Re: EFI boot problem with MBP 5,4

[Daire Byrne]

Robert,

I had some time to do the bisect work. It looks like this kernel patch
broke booting by grub-efi on MBP5,4 from v2.6.31 onwards:

ceefccc93932b920a8ec6f35f596db05202a12fe is the first bad commit
commit ceefccc93932b920a8ec6f35f596db05202a12fe
Author: H. Peter Anvin <address@hidden>
Date:   Mon May 11 16:12:16 2009 -0700

    x86: default CONFIG_PHYSICAL_START and CONFIG_PHYSICAL_ALIGN to 16 MB

    Default CONFIG_PHYSICAL_START and CONFIG_PHYSICAL_ALIGN each to 16 MB,
    so that both non-relocatable and relocatable kernels are loaded at
    16 MB by a non-relocating bootloader.  This is somewhat hacky, but it
    appears to be the only way to do this that does not break some some
    set of existing bootloaders.

    We want to avoid the bottom 16 MB because of large page breakup,
    memory holes, and ZONE_DMA.  Embedded systems may need to reduce this,
    or update their bootloaders to be aware of the new min_alignment field.

    [ Impact: performance improvement, avoids problems on some systems ]

    Signed-off-by: H. Peter Anvin <address@hidden>

Is this just a kernel bug that I should report or can grub work around
this? It seems to be a unique incomparability with the EFI firmware in
the latest MBP because grub-efi can boot kernels with this patch on
earlier macbooks (e.g. MBP4,1).

Daire

On Mon, Jan 11, 2010 at 5:17 PM, Daire Byrne <address@hidden> wrote:
> Robert,
>
> On Tue, Jan 5, 2010 at 5:15 PM, Daire Byrne <address@hidden> wrote:
>>> It looks like recent kernels (namely 2.6.31 and 2.6.32) can't find the
>>> initrd when booting from grub (efi) on a macbook pro 5,4 giving the
>>> error:
>>>
>>>  RAMDISK: Couldn't find valid RAM disk image starting at 0.
>>>
>>> The 2.6.30 kernel does not seem to suffer from this so something must
>>> have changed in 2.6.31+. I also tested all three kernels on a MBP4,1
>>> and they all boot fine from grub - they find and load the initrd. In all
>>> cases the grub.cfg and initrd remained the same and only the kernel
>>> was changed. This could be a kernel issue but the fact that they all
>>> boot on a MBP4,1 makes me think it is a EFI1.1 vs EFI 2.0 thing.
>>>
>>> As an aside I can get the nvidia module working with MBP5,4+2.6.30 but
>>> a MBP4,1+2.6.30 can't find the vbios. Is this a known thing and is
>>> there a possible fix for it?
>>>
>>
>>Perhaps you could run a regression test (aka bisect) on Linux source?  This
>>would help us figure out the problem (whether it's a bug in Linux, in GRUB
>>or in your firmware).
>
> I haven't yet had time to dig into this deeper and do the required
> bisect work yet but when searching for similar reports I did find a
> report about elilo not being able to boot since kernel 2.6.30.
>
>  http://bugzilla.kernel.org/show_bug.cgi?id=14658
>  http://marc.info/?l=linux-kernel&m=125951310317648&w=4
>
> It doesn't look like the proposed patch actually fixed the issue.
> Either way that patch is now in 2.6.32 and yet it does not resolve the
> ramdisk loading error on my MPB5,4 (btw I'm using a 64bit kernel +
> EFI). I will try and do the bisect investigation but I'm still quite
> new to git+kernel.org repos so it may take me some time. Because
> 2.6.32 boots on a MBP4,1 I assume that there is something in 2.6.31+
> that does not like newer 64bit EFI Mac firmwares.
>
> Daire
>