Re: hd numbering in 9.0beta1

On Mon, 29 Aug 2011, Roger Genre wrote:

> Hi everybody,
>
> I would point out a problem related with the new way, coming in 9.0, to hard 
> disks numbering.
>
> As far I remember (5.0 ?), hardware detection  of H.D.'s at O.S. boot-up 
> numbered every channel potentially able to attach a disk to, and tagged the 
> disks really attached with the number of his control channel; the sequence 
> (from lowers to highers numbers) begins with scsi or scsi-like (e-sata, usb, 
> fire-wire,...) controllers and ends with the controllers directly depending 
> from the chipset (sata at this time).
>
> Such strategy allows to attach easily a new mass-storage device without 
> modifying the disks numbering and thus the relevant fstab files.
>
> 9.0beta1 use a different numbering strategy,(with a similar sequence in 
> harware detection) tagging succesively detected disks with adjacent numbers.

The best way I can put it has already been noted in the archives several 
months back:

- http://lists.freebsd.org/pipermail/freebsd-current/2011-April/024110.html
- http://lists.freebsd.org/pipermail/freebsd-current/2011-May/024495.html
- http://lists.freebsd.org/pipermail/freebsd-current/2011-April/024233.html

I don't have the corresponding commits right now, but I could dig them up 
as they spawned a large discussion thread as well.

The basic gist is that several folks agreed that:

1. GEOM/UFS labels were the only way to go.
2. There are some caveats to using GEOM labels that discourages use as a
    means of deterministically determining mountpoints.
3. A compatibility shim was added for ata -> atacam transitioning; see
    kern.cam.ada.legacy_aliases in /sys/cam/ata/ata_da.c

Cheers,
-Garrett