On 5/15/25 4:30 PM, Tom Rini wrote:
On Thu, May 15, 2025 at 03:26:18PM +0200, Quentin Schulz wrote:
Hi Dario,
On 5/15/25 3:12 PM, Dario Binacchi wrote:
Commit 2a00d73d081a ("spl: mmc: Try to clean up raw-mode options") breaks
the boot of the BSH SMM S2 board. As stated in the commit itself, "Some
boards use this value even though MMC is not enabled in SPL, for example
imx8mn_bsh_smm_s2". The same commit makes SPL_SYS_MMCSD_RAW_MODE depend
on SPL_DM_MMC || SPL_MMC. With SPL_SYS_MMCSD_RAW_MODE unset, it is not
possible to enable SYS_MMCSD_RAW_MODE_U_BOOT_USE_SECTOR, thus breaking
the board's boot process.
The patch once again allows SYS_MMCSD_RAW_MODE_U_BOOT_USE_SECTOR to be
selected also for boards that do not use MMC, in this case on the i.MX8M
platform.
Why would you use SYS_MMCSD_RAW_MODE_U_BOOT_USE_SECTOR if you don't have any
MMC? Shouldn't we rather fix the incorrect usage of this variable to be
something else (or better guarded?).
It appears that arch/arm/mach-imx/spl_imx_romapi.c::spl_romapi_get_uboot_base()
will use the value so that we (apparently) go ROM->U-Boot, or at least
ROM->some series of prior stage loaders->ROM->U-Boot. I'm not sure I
love overloading the value like this, but I can't say it's wrong.
Perhaps the help needs to be updated too?
Isn't it rather U-Boot SPL asking the BootROM to load the image for it
instead of doing it by itself? Some MMC is technically present then,
just that because the BootROM handles it and not U-Boot, we need to
specify the offset in the MMC even though we only return that value to
the BootROM.
If that is correct, then I'm not sure we want to use
SYS_MMCSD_RAW_MODE_U_BOOT_SECTOR considering it selects SPL_LOAD_BLOCK
which I would assume to be useless if the BootROM is doing all the work
for the SPL?
Cheers,
Quentin
