[Bug 1272604] Re: Ubuntu install fails to correctly configure GRUB with RAID + LUKS in UEFI mode
Anatoli
1272604 at bugs.launchpad.net
Thu Jan 30 17:37:59 UTC 2014
This problem was actually composed of two independent bugs. The first
one was related to grub: here are the details and the solution:
http://ubuntuforums.org/showthread.php?t=2201239.
Another bug was in the Ubuntu installer incorrectly informing initramfs-
tools about the necessary drivers to load (the USB keyboard driver was
not loaded at the LUKS passphrase stage); here are the details and the
solution: http://ubuntuforums.org/showthread.php?t=2201983.
With these 2 bugs fixed we could correctly boot the server.
Nevertheless, we found another problem when we tried to test the RAID
for automatic recovery from a single disk failure: we powered off the
server, removed one disk and booted it again. The mdadm could not
correctly activate the degraded arrays in spite of the bootdegraded
options in mdadm config for initramfs and the kernel param, and the
cryptosetup stuck for 180s waiting for the root device, then dropped to
the initramfs prompt. We finally solved it as well and later found that
there are a number of open bug reports (one from 2008!!) for this issue.
Here are the details:
https://bugs.launchpad.net/ubuntu/+s...up/+bug/251164. For the solution,
check the comment #30.
--
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to grub-installer in Ubuntu.
https://bugs.launchpad.net/bugs/1272604
Title:
Ubuntu install fails to correctly configure GRUB with RAID + LUKS in
UEFI mode
Status in “grub-installer” package in Ubuntu:
New
Bug description:
We are trying to install Ubuntu 12.04.3 x64 Alternate (ubuntu-12.04.3
-alternate-amd64.iso) on one IBM machine (IBM System x3100 M4) with 2
x SATA 1TB WD RE4 booting from an USB drive in UEFI mode. At all steps
we choose the most basic/default options.
At the Partition disks step we create 3 partitions on each disk (GPT
partition table): the 1st one is EFI Boot Partition (200Mb), a 2nd
partition of 100Mb, and a 3rd partition of 10Gb. Then we create 2
RAID1 arrays on the 2nd and 3rd partitions and then we create an
encrypted partition (with the default options) on top of the 2nd
array. We format the 1st RAID array in ext2 and assign it /boot
mounting point and the 2nd one is formatted in ext4 for / mounting
point. The layout looks this way:
Encrypted volume (md1_crypt) - 10.0 GB Linux device-mapper (crypt)
> #1 10.0 GB F ext4 /
RAID1 device #0 - 99.5 MB Linux Software RAID Array
> #1 99.5 MB F ext2 /boot
RAID1 device #1 - 10.0 GB Software RAID device
> #1 10.0 GB K crypto (md1_crypt)
> 512.0 B unusable
SCSI2 (0,0,0) (sda) - 1.0 TB ATA WDC WD1003FBYZ-0
> 1.0 MB FREE SPACE
> #1 199.2 MB B K EFIboot
> #2 99.6 MB K raid
> #3 10.0 GB K raid
> 989.9 GB FREE SPACE
SCSI4 (0,0,0) (sdb) - 1.0 TB ATA WDC WD1003FBYZ-0
> 1.0 MB FREE SPACE
> #1 199.2 MB B K EFIboot
> #2 99.6 MB K raid
> #3 10.0 GB K raid
> 989.9 GB FREE SPACE
Then we proceed with the rest of the install using the most basic
options. After the reboot an error flashes for an instant: "error:
efidisk read error" and we are brought to the grub> prompt.
We have tried boot-repair and a lot of other solutions recommended for
similar issues on the net, nothing solves the problem.
Same install on a generic PC (without UEFI) or in VMWare boots
correctly without any special interventions. Same install on the IBM
machine without RAID/encryption boots correctly as well. Same install
without encryption but with RAID produces the grub> prompt, but is
fixable with boot-repair.
We know that the same or very similar problem happens on a number of
IBM System x machines, looks like the lines x3100 and x3200 M3/M4 are
all affected. It looks like the problem is related in some way to UEFI
mode and GRUB installer, but we aren't sure.
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/grub-installer/+bug/1272604/+subscriptions
More information about the foundations-bugs
mailing list