Jonathan Gray <jsg@jsg.id.au> wrote:

> On Thu, Jan 16, 2025 at 07:08:32AM +0000, jon@elytron.openbsd.amsterdam wrote:
> > Hello, I would like to share a trivial patch to enable some ahci
> > devices.
> 
> The RAID devices are intentionally not listed.
> 
> sys/dev/pci/ahci.c rev 1.163
> 
> ----------------------------
> revision 1.163
> date: 2010/04/16 22:15:39;  author: kettenis;  state: Exp;  lines: +1 -10;
> Remove Intel RAID IDs for now.  They can be added back when code has been
> written to detect the Intel RAID metadata format such that we no longer risk
> overwriting it.  If this makes you lose your disks, you'll need to change the
> controller type from RAID to AHCI in the BIOS to get them back.
> 
> ok marco@, deraadt@
> ----------------------------

The problem is that these RAID modes use sectors at the end of the disk for
a specific purpose.  If we expose this mode as a regular OpenBSD disk, it
becomes possible for GPT/MBR/disklabel to cover those last few sectors,
and they could get changed.  If you are expecting to dual-boot, you are
in for surprises.

A long long time ago, I had a theory that we could detect this situation,
parse those RAID configuration offsets, and automatically create a softraid
in that instance.  It would merge multiple disks in this mode into a softraid
sd.  If a sole, or unconfigured disk was found, it would create a single-raid.
This type of softraid would only read the BIOS/disk configuration, to create
a softraid disk-span.  It would probably not update the configuration, because
that's a lot more code.