Re: booting to install - GRUB

Hi Adrian,

John Paul Adrian Glaubitz wrote:
> GRUB probably doesn't work on these very UltraSPARCs, so I'm afraid you
> have to pick an older image which still uses SILO.

For the record on this thread, the "workaround" is to remove the 
framebuffer. Apparently GRUB, even if booted via seriall console, 
enables itself if a FB is present (even if neither monitor nor keyboard 
are connected).


>> Here I went all with the most default setup, I got this warning:
>>
>>       x The disk has 69369 cylinders, which is greater than the maximum
>> of  x
>>       x 65536. x
> Yes, this is a warning that you can safely ignore. I think it came because
> the /boot partition size was increased recently without checking first whether
> this size would work on sun4u systems and/or without adjusting the upper
> limit for this warning.

Yes, I have seen a 256MB boot partition, useful when testing a couple of 
kernels nowadays.
The hard disk has been totally wiped from Solaris to Linux, so no 
"fears" in that case, except that maybe some boundary is off and data 
loss could occur. We shall see through testing.


> Please keep in mind that we're using a shared codebase and if a 
> maintainer
> thinks the change he made on x86_64 will just work on any other architecture,
> we will get regressions like these.

True, but it is also the "avantage" of using shared setup and GRUB 
instead of e.g. SILO with different boot partition or such things.

>> Now the long process of setting it up from the minimal CD started: no
>> sshd, no proper keys, no gpg, wrong sources... exhausting.
> What installation CD is that? Please always link the ISO so I know what
> you're talking about.

last Debian 12 ISO I found:

https://cdimage.debian.org/cdimage/ports/12.0/sparc64/

dated 2023-05

I found no ready 13 for sparc64 and I wanted to start with a "stable 
kernel" which has proven advantageous.
Anyway, no big hassle. I set the repository as trusted, then slowly 
worked myself up.


> FWIW, this is an unpatched kernel. It does not contain any of the recent
> fixes, so you're not really testing anything there.

testing how it "doesn't work" as expected! Ah ah. Good to have reference 
points.

Riccardo