On Sun, May 22, 2005 at 12:06:27PM +0200, Marco d'Itri wrote: > On May 22, martin f krafft <[EMAIL PROTECTED]> wrote: > > > Erik said that if initrd brings up /dev/md0, partitions from other > > software RAID devices (e.g. /dev/md7) cannot be brought up since the > > above code skips the MAKEDEV call in the case of presence of > > /dev/md0. > I do not know how the initrd work, but does it actually do this?
Quoting Maks from #304483: > > I can't try this right now but I doubt this would help. The reason is that > > when > > the system boots initrd would not know how to assemble /dev/md1 because > > .../initrimg/script does not contain "mdadm -A" record for /dev/md1. This is > > the root of the problem. > > no it is not! > initrd-tools enables the root partition for the pivot_root, > and if existing the swap partition, > everything else need to be done by mdadm init scripts. I guess that means it actually does this; this matches with what the code says. For what it's worth, I think that's proper behaviour for mkinitrd. > > 1 check for another device node, e.g. /dev/md5 and hope that the > > initrd only ever configures /dev/md0 > I think that this is the best solution. If the initrd only activates the > / array then you can check for md1 and md/1. Users can dedicate any partition to swap, and it will be activated by initrd. In 304483, user has md2 and md3 as root and swap, these are activated, md1 is /boot and remains inactive. If you have to pick just one device, use /dev/md15. That's generated by MAKEDEV but not terribly likely to be in use. Regards, Erik -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
