it way safe for later use.

	Andrew