initramfs support
Alexander E. Patrakov
patrakov at ums.usu.ru
Sat Jul 21 09:01:31 MDT 2007
Bryan Kadzban wrote:
> Not sure what the testing came up with, but I figured out a way to get
> it to work:
>
Congratulations - you did more than Debian. They work only with
/dev/md<number> raids, not with named ones.
> mkdir /dev/md
>
> for array in /dev/md[0-9]* ; do
> # subshell so the $MD_* variables don't carry between arrays
> (
> eval `mdadm --detail --export $array`
> [ -z "$MD_NAME" ] && exit 0 # v0.9 metadata has no name
>
> real_name=`echo "$MD_NAME" | cut -d: -f2`
> # if there's no host field (no colon), use the full name
> [ -z "$real_name" ] && real_name="$MD_NAME"
>
> cp -a $array "/dev/md/$real_name"
> cp -a $array "/dev/md_$real_name"
>
Isn't /dev/md_$real_name normally a symlink to /dev/md/$real_name?
> # might or might not be needed; my setup works without:
> rm $array
>
Please don't remove this device node.
> )
> done
>
> I have not tested this on a metadata=0.9 array, but as long as those
> don't provide an MD_NAME in --detail --export, it should work fine.
>
--
Alexander E. Patrakov
More information about the lfs-dev
mailing list