[PATCH 4/8] Revert "fdt: Allow the devicetree to come from a bloblist"

[Simon Glass]

Hi Tom,

On Mon, 2 Dec 2024 at 17:55, Tom Rini <trini at konsulko.com> wrote:
>
> On Mon, Dec 02, 2024 at 05:22:35PM -0700, Simon Glass wrote:
> > Hi Tom,
> >
> > On Mon, 2 Dec 2024 at 09:22, Tom Rini <trini at konsulko.com> wrote:
> > >
> > > On Mon, Dec 02, 2024 at 09:26:42AM -0500, Raymond Mao wrote:
> > > > Hi Simon,
> > > >
> > > > On Sun, 1 Dec 2024 at 09:43, Simon Glass <sjg at chromium.org> wrote:
> > > >
> > > > > This stops coral, bob and kevin from booting.
> > > > >
> > > > > The correct way to do this was always to use a Kconfig option, so let's
> > > > > first revert this broken idea.
> > > > >
> > > > > OF_BOARD and BLOBLIST should be general and hardware agnostic,
> > > > do you have more information on why they don't work for a few boards?
> > > > Maybe the way to handle these logic within certain boards should be
> > > > changed other than to revert this?
> > >
> > > Yes, I'd like to understand the whole problem better since we had a
> > > forever-and-ever thread, compromised to what's in tree now and now you
> > > want to revert the compromise and bring is back to what was argued
> > > against for forever-and-ever, at least from a very quick read.
> >
> > I'm sorry but I believe I have explained the problem in that forever
> > and-ever thread, so I don't really want to re-run it. It may be that I
> > was right all along?
>
> Or maybe your use case is still badly explained, and your changes will
> instead break other platforms.
>
> > Raymond, I cannot pass the DT in the bloblist from TPL to SPL on x86.
> > There is not enough cache-as-RAM, for a start. Plus the DT is in the
> > memory-mapped SPI flash, so it makes no sense...
>
> DT using x86 is a pretty odd case. I don't recall if you mentioned that
> before, and it's likely everyone else assumed you were talking about
> something more standard. Of course, you're listing one x86 platform and
> two ARM platforms as needing this change. Non-chromebook Rockchip isn't
> generally broken (or several other people would be complaining).
>
> So something that may or may not be explained in the longer older
> threads is how the firmware in these platforms are doing anything with
> the device tree?
>
> But, the biggest problem I see with this portion of the series is that
> patch #4 is 'Revert "fdt: Allow the devicetree to come from a bloblist"'
> and patch #5 is "fdt: Allow the devicetree to come from a bloblist".
> What you need to do instead is squash all 3 of these in to one commit
> and then write a new message that explains what exactly you are
> changing, and why, instead of making someone else do that in order to
> review  your changes here.

As I said, I do not wish to go through all of that again. It is still
a bad memory every though it was a year ago. I would like to build on
what I have here, to resolve any future breakages.

Regards,
SImon