[PATCH] arm: imx: imx9: scmi: Fix booting from USB

[Frieder Schrempf]

Am 31.07.25 um 20:47 schrieb João Paulo Gonçalves:
> Hello,
> 
> On Mon, Jul 28, 2025 at 11:25:04AM -0300, Fabio Estevam wrote:
>> What confuses me about João's patch is that it was not tested against
>> the mainline U-Boot.
> 
> In the end, the code is nearly identical between the downstream and
> mainline U-Boot.
> 
> On Wed, Jul 30, 2025 at 08:35:40AM +0200, Frieder Schrempf wrote:
>> I have the below patch in our downstream fork to fix the issue mentioned
>> here for i.MX93. It is derived from Fabio's patch [1]. I'm not sure if
>> it's correct, though.
> 
> We are also doing something similar on our branch.
> 
> I have two proposals. My goal is to prevent the board from hanging when
> booting from USB if CONFIG_ENV_IS_NOWHERE=n (since the current code
> assumes it is always enabled) and/or if the environment is defined
> elsewhere:
> 
> 1) Imply CONFIG_ENV_IS_NOWHERE for i.MX9 (as mentioned earlier), to make
> it clear that it is required by ARCH.
> 
> or
> 
> 2) Doing something similar to what Frieder showed for USB boot (which is
> almost the same on i.MX8MM/i.MX8MP), but preserving the current
> behavior:
> 
> diff --git a/arch/arm/mach-imx/imx9/scmi/soc.c b/arch/arm/mach-imx/imx9/scmi/soc.c
> index 13f13ca7d105..f973652d0cbe 100644
> --- a/arch/arm/mach-imx/imx9/scmi/soc.c
> +++ b/arch/arm/mach-imx/imx9/scmi/soc.c
> @@ -635,7 +635,8 @@ enum env_location env_get_location(enum env_operation op, int prio)
> 
>         switch (dev) {
>         case QSPI_BOOT:
> -               env_loc = ENVL_SPI_FLASH;
> +               if (IS_ENABLED(CONFIG_ENV_IS_IN_SPI_FLASH))
> +                       env_loc = ENVL_SPI_FLASH;
>                 break;
>         case SD1_BOOT:
>         case SD2_BOOT:
> @@ -643,10 +644,16 @@ enum env_location env_get_location(enum env_operation op, int prio)
>         case MMC1_BOOT:
>         case MMC2_BOOT:
>         case MMC3_BOOT:
> -               env_loc =  ENVL_MMC;
> +               if (IS_ENABLED(CONFIG_ENV_IS_IN_MMC))
> +                       env_loc =  ENVL_MMC;
>                 break;
>         default:
> -               env_loc = ENVL_NOWHERE;
> +               if (IS_ENABLED(CONFIG_ENV_IS_NOWHERE))
> +                       env_loc = ENVL_NOWHERE;
> +               else if (IS_ENABLED(CONFIG_ENV_IS_IN_SPI_FLASH))
> +                       env_loc = ENVL_SPI_FLASH;
> +               else if (IS_ENABLED(CONFIG_ENV_IS_IN_MMC))
> +                       env_loc = ENVL_MMC;
>                 break;
>         }
> 
> With this patch, I believe Peng's concerns are addressed, and it also
> works for our use case.
> 
> What do you think?

On first glance, this looks good to me. If possible, can you please also
send a patch for i.MX93 (arch/arm/mach-imx/imx9/soc.c) to align the
behavior of both platforms?

Thanks!