Hi Vinicius,
On 10/9/26 18:14, Vinicius Pedrosa wrote:
Hi Andre,
Sorry about the wording. I kept repeating it to give context to people
who might be reading the thread for the first time, but I see how it
came across. I'm building on Yixun's work and yours, and I should have
made that clearer.
no worries, I got you ;-)
Honest question, is it really *you*, or is it some LLM?
Yes, it's really me. I've been working with embedded systems for many
years, but I'm new to public contributions, as you've probably noticed.
I do use AI to help me research tools, handle some tedious tasks, and
improve my English when writing emails. English isn't my native
language, and I sometimes rely on it too much for the wording.
It's not the first language for most of the sunxi folks either
(including me). Just try your best, we are used to decipher slightly
broken English on a regular base. If you really must, feel free to let
AI help you with the wording, but try to not let it wash down your style
too much, or make it too perfect. Translating incoming emails into
Portuguese(?) is fine, of course.
But maybe *you* learn a bit or two by doing it yourself ...
But the
hardware work is real, and I take responsibility for what I send.
Cool, that's good. I already figured as much, just wanted to double check.
Why not? Probing DRAM is somewhat the idea of that routine, and it
should restore the previous values it overwrote? So have you tried
running that?
I haven't. I was concerned about overlapping BL31, but I should have
looked more carefully at how the routine works before ruling it out.
I'll check the probe range and try it.
It's not the nicest test in the world, and it stomping over BL31 is
sketchy at best, and just works because we don't really protect BL31 if
it lives in DRAM. If we would (and we should!) it might actually trigger
an EL3 exception, so the routine really should observe reserved memory.
Also I am not sure what exactly happens when it goes beyond the
installed size, I guess that differs by system.
BUT: it might just work, and then it's good enough for that time where
we rely on boot0 for DRAM init.
Cheers,
Andre
Understood that the FEL workaround is temporary, so I'll drop the size
header idea. If it's useful, I can test your U-Boot proper series on the
A7Z when you post it.
Thanks for your time and guidance.
Vinicius
