[PATCH] board: amlogic: vim3: fix setup ethernet mac from efuse

[Marek Szyprowski]

Hi Artem,

On 12.01.2021 06:03, Artem Lapkin wrote:
> Fix reading built-in ethernet MAC address from efuse
>
> NOTE: MAC is stored in ASCII format, 1bytes = 2characters by 0 offset
>
> if mac from efuse not valid we use meson_generate_serial_ethaddr
>
> NOTE: remake odroid-n2.c variant from Neil Armstrong <narmstrong at baylibre.com>
>
> Signed-off-by: Artem Lapkin <art at khadas.com>

First of all, thanks pointing again at this issue.

I've implemented reading of MAC address from EFUSE for VIM boards and I 
was convinced that it is stored in binary format. I've based my 
assumptions on the vendor code, especially the message from the vendor 
kernel:

[    4.028236] efusekeynum: 1
[    4.031072] efusekeyname:             mac    offset:     0 size:     6
[    4.037542] efuse efuse: probe OK!

and the vendors dts:

https://github.com/khadas/linux/blob/khadas-vims-4.9.y/arch/arm64/boot/dts/amlogic/kvim3_linux.dts#L506

which both points that the EFUSE key has only 6 bytes, what means binary 
format. I didn't check it further what is really stored there.

Now I checked again and it looks that the DTS (and thus vendor's kernel 
message) is incorrect, as the MAC is stored in ASCII and occupies 12 
bytes in efuse. One can check that with vendor's kernel:

# cat /sys/class/efuse/userdata
0x00: 43 38 36 33 31 34 37 30 41 38 44 37 00 00 00 00
0x10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
...

This means that your patch is correct, although I would change it a bit.

BTW, the Odroid N2/C4 boards don't have MAC stored explicitly in the 
EFUSE, but the boards UUID. The original code use the last bytes of the 
UUID as the MAC address.

> ---
>   board/amlogic/vim3/vim3.c | 22 +++++++++++++++++++---
>   1 file changed, 19 insertions(+), 3 deletions(-)
>
> diff --git a/board/amlogic/vim3/vim3.c b/board/amlogic/vim3/vim3.c
> index 824fff8262..87d9fe1f02 100644
> --- a/board/amlogic/vim3/vim3.c
> +++ b/board/amlogic/vim3/vim3.c
> @@ -139,26 +139,42 @@ int meson_ft_board_setup(void *blob, struct bd_info *bd)
>   }
>   
>   #define EFUSE_MAC_OFFSET	0
> -#define EFUSE_MAC_SIZE		6
> +#define EFUSE_MAC_SIZE		12
> +#define MAC_ADDR_LEN		6
>   
>   int misc_init_r(void)
>   {
> -	uint8_t mac_addr[EFUSE_MAC_SIZE];
> +	uint8_t mac_addr[MAC_ADDR_LEN];
> +	char efuse_mac_addr[EFUSE_MAC_SIZE], tmp[3];
>   	ssize_t len;
>   
>   	meson_eth_init(PHY_INTERFACE_MODE_RGMII, 0);
>   
>   	if (!eth_env_get_enetaddr("ethaddr", mac_addr)) {
>   		len = meson_sm_read_efuse(EFUSE_MAC_OFFSET,
> -					  mac_addr, EFUSE_MAC_SIZE);
> +					  efuse_mac_addr, EFUSE_MAC_SIZE);
>   		if (len != EFUSE_MAC_SIZE)
>   			return 0;
>   
> +		/* MAC is stored in ASCII format, 1bytes = 2characters */
> +		for (int i = 0; i < 6; i++) {
> +			tmp[0] = efuse_mac_addr[i * 2];
> +			tmp[1] = efuse_mac_addr[i * 2 + 1];
> +			tmp[2] = '\0';
> +			mac_addr[i] = simple_strtoul(tmp, NULL, 16);
> +		}
> +
>   		if (is_valid_ethaddr(mac_addr))
>   			eth_env_set_enetaddr("ethaddr", mac_addr);
>   		else
>   			meson_generate_serial_ethaddr();

Now the above code is the same as for Odroid N2/C4, so it can be moved 
to arch/arm/mach-meson/board-common.c, maybe as 
meson_read_efuse_mac_addr() with the offset as parameter, so it can be 
directly used by Odroids and VIMs.

> +
> +		eth_env_get_enetaddr("ethaddr", mac_addr);
> +		printf("[i] setup onboard mac %02X:%02X:%02X:%02X:%02X:%02X\n",
> +			mac_addr[0],mac_addr[1],mac_addr[2],
> +			mac_addr[3],mac_addr[4],mac_addr[5]);

I'm not really convinced that this message has to be displayed. If so, 
then use "%pM" for printing MAC address.

>   	}
>   
>   	return 0;
>   }
> +

Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland