Re: 8.0-BETA1 bsdlabel broken?

On Jul 11, 2009, at 3:22 PM, Eygene Ryabinkin wrote:

> Aragon, good day.
>
> Sun, Jul 12, 2009 at 12:13:19AM +0200, Aragon Gouveia wrote:
>> Eygene Ryabinkin wrote:
>>> Please, try the attached patch -- it should heal bsdlabel and it  
>>> will
>>> show/use the proper offsets everywhere.  Works fine for me --  
>>> offsets
>>> for 7.x and 8.x are the same.
>>
>> I've seen the reported problem on my BETA1 systems too.  I've also
>> noticed this in my bootup dmesg:
>>
>> GEOM: ad4s2: geometry does not match label (255h,63s != 16h,63s).
>
> Different and mostly harmful (as per UPDATING, 20090320) problem.

I think you meant to say harmless and not harmful.

>>
>> Is it possible sysinstall is a culprit in all this?
>
> No, the problem is that bsdlabel is missing the so-called "MBR offset"
> from GEOM, because in 8.x GEOM_PART_* are used and they have no such
> geom verb (and anyway, 8.x's bsdlabel queries BSD geoms for it, not  
> the
> PART_BSD ones).

Actually, yes. This is a problem in sysinstall. The "MBR offset"
you're talking about does not relate to the geometry of a device.
It relates to offsets in the disk label.

The warning Aragon talks about is the result of having the underlying
disk report 16 heads and 63 sectors/track, whereas the BSD disklabel
itself contain values of 255 for heads and 63 for sectors/track. This
means that the BSD disklabel was constructed with values that do not
correspond to what the disk reports. This is where sysinstall comes
into the picture.

Note that it's also possible that the underlying disk changes geometry.
If I install over USB onto a SATA disk (using a USB to SATA converter)
and then later install the SATA disk in a machine and connected to a
SATA controller, the geometry may change. In the first case the disk
is accessed as daX and in the second case the disk is accessed as adX.
SCSI/CAM and ATA do not agree on default geometries AFAICT...

-- 
Marcel Moolenaar
xcllnt_at_mac.com