Om2009 testing release 4

[Al Johnson]

On Wednesday 03 June 2009, Kevin Day wrote:
> On Tue, Jun 2, 2009 at 9:50 AM, Al Johnson
>
> <openmoko at mazikeen.demon.co.uk> wrote:
> > On Tuesday 02 June 2009, Robin Paulson wrote:
> >> 2009/6/2 Al Johnson <openmoko at mazikeen.demon.co.uk>:
> >> > On Tuesday 26 May 2009, Robin Paulson wrote:
> >> >> there's no incentive i can see to stay with uboot
> >> >
> >> > You probably don't have 4 different distros on your SD then ;-)
> >>
> >> no, i don't.
> >>
> >> but it doesn't look like it will be long before qi can do multi-boot
> >>
> >> http://wiki.openmoko.org/wiki/Qi#Boot_Menu
> >
> > It's been done already - there was a thread about building the minimal
> > kernel and rootfs, and using it to kexec into the one you really want to
> > boot. Unfortunately that takes up another partition that I just don't
> > have spare due to the kernel limitation on partitions in mmcblk devices,
> > and android's thirst for partitions. This could be worked around too, but
> > as I already have  a working uboot setup there doesn't seem much point.
> > I've nothing against Qi, but in my case uboot is the better tool for the
> > job.
>
> What one could do is create an initramfs and build it directly into the
> kernel. The small rootfs would exist with the boot kernel on any desired
> partition (say partition 0), taking up no more partitions than already
> used.

That's one of several possible workarounds, but it's not very helpful given 
that the partitions tend to get reformatted and have a new tarball extracted 
over them fairly often. The workarounds all involve extra work for little 
gain.

One thing that would fix it would be if Qi would check the last partition, 
which may be a partition that is outside the kernel partition limit.