cvs commit: src/sbin/gpt gpt.8 gpt.c show.c

[Maxim Sobolev]

Marcel Moolenaar wrote:
> 
> On Jun 22, 2006, at 5:00 PM, Maxim Sobolev wrote:
> 
>> Marcel Moolenaar wrote:
>>> marcel      2006-06-22 22:22:32 UTC
>>>   FreeBSD src repository
>>>   Modified files:
>>>     sbin/gpt             gpt.8 gpt.c show.c   Log:
>>>   Improve support for Intel based Macs:
>>>   o  Accept hfs as partition type.
>>>   o  Print Apple HFS partitions using a friendly name.
>>
>> While we are on subject, I have a patch which makes geom_gpt friendly 
>> to the disks partitioned using BootCamp. Basically, for compatibility 
>> with non-GPT operating systems, BootCamp creates three records in the 
>> MBR: one GPT protective, covering the whole unit, the second one for 
>> HFS and the third one for non-Apple operating system. However, since 
>> geom_gpt considers protective MBR to be invalid unless it's the only 
>> valid MBR record, it doesn't attaches to such disk.
>>
>> Do you have any objections to lifting such check?
> 
> I don't mind changing geom_gpt to work for Intel based Macs, but I'd
> like to avoid simply removing checks and thus accepting just about
> anything. I'd like to keep the checks strict enough that we don't try
> to interpret garbage as partition information. If we allow the PMBR
> to have actual partitions, I'd like to extend the checks to cover
> the partition information itself. For example, we now check that the
> table only has partitions of type 0xEE (if at all). This is a sloppy
> test by any standard. If we allow "normal" partitions, I'd like to
> enforce that there's exactly 1 partition of type 0xEE and it covers
> the whole disk. This way we make sure that it's still pretty much a
> PMBR by intend, even though it really is used as an MBR.
> 
> Does that sound reasonable to you?

Thanks, that's exactly what my patch does!

Not that it matter much, but linux uses the same approach.

-Maxim