Subject: Re: Experimental support for ATA "RAID" volumes
To: Jason R Thorpe <>
From: Manuel Bouyer <>
List: current-users
Date: 01/30/2003 21:28:44
On Mon, Jan 27, 2003 at 10:53:41AM -0800, Jason R Thorpe wrote:
> Folks...
> I just checked in some experimental support for RAID volumes found
> on ATA "RAID" controllers; not the smart controllers (like e.g. Adaptec
> and 3Ware), but the dumb ones ... in this particular case, Promise.
> These controllers are just IDE controllers with a BIOS that can configure
> RAID volumes, write config blocks to the disks, and do I/O to the volumes,
> for the purpose of booting.  The OS has to implement the RAID in software,
> using the configuration data that is written to the disks by the BIOS.
> The functionality is pretty basic at this point, but it works well
> enough for me to use the RAID0 volume on this new Intel server I have
> (which has a Win2k Advanced Server installation that I need to preserve).
> Attached are boot messages from my machine (a dual 2.4GHz Xeon w/ HT,
> though I am only using a uniprocessor kernel on it at the moment); it
> is using an NFS root right now, since I was using PXE to load kernels
> while working on the ATA RAID code.

I've been thinking about this each time I reboot a machine with on-board
promise ...
What are your plans for raid1 or more ?
We should probably be using the raidframe framework for this, and maybe for
raid0 too. My idea for this was to make the raidframe autoconfig stuff more
flexible, so that it could be autoconfigured from its own on-disk data,
or from the ataraid pseudo-device.

Manuel Bouyer <>
     NetBSD: 24 ans d'experience feront toujours la difference