> This is not QSPI, but HF.

Ah, okay.

> Building a new firmware for R-Car H3 ES1.0 with HF unlocked will be
> complicated, as it is not supported by upstream TF-A.

You mean QSPI here?

> Note that HF also fails to probe on R-Car M3-W and M3-N ES1.0.

Do you have this patch form Andrew in your tree:

[PATCH] memory: renesas-rpc-if: Avoid unaligned bus access for HyperFlash

Even if so, I don't think that reverting patches is the solution. As you
could see from Andrew's patch, HyperFlash was also broken before and it
just may need more fixes for Gen3 perhaps? IIRC my patches didn't break
Andrew's tests but maybe we should ask him again. Maybe Andrew has also
some more ideas, I only did QSPI.