luks related change is prevented boots of system with encryption over raid

Bruno Wolff III bruno at wolff.to
Wed Oct 29 01:26:56 UTC 2008


On Tue, Oct 28, 2008 at 17:40:51 -0700,
  Tom London <selinux at gmail.com> wrote:
> 
> I just reverted mkinitrd (and nash and libbdevid-python), rebuilt the
> initrd and all was OK.

Yeah, I meant mkinitrd needed to be reverted. I had to access the bugzilla
through lynx and have been bouncing back and forth.

However, there are two separate problems. One is not running crypt setup
another is that plymouth is using a different way to name encrypted raid
devices than udev.

I got my machine booted up into its (mostly) normal state now but there
are a lot of errors I need to skip during the boot.

To fix things I needed to name / and swap with /dev/mapper/luks-*long luks id*
in /etc/fstab and /boot/grub/grub.conf and leave /home and /play as
/dev/mapper/luks-md? . When udev runs it tries to mount the old style names
for / and swap again and things fail. But I can hit return a bunch of times
and eventually the system finishes booting.

This is what my device names look like:
[bruno at cerberus ~]$ ls /dev/mapper
control                                    luks-md3
luks-9979e281-5710-476f-8536-e49846d53398  luks-md4
luks-9cfcdd2a-d2ff-4c30-818c-344764c7cc21

I wouldn't mind seeing a switch over to using the luks ids but that is probably
going to bite people with existing setups and its probably too late to
develop this since we are at the freeze. So I suggest changing plymouth
(or whatever) back to using names like udev currently does and consider
changing later with some way of simply updating for people.




More information about the fedora-test-list mailing list