On 01/27/10 06:34 AM, Jan Damborsky wrote:
> On 01/21/10 08:07 PM, Dave Miner wrote:
>> On 01/21/10 12:07 PM, William Schumann wrote:
>>> initially misnamed: auto-install creates bogus partition table
>>>
>>> http://cr.opensolaris.org/~wmsch/bug-13331/
>>> http://defect.opensolaris.org/bz/show_bug.cgi?id=13331
>>>
>>> Interoperability issue: in some cases, Jumpstart will create an
>>> alternates, or ALT slice. OpenSolaris Target Instantiation is not
>>> preserving the ALT slice presently. The fix starts new slice creation
>>> after the BOOT slice and the ALT slice, if present.
>>
>> Looks OK. I do think we need to assess whether creating alternate
>> slices on non-SCSI disks is required in the installer, though.
>
> I had discussion some time ago with Shidokht Yadegari about how
> user could get rid of existing ALT slice (format -e allows this)
> he pointed out with respect to the ALT slice during that discussion:
>
> ...
> Yes, We don't create slice 9 for scsi disks on x86.
> DKIOCADDBAD ioctl has not been supported on scsi disks since
> PSARC 1997/281 (guess who submitted it ;-)), when we switched to use
> sd instead of cmdk for scsi disks. Later on, creating the alternate slice
> for scsi disks in sd was removed all together.
>
> I agree with this slice being kind of useless and probably nobody
> uses addbadsec anymore. However imho, the installer has to agree
> with kernel and disk partitioning utility. If we want to not make
> slice 9 on these devices anymore, I believe the proper way of fixing
> it would be by EOLing the addbadsec, DKIOCADDBAD ioctl and changing
> target driver, utilities and installer to all agree on that.
> Otherwise if addbadsec is still not ready to be EOLed, then installer
> should create the slice and be consistent.
>
> I think we need a low pri bug to be filed for caiman. We may need a
> separate RFE later probably against cmdk and format to get rid of alt
> slice.
> Perhaps it should have dealt with when sata framework was introduced and
> some ide devices migrated from cmdk to sd.
> ...
>

If you wouldn't mind following up with Shidokht and agreeing on where 
this needs to go, I'd appreciate it.

Dave

Reply via email to