intel fakeraid (imsm) + raid0 root partition + mdadm + genke

[Ginta]

HI there, I am two days trying to solve this nasty issue with my raid0 setup on the root partition.

I've read and read, without coming to a solution, but now I can manually boot my system by dropping to a rescue shell, and issuing the command:
mdadm -I /dev/md127

Once I do that, I hit ctrl+D and the system boots, but.. isn't the purpose of genkernel-next and --mdadm option to automatically do this for me before init?

Any hint so the system can boot automatically?

Thanks
_________________
Some people should use emerge brain on theirselves...

[frostschutz]

Are you dual boot with Windows? Otherwise just get rid of the imsm and use pure mdadm.

What dost your mdadm.conf look like?

[Ginta]

Yes, I do use dual boot with windows, otherwise Id just use btrfs raid, which is not so complicated.

Anyway, this is my file

[Ginta]

What more info do you need from my system?

I can also tell that once I type shell when it says no / partition found, and execute the command manually, after boot, my /proc/mdstat seems good:

[frostschutz]

That does not look wrong to me. So for some reason something goes awry in your Initramfs. Unfortunately I don't have first hand experience with genkernel(-next)/imsm so maybe somebody else has some ideas.

Edit:

You could try changing /dev/md/imsm0 and /dev/md/vol0 to /dev/md126 /dev/md127 (or /dev/md0 /dev/md1 if you prefer those numbers) resp. in your mdadm.conf just in case initramfs gets confused about /dev/md/NAME style. Then update initramfs and try again.

[buddyabaddon]

I had the same issue with my root imsm raid0 array as well.

My initramfs is created via 'genkernel-next --udev --mdadm --no-dmraid' and it too would always drop me into an emergency shell upon boot. From there I would run 'mdadm --run /dev/md127' and my root partition would be forced active and the system would continue to boot. Back before systemd with the old genkernel and initrc, everything worked as expected.

It turns out the issue is a bug in genkernel-next. The udev rules that are part of the sys-fs/mdadm package are not appended to the initramfs causing the behavior you are seeing.

I modified genkernel-next's 'gen_initramfs.sh' script to include these udev rules (63-md-raid-arrays.rules and 64-md-raid-assembly.rules) in the initramfs and everything begins to work as it should. The container and arrays assemble automatically and the kernel successfully loads root.

Here's the submitted fix to genkernel-next:
https://github.com/Sabayon/genkernel-next/pull/57