[PATCH] env/fat.c: allow loading from a FAT partition on the MMC boot device

Tom Rini trini at konsulko.com
Sat Jul 25 19:03:37 CEST 2020


On Fri, Jun 19, 2020 at 11:07:17PM +0100, David Woodhouse wrote:

> I don't want to have to specify the device; only the partition.
> 
> This allows me to use the same image on internal eMMC or SD card for
> Banana Pi R2, and it finds its own environment either way.
> 
> Signed-off-by: David Woodhouse <dwmw2 at infradead.org>
> ---
>  env/Kconfig |  4 ++++
>  env/fat.c   | 29 +++++++++++++++++++++++++++--
>  2 files changed, 31 insertions(+), 2 deletions(-)
> 
> diff --git a/env/Kconfig b/env/Kconfig
> index 38e7fadbb9..5784136674 100644
> --- a/env/Kconfig
> +++ b/env/Kconfig
> @@ -434,6 +434,10 @@ config ENV_FAT_DEVICE_AND_PART
>  	                   If none, first valid partition in device D. If no
>  	                   partition table then means device D.
>  
> +	  If ENV_FAT_INTERFACE is set to "mmc" then device 'D' can be omitted,
> +	  leaving the string starting with a colon, and the boot device will
> +	  be used.
> +
>  config ENV_FAT_FILE
>  	string "Name of the FAT file to use for the environment"
>  	depends on ENV_IS_IN_FAT
> diff --git a/env/fat.c b/env/fat.c
> index 35a1955e63..37c15d16cd 100644
> --- a/env/fat.c
> +++ b/env/fat.c
> @@ -29,6 +29,31 @@
>  # define LOADENV
>  #endif
>  
> +__weak int mmc_get_env_dev(void)
> +{
> +        return CONFIG_SYS_MMC_ENV_DEV;
> +}
> +
> +static char *env_fat_device_and_part(void)
> +{
> +#ifdef CONFIG_MMC
> +	static char *part_str;
> +
> +	if (!part_str) {
> +		part_str = CONFIG_ENV_FAT_DEVICE_AND_PART;
> +		if (!strcmp(CONFIG_ENV_FAT_INTERFACE, "mmc")
> +		    && part_str[0] == ':') {
> +			part_str = "0" CONFIG_ENV_FAT_DEVICE_AND_PART;
> +			part_str[0] += mmc_get_env_dev();
> +		}
> +	}
> +
> +	return part_str;
> +#else
> +	return CONFIG_ENV_FAT_DEVICE_AND_PART;
> +#endif
> +}
> +
>  static int env_fat_save(void)
>  {
>  	env_t __aligned(ARCH_DMA_MINALIGN) env_new;
> @@ -43,7 +68,7 @@ static int env_fat_save(void)
>  		return err;
>  
>  	part = blk_get_device_part_str(CONFIG_ENV_FAT_INTERFACE,
> -					CONFIG_ENV_FAT_DEVICE_AND_PART,
> +					env_fat_device_and_part(),
>  					&dev_desc, &info, 1);
>  	if (part < 0)
>  		return 1;
> @@ -89,7 +114,7 @@ static int env_fat_load(void)
>  #endif
>  
>  	part = blk_get_device_part_str(CONFIG_ENV_FAT_INTERFACE,
> -					CONFIG_ENV_FAT_DEVICE_AND_PART,
> +					env_fat_device_and_part(),
>  					&dev_desc, &info, 1);
>  	if (part < 0)
>  		goto err_env_relocate;

So, the biggest problem here, which I found while testing, is that since
CONFIG_SYS_MMC_ENV_DEV isn't always defined, this blows up a few cases.
Migrating this to Kconfig needs to happen anyhow, so I started on that.
Doing so however isn't trivial as we have cases where we set the value
to 0 (along with SYS_MMC_ENV_PART), and setting it to 0 isn't the same
code paths as unset.  I've sent out patches to deal with that now.  For
the moment, I'm adding a #ifndef / #define 0 / #endif to this patch.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200725/74be937a/attachment.sig>


More information about the U-Boot mailing list