Re: GEOM and moving to CURRENT from 7.1

On Jan 15, 2009, at 11:24 AM, Dimitry Andric wrote:

> On 2009-01-15 19:05, Marcel Moolenaar wrote:
>>> Hmm, strangely enough I have multiple systems which really do use  
>>> ad0s1
>>> while installed DD using sysinstall, and have been using that  
>>> since years...
>> This is a problem with sysinstall. It looks like DD is broken.
>> MBR partitions are always created.
>
> Indeed, and the MBR itself is filled with /boot/boot1, plus a  
> partition
> table.  It is possible this was done to appease some PC BIOSes,  
> because
> some of those refuse to boot, if there is no "valid" MBR with an  
> active
> partition...

You need a boot sector for the BIOS to jump into, but
it doesn't have to be a MBR with slices. That's why
a DD has something that looks an awful lot like a MBR
in the first sector. It could very well be boot sector
of an FAT* file system.
>
>> GEOM_PART correctly uses the
>> BSD disklabel that's in the 2nd sector, but GEOM_MBR claims the
>> MBR being unaware that the disk is DD.
>
> But how can this be, if I don't have GEOM_MBR in my kernel config?   
> (I'm
> using GENERIC, in fact.)

sys/i386/config DEFAULTS is included implicitly. It
used to have GEOM_BSD and GEOM_MBR.

>> libdisk is broken in that it should
>> not create MBR slices for DD configurations to begin with...
>
> As stated above, this might be on purpose; enough braindead BIOSes out
> there... :)

Creating ambiguous, conflicting and/or overlapping partition
information to work around broken BIOSes to support DD, is
not a solution *if* it was on purpose. You want it to fail
so that the user partitions the disk in the normal way.

Put differently: if there are enough braindead BIOSes out
there that won't support DD, then DD should be de-supported
by FreeBSD.

-- 
Marcel Moolenaar
xcllnt_at_mac.com