[Usb-creator-hackers] about foundations-m-uefi-support

[Evan Dandrea]

On Tue, Jun 1, 2010 at 7:11 PM, Dmitrijs Ledkovs
<dmitrij.ledkov at ubuntu.com> wrote:
> Not sure if it's appropriate to raise this question here about
> foundationa-m-uefi-support [1]

It sure is.

> Quote:
> "Intel Macs should *not* boot in EFI mode, because this breaks
> proprietary driver bits; legacy mode will work better here. In other
> words, we'll leave Intel Macs as they are for the moment."
>
> Can someone expand on the "breaks proprietary driver bits" with
> respect to Intel Macs?

For what it's worth, you can catch the audio recordings of most of the
UDS sessions at:
http://uds.ubuntu.com/audio/uds-m/

This specific talk is at:
http://uds.ubuntu.com/audio/uds-m/2010-05-10/2010-05-10_111133_ebene_.ogg
11:44:00 - Discussion starts
11:50:30 - Blessing
11:53:00 - Why we wont enable on Intel Macs

> I'm owning Early 2006 Macbook model aka 1.1 [2] and all models upto
> Mid-2007 model use Intel GMA 950 and do not require "proprietary
> driver bits" except for iSight firmware for the webcam. You can push
> up to Late 2008 Macbooks which have Intel GMA X3100  (not sure the
> sate of the drivers there).

Some machines have Nvidia chipsets, and the nvidia driver does not
play well in this environment just yet.  My 2006 iMac, for example,
requires that I use vesafb when booting with grub-efi, otherwise it
hard locks once you hit nvidia.ko.

> I have succesfully booted my macbook with grub-efi when installed on
> the hfs+ mac os x partition and "blessed" when booted in mac os x.
> After i could restart mac and boot it with grub-efi, or boot into
> rEFIt [3] menu and choose grub-efi from there.
>
> I was planning on experimenting installing grub-efi directly in ubuntu
> and getting that to boot. Also note that rEFIt can detected bootable
> partitions/bootloaders on reiserFS, ext2 & ext3 (haven't tried ext4
> and it's not explicitly mentioned in the documentation but shouldn't
> be that hard to implement). If you are interested you can poke the
> bzr-import of the upstream svn ;-) here [4].
>
> It is good enough assumption that Intel macs should *not* boot in EFI
> mode, but these older Macbooks are very valid candidates for testing
> and even having it by-default / easy-to-switch to. (Fancy pants magic
> in the ubiquity advanced mode ;-) or alternative install cd).

Colin, do you see any value in a preseed to force EFI in the Mac case?

> At least the current plan I have in mind for the usb-creator Mac port
> is to partition usb-stick using GPT partition table, install rEFIt on
> it and use that to boot LiveUbuntu (my mileage may vary but this is
> currently the most optimal and future proof way in the light of grub2
> on cd's spec). Alternative to do mdos partitioning table and create
> sticks like the udev backend does right now should be supported as
> well cause you do want to boot regular bios PC as well.

Indeed, I believe you'll need to go the rEFIt route as Intel Macs do
not support booting from an MBR on a USB disk.

> Personally booting intel macs using EFI will shave 8-10s of boot time
> if measured from power button press, instead of from the point where
> you have chainloaded into grub-pc & bootchart kicks in.
>
> Have folks from mactel support team participated in discussing this spec?

Not to my knowledge.  Feel free to include them in the discussion
moving forward.

For what it's worth, I've just spoken with Colin and he's in the midst
of working out the finer details of how EFI booting on Ubuntu will
work, but there are still some as-yet unanswered questions.  How do we
avoid using grub-efi on the CD for an Intel Mac, for example.