On 2010-07-27 11:26 AM, Lennart Sorensen wrote:
> On Mon, Jul 26, 2010 at 09:00:53PM -0400, Doug Nazar wrote:
>
>> I'll run it through the grinder in a bit.
> So with my patch I am booting of the md raid1 very successfully (other
> than manually having to fix grub-mkimage run since grub-install doesn't
> understand the machine yet.  I am working on fixing that now.)

Ok, finally made some progress. Ran into several issues, some of them 
obviously QEMU/OpenBios that I'm not sure if GRUB should work around. 
With your patch the raid mostly worked, small problem with too many 
devices because OpenBios creates several aliases for some devices. Also 
I think you missed the endianess when setting the level.

This patch includes the following:

- Fix the ofdisk_hash system. We weren't making a copy of the devpath so 
never found the cached item again.
- Extend the ofdisk_hash to cache the disk size
- Scan for a disk size using seek (probably want to set a different 
start size). Required for metadata 1.0 arrays
- Optimize checking of raid level
- If we find a duplicate disk (claims to be same index in the array), 
skip it or else level 0 arrays wont be found
- QEMU/OpenBios doesn't seem to like if the prev & name parameters of 
ieee1275_next_property are the same pointer which caused no devices to 
be found

The issues that I came across which are in QEMU/OpenBios:

- The rows are misreported. screen-#rows is set to 75 when in fact there 
are only 60 rows. Worked around using -prom-env parameter
- Aliases don't take into account the index (i.e. disk@1). I ended up with
     disk /pci/pci-ata/ata-1/disk
     hd /pci/pci-ata/ata-1/disk
     ide0 /pci/pci-ata/ata-1/disk
     ide1 /pci/mac-io/ata-3/disk
     ide2 /pci/mac-io/ata-3/disk

when ide1 should be disk@0 and ide2 should be disk@1
- boot command hangs when passed wrong disk or used from boot-command. 
Worked around by using load & go

Things do work, and fixing QEMU/OpenBios is a bit further down the 
rabbit hole than I want to go. ;-)

Thanks,
Doug