[PATCH 3/4] arch: arm: mach-socfpga: Adjust a raw sectors for MMC loading of U-Boot from SPL

[Jan Kiszka]

On 25.11.25 08:23, Sune Brian wrote:
>> But that boot approach fell apart after the refactorings of raw mode mmc
>> loading and now starts to work again with my patches. Not saying they
>> are perfect, but at least the board defconfig gives something that runs.
> 
> Hi Jan,
> 
> No it did not.
> 
> My latest since 2025.07 built on Uboot not even used what the hard code setups.
> 
> grep -r "SYS_MMCSD_RAW_MODE_U_BOOT_USE_SECTOR" .config
> # CONFIG_SYS_MMCSD_RAW_MODE_U_BOOT_USE_SECTOR is not set

That's because it never used that (intentionally). We always booted from
a partition ID, but the old versions had all methods enabled, and that
papered over the fact that they were mangled into something that was
supposed to use partition ID but used bits from the other paths as well
to compensate mistakes.

> 
> I simply used the A2 partition settings on SPL and it booted properly w/o issue.

At least for the CycloneV with DE0 Nano SoC, this was broken. Maybe your
setup was lucky to still work, but I doubt that was by design - see all
the changes I had to made.

> All you need is to split the u-boot.img to fat partition.
> I build this and boot almost w/o any major issue since the changes from
> mixed setup to 3 individual setups boot by "offset", "#", and "Type".
> Still not sure what you have trouble with in the first place?
> So I really dont understand what this is trying to fix.
> 
> Maybe T.F. can have more better comment but u-boot after
> 2025.07 no issue on boot by type=A2. I did not try boot by#,
> b.c. raw type must be sought from first.
> 

Just study my series to see all the breakages. It was never clean,
though, and that may explain working setups by luck.

Jan

-- 
Siemens AG, Foundational Technologies
Linux Expert Center