[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [PATCH 3/5] Enforce GPT on EFI and MSDOS on non-EFI x86 boot disks.

On 12/10/2009 10:01 AM, Peter Jones wrote:
> On 12/09/2009 08:59 AM, Jeremy Katz wrote:
>> On Wed, Dec 9, 2009 at 6:48 AM, Hans de Goede<hdegoede redhat com> 
>> wrote:
>>> On 12/09/2009 09:00 AM, David Lehman wrote:
>>>> +        if not self.isEfi and disk.type != "msdos":
>>>> +            errors.append(_("%s must have an MSDOS disk label.") %
>>>> req.disk.name)
>>>>           return errors
>>>>       def setDefaultPartitioning(self):
>>> I'm not quite sure about this one, GPT disks can have an msdos
>>> compatibility
>>> table at the beginning of the disc (created using gptsync), and some
>>> EFI machines
>>> can boot from msdos labelled disks.
>> The first chunk should be fine (even if the MSDOS compat table is
>> there, we'll see it as GPT).  The second one, yes, having this will
>> break installing on Intel Macs booting via the BIOS emulation vs going
>> through EFI.  And given some of the oddities of their EFI
>> implementation, there are some advantages of not booting from it
>> directly
> Hrm?  How will it break booting through their CSM?  self.isEfi will be
> /False/ in that case.

I was assuming here that x86 has its own checkBootRequest method, which it currently doesn't.

> Hans is right that some UEFI machines can boot from MBR/MS-DOS partition
> tables, but a) those machines can also boot from GPT based tables and b)
> we don't have any way to know if we're running on such a machine.  I
> think always requiring GPT for /boot/efi is probably the right thing to
> do, especially since we can't actually tell when it is (or isn't)
> required; the safe plan is to assume it's always required.

This still holds true for the actual EFI case, though.


I number the Linux folks among my personal heroes.
		-- Donald Knuth

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]