[PATCH 0/8] QI: make phase2.c easier to change

[Werner Almesberger]

Andy Green wrote:
> Hm after completely wasting last night on a thread where people --
> including you -- come out of the woodwork to explain the value of
> talking to upstream before doing any work, maybe we should have had the
> larger discussion here first.

Heh, I thought you preferred patches over discussion :-) Anyway,
what I was after was checking that booting zImage really worked as
I had described. But when I looked at phase2.c, I realized that I
should do a bit of cleanup before trying to understand the logic.

The patches may look impressive but what most of them do is move
code around and change whitespace, so it wasn't all that much work.
If you don't like the changes, please feel free to ignore them. But
I think phase2.c is unnecessarily hard to understand in its present
form, so I think that it would benefit from something at least along
the lines of the restructuring I did.

> In the thread you mention, you point out that pivot_root will do in the
> shared kernel case and then kexec and zImage is not needed.

Yes, but that's just for this specific case. If we want to have an
environment that gives the full set of choices - including loading
a new kernel - then we still need kexec.

> Since 9 of those 11 patches are just whitespace / breaking out existing
> working code into the same code in subfunctions, and "supporting zImage"
> in patch 3/3 is a very small and simple piece of additional code mainly
> checking a magic I guess they're fine anyway.

Did you get them all ? Only three of the mails have made it back to
me so far :-( I see them in the list archive, though. Maybe our list
processor doesn't like me today ..

- Werner