Bug#1088151: Bug: GRUB fails to boot after minimal install on Dell 2950 RAID; XFCE resolves issue

[Pascal Hambourg]

Control: found -1 2.06-13+deb12u1

On 24/11/2024 at 08:23, CharlesWorks Support wrote:
> Version: 2.06-13+deb12u1  (I do not know if this is the exact version used in the minimal install)

You should not add any comment to pseudo-headers, it prevented the BTS 
from identifying the affected version.

> Reinstalling Debian 12 using the XFCE desktop environment option resolves the issue, and the server boots successfully. This suggests a missing dependency, module, or configuration step in the minimal installation process that is addressed when XFCE is selected.

AFAIK the only differences when installing a desktop environment are:
- extra files are written to the root filesystem before GRUB packages 
are installed;
- package desktop-base installs grub background images which can be used 
in grub.cfg.

> ________________________________
> Steps to Reproduce
> 
>    1.  Install Debian 12 (Bookworm) on a Dell PowerEdge 2950 server with the following configuration:
>       *   PERC 6/i RAID controller configured as RAID 5 with six 2TB SAS drives.
>       *   Partitioning set to use the entire disk with default options.
>       *   Choose the minimal installation option during the installer.

There is no "minimal installation" option in the Debian installer. Which 
installation media did you use ?

> ________________________________
> Actual Behavior
> 
>    *
> GRUB fails to boot, displaying the error message:
> 
> GRUB Loading.
> Welcome to GRUB!
> error: attempt to read or write outside of 'hd0'.
> grub rescue>
> 
> 
>    *
> Running ls in the GRUB rescue shell outputs:
> 
> grub rescue> ls
> (hd0) (hd0,gpt3) (hd0,gpt2) (hd0,gpt1)
> 
> 
>       *   Only (hd0,gpt2) contains the /boot directory and shows a valid filesystem.
>       *   (hd0,gpt1) and (hd0,gpt3) return "unknown filesystem."

If you selected "Guided - all in one partition" partitioning, then 
(hd0,gpt1) is the BIOS boot partition (for grub-pc on GPT), (hd0,gpt2) 
is the root partition and (hd0,gpt3) is the swap partition.

> Attempting to set the root and prefix in GRUB rescue mode (set root=(hd0,gpt2) and set prefix=(hd0,gpt2)/boot/grub) does not resolve the issue.

It won't help if these values were already set. Did you check the 
original values ?

> Commands like insmod normal or insmod ext2 also fail with the same error.

This means that GRUB ext4 driver fails to read requested files. Possible 
causes include a bug in GRUB disk or ext2 driver or a bug in the BIOS 
RAID disk driver. It can happen on very large disks.

Did you try to partition the RAID array with a separate 1GB /boot 
partition at the beginning ?

> ________________________________
> Workaround
> Reinstalling Debian 12 with the XFCE desktop environment option allows the system to boot correctly.

I am afraid this is only by luck and any grub update may break the boot.