From mboxrd@z Thu Jan 1 00:00:00 1970 From: Matt Darcy Subject: Linux Raid boot - stuck in a raid start loop - Advice Date: Wed, 30 Nov 2005 09:41:54 +0000 Message-ID: <438D73E2.70502@projecthugo.co.uk> Reply-To: kernel-lists@projecthugo.co.uk Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Sender: linux-raid-owner@vger.kernel.org To: linux-raid@vger.kernel.org List-Id: linux-raid.ids Hi all, this is not a bug, this is a tricky situation due to a few bad choices on my part. I am using a 8 port Sata controller (MV88SX5081 8-port SATA I PCI-X Controller) with the 2.6.12.6 kernel. I am using the 3rd party sata_mv drivers as the sata_mv drivers included in the 2.6.15 kernel are not stable enough yet - This problem will go away when these drivers become stable. The 3rd party driver limits me to 2.6.12.6. The 3rd party driver can only be built as a module, and not directly into the kernel - just making that clear before people suggest building it into the kernel. I have 6 Sata disks hanging off this controller that I am trying to put into a raid 5 configurations, which can then be included into LVM2. I am using mdadm to try to migrate away from raidtools. The problem I have is that the kernel boots sees the sata raid boot disks (on a diffect controller) and raid starts them - sysfs sees the raid devices as block devices and advertises them to udev - which then creates the /dev/md* entried (1 - 5) for the root disks, the file system is mounted and the init scripts run - which then loads the 3rd party module for the sata controller - which then makes the 6 sata disks visable, but becase the kernel was not aware of these disks at boot/raid time it does not start the raid array - which does not allow /dev/md6 to be created. I then have to manuall create /dev/md6 - but even then my attempt to start the existing raid device fails (I believe this is because the kernel is not aware of these disks as raid devices). I can however recreate the raid device (mdadm --create) but this obviously wipes all the data - but it does however prove the the /dev/md6 device is functioning correctly. Any advice on a process to get around this poor loop would be most appriciated, I have tested against the 2.6.15-rc2 kernel using the inbuild drivers and this problem goes away, /dev/md6 is created and the raid5 device is started and running well. The machine just becomes unstable when accessing the raid5 device using the driver from 2.6.15-rc2. Many thanks, Matt