[coreboot] resuming a kernel directly from coreboot?

[Matt Price]

On Sat, 2008-10-18 at 21:27 -0700, ron minnich wrote:
> On Fri, Oct 17, 2008 at 7:33 PM, Matt Price <matt.price at utoronto.ca> wrote:
> > On Fri, 2008-10-17 at 16:55 -0700, ron minnich wrote:
> >> I think this is possible, given someone who is motivated :-)
> >>
> > can you give me a sense of what "motivation" entails?  where i should i
> > start digging to try to figure this out?  thanks,
> >
> 
> Where is the suspended kernel saved? What does kernel A do to boot
> Kernel B? Basic kexec?
> 
> we need those three answers first.
> 
here's what i think so far:
- the suspened image is saved to a swap partition, whose location needs
to be specified within the initramfs which will be part of the payload.
i don't know if that's a problem or not; presumably one would set it up
outside of any LVM, RAID, or more complex arrangement.  
- as i understand it (and i'm not sure about this) kernels since 2.6.26
can use a hibernation image written by another kernel.  So when there is
a valid suspended image in the right place, kernel A simply activates
the normal resume sequence, and we automatically end up with kernel B at
the end of the resume process.
- one would need some other process for booting kernel B in cases where
there's no good hibernation image -- like on the system's first boot! --
i guess one could use kexec in an initramfs script.  I haven't put this
system together yet so I don't know whether there might be driver issues
using kexec.  

i guess one could start with a proof-of-concept minimal kernel in qemu &
see if things work?  i'm not sure how well the in-kernel suspend works
with qemu, though -- i've always used the "save state" option there...

anyway does that sound about right?

matt




> ron
> 
> --
> coreboot mailing list: coreboot at coreboot.org
> http://www.coreboot.org/mailman/listinfo/coreboot
-- 
Matt Price
matt.price at utoronto.ca